MyFavMirrors/minio
1
0
Fork 0
mirror of https://github.com/minio/minio.git synced 2024-12-27 15:45:55 -05:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #440 from fkautz/pr_out_removing_godeps_from_pkg_api

Browse Source
This commit is contained in:
Frederick F. Kautz IV 2015-04-03 20:40:37 -07:00
commit 2abb4bba3a
153 changed files with 0 additions and 49741 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
pkg/api/Godeps/Godeps.json generated
View File

@ -1,41 +0,0 @@
{
"ImportPath": "github.com/minio-io/minio/pkg/api",
"GoVersion": "go1.4",
"Packages": [
"./..."
],
"Deps": [
{
"ImportPath": "github.com/gorilla/context",
"Rev": "50c25fb3b2b3b3cc724e9b6ac75fb44b3bccd0da"
},
{
"ImportPath": "github.com/gorilla/mux",
"Rev": "e444e69cbd2e2e3e0749a2f3c717cec491552bbf"
},
{
"ImportPath": "github.com/minio-io/erasure",
"Rev": "3cece1a107115563682604b1430418e28f65dd80"
},
{
"ImportPath": "github.com/minio-io/iodine",
"Rev": "55cc4d4256c68fbd6f0775f1a25e37e6a2f6457e"
},
{
"ImportPath": "github.com/stretchr/objx",
"Rev": "cbeaeb16a013161a98496fad62933b1d21786672"
},
{
"ImportPath": "github.com/stretchr/testify/assert",
"Rev": "e4ec8152c15fc46bd5056ce65997a07c7d415325"
},
{
"ImportPath": "github.com/stretchr/testify/mock",
"Rev": "e4ec8152c15fc46bd5056ce65997a07c7d415325"
},
{
"ImportPath": "gopkg.in/check.v1",
"Rev": "64131543e7896d5bcc6bd5a76287eb75ea96c673"
}
]
}

5
pkg/api/Godeps/Readme generated
View File

@ -1,5 +0,0 @@
This directory tree is generated automatically by godep.
Please do not edit.
See https://github.com/tools/godep for more information.

2
pkg/api/Godeps/_workspace/.gitignore generated vendored
View File

@ -1,2 +0,0 @@
/pkg
/bin

7
pkg/api/Godeps/_workspace/src/github.com/gorilla/context/.travis.yml generated vendored
View File

@ -1,7 +0,0 @@
language: go
go:
- 1.0
- 1.1
- 1.2
- tip

27
pkg/api/Godeps/_workspace/src/github.com/gorilla/context/LICENSE generated vendored
View File

@ -1,27 +0,0 @@
Copyright (c) 2012 Rodrigo Moraes. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following disclaimer
in the documentation and/or other materials provided with the
distribution.
* Neither the name of Google Inc. nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

7
pkg/api/Godeps/_workspace/src/github.com/gorilla/context/README.md generated vendored
View File

@ -1,7 +0,0 @@
context
=======
[![Build Status](https://travis-ci.org/gorilla/context.png?branch=master)](https://travis-ci.org/gorilla/context)
gorilla/context is a general purpose registry for global request variables.
Read the full documentation here: http://www.gorillatoolkit.org/pkg/context

143
pkg/api/Godeps/_workspace/src/github.com/gorilla/context/context.go generated vendored
View File

@ -1,143 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package context
import (
"net/http"
"sync"
"time"
)
var (
mutex sync.RWMutex
data = make(map[*http.Request]map[interface{}]interface{})
datat = make(map[*http.Request]int64)
)
// Set stores a value for a given key in a given request.
func Set(r *http.Request, key, val interface{}) {
mutex.Lock()
if data[r] == nil {
data[r] = make(map[interface{}]interface{})
datat[r] = time.Now().Unix()
}
data[r][key] = val
mutex.Unlock()
}
// Get returns a value stored for a given key in a given request.
func Get(r *http.Request, key interface{}) interface{} {
mutex.RLock()
if ctx := data[r]; ctx != nil {
value := ctx[key]
mutex.RUnlock()
return value
}
mutex.RUnlock()
return nil
}
// GetOk returns stored value and presence state like multi-value return of map access.
func GetOk(r *http.Request, key interface{}) (interface{}, bool) {
mutex.RLock()
if _, ok := data[r]; ok {
value, ok := data[r][key]
mutex.RUnlock()
return value, ok
}
mutex.RUnlock()
return nil, false
}
// GetAll returns all stored values for the request as a map. Nil is returned for invalid requests.
func GetAll(r *http.Request) map[interface{}]interface{} {
mutex.RLock()
if context, ok := data[r]; ok {
result := make(map[interface{}]interface{}, len(context))
for k, v := range context {
result[k] = v
}
mutex.RUnlock()
return result
}
mutex.RUnlock()
return nil
}
// GetAllOk returns all stored values for the request as a map and a boolean value that indicates if
// the request was registered.
func GetAllOk(r *http.Request) (map[interface{}]interface{}, bool) {
mutex.RLock()
context, ok := data[r]
result := make(map[interface{}]interface{}, len(context))
for k, v := range context {
result[k] = v
}
mutex.RUnlock()
return result, ok
}
// Delete removes a value stored for a given key in a given request.
func Delete(r *http.Request, key interface{}) {
mutex.Lock()
if data[r] != nil {
delete(data[r], key)
}
mutex.Unlock()
}
// Clear removes all values stored for a given request.
//
// This is usually called by a handler wrapper to clean up request
// variables at the end of a request lifetime. See ClearHandler().
func Clear(r *http.Request) {
mutex.Lock()
clear(r)
mutex.Unlock()
}
// clear is Clear without the lock.
func clear(r *http.Request) {
delete(data, r)
delete(datat, r)
}
// Purge removes request data stored for longer than maxAge, in seconds.
// It returns the amount of requests removed.
//
// If maxAge <= 0, all request data is removed.
//
// This is only used for sanity check: in case context cleaning was not
// properly set some request data can be kept forever, consuming an increasing
// amount of memory. In case this is detected, Purge() must be called
// periodically until the problem is fixed.
func Purge(maxAge int) int {
mutex.Lock()
count := 0
if maxAge <= 0 {
count = len(data)
data = make(map[*http.Request]map[interface{}]interface{})
datat = make(map[*http.Request]int64)
} else {
min := time.Now().Unix() - int64(maxAge)
for r := range data {
if datat[r] < min {
clear(r)
count++
}
}
}
mutex.Unlock()
return count
}
// ClearHandler wraps an http.Handler and clears request values at the end
// of a request lifetime.
func ClearHandler(h http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
defer Clear(r)
h.ServeHTTP(w, r)
})
}

161
pkg/api/Godeps/_workspace/src/github.com/gorilla/context/context_test.go generated vendored
View File

@ -1,161 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package context
import (
"net/http"
"testing"
)
type keyType int
const (
key1 keyType = iota
key2
)
func TestContext(t *testing.T) {
assertEqual := func(val interface{}, exp interface{}) {
if val != exp {
t.Errorf("Expected %v, got %v.", exp, val)
}
}
r, _ := http.NewRequest("GET", "http://localhost:8080/", nil)
emptyR, _ := http.NewRequest("GET", "http://localhost:8080/", nil)
// Get()
assertEqual(Get(r, key1), nil)
// Set()
Set(r, key1, "1")
assertEqual(Get(r, key1), "1")
assertEqual(len(data[r]), 1)
Set(r, key2, "2")
assertEqual(Get(r, key2), "2")
assertEqual(len(data[r]), 2)
//GetOk
value, ok := GetOk(r, key1)
assertEqual(value, "1")
assertEqual(ok, true)
value, ok = GetOk(r, "not exists")
assertEqual(value, nil)
assertEqual(ok, false)
Set(r, "nil value", nil)
value, ok = GetOk(r, "nil value")
assertEqual(value, nil)
assertEqual(ok, true)
// GetAll()
values := GetAll(r)
assertEqual(len(values), 3)
// GetAll() for empty request
values = GetAll(emptyR)
if values != nil {
t.Error("GetAll didn't return nil value for invalid request")
}
// GetAllOk()
values, ok = GetAllOk(r)
assertEqual(len(values), 3)
assertEqual(ok, true)
// GetAllOk() for empty request
values, ok = GetAllOk(emptyR)
assertEqual(value, nil)
assertEqual(ok, false)
// Delete()
Delete(r, key1)
assertEqual(Get(r, key1), nil)
assertEqual(len(data[r]), 2)
Delete(r, key2)
assertEqual(Get(r, key2), nil)
assertEqual(len(data[r]), 1)
// Clear()
Clear(r)
assertEqual(len(data), 0)
}
func parallelReader(r *http.Request, key string, iterations int, wait, done chan struct{}) {
<-wait
for i := 0; i < iterations; i++ {
Get(r, key)
}
done <- struct{}{}
}
func parallelWriter(r *http.Request, key, value string, iterations int, wait, done chan struct{}) {
<-wait
for i := 0; i < iterations; i++ {
Set(r, key, value)
}
done <- struct{}{}
}
func benchmarkMutex(b *testing.B, numReaders, numWriters, iterations int) {
b.StopTimer()
r, _ := http.NewRequest("GET", "http://localhost:8080/", nil)
done := make(chan struct{})
b.StartTimer()
for i := 0; i < b.N; i++ {
wait := make(chan struct{})
for i := 0; i < numReaders; i++ {
go parallelReader(r, "test", iterations, wait, done)
}
for i := 0; i < numWriters; i++ {
go parallelWriter(r, "test", "123", iterations, wait, done)
}
close(wait)
for i := 0; i < numReaders+numWriters; i++ {
<-done
}
}
}
func BenchmarkMutexSameReadWrite1(b *testing.B) {
benchmarkMutex(b, 1, 1, 32)
}
func BenchmarkMutexSameReadWrite2(b *testing.B) {
benchmarkMutex(b, 2, 2, 32)
}
func BenchmarkMutexSameReadWrite4(b *testing.B) {
benchmarkMutex(b, 4, 4, 32)
}
func BenchmarkMutex1(b *testing.B) {
benchmarkMutex(b, 2, 8, 32)
}
func BenchmarkMutex2(b *testing.B) {
benchmarkMutex(b, 16, 4, 64)
}
func BenchmarkMutex3(b *testing.B) {
benchmarkMutex(b, 1, 2, 128)
}
func BenchmarkMutex4(b *testing.B) {
benchmarkMutex(b, 128, 32, 256)
}
func BenchmarkMutex5(b *testing.B) {
benchmarkMutex(b, 1024, 2048, 64)
}
func BenchmarkMutex6(b *testing.B) {
benchmarkMutex(b, 2048, 1024, 512)
}

82
pkg/api/Godeps/_workspace/src/github.com/gorilla/context/doc.go generated vendored
View File

@ -1,82 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
/*
Package context stores values shared during a request lifetime.
For example, a router can set variables extracted from the URL and later
application handlers can access those values, or it can be used to store
sessions values to be saved at the end of a request. There are several
others common uses.
The idea was posted by Brad Fitzpatrick to the go-nuts mailing list:
http://groups.google.com/group/golang-nuts/msg/e2d679d303aa5d53
Here's the basic usage: first define the keys that you will need. The key
type is interface{} so a key can be of any type that supports equality.
Here we define a key using a custom int type to avoid name collisions:
package foo
import (
"github.com/gorilla/context"
)
type key int
const MyKey key = 0
Then set a variable. Variables are bound to an http.Request object, so you
need a request instance to set a value:
context.Set(r, MyKey, "bar")
The application can later access the variable using the same key you provided:
func MyHandler(w http.ResponseWriter, r *http.Request) {
// val is "bar".
val := context.Get(r, foo.MyKey)
// returns ("bar", true)
val, ok := context.GetOk(r, foo.MyKey)
// ...
}
And that's all about the basic usage. We discuss some other ideas below.
Any type can be stored in the context. To enforce a given type, make the key
private and wrap Get() and Set() to accept and return values of a specific
type:
type key int
const mykey key = 0
// GetMyKey returns a value for this package from the request values.
func GetMyKey(r *http.Request) SomeType {
if rv := context.Get(r, mykey); rv != nil {
return rv.(SomeType)
}
return nil
}
// SetMyKey sets a value for this package in the request values.
func SetMyKey(r *http.Request, val SomeType) {
context.Set(r, mykey, val)
}
Variables must be cleared at the end of a request, to remove all values
that were stored. This can be done in an http.Handler, after a request was
served. Just call Clear() passing the request:
context.Clear(r)
...or use ClearHandler(), which conveniently wraps an http.Handler to clear
variables at the end of a request lifetime.
The Routers from the packages gorilla/mux and gorilla/pat call Clear()
so if you are using either of them you don't need to clear the context manually.
*/
package context

7
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/.travis.yml generated vendored
View File

@ -1,7 +0,0 @@
language: go
go:
- 1.0
- 1.1
- 1.2
- tip

27
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/LICENSE generated vendored
View File

@ -1,27 +0,0 @@
Copyright (c) 2012 Rodrigo Moraes. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following disclaimer
in the documentation and/or other materials provided with the
distribution.
* Neither the name of Google Inc. nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

7
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/README.md generated vendored
View File

@ -1,7 +0,0 @@
mux
===
[![Build Status](https://travis-ci.org/gorilla/mux.png?branch=master)](https://travis-ci.org/gorilla/mux)
gorilla/mux is a powerful URL router and dispatcher.
Read the full documentation here: http://www.gorillatoolkit.org/pkg/mux

21
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/bench_test.go generated vendored
View File

@ -1,21 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package mux
import (
"net/http"
"testing"
)
func BenchmarkMux(b *testing.B) {
router := new(Router)
handler := func(w http.ResponseWriter, r *http.Request) {}
router.HandleFunc("/v1/{v1}", handler)
request, _ := http.NewRequest("GET", "/v1/anything", nil)
for i := 0; i < b.N; i++ {
router.ServeHTTP(nil, request)
}
}

199
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/doc.go generated vendored
View File

@ -1,199 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
/*
Package gorilla/mux implements a request router and dispatcher.
The name mux stands for "HTTP request multiplexer". Like the standard
http.ServeMux, mux.Router matches incoming requests against a list of
registered routes and calls a handler for the route that matches the URL
or other conditions. The main features are:
* Requests can be matched based on URL host, path, path prefix, schemes,
header and query values, HTTP methods or using custom matchers.
* URL hosts and paths can have variables with an optional regular
expression.
* Registered URLs can be built, or "reversed", which helps maintaining
references to resources.
* Routes can be used as subrouters: nested routes are only tested if the
parent route matches. This is useful to define groups of routes that
share common conditions like a host, a path prefix or other repeated
attributes. As a bonus, this optimizes request matching.
* It implements the http.Handler interface so it is compatible with the
standard http.ServeMux.
Let's start registering a couple of URL paths and handlers:
func main() {
r := mux.NewRouter()
r.HandleFunc("/", HomeHandler)
r.HandleFunc("/products", ProductsHandler)
r.HandleFunc("/articles", ArticlesHandler)
http.Handle("/", r)
}
Here we register three routes mapping URL paths to handlers. This is
equivalent to how http.HandleFunc() works: if an incoming request URL matches
one of the paths, the corresponding handler is called passing
(http.ResponseWriter, *http.Request) as parameters.
Paths can have variables. They are defined using the format {name} or
{name:pattern}. If a regular expression pattern is not defined, the matched
variable will be anything until the next slash. For example:
r := mux.NewRouter()
r.HandleFunc("/products/{key}", ProductHandler)
r.HandleFunc("/articles/{category}/", ArticlesCategoryHandler)
r.HandleFunc("/articles/{category}/{id:[0-9]+}", ArticleHandler)
The names are used to create a map of route variables which can be retrieved
calling mux.Vars():
vars := mux.Vars(request)
category := vars["category"]
And this is all you need to know about the basic usage. More advanced options
are explained below.
Routes can also be restricted to a domain or subdomain. Just define a host
pattern to be matched. They can also have variables:
r := mux.NewRouter()
// Only matches if domain is "www.domain.com".
r.Host("www.domain.com")
// Matches a dynamic subdomain.
r.Host("{subdomain:[a-z]+}.domain.com")
There are several other matchers that can be added. To match path prefixes:
r.PathPrefix("/products/")
...or HTTP methods:
r.Methods("GET", "POST")
...or URL schemes:
r.Schemes("https")
...or header values:
r.Headers("X-Requested-With", "XMLHttpRequest")
...or query values:
r.Queries("key", "value")
...or to use a custom matcher function:
r.MatcherFunc(func(r *http.Request, rm *RouteMatch) bool {
return r.ProtoMajor == 0
})
...and finally, it is possible to combine several matchers in a single route:
r.HandleFunc("/products", ProductsHandler).
Host("www.domain.com").
Methods("GET").
Schemes("http")
Setting the same matching conditions again and again can be boring, so we have
a way to group several routes that share the same requirements.
We call it "subrouting".
For example, let's say we have several URLs that should only match when the
host is "www.domain.com". Create a route for that host and get a "subrouter"
from it:
r := mux.NewRouter()
s := r.Host("www.domain.com").Subrouter()
Then register routes in the subrouter:
s.HandleFunc("/products/", ProductsHandler)
s.HandleFunc("/products/{key}", ProductHandler)
s.HandleFunc("/articles/{category}/{id:[0-9]+}"), ArticleHandler)
The three URL paths we registered above will only be tested if the domain is
"www.domain.com", because the subrouter is tested first. This is not
only convenient, but also optimizes request matching. You can create
subrouters combining any attribute matchers accepted by a route.
Subrouters can be used to create domain or path "namespaces": you define
subrouters in a central place and then parts of the app can register its
paths relatively to a given subrouter.
There's one more thing about subroutes. When a subrouter has a path prefix,
the inner routes use it as base for their paths:
r := mux.NewRouter()
s := r.PathPrefix("/products").Subrouter()
// "/products/"
s.HandleFunc("/", ProductsHandler)
// "/products/{key}/"
s.HandleFunc("/{key}/", ProductHandler)
// "/products/{key}/details"
s.HandleFunc("/{key}/details", ProductDetailsHandler)
Now let's see how to build registered URLs.
Routes can be named. All routes that define a name can have their URLs built,
or "reversed". We define a name calling Name() on a route. For example:
r := mux.NewRouter()
r.HandleFunc("/articles/{category}/{id:[0-9]+}", ArticleHandler).
Name("article")
To build a URL, get the route and call the URL() method, passing a sequence of
key/value pairs for the route variables. For the previous route, we would do:
url, err := r.Get("article").URL("category", "technology", "id", "42")
...and the result will be a url.URL with the following path:
"/articles/technology/42"
This also works for host variables:
r := mux.NewRouter()
r.Host("{subdomain}.domain.com").
Path("/articles/{category}/{id:[0-9]+}").
HandlerFunc(ArticleHandler).
Name("article")
// url.String() will be "http://news.domain.com/articles/technology/42"
url, err := r.Get("article").URL("subdomain", "news",
"category", "technology",
"id", "42")
All variables defined in the route are required, and their values must
conform to the corresponding patterns. These requirements guarantee that a
generated URL will always match a registered route -- the only exception is
for explicitly defined "build-only" routes which never match.
There's also a way to build only the URL host or path for a route:
use the methods URLHost() or URLPath() instead. For the previous route,
we would do:
// "http://news.domain.com/"
host, err := r.Get("article").URLHost("subdomain", "news")
// "/articles/technology/42"
path, err := r.Get("article").URLPath("category", "technology", "id", "42")
And if you use subrouters, host and path defined separately can be built
as well:
r := mux.NewRouter()
s := r.Host("{subdomain}.domain.com").Subrouter()
s.Path("/articles/{category}/{id:[0-9]+}").
HandlerFunc(ArticleHandler).
Name("article")
// "http://news.domain.com/articles/technology/42"
url, err := r.Get("article").URL("subdomain", "news",
"category", "technology",
"id", "42")
*/
package mux

353
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/mux.go generated vendored
View File

@ -1,353 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package mux
import (
"fmt"
"net/http"
"path"
"github.com/gorilla/context"
)
// NewRouter returns a new router instance.
func NewRouter() *Router {
return &Router{namedRoutes: make(map[string]*Route), KeepContext: false}
}
// Router registers routes to be matched and dispatches a handler.
//
// It implements the http.Handler interface, so it can be registered to serve
// requests:
//
// var router = mux.NewRouter()
//
// func main() {
// http.Handle("/", router)
// }
//
// Or, for Google App Engine, register it in a init() function:
//
// func init() {
// http.Handle("/", router)
// }
//
// This will send all incoming requests to the router.
type Router struct {
// Configurable Handler to be used when no route matches.
NotFoundHandler http.Handler
// Parent route, if this is a subrouter.
parent parentRoute
// Routes to be matched, in order.
routes []*Route
// Routes by name for URL building.
namedRoutes map[string]*Route
// See Router.StrictSlash(). This defines the flag for new routes.
strictSlash bool
// If true, do not clear the request context after handling the request
KeepContext bool
}
// Match matches registered routes against the request.
func (r *Router) Match(req *http.Request, match *RouteMatch) bool {
for _, route := range r.routes {
if route.Match(req, match) {
return true
}
}
return false
}
// ServeHTTP dispatches the handler registered in the matched route.
//
// When there is a match, the route variables can be retrieved calling
// mux.Vars(request).
func (r *Router) ServeHTTP(w http.ResponseWriter, req *http.Request) {
// Clean path to canonical form and redirect.
if p := cleanPath(req.URL.Path); p != req.URL.Path {
// Added 3 lines (Philip Schlump) - It was droping the query string and #whatever from query.
// This matches with fix in go 1.2 r.c. 4 for same problem. Go Issue:
// http://code.google.com/p/go/issues/detail?id=5252
url := *req.URL
url.Path = p
p = url.String()
w.Header().Set("Location", p)
w.WriteHeader(http.StatusMovedPermanently)
return
}
var match RouteMatch
var handler http.Handler
if r.Match(req, &match) {
handler = match.Handler
setVars(req, match.Vars)
setCurrentRoute(req, match.Route)
}
if handler == nil {
handler = r.NotFoundHandler
if handler == nil {
handler = http.NotFoundHandler()
}
}
if !r.KeepContext {
defer context.Clear(req)
}
handler.ServeHTTP(w, req)
}
// Get returns a route registered with the given name.
func (r *Router) Get(name string) *Route {
return r.getNamedRoutes()[name]
}
// GetRoute returns a route registered with the given name. This method
// was renamed to Get() and remains here for backwards compatibility.
func (r *Router) GetRoute(name string) *Route {
return r.getNamedRoutes()[name]
}
// StrictSlash defines the trailing slash behavior for new routes. The initial
// value is false.
//
// When true, if the route path is "/path/", accessing "/path" will redirect
// to the former and vice versa. In other words, your application will always
// see the path as specified in the route.
//
// When false, if the route path is "/path", accessing "/path/" will not match
// this route and vice versa.
//
// Special case: when a route sets a path prefix using the PathPrefix() method,
// strict slash is ignored for that route because the redirect behavior can't
// be determined from a prefix alone. However, any subrouters created from that
// route inherit the original StrictSlash setting.
func (r *Router) StrictSlash(value bool) *Router {
r.strictSlash = value
return r
}
// ----------------------------------------------------------------------------
// parentRoute
// ----------------------------------------------------------------------------
// getNamedRoutes returns the map where named routes are registered.
func (r *Router) getNamedRoutes() map[string]*Route {
if r.namedRoutes == nil {
if r.parent != nil {
r.namedRoutes = r.parent.getNamedRoutes()
} else {
r.namedRoutes = make(map[string]*Route)
}
}
return r.namedRoutes
}
// getRegexpGroup returns regexp definitions from the parent route, if any.
func (r *Router) getRegexpGroup() *routeRegexpGroup {
if r.parent != nil {
return r.parent.getRegexpGroup()
}
return nil
}
// ----------------------------------------------------------------------------
// Route factories
// ----------------------------------------------------------------------------
// NewRoute registers an empty route.
func (r *Router) NewRoute() *Route {
route := &Route{parent: r, strictSlash: r.strictSlash}
r.routes = append(r.routes, route)
return route
}
// Handle registers a new route with a matcher for the URL path.
// See Route.Path() and Route.Handler().
func (r *Router) Handle(path string, handler http.Handler) *Route {
return r.NewRoute().Path(path).Handler(handler)
}
// HandleFunc registers a new route with a matcher for the URL path.
// See Route.Path() and Route.HandlerFunc().
func (r *Router) HandleFunc(path string, f func(http.ResponseWriter,
*http.Request)) *Route {
return r.NewRoute().Path(path).HandlerFunc(f)
}
// Headers registers a new route with a matcher for request header values.
// See Route.Headers().
func (r *Router) Headers(pairs ...string) *Route {
return r.NewRoute().Headers(pairs...)
}
// Host registers a new route with a matcher for the URL host.
// See Route.Host().
func (r *Router) Host(tpl string) *Route {
return r.NewRoute().Host(tpl)
}
// MatcherFunc registers a new route with a custom matcher function.
// See Route.MatcherFunc().
func (r *Router) MatcherFunc(f MatcherFunc) *Route {
return r.NewRoute().MatcherFunc(f)
}
// Methods registers a new route with a matcher for HTTP methods.
// See Route.Methods().
func (r *Router) Methods(methods ...string) *Route {
return r.NewRoute().Methods(methods...)
}
// Path registers a new route with a matcher for the URL path.
// See Route.Path().
func (r *Router) Path(tpl string) *Route {
return r.NewRoute().Path(tpl)
}
// PathPrefix registers a new route with a matcher for the URL path prefix.
// See Route.PathPrefix().
func (r *Router) PathPrefix(tpl string) *Route {
return r.NewRoute().PathPrefix(tpl)
}
// Queries registers a new route with a matcher for URL query values.
// See Route.Queries().
func (r *Router) Queries(pairs ...string) *Route {
return r.NewRoute().Queries(pairs...)
}
// Schemes registers a new route with a matcher for URL schemes.
// See Route.Schemes().
func (r *Router) Schemes(schemes ...string) *Route {
return r.NewRoute().Schemes(schemes...)
}
// ----------------------------------------------------------------------------
// Context
// ----------------------------------------------------------------------------
// RouteMatch stores information about a matched route.
type RouteMatch struct {
Route *Route
Handler http.Handler
Vars map[string]string
}
type contextKey int
const (
varsKey contextKey = iota
routeKey
)
// Vars returns the route variables for the current request, if any.
func Vars(r *http.Request) map[string]string {
if rv := context.Get(r, varsKey); rv != nil {
return rv.(map[string]string)
}
return nil
}
// CurrentRoute returns the matched route for the current request, if any.
func CurrentRoute(r *http.Request) *Route {
if rv := context.Get(r, routeKey); rv != nil {
return rv.(*Route)
}
return nil
}
func setVars(r *http.Request, val interface{}) {
context.Set(r, varsKey, val)
}
func setCurrentRoute(r *http.Request, val interface{}) {
context.Set(r, routeKey, val)
}
// ----------------------------------------------------------------------------
// Helpers
// ----------------------------------------------------------------------------
// cleanPath returns the canonical path for p, eliminating . and .. elements.
// Borrowed from the net/http package.
func cleanPath(p string) string {
if p == "" {
return "/"
}
if p[0] != '/' {
p = "/" + p
}
np := path.Clean(p)
// path.Clean removes trailing slash except for root;
// put the trailing slash back if necessary.
if p[len(p)-1] == '/' && np != "/" {
np += "/"
}
return np
}
// uniqueVars returns an error if two slices contain duplicated strings.
func uniqueVars(s1, s2 []string) error {
for _, v1 := range s1 {
for _, v2 := range s2 {
if v1 == v2 {
return fmt.Errorf("mux: duplicated route variable %q", v2)
}
}
}
return nil
}
// mapFromPairs converts variadic string parameters to a string map.
func mapFromPairs(pairs ...string) (map[string]string, error) {
length := len(pairs)
if length%2 != 0 {
return nil, fmt.Errorf(
"mux: number of parameters must be multiple of 2, got %v", pairs)
}
m := make(map[string]string, length/2)
for i := 0; i < length; i += 2 {
m[pairs[i]] = pairs[i+1]
}
return m, nil
}
// matchInArray returns true if the given string value is in the array.
func matchInArray(arr []string, value string) bool {
for _, v := range arr {
if v == value {
return true
}
}
return false
}
// matchMap returns true if the given key/value pairs exist in a given map.
func matchMap(toCheck map[string]string, toMatch map[string][]string,
canonicalKey bool) bool {
for k, v := range toCheck {
// Check if key exists.
if canonicalKey {
k = http.CanonicalHeaderKey(k)
}
if values := toMatch[k]; values == nil {
return false
} else if v != "" {
// If value was defined as an empty string we only check that the
// key exists. Otherwise we also check for equality.
valueExists := false
for _, value := range values {
if v == value {
valueExists = true
break
}
}
if !valueExists {
return false
}
}
}
return true
}

943
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/mux_test.go generated vendored
View File

@ -1,943 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package mux
import (
"fmt"
"net/http"
"testing"
"github.com/gorilla/context"
)
type routeTest struct {
title string // title of the test
route *Route // the route being tested
request *http.Request // a request to test the route
vars map[string]string // the expected vars of the match
host string // the expected host of the match
path string // the expected path of the match
shouldMatch bool // whether the request is expected to match the route at all
shouldRedirect bool // whether the request should result in a redirect
}
func TestHost(t *testing.T) {
// newRequestHost a new request with a method, url, and host header
newRequestHost := func(method, url, host string) *http.Request {
req, err := http.NewRequest(method, url, nil)
if err != nil {
panic(err)
}
req.Host = host
return req
}
tests := []routeTest{
{
title: "Host route match",
route: new(Route).Host("aaa.bbb.ccc"),
request: newRequest("GET", "http://aaa.bbb.ccc/111/222/333"),
vars: map[string]string{},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: true,
},
{
title: "Host route, wrong host in request URL",
route: new(Route).Host("aaa.bbb.ccc"),
request: newRequest("GET", "http://aaa.222.ccc/111/222/333"),
vars: map[string]string{},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: false,
},
{
title: "Host route with port, match",
route: new(Route).Host("aaa.bbb.ccc:1234"),
request: newRequest("GET", "http://aaa.bbb.ccc:1234/111/222/333"),
vars: map[string]string{},
host: "aaa.bbb.ccc:1234",
path: "",
shouldMatch: true,
},
{
title: "Host route with port, wrong port in request URL",
route: new(Route).Host("aaa.bbb.ccc:1234"),
request: newRequest("GET", "http://aaa.bbb.ccc:9999/111/222/333"),
vars: map[string]string{},
host: "aaa.bbb.ccc:1234",
path: "",
shouldMatch: false,
},
{
title: "Host route, match with host in request header",
route: new(Route).Host("aaa.bbb.ccc"),
request: newRequestHost("GET", "/111/222/333", "aaa.bbb.ccc"),
vars: map[string]string{},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: true,
},
{
title: "Host route, wrong host in request header",
route: new(Route).Host("aaa.bbb.ccc"),
request: newRequestHost("GET", "/111/222/333", "aaa.222.ccc"),
vars: map[string]string{},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: false,
},
// BUG {new(Route).Host("aaa.bbb.ccc:1234"), newRequestHost("GET", "/111/222/333", "aaa.bbb.ccc:1234"), map[string]string{}, "aaa.bbb.ccc:1234", "", true},
{
title: "Host route with port, wrong host in request header",
route: new(Route).Host("aaa.bbb.ccc:1234"),
request: newRequestHost("GET", "/111/222/333", "aaa.bbb.ccc:9999"),
vars: map[string]string{},
host: "aaa.bbb.ccc:1234",
path: "",
shouldMatch: false,
},
{
title: "Host route with pattern, match",
route: new(Route).Host("aaa.{v1:[a-z]{3}}.ccc"),
request: newRequest("GET", "http://aaa.bbb.ccc/111/222/333"),
vars: map[string]string{"v1": "bbb"},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: true,
},
{
title: "Host route with pattern, wrong host in request URL",
route: new(Route).Host("aaa.{v1:[a-z]{3}}.ccc"),
request: newRequest("GET", "http://aaa.222.ccc/111/222/333"),
vars: map[string]string{"v1": "bbb"},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: false,
},
{
title: "Host route with multiple patterns, match",
route: new(Route).Host("{v1:[a-z]{3}}.{v2:[a-z]{3}}.{v3:[a-z]{3}}"),
request: newRequest("GET", "http://aaa.bbb.ccc/111/222/333"),
vars: map[string]string{"v1": "aaa", "v2": "bbb", "v3": "ccc"},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: true,
},
{
title: "Host route with multiple patterns, wrong host in request URL",
route: new(Route).Host("{v1:[a-z]{3}}.{v2:[a-z]{3}}.{v3:[a-z]{3}}"),
request: newRequest("GET", "http://aaa.222.ccc/111/222/333"),
vars: map[string]string{"v1": "aaa", "v2": "bbb", "v3": "ccc"},
host: "aaa.bbb.ccc",
path: "",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestPath(t *testing.T) {
tests := []routeTest{
{
title: "Path route, match",
route: new(Route).Path("/111/222/333"),
request: newRequest("GET", "http://localhost/111/222/333"),
vars: map[string]string{},
host: "",
path: "/111/222/333",
shouldMatch: true,
},
{
title: "Path route, match with trailing slash in request and path",
route: new(Route).Path("/111/"),
request: newRequest("GET", "http://localhost/111/"),
vars: map[string]string{},
host: "",
path: "/111/",
shouldMatch: true,
},
{
title: "Path route, do not match with trailing slash in path",
route: new(Route).Path("/111/"),
request: newRequest("GET", "http://localhost/111"),
vars: map[string]string{},
host: "",
path: "/111",
shouldMatch: false,
},
{
title: "Path route, do not match with trailing slash in request",
route: new(Route).Path("/111"),
request: newRequest("GET", "http://localhost/111/"),
vars: map[string]string{},
host: "",
path: "/111/",
shouldMatch: false,
},
{
title: "Path route, wrong path in request in request URL",
route: new(Route).Path("/111/222/333"),
request: newRequest("GET", "http://localhost/1/2/3"),
vars: map[string]string{},
host: "",
path: "/111/222/333",
shouldMatch: false,
},
{
title: "Path route with pattern, match",
route: new(Route).Path("/111/{v1:[0-9]{3}}/333"),
request: newRequest("GET", "http://localhost/111/222/333"),
vars: map[string]string{"v1": "222"},
host: "",
path: "/111/222/333",
shouldMatch: true,
},
{
title: "Path route with pattern, URL in request does not match",
route: new(Route).Path("/111/{v1:[0-9]{3}}/333"),
request: newRequest("GET", "http://localhost/111/aaa/333"),
vars: map[string]string{"v1": "222"},
host: "",
path: "/111/222/333",
shouldMatch: false,
},
{
title: "Path route with multiple patterns, match",
route: new(Route).Path("/{v1:[0-9]{3}}/{v2:[0-9]{3}}/{v3:[0-9]{3}}"),
request: newRequest("GET", "http://localhost/111/222/333"),
vars: map[string]string{"v1": "111", "v2": "222", "v3": "333"},
host: "",
path: "/111/222/333",
shouldMatch: true,
},
{
title: "Path route with multiple patterns, URL in request does not match",
route: new(Route).Path("/{v1:[0-9]{3}}/{v2:[0-9]{3}}/{v3:[0-9]{3}}"),
request: newRequest("GET", "http://localhost/111/aaa/333"),
vars: map[string]string{"v1": "111", "v2": "222", "v3": "333"},
host: "",
path: "/111/222/333",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestPathPrefix(t *testing.T) {
tests := []routeTest{
{
title: "PathPrefix route, match",
route: new(Route).PathPrefix("/111"),
request: newRequest("GET", "http://localhost/111/222/333"),
vars: map[string]string{},
host: "",
path: "/111",
shouldMatch: true,
},
{
title: "PathPrefix route, match substring",
route: new(Route).PathPrefix("/1"),
request: newRequest("GET", "http://localhost/111/222/333"),
vars: map[string]string{},
host: "",
path: "/1",
shouldMatch: true,
},
{
title: "PathPrefix route, URL prefix in request does not match",
route: new(Route).PathPrefix("/111"),
request: newRequest("GET", "http://localhost/1/2/3"),
vars: map[string]string{},
host: "",
path: "/111",
shouldMatch: false,
},
{
title: "PathPrefix route with pattern, match",
route: new(Route).PathPrefix("/111/{v1:[0-9]{3}}"),
request: newRequest("GET", "http://localhost/111/222/333"),
vars: map[string]string{"v1": "222"},
host: "",
path: "/111/222",
shouldMatch: true,
},
{
title: "PathPrefix route with pattern, URL prefix in request does not match",
route: new(Route).PathPrefix("/111/{v1:[0-9]{3}}"),
request: newRequest("GET", "http://localhost/111/aaa/333"),
vars: map[string]string{"v1": "222"},
host: "",
path: "/111/222",
shouldMatch: false,
},
{
title: "PathPrefix route with multiple patterns, match",
route: new(Route).PathPrefix("/{v1:[0-9]{3}}/{v2:[0-9]{3}}"),
request: newRequest("GET", "http://localhost/111/222/333"),
vars: map[string]string{"v1": "111", "v2": "222"},
host: "",
path: "/111/222",
shouldMatch: true,
},
{
title: "PathPrefix route with multiple patterns, URL prefix in request does not match",
route: new(Route).PathPrefix("/{v1:[0-9]{3}}/{v2:[0-9]{3}}"),
request: newRequest("GET", "http://localhost/111/aaa/333"),
vars: map[string]string{"v1": "111", "v2": "222"},
host: "",
path: "/111/222",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestHostPath(t *testing.T) {
tests := []routeTest{
{
title: "Host and Path route, match",
route: new(Route).Host("aaa.bbb.ccc").Path("/111/222/333"),
request: newRequest("GET", "http://aaa.bbb.ccc/111/222/333"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Host and Path route, wrong host in request URL",
route: new(Route).Host("aaa.bbb.ccc").Path("/111/222/333"),
request: newRequest("GET", "http://aaa.222.ccc/111/222/333"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: false,
},
{
title: "Host and Path route with pattern, match",
route: new(Route).Host("aaa.{v1:[a-z]{3}}.ccc").Path("/111/{v2:[0-9]{3}}/333"),
request: newRequest("GET", "http://aaa.bbb.ccc/111/222/333"),
vars: map[string]string{"v1": "bbb", "v2": "222"},
host: "aaa.bbb.ccc",
path: "/111/222/333",
shouldMatch: true,
},
{
title: "Host and Path route with pattern, URL in request does not match",
route: new(Route).Host("aaa.{v1:[a-z]{3}}.ccc").Path("/111/{v2:[0-9]{3}}/333"),
request: newRequest("GET", "http://aaa.222.ccc/111/222/333"),
vars: map[string]string{"v1": "bbb", "v2": "222"},
host: "aaa.bbb.ccc",
path: "/111/222/333",
shouldMatch: false,
},
{
title: "Host and Path route with multiple patterns, match",
route: new(Route).Host("{v1:[a-z]{3}}.{v2:[a-z]{3}}.{v3:[a-z]{3}}").Path("/{v4:[0-9]{3}}/{v5:[0-9]{3}}/{v6:[0-9]{3}}"),
request: newRequest("GET", "http://aaa.bbb.ccc/111/222/333"),
vars: map[string]string{"v1": "aaa", "v2": "bbb", "v3": "ccc", "v4": "111", "v5": "222", "v6": "333"},
host: "aaa.bbb.ccc",
path: "/111/222/333",
shouldMatch: true,
},
{
title: "Host and Path route with multiple patterns, URL in request does not match",
route: new(Route).Host("{v1:[a-z]{3}}.{v2:[a-z]{3}}.{v3:[a-z]{3}}").Path("/{v4:[0-9]{3}}/{v5:[0-9]{3}}/{v6:[0-9]{3}}"),
request: newRequest("GET", "http://aaa.222.ccc/111/222/333"),
vars: map[string]string{"v1": "aaa", "v2": "bbb", "v3": "ccc", "v4": "111", "v5": "222", "v6": "333"},
host: "aaa.bbb.ccc",
path: "/111/222/333",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestHeaders(t *testing.T) {
// newRequestHeaders creates a new request with a method, url, and headers
newRequestHeaders := func(method, url string, headers map[string]string) *http.Request {
req, err := http.NewRequest(method, url, nil)
if err != nil {
panic(err)
}
for k, v := range headers {
req.Header.Add(k, v)
}
return req
}
tests := []routeTest{
{
title: "Headers route, match",
route: new(Route).Headers("foo", "bar", "baz", "ding"),
request: newRequestHeaders("GET", "http://localhost", map[string]string{"foo": "bar", "baz": "ding"}),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Headers route, bad header values",
route: new(Route).Headers("foo", "bar", "baz", "ding"),
request: newRequestHeaders("GET", "http://localhost", map[string]string{"foo": "bar", "baz": "dong"}),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestMethods(t *testing.T) {
tests := []routeTest{
{
title: "Methods route, match GET",
route: new(Route).Methods("GET", "POST"),
request: newRequest("GET", "http://localhost"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Methods route, match POST",
route: new(Route).Methods("GET", "POST"),
request: newRequest("POST", "http://localhost"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Methods route, bad method",
route: new(Route).Methods("GET", "POST"),
request: newRequest("PUT", "http://localhost"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestQueries(t *testing.T) {
tests := []routeTest{
{
title: "Queries route, match",
route: new(Route).Queries("foo", "bar", "baz", "ding"),
request: newRequest("GET", "http://localhost?foo=bar&baz=ding"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Queries route, match with a query string",
route: new(Route).Host("www.example.com").Path("/api").Queries("foo", "bar", "baz", "ding"),
request: newRequest("GET", "http://www.example.com/api?foo=bar&baz=ding"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Queries route, match with a query string out of order",
route: new(Route).Host("www.example.com").Path("/api").Queries("foo", "bar", "baz", "ding"),
request: newRequest("GET", "http://www.example.com/api?baz=ding&foo=bar"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Queries route, bad query",
route: new(Route).Queries("foo", "bar", "baz", "ding"),
request: newRequest("GET", "http://localhost?foo=bar&baz=dong"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: false,
},
{
title: "Queries route with pattern, match",
route: new(Route).Queries("foo", "{v1}"),
request: newRequest("GET", "http://localhost?foo=bar"),
vars: map[string]string{"v1": "bar"},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Queries route with multiple patterns, match",
route: new(Route).Queries("foo", "{v1}", "baz", "{v2}"),
request: newRequest("GET", "http://localhost?foo=bar&baz=ding"),
vars: map[string]string{"v1": "bar", "v2": "ding"},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Queries route with regexp pattern, match",
route: new(Route).Queries("foo", "{v1:[0-9]+}"),
request: newRequest("GET", "http://localhost?foo=10"),
vars: map[string]string{"v1": "10"},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Queries route with regexp pattern, regexp does not match",
route: new(Route).Queries("foo", "{v1:[0-9]+}"),
request: newRequest("GET", "http://localhost?foo=a"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestSchemes(t *testing.T) {
tests := []routeTest{
// Schemes
{
title: "Schemes route, match https",
route: new(Route).Schemes("https", "ftp"),
request: newRequest("GET", "https://localhost"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Schemes route, match ftp",
route: new(Route).Schemes("https", "ftp"),
request: newRequest("GET", "ftp://localhost"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "Schemes route, bad scheme",
route: new(Route).Schemes("https", "ftp"),
request: newRequest("GET", "http://localhost"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestMatcherFunc(t *testing.T) {
m := func(r *http.Request, m *RouteMatch) bool {
if r.URL.Host == "aaa.bbb.ccc" {
return true
}
return false
}
tests := []routeTest{
{
title: "MatchFunc route, match",
route: new(Route).MatcherFunc(m),
request: newRequest("GET", "http://aaa.bbb.ccc"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: true,
},
{
title: "MatchFunc route, non-match",
route: new(Route).MatcherFunc(m),
request: newRequest("GET", "http://aaa.222.ccc"),
vars: map[string]string{},
host: "",
path: "",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestSubRouter(t *testing.T) {
subrouter1 := new(Route).Host("{v1:[a-z]+}.google.com").Subrouter()
subrouter2 := new(Route).PathPrefix("/foo/{v1}").Subrouter()
tests := []routeTest{
{
route: subrouter1.Path("/{v2:[a-z]+}"),
request: newRequest("GET", "http://aaa.google.com/bbb"),
vars: map[string]string{"v1": "aaa", "v2": "bbb"},
host: "aaa.google.com",
path: "/bbb",
shouldMatch: true,
},
{
route: subrouter1.Path("/{v2:[a-z]+}"),
request: newRequest("GET", "http://111.google.com/111"),
vars: map[string]string{"v1": "aaa", "v2": "bbb"},
host: "aaa.google.com",
path: "/bbb",
shouldMatch: false,
},
{
route: subrouter2.Path("/baz/{v2}"),
request: newRequest("GET", "http://localhost/foo/bar/baz/ding"),
vars: map[string]string{"v1": "bar", "v2": "ding"},
host: "",
path: "/foo/bar/baz/ding",
shouldMatch: true,
},
{
route: subrouter2.Path("/baz/{v2}"),
request: newRequest("GET", "http://localhost/foo/bar"),
vars: map[string]string{"v1": "bar", "v2": "ding"},
host: "",
path: "/foo/bar/baz/ding",
shouldMatch: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
func TestNamedRoutes(t *testing.T) {
r1 := NewRouter()
r1.NewRoute().Name("a")
r1.NewRoute().Name("b")
r1.NewRoute().Name("c")
r2 := r1.NewRoute().Subrouter()
r2.NewRoute().Name("d")
r2.NewRoute().Name("e")
r2.NewRoute().Name("f")
r3 := r2.NewRoute().Subrouter()
r3.NewRoute().Name("g")
r3.NewRoute().Name("h")
r3.NewRoute().Name("i")
if r1.namedRoutes == nil || len(r1.namedRoutes) != 9 {
t.Errorf("Expected 9 named routes, got %v", r1.namedRoutes)
} else if r1.Get("i") == nil {
t.Errorf("Subroute name not registered")
}
}
func TestStrictSlash(t *testing.T) {
r := NewRouter()
r.StrictSlash(true)
tests := []routeTest{
{
title: "Redirect path without slash",
route: r.NewRoute().Path("/111/"),
request: newRequest("GET", "http://localhost/111"),
vars: map[string]string{},
host: "",
path: "/111/",
shouldMatch: true,
shouldRedirect: true,
},
{
title: "Do not redirect path with slash",
route: r.NewRoute().Path("/111/"),
request: newRequest("GET", "http://localhost/111/"),
vars: map[string]string{},
host: "",
path: "/111/",
shouldMatch: true,
shouldRedirect: false,
},
{
title: "Redirect path with slash",
route: r.NewRoute().Path("/111"),
request: newRequest("GET", "http://localhost/111/"),
vars: map[string]string{},
host: "",
path: "/111",
shouldMatch: true,
shouldRedirect: true,
},
{
title: "Do not redirect path without slash",
route: r.NewRoute().Path("/111"),
request: newRequest("GET", "http://localhost/111"),
vars: map[string]string{},
host: "",
path: "/111",
shouldMatch: true,
shouldRedirect: false,
},
{
title: "Propagate StrictSlash to subrouters",
route: r.NewRoute().PathPrefix("/static/").Subrouter().Path("/images/"),
request: newRequest("GET", "http://localhost/static/images"),
vars: map[string]string{},
host: "",
path: "/static/images/",
shouldMatch: true,
shouldRedirect: true,
},
{
title: "Ignore StrictSlash for path prefix",
route: r.NewRoute().PathPrefix("/static/"),
request: newRequest("GET", "http://localhost/static/logo.png"),
vars: map[string]string{},
host: "",
path: "/static/",
shouldMatch: true,
shouldRedirect: false,
},
}
for _, test := range tests {
testRoute(t, test)
}
}
// ----------------------------------------------------------------------------
// Helpers
// ----------------------------------------------------------------------------
func getRouteTemplate(route *Route) string {
host, path := "none", "none"
if route.regexp != nil {
if route.regexp.host != nil {
host = route.regexp.host.template
}
if route.regexp.path != nil {
path = route.regexp.path.template
}
}
return fmt.Sprintf("Host: %v, Path: %v", host, path)
}
func testRoute(t *testing.T, test routeTest) {
request := test.request
route := test.route
vars := test.vars
shouldMatch := test.shouldMatch
host := test.host
path := test.path
url := test.host + test.path
shouldRedirect := test.shouldRedirect
var match RouteMatch
ok := route.Match(request, &match)
if ok != shouldMatch {
msg := "Should match"
if !shouldMatch {
msg = "Should not match"
}
t.Errorf("(%v) %v:\nRoute: %#v\nRequest: %#v\nVars: %v\n", test.title, msg, route, request, vars)
return
}
if shouldMatch {
if test.vars != nil && !stringMapEqual(test.vars, match.Vars) {
t.Errorf("(%v) Vars not equal: expected %v, got %v", test.title, vars, match.Vars)
return
}
if host != "" {
u, _ := test.route.URLHost(mapToPairs(match.Vars)...)
if host != u.Host {
t.Errorf("(%v) URLHost not equal: expected %v, got %v -- %v", test.title, host, u.Host, getRouteTemplate(route))
return
}
}
if path != "" {
u, _ := route.URLPath(mapToPairs(match.Vars)...)
if path != u.Path {
t.Errorf("(%v) URLPath not equal: expected %v, got %v -- %v", test.title, path, u.Path, getRouteTemplate(route))
return
}
}
if url != "" {
u, _ := route.URL(mapToPairs(match.Vars)...)
if url != u.Host+u.Path {
t.Errorf("(%v) URL not equal: expected %v, got %v -- %v", test.title, url, u.Host+u.Path, getRouteTemplate(route))
return
}
}
if shouldRedirect && match.Handler == nil {
t.Errorf("(%v) Did not redirect", test.title)
return
}
if !shouldRedirect && match.Handler != nil {
t.Errorf("(%v) Unexpected redirect", test.title)
return
}
}
}
// Tests that the context is cleared or not cleared properly depending on
// the configuration of the router
func TestKeepContext(t *testing.T) {
func1 := func(w http.ResponseWriter, r *http.Request) {}
r := NewRouter()
r.HandleFunc("/", func1).Name("func1")
req, _ := http.NewRequest("GET", "http://localhost/", nil)
context.Set(req, "t", 1)
res := new(http.ResponseWriter)
r.ServeHTTP(*res, req)
if _, ok := context.GetOk(req, "t"); ok {
t.Error("Context should have been cleared at end of request")
}
r.KeepContext = true
req, _ = http.NewRequest("GET", "http://localhost/", nil)
context.Set(req, "t", 1)
r.ServeHTTP(*res, req)
if _, ok := context.GetOk(req, "t"); !ok {
t.Error("Context should NOT have been cleared at end of request")
}
}
type TestA301ResponseWriter struct {
hh http.Header
status int
}
func (ho TestA301ResponseWriter) Header() http.Header {
return http.Header(ho.hh)
}
func (ho TestA301ResponseWriter) Write(b []byte) (int, error) {
return 0, nil
}
func (ho TestA301ResponseWriter) WriteHeader(code int) {
ho.status = code
}
func Test301Redirect(t *testing.T) {
m := make(http.Header)
func1 := func(w http.ResponseWriter, r *http.Request) {}
func2 := func(w http.ResponseWriter, r *http.Request) {}
r := NewRouter()
r.HandleFunc("/api/", func2).Name("func2")
r.HandleFunc("/", func1).Name("func1")
req, _ := http.NewRequest("GET", "http://localhost//api/?abc=def", nil)
res := TestA301ResponseWriter{
hh: m,
status: 0,
}
r.ServeHTTP(&res, req)
if "http://localhost/api/?abc=def" != res.hh["Location"][0] {
t.Errorf("Should have complete URL with query string")
}
}
// https://plus.google.com/101022900381697718949/posts/eWy6DjFJ6uW
func TestSubrouterHeader(t *testing.T) {
expected := "func1 response"
func1 := func(w http.ResponseWriter, r *http.Request) {
fmt.Fprint(w, expected)
}
func2 := func(http.ResponseWriter, *http.Request) {}
r := NewRouter()
s := r.Headers("SomeSpecialHeader", "").Subrouter()
s.HandleFunc("/", func1).Name("func1")
r.HandleFunc("/", func2).Name("func2")
req, _ := http.NewRequest("GET", "http://localhost/", nil)
req.Header.Add("SomeSpecialHeader", "foo")
match := new(RouteMatch)
matched := r.Match(req, match)
if !matched {
t.Errorf("Should match request")
}
if match.Route.GetName() != "func1" {
t.Errorf("Expecting func1 handler, got %s", match.Route.GetName())
}
resp := NewRecorder()
match.Handler.ServeHTTP(resp, req)
if resp.Body.String() != expected {
t.Errorf("Expecting %q", expected)
}
}
// mapToPairs converts a string map to a slice of string pairs
func mapToPairs(m map[string]string) []string {
var i int
p := make([]string, len(m)*2)
for k, v := range m {
p[i] = k
p[i+1] = v
i += 2
}
return p
}
// stringMapEqual checks the equality of two string maps
func stringMapEqual(m1, m2 map[string]string) bool {
nil1 := m1 == nil
nil2 := m2 == nil
if nil1 != nil2 || len(m1) != len(m2) {
return false
}
for k, v := range m1 {
if v != m2[k] {
return false
}
}
return true
}
// newRequest is a helper function to create a new request with a method and url
func newRequest(method, url string) *http.Request {
req, err := http.NewRequest(method, url, nil)
if err != nil {
panic(err)
}
return req
}

714
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/old_test.go generated vendored
View File

@ -1,714 +0,0 @@
// Old tests ported to Go1. This is a mess. Want to drop it one day.
// Copyright 2011 Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package mux
import (
"bytes"
"net/http"
"testing"
)
// ----------------------------------------------------------------------------
// ResponseRecorder
// ----------------------------------------------------------------------------
// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// ResponseRecorder is an implementation of http.ResponseWriter that
// records its mutations for later inspection in tests.
type ResponseRecorder struct {
Code int // the HTTP response code from WriteHeader
HeaderMap http.Header // the HTTP response headers
Body *bytes.Buffer // if non-nil, the bytes.Buffer to append written data to
Flushed bool
}
// NewRecorder returns an initialized ResponseRecorder.
func NewRecorder() *ResponseRecorder {
return &ResponseRecorder{
HeaderMap: make(http.Header),
Body: new(bytes.Buffer),
}
}
// DefaultRemoteAddr is the default remote address to return in RemoteAddr if
// an explicit DefaultRemoteAddr isn't set on ResponseRecorder.
const DefaultRemoteAddr = "1.2.3.4"
// Header returns the response headers.
func (rw *ResponseRecorder) Header() http.Header {
return rw.HeaderMap
}
// Write always succeeds and writes to rw.Body, if not nil.
func (rw *ResponseRecorder) Write(buf []byte) (int, error) {
if rw.Body != nil {
rw.Body.Write(buf)
}
if rw.Code == 0 {
rw.Code = http.StatusOK
}
return len(buf), nil
}
// WriteHeader sets rw.Code.
func (rw *ResponseRecorder) WriteHeader(code int) {
rw.Code = code
}
// Flush sets rw.Flushed to true.
func (rw *ResponseRecorder) Flush() {
rw.Flushed = true
}
// ----------------------------------------------------------------------------
func TestRouteMatchers(t *testing.T) {
var scheme, host, path, query, method string
var headers map[string]string
var resultVars map[bool]map[string]string
router := NewRouter()
router.NewRoute().Host("{var1}.google.com").
Path("/{var2:[a-z]+}/{var3:[0-9]+}").
Queries("foo", "bar").
Methods("GET").
Schemes("https").
Headers("x-requested-with", "XMLHttpRequest")
router.NewRoute().Host("www.{var4}.com").
PathPrefix("/foo/{var5:[a-z]+}/{var6:[0-9]+}").
Queries("baz", "ding").
Methods("POST").
Schemes("http").
Headers("Content-Type", "application/json")
reset := func() {
// Everything match.
scheme = "https"
host = "www.google.com"
path = "/product/42"
query = "?foo=bar"
method = "GET"
headers = map[string]string{"X-Requested-With": "XMLHttpRequest"}
resultVars = map[bool]map[string]string{
true: {"var1": "www", "var2": "product", "var3": "42"},
false: {},
}
}
reset2 := func() {
// Everything match.
scheme = "http"
host = "www.google.com"
path = "/foo/product/42/path/that/is/ignored"
query = "?baz=ding"
method = "POST"
headers = map[string]string{"Content-Type": "application/json"}
resultVars = map[bool]map[string]string{
true: {"var4": "google", "var5": "product", "var6": "42"},
false: {},
}
}
match := func(shouldMatch bool) {
url := scheme + "://" + host + path + query
request, _ := http.NewRequest(method, url, nil)
for key, value := range headers {
request.Header.Add(key, value)
}
var routeMatch RouteMatch
matched := router.Match(request, &routeMatch)
if matched != shouldMatch {
// Need better messages. :)
if matched {
t.Errorf("Should match.")
} else {
t.Errorf("Should not match.")
}
}
if matched {
currentRoute := routeMatch.Route
if currentRoute == nil {
t.Errorf("Expected a current route.")
}
vars := routeMatch.Vars
expectedVars := resultVars[shouldMatch]
if len(vars) != len(expectedVars) {
t.Errorf("Expected vars: %v Got: %v.", expectedVars, vars)
}
for name, value := range vars {
if expectedVars[name] != value {
t.Errorf("Expected vars: %v Got: %v.", expectedVars, vars)
}
}
}
}
// 1st route --------------------------------------------------------------
// Everything match.
reset()
match(true)
// Scheme doesn't match.
reset()
scheme = "http"
match(false)
// Host doesn't match.
reset()
host = "www.mygoogle.com"
match(false)
// Path doesn't match.
reset()
path = "/product/notdigits"
match(false)
// Query doesn't match.
reset()
query = "?foo=baz"
match(false)
// Method doesn't match.
reset()
method = "POST"
match(false)
// Header doesn't match.
reset()
headers = map[string]string{}
match(false)
// Everything match, again.
reset()
match(true)
// 2nd route --------------------------------------------------------------
// Everything match.
reset2()
match(true)
// Scheme doesn't match.
reset2()
scheme = "https"
match(false)
// Host doesn't match.
reset2()
host = "sub.google.com"
match(false)
// Path doesn't match.
reset2()
path = "/bar/product/42"
match(false)
// Query doesn't match.
reset2()
query = "?foo=baz"
match(false)
// Method doesn't match.
reset2()
method = "GET"
match(false)
// Header doesn't match.
reset2()
headers = map[string]string{}
match(false)
// Everything match, again.
reset2()
match(true)
}
type headerMatcherTest struct {
matcher headerMatcher
headers map[string]string
result bool
}
var headerMatcherTests = []headerMatcherTest{
{
matcher: headerMatcher(map[string]string{"x-requested-with": "XMLHttpRequest"}),
headers: map[string]string{"X-Requested-With": "XMLHttpRequest"},
result: true,
},
{
matcher: headerMatcher(map[string]string{"x-requested-with": ""}),
headers: map[string]string{"X-Requested-With": "anything"},
result: true,
},
{
matcher: headerMatcher(map[string]string{"x-requested-with": "XMLHttpRequest"}),
headers: map[string]string{},
result: false,
},
}
type hostMatcherTest struct {
matcher *Route
url string
vars map[string]string
result bool
}
var hostMatcherTests = []hostMatcherTest{
{
matcher: NewRouter().NewRoute().Host("{foo:[a-z][a-z][a-z]}.{bar:[a-z][a-z][a-z]}.{baz:[a-z][a-z][a-z]}"),
url: "http://abc.def.ghi/",
vars: map[string]string{"foo": "abc", "bar": "def", "baz": "ghi"},
result: true,
},
{
matcher: NewRouter().NewRoute().Host("{foo:[a-z][a-z][a-z]}.{bar:[a-z][a-z][a-z]}.{baz:[a-z][a-z][a-z]}"),
url: "http://a.b.c/",
vars: map[string]string{"foo": "abc", "bar": "def", "baz": "ghi"},
result: false,
},
}
type methodMatcherTest struct {
matcher methodMatcher
method string
result bool
}
var methodMatcherTests = []methodMatcherTest{
{
matcher: methodMatcher([]string{"GET", "POST", "PUT"}),
method: "GET",
result: true,
},
{
matcher: methodMatcher([]string{"GET", "POST", "PUT"}),
method: "POST",
result: true,
},
{
matcher: methodMatcher([]string{"GET", "POST", "PUT"}),
method: "PUT",
result: true,
},
{
matcher: methodMatcher([]string{"GET", "POST", "PUT"}),
method: "DELETE",
result: false,
},
}
type pathMatcherTest struct {
matcher *Route
url string
vars map[string]string
result bool
}
var pathMatcherTests = []pathMatcherTest{
{
matcher: NewRouter().NewRoute().Path("/{foo:[0-9][0-9][0-9]}/{bar:[0-9][0-9][0-9]}/{baz:[0-9][0-9][0-9]}"),
url: "http://localhost:8080/123/456/789",
vars: map[string]string{"foo": "123", "bar": "456", "baz": "789"},
result: true,
},
{
matcher: NewRouter().NewRoute().Path("/{foo:[0-9][0-9][0-9]}/{bar:[0-9][0-9][0-9]}/{baz:[0-9][0-9][0-9]}"),
url: "http://localhost:8080/1/2/3",
vars: map[string]string{"foo": "123", "bar": "456", "baz": "789"},
result: false,
},
}
type schemeMatcherTest struct {
matcher schemeMatcher
url string
result bool
}
var schemeMatcherTests = []schemeMatcherTest{
{
matcher: schemeMatcher([]string{"http", "https"}),
url: "http://localhost:8080/",
result: true,
},
{
matcher: schemeMatcher([]string{"http", "https"}),
url: "https://localhost:8080/",
result: true,
},
{
matcher: schemeMatcher([]string{"https"}),
url: "http://localhost:8080/",
result: false,
},
{
matcher: schemeMatcher([]string{"http"}),
url: "https://localhost:8080/",
result: false,
},
}
type urlBuildingTest struct {
route *Route
vars []string
url string
}
var urlBuildingTests = []urlBuildingTest{
{
route: new(Route).Host("foo.domain.com"),
vars: []string{},
url: "http://foo.domain.com",
},
{
route: new(Route).Host("{subdomain}.domain.com"),
vars: []string{"subdomain", "bar"},
url: "http://bar.domain.com",
},
{
route: new(Route).Host("foo.domain.com").Path("/articles"),
vars: []string{},
url: "http://foo.domain.com/articles",
},
{
route: new(Route).Path("/articles"),
vars: []string{},
url: "/articles",
},
{
route: new(Route).Path("/articles/{category}/{id:[0-9]+}"),
vars: []string{"category", "technology", "id", "42"},
url: "/articles/technology/42",
},
{
route: new(Route).Host("{subdomain}.domain.com").Path("/articles/{category}/{id:[0-9]+}"),
vars: []string{"subdomain", "foo", "category", "technology", "id", "42"},
url: "http://foo.domain.com/articles/technology/42",
},
}
func TestHeaderMatcher(t *testing.T) {
for _, v := range headerMatcherTests {
request, _ := http.NewRequest("GET", "http://localhost:8080/", nil)
for key, value := range v.headers {
request.Header.Add(key, value)
}
var routeMatch RouteMatch
result := v.matcher.Match(request, &routeMatch)
if result != v.result {
if v.result {
t.Errorf("%#v: should match %v.", v.matcher, request.Header)
} else {
t.Errorf("%#v: should not match %v.", v.matcher, request.Header)
}
}
}
}
func TestHostMatcher(t *testing.T) {
for _, v := range hostMatcherTests {
request, _ := http.NewRequest("GET", v.url, nil)
var routeMatch RouteMatch
result := v.matcher.Match(request, &routeMatch)
vars := routeMatch.Vars
if result != v.result {
if v.result {
t.Errorf("%#v: should match %v.", v.matcher, v.url)
} else {
t.Errorf("%#v: should not match %v.", v.matcher, v.url)
}
}
if result {
if len(vars) != len(v.vars) {
t.Errorf("%#v: vars length should be %v, got %v.", v.matcher, len(v.vars), len(vars))
}
for name, value := range vars {
if v.vars[name] != value {
t.Errorf("%#v: expected value %v for key %v, got %v.", v.matcher, v.vars[name], name, value)
}
}
} else {
if len(vars) != 0 {
t.Errorf("%#v: vars length should be 0, got %v.", v.matcher, len(vars))
}
}
}
}
func TestMethodMatcher(t *testing.T) {
for _, v := range methodMatcherTests {
request, _ := http.NewRequest(v.method, "http://localhost:8080/", nil)
var routeMatch RouteMatch
result := v.matcher.Match(request, &routeMatch)
if result != v.result {
if v.result {
t.Errorf("%#v: should match %v.", v.matcher, v.method)
} else {
t.Errorf("%#v: should not match %v.", v.matcher, v.method)
}
}
}
}
func TestPathMatcher(t *testing.T) {
for _, v := range pathMatcherTests {
request, _ := http.NewRequest("GET", v.url, nil)
var routeMatch RouteMatch
result := v.matcher.Match(request, &routeMatch)
vars := routeMatch.Vars
if result != v.result {
if v.result {
t.Errorf("%#v: should match %v.", v.matcher, v.url)
} else {
t.Errorf("%#v: should not match %v.", v.matcher, v.url)
}
}
if result {
if len(vars) != len(v.vars) {
t.Errorf("%#v: vars length should be %v, got %v.", v.matcher, len(v.vars), len(vars))
}
for name, value := range vars {
if v.vars[name] != value {
t.Errorf("%#v: expected value %v for key %v, got %v.", v.matcher, v.vars[name], name, value)
}
}
} else {
if len(vars) != 0 {
t.Errorf("%#v: vars length should be 0, got %v.", v.matcher, len(vars))
}
}
}
}
func TestSchemeMatcher(t *testing.T) {
for _, v := range schemeMatcherTests {
request, _ := http.NewRequest("GET", v.url, nil)
var routeMatch RouteMatch
result := v.matcher.Match(request, &routeMatch)
if result != v.result {
if v.result {
t.Errorf("%#v: should match %v.", v.matcher, v.url)
} else {
t.Errorf("%#v: should not match %v.", v.matcher, v.url)
}
}
}
}
func TestUrlBuilding(t *testing.T) {
for _, v := range urlBuildingTests {
u, _ := v.route.URL(v.vars...)
url := u.String()
if url != v.url {
t.Errorf("expected %v, got %v", v.url, url)
/*
reversePath := ""
reverseHost := ""
if v.route.pathTemplate != nil {
reversePath = v.route.pathTemplate.Reverse
}
if v.route.hostTemplate != nil {
reverseHost = v.route.hostTemplate.Reverse
}
t.Errorf("%#v:\nexpected: %q\ngot: %q\nreverse path: %q\nreverse host: %q", v.route, v.url, url, reversePath, reverseHost)
*/
}
}
ArticleHandler := func(w http.ResponseWriter, r *http.Request) {
}
router := NewRouter()
router.HandleFunc("/articles/{category}/{id:[0-9]+}", ArticleHandler).Name("article")
url, _ := router.Get("article").URL("category", "technology", "id", "42")
expected := "/articles/technology/42"
if url.String() != expected {
t.Errorf("Expected %v, got %v", expected, url.String())
}
}
func TestMatchedRouteName(t *testing.T) {
routeName := "stock"
router := NewRouter()
route := router.NewRoute().Path("/products/").Name(routeName)
url := "http://www.domain.com/products/"
request, _ := http.NewRequest("GET", url, nil)
var rv RouteMatch
ok := router.Match(request, &rv)
if !ok || rv.Route != route {
t.Errorf("Expected same route, got %+v.", rv.Route)
}
retName := rv.Route.GetName()
if retName != routeName {
t.Errorf("Expected %q, got %q.", routeName, retName)
}
}
func TestSubRouting(t *testing.T) {
// Example from docs.
router := NewRouter()
subrouter := router.NewRoute().Host("www.domain.com").Subrouter()
route := subrouter.NewRoute().Path("/products/").Name("products")
url := "http://www.domain.com/products/"
request, _ := http.NewRequest("GET", url, nil)
var rv RouteMatch
ok := router.Match(request, &rv)
if !ok || rv.Route != route {
t.Errorf("Expected same route, got %+v.", rv.Route)
}
u, _ := router.Get("products").URL()
builtUrl := u.String()
// Yay, subroute aware of the domain when building!
if builtUrl != url {
t.Errorf("Expected %q, got %q.", url, builtUrl)
}
}
func TestVariableNames(t *testing.T) {
route := new(Route).Host("{arg1}.domain.com").Path("/{arg1}/{arg2:[0-9]+}")
if route.err == nil {
t.Errorf("Expected error for duplicated variable names")
}
}
func TestRedirectSlash(t *testing.T) {
var route *Route
var routeMatch RouteMatch
r := NewRouter()
r.StrictSlash(false)
route = r.NewRoute()
if route.strictSlash != false {
t.Errorf("Expected false redirectSlash.")
}
r.StrictSlash(true)
route = r.NewRoute()
if route.strictSlash != true {
t.Errorf("Expected true redirectSlash.")
}
route = new(Route)
route.strictSlash = true
route.Path("/{arg1}/{arg2:[0-9]+}/")
request, _ := http.NewRequest("GET", "http://localhost/foo/123", nil)
routeMatch = RouteMatch{}
_ = route.Match(request, &routeMatch)
vars := routeMatch.Vars
if vars["arg1"] != "foo" {
t.Errorf("Expected foo.")
}
if vars["arg2"] != "123" {
t.Errorf("Expected 123.")
}
rsp := NewRecorder()
routeMatch.Handler.ServeHTTP(rsp, request)
if rsp.HeaderMap.Get("Location") != "http://localhost/foo/123/" {
t.Errorf("Expected redirect header.")
}
route = new(Route)
route.strictSlash = true
route.Path("/{arg1}/{arg2:[0-9]+}")
request, _ = http.NewRequest("GET", "http://localhost/foo/123/", nil)
routeMatch = RouteMatch{}
_ = route.Match(request, &routeMatch)
vars = routeMatch.Vars
if vars["arg1"] != "foo" {
t.Errorf("Expected foo.")
}
if vars["arg2"] != "123" {
t.Errorf("Expected 123.")
}
rsp = NewRecorder()
routeMatch.Handler.ServeHTTP(rsp, request)
if rsp.HeaderMap.Get("Location") != "http://localhost/foo/123" {
t.Errorf("Expected redirect header.")
}
}
// Test for the new regexp library, still not available in stable Go.
func TestNewRegexp(t *testing.T) {
var p *routeRegexp
var matches []string
tests := map[string]map[string][]string{
"/{foo:a{2}}": {
"/a": nil,
"/aa": {"aa"},
"/aaa": nil,
"/aaaa": nil,
},
"/{foo:a{2,}}": {
"/a": nil,
"/aa": {"aa"},
"/aaa": {"aaa"},
"/aaaa": {"aaaa"},
},
"/{foo:a{2,3}}": {
"/a": nil,
"/aa": {"aa"},
"/aaa": {"aaa"},
"/aaaa": nil,
},
"/{foo:[a-z]{3}}/{bar:[a-z]{2}}": {
"/a": nil,
"/ab": nil,
"/abc": nil,
"/abcd": nil,
"/abc/ab": {"abc", "ab"},
"/abc/abc": nil,
"/abcd/ab": nil,
},
`/{foo:\w{3,}}/{bar:\d{2,}}`: {
"/a": nil,
"/ab": nil,
"/abc": nil,
"/abc/1": nil,
"/abc/12": {"abc", "12"},
"/abcd/12": {"abcd", "12"},
"/abcd/123": {"abcd", "123"},
},
}
for pattern, paths := range tests {
p, _ = newRouteRegexp(pattern, false, false, false, false)
for path, result := range paths {
matches = p.regexp.FindStringSubmatch(path)
if result == nil {
if matches != nil {
t.Errorf("%v should not match %v.", pattern, path)
}
} else {
if len(matches) != len(result)+1 {
t.Errorf("Expected %v matches, got %v.", len(result)+1, len(matches))
} else {
for k, v := range result {
if matches[k+1] != v {
t.Errorf("Expected %v, got %v.", v, matches[k+1])
}
}
}
}
}
}
}

276
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/regexp.go generated vendored
View File

@ -1,276 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package mux
import (
"bytes"
"fmt"
"net/http"
"net/url"
"regexp"
"strings"
)
// newRouteRegexp parses a route template and returns a routeRegexp,
// used to match a host, a path or a query string.
//
// It will extract named variables, assemble a regexp to be matched, create
// a "reverse" template to build URLs and compile regexps to validate variable
// values used in URL building.
//
// Previously we accepted only Python-like identifiers for variable
// names ([a-zA-Z_][a-zA-Z0-9_]*), but currently the only restriction is that
// name and pattern can't be empty, and names can't contain a colon.
func newRouteRegexp(tpl string, matchHost, matchPrefix, matchQuery, strictSlash bool) (*routeRegexp, error) {
// Check if it is well-formed.
idxs, errBraces := braceIndices(tpl)
if errBraces != nil {
return nil, errBraces
}
// Backup the original.
template := tpl
// Now let's parse it.
defaultPattern := "[^/]+"
if matchQuery {
defaultPattern = "[^?&]+"
matchPrefix = true
} else if matchHost {
defaultPattern = "[^.]+"
matchPrefix = false
}
// Only match strict slash if not matching
if matchPrefix || matchHost || matchQuery {
strictSlash = false
}
// Set a flag for strictSlash.
endSlash := false
if strictSlash && strings.HasSuffix(tpl, "/") {
tpl = tpl[:len(tpl)-1]
endSlash = true
}
varsN := make([]string, len(idxs)/2)
varsR := make([]*regexp.Regexp, len(idxs)/2)
pattern := bytes.NewBufferString("")
if !matchQuery {
pattern.WriteByte('^')
}
reverse := bytes.NewBufferString("")
var end int
var err error
for i := 0; i < len(idxs); i += 2 {
// Set all values we are interested in.
raw := tpl[end:idxs[i]]
end = idxs[i+1]
parts := strings.SplitN(tpl[idxs[i]+1:end-1], ":", 2)
name := parts[0]
patt := defaultPattern
if len(parts) == 2 {
patt = parts[1]
}
// Name or pattern can't be empty.
if name == "" || patt == "" {
return nil, fmt.Errorf("mux: missing name or pattern in %q",
tpl[idxs[i]:end])
}
// Build the regexp pattern.
fmt.Fprintf(pattern, "%s(%s)", regexp.QuoteMeta(raw), patt)
// Build the reverse template.
fmt.Fprintf(reverse, "%s%%s", raw)
// Append variable name and compiled pattern.
varsN[i/2] = name
varsR[i/2], err = regexp.Compile(fmt.Sprintf("^%s$", patt))
if err != nil {
return nil, err
}
}
// Add the remaining.
raw := tpl[end:]
pattern.WriteString(regexp.QuoteMeta(raw))
if strictSlash {
pattern.WriteString("[/]?")
}
if !matchPrefix {
pattern.WriteByte('$')
}
reverse.WriteString(raw)
if endSlash {
reverse.WriteByte('/')
}
// Compile full regexp.
reg, errCompile := regexp.Compile(pattern.String())
if errCompile != nil {
return nil, errCompile
}
// Done!
return &routeRegexp{
template: template,
matchHost: matchHost,
matchQuery: matchQuery,
strictSlash: strictSlash,
regexp: reg,
reverse: reverse.String(),
varsN: varsN,
varsR: varsR,
}, nil
}
// routeRegexp stores a regexp to match a host or path and information to
// collect and validate route variables.
type routeRegexp struct {
// The unmodified template.
template string
// True for host match, false for path or query string match.
matchHost bool
// True for query string match, false for path and host match.
matchQuery bool
// The strictSlash value defined on the route, but disabled if PathPrefix was used.
strictSlash bool
// Expanded regexp.
regexp *regexp.Regexp
// Reverse template.
reverse string
// Variable names.
varsN []string
// Variable regexps (validators).
varsR []*regexp.Regexp
}
// Match matches the regexp against the URL host or path.
func (r *routeRegexp) Match(req *http.Request, match *RouteMatch) bool {
if !r.matchHost {
if r.matchQuery {
return r.regexp.MatchString(req.URL.RawQuery)
} else {
return r.regexp.MatchString(req.URL.Path)
}
}
return r.regexp.MatchString(getHost(req))
}
// url builds a URL part using the given values.
func (r *routeRegexp) url(pairs ...string) (string, error) {
values, err := mapFromPairs(pairs...)
if err != nil {
return "", err
}
urlValues := make([]interface{}, len(r.varsN))
for k, v := range r.varsN {
value, ok := values[v]
if !ok {
return "", fmt.Errorf("mux: missing route variable %q", v)
}
urlValues[k] = value
}
rv := fmt.Sprintf(r.reverse, urlValues...)
if !r.regexp.MatchString(rv) {
// The URL is checked against the full regexp, instead of checking
// individual variables. This is faster but to provide a good error
// message, we check individual regexps if the URL doesn't match.
for k, v := range r.varsN {
if !r.varsR[k].MatchString(values[v]) {
return "", fmt.Errorf(
"mux: variable %q doesn't match, expected %q", values[v],
r.varsR[k].String())
}
}
}
return rv, nil
}
// braceIndices returns the first level curly brace indices from a string.
// It returns an error in case of unbalanced braces.
func braceIndices(s string) ([]int, error) {
var level, idx int
idxs := make([]int, 0)
for i := 0; i < len(s); i++ {
switch s[i] {
case '{':
if level++; level == 1 {
idx = i
}
case '}':
if level--; level == 0 {
idxs = append(idxs, idx, i+1)
} else if level < 0 {
return nil, fmt.Errorf("mux: unbalanced braces in %q", s)
}
}
}
if level != 0 {
return nil, fmt.Errorf("mux: unbalanced braces in %q", s)
}
return idxs, nil
}
// ----------------------------------------------------------------------------
// routeRegexpGroup
// ----------------------------------------------------------------------------
// routeRegexpGroup groups the route matchers that carry variables.
type routeRegexpGroup struct {
host *routeRegexp
path *routeRegexp
queries []*routeRegexp
}
// setMatch extracts the variables from the URL once a route matches.
func (v *routeRegexpGroup) setMatch(req *http.Request, m *RouteMatch, r *Route) {
// Store host variables.
if v.host != nil {
hostVars := v.host.regexp.FindStringSubmatch(getHost(req))
if hostVars != nil {
for k, v := range v.host.varsN {
m.Vars[v] = hostVars[k+1]
}
}
}
// Store path variables.
if v.path != nil {
pathVars := v.path.regexp.FindStringSubmatch(req.URL.Path)
if pathVars != nil {
for k, v := range v.path.varsN {
m.Vars[v] = pathVars[k+1]
}
// Check if we should redirect.
if v.path.strictSlash {
p1 := strings.HasSuffix(req.URL.Path, "/")
p2 := strings.HasSuffix(v.path.template, "/")
if p1 != p2 {
u, _ := url.Parse(req.URL.String())
if p1 {
u.Path = u.Path[:len(u.Path)-1]
} else {
u.Path += "/"
}
m.Handler = http.RedirectHandler(u.String(), 301)
}
}
}
}
// Store query string variables.
rawQuery := req.URL.RawQuery
for _, q := range v.queries {
queryVars := q.regexp.FindStringSubmatch(rawQuery)
if queryVars != nil {
for k, v := range q.varsN {
m.Vars[v] = queryVars[k+1]
}
}
}
}
// getHost tries its best to return the request host.
func getHost(r *http.Request) string {
if r.URL.IsAbs() {
return r.URL.Host
}
host := r.Host
// Slice off any port information.
if i := strings.Index(host, ":"); i != -1 {
host = host[:i]
}
return host
}

524
pkg/api/Godeps/_workspace/src/github.com/gorilla/mux/route.go generated vendored
View File

@ -1,524 +0,0 @@
// Copyright 2012 The Gorilla Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package mux
import (
"errors"
"fmt"
"net/http"
"net/url"
"strings"
)
// Route stores information to match a request and build URLs.
type Route struct {
// Parent where the route was registered (a Router).
parent parentRoute
// Request handler for the route.
handler http.Handler
// List of matchers.
matchers []matcher
// Manager for the variables from host and path.
regexp *routeRegexpGroup
// If true, when the path pattern is "/path/", accessing "/path" will
// redirect to the former and vice versa.
strictSlash bool
// If true, this route never matches: it is only used to build URLs.
buildOnly bool
// The name used to build URLs.
name string
// Error resulted from building a route.
err error
}
// Match matches the route against the request.
func (r *Route) Match(req *http.Request, match *RouteMatch) bool {
if r.buildOnly || r.err != nil {
return false
}
// Match everything.
for _, m := range r.matchers {
if matched := m.Match(req, match); !matched {
return false
}
}
// Yay, we have a match. Let's collect some info about it.
if match.Route == nil {
match.Route = r
}
if match.Handler == nil {
match.Handler = r.handler
}
if match.Vars == nil {
match.Vars = make(map[string]string)
}
// Set variables.
if r.regexp != nil {
r.regexp.setMatch(req, match, r)
}
return true
}
// ----------------------------------------------------------------------------
// Route attributes
// ----------------------------------------------------------------------------
// GetError returns an error resulted from building the route, if any.
func (r *Route) GetError() error {
return r.err
}
// BuildOnly sets the route to never match: it is only used to build URLs.
func (r *Route) BuildOnly() *Route {
r.buildOnly = true
return r
}
// Handler --------------------------------------------------------------------
// Handler sets a handler for the route.
func (r *Route) Handler(handler http.Handler) *Route {
if r.err == nil {
r.handler = handler
}
return r
}
// HandlerFunc sets a handler function for the route.
func (r *Route) HandlerFunc(f func(http.ResponseWriter, *http.Request)) *Route {
return r.Handler(http.HandlerFunc(f))
}
// GetHandler returns the handler for the route, if any.
func (r *Route) GetHandler() http.Handler {
return r.handler
}
// Name -----------------------------------------------------------------------
// Name sets the name for the route, used to build URLs.
// If the name was registered already it will be overwritten.
func (r *Route) Name(name string) *Route {
if r.name != "" {
r.err = fmt.Errorf("mux: route already has name %q, can't set %q",
r.name, name)
}
if r.err == nil {
r.name = name
r.getNamedRoutes()[name] = r
}
return r
}
// GetName returns the name for the route, if any.
func (r *Route) GetName() string {
return r.name
}
// ----------------------------------------------------------------------------
// Matchers
// ----------------------------------------------------------------------------
// matcher types try to match a request.
type matcher interface {
Match(*http.Request, *RouteMatch) bool
}
// addMatcher adds a matcher to the route.
func (r *Route) addMatcher(m matcher) *Route {
if r.err == nil {
r.matchers = append(r.matchers, m)
}
return r
}
// addRegexpMatcher adds a host or path matcher and builder to a route.
func (r *Route) addRegexpMatcher(tpl string, matchHost, matchPrefix, matchQuery bool) error {
if r.err != nil {
return r.err
}
r.regexp = r.getRegexpGroup()
if !matchHost && !matchQuery {
if len(tpl) == 0 || tpl[0] != '/' {
return fmt.Errorf("mux: path must start with a slash, got %q", tpl)
}
if r.regexp.path != nil {
tpl = strings.TrimRight(r.regexp.path.template, "/") + tpl
}
}
rr, err := newRouteRegexp(tpl, matchHost, matchPrefix, matchQuery, r.strictSlash)
if err != nil {
return err
}
for _, q := range r.regexp.queries {
if err = uniqueVars(rr.varsN, q.varsN); err != nil {
return err
}
}
if matchHost {
if r.regexp.path != nil {
if err = uniqueVars(rr.varsN, r.regexp.path.varsN); err != nil {
return err
}
}
r.regexp.host = rr
} else {
if r.regexp.host != nil {
if err = uniqueVars(rr.varsN, r.regexp.host.varsN); err != nil {
return err
}
}
if matchQuery {
r.regexp.queries = append(r.regexp.queries, rr)
} else {
r.regexp.path = rr
}
}
r.addMatcher(rr)
return nil
}
// Headers --------------------------------------------------------------------
// headerMatcher matches the request against header values.
type headerMatcher map[string]string
func (m headerMatcher) Match(r *http.Request, match *RouteMatch) bool {
return matchMap(m, r.Header, true)
}
// Headers adds a matcher for request header values.
// It accepts a sequence of key/value pairs to be matched. For example:
//
// r := mux.NewRouter()
// r.Headers("Content-Type", "application/json",
// "X-Requested-With", "XMLHttpRequest")
//
// The above route will only match if both request header values match.
//
// It the value is an empty string, it will match any value if the key is set.
func (r *Route) Headers(pairs ...string) *Route {
if r.err == nil {
var headers map[string]string
headers, r.err = mapFromPairs(pairs...)
return r.addMatcher(headerMatcher(headers))
}
return r
}
// Host -----------------------------------------------------------------------
// Host adds a matcher for the URL host.
// It accepts a template with zero or more URL variables enclosed by {}.
// Variables can define an optional regexp pattern to me matched:
//
// - {name} matches anything until the next dot.
//
// - {name:pattern} matches the given regexp pattern.
//
// For example:
//
// r := mux.NewRouter()
// r.Host("www.domain.com")
// r.Host("{subdomain}.domain.com")
// r.Host("{subdomain:[a-z]+}.domain.com")
//
// Variable names must be unique in a given route. They can be retrieved
// calling mux.Vars(request).
func (r *Route) Host(tpl string) *Route {
r.err = r.addRegexpMatcher(tpl, true, false, false)
return r
}
// MatcherFunc ----------------------------------------------------------------
// MatcherFunc is the function signature used by custom matchers.
type MatcherFunc func(*http.Request, *RouteMatch) bool
func (m MatcherFunc) Match(r *http.Request, match *RouteMatch) bool {
return m(r, match)
}
// MatcherFunc adds a custom function to be used as request matcher.
func (r *Route) MatcherFunc(f MatcherFunc) *Route {
return r.addMatcher(f)
}
// Methods --------------------------------------------------------------------
// methodMatcher matches the request against HTTP methods.
type methodMatcher []string
func (m methodMatcher) Match(r *http.Request, match *RouteMatch) bool {
return matchInArray(m, r.Method)
}
// Methods adds a matcher for HTTP methods.
// It accepts a sequence of one or more methods to be matched, e.g.:
// "GET", "POST", "PUT".
func (r *Route) Methods(methods ...string) *Route {
for k, v := range methods {
methods[k] = strings.ToUpper(v)
}
return r.addMatcher(methodMatcher(methods))
}
// Path -----------------------------------------------------------------------
// Path adds a matcher for the URL path.
// It accepts a template with zero or more URL variables enclosed by {}. The
// template must start with a "/".
// Variables can define an optional regexp pattern to me matched:
//
// - {name} matches anything until the next slash.
//
// - {name:pattern} matches the given regexp pattern.
//
// For example:
//
// r := mux.NewRouter()
// r.Path("/products/").Handler(ProductsHandler)
// r.Path("/products/{key}").Handler(ProductsHandler)
// r.Path("/articles/{category}/{id:[0-9]+}").
// Handler(ArticleHandler)
//
// Variable names must be unique in a given route. They can be retrieved
// calling mux.Vars(request).
func (r *Route) Path(tpl string) *Route {
r.err = r.addRegexpMatcher(tpl, false, false, false)
return r
}
// PathPrefix -----------------------------------------------------------------
// PathPrefix adds a matcher for the URL path prefix. This matches if the given
// template is a prefix of the full URL path. See Route.Path() for details on
// the tpl argument.
//
// Note that it does not treat slashes specially ("/foobar/" will be matched by
// the prefix "/foo") so you may want to use a trailing slash here.
//
// Also note that the setting of Router.StrictSlash() has no effect on routes
// with a PathPrefix matcher.
func (r *Route) PathPrefix(tpl string) *Route {
r.err = r.addRegexpMatcher(tpl, false, true, false)
return r
}
// Query ----------------------------------------------------------------------
// Queries adds a matcher for URL query values.
// It accepts a sequence of key/value pairs. Values may define variables.
// For example:
//
// r := mux.NewRouter()
// r.Queries("foo", "bar", "id", "{id:[0-9]+}")
//
// The above route will only match if the URL contains the defined queries
// values, e.g.: ?foo=bar&id=42.
//
// It the value is an empty string, it will match any value if the key is set.
//
// Variables can define an optional regexp pattern to me matched:
//
// - {name} matches anything until the next slash.
//
// - {name:pattern} matches the given regexp pattern.
func (r *Route) Queries(pairs ...string) *Route {
length := len(pairs)
if length%2 != 0 {
r.err = fmt.Errorf(
"mux: number of parameters must be multiple of 2, got %v", pairs)
return nil
}
for i := 0; i < length; i += 2 {
if r.err = r.addRegexpMatcher(pairs[i]+"="+pairs[i+1], false, true, true); r.err != nil {
return r
}
}
return r
}
// Schemes --------------------------------------------------------------------
// schemeMatcher matches the request against URL schemes.
type schemeMatcher []string
func (m schemeMatcher) Match(r *http.Request, match *RouteMatch) bool {
return matchInArray(m, r.URL.Scheme)
}
// Schemes adds a matcher for URL schemes.
// It accepts a sequence of schemes to be matched, e.g.: "http", "https".
func (r *Route) Schemes(schemes ...string) *Route {
for k, v := range schemes {
schemes[k] = strings.ToLower(v)
}
return r.addMatcher(schemeMatcher(schemes))
}
// Subrouter ------------------------------------------------------------------
// Subrouter creates a subrouter for the route.
//
// It will test the inner routes only if the parent route matched. For example:
//
// r := mux.NewRouter()
// s := r.Host("www.domain.com").Subrouter()
// s.HandleFunc("/products/", ProductsHandler)
// s.HandleFunc("/products/{key}", ProductHandler)
// s.HandleFunc("/articles/{category}/{id:[0-9]+}"), ArticleHandler)
//
// Here, the routes registered in the subrouter won't be tested if the host
// doesn't match.
func (r *Route) Subrouter() *Router {
router := &Router{parent: r, strictSlash: r.strictSlash}
r.addMatcher(router)
return router
}
// ----------------------------------------------------------------------------
// URL building
// ----------------------------------------------------------------------------
// URL builds a URL for the route.
//
// It accepts a sequence of key/value pairs for the route variables. For
// example, given this route:
//
// r := mux.NewRouter()
// r.HandleFunc("/articles/{category}/{id:[0-9]+}", ArticleHandler).
// Name("article")
//
// ...a URL for it can be built using:
//
// url, err := r.Get("article").URL("category", "technology", "id", "42")
//
// ...which will return an url.URL with the following path:
//
// "/articles/technology/42"
//
// This also works for host variables:
//
// r := mux.NewRouter()
// r.Host("{subdomain}.domain.com").
// HandleFunc("/articles/{category}/{id:[0-9]+}", ArticleHandler).
// Name("article")
//
// // url.String() will be "http://news.domain.com/articles/technology/42"
// url, err := r.Get("article").URL("subdomain", "news",
// "category", "technology",
// "id", "42")
//
// All variables defined in the route are required, and their values must
// conform to the corresponding patterns.
func (r *Route) URL(pairs ...string) (*url.URL, error) {
if r.err != nil {
return nil, r.err
}
if r.regexp == nil {
return nil, errors.New("mux: route doesn't have a host or path")
}
var scheme, host, path string
var err error
if r.regexp.host != nil {
// Set a default scheme.
scheme = "http"
if host, err = r.regexp.host.url(pairs...); err != nil {
return nil, err
}
}
if r.regexp.path != nil {
if path, err = r.regexp.path.url(pairs...); err != nil {
return nil, err
}
}
return &url.URL{
Scheme: scheme,
Host: host,
Path: path,
}, nil
}
// URLHost builds the host part of the URL for a route. See Route.URL().
//
// The route must have a host defined.
func (r *Route) URLHost(pairs ...string) (*url.URL, error) {
if r.err != nil {
return nil, r.err
}
if r.regexp == nil || r.regexp.host == nil {
return nil, errors.New("mux: route doesn't have a host")
}
host, err := r.regexp.host.url(pairs...)
if err != nil {
return nil, err
}
return &url.URL{
Scheme: "http",
Host: host,
}, nil
}
// URLPath builds the path part of the URL for a route. See Route.URL().
//
// The route must have a path defined.
func (r *Route) URLPath(pairs ...string) (*url.URL, error) {
if r.err != nil {
return nil, r.err
}
if r.regexp == nil || r.regexp.path == nil {
return nil, errors.New("mux: route doesn't have a path")
}
path, err := r.regexp.path.url(pairs...)
if err != nil {
return nil, err
}
return &url.URL{
Path: path,
}, nil
}
// ----------------------------------------------------------------------------
// parentRoute
// ----------------------------------------------------------------------------
// parentRoute allows routes to know about parent host and path definitions.
type parentRoute interface {
getNamedRoutes() map[string]*Route
getRegexpGroup() *routeRegexpGroup
}
// getNamedRoutes returns the map where named routes are registered.
func (r *Route) getNamedRoutes() map[string]*Route {
if r.parent == nil {
// During tests router is not always set.
r.parent = NewRouter()
}
return r.parent.getNamedRoutes()
}
// getRegexpGroup returns regexp definitions from this route.
func (r *Route) getRegexpGroup() *routeRegexpGroup {
if r.regexp == nil {
if r.parent == nil {
// During tests router is not always set.
r.parent = NewRouter()
}
regexp := r.parent.getRegexpGroup()
if regexp == nil {
r.regexp = new(routeRegexpGroup)
} else {
// Copy.
r.regexp = &routeRegexpGroup{
host: regexp.host,
path: regexp.path,
queries: regexp.queries,
}
}
}
return r.regexp
}

6
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/.gitignore generated vendored
View File

@ -1,6 +0,0 @@
*.o
*.a
*.so
*~
*.dSYM
*.syso

80
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/BUILDDEPS.md generated vendored
View File

@ -1,80 +0,0 @@
## Ubuntu (Kylin) 14.04
### Build Dependencies
This installation document assumes Ubuntu 14.04 or later on x86-64 platform.
##### Install YASM
Erasure depends on Intel ISAL library, ISAL uses Intel AVX2 processor instructions, to compile these files one needs to install ``yasm`` which supports AVX2 instructions. AVX2 support only ended in ``yasm`` from version ``1.2.0``, any version below ``1.2.0`` will throw a build error.
```sh
$ sudo apt-get install yasm
```
##### Install Go 1.4+
Download Go 1.4+ from [https://golang.org/dl/](https://golang.org/dl/) and extract it into ``${HOME}/local`` and setup ``${HOME}/mygo`` as your project workspace folder.
For example:
```sh
.... Extract and install golang ....
$ wget https://storage.googleapis.com/golang/go1.4.linux-amd64.tar.gz
$ mkdir -p ${HOME}/local
$ mkdir -p $HOME/mygo
$ tar -C ${HOME}/local -xzf go1.4.linux-amd64.tar.gz
.... Export necessary environment variables ....
$ export PATH=$PATH:${HOME}/local/go/bin
$ export GOROOT=${HOME}/local/go
$ export GOPATH=$HOME/mygo
$ export PATH=$PATH:$GOPATH/bin
.... Add paths to your bashrc ....
$ echo "export PATH=$PATH:${HOME}/local/go/bin" >> ${HOME}/.bashrc
$ echo "export GOROOT=${HOME}/local/go" >> ${HOME}/.bashrc
$ echo "export GOPATH=$HOME/mygo" >> ${HOME}/.bashrc
$ echo "export PATH=$PATH:$GOPATH/bin" >> ${HOME}/.bashrc
```
## Mac OSX (Yosemite) 10.10
### Build Dependencies
This installation document assumes Mac OSX Yosemite 10.10 or later on x86-64 platform.
##### Install brew
```sh
$ ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
```
##### Install Git
```sh
$ brew install git
```
##### Install YASM
Erasure depends on Intel ISAL library, ISAL uses Intel AVX2 processor instructions, to compile these files one needs to install ``yasm`` which supports AVX2 instructions. AVX2 support only ended in ``yasm`` from version ``1.2.0``, any version below ``1.2.0`` will throw a build error.
```sh
$ brew install yasm
```
##### Install Go 1.4+
On MacOSX ``brew.sh`` is the best way to install golang
For example:
```sh
.... Install golang using `brew` ....
$ brew install go
$ mkdir -p $HOME/mygo
.... Export necessary environment variables ....
$ export GOPATH=$HOME/mygo
$ export PATH=$PATH:$GOPATH/bin
.... Add paths to your bashrc ....
$ echo "export GOPATH=$HOME/mygo" >> ${HOME}/.bashrc
$ echo "export PATH=$PATH:$GOPATH/bin" >> ${HOME}/.bashrc
```

30
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/CONTRIBUTING.md generated vendored
View File

@ -1,30 +0,0 @@
### Setup your Erasure Github Repository
Fork [Erasure upstream](https://github.com/minio-io/erasure/fork) source repository to your own personal repository. Copy the URL and pass it to ``go get`` command. Go uses git to clone a copy into your project workspace folder.
```sh
$ git clone https://github.com/$USER_ID/erasure
$ cd erasure
$ mkdir -p ${GOPATH}/src/github.com/minio-io
$ ln -s ${PWD} $GOPATH/src/github.com/minio-io/
```
### Compiling Erasure from source
```sh
$ go generate
$ go build
```
### Developer Guidelines
To make the process as seamless as possible, we ask for the following:
* Go ahead and fork the project and make your changes. We encourage pull requests to discuss code changes.
- Fork it
- Create your feature branch (git checkout -b my-new-feature)
- Commit your changes (git commit -am 'Add some feature')
- Push to the branch (git push origin my-new-feature)
- Create new Pull Request
* When you're ready to create a pull request, be sure to:
- Have test cases for the new code. If you have questions about how to do it, please ask in your pull request.
- Run `go fmt`
- Squash your commits into a single commit. `git rebase -i`. It's okay to force update your pull request.
- Make sure `go test -race ./...` and `go build` completes.
* Read [Effective Go](https://github.com/golang/go/wiki/CodeReviewComments) article from Golang project
- `Erasure` project is strictly conformant with Golang style
- if you happen to observe offending code, please feel free to send a pull request

26
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/LICENSE.INTEL generated vendored
View File

@ -1,26 +0,0 @@
Copyright(c) 2011-2014 Intel Corporation All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

202
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/LICENSE.MINIO generated vendored
View File

@ -1,202 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

25
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/README.md generated vendored
View File

@ -1,25 +0,0 @@
## Introduction
Erasure is an open source Golang library written on top of ISAL (Intel Intelligent Storage Library) released under [Apache license v2](./LICENSE)
### Developers
* [Get Source](./CONTRIBUTING.md)
* [Build Dependencies](./BUILDDEPS.md)
* [Development Workflow](./CONTRIBUTING.md#developer-guidelines)
* [Developer discussions and bugs](https://github.com/Minio-io/erasure/issues)
### Supported platforms
| Name | Supported |
| ------------- | ------------- |
| Linux | Yes |
| Windows | Not yet |
| Mac OSX | Yes |
### Supported architectures
| Arch | Supported |
| ------------- | ------------- |
| x86-64 | Yes |
| arm64 | Not yet|
| i386 | Never |

49
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/RELEASE-NOTES.INTEL generated vendored
View File

@ -1,49 +0,0 @@
================================================================================
v2.10 Intel Intelligent Storage Acceleration Library Release Notes
Open Source Version
================================================================================
================================================================================
RELEASE NOTE CONTENTS
================================================================================
1. KNOWN ISSUES
2. FIXED ISSUES
3. CHANGE LOG & FEATURES ADDED
================================================================================
1. KNOWN ISSUES
================================================================================
* Only erasure code unit included in open source version at this time.
* Perf tests do not run in Windows environment.
* Leaving <unit>/bin directories from builds in unit directories will cause the
top-level make build to fail. Build only in top-level or ensure unit
directories are clean of objects and /bin.
* 32-bit lib is not supported in Windows.
================================================================================
2. FIXED ISSUES
================================================================================
v2.10
* Fix for windows register save overlap in gf_{3-6}vect_dot_prod_sse.asm. Only
affects windows versions of erasure code. GP register saves/restore were
pushed to same stack area as XMM.
================================================================================
3. CHANGE LOG & FEATURES ADDED
================================================================================
v2.10
* Erasure code updates
- New AVX and AVX2 support functions.
- Changes min len requirement on gf_vect_dot_prod() to 32 from 16.
- Tests include both source and parity recovery with ec_encode_data().
- New encoding examples with Vandermonde or Cauchy matrix.
v2.8
* First open release of erasure code unit that is part of ISA-L.

3
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/RELEASE-NOTES.MINIO generated vendored
View File

@ -1,3 +0,0 @@
v1.0 - Erasure Golang Package
============================
- First release, supports only amd64 or x86-64 architecture

71
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/cauchy_test.go generated vendored
View File

@ -1,71 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
import (
"bytes"
"testing"
. "gopkg.in/check.v1"
)
type MySuite struct{}
var _ = Suite(&MySuite{})
func Test(t *testing.T) { TestingT(t) }
const (
k = 10
m = 5
)
func (s *MySuite) TestCauchyEncodeDecodeFailure(c *C) {
ep, _ := ValidateParams(k, m, Cauchy)
data := []byte("Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry's standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum.")
e := NewErasure(ep)
chunks, err := e.Encode(data)
c.Assert(err, IsNil)
errorIndex := []int{0, 3, 5, 9, 11, 13}
chunks = corruptChunks(chunks, errorIndex)
_, err = e.Decode(chunks, len(data))
c.Assert(err, Not(IsNil))
}
func (s *MySuite) TestCauchyEncodeDecodeSuccess(c *C) {
ep, _ := ValidateParams(k, m, Cauchy)
data := []byte("Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry's standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum.")
e := NewErasure(ep)
chunks, err := e.Encode(data)
c.Assert(err, IsNil)
errorIndex := []int{0, 3, 5, 9, 13}
chunks = corruptChunks(chunks, errorIndex)
recoveredData, err := e.Decode(chunks, len(data))
c.Assert(err, IsNil)
if !bytes.Equal(data, recoveredData) {
c.Fatalf("Recovered data mismatches with original data")
}
}

59
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ctypes.go generated vendored
View File

@ -1,59 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
// #include <stdint.h>
import "C"
import (
"fmt"
"unsafe"
)
// intSlice2CIntArray converts Go int slice to C int array
func intSlice2CIntArray(srcErrList []int) *C.int32_t {
var sizeErrInt = int(unsafe.Sizeof(srcErrList[0]))
switch sizeInt {
case sizeErrInt:
return (*C.int32_t)(unsafe.Pointer(&srcErrList[0]))
case sizeInt8:
int8Array := make([]int8, len(srcErrList))
for i, v := range srcErrList {
int8Array[i] = int8(v)
}
return (*C.int32_t)(unsafe.Pointer(&int8Array[0]))
case sizeInt16:
int16Array := make([]int16, len(srcErrList))
for i, v := range srcErrList {
int16Array[i] = int16(v)
}
return (*C.int32_t)(unsafe.Pointer(&int16Array[0]))
case sizeInt32:
int32Array := make([]int32, len(srcErrList))
for i, v := range srcErrList {
int32Array[i] = int32(v)
}
return (*C.int32_t)(unsafe.Pointer(&int32Array[0]))
case sizeInt64:
int64Array := make([]int64, len(srcErrList))
for i, v := range srcErrList {
int64Array[i] = int64(v)
}
return (*C.int32_t)(unsafe.Pointer(&int64Array[0]))
default:
panic(fmt.Sprintf("Unsupported: %d", sizeInt))
}
}

67
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/doc.go generated vendored
View File

@ -1,67 +0,0 @@
// Package erasure is a Go wrapper for the Intel Intelligent Storage
// Acceleration Library (Intel ISA-L). Intel ISA-L is a CPU optimized
// implementation of erasure coding algorithms.
//
// For more information on Intel ISA-L, please visit:
// https://01.org/intel%C2%AE-storage-acceleration-library-open-source-version
//
// Usage:
//
// Encode encodes a block of data. The input is the original data. The output
// is a 2 tuple containing (k + m) chunks of erasure encoded data and the
// length of the original object.
//
// Decode decodes 2 tuple data containing (k + m) chunks back into its original form.
// Additionally original block length should also be provided as input.
//
// Decoded data is exactly similar in length and content as the original data.
//
// Encoding data may be performed in 3 steps.
//
// 1. Create a parse set of encoder parameters
// 2. Create a new encoder
// 3. Encode data
//
// Decoding data is also performed in 3 steps.
//
// 1. Create a parse set of encoder parameters for validation
// 2. Create a new encoder
// 3. Decode data
//
// Erasure parameters contain three configurable elements:
// ValidateParams(k, m, technique int) (ErasureParams, error)
// k - Number of rows in matrix
// m - Number of colums in matrix
// technique - Matrix type, can be either Cauchy (recommended) or Vandermonde
// constraints: k + m < Galois Field (2^8)
//
// Choosing right parity and matrix technique is left for application to decide.
//
// But here are the few points to keep in mind
//
// Techniques:
// - Vandermonde is most commonly used method for choosing coefficients in erasure
// encoding but does not guarantee invertable for every sub matrix.
// Users may want to adjust for k > 5. (k is data blocks)
// - Whereas Cauchy is our recommended method for choosing coefficients in erasure coding.
// Since any sub-matrix of a Cauchy matrix is invertable.
//
// Total blocks:
// - Data blocks and Parity blocks should not be greater than 'Galois Field' (2^8)
//
// Example
//
// Creating and using an encoder
// var bytes []byte
// params := erasure.ValidateParams(10, 5, erasure.Cauchy)
// encoder := erasure.NewErasure(params)
// encodedData, length := encoder.Encode(bytes)
//
// Creating and using a decoder
// var encodedData [][]byte
// var length int
// params := erasure.ValidateParams(10, 5, erasure.Cauchy)
// encoder := erasure.NewErasure(params)
// originalData, err := encoder.Decode(encodedData, length)
//
package erasure

5429
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/docs/isa-l_open_src_2.10.pdf generated vendored
View File

File diff suppressed because it is too large Load Diff

BIN
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/docs/isa-l_open_src_2.13.pdf generated vendored
View File

Binary file not shown.

348
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_base.c generated vendored
View File

@ -1,348 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**********************************************************************/
#include <limits.h>
#include <string.h> // for memset
#include "ec_code.h"
#include "ec_base.h" // for GF tables
#include "ec_types.h"
unsigned char gf_mul(unsigned char a, unsigned char b)
{
#ifndef GF_LARGE_TABLES
int i;
if ((a == 0) || (b == 0))
return 0;
return gff_base[(i = gflog_base[a] + gflog_base[b]) > 254 ? i - 255 : i];
#else
return gf_mul_table_base[b * 256 + a];
#endif
}
unsigned char gf_inv(unsigned char a)
{
#ifndef GF_LARGE_TABLES
if (a == 0)
return 0;
return gff_base[255 - gflog_base[a]];
#else
return gf_inv_table_base[a];
#endif
}
void gf_gen_rs_matrix(unsigned char *a, int m, int k)
{
int i, j;
unsigned char p, gen = 1;
memset(a, 0, k * m);
for (i = 0; i < k; i++)
a[k * i + i] = 1;
for (i = k; i < m; i++) {
p = 1;
for (j = 0; j < k; j++) {
a[k * i + j] = p;
p = gf_mul(p, gen);
}
gen = gf_mul(gen, 2);
}
}
void gf_gen_cauchy1_matrix(unsigned char *a, int m, int k)
{
int i, j;
unsigned char *p;
// Identity matrix in high position
memset(a, 0, k * m);
for (i = 0; i < k; i++)
a[k * i + i] = 1;
// For the rest choose 1/(i + j) | i != j
p = &a[k * k];
for (i = k; i < m; i++)
for (j = 0; j < k; j++)
*p++ = gf_inv(i ^ j);
}
int gf_invert_matrix(unsigned char *in_mat, unsigned char *out_mat, const int n)
{
int i, j, k;
unsigned char temp;
// Set out_mat[] to the identity matrix
for (i = 0; i < n * n; i++) // memset(out_mat, 0, n*n)
out_mat[i] = 0;
for (i = 0; i < n; i++)
out_mat[i * n + i] = 1;
// Inverse
for (i = 0; i < n; i++) {
// Check for 0 in pivot element
if (in_mat[i * n + i] == 0) {
// Find a row with non-zero in current column and swap
for (j = i + 1; j < n; j++)
if (in_mat[j * n + i])
break;
if (j == n) // Couldn't find means it's singular
return -1;
for (k = 0; k < n; k++) { // Swap rows i,j
temp = in_mat[i * n + k];
in_mat[i * n + k] = in_mat[j * n + k];
in_mat[j * n + k] = temp;
temp = out_mat[i * n + k];
out_mat[i * n + k] = out_mat[j * n + k];
out_mat[j * n + k] = temp;
}
}
temp = gf_inv(in_mat[i * n + i]); // 1/pivot
for (j = 0; j < n; j++) { // Scale row i by 1/pivot
in_mat[i * n + j] = gf_mul(in_mat[i * n + j], temp);
out_mat[i * n + j] = gf_mul(out_mat[i * n + j], temp);
}
for (j = 0; j < n; j++) {
if (j == i)
continue;
temp = in_mat[j * n + i];
for (k = 0; k < n; k++) {
out_mat[j * n + k] ^= gf_mul(temp, out_mat[i * n + k]);
in_mat[j * n + k] ^= gf_mul(temp, in_mat[i * n + k]);
}
}
}
return 0;
}
// Calculates const table gftbl in GF(2^8) from single input A
// gftbl(A) = {A{00}, A{01}, A{02}, ... , A{0f} }, {A{00}, A{10}, A{20}, ... , A{f0} }
void gf_vect_mul_init(unsigned char c, unsigned char *tbl)
{
unsigned char c2 = (c << 1) ^ ((c & 0x80) ? 0x1d : 0); //Mult by GF{2}
unsigned char c4 = (c2 << 1) ^ ((c2 & 0x80) ? 0x1d : 0); //Mult by GF{2}
unsigned char c8 = (c4 << 1) ^ ((c4 & 0x80) ? 0x1d : 0); //Mult by GF{2}
#if __WORDSIZE == 64 || _WIN64 || __x86_64__
unsigned long long v1, v2, v4, v8, *t;
unsigned long long v10, v20, v40, v80;
unsigned char c17, c18, c20, c24;
t = (unsigned long long *)tbl;
v1 = c * 0x0100010001000100ull;
v2 = c2 * 0x0101000001010000ull;
v4 = c4 * 0x0101010100000000ull;
v8 = c8 * 0x0101010101010101ull;
v4 = v1 ^ v2 ^ v4;
t[0] = v4;
t[1] = v8 ^ v4;
c17 = (c8 << 1) ^ ((c8 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c18 = (c17 << 1) ^ ((c17 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c20 = (c18 << 1) ^ ((c18 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c24 = (c20 << 1) ^ ((c20 & 0x80) ? 0x1d : 0); //Mult by GF{2}
v10 = c17 * 0x0100010001000100ull;
v20 = c18 * 0x0101000001010000ull;
v40 = c20 * 0x0101010100000000ull;
v80 = c24 * 0x0101010101010101ull;
v40 = v10 ^ v20 ^ v40;
t[2] = v40;
t[3] = v80 ^ v40;
#else // 32-bit or other
unsigned char c3, c5, c6, c7, c9, c10, c11, c12, c13, c14, c15;
unsigned char c17, c18, c19, c20, c21, c22, c23, c24, c25, c26, c27, c28, c29, c30,
c31;
c3 = c2 ^ c;
c5 = c4 ^ c;
c6 = c4 ^ c2;
c7 = c4 ^ c3;
c9 = c8 ^ c;
c10 = c8 ^ c2;
c11 = c8 ^ c3;
c12 = c8 ^ c4;
c13 = c8 ^ c5;
c14 = c8 ^ c6;
c15 = c8 ^ c7;
tbl[0] = 0;
tbl[1] = c;
tbl[2] = c2;
tbl[3] = c3;
tbl[4] = c4;
tbl[5] = c5;
tbl[6] = c6;
tbl[7] = c7;
tbl[8] = c8;
tbl[9] = c9;
tbl[10] = c10;
tbl[11] = c11;
tbl[12] = c12;
tbl[13] = c13;
tbl[14] = c14;
tbl[15] = c15;
c17 = (c8 << 1) ^ ((c8 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c18 = (c17 << 1) ^ ((c17 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c19 = c18 ^ c17;
c20 = (c18 << 1) ^ ((c18 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c21 = c20 ^ c17;
c22 = c20 ^ c18;
c23 = c20 ^ c19;
c24 = (c20 << 1) ^ ((c20 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c25 = c24 ^ c17;
c26 = c24 ^ c18;
c27 = c24 ^ c19;
c28 = c24 ^ c20;
c29 = c24 ^ c21;
c30 = c24 ^ c22;
c31 = c24 ^ c23;
tbl[16] = 0;
tbl[17] = c17;
tbl[18] = c18;
tbl[19] = c19;
tbl[20] = c20;
tbl[21] = c21;
tbl[22] = c22;
tbl[23] = c23;
tbl[24] = c24;
tbl[25] = c25;
tbl[26] = c26;
tbl[27] = c27;
tbl[28] = c28;
tbl[29] = c29;
tbl[30] = c30;
tbl[31] = c31;
#endif //__WORDSIZE == 64 || _WIN64 || __x86_64__
}
void gf_vect_dot_prod_base(int len, int vlen, unsigned char *v,
unsigned char **src, unsigned char *dest)
{
int i, j;
unsigned char s;
for (i = 0; i < len; i++) {
s = 0;
for (j = 0; j < vlen; j++)
s ^= gf_mul(src[j][i], v[j * 32 + 1]);
dest[i] = s;
}
}
void gf_vect_mad_base(int len, int vec, int vec_i,
unsigned char *v, unsigned char *src, unsigned char *dest)
{
int i;
unsigned char s;
for (i = 0; i < len; i++) {
s = dest[i];
s ^= gf_mul(src[i], v[vec_i * 32 + 1]);
dest[i] = s;
}
}
void ec_encode_data_base(int len, int srcs, int dests, unsigned char *v,
unsigned char **src, unsigned char **dest)
{
int i, j, l;
unsigned char s;
for (l = 0; l < dests; l++) {
for (i = 0; i < len; i++) {
s = 0;
for (j = 0; j < srcs; j++)
s ^= gf_mul(src[j][i], v[j * 32 + l * srcs * 32 + 1]);
dest[l][i] = s;
}
}
}
void ec_encode_data_update_base(int len, int k, int rows, int vec_i, unsigned char *v,
unsigned char *data, unsigned char **dest)
{
int i, l;
unsigned char s;
for (l = 0; l < rows; l++) {
for (i = 0; i < len; i++) {
s = dest[l][i];
s ^= gf_mul(data[i], v[vec_i * 32 + l * k * 32 + 1]);
dest[l][i] = s;
}
}
}
void gf_vect_mul_base(int len, unsigned char *a, unsigned char *src, unsigned char *dest)
{
//2nd element of table array is ref value used to fill it in
unsigned char c = a[1];
while (len-- > 0)
*dest++ = gf_mul(c, *src++);
}
struct slver {
UINT16 snum;
UINT8 ver;
UINT8 core;
};
// Version info
struct slver gf_vect_mul_init_slver_00020035;
struct slver gf_vect_mul_init_slver = { 0x0035, 0x02, 0x00 };
struct slver ec_encode_data_base_slver_00010135;
struct slver ec_encode_data_base_slver = { 0x0135, 0x01, 0x00 };
struct slver gf_vect_mul_base_slver_00010136;
struct slver gf_vect_mul_base_slver = { 0x0136, 0x01, 0x00 };
struct slver gf_vect_dot_prod_base_slver_00010137;
struct slver gf_vect_dot_prod_base_slver = { 0x0137, 0x01, 0x00 };

6680
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_base.h generated vendored
View File

File diff suppressed because it is too large Load Diff

933
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_code.h generated vendored
View File

@ -1,933 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**********************************************************************/
#ifndef _ERASURE_CODE_H_
#define _ERASURE_CODE_H_
/**
* @file erasure_code.h
* @brief Interface to functions supporting erasure code encode and decode.
*
* This file defines the interface to optimized functions used in erasure
* codes. Encode and decode of erasures in GF(2^8) are made by calculating the
* dot product of the symbols (bytes in GF(2^8)) across a set of buffers and a
* set of coefficients. Values for the coefficients are determined by the type
* of erasure code. Using a general dot product means that any sequence of
* coefficients may be used including erasure codes based on random
* coefficients.
* Multiple versions of dot product are supplied to calculate 1-6 output
* vectors in one pass.
* Base GF multiply and divide functions can be sped up by defining
* GF_LARGE_TABLES at the expense of memory size.
*
*/
#include "gf_vect_mul.h"
#ifdef __cplusplus
extern "C" {
#endif
/**
* @brief Initialize tables for fast Erasure Code encode and decode.
*
* Generates the expanded tables needed for fast encode or decode for erasure
* codes on blocks of data. 32bytes is generated for each input coefficient.
*
* @param k The number of vector sources or rows in the generator matrix
* for coding.
* @param rows The number of output vectors to concurrently encode/decode.
* @param a Pointer to sets of arrays of input coefficients used to encode
* or decode data.
* @param gftbls Pointer to start of space for concatenated output tables
* generated from input coefficients. Must be of size 32*k*rows.
* @returns none
*/
void ec_init_tables(int k, int rows, unsigned char* a, unsigned char* gftbls);
/**
* @brief Generate or decode erasure codes on blocks of data, runs appropriate version.
*
* Given a list of source data blocks, generate one or multiple blocks of
* encoded data as specified by a matrix of GF(2^8) coefficients. When given a
* suitable set of coefficients, this function will perform the fast generation
* or decoding of Reed-Solomon type erasure codes.
*
* This function determines what instruction sets are enabled and
* selects the appropriate version at runtime.
*
* @param len Length of each block of data (vector) of source or dest data.
* @param k The number of vector sources or rows in the generator matrix
* for coding.
* @param rows The number of output vectors to concurrently encode/decode.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*k*rows
* @param data Array of pointers to source input buffers.
* @param coding Array of pointers to coded output buffers.
* @returns none
*/
void ec_encode_data(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data.
*
* Arch specific version of ec_encode_data() with same parameters.
* @requires SSE4.1
*/
void ec_encode_data_sse(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data.
*
* Arch specific version of ec_encode_data() with same parameters.
* @requires AVX
*/
void ec_encode_data_avx(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data.
*
* Arch specific version of ec_encode_data() with same parameters.
* @requires AVX2
*/
void ec_encode_data_avx2(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data, runs baseline version.
*
* Baseline version of ec_encode_data() with same parameters.
*/
void ec_encode_data_base(int len, int srcs, int dests, unsigned char *v, unsigned char **src,
unsigned char **dest);
/**
* @brief Generate update for encode or decode of erasure codes from single source, runs appropriate version.
*
* Given one source data block, update one or multiple blocks of encoded data as
* specified by a matrix of GF(2^8) coefficients. When given a suitable set of
* coefficients, this function will perform the fast generation or decoding of
* Reed-Solomon type erasure codes from one input source at a time.
*
* This function determines what instruction sets are enabled and selects the
* appropriate version at runtime.
*
* @param len Length of each block of data (vector) of source or dest data.
* @param k The number of vector sources or rows in the generator matrix
* for coding.
* @param rows The number of output vectors to concurrently encode/decode.
* @param vec_i The vector index corresponding to the single input source.
* @param g_tbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*k*rows
* @param data Pointer to single input source used to update output parity.
* @param coding Array of pointers to coded output buffers.
* @returns none
*/
void ec_encode_data_update(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Arch specific version of ec_encode_data_update() with same parameters.
* @requires SSE4.1
*/
void ec_encode_data_update_sse(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Arch specific version of ec_encode_data_update() with same parameters.
* @requires AVX
*/
void ec_encode_data_update_avx(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Arch specific version of ec_encode_data_update() with same parameters.
* @requires AVX2
*/
void ec_encode_data_update_avx2(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Baseline version of ec_encode_data_update().
*/
void ec_encode_data_update_base(int len, int k, int rows, int vec_i, unsigned char *v,
unsigned char *data, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product with two outputs.
*
* Vector dot product optimized to calculate two ouputs at a time. Does two
* GF(2^8) dot products across each byte of the input array and two constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 2*32*vlen byte constant array based on the two sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 2*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_2vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with two outputs.
*
* Vector dot product optimized to calculate two ouputs at a time. Does two
* GF(2^8) dot products across each byte of the input array and two constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 2*32*vlen byte constant array based on the two sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 2*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_2vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with two outputs.
*
* Vector dot product optimized to calculate two ouputs at a time. Does two
* GF(2^8) dot products across each byte of the input array and two constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 2*32*vlen byte constant array based on the two sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 2*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_2vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with three outputs.
*
* Vector dot product optimized to calculate three ouputs at a time. Does three
* GF(2^8) dot products across each byte of the input array and three constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 3*32*vlen byte constant array based on the three sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 3*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_3vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with three outputs.
*
* Vector dot product optimized to calculate three ouputs at a time. Does three
* GF(2^8) dot products across each byte of the input array and three constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 3*32*vlen byte constant array based on the three sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 3*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_3vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with three outputs.
*
* Vector dot product optimized to calculate three ouputs at a time. Does three
* GF(2^8) dot products across each byte of the input array and three constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 3*32*vlen byte constant array based on the three sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 3*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_3vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with four outputs.
*
* Vector dot product optimized to calculate four ouputs at a time. Does four
* GF(2^8) dot products across each byte of the input array and four constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 4*32*vlen byte constant array based on the four sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 4*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_4vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with four outputs.
*
* Vector dot product optimized to calculate four ouputs at a time. Does four
* GF(2^8) dot products across each byte of the input array and four constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 4*32*vlen byte constant array based on the four sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 4*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_4vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with four outputs.
*
* Vector dot product optimized to calculate four ouputs at a time. Does four
* GF(2^8) dot products across each byte of the input array and four constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 4*32*vlen byte constant array based on the four sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 4*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_4vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with five outputs.
*
* Vector dot product optimized to calculate five ouputs at a time. Does five
* GF(2^8) dot products across each byte of the input array and five constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 5*32*vlen byte constant array based on the five sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 5*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_5vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with five outputs.
*
* Vector dot product optimized to calculate five ouputs at a time. Does five
* GF(2^8) dot products across each byte of the input array and five constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 5*32*vlen byte constant array based on the five sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 5*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_5vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with five outputs.
*
* Vector dot product optimized to calculate five ouputs at a time. Does five
* GF(2^8) dot products across each byte of the input array and five constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 5*32*vlen byte constant array based on the five sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 5*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_5vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with six outputs.
*
* Vector dot product optimized to calculate six ouputs at a time. Does six
* GF(2^8) dot products across each byte of the input array and six constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 6*32*vlen byte constant array based on the six sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 6*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_6vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with six outputs.
*
* Vector dot product optimized to calculate six ouputs at a time. Does six
* GF(2^8) dot products across each byte of the input array and six constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 6*32*vlen byte constant array based on the six sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 6*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_6vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with six outputs.
*
* Vector dot product optimized to calculate six ouputs at a time. Does six
* GF(2^8) dot products across each byte of the input array and six constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 6*32*vlen byte constant array based on the six sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 6*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_6vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product, runs baseline version.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients. Only elements 32*CONST*j + 1
* of this array are used, where j = (0, 1, 2...) and CONST is the
* number of elements in the array of input coefficients. The
* elements used correspond to the original input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_base(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product, runs appropriate version.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
*
* This function determines what instruction sets are enabled and
* selects the appropriate version at runtime.
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, runs appropriate version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constant and add to destination array. Can be used for erasure coding encode
* and decode update when only one source is available at a time. Function
* requires pre-calculation of a 32*vec byte constant array based on the input
* coefficients.
*
* This function determines what instruction sets are enabled and selects the
* appropriate version at runtime.
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_mad(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, arch specific version.
*
* Arch specific version of gf_vect_mad() with same parameters.
* @requires SSE4.1
*/
void gf_vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, arch specific version.
*
* Arch specific version of gf_vect_mad() with same parameters.
* @requires AVX
*/
void gf_vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, arch specific version.
*
* Arch specific version of gf_vect_mad() with same parameters.
* @requires AVX2
*/
void gf_vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, baseline version.
*
* Baseline version of gf_vect_mad() with same parameters.
*/
void gf_vect_mad_base(int len, int vec, int vec_i, unsigned char *v, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply with 2 accumulate. SSE version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constants and add to destination arrays. Can be used for erasure coding
* encode and decode update when only one source is available at a
* time. Function requires pre-calculation of a 32*vec byte constant array based
* on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Pointer to source input array.
* @param dest Array of pointers to destination input/outputs.
* @returns none
*/
void gf_2vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 2 accumulate. AVX version of gf_2vect_mad_sse().
* @requires AVX
*/
void gf_2vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 2 accumulate. AVX2 version of gf_2vect_mad_sse().
* @requires AVX2
*/
void gf_2vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 3 accumulate. SSE version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constants and add to destination arrays. Can be used for erasure coding
* encode and decode update when only one source is available at a
* time. Function requires pre-calculation of a 32*vec byte constant array based
* on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Pointer to source input array.
* @param dest Array of pointers to destination input/outputs.
* @returns none
*/
void gf_3vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 3 accumulate. AVX version of gf_3vect_mad_sse().
* @requires AVX
*/
void gf_3vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 3 accumulate. AVX2 version of gf_3vect_mad_sse().
* @requires AVX2
*/
void gf_3vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 4 accumulate. SSE version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constants and add to destination arrays. Can be used for erasure coding
* encode and decode update when only one source is available at a
* time. Function requires pre-calculation of a 32*vec byte constant array based
* on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Pointer to source input array.
* @param dest Array of pointers to destination input/outputs.
* @returns none
*/
void gf_4vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 4 accumulate. AVX version of gf_4vect_mad_sse().
* @requires AVX
*/
void gf_4vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 4 accumulate. AVX2 version of gf_4vect_mad_sse().
* @requires AVX2
*/
void gf_4vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 5 accumulate. SSE version.
* @requires SSE4.1
*/
void gf_5vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 5 accumulate. AVX version.
* @requires AVX
*/
void gf_5vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 5 accumulate. AVX2 version.
* @requires AVX2
*/
void gf_5vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 6 accumulate. SSE version.
* @requires SSE4.1
*/
void gf_6vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 6 accumulate. AVX version.
* @requires AVX
*/
void gf_6vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 6 accumulate. AVX2 version.
* @requires AVX2
*/
void gf_6vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**********************************************************************
* The remaining are lib support functions used in GF(2^8) operations.
*/
/**
* @brief Single element GF(2^8) multiply.
*
* @param a Multiplicand a
* @param b Multiplicand b
* @returns Product of a and b in GF(2^8)
*/
unsigned char gf_mul(unsigned char a, unsigned char b);
/**
* @brief Single element GF(2^8) inverse.
*
* @param a Input element
* @returns Field element b such that a x b = {1}
*/
unsigned char gf_inv(unsigned char a);
/**
* @brief Generate a matrix of coefficients to be used for encoding.
*
* Vandermonde matrix example of encoding coefficients where high portion of
* matrix is identity matrix I and lower portion is constructed as 2^{i*(j-k+1)}
* i:{0,k-1} j:{k,m-1}. Commonly used method for choosing coefficients in
* erasure encoding but does not guarantee invertable for every sub matrix. For
* large k it is possible to find cases where the decode matrix chosen from
* sources and parity not in erasure are not invertable. Users may want to
* adjust for k > 5.
*
* @param a [mxk] array to hold coefficients
* @param m number of rows in matrix corresponding to srcs + parity.
* @param k number of columns in matrix corresponding to srcs.
* @returns none
*/
void gf_gen_rs_matrix(unsigned char *a, int m, int k);
/**
* @brief Generate a Cauchy matrix of coefficients to be used for encoding.
*
* Cauchy matrix example of encoding coefficients where high portion of matrix
* is identity matrix I and lower portion is constructed as 1/(i + j) | i != j,
* i:{0,k-1} j:{k,m-1}. Any sub-matrix of a Cauchy matrix should be invertable.
*
* @param a [mxk] array to hold coefficients
* @param m number of rows in matrix corresponding to srcs + parity.
* @param k number of columns in matrix corresponding to srcs.
* @returns none
*/
void gf_gen_cauchy1_matrix(unsigned char *a, int m, int k);
/**
* @brief Invert a matrix in GF(2^8)
*
* @param in input matrix
* @param out output matrix such that [in] x [out] = [I] - identity matrix
* @param n size of matrix [nxn]
* @returns 0 successful, other fail on singular input matrix
*/
int gf_invert_matrix(unsigned char *in, unsigned char *out, const int n);
/*************************************************************/
#ifdef __cplusplus
}
#endif
#endif //_ERASURE_CODE_H_

267
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_highlevel_func.c generated vendored
View File

@ -1,267 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**********************************************************************/
#include <limits.h>
#include "ec_code.h"
#include "ec_types.h"
void ec_init_tables(int k, int rows, unsigned char *a, unsigned char *g_tbls)
{
int i, j;
for (i = 0; i < rows; i++) {
for (j = 0; j < k; j++) {
gf_vect_mul_init(*a++, g_tbls);
g_tbls += 32;
}
}
}
void ec_encode_data_sse(int len, int k, int rows, unsigned char *g_tbls, unsigned char **data,
unsigned char **coding)
{
if (len < 16) {
ec_encode_data_base(len, k, rows, g_tbls, data, coding);
return;
}
while (rows >= 4) {
gf_4vect_dot_prod_sse(len, k, g_tbls, data, coding);
g_tbls += 4 * k * 32;
coding += 4;
rows -= 4;
}
switch (rows) {
case 3:
gf_3vect_dot_prod_sse(len, k, g_tbls, data, coding);
break;
case 2:
gf_2vect_dot_prod_sse(len, k, g_tbls, data, coding);
break;
case 1:
gf_vect_dot_prod_sse(len, k, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_avx(int len, int k, int rows, unsigned char *g_tbls, unsigned char **data,
unsigned char **coding)
{
if (len < 16) {
ec_encode_data_base(len, k, rows, g_tbls, data, coding);
return;
}
while (rows >= 4) {
gf_4vect_dot_prod_avx(len, k, g_tbls, data, coding);
g_tbls += 4 * k * 32;
coding += 4;
rows -= 4;
}
switch (rows) {
case 3:
gf_3vect_dot_prod_avx(len, k, g_tbls, data, coding);
break;
case 2:
gf_2vect_dot_prod_avx(len, k, g_tbls, data, coding);
break;
case 1:
gf_vect_dot_prod_avx(len, k, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_avx2(int len, int k, int rows, unsigned char *g_tbls, unsigned char **data,
unsigned char **coding)
{
if (len < 32) {
ec_encode_data_base(len, k, rows, g_tbls, data, coding);
return;
}
while (rows >= 4) {
gf_4vect_dot_prod_avx2(len, k, g_tbls, data, coding);
g_tbls += 4 * k * 32;
coding += 4;
rows -= 4;
}
switch (rows) {
case 3:
gf_3vect_dot_prod_avx2(len, k, g_tbls, data, coding);
break;
case 2:
gf_2vect_dot_prod_avx2(len, k, g_tbls, data, coding);
break;
case 1:
gf_vect_dot_prod_avx2(len, k, g_tbls, data, *coding);
break;
case 0:
break;
}
}
#if __WORDSIZE == 64 || _WIN64 || __x86_64__
void ec_encode_data_update_sse(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding)
{
if (len < 16) {
ec_encode_data_update_base(len, k, rows, vec_i, g_tbls, data, coding);
return;
}
while (rows > 6) {
gf_6vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
g_tbls += 6 * k * 32;
coding += 6;
rows -= 6;
}
switch (rows) {
case 6:
gf_6vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 5:
gf_5vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 4:
gf_4vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 3:
gf_3vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 2:
gf_2vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 1:
gf_vect_mad_sse(len, k, vec_i, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_update_avx(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding)
{
if (len < 16) {
ec_encode_data_update_base(len, k, rows, vec_i, g_tbls, data, coding);
return;
}
while (rows > 6) {
gf_6vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
g_tbls += 6 * k * 32;
coding += 6;
rows -= 6;
}
switch (rows) {
case 6:
gf_6vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 5:
gf_5vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 4:
gf_4vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 3:
gf_3vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 2:
gf_2vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 1:
gf_vect_mad_avx(len, k, vec_i, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_update_avx2(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding)
{
if (len < 32) {
ec_encode_data_update_base(len, k, rows, vec_i, g_tbls, data, coding);
return;
}
while (rows > 6) {
gf_6vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
g_tbls += 6 * k * 32;
coding += 6;
rows -= 6;
}
switch (rows) {
case 6:
gf_6vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 5:
gf_5vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 4:
gf_4vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 3:
gf_3vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 2:
gf_2vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 1:
gf_vect_mad_avx2(len, k, vec_i, g_tbls, data, *coding);
break;
case 0:
break;
}
}
#endif //__WORDSIZE == 64 || _WIN64 || __x86_64__
struct slver {
UINT16 snum;
UINT8 ver;
UINT8 core;
};
// Version info
struct slver ec_init_tables_slver_00010068;
struct slver ec_init_tables_slver = { 0x0068, 0x01, 0x00 };
struct slver ec_encode_data_sse_slver_00020069;
struct slver ec_encode_data_sse_slver = { 0x0069, 0x02, 0x00 };

41
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_isal-l.h generated vendored
View File

@ -1,41 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**********************************************************************/
#ifndef _ISAL_H_
#define _ISAL_H_
#define ISAL_MAJOR_VERSION 2
#define ISAL_MINOR_VERSION 13
#define ISAL_PATCH_VERSION 0
#define ISAL_MAKE_VERSION(maj, min, patch) ((maj) * 0x10000 + (min) * 0x100 + (patch))
#define ISAL_VERSION ISAL_MAKE_VERSION(ISAL_MAJOR_VERSION, ISAL_MINOR_VERSION, ISAL_PATCH_VERSION)
#include "ec_code.h"
#include "gf_vect_mul.h"
#endif //_ISAL_H_

39
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_minio_common.h generated vendored
View File

@ -1,39 +0,0 @@
/*
* Mini Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef __COMMON_H__
#define __COMMON_H__
#include <stdint.h>
int32_t minio_init_encoder (int technique, int k, int m,
unsigned char **encode_matrix,
unsigned char **encode_tbls);
int32_t minio_init_decoder (int32_t *error_index,
int k, int n, int errs,
unsigned char *encoding_matrix,
unsigned char **decode_matrix,
unsigned char **decode_tbls,
uint32_t **decode_index);
int32_t minio_get_source_target (int errs, int k, int m,
int32_t *error_index,
uint32_t *decode_index,
unsigned char **buffs,
unsigned char ***source,
unsigned char ***target);
#endif /* __COMMON_H__ */

134
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_minio_decode.c generated vendored
View File

@ -1,134 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "ec_isal-l.h"
#include "ec_minio_common.h"
static
int32_t _minio_src_index_in_error (int r, int32_t *error_index)
{
int i;
for (i = 0; error_index[i] != -1; i++) {
if (error_index[i] == r) {
// true
return 1;
}
}
// false
return 0;
}
// Separate out source data and target buffers
int32_t minio_get_source_target (int errs, int k, int m,
int32_t *error_index,
uint32_t *decode_index,
unsigned char **buffs,
unsigned char ***source,
unsigned char ***target)
{
int i;
unsigned char *tmp_source[k];
unsigned char *tmp_target[m];
if (k < 0 || m < 0) {
return -1;
}
memset (tmp_source, 0, k);
memset (tmp_target, 0, m);
for (i = 0; i < k; i++) {
tmp_source[i] = (unsigned char *) buffs[decode_index[i]];
}
for (i = 0; i < m; i++) {
if (i < errs)
tmp_target[i] = (unsigned char *) buffs[error_index[i]];
}
*source = tmp_source;
*target = tmp_target;
return 0;
}
/*
Generate decode matrix during the decoding phase
*/
int minio_init_decoder (int32_t *error_index,
int k, int n, int errs,
unsigned char *encode_matrix,
unsigned char **decode_matrix,
unsigned char **decode_tbls,
uint32_t **decode_index)
{
int i, j, r, s, l, z;
unsigned char input_matrix[k * n];
unsigned char inverse_matrix[k * n];
unsigned char tmp_decode_matrix[k * n];
unsigned char tmp_decode_tbls[k * n * 32];
uint32_t tmp_decode_index[k];
for (i = 0, r = 0; i < k; i++, r++) {
while (_minio_src_index_in_error(r, error_index))
r++;
for (j = 0; j < k; j++) {
input_matrix[k * i + j] = encode_matrix[k * r + j];
}
tmp_decode_index[i] = r;
}
// Not all vandermonde matrix can be inverted
if (gf_invert_matrix(input_matrix, inverse_matrix, k) < 0) {
return -1;
}
for (l = 0; l < errs; l++) {
if (error_index[l] < k) {
// decoding matrix elements for data chunks
for (j = 0; j < k; j++) {
tmp_decode_matrix[k * l + j] =
inverse_matrix[k *
error_index[l] + j];
}
} else {
int s = 0;
// decoding matrix element for coding chunks
for (i = 0; i < k; i++) {
s = 0;
for (j = 0; j < k; j++) {
s ^= gf_mul(inverse_matrix[j * k + i],
encode_matrix[k *
error_index[l] + j]);
}
tmp_decode_matrix[k * l + i] = s;
}
}
}
ec_init_tables (k, errs, tmp_decode_matrix, tmp_decode_tbls);
*decode_matrix = tmp_decode_matrix;
*decode_tbls = tmp_decode_tbls;
*decode_index = tmp_decode_index;
return 0;
}

59
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_minio_encode.c generated vendored
View File

@ -1,59 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdlib.h>
#include <stdio.h>
#include "ec_isal-l.h"
#include "ec_minio_common.h"
/*
Generate encode matrix during the encoding phase
*/
int32_t minio_init_encoder (int technique, int k, int m,
unsigned char **encode_matrix,
unsigned char **encode_tbls)
{
size_t encode_matrix_size;
size_t encode_tbls_size;
unsigned char *tmp_matrix;
unsigned char *tmp_tbls;
tmp_matrix = (unsigned char *) malloc (k * (k + m));
tmp_tbls = (unsigned char *) malloc (k * (k + m) * 32);
if (technique == 0) {
/*
Commonly used method for choosing coefficients in erasure
encoding but does not guarantee invertable for every sub
matrix. For large k it is possible to find cases where the
decode matrix chosen from sources and parity not in erasure
are not invertable. Users may want to adjust for k > 5.
-- Intel
*/
gf_gen_rs_matrix (tmp_matrix, k + m, k);
} else if (technique == 1) {
gf_gen_cauchy1_matrix (tmp_matrix, k + m, k);
}
ec_init_tables(k, m, &tmp_matrix[k * k], tmp_tbls);
*encode_matrix = tmp_matrix;
*encode_tbls = tmp_tbls;
return 0;
}

462
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_multibinary.asm generated vendored
View File

@ -1,462 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
%ifidn __OUTPUT_FORMAT__, elf64
%define WRT_OPT wrt ..plt
%else
%define WRT_OPT
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define EC_ENCODE_DATA_SSE _ec_encode_data_sse
%define EC_ENCODE_DATA_AVX _ec_encode_data_avx
%define EC_ENCODE_DATA_AVX2 _ec_encode_data_avx2
%define EC_ENCODE_DATA_BASE _ec_encode_data_base
%define EC_ENCODE_DATA _ec_encode_data
%define EC_ENCODE_DATA_UPDATE_BASE _ec_encode_data_update_base
%define EC_ENCODE_DATA_UPDATE_SSE _ec_encode_data_update_sse
%define EC_ENCODE_DATA_UPDATE_AVX _ec_encode_data_update_avx
%define EC_ENCODE_DATA_UPDATE_AVX2 _ec_encode_data_update_avx2
%define GF_VECT_MAD_BASE _gf_vect_mad_base
%define GF_VECT_MAD_SSE _gf_vect_mad_sse
%define GF_VECT_MAD_AVX _gf_vect_mad_avx
%define GF_VECT_MAD_AVX2 _gf_vect_mad_avx2
%define GF_VECT_MUL_SSE _gf_vect_mul_sse
%define GF_VECT_MUL_AVX _gf_vect_mul_avx
%define GF_VECT_MUL_BASE _gf_vect_mul_base
%define GF_VECT_MUL _gf_vect_mul
%define GF_VECT_DOT_PROD_SSE _gf_vect_dot_prod_sse
%define GF_VECT_DOT_PROD_AVX _gf_vect_dot_prod_avx
%define GF_VECT_DOT_PROD_AVX2 _gf_vect_dot_prod_avx2
%define GF_VECT_DOT_PROD_BASE _gf_vect_dot_prod_base
%define GF_VECT_DOT_PROD _gf_vect_dot_prod
%else
%define EC_ENCODE_DATA_SSE ec_encode_data_sse
%define EC_ENCODE_DATA_AVX ec_encode_data_avx
%define EC_ENCODE_DATA_AVX2 ec_encode_data_avx2
%define EC_ENCODE_DATA_BASE ec_encode_data_base
%define EC_ENCODE_DATA ec_encode_data
%define EC_ENCODE_DATA_UPDATE_BASE ec_encode_data_update_base
%define EC_ENCODE_DATA_UPDATE_SSE ec_encode_data_update_sse
%define EC_ENCODE_DATA_UPDATE_AVX ec_encode_data_update_avx
%define EC_ENCODE_DATA_UPDATE_AVX2 ec_encode_data_update_avx2
%define GF_VECT_MAD_BASE gf_vect_mad_base
%define GF_VECT_MAD_SSE gf_vect_mad_sse
%define GF_VECT_MAD_AVX gf_vect_mad_avx
%define GF_VECT_MAD_AVX2 gf_vect_mad_avx2
%define GF_VECT_MUL_SSE gf_vect_mul_sse
%define GF_VECT_MUL_AVX gf_vect_mul_avx
%define GF_VECT_MUL_BASE gf_vect_mul_base
%define GF_VECT_MUL gf_vect_mul
%define GF_VECT_DOT_PROD_SSE gf_vect_dot_prod_sse
%define GF_VECT_DOT_PROD_AVX gf_vect_dot_prod_avx
%define GF_VECT_DOT_PROD_AVX2 gf_vect_dot_prod_avx2
%define GF_VECT_DOT_PROD_BASE gf_vect_dot_prod_base
%define GF_VECT_DOT_PROD gf_vect_dot_prod
%endif
%include "ec_reg_sizes.asm"
%ifidn __OUTPUT_FORMAT__, elf32
[bits 32]
%define def_wrd dd
%define wrd_sz dword
%define arg1 esi
%define arg2 eax
%define arg3 ebx
%define arg4 ecx
%define arg5 edx
%else
default rel
[bits 64]
%define def_wrd dq
%define wrd_sz qword
%define arg1 rsi
%define arg2 rax
%define arg3 rbx
%define arg4 rcx
%define arg5 rdx
extern EC_ENCODE_DATA_UPDATE_SSE
extern EC_ENCODE_DATA_UPDATE_AVX
extern EC_ENCODE_DATA_UPDATE_AVX2
extern GF_VECT_MUL_SSE
extern GF_VECT_MUL_AVX
extern GF_VECT_MAD_SSE
extern GF_VECT_MAD_AVX
extern GF_VECT_MAD_AVX2
%endif
extern GF_VECT_MUL_BASE
extern EC_ENCODE_DATA_BASE
extern EC_ENCODE_DATA_UPDATE_BASE
extern GF_VECT_DOT_PROD_BASE
extern GF_VECT_MAD_BASE
extern GF_VECT_DOT_PROD_SSE
extern GF_VECT_DOT_PROD_AVX
extern GF_VECT_DOT_PROD_AVX2
extern EC_ENCODE_DATA_SSE
extern EC_ENCODE_DATA_AVX
extern EC_ENCODE_DATA_AVX2
section .data
;;; *_mbinit are initial values for *_dispatched; is updated on first call.
;;; Therefore, *_dispatch_init is only executed on first call.
ec_encode_data_dispatched:
def_wrd ec_encode_data_mbinit
gf_vect_mul_dispatched:
def_wrd gf_vect_mul_mbinit
gf_vect_dot_prod_dispatched:
def_wrd gf_vect_dot_prod_mbinit
ec_encode_data_update_dispatched:
def_wrd ec_encode_data_update_mbinit
gf_vect_mad_dispatched:
def_wrd gf_vect_mad_mbinit
section .text
;;;;
; ec_encode_data multibinary function
;;;;
global EC_ENCODE_DATA:function
ec_encode_data_mbinit:
call ec_encode_data_dispatch_init
EC_ENCODE_DATA:
jmp wrd_sz [ec_encode_data_dispatched]
ec_encode_data_dispatch_init:
push arg1
push arg2
push arg3
push arg4
push arg5
lea arg1, [EC_ENCODE_DATA_BASE WRT_OPT] ; Default
mov eax, 1
cpuid
lea arg3, [EC_ENCODE_DATA_SSE WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, arg3
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea arg3, [EC_ENCODE_DATA_AVX WRT_OPT]
jne _done_ec_encode_data_init
mov arg1, arg3
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea arg3, [EC_ENCODE_DATA_AVX2 WRT_OPT]
cmovne arg1, arg3
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_ec_encode_data_init
lea arg1, [EC_ENCODE_DATA_SSE WRT_OPT]
_done_ec_encode_data_init:
pop arg5
pop arg4
pop arg3
pop arg2
mov [ec_encode_data_dispatched], arg1
pop arg1
ret
;;;;
; gf_vect_mul multibinary function
;;;;
global GF_VECT_MUL:function
gf_vect_mul_mbinit:
call gf_vect_mul_dispatch_init
GF_VECT_MUL:
jmp wrd_sz [gf_vect_mul_dispatched]
gf_vect_mul_dispatch_init:
push arg1
%ifidn __OUTPUT_FORMAT__, elf32 ;; 32-bit check
lea arg1, [GF_VECT_MUL_BASE]
%else
push rax
push rbx
push rcx
push rdx
lea arg1, [GF_VECT_MUL_BASE WRT_OPT] ; Default
mov eax, 1
cpuid
test ecx, FLAG_CPUID1_ECX_SSE4_2
lea rbx, [GF_VECT_MUL_SSE WRT_OPT]
je _done_gf_vect_mul_dispatch_init
mov arg1, rbx
;; Try for AVX
and ecx, (FLAG_CPUID1_ECX_OSXSAVE | FLAG_CPUID1_ECX_AVX)
cmp ecx, (FLAG_CPUID1_ECX_OSXSAVE | FLAG_CPUID1_ECX_AVX)
jne _done_gf_vect_mul_dispatch_init
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
jne _done_gf_vect_mul_dispatch_init
lea arg1, [GF_VECT_MUL_AVX WRT_OPT]
_done_gf_vect_mul_dispatch_init:
pop rdx
pop rcx
pop rbx
pop rax
%endif ;; END 32-bit check
mov [gf_vect_mul_dispatched], arg1
pop arg1
ret
;;;;
; ec_encode_data_update multibinary function
;;;;
global EC_ENCODE_DATA_UPDATE:function
ec_encode_data_update_mbinit:
call ec_encode_data_update_dispatch_init
EC_ENCODE_DATA_UPDATE:
jmp wrd_sz [ec_encode_data_update_dispatched]
ec_encode_data_update_dispatch_init:
push arg1
%ifidn __OUTPUT_FORMAT__, elf32 ;; 32-bit check
lea arg1, [ec_encode_data_update_base]
%else
push rax
push rbx
push rcx
push rdx
lea arg1, [EC_ENCODE_DATA_UPDATE_BASE WRT_OPT] ; Default
mov eax, 1
cpuid
lea rbx, [EC_ENCODE_DATA_UPDATE_SSE WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, rbx
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea rbx, [EC_ENCODE_DATA_UPDATE_AVX WRT_OPT]
jne _done_ec_encode_data_update_init
mov rsi, rbx
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea rbx, [EC_ENCODE_DATA_UPDATE_AVX2 WRT_OPT]
cmovne rsi, rbx
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_ec_encode_data_update_init
lea rsi, [EC_ENCODE_DATA_UPDATE_SSE WRT_OPT]
_done_ec_encode_data_update_init:
pop rdx
pop rcx
pop rbx
pop rax
%endif ;; END 32-bit check
mov [ec_encode_data_update_dispatched], arg1
pop arg1
ret
;;;;
; gf_vect_dot_prod multibinary function
;;;;
global GF_VECT_DOT_PROD:function
gf_vect_dot_prod_mbinit:
call gf_vect_dot_prod_dispatch_init
GF_VECT_DOT_PROD:
jmp wrd_sz [gf_vect_dot_prod_dispatched]
gf_vect_dot_prod_dispatch_init:
push arg1
push arg2
push arg3
push arg4
push arg5
lea arg1, [GF_VECT_DOT_PROD_BASE WRT_OPT] ; Default
mov eax, 1
cpuid
lea arg3, [GF_VECT_DOT_PROD_SSE WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, arg3
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea arg3, [GF_VECT_DOT_PROD_AVX WRT_OPT]
jne _done_gf_vect_dot_prod_init
mov arg1, arg3
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea arg3, [GF_VECT_DOT_PROD_AVX2 WRT_OPT]
cmovne arg1, arg3
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_gf_vect_dot_prod_init
lea arg1, [GF_VECT_DOT_PROD_SSE WRT_OPT]
_done_gf_vect_dot_prod_init:
pop arg5
pop arg4
pop arg3
pop arg2
mov [gf_vect_dot_prod_dispatched], arg1
pop arg1
ret
;;;;
; gf_vect_mad multibinary function
;;;;
global GF_VECT_MAD:function
gf_vect_mad_mbinit:
call gf_vect_mad_dispatch_init
GF_VECT_MAD:
jmp wrd_sz [gf_vect_mad_dispatched]
gf_vect_mad_dispatch_init:
push arg1
%ifidn __OUTPUT_FORMAT__, elf32 ;; 32-bit check
lea arg1, [gf_vect_mad_base]
%else
push rax
push rbx
push rcx
push rdx
lea arg1, [GF_VECT_MAD_BASE WRT_OPT] ; Default
mov eax, 1
cpuid
lea rbx, [GF_VECT_MAD_SSE WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, rbx
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea rbx, [GF_VECT_MAD_AVX WRT_OPT]
jne _done_gf_vect_mad_init
mov rsi, rbx
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea rbx, [GF_VECT_MAD_AVX2 WRT_OPT]
cmovne rsi, rbx
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_gf_vect_mad_init
lea rsi, [GF_VECT_MAD_SSE WRT_OPT]
_done_gf_vect_mad_init:
pop rdx
pop rcx
pop rbx
pop rax
%endif ;; END 32-bit check
mov [gf_vect_mad_dispatched], arg1
pop arg1
ret
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion EC_ENCODE_DATA, 00, 03, 0133
slversion GF_VECT_MUL, 00, 02, 0134
slversion EC_ENCODE_DATA_UPDATE, 00, 02, 0212
slversion GF_VECT_DOT_PROD, 00, 02, 0138
slversion GF_VECT_MAD, 00, 01, 0213

96
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_reg_sizes.asm generated vendored
View File

@ -1,96 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
%define EFLAGS_HAS_CPUID (1<<21)
%define FLAG_CPUID1_ECX_CLMUL (1<<1)
%define FLAG_CPUID1_EDX_SSE2 (1<<26)
%define FLAG_CPUID1_ECX_SSE3 (1)
%define FLAG_CPUID1_ECX_SSE4_1 (1<<19)
%define FLAG_CPUID1_ECX_SSE4_2 (1<<20)
%define FLAG_CPUID1_ECX_POPCNT (1<<23)
%define FLAG_CPUID1_ECX_AESNI (1<<25)
%define FLAG_CPUID1_ECX_OSXSAVE (1<<27)
%define FLAG_CPUID1_ECX_AVX (1<<28)
%define FLAG_CPUID1_EBX_AVX2 (1<<5)
%define FLAG_XGETBV_EAX_XMM_YMM 0x6
%define FLAG_CPUID1_EAX_AVOTON 0x000406d0
; define d and w variants for registers
%define raxd eax
%define raxw ax
%define raxb al
%define rbxd ebx
%define rbxw bx
%define rbxb bl
%define rcxd ecx
%define rcxw cx
%define rcxb cl
%define rdxd edx
%define rdxw dx
%define rdxb dl
%define rsid esi
%define rsiw si
%define rsib sil
%define rdid edi
%define rdiw di
%define rdib dil
%define rbpd ebp
%define rbpw bp
%define rbpb bpl
%define ymm0x xmm0
%define ymm1x xmm1
%define ymm2x xmm2
%define ymm3x xmm3
%define ymm4x xmm4
%define ymm5x xmm5
%define ymm6x xmm6
%define ymm7x xmm7
%define ymm8x xmm8
%define ymm9x xmm9
%define ymm10x xmm10
%define ymm11x xmm11
%define ymm12x xmm12
%define ymm13x xmm13
%define ymm14x xmm14
%define ymm15x xmm15
%define DWORD(reg) reg %+ d
%define WORD(reg) reg %+ w
%define BYTE(reg) reg %+ b
%define XWORD(reg) reg %+ x

80
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/ec_types.h generated vendored
View File

@ -1,80 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**********************************************************************/
/**
* @file types.h
* @brief Defines standard width types.
*
*/
#ifndef __TYPES_H
#define __TYPES_H
#ifdef __cplusplus
extern "C" {
#endif
#if !defined(__unix__) && !defined(__APPLE__)
#ifdef __MINGW32__
# include <_mingw.h>
#endif
typedef unsigned __int64 UINT64;
typedef __int64 INT64;
typedef unsigned __int32 UINT32;
typedef unsigned __int16 UINT16;
typedef unsigned char UINT8;
#else
typedef unsigned long int UINT64;
typedef long int INT64;
typedef unsigned int UINT32;
typedef unsigned short int UINT16;
typedef unsigned char UINT8;
#endif
#if defined(__unix__) || defined(__APPLE__)
# define DECLARE_ALIGNED(decl, alignval) decl __attribute__((aligned(alignval)))
# define __forceinline static inline
#else
# define DECLARE_ALIGNED(decl, alignval) __declspec(align(alignval)) decl
# define posix_memalign(p, algn, len) (NULL == (*((char**)(p)) = (void*) _aligned_malloc(len, algn)))
#endif
#ifdef DEBUG
# define DEBUG_PRINT(x) printf x
#else
# define DEBUG_PRINT(x) do {} while (0)
#endif
#ifdef __cplusplus
}
#endif
#endif //__TYPES_H

122
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/erasure_decode.go generated vendored
View File

@ -1,122 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
// #cgo CFLAGS: -O0
// #include <stdlib.h>
// #include "ec_isal-l.h"
// #include "ec_minio_common.h"
import "C"
import (
"errors"
"fmt"
"unsafe"
)
// Decode decodes erasure coded blocks of data into its original
// form. Erasure coded data contains K data blocks and M parity
// blocks. Decode can withstand data loss up to any M number of blocks.
//
// "encodedDataBlocks" is an array of K data blocks and M parity
// blocks. Data blocks are position and order dependent. Missing blocks
// are set to "nil". There must be at least "K" number of data|parity
// blocks.
//
// "dataLen" is the length of original source data
func (e *Erasure) Decode(encodedDataBlocks [][]byte, dataLen int) (decodedData []byte, err error) {
var source, target **C.uchar
k := int(e.params.K)
m := int(e.params.M)
n := k + m
// We need the data and parity blocks preserved in the same order. Missing blocks are set to nil.
if len(encodedDataBlocks) != n {
msg := fmt.Sprintf("Encoded data blocks slice must of length [%d]", n)
return nil, errors.New(msg)
}
// Length of a single encoded block
encodedBlockLen := GetEncodedBlockLen(dataLen, uint8(k))
// Keep track of errors per block.
missingEncodedBlocks := make([]int, n+1)
var missingEncodedBlocksCount int
// Check for the missing encoded blocks
for i := range encodedDataBlocks {
if encodedDataBlocks[i] == nil || len(encodedDataBlocks[i]) == 0 {
missingEncodedBlocks[missingEncodedBlocksCount] = i
missingEncodedBlocksCount++
}
}
missingEncodedBlocks[missingEncodedBlocksCount] = -1
missingEncodedBlocksCount++
// Cannot reconstruct original data. Need at least M number of data or parity blocks.
if missingEncodedBlocksCount-1 > m {
return nil, fmt.Errorf("Cannot reconstruct original data. Need at least [%d] data or parity blocks", m)
}
// Convert from Go int slice to C int array
missingEncodedBlocksC := intSlice2CIntArray(missingEncodedBlocks[:missingEncodedBlocksCount])
// Allocate buffer for the missing blocks
for i := range encodedDataBlocks {
if encodedDataBlocks[i] == nil || len(encodedDataBlocks[i]) == 0 {
encodedDataBlocks[i] = make([]byte, encodedBlockLen)
}
}
// If not already initialized, recompute and cache
if e.decodeMatrix == nil || e.decodeTbls == nil || e.decodeIndex == nil {
var decodeMatrix, decodeTbls *C.uchar
var decodeIndex *C.uint32_t
C.minio_init_decoder(missingEncodedBlocksC, C.int(k), C.int(n), C.int(missingEncodedBlocksCount-1),
e.encodeMatrix, &decodeMatrix, &decodeTbls, &decodeIndex)
// cache this for future needs
e.decodeMatrix = decodeMatrix
e.decodeTbls = decodeTbls
e.decodeIndex = decodeIndex
}
// Make a slice of pointers to encoded blocks. Necessary to bridge to the C world.
pointers := make([]*byte, n)
for i := range encodedDataBlocks {
pointers[i] = &encodedDataBlocks[i][0]
}
// Get pointers to source "data" and target "parity" blocks from the output byte array.
ret := C.minio_get_source_target(C.int(missingEncodedBlocksCount-1), C.int(k), C.int(m), missingEncodedBlocksC,
e.decodeIndex, (**C.uchar)(unsafe.Pointer(&pointers[0])), &source, &target)
if int(ret) == -1 {
return nil, errors.New("Unable to decode data")
}
// Decode data
C.ec_encode_data(C.int(encodedBlockLen), C.int(k), C.int(missingEncodedBlocksCount-1), e.decodeTbls,
source, target)
// Allocate buffer to output buffer
decodedData = make([]byte, 0, encodedBlockLen*int(k))
for i := 0; i < int(k); i++ {
decodedData = append(decodedData, encodedDataBlocks[i]...)
}
return decodedData[:dataLen], nil
}

197
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/erasure_encode.go generated vendored
View File

@ -1,197 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
// #cgo CFLAGS: -O0
// #include <stdlib.h>
// #include "ec_isal-l.h"
// #include "ec_minio_common.h"
import "C"
import (
"errors"
"unsafe"
)
// Technique - type of matrix type used in encoding
type Technique uint8
// Different types of supported matrix types
const (
Vandermonde Technique = iota
Cauchy
None
)
// Default Data and Parity blocks
const (
K = 10
M = 3
)
// Block alignment
const (
SIMDAlign = 32
)
// ErasureParams is a configuration set for building an encoder. It is created using ValidateParams().
type ErasureParams struct {
K uint8
M uint8
Technique Technique // cauchy or vandermonde matrix (RS)
}
// Erasure is an object used to encode and decode data.
type Erasure struct {
params *ErasureParams
encodeMatrix, encodeTbls *C.uchar
decodeMatrix, decodeTbls *C.uchar
decodeIndex *C.uint32_t
}
// ValidateParams creates an ErasureParams object.
//
// k and m represent the matrix size, which corresponds to the protection level
// technique is the matrix type. Valid inputs are Cauchy (recommended) or Vandermonde.
//
func ValidateParams(k, m uint8, technique Technique) (*ErasureParams, error) {
if k < 1 {
return nil, errors.New("k cannot be zero")
}
if m < 1 {
return nil, errors.New("m cannot be zero")
}
if k+m > 255 {
return nil, errors.New("(k + m) cannot be bigger than Galois field GF(2^8) - 1")
}
switch technique {
case Vandermonde:
break
case Cauchy:
break
default:
return nil, errors.New("Technique can be either vandermonde or cauchy")
}
return &ErasureParams{
K: k,
M: m,
Technique: technique,
}, nil
}
// NewErasure creates an encoder object with a given set of parameters.
func NewErasure(ep *ErasureParams) *Erasure {
var k = C.int(ep.K)
var m = C.int(ep.M)
var encodeMatrix *C.uchar
var encodeTbls *C.uchar
C.minio_init_encoder(C.int(ep.Technique), k, m, &encodeMatrix,
&encodeTbls)
return &Erasure{
params: ep,
encodeMatrix: encodeMatrix,
encodeTbls: encodeTbls,
decodeMatrix: nil,
decodeTbls: nil,
decodeIndex: nil,
}
}
// GetEncodedBlocksLen - total length of all encoded blocks
func GetEncodedBlocksLen(inputLen int, k, m uint8) (outputLen int) {
outputLen = GetEncodedBlockLen(inputLen, k) * int(k+m)
return outputLen
}
// GetEncodedBlockLen - length per block of encoded blocks
func GetEncodedBlockLen(inputLen int, k uint8) (encodedOutputLen int) {
alignment := int(k) * SIMDAlign
remainder := inputLen % alignment
paddedInputLen := inputLen
if remainder != 0 {
paddedInputLen = inputLen + (alignment - remainder)
}
encodedOutputLen = paddedInputLen / int(k)
return encodedOutputLen
}
// Encode erasure codes a block of data in "k" data blocks and "m" parity blocks.
// Output is [k+m][]blocks of data and parity slices.
func (e *Erasure) Encode(inputData []byte) (encodedBlocks [][]byte, err error) {
k := int(e.params.K) // "k" data blocks
m := int(e.params.M) // "m" parity blocks
n := k + m // "n" total encoded blocks
// Length of a single encoded chunk.
// Total number of encoded chunks = "k" data + "m" parity blocks
encodedBlockLen := GetEncodedBlockLen(len(inputData), uint8(k))
// Length of total number of "k" data chunks
encodedDataBlocksLen := encodedBlockLen * k
// Length of extra padding required for the data blocks.
encodedDataBlocksPadLen := encodedDataBlocksLen - len(inputData)
// Extend inputData buffer to accommodate coded data blocks if necesssary
if encodedDataBlocksPadLen > 0 {
padding := make([]byte, encodedDataBlocksPadLen)
// Expand with new padded blocks to the byte array
inputData = append(inputData, padding...)
}
// Extend inputData buffer to accommodate coded parity blocks
{ // Local Scope
encodedParityBlocksLen := encodedBlockLen * m
parityBlocks := make([]byte, encodedParityBlocksLen)
inputData = append(inputData, parityBlocks...)
}
// Allocate memory to the "encoded blocks" return buffer
encodedBlocks = make([][]byte, n) // Return buffer
// Nessary to bridge Go to the C world. C requires 2D arry of pointers to
// byte array. "encodedBlocks" is a 2D slice.
pointersToEncodedBlock := make([]*byte, n) // Pointers to encoded blocks.
// Copy data block slices to encoded block buffer
for i := 0; i < k; i++ {
encodedBlocks[i] = inputData[i*encodedBlockLen : (i+1)*encodedBlockLen]
pointersToEncodedBlock[i] = &encodedBlocks[i][0]
}
// Copy erasure block slices to encoded block buffer
for i := k; i < n; i++ {
encodedBlocks[i] = make([]byte, encodedBlockLen)
pointersToEncodedBlock[i] = &encodedBlocks[i][0]
}
// Erasure code the data into K data blocks and M parity
// blocks. Only the parity blocks are filled. Data blocks remain
// intact.
C.ec_encode_data(C.int(encodedBlockLen), C.int(k), C.int(m), e.encodeTbls,
(**C.uchar)(unsafe.Pointer(&pointersToEncodedBlock[:k][0])), // Pointers to data blocks
(**C.uchar)(unsafe.Pointer(&pointersToEncodedBlock[k:][0]))) // Pointers to parity blocks
return encodedBlocks, nil
}

43
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/erasure_yasm_darwin.go generated vendored
View File

@ -1,43 +0,0 @@
// !build amd64
package erasure
//go:generate yasm -f macho64 ec_multibinary.asm -o ec_multibinary.syso
//go:generate yasm -f macho64 gf_2vect_mad_avx2.asm -o gf_2vect_mad_avx2.syso
//go:generate yasm -f macho64 gf_2vect_mad_avx.asm -o gf_2vect_mad_avx.syso
//go:generate yasm -f macho64 gf_2vect_mad_sse.asm -o gf_2vect_mad_sse.syso
//go:generate yasm -f macho64 gf_3vect_mad_avx2.asm -o gf_3vect_mad_avx2.syso
//go:generate yasm -f macho64 gf_3vect_mad_avx.asm -o gf_3vect_mad_avx.syso
//go:generate yasm -f macho64 gf_3vect_mad_sse.asm -o gf_3vect_mad_sse.syso
//go:generate yasm -f macho64 gf_4vect_mad_avx2.asm -o gf_4vect_mad_avx2.syso
//go:generate yasm -f macho64 gf_4vect_mad_avx.asm -o gf_4vect_mad_avx.syso
//go:generate yasm -f macho64 gf_4vect_mad_sse.asm -o gf_4vect_mad_sse.syso
//go:generate yasm -f macho64 gf_5vect_mad_avx2.asm -o gf_5vect_mad_avx2.syso
//go:generate yasm -f macho64 gf_5vect_mad_avx.asm -o gf_5vect_mad_avx.syso
//go:generate yasm -f macho64 gf_5vect_mad_sse.asm -o gf_5vect_mad_sse.syso
//go:generate yasm -f macho64 gf_6vect_mad_avx2.asm -o gf_6vect_mad_avx2.syso
//go:generate yasm -f macho64 gf_6vect_mad_avx.asm -o gf_6vect_mad_avx.syso
//go:generate yasm -f macho64 gf_6vect_mad_sse.asm -o gf_6vect_mad_sse.syso
//go:generate yasm -f macho64 gf_vect_mad_avx2.asm -o gf_vect_mad_avx2.syso
//go:generate yasm -f macho64 gf_vect_mad_avx.asm -o gf_vect_mad_avx.syso
//go:generate yasm -f macho64 gf_vect_mad_sse.asm -o gf_vect_mad_sse.syso
//go:generate yasm -f macho64 gf_2vect_dot_prod_avx2.asm -o gf_2vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 gf_2vect_dot_prod_avx.asm -o gf_2vect_dot_prod_avx.syso
//go:generate yasm -f macho64 gf_2vect_dot_prod_sse.asm -o gf_2vect_dot_prod_sse.syso
//go:generate yasm -f macho64 gf_3vect_dot_prod_avx2.asm -o gf_3vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 gf_3vect_dot_prod_avx.asm -o gf_3vect_dot_prod_avx.syso
//go:generate yasm -f macho64 gf_3vect_dot_prod_sse.asm -o gf_3vect_dot_prod_sse.syso
//go:generate yasm -f macho64 gf_4vect_dot_prod_avx2.asm -o gf_4vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 gf_4vect_dot_prod_avx.asm -o gf_4vect_dot_prod_avx.syso
//go:generate yasm -f macho64 gf_4vect_dot_prod_sse.asm -o gf_4vect_dot_prod_sse.syso
//go:generate yasm -f macho64 gf_5vect_dot_prod_avx2.asm -o gf_5vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 gf_5vect_dot_prod_avx.asm -o gf_5vect_dot_prod_avx.syso
//go:generate yasm -f macho64 gf_5vect_dot_prod_sse.asm -o gf_5vect_dot_prod_sse.syso
//go:generate yasm -f macho64 gf_6vect_dot_prod_avx2.asm -o gf_6vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 gf_6vect_dot_prod_avx.asm -o gf_6vect_dot_prod_avx.syso
//go:generate yasm -f macho64 gf_6vect_dot_prod_sse.asm -o gf_6vect_dot_prod_sse.syso
//go:generate yasm -f macho64 gf_vect_dot_prod_avx2.asm -o gf_vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 gf_vect_dot_prod_avx.asm -o gf_vect_dot_prod_avx.syso
//go:generate yasm -f macho64 gf_vect_dot_prod_sse.asm -o gf_vect_dot_prod_sse.syso
//go:generate yasm -f macho64 gf_vect_mul_avx.asm -o gf_vect_mul_avx.syso
//go:generate yasm -f macho64 gf_vect_mul_sse.asm -o gf_vect_mul_sse.syso

43
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/erasure_yasm_linux.go generated vendored
View File

@ -1,43 +0,0 @@
// !build amd64
package erasure
//go:generate yasm -f elf64 ec_multibinary.asm -o ec_multibinary.syso
//go:generate yasm -f elf64 gf_2vect_mad_avx2.asm -o gf_2vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_2vect_mad_avx.asm -o gf_2vect_mad_avx.syso
//go:generate yasm -f elf64 gf_2vect_mad_sse.asm -o gf_2vect_mad_sse.syso
//go:generate yasm -f elf64 gf_3vect_mad_avx2.asm -o gf_3vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_3vect_mad_avx.asm -o gf_3vect_mad_avx.syso
//go:generate yasm -f elf64 gf_3vect_mad_sse.asm -o gf_3vect_mad_sse.syso
//go:generate yasm -f elf64 gf_4vect_mad_avx2.asm -o gf_4vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_4vect_mad_avx.asm -o gf_4vect_mad_avx.syso
//go:generate yasm -f elf64 gf_4vect_mad_sse.asm -o gf_4vect_mad_sse.syso
//go:generate yasm -f elf64 gf_5vect_mad_avx2.asm -o gf_5vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_5vect_mad_avx.asm -o gf_5vect_mad_avx.syso
//go:generate yasm -f elf64 gf_5vect_mad_sse.asm -o gf_5vect_mad_sse.syso
//go:generate yasm -f elf64 gf_6vect_mad_avx2.asm -o gf_6vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_6vect_mad_avx.asm -o gf_6vect_mad_avx.syso
//go:generate yasm -f elf64 gf_6vect_mad_sse.asm -o gf_6vect_mad_sse.syso
//go:generate yasm -f elf64 gf_vect_mad_avx2.asm -o gf_vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_vect_mad_avx.asm -o gf_vect_mad_avx.syso
//go:generate yasm -f elf64 gf_vect_mad_sse.asm -o gf_vect_mad_sse.syso
//go:generate yasm -f elf64 gf_2vect_dot_prod_avx2.asm -o gf_2vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_2vect_dot_prod_avx.asm -o gf_2vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_2vect_dot_prod_sse.asm -o gf_2vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_3vect_dot_prod_avx2.asm -o gf_3vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_3vect_dot_prod_avx.asm -o gf_3vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_3vect_dot_prod_sse.asm -o gf_3vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_4vect_dot_prod_avx2.asm -o gf_4vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_4vect_dot_prod_avx.asm -o gf_4vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_4vect_dot_prod_sse.asm -o gf_4vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_5vect_dot_prod_avx2.asm -o gf_5vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_5vect_dot_prod_avx.asm -o gf_5vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_5vect_dot_prod_sse.asm -o gf_5vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_6vect_dot_prod_avx2.asm -o gf_6vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_6vect_dot_prod_avx.asm -o gf_6vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_6vect_dot_prod_sse.asm -o gf_6vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_vect_dot_prod_avx2.asm -o gf_vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_vect_dot_prod_avx.asm -o gf_vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_vect_dot_prod_sse.asm -o gf_vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_vect_mul_avx.asm -o gf_vect_mul_avx.syso
//go:generate yasm -f elf64 gf_vect_mul_sse.asm -o gf_vect_mul_sse.syso

43
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/erasure_yasm_windows.go generated vendored
View File

@ -1,43 +0,0 @@
// !build amd64
package erasure
//go:generate yasm -f win64 ec_multibinary.asm -o ec_multibinary.syso
//go:generate yasm -f win64 gf_2vect_mad_avx2.asm -o gf_2vect_mad_avx2.syso
//go:generate yasm -f win64 gf_2vect_mad_avx.asm -o gf_2vect_mad_avx.syso
//go:generate yasm -f win64 gf_2vect_mad_sse.asm -o gf_2vect_mad_sse.syso
//go:generate yasm -f win64 gf_3vect_mad_avx2.asm -o gf_3vect_mad_avx2.syso
//go:generate yasm -f win64 gf_3vect_mad_avx.asm -o gf_3vect_mad_avx.syso
//go:generate yasm -f win64 gf_3vect_mad_sse.asm -o gf_3vect_mad_sse.syso
//go:generate yasm -f win64 gf_4vect_mad_avx2.asm -o gf_4vect_mad_avx2.syso
//go:generate yasm -f win64 gf_4vect_mad_avx.asm -o gf_4vect_mad_avx.syso
//go:generate yasm -f win64 gf_4vect_mad_sse.asm -o gf_4vect_mad_sse.syso
//go:generate yasm -f win64 gf_5vect_mad_avx2.asm -o gf_5vect_mad_avx2.syso
//go:generate yasm -f win64 gf_5vect_mad_avx.asm -o gf_5vect_mad_avx.syso
//go:generate yasm -f win64 gf_5vect_mad_sse.asm -o gf_5vect_mad_sse.syso
//go:generate yasm -f win64 gf_6vect_mad_avx2.asm -o gf_6vect_mad_avx2.syso
//go:generate yasm -f win64 gf_6vect_mad_avx.asm -o gf_6vect_mad_avx.syso
//go:generate yasm -f win64 gf_6vect_mad_sse.asm -o gf_6vect_mad_sse.syso
//go:generate yasm -f win64 gf_vect_mad_avx2.asm -o gf_vect_mad_avx2.syso
//go:generate yasm -f win64 gf_vect_mad_avx.asm -o gf_vect_mad_avx.syso
//go:generate yasm -f win64 gf_vect_mad_sse.asm -o gf_vect_mad_sse.syso
//go:generate yasm -f win64 gf_2vect_dot_prod_avx2.asm -o gf_2vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_2vect_dot_prod_avx.asm -o gf_2vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_2vect_dot_prod_sse.asm -o gf_2vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_3vect_dot_prod_avx2.asm -o gf_3vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_3vect_dot_prod_avx.asm -o gf_3vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_3vect_dot_prod_sse.asm -o gf_3vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_4vect_dot_prod_avx2.asm -o gf_4vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_4vect_dot_prod_avx.asm -o gf_4vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_4vect_dot_prod_sse.asm -o gf_4vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_5vect_dot_prod_avx2.asm -o gf_5vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_5vect_dot_prod_avx.asm -o gf_5vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_5vect_dot_prod_sse.asm -o gf_5vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_6vect_dot_prod_avx2.asm -o gf_6vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_6vect_dot_prod_avx.asm -o gf_6vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_6vect_dot_prod_sse.asm -o gf_6vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_vect_dot_prod_avx2.asm -o gf_vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_vect_dot_prod_avx.asm -o gf_vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_vect_dot_prod_sse.asm -o gf_vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_vect_mul_avx.asm -o gf_vect_mul_avx.syso
//go:generate yasm -f win64 gf_vect_mul_sse.asm -o gf_vect_mul_sse.syso

381
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_2vect_dot_prod_avx.asm generated vendored
View File

@ -1,381 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_dot_prod_avx(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_2VECT_DOT_PROD_AVX _gf_2vect_dot_prod_avx
%else
%define GF_2VECT_DOT_PROD_AVX gf_2vect_dot_prod_avx
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 3*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_reg r12, 3*16 + 0*8
save_reg r13, 3*16 + 1*8
save_reg r14, 3*16 + 2*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
mov r12, [rsp + 3*16 + 0*8]
mov r13, [rsp + 3*16 + 1*8]
mov r14, [rsp + 3*16 + 2*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp4 trans2
%define tmp4_m var(0)
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*1 ;1 local variable
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*1 ;1 local variable
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define vec_i tmp2
%define ptr tmp3
%define dest2 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ; 64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm8
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xmm5
%define xgft2_hi xmm4
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%else ;32-bit code
%define xmask0f xmm4
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%endif
align 16
global GF_2VECT_DOT_PROD_AVX:function
func(GF_2VECT_DOT_PROD_AVX)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16
vpxor xp1, xp1
vpxor xp2, xp2
mov tmp, mul_array
xor vec_i, vec_i
.next_vect
SLDR src, src_m
mov ptr, [src+vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%ifidn PS,8 ; 64-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
XLDR x0, [ptr+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ; 32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_2VECT_DOT_PROD_AVX, 02, 04, 0191

398
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_2vect_dot_prod_avx2.asm generated vendored
View File

@ -1,398 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_2VECT_DOT_PROD_AVX2 _gf_2vect_dot_prod_avx2
%else
%define GF_2VECT_DOT_PROD_AVX2 gf_2vect_dot_prod_avx2
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 3*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
save_reg r12, 3*16 + 0*8
save_reg r13, 3*16 + 1*8
save_reg r14, 3*16 + 2*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
mov r12, [rsp + 3*16 + 0*8]
mov r13, [rsp + 3*16 + 1*8]
mov r14, [rsp + 3*16 + 2*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp.w edx
%define tmp.b dl
%define tmp2 edi
%define tmp3 trans2
%define tmp4 trans2
%define tmp4_m var(0)
%define return eax
%macro SLDR 2 ;stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*1 ;1 local variable
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*1 ;1 local variable
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define vec_i tmp2
%define ptr tmp3
%define dest2 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f ymm8
%define xmask0fx xmm8
%define xgft1_lo ymm7
%define xgft1_hi ymm6
%define xgft2_lo ymm5
%define xgft2_hi ymm4
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%else ;32-bit code
%define xmask0f ymm7
%define xmask0fx xmm7
%define xgft1_lo ymm5
%define xgft1_hi ymm4
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%endif
align 16
global GF_2VECT_DOT_PROD_AVX2:function
func(GF_2VECT_DOT_PROD_AVX2)
FUNC_SAVE
SLDR len, len_m
sub len, 32
SSTR len_m, len
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop32
vpxor xp1, xp1
vpxor xp2, xp2
mov tmp, mul_array
xor vec_i, vec_i
.next_vect
SLDR src, src_m
mov ptr, [src+vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft1_lo, xgft1_lo, xgft1_lo, 0x00 ; swapped to lo | lo
%ifidn PS,8 ; 64-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
XLDR x0, [ptr+pos] ;Get next source vector
add tmp, 32
add vec_i, PS
%else
XLDR x0, [ptr+pos] ;Get next source vector
%endif
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ; 32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
add tmp, 32
add vec_i, PS
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR len, len_m
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop32 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_2VECT_DOT_PROD_AVX2, 04, 04, 0196

383
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_2vect_dot_prod_sse.asm generated vendored
View File

@ -1,383 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_dot_prod_sse(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_2VECT_DOT_PROD_SSE _gf_2vect_dot_prod_sse
%else
%define GF_2VECT_DOT_PROD_SSE gf_2vect_dot_prod_sse
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 3*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_reg r12, 3*16 + 0*8
save_reg r13, 3*16 + 1*8
save_reg r14, 3*16 + 2*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp + 0*16]
movdqa xmm7, [rsp + 1*16]
movdqa xmm8, [rsp + 2*16]
mov r12, [rsp + 3*16 + 0*8]
mov r13, [rsp + 3*16 + 1*8]
mov r14, [rsp + 3*16 + 2*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp4 trans2
%define tmp4_m var(0)
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*1 ;1 local variable
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*1 ;1 local variable
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define vec_i tmp2
%define ptr tmp3
%define dest2 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm8
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xmm5
%define xgft2_hi xmm4
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%else ;32-bit code
%define xmask0f xmm4
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%endif
align 16
global GF_2VECT_DOT_PROD_SSE:function
func(GF_2VECT_DOT_PROD_SSE)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16
pxor xp1, xp1
pxor xp2, xp2
mov tmp, mul_array
xor vec_i, vec_i
.next_vect
SLDR src, src_m
mov ptr, [src+vec_i]
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%ifidn PS,8 ;64-bit code
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
XLDR x0, [ptr+pos] ;Get next source vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ;32-bit code
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pxor xp2, xgft2_hi ;xp2 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_2VECT_DOT_PROD_SSE, 00, 03, 0062

264
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_2vect_mad_avx.asm generated vendored
View File

@ -1,264 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_mad_avx(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_2VECT_MAD_AVX _gf_2vect_mad_avx
%else
%define GF_2VECT_MAD_AVX gf_2vect_mad_avx
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define stack_size 16*9 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
save_reg r12, 9*16 + 0*8
save_reg r15, 9*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
mov r12, [rsp + 9*16 + 0*8]
mov r15, [rsp + 9*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_2vect_mad_avx(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp2
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm14
%define xgft1_lo xmm13
%define xgft1_hi xmm12
%define xgft2_lo xmm11
%define xgft2_hi xmm10
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xd1 xmm6
%define xd2 xmm7
%define xtmpd1 xmm8
%define xtmpd2 xmm9
align 16
global GF_2VECT_MAD_AVX:function
func(GF_2VECT_MAD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xgft2_hi, [tmp+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
mov dest2, [dest1+PS]
mov dest1, [dest1]
XLDR xtmpd1, [dest1+len] ;backup the last 16 bytes in dest
XLDR xtmpd2, [dest2+len] ;backup the last 16 bytes in dest
.loop16
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
.loop16_overlap:
XLDR x0, [src+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
vpshufb xtmph2, xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2 ;xd2 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
vmovdqa xd1, xtmpd1 ;Restore xd1
vmovdqa xd2, xtmpd2 ;Restore xd2
jmp .loop16_overlap ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_2VECT_MAD_AVX, 02, 00, 0204

280
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_2vect_mad_avx2.asm generated vendored
View File

@ -1,280 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_2VECT_MAD_AVX2 _gf_2vect_mad_avx2
%else
%define GF_2VECT_MAD_AVX2 gf_2vect_mad_avx2
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define stack_size 16*9 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
save_reg r12, 9*16 + 0*8
save_reg r15, 9*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
mov r12, [rsp + 9*16 + 0*8]
mov r15, [rsp + 9*16 + 1*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_2vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp2
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm14
%define xmask0fx xmm14
%define xgft1_lo ymm13
%define xgft1_hi ymm12
%define xgft2_lo ymm11
%define xgft2_hi ymm10
%define x0 ymm0
%define xtmpa ymm1
%define xtmph1 ymm2
%define xtmpl1 ymm3
%define xtmph2 ymm4
%define xtmpl2 ymm5
%define xd1 ymm6
%define xd2 ymm7
%define xtmpd1 ymm8
%define xtmpd2 ymm9
align 16
global GF_2VECT_MAD_AVX2:function
func(GF_2VECT_MAD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft1_lo, xgft1_lo, xgft1_lo, 0x00 ; swapped to lo | lo
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
mov dest2, [dest1+PS] ; reuse mul_array
mov dest1, [dest1]
XLDR xtmpd1, [dest1+len] ;backup the last 16 bytes in dest
XLDR xtmpd2, [dest2+len] ;backup the last 16 bytes in dest
.loop32
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
.loop32_overlap:
XLDR x0, [src+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
vpshufb xtmph2, xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2 ;xd2 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-32
vmovdqa xd1, xtmpd1 ;Restore xd1
vmovdqa xd2, xtmpd2 ;Restore xd2
jmp .loop32_overlap ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_2VECT_MAD_AVX2, 04, 00, 0205

267
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_2vect_mad_sse.asm generated vendored
View File

@ -1,267 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_mad_sse(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_2VECT_MAD_SSE _gf_2vect_mad_sse
%else
%define GF_2VECT_MAD_SSE gf_2vect_mad_sse
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define stack_size 16*9 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
save_reg r12, 9*16 + 0*8
save_reg r15, 9*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
mov r12, [rsp + 9*16 + 0*8]
mov r15, [rsp + 9*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_2vect_mad_sse(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp2
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm14
%define xgft1_lo xmm13
%define xgft1_hi xmm12
%define xgft2_lo xmm11
%define xgft2_hi xmm10
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xd1 xmm6
%define xd2 xmm7
%define xtmpd1 xmm8
%define xtmpd2 xmm9
align 16
global GF_2VECT_MAD_SSE:function
func(GF_2VECT_MAD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
movdqu xgft1_lo,[tmp] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xgft2_hi, [tmp+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
mov dest2, [dest1+PS]
mov dest1, [dest1]
XLDR xtmpd1, [dest1+len] ;backup the last 16 bytes in dest
XLDR xtmpd2, [dest2+len] ;backup the last 16 bytes in dest
.loop16:
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
.loop16_overlap:
XLDR x0, [src+pos] ;Get next source vector
movdqa xtmph1, xgft1_hi ;Reload const array registers
movdqa xtmpl1, xgft1_lo
movdqa xtmph2, xgft2_hi ;Reload const array registers
movdqa xtmpl2, xgft2_lo
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pxor xd1, xtmph1
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pxor xd2, xtmph2
XSTR [dest1+pos], xd1 ;Store result
XSTR [dest2+pos], xd2 ;Store result
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
movdqa xd1, xtmpd1 ;Restore xd1
movdqa xd2, xtmpd2 ;Restore xd2
jmp .loop16_overlap ;Do one more overlap pass
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f:
ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_2VECT_MAD_SSE, 00, 00, 0203

421
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_3vect_dot_prod_avx.asm generated vendored
View File

@ -1,421 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_3vect_dot_prod_avx(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_DOT_PROD_AVX _gf_3vect_dot_prod_avx
%else
%define GF_3VECT_DOT_PROD_AVX gf_3vect_dot_prod_avx
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 6*16 + 5*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_reg r12, 6*16 + 0*8
save_reg r13, 6*16 + 1*8
save_reg r14, 6*16 + 2*8
save_reg r15, 6*16 + 3*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
mov r12, [rsp + 6*16 + 0*8]
mov r13, [rsp + 6*16 + 1*8]
mov r14, [rsp + 6*16 + 2*8]
mov r15, [rsp + 6*16 + 3*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; var1
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define arg5 trans2
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp3_m var(0)
%define tmp4 trans2
%define tmp4_m var(1)
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*2 ;2 local variables
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*2 ;2 local variables
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define dest2 tmp3
%define dest3 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp3_m
%define dest3_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ; 64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm11
%define xgft1_lo xmm10
%define xgft1_hi xmm9
%define xgft2_lo xmm8
%define xgft2_hi xmm7
%define xgft3_lo xmm6
%define xgft3_hi xmm5
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%else
%define xmask0f xmm7
%define xgft1_lo xmm6
%define xgft1_hi xmm5
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define xgft3_lo xgft1_lo
%define xgft3_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%endif
align 16
global GF_3VECT_DOT_PROD_AVX:function
func(GF_3VECT_DOT_PROD_AVX)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest3, [dest1+2*PS]
SSTR dest3_m, dest3
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16:
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
SLDR src, src_m
mov ptr, [src+vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%ifidn PS,8 ; 64-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+vec*(64/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
vmovdqu xgft3_hi, [tmp+vec*(64/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
add tmp, 32
add vec_i, PS
%endif
XLDR x0, [ptr+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ; 32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
%ifidn PS,4 ; 32-bit code
sal vec, 1
vmovdqu xgft3_lo, [tmp+vec*(32/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
vmovdqu xgft3_hi, [tmp+vec*(32/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
sar vec, 1
add tmp, 32
add vec_i, PS
%endif
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR dest3, dest3_m
XSTR [dest3+pos], xp3
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_3VECT_DOT_PROD_AVX, 02, 04, 0192

441
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_3vect_dot_prod_avx2.asm generated vendored
View File

@ -1,441 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_3vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_DOT_PROD_AVX2 _gf_3vect_dot_prod_avx2
%else
%define GF_3VECT_DOT_PROD_AVX2 gf_3vect_dot_prod_avx2
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 6*16 + 5*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
vmovdqa [rsp + 3*16], xmm9
vmovdqa [rsp + 4*16], xmm10
vmovdqa [rsp + 5*16], xmm11
save_reg r12, 6*16 + 0*8
save_reg r13, 6*16 + 1*8
save_reg r14, 6*16 + 2*8
save_reg r15, 6*16 + 3*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
mov r12, [rsp + 6*16 + 0*8]
mov r13, [rsp + 6*16 + 1*8]
mov r14, [rsp + 6*16 + 2*8]
mov r15, [rsp + 6*16 + 3*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; var1
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define arg5 trans2
%define tmp edx
%define tmp.w edx
%define tmp.b dl
%define tmp2 edi
%define tmp3 trans2
%define tmp3_m var(0)
%define tmp4 trans2
%define tmp4_m var(1)
%define return eax
%macro SLDR 2 ;stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*2 ;2 local variables
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*2 ;2 local variables
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define dest2 tmp3
%define dest3 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp3_m
%define dest3_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f ymm11
%define xmask0fx xmm11
%define xgft1_lo ymm10
%define xgft1_hi ymm9
%define xgft2_lo ymm8
%define xgft2_hi ymm7
%define xgft3_lo ymm6
%define xgft3_hi ymm5
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%define xp3 ymm4
%else
%define xmask0f ymm7
%define xmask0fx xmm7
%define xgft1_lo ymm6
%define xgft1_hi ymm5
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define xgft3_lo xgft1_lo
%define xgft3_hi xgft1_hi
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%define xp3 ymm4
%endif
align 16
global GF_3VECT_DOT_PROD_AVX2:function
func(GF_3VECT_DOT_PROD_AVX2)
FUNC_SAVE
SLDR len, len_m
sub len, 32
SSTR len_m, len
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest3, [dest1+2*PS]
SSTR dest3_m, dest3
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop32:
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
SLDR src, src_m
mov ptr, [src+vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft1_lo, xgft1_lo, xgft1_lo, 0x00 ; swapped to lo | lo
%ifidn PS,8 ; 64-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
vmovdqu xgft3_lo, [tmp+vec*(64/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft3_lo, xgft3_lo, xgft3_lo, 0x00 ; swapped to lo | lo
add tmp, 32
add vec_i, PS
%endif
XLDR x0, [ptr+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ; 32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
%ifidn PS,4 ; 32-bit code
sal vec, 1
vmovdqu xgft3_lo, [tmp+vec*(32/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft3_lo, xgft3_lo, xgft3_lo, 0x00 ; swapped to lo | lo
sar vec, 1
add tmp, 32
add vec_i, PS
%endif
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR dest3, dest3_m
XSTR [dest3+pos], xp3
SLDR len, len_m
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop32 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_3VECT_DOT_PROD_AVX2, 04, 04, 0197

422
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_3vect_dot_prod_sse.asm generated vendored
View File

@ -1,422 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_3vect_dot_prod_sse(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_DOT_PROD_SSE _gf_3vect_dot_prod_sse
%else
%define GF_3VECT_DOT_PROD_SSE gf_3vect_dot_prod_sse
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 6*16 + 5*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_reg r12, 6*16 + 0*8
save_reg r13, 6*16 + 1*8
save_reg r14, 6*16 + 2*8
save_reg r15, 6*16 + 3*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp + 0*16]
movdqa xmm7, [rsp + 1*16]
movdqa xmm8, [rsp + 2*16]
movdqa xmm9, [rsp + 3*16]
movdqa xmm10, [rsp + 4*16]
movdqa xmm11, [rsp + 5*16]
mov r12, [rsp + 6*16 + 0*8]
mov r13, [rsp + 6*16 + 1*8]
mov r14, [rsp + 6*16 + 2*8]
mov r15, [rsp + 6*16 + 3*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; var1
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define arg5 trans2
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp3_m var(0)
%define tmp4 trans2
%define tmp4_m var(1)
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*2 ;2 local variables
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*2 ;2 local variables
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define dest2 tmp3
%define dest3 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp3_m
%define dest3_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
%ifidn PS,8 ; 64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm11
%define xgft1_lo xmm2
%define xgft1_hi xmm3
%define xgft2_lo xmm4
%define xgft2_hi xmm7
%define xgft3_lo xmm6
%define xgft3_hi xmm5
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm10
%define xp2 xmm9
%define xp3 xmm8
%else
%define xmask0f xmm7
%define xgft1_lo xmm6
%define xgft1_hi xmm5
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define xgft3_lo xgft1_lo
%define xgft3_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%endif
align 16
global GF_3VECT_DOT_PROD_SSE:function
func(GF_3VECT_DOT_PROD_SSE)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest3, [dest1+2*PS]
SSTR dest3_m, dest3
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16:
pxor xp1, xp1
pxor xp2, xp2
pxor xp3, xp3
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
SLDR src, src_m
mov ptr, [src+vec_i]
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%ifidn PS,8 ;64-bit code
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
movdqu xgft3_lo, [tmp+vec*(64/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
movdqu xgft3_hi, [tmp+vec*(64/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
add tmp, 32
add vec_i, PS
%endif
XLDR x0, [ptr+pos] ;Get next source vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ;32-bit code
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
%endif
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pxor xp2, xgft2_hi ;xp2 += partial
%ifidn PS,4 ;32-bit code
sal vec, 1
movdqu xgft3_lo, [tmp+vec*(32/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
movdqu xgft3_hi, [tmp+vec*(32/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
sar vec, 1
add tmp, 32
add vec_i, PS
%endif
pshufb xgft3_hi, x0 ;Lookup mul table of high nibble
pshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft3_hi, xgft3_lo ;GF add high and low partials
pxor xp3, xgft3_hi ;xp3 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR dest3, dest3_m
XSTR [dest3+pos], xp3
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_3VECT_DOT_PROD_SSE, 00, 05, 0063

315
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_3vect_mad_avx.asm generated vendored
View File

@ -1,315 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_3vect_mad_avx(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_MAD_AVX _gf_3vect_mad_avx
%else
%define GF_3VECT_MAD_AVX gf_3vect_mad_avx
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
vmovdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
vmovdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_3vect_mad_avx(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft1_lo xmm14
%define xgft1_hi xmm13
%define xgft2_lo xmm12
%define xgft2_hi xmm11
%define xgft3_lo xmm10
%define xgft3_hi xmm9
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xd1 xmm8
%define xd2 xtmpl1
%define xd3 xtmph1
align 16
global GF_3VECT_MAD_AVX:function
func(GF_3VECT_MAD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xgft2_hi, [tmp+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xgft3_lo, [tmp+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
vmovdqu xgft3_hi, [tmp+2*vec+16]; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
mov dest2, [dest1+PS] ; reuse mul_array
mov dest3, [dest1+2*PS] ; reuse vec_i
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
XLDR xd1, [dest1+pos] ;Get next dest vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
XLDR xd2, [dest2+pos] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+pos] ;reuse xtmph1. Get next dest vector
; dest2
vpshufb xtmph2, xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2 ;xd2 += partial
; dest3
vpshufb xtmph3, xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xtmpl3 ;GF add high and low partials
vpxor xd3, xd3, xtmph3 ;xd3 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
XSTR [dest3+pos], xd3
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
mov tmp, len ;Overlapped offset length-16
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+tmp] ;reuse xtmph1. Get next dest vector
sub len, pos
movdqa xtmph3, [constip16] ;Load const of i + 16
vpinsrb xtmpl3, xtmpl3, len.w, 15
vpshufb xtmpl3, xtmpl3, xmask0f ;Broadcast len to all bytes
vpcmpgtb xtmpl3, xtmpl3, xtmph3
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xgft1_hi, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_hi, xgft1_lo ;GF add high and low partials
vpand xgft1_hi, xgft1_hi, xtmpl3
vpxor xd1, xd1, xgft1_hi
; dest2
vpshufb xgft2_hi, xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_hi, xgft2_lo ;GF add high and low partials
vpand xgft2_hi, xgft2_hi, xtmpl3
vpxor xd2, xd2, xgft2_hi
; dest3
vpshufb xgft3_hi, xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_hi, xgft3_lo ;GF add high and low partials
vpand xgft3_hi, xgft3_hi, xtmpl3
vpxor xd3, xd3, xgft3_hi
XSTR [dest1+tmp], xd1
XSTR [dest2+tmp], xd2
XSTR [dest3+tmp], xd3
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_3VECT_MAD_AVX, 02, 00, 0207

347
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_3vect_mad_avx2.asm generated vendored
View File

@ -1,347 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_3vect_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_MAD_AVX2 _gf_3vect_mad_avx2
%else
%define GF_3VECT_MAD_AVX2 gf_3vect_mad_avx2
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
vmovdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
vmovdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_3vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm15
%define xmask0fx xmm15
%define xgft1_lo ymm14
%define xgft1_hi ymm13
%define xgft2_lo ymm12
%define xgft3_lo ymm11
%define x0 ymm0
%define xtmpa ymm1
%define xtmph1 ymm2
%define xtmpl1 ymm3
%define xtmph2 ymm4
%define xtmpl2 ymm5
%define xtmpl2x xmm5
%define xtmph3 ymm6
%define xtmpl3 ymm7
%define xtmpl3x xmm7
%define xd1 ymm8
%define xd2 ymm9
%define xd3 ymm10
align 16
global GF_3VECT_MAD_AVX2:function
func(GF_3VECT_MAD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft1_lo, xgft1_lo, xgft1_lo, 0x00 ; swapped to lo | lo
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xgft3_lo, [tmp+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
mov dest2, [dest1+PS] ; reuse mul_array
mov dest3, [dest1+2*PS] ; reuse vec_i
mov dest1, [dest1]
.loop32:
XLDR x0, [src+pos] ;Get next source vector
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
XLDR xd3, [dest3+pos] ;Get next dest vector
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xtmpl2, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
vperm2i128 xtmph3, xgft3_lo, xgft3_lo, 0x11 ; swapped to hi | hi
vperm2i128 xtmpl3, xgft3_lo, xgft3_lo, 0x00 ; swapped to lo | lo
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
; dest2
vpshufb xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xtmph2 ;xd2 += partial
; dest3
vpshufb xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmpl3 ;GF add high and low partials
vpxor xd3, xtmph3 ;xd3 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
XSTR [dest3+pos], xd3
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
.lessthan32:
;; Tail len
;; Do one more overlap pass
mov tmp.b, 0x1f
vpinsrb xtmpl2x, xtmpl2x, tmp.w, 0
vpbroadcastb xtmpl2, xtmpl2x ;Construct mask 0x1f1f1f...
mov tmp, len ;Overlapped offset length-32
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
XLDR xd3, [dest3+tmp] ;Get next dest vector
sub len, pos
vmovdqa xtmph3, [constip32] ;Load const of i + 32
vpinsrb xtmpl3x, xtmpl3x, len.w, 15
vinserti128 xtmpl3, xtmpl3, xtmpl3x, 1 ;swapped to xtmpl3x | xtmpl3x
vpshufb xtmpl3, xtmpl3, xtmpl2 ;Broadcast len to all bytes. xtmpl2=0x1f1f1f...
vpcmpgtb xtmpl3, xtmpl3, xtmph3
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
vperm2i128 xtmph3, xgft3_lo, xgft3_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft3_lo, xgft3_lo, xgft3_lo, 0x00 ; swapped to lo | lo
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpand xtmph1, xtmph1, xtmpl3
vpxor xd1, xd1, xtmph1 ;xd1 += partial
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xgft2_lo ;GF add high and low partials
vpand xtmph2, xtmph2, xtmpl3
vpxor xd2, xd2, xtmph2 ;xd2 += partial
; dest3
vpshufb xtmph3, xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xgft3_lo ;GF add high and low partials
vpand xtmph3, xtmph3, xtmpl3
vpxor xd3, xd3, xtmph3 ;xd3 += partial
XSTR [dest1+tmp], xd1
XSTR [dest2+tmp], xd2
XSTR [dest3+tmp], xd3
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 32
constip32:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
ddq 0xe0e1e2e3e4e5e6e7e8e9eaebecedeeef
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_3VECT_MAD_AVX2, 04, 00, 0208

326
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_3vect_mad_sse.asm generated vendored
View File

@ -1,326 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_3vect_mad_sse(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_MAD_SSE _gf_3vect_mad_sse
%else
%define GF_3VECT_MAD_SSE gf_3vect_mad_sse
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_3vect_mad_sse(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft1_lo xmm14
%define xgft1_hi xmm13
%define xgft2_lo xmm12
%define xgft2_hi xmm11
%define xgft3_lo xmm10
%define xgft3_hi xmm9
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xd1 xmm8
%define xd2 xtmpl1
%define xd3 xtmph1
align 16
global GF_3VECT_MAD_SSE:function
func(GF_3VECT_MAD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xgft2_hi, [tmp+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
movdqu xgft3_lo, [tmp+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqu xgft3_hi, [tmp+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
mov dest2, [dest1+PS] ; reuse mul_array
mov dest3, [dest1+2*PS] ; reuse vec_i
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
movdqa xtmph1, xgft1_hi ;Reload const array registers
movdqa xtmpl1, xgft1_lo
movdqa xtmph2, xgft2_hi ;Reload const array registers
movdqa xtmpl2, xgft2_lo
movdqa xtmph3, xgft3_hi ;Reload const array registers
movdqa xtmpl3, xgft3_lo
XLDR xd1, [dest1+pos] ;Get next dest vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
; dest1
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pxor xd1, xtmph1
XLDR xd2, [dest2+pos] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+pos] ;reuse xtmph1. Get next dest vector
; dest2
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pxor xd2, xtmph2
; dest3
pshufb xtmph3, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xtmph3, xtmpl3 ;GF add high and low partials
pxor xd3, xtmph3
XSTR [dest1+pos], xd1 ;Store result
XSTR [dest2+pos], xd2 ;Store result
XSTR [dest3+pos], xd3 ;Store result
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
mov tmp, len ;Overlapped offset length-16
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+tmp] ;reuse xtmph1. Get next dest vector
sub len, pos
movdqa xtmph3, [constip16] ;Load const of i + 16
pinsrb xtmpl3, len.w, 15
pshufb xtmpl3, xmask0f ;Broadcast len to all bytes
pcmpgtb xtmpl3, xtmph3
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
; dest1
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pand xgft1_hi, xtmpl3
pxor xd1, xgft1_hi
; dest2
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pand xgft2_hi, xtmpl3
pxor xd2, xgft2_hi
; dest3
pshufb xgft3_hi, x0 ;Lookup mul table of high nibble
pshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft3_hi, xgft3_lo ;GF add high and low partials
pand xgft3_hi, xtmpl3
pxor xd3, xgft3_hi
XSTR [dest1+tmp], xd1 ;Store result
XSTR [dest2+tmp], xd2 ;Store result
XSTR [dest3+tmp], xd3 ;Store result
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f:
ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_3VECT_MAD_SSE, 00, 00, 0206

489
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_4vect_dot_prod_avx.asm generated vendored
View File

@ -1,489 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_4vect_dot_prod_avx(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_4VECT_DOT_PROD_AVX _gf_4vect_dot_prod_avx
%else
%define GF_4VECT_DOT_PROD_AVX gf_4vect_dot_prod_avx
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 9*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_xmm128 xmm12, 6*16
save_xmm128 xmm13, 7*16
save_xmm128 xmm14, 8*16
save_reg r12, 9*16 + 0*8
save_reg r13, 9*16 + 1*8
save_reg r14, 9*16 + 2*8
save_reg r15, 9*16 + 3*8
save_reg rdi, 9*16 + 4*8
save_reg rsi, 9*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
mov r12, [rsp + 9*16 + 0*8]
mov r13, [rsp + 9*16 + 1*8]
mov r14, [rsp + 9*16 + 2*8]
mov r15, [rsp + 9*16 + 3*8]
mov rdi, [rsp + 9*16 + 4*8]
mov rsi, [rsp + 9*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; var1
;;; var2
;;; var3
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define arg5 trans2
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp3_m var(0)
%define tmp4 trans2
%define tmp4_m var(1)
%define tmp5 trans2
%define tmp5_m var(2)
%define tmp6 trans2
%define tmp6_m var(3)
%define return eax
%macro SLDR 2 ;stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*4 ;4 local variables
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*4 ;4 local variables
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define dest2 tmp3
%define dest3 tmp4
%define dest4 tmp5
%define vskip3 tmp6
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp3_m
%define dest3_m tmp4_m
%define dest4_m tmp5_m
%define vskip3_m tmp6_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ; 64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm14
%define xgft1_lo xmm13
%define xgft1_hi xmm12
%define xgft2_lo xmm11
%define xgft2_hi xmm10
%define xgft3_lo xmm9
%define xgft3_hi xmm8
%define xgft4_lo xmm7
%define xgft4_hi xmm6
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%define xp4 xmm5
%else
%define xmm_trans xmm7 ;reuse xmask0f and xgft1_lo
%define xmask0f xmm_trans
%define xgft1_lo xmm_trans
%define xgft1_hi xmm6
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define xgft3_lo xgft1_lo
%define xgft3_hi xgft1_hi
%define xgft4_lo xgft1_lo
%define xgft4_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%define xp4 xmm5
%endif
align 16
global GF_4VECT_DOT_PROD_AVX:function
func(GF_4VECT_DOT_PROD_AVX)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov vskip3, vec
imul vskip3, 96
SSTR vskip3_m, vskip3
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest3, [dest1+2*PS]
SSTR dest3_m, dest3
mov dest4, [dest1+3*PS]
SSTR dest4_m, dest4
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16:
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
vpxor xp4, xp4
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
SLDR src, src_m
mov ptr, [src+vec_i]
%ifidn PS,8 ;64-bit code
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+vec*(64/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
vmovdqu xgft3_hi, [tmp+vec*(64/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
vmovdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
vmovdqu xgft4_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
XLDR x0, [ptr+pos] ;Get next source vector
add tmp, 32
add vec_i, PS
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
%else ;32-bit code
XLDR x0, [ptr+pos] ;Get next source vector
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%endif
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ;32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
%ifidn PS,4 ;32-bit code
sal vec, 1
vmovdqu xgft3_lo, [tmp+vec*(32/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
vmovdqu xgft3_hi, [tmp+vec*(32/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
sar vec, 1
%endif
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
%ifidn PS,4 ;32-bit code
SLDR vskip3, vskip3_m
vmovdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
vmovdqu xgft4_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
add tmp, 32
add vec_i, PS
%endif
vpshufb xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_lo ;GF add high and low partials
vpxor xp4, xgft4_hi ;xp4 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR dest3, dest3_m
XSTR [dest3+pos], xp3
SLDR dest4, dest4_m
XSTR [dest4+pos], xp4
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_4VECT_DOT_PROD_AVX, 02, 04, 0193

510
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_4vect_dot_prod_avx2.asm generated vendored
View File

@ -1,510 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_4vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_4VECT_DOT_PROD_AVX2 _gf_4vect_dot_prod_avx2
%else
%define GF_4VECT_DOT_PROD_AVX2 gf_4vect_dot_prod_avx2
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 9*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
vmovdqa [rsp + 3*16], xmm9
vmovdqa [rsp + 4*16], xmm10
vmovdqa [rsp + 5*16], xmm11
vmovdqa [rsp + 6*16], xmm12
vmovdqa [rsp + 7*16], xmm13
vmovdqa [rsp + 8*16], xmm14
save_reg r12, 9*16 + 0*8
save_reg r13, 9*16 + 1*8
save_reg r14, 9*16 + 2*8
save_reg r15, 9*16 + 3*8
save_reg rdi, 9*16 + 4*8
save_reg rsi, 9*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
mov r12, [rsp + 9*16 + 0*8]
mov r13, [rsp + 9*16 + 1*8]
mov r14, [rsp + 9*16 + 2*8]
mov r15, [rsp + 9*16 + 3*8]
mov rdi, [rsp + 9*16 + 4*8]
mov rsi, [rsp + 9*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; var1
;;; var2
;;; var3
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define arg5 trans2
%define tmp edx
%define tmp.w edx
%define tmp.b dl
%define tmp2 edi
%define tmp3 trans2
%define tmp3_m var(0)
%define tmp4 trans2
%define tmp4_m var(1)
%define tmp5 trans2
%define tmp5_m var(2)
%define tmp6 trans2
%define tmp6_m var(3)
%define return eax
%macro SLDR 2 ;stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*4 ;4 local variables
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*4 ;4 local variables
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define dest2 tmp3
%define dest3 tmp4
%define dest4 tmp5
%define vskip3 tmp6
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp3_m
%define dest3_m tmp4_m
%define dest4_m tmp5_m
%define vskip3_m tmp6_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f ymm14
%define xmask0fx xmm14
%define xgft1_lo ymm13
%define xgft1_hi ymm12
%define xgft2_lo ymm11
%define xgft2_hi ymm10
%define xgft3_lo ymm9
%define xgft3_hi ymm8
%define xgft4_lo ymm7
%define xgft4_hi ymm6
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%define xp3 ymm4
%define xp4 ymm5
%else
%define ymm_trans ymm7 ;reuse xmask0f and xgft1_hi
%define xmask0f ymm_trans
%define xmask0fx xmm7
%define xgft1_lo ymm6
%define xgft1_hi ymm_trans
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define xgft3_lo xgft1_lo
%define xgft3_hi xgft1_hi
%define xgft4_lo xgft1_lo
%define xgft4_hi xgft1_hi
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%define xp3 ymm4
%define xp4 ymm5
%endif
align 16
global GF_4VECT_DOT_PROD_AVX2:function
func(GF_4VECT_DOT_PROD_AVX2)
FUNC_SAVE
SLDR len, len_m
sub len, 32
SSTR len_m, len
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
mov vskip3, vec
imul vskip3, 96
SSTR vskip3_m, vskip3
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest3, [dest1+2*PS]
SSTR dest3_m, dest3
mov dest4, [dest1+3*PS]
SSTR dest4_m, dest4
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop32:
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
vpxor xp4, xp4
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
SLDR src, src_m
mov ptr, [src+vec_i]
XLDR x0, [ptr+pos] ;Get next source vector
add vec_i, PS
%ifidn PS,8 ;64-bit code
vpand xgft4_lo, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xgft4_lo, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xgft4_lo, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+vec*(64/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vmovdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
; " Dx{00}, Dx{10}, ..., Dx{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft4_hi, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
add tmp, 32
%else ;32-bit code
mov cl, 0x0f ;use ecx as a temp variable
vpinsrb xmask0fx, xmask0fx, ecx, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
vpand xgft4_lo, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xgft4_lo, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xgft4_lo, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
%endif
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ; 32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
%ifidn PS,4 ; 32-bit code
sal vec, 1
vmovdqu xgft3_lo, [tmp+vec*(32/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
sar vec, 1
%endif
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
%ifidn PS,4 ; 32-bit code
SLDR vskip3, vskip3_m
vmovdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
; " DX{00}, Dx{10}, ..., Dx{f0}
vperm2i128 xgft4_hi, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
add tmp, 32
%endif
vpshufb xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_lo ;GF add high and low partials
vpxor xp4, xgft4_hi ;xp4 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR dest3, dest3_m
XSTR [dest3+pos], xp3
SLDR dest4, dest4_m
XSTR [dest4+pos], xp4
SLDR len, len_m
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-32
jmp .loop32 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_4VECT_DOT_PROD_AVX2, 04, 04, 0198

491
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_4vect_dot_prod_sse.asm generated vendored
View File

@ -1,491 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_4vect_dot_prod_sse(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_4VECT_DOT_PROD_SSE _gf_4vect_dot_prod_sse
%else
%define GF_4VECT_DOT_PROD_SSE gf_4vect_dot_prod_sse
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 9*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_xmm128 xmm12, 6*16
save_xmm128 xmm13, 7*16
save_xmm128 xmm14, 8*16
save_reg r12, 9*16 + 0*8
save_reg r13, 9*16 + 1*8
save_reg r14, 9*16 + 2*8
save_reg r15, 9*16 + 3*8
save_reg rdi, 9*16 + 4*8
save_reg rsi, 9*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp + 0*16]
movdqa xmm7, [rsp + 1*16]
movdqa xmm8, [rsp + 2*16]
movdqa xmm9, [rsp + 3*16]
movdqa xmm10, [rsp + 4*16]
movdqa xmm11, [rsp + 5*16]
movdqa xmm12, [rsp + 6*16]
movdqa xmm13, [rsp + 7*16]
movdqa xmm14, [rsp + 8*16]
mov r12, [rsp + 9*16 + 0*8]
mov r13, [rsp + 9*16 + 1*8]
mov r14, [rsp + 9*16 + 2*8]
mov r15, [rsp + 9*16 + 3*8]
mov rdi, [rsp + 9*16 + 4*8]
mov rsi, [rsp + 9*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; var1
;;; var2
;;; var3
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define arg5 trans2
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp3_m var(0)
%define tmp4 trans2
%define tmp4_m var(1)
%define tmp5 trans2
%define tmp5_m var(2)
%define tmp6 trans2
%define tmp6_m var(3)
%define return eax
%macro SLDR 2 ;stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*4 ;4 local variables
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*4 ;4 local variables
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define dest2 tmp3
%define dest3 tmp4
%define dest4 tmp5
%define vskip3 tmp6
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp3_m
%define dest3_m tmp4_m
%define dest4_m tmp5_m
%define vskip3_m tmp6_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
%ifidn PS,8 ; 64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm14
%define xgft1_lo xmm2
%define xgft1_hi xmm3
%define xgft2_lo xmm11
%define xgft2_hi xmm4
%define xgft3_lo xmm9
%define xgft3_hi xmm5
%define xgft4_lo xmm7
%define xgft4_hi xmm6
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm8
%define xp2 xmm10
%define xp3 xmm12
%define xp4 xmm13
%else
%define xmm_trans xmm7 ;reuse xmask0f and xgft1_lo
%define xmask0f xmm_trans
%define xgft1_lo xmm_trans
%define xgft1_hi xmm6
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define xgft3_lo xgft1_lo
%define xgft3_hi xgft1_hi
%define xgft4_lo xgft1_lo
%define xgft4_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%define xp4 xmm5
%endif
align 16
global GF_4VECT_DOT_PROD_SSE:function
func(GF_4VECT_DOT_PROD_SSE)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov vskip3, vec
imul vskip3, 96
SSTR vskip3_m, vskip3
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest3, [dest1+2*PS]
SSTR dest3_m, dest3
mov dest4, [dest1+3*PS]
SSTR dest4_m, dest4
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16:
pxor xp1, xp1
pxor xp2, xp2
pxor xp3, xp3
pxor xp4, xp4
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
SLDR src, src_m
mov ptr, [src+vec_i]
%ifidn PS,8 ;64-bit code
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
movdqu xgft3_lo, [tmp+vec*(64/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
movdqu xgft3_hi, [tmp+vec*(64/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
movdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
movdqu xgft4_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
XLDR x0, [ptr+pos] ;Get next source vector
add tmp, 32
add vec_i, PS
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
%else ;32-bit code
XLDR x0, [ptr+pos] ;Get next source vector
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%endif
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ;32-bit code
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
%endif
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pxor xp2, xgft2_hi ;xp2 += partial
%ifidn PS,4 ;32-bit code
sal vec, 1
movdqu xgft3_lo, [tmp+vec*(32/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
movdqu xgft3_hi, [tmp+vec*(32/PS)+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
sar vec, 1
%endif
pshufb xgft3_hi, x0 ;Lookup mul table of high nibble
pshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft3_hi, xgft3_lo ;GF add high and low partials
pxor xp3, xgft3_hi ;xp3 += partial
%ifidn PS,4 ;32-bit code
SLDR vskip3, vskip3_m
movdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
movdqu xgft4_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
add tmp, 32
add vec_i, PS
%endif
pshufb xgft4_hi, x0 ;Lookup mul table of high nibble
pshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft4_hi, xgft4_lo ;GF add high and low partials
pxor xp4, xgft4_hi ;xp4 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR dest3, dest3_m
XSTR [dest3+pos], xp3
SLDR dest4, dest4_m
XSTR [dest4+pos], xp4
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_4VECT_DOT_PROD_SSE, 00, 05, 0064

370
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_4vect_mad_avx.asm generated vendored
View File

@ -1,370 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_4vect_mad_avx(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_4VECT_MAD_AVX _gf_4vect_mad_avx
%else
%define GF_4VECT_MAD_AVX gf_4vect_mad_avx
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define tmp3 r13
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r15, 10*16 + 2*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r15, [rsp + 10*16 + 2*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
;;; gf_4vect_mad_avx(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 tmp2
%define dest4 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft3_hi xmm14
%define xgft4_hi xmm13
%define xgft4_lo xmm12
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xtmph4 xmm8
%define xtmpl4 xmm9
%define xd1 xmm10
%define xd2 xmm11
%define xd3 xtmph1
%define xd4 xtmpl1
align 16
global GF_4VECT_MAD_AVX:function
func(GF_4VECT_MAD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov tmp, vec
sal vec_i, 5 ;Multiply by 32
lea tmp3, [mul_array + vec_i]
sal tmp, 6 ;Multiply by 64
vmovdqu xgft3_hi, [tmp3+tmp+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
sal vec, 5 ;Multiply by 32
add tmp, vec
vmovdqu xgft4_lo, [tmp3+tmp] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
vmovdqu xgft4_hi, [tmp3+tmp+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
mov dest2, [dest1+PS] ; reuse mul_array
mov dest3, [dest1+2*PS]
mov dest4, [dest1+3*PS] ; reuse vec_i
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
vmovdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xtmpl1, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1
XLDR xd3, [dest3+pos] ;Reuse xtmph1, Get next dest vector
XLDR xd4, [dest4+pos] ;Reuse xtmpl1, Get next dest vector
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xtmpl2, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2
; dest3
vpshufb xtmph3, xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xtmpl3, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xtmpl3 ;GF add high and low partials
vpxor xd3, xd3, xtmph3
; dest4
vpshufb xtmph4, xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl4, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph4, xtmph4, xtmpl4 ;GF add high and low partials
vpxor xd4, xd4, xtmph4
XSTR [dest1+pos], xd1 ;Store result
XSTR [dest2+pos], xd2 ;Store result
XSTR [dest3+pos], xd3 ;Store result
XSTR [dest4+pos], xd4 ;Store result
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
mov tmp, len ;Overlapped offset length-16
XLDR x0, [src+tmp] ;Get next source vector
vmovdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
XLDR xtmph4, [dest3+tmp] ;Get next dest vector
sub len, pos
vmovdqa xtmpl4, [constip16] ;Load const of i + 16
vpinsrb xtmph3, xtmph3, len.w, 15
vpshufb xtmph3, xtmph3, xmask0f ;Broadcast len to all bytes
vpcmpgtb xtmph3, xtmph3, xtmpl4
XLDR xtmpl4, [dest4+tmp] ;Get next dest vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xtmpl1, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpand xtmph1, xtmph1, xtmph3
vpxor xd1, xd1, xtmph1
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xtmpl2, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpand xtmph2, xtmph2, xtmph3
vpxor xd2, xd2, xtmph2
; dest3
vpshufb xgft3_hi, xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xtmpl3, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_hi, xtmpl3 ;GF add high and low partials
vpand xgft3_hi, xgft3_hi, xtmph3
vpxor xtmph4, xtmph4, xgft3_hi
; dest4
vpshufb xgft4_hi, xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_hi, xgft4_lo ;GF add high and low partials
vpand xgft4_hi, xgft4_hi, xtmph3
vpxor xtmpl4, xtmpl4, xgft4_hi
XSTR [dest1+tmp], xd1 ;Store result
XSTR [dest2+tmp], xd2 ;Store result
XSTR [dest3+tmp], xtmph4 ;Store result
XSTR [dest4+tmp], xtmpl4 ;Store result
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_4VECT_MAD_AVX, 02, 00, 020a

371
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_4vect_mad_avx2.asm generated vendored
View File

@ -1,371 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_4vect_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_4VECT_MAD_AVX2 _gf_4vect_mad_avx2
%else
%define GF_4VECT_MAD_AVX2 gf_4vect_mad_avx2
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_4vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 vec
%define dest4 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm15
%define xmask0fx xmm15
%define xgft1_lo ymm14
%define xgft2_lo ymm13
%define xgft3_lo ymm12
%define xgft4_lo ymm11
%define x0 ymm0
%define xtmpa ymm1
%define xtmpl ymm2
%define xtmplx xmm2
%define xtmph1 ymm3
%define xtmph1x xmm3
%define xtmph2 ymm4
%define xtmph3 ymm5
%define xtmph4 ymm6
%define xd1 ymm7
%define xd2 ymm8
%define xd3 ymm9
%define xd4 ymm10
align 16
global GF_4VECT_MAD_AVX2:function
func(GF_4VECT_MAD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec_i, 5 ;Multiply by 32
sal vec, 5 ;Multiply by 32
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xgft3_lo, [tmp+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
add tmp, vec
vmovdqu xgft4_lo, [tmp+2*vec] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
mov dest2, [dest1+PS] ; reuse mul_array
mov dest3, [dest1+2*PS] ; reuse vec
mov dest4, [dest1+3*PS] ; reuse vec_i
mov dest1, [dest1]
.loop32:
XLDR x0, [src+pos] ;Get next source vector
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
XLDR xd3, [dest3+pos] ;Get next dest vector
XLDR xd4, [dest4+pos] ;reuse xtmpl1. Get next dest vector
vpand xtmpl, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xtmpl, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xtmpl, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vperm2i128 xtmph1, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph3, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph4, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl ;GF add high and low partials
vpxor xd2, xd2, xtmph2 ;xd2 += partial
; dest3
vpshufb xtmph3, xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xtmpl ;GF add high and low partials
vpxor xd3, xd3, xtmph3 ;xd3 += partial
; dest4
vpshufb xtmph4, xtmph4, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph4, xtmph4, xtmpl ;GF add high and low partials
vpxor xd4, xd4, xtmph4 ;xd4 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
XSTR [dest3+pos], xd3
XSTR [dest4+pos], xd4
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
.lessthan32:
;; Tail len
;; Do one more overlap pass
mov tmp.b, 0x1f
vpinsrb xtmph1x, xtmph1x, tmp.w, 0
vpbroadcastb xtmph1, xtmph1x ;Construct mask 0x1f1f1f...
mov tmp, len ;Overlapped offset length-32
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
XLDR xd3, [dest3+tmp] ;Get next dest vector
XLDR xd4, [dest4+tmp] ;Get next dest vector
sub len, pos
vmovdqa xtmph2, [constip32] ;Load const of i + 32
vpinsrb xtmplx, xtmplx, len.w, 15
vinserti128 xtmpl, xtmpl, xtmplx, 1 ;swapped to xtmplx | xtmplx
vpshufb xtmpl, xtmpl, xtmph1 ;Broadcast len to all bytes. xtmph1=0x1f1f1f...
vpcmpgtb xtmpl, xtmpl, xtmph2
vpand xtmph1, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xtmph1, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xtmph1, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vperm2i128 xtmph1, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph3, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph4, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xgft1_lo ;GF add high and low partials
vpand xtmph1, xtmph1, xtmpl
vpxor xd1, xd1, xtmph1 ;xd1 += partial
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xgft2_lo ;GF add high and low partials
vpand xtmph2, xtmph2, xtmpl
vpxor xd2, xd2, xtmph2 ;xd2 += partial
; dest3
vpshufb xtmph3, xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xgft3_lo ;GF add high and low partials
vpand xtmph3, xtmph3, xtmpl
vpxor xd3, xd3, xtmph3 ;xd3 += partial
; dest4
vpshufb xtmph4, xtmph4, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph4, xtmph4, xgft4_lo ;GF add high and low partials
vpand xtmph4, xtmph4, xtmpl
vpxor xd4, xd4, xtmph4 ;xd4 += partial
XSTR [dest1+tmp], xd1
XSTR [dest2+tmp], xd2
XSTR [dest3+tmp], xd3
XSTR [dest4+tmp], xd4
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 32
constip32:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
ddq 0xe0e1e2e3e4e5e6e7e8e9eaebecedeeef
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_4VECT_MAD_AVX2, 04, 00, 020b

375
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_4vect_mad_sse.asm generated vendored
View File

@ -1,375 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_4vect_mad_sse(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_4VECT_MAD_SSE _gf_4vect_mad_sse
%else
%define GF_4VECT_MAD_SSE gf_4vect_mad_sse
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define tmp3 r13
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r15, 10*16 + 2*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r15, [rsp + 10*16 + 2*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
;;; gf_4vect_mad_sse(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 tmp2
%define dest4 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft3_hi xmm14
%define xgft4_hi xmm13
%define xgft4_lo xmm12
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xtmph4 xmm8
%define xtmpl4 xmm9
%define xd1 xmm10
%define xd2 xmm11
%define xd3 xtmph1
%define xd4 xtmpl1
align 16
global GF_4VECT_MAD_SSE:function
func(GF_4VECT_MAD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov tmp, vec
sal vec_i, 5 ;Multiply by 32
lea tmp3, [mul_array + vec_i]
sal tmp, 6 ;Multiply by 64
movdqu xgft3_hi, [tmp3+tmp+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
sal vec, 5 ;Multiply by 32
add tmp, vec
movdqu xgft4_lo, [tmp3+tmp] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
movdqu xgft4_hi, [tmp3+tmp+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
mov dest2, [dest1+PS] ; reuse mul_array
mov dest3, [dest1+2*PS]
mov dest4, [dest1+3*PS] ; reuse vec_i
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
movdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
movdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqa xtmph3, xgft3_hi
movdqa xtmpl4, xgft4_lo
movdqa xtmph4, xgft4_hi
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
; dest1
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pxor xd1, xtmph1
XLDR xd3, [dest3+pos] ;Reuse xtmph1, Get next dest vector
XLDR xd4, [dest4+pos] ;Reuse xtmpl1, Get next dest vector
; dest2
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pxor xd2, xtmph2
; dest3
pshufb xtmph3, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xtmph3, xtmpl3 ;GF add high and low partials
pxor xd3, xtmph3
; dest4
pshufb xtmph4, x0 ;Lookup mul table of high nibble
pshufb xtmpl4, xtmpa ;Lookup mul table of low nibble
pxor xtmph4, xtmpl4 ;GF add high and low partials
pxor xd4, xtmph4
XSTR [dest1+pos], xd1 ;Store result
XSTR [dest2+pos], xd2 ;Store result
XSTR [dest3+pos], xd3 ;Store result
XSTR [dest4+pos], xd4 ;Store result
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
mov tmp, len ;Overlapped offset length-16
XLDR x0, [src+tmp] ;Get next source vector
movdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
movdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
XLDR xtmph4, [dest3+tmp] ;Reuse xtmph1. Get next dest vector
sub len, pos
movdqa xtmpl4, [constip16] ;Load const of i + 16
pinsrb xtmph3, len.w, 15
pshufb xtmph3, xmask0f ;Broadcast len to all bytes
pcmpgtb xtmph3, xtmpl4
XLDR xtmpl4, [dest4+tmp] ;Get next dest vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
; dest1
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pand xtmph1, xtmph3
pxor xd1, xtmph1
; dest2
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pand xtmph2, xtmph3
pxor xd2, xtmph2
; dest3
pshufb xgft3_hi, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xgft3_hi, xtmpl3 ;GF add high and low partials
pand xgft3_hi, xtmph3
pxor xtmph4, xgft3_hi
; dest4
pshufb xgft4_hi, x0 ;Lookup mul table of high nibble
pshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft4_hi, xgft4_lo ;GF add high and low partials
pand xgft4_hi, xtmph3
pxor xtmpl4, xgft4_hi
XSTR [dest1+tmp], xd1 ;Store result
XSTR [dest2+tmp], xd2 ;Store result
XSTR [dest3+tmp], xtmph4 ;Store result
XSTR [dest4+tmp], xtmpl4 ;Store result
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f:
ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_4VECT_MAD_SSE, 00, 00, 0209

348
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_5vect_dot_prod_avx.asm generated vendored
View File

@ -1,348 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_5vect_dot_prod_avx(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_5VECT_DOT_PROD_AVX _gf_5vect_dot_prod_avx
%else
%define GF_5VECT_DOT_PROD_AVX gf_5vect_dot_prod_avx
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 10*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_xmm128 xmm12, 6*16
save_xmm128 xmm13, 7*16
save_xmm128 xmm14, 8*16
save_xmm128 xmm15, 9*16
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
save_reg rsi, 10*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
vmovdqa xmm15, [rsp + 9*16]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
mov rsi, [rsp + 10*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define ptr arg5
%define vec_i tmp2
%define dest1 tmp3
%define dest2 tmp4
%define vskip1 tmp5
%define vskip3 tmp6
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft1_lo xmm14
%define xgft1_hi xmm13
%define xgft2_lo xmm12
%define xgft2_hi xmm11
%define xgft3_lo xmm10
%define xgft3_hi xmm9
%define xgft4_lo xmm8
%define xgft4_hi xmm7
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%define xp4 xmm5
%define xp5 xmm6
align 16
global GF_5VECT_DOT_PROD_AVX:function
func(GF_5VECT_DOT_PROD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov vskip1, vec
imul vskip1, 32
mov vskip3, vec
imul vskip3, 96
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
mov dest1, [dest]
mov dest2, [dest+PS]
.loop16:
mov tmp, mul_array
xor vec_i, vec_i
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
vpxor xp4, xp4
vpxor xp5, xp5
.next_vect:
mov ptr, [src+vec_i]
add vec_i, PS
XLDR x0, [ptr+pos] ;Get next source vector
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vskip1*1] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vskip1*1+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+vskip1*2] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
vmovdqu xgft3_hi, [tmp+vskip1*2+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
vmovdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
vmovdqu xgft4_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
vmovdqu xgft1_lo, [tmp+vskip1*4] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
vmovdqu xgft1_hi, [tmp+vskip1*4+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
add tmp, 32
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
vpshufb xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_lo ;GF add high and low partials
vpxor xp4, xgft4_hi ;xp4 += partial
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp5, xgft1_hi ;xp5 += partial
cmp vec_i, vec
jl .next_vect
mov tmp, [dest+2*PS]
mov ptr, [dest+3*PS]
mov vec_i, [dest+4*PS]
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
XSTR [tmp+pos], xp3
XSTR [ptr+pos], xp4
XSTR [vec_i+pos], xp5
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_5VECT_DOT_PROD_AVX, 02, 03, 0194

362
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_5vect_dot_prod_avx2.asm generated vendored
View File

@ -1,362 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_5vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_5VECT_DOT_PROD_AVX2 _gf_5vect_dot_prod_avx2
%else
%define GF_5VECT_DOT_PROD_AVX2 gf_5vect_dot_prod_avx2
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 10*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
vmovdqa [rsp + 3*16], xmm9
vmovdqa [rsp + 4*16], xmm10
vmovdqa [rsp + 5*16], xmm11
vmovdqa [rsp + 6*16], xmm12
vmovdqa [rsp + 7*16], xmm13
vmovdqa [rsp + 8*16], xmm14
vmovdqa [rsp + 9*16], xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
save_reg rsi, 10*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
vmovdqa xmm15, [rsp + 9*16]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
mov rsi, [rsp + 10*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define ptr arg5
%define vec_i tmp2
%define dest1 tmp3
%define dest2 tmp4
%define vskip1 tmp5
%define vskip3 tmp6
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm15
%define xmask0fx xmm15
%define xgft1_lo ymm14
%define xgft1_hi ymm13
%define xgft2_lo ymm12
%define xgft2_hi ymm11
%define xgft3_lo ymm10
%define xgft3_hi ymm9
%define xgft4_lo ymm8
%define xgft4_hi ymm7
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%define xp3 ymm4
%define xp4 ymm5
%define xp5 ymm6
align 16
global GF_5VECT_DOT_PROD_AVX2:function
func(GF_5VECT_DOT_PROD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
mov vskip1, vec
imul vskip1, 32
mov vskip3, vec
imul vskip3, 96
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
mov dest1, [dest]
mov dest2, [dest+PS]
.loop32:
mov tmp, mul_array
xor vec_i, vec_i
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
vpxor xp4, xp4
vpxor xp5, xp5
.next_vect:
mov ptr, [src+vec_i]
XLDR x0, [ptr+pos] ;Get next source vector
add vec_i, PS
vpand xgft4_lo, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xgft4_lo, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xgft4_lo, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vskip1*1] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+vskip1*2] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vmovdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
; " Dx{00}, Dx{10}, ..., Dx{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft4_hi, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
vmovdqu xgft1_lo, [tmp+vskip1*4] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
; " Ex{00}, Ex{10}, ..., Ex{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
add tmp, 32
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
vpshufb xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_lo ;GF add high and low partials
vpxor xp4, xgft4_hi ;xp4 += partial
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp5, xgft1_hi ;xp5 += partial
cmp vec_i, vec
jl .next_vect
mov tmp, [dest+2*PS]
mov ptr, [dest+3*PS]
mov vec_i, [dest+4*PS]
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
XSTR [tmp+pos], xp3
XSTR [ptr+pos], xp4
XSTR [vec_i+pos], xp5
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop32 ;Do one more overlap pass
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_5VECT_DOT_PROD_AVX2, 04, 03, 0199

349
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_5vect_dot_prod_sse.asm generated vendored
View File

@ -1,349 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_5vect_dot_prod_sse(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_5VECT_DOT_PROD_SSE _gf_5vect_dot_prod_sse
%else
%define GF_5VECT_DOT_PROD_SSE gf_5vect_dot_prod_sse
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 10*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_xmm128 xmm12, 6*16
save_xmm128 xmm13, 7*16
save_xmm128 xmm14, 8*16
save_xmm128 xmm15, 9*16
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
save_reg rsi, 10*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp + 0*16]
movdqa xmm7, [rsp + 1*16]
movdqa xmm8, [rsp + 2*16]
movdqa xmm9, [rsp + 3*16]
movdqa xmm10, [rsp + 4*16]
movdqa xmm11, [rsp + 5*16]
movdqa xmm12, [rsp + 6*16]
movdqa xmm13, [rsp + 7*16]
movdqa xmm14, [rsp + 8*16]
movdqa xmm15, [rsp + 9*16]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
mov rsi, [rsp + 10*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define ptr arg5
%define vec_i tmp2
%define dest1 tmp3
%define dest2 tmp4
%define vskip1 tmp5
%define vskip3 tmp6
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft1_lo xmm2
%define xgft1_hi xmm3
%define xgft2_lo xmm4
%define xgft2_hi xmm5
%define xgft3_lo xmm10
%define xgft3_hi xmm6
%define xgft4_lo xmm8
%define xgft4_hi xmm7
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm9
%define xp2 xmm11
%define xp3 xmm12
%define xp4 xmm13
%define xp5 xmm14
align 16
global GF_5VECT_DOT_PROD_SSE:function
func(GF_5VECT_DOT_PROD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov vskip1, vec
imul vskip1, 32
mov vskip3, vec
imul vskip3, 96
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
mov dest1, [dest]
mov dest2, [dest+PS]
.loop16:
mov tmp, mul_array
xor vec_i, vec_i
pxor xp1, xp1
pxor xp2, xp2
pxor xp3, xp3
pxor xp4, xp4
pxor xp5, xp5
.next_vect:
mov ptr, [src+vec_i]
add vec_i, PS
XLDR x0, [ptr+pos] ;Get next source vector
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
movdqu xgft2_lo, [tmp+vskip1*1] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vskip1*1+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
movdqu xgft3_lo, [tmp+vskip1*2] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
movdqu xgft3_hi, [tmp+vskip1*2+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
movdqu xgft4_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
movdqu xgft4_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp1, xgft1_hi ;xp1 += partial
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pxor xp2, xgft2_hi ;xp2 += partial
movdqu xgft1_lo, [tmp+vskip1*4] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
movdqu xgft1_hi, [tmp+vskip1*4+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
add tmp, 32
pshufb xgft3_hi, x0 ;Lookup mul table of high nibble
pshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft3_hi, xgft3_lo ;GF add high and low partials
pxor xp3, xgft3_hi ;xp3 += partial
pshufb xgft4_hi, x0 ;Lookup mul table of high nibble
pshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft4_hi, xgft4_lo ;GF add high and low partials
pxor xp4, xgft4_hi ;xp4 += partial
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp5, xgft1_hi ;xp5 += partial
cmp vec_i, vec
jl .next_vect
mov tmp, [dest+2*PS]
mov ptr, [dest+3*PS]
mov vec_i, [dest+4*PS]
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
XSTR [tmp+pos], xp3
XSTR [ptr+pos], xp4
XSTR [vec_i+pos], xp5
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_5VECT_DOT_PROD_SSE, 00, 04, 0065

401
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_5vect_mad_avx.asm generated vendored
View File

@ -1,401 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_5vect_mad_avx(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_5VECT_MAD_AVX _gf_5vect_mad_avx
%else
%define GF_5VECT_MAD_AVX gf_5vect_mad_avx
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define tmp3 r13
%define tmp4 r14
%define return rax
%define return.w eax
%define stack_size 16*10 + 5*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
;;; gf_5vect_mad_avx(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp4
%define dest3 mul_array
%define dest4 tmp2
%define dest5 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft5_hi xmm14
%define xgft4_lo xmm13
%define xgft4_hi xmm12
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xtmph5 xmm8
%define xtmpl5 xmm9
%define xd1 xmm10
%define xd2 xmm11
%define xd3 xtmpl1
%define xd4 xtmph1
%define xd5 xtmpl2
align 16
global GF_5VECT_MAD_AVX:function
func(GF_5VECT_MAD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov tmp, vec
sal vec_i, 5 ;Multiply by 32
lea tmp3, [mul_array + vec_i]
sal tmp, 6 ;Multiply by 64
vmovdqu xgft5_hi, [tmp3+2*tmp+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
sal vec, 5 ;Multiply by 32
add tmp, vec
vmovdqu xgft4_hi, [tmp3+tmp+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
vmovdqu xgft4_lo, [tmp3+tmp] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
mov dest3, [dest1+2*PS] ; reuse mul_array
mov dest4, [dest1+3*PS]
mov dest5, [dest1+4*PS] ; reuse vec_i
mov dest2, [dest1+PS]
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
vmovdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
vmovdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
vmovdqu xtmpl5, [tmp3+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xtmpl1, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1
XLDR xd3, [dest3+pos] ;Reuse xtmpl1, Get next dest vector
XLDR xd4, [dest4+pos] ;Reuse xtmph1, Get next dest vector
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xtmpl2, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2
XLDR xd5, [dest5+pos] ;Reuse xtmpl2. Get next dest vector
; dest3
vpshufb xtmph3, xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xtmpl3, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xtmpl3 ;GF add high and low partials
vpxor xd3, xd3, xtmph3
; dest4
vpshufb xtmph2, xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl3 ;GF add high and low partials
vpxor xd4, xd4, xtmph2
; dest5
vpshufb xtmph5, xgft5_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl5, xtmpl5, xtmpa ;Lookup mul table of low nibble
vpxor xtmph5, xtmph5, xtmpl5 ;GF add high and low partials
vpxor xd5, xd5, xtmph5
XSTR [dest1+pos], xd1 ;Store result into dest1
XSTR [dest2+pos], xd2 ;Store result into dest2
XSTR [dest3+pos], xd3 ;Store result into dest3
XSTR [dest4+pos], xd4 ;Store result into dest4
XSTR [dest5+pos], xd5 ;Store result into dest5
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
mov tmp, len ;Overlapped offset length-16
XLDR x0, [src+tmp] ;Get next source vector
sub len, pos
vmovdqa xtmph1, [constip16] ;Load const of i + 16
vpinsrb xtmph5, len.w, 15
vpshufb xtmph5, xmask0f ;Broadcast len to all bytes
vpcmpgtb xtmph5, xtmph5, xtmph1
vmovdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
vmovdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
vmovdqu xtmpl5, [tmp3+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xtmpl1, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpand xtmph1, xtmph1, xtmph5
vpxor xd1, xd1, xtmph1
XLDR xd3, [dest3+tmp] ;Reuse xtmpl1, Get next dest vector
XLDR xd4, [dest4+tmp] ;Reuse xtmph1, Get next dest vector
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xtmpl2, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpand xtmph2, xtmph2, xtmph5
vpxor xd2, xd2, xtmph2
XLDR xd5, [dest5+tmp] ;Reuse xtmpl2. Get next dest vector
; dest3
vpshufb xtmph3, xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xtmpl3, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xtmpl3 ;GF add high and low partials
vpand xtmph3, xtmph3, xtmph5
vpxor xd3, xd3, xtmph3
; dest4
vpshufb xgft4_hi, xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_hi, xgft4_lo ;GF add high and low partials
vpand xgft4_hi, xgft4_hi, xtmph5
vpxor xd4, xd4, xgft4_hi
; dest5
vpshufb xgft5_hi, xgft5_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl5, xtmpl5, xtmpa ;Lookup mul table of low nibble
vpxor xgft5_hi, xgft5_hi, xtmpl5 ;GF add high and low partials
vpand xgft5_hi, xgft5_hi, xtmph5
vpxor xd5, xd5, xgft5_hi
XSTR [dest1+tmp], xd1 ;Store result into dest1
XSTR [dest2+tmp], xd2 ;Store result into dest2
XSTR [dest3+tmp], xd3 ;Store result into dest3
XSTR [dest4+tmp], xd4 ;Store result into dest4
XSTR [dest5+tmp], xd5 ;Store result into dest5
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_5VECT_MAD_AVX, 02, 00, 020d

393
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_5vect_mad_avx2.asm generated vendored
View File

@ -1,393 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_5vect_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_5VECT_MAD_AVX2 _gf_5vect_mad_avx2
%else
%define GF_5VECT_MAD_AVX2 gf_5vect_mad_avx2
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_5vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp2
%define dest3 mul_array
%define dest4 vec
%define dest5 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm15
%define xmask0fx xmm15
%define xgft1_lo ymm14
%define xgft2_lo ymm13
%define xgft3_lo ymm12
%define xgft4_lo ymm11
%define xgft5_lo ymm10
%define x0 ymm0
%define xtmpa ymm1
%define xtmpl ymm2
%define xtmplx xmm2
%define xtmph1 ymm3
%define xtmph1x xmm3
%define xtmph2 ymm4
%define xd1 ymm5
%define xd2 ymm6
%define xd3 ymm7
%define xd4 ymm8
%define xd5 ymm9
align 16
global GF_5VECT_MAD_AVX2:function
func(GF_5VECT_MAD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec_i, 5 ;Multiply by 32
sal vec, 5 ;Multiply by 32
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+2*vec] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vmovdqu xgft5_lo, [tmp+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
; " Ex{00}, Ex{10}, ..., Ex{f0}
add tmp, vec
vmovdqu xgft4_lo, [tmp+2*vec] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
; " Dx{00}, Dx{10}, ..., Dx{f0}
mov dest3, [dest1+2*PS] ; reuse mul_array
mov dest4, [dest1+3*PS] ; reuse vec
mov dest5, [dest1+4*PS] ; reuse vec_i
mov dest2, [dest1+PS]
mov dest1, [dest1]
.loop32:
XLDR x0, [src+pos] ;Get next source vector
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
XLDR xd3, [dest3+pos] ;Get next dest vector
XLDR xd4, [dest4+pos] ;Get next dest vector
XLDR xd5, [dest5+pos] ;Get next dest vector
vpand xtmpl, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xtmpl, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xtmpl, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vperm2i128 xtmph1, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
vperm2i128 xtmph1, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl ;GF add high and low partials
vpxor xd2, xd2, xtmph2 ;xd2 += partial
vperm2i128 xtmph2, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
; dest3
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl ;GF add high and low partials
vpxor xd3, xd3, xtmph1 ;xd3 += partial
vperm2i128 xtmph1, xgft5_lo, xgft5_lo, 0x01 ; swapped to hi | lo
; dest4
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl ;GF add high and low partials
vpxor xd4, xd4, xtmph2 ;xd4 += partial
; dest5
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft5_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl ;GF add high and low partials
vpxor xd5, xd5, xtmph1 ;xd5 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
XSTR [dest3+pos], xd3
XSTR [dest4+pos], xd4
XSTR [dest5+pos], xd5
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
.lessthan32:
;; Tail len
;; Do one more overlap pass
mov tmp.b, 0x1f
vpinsrb xtmph1x, xtmph1x, tmp.w, 0
vpbroadcastb xtmph1, xtmph1x ;Construct mask 0x1f1f1f...
mov tmp, len ;Overlapped offset length-32
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
XLDR xd3, [dest3+tmp] ;Get next dest vector
XLDR xd4, [dest4+tmp] ;Get next dest vector
XLDR xd5, [dest5+tmp] ;Get next dest vector
sub len, pos
vmovdqa xtmph2, [constip32] ;Load const of i + 32
vpinsrb xtmplx, xtmplx, len.w, 15
vinserti128 xtmpl, xtmpl, xtmplx, 1 ;swapped to xtmplx | xtmplx
vpshufb xtmpl, xtmpl, xtmph1 ;Broadcast len to all bytes. xtmph1=0x1f1f1f...
vpcmpgtb xtmpl, xtmpl, xtmph2
vpand xtmph1, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xtmph1, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xtmph1, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vperm2i128 xtmph1, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
; dest1
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xgft1_lo ;GF add high and low partials
vpand xtmph1, xtmph1, xtmpl
vpxor xd1, xd1, xtmph1 ;xd1 += partial
vperm2i128 xtmph1, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xgft2_lo ;GF add high and low partials
vpand xtmph2, xtmph2, xtmpl
vpxor xd2, xd2, xtmph2 ;xd2 += partial
vperm2i128 xtmph2, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
; dest3
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xgft3_lo ;GF add high and low partials
vpand xtmph1, xtmph1, xtmpl
vpxor xd3, xd3, xtmph1 ;xd3 += partial
vperm2i128 xtmph1, xgft5_lo, xgft5_lo, 0x01 ; swapped to hi | lo
; dest4
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xgft4_lo ;GF add high and low partials
vpand xtmph2, xtmph2, xtmpl
vpxor xd4, xd4, xtmph2 ;xd4 += partial
; dest5
vpshufb xtmph1, xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xgft5_lo, xgft5_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xgft5_lo ;GF add high and low partials
vpand xtmph1, xtmph1, xtmpl
vpxor xd5, xd5, xtmph1 ;xd5 += partial
XSTR [dest1+tmp], xd1
XSTR [dest2+tmp], xd2
XSTR [dest3+tmp], xd3
XSTR [dest4+tmp], xd4
XSTR [dest5+tmp], xd5
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 32
constip32:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
ddq 0xe0e1e2e3e4e5e6e7e8e9eaebecedeeef
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_5VECT_MAD_AVX2, 04, 00, 020e

409
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_5vect_mad_sse.asm generated vendored
View File

@ -1,409 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_5vect_mad_sse(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_5VECT_MAD_SSE _gf_5vect_mad_sse
%else
%define GF_5VECT_MAD_SSE gf_5vect_mad_sse
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define tmp3 r13
%define tmp4 r14
%define return rax
%define return.w eax
%define stack_size 16*10 + 5*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
;;; gf_5vect_mad_sse(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp4
%define dest3 mul_array
%define dest4 tmp2
%define dest5 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft5_hi xmm14
%define xgft4_lo xmm13
%define xgft4_hi xmm12
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xtmph5 xmm8
%define xtmpl5 xmm9
%define xd1 xmm10
%define xd2 xmm11
%define xd3 xtmpl1
%define xd4 xtmph1
%define xd5 xtmpl2
align 16
global GF_5VECT_MAD_SSE:function
func(GF_5VECT_MAD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov tmp, vec
sal vec_i, 5 ;Multiply by 32
lea tmp3, [mul_array + vec_i]
sal tmp, 6 ;Multiply by 64
movdqu xgft5_hi, [tmp3+2*tmp+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
sal vec, 5 ;Multiply by 32
add tmp, vec
movdqu xgft4_hi, [tmp3+tmp+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
movdqu xgft4_lo, [tmp3+tmp] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
mov dest3, [dest1+2*PS] ; reuse mul_array
mov dest4, [dest1+3*PS]
mov dest5, [dest1+4*PS] ; reuse vec_i
mov dest2, [dest1+PS]
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
movdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
movdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
movdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqu xtmpl5, [tmp3+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
movdqa xtmph5, xgft5_hi ;Reload const array registers
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
; dest1
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pxor xd1, xtmph1
XLDR xd3, [dest3+pos] ;Reuse xtmpl1, Get next dest vector
XLDR xd4, [dest4+pos] ;Reuse xtmph1. Get next dest vector
; dest2
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pxor xd2, xtmph2
XLDR xd5, [dest5+pos] ;Reuse xtmpl2. Get next dest vector
; dest3
pshufb xtmph3, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xtmph3, xtmpl3 ;GF add high and low partials
pxor xd3, xtmph3
movdqa xtmph2, xgft4_hi ;Reload const array registers
movdqa xtmpl3, xgft4_lo ;Reload const array registers
; dest5
pshufb xtmph5, x0 ;Lookup mul table of high nibble
pshufb xtmpl5, xtmpa ;Lookup mul table of low nibble
pxor xtmph5, xtmpl5 ;GF add high and low partials
pxor xd5, xtmph5
; dest4
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl3 ;GF add high and low partials
pxor xd4, xtmph2
XSTR [dest1+pos], xd1 ;Store result into dest1
XSTR [dest2+pos], xd2 ;Store result into dest2
XSTR [dest3+pos], xd3 ;Store result into dest3
XSTR [dest4+pos], xd4 ;Store result into dest4
XSTR [dest5+pos], xd5 ;Store result into dest5
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
mov tmp, len ;Overlapped offset length-16
XLDR x0, [src+tmp] ;Get next source vector
sub len, pos
movdqa xtmpl1, [constip16] ;Load const of i + 16
pinsrb xtmph5, len.w, 15
pshufb xtmph5, xmask0f ;Broadcast len to all bytes
pcmpgtb xtmph5, xtmpl1
movdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
movdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
movdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqu xtmpl5, [tmp3+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
; dest1
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pand xtmph1, xtmph5
pxor xd1, xtmph1
XLDR xd3, [dest3+tmp] ;Reuse xtmpl1, Get next dest vector
XLDR xd4, [dest4+tmp] ;Reuse xtmph1. Get next dest vector
; dest2
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pand xtmph2, xtmph5
pxor xd2, xtmph2
XLDR xd5, [dest5+tmp] ;Reuse xtmpl2. Get next dest vector
; dest3
pshufb xtmph3, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xtmph3, xtmpl3 ;GF add high and low partials
pand xtmph3, xtmph5
pxor xd3, xtmph3
; dest4
pshufb xgft4_hi, x0 ;Lookup mul table of high nibble
pshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft4_hi, xgft4_lo ;GF add high and low partials
pand xgft4_hi, xtmph5
pxor xd4, xgft4_hi
; dest5
pshufb xgft5_hi, x0 ;Lookup mul table of high nibble
pshufb xtmpl5, xtmpa ;Lookup mul table of low nibble
pxor xgft5_hi, xtmpl5 ;GF add high and low partials
pand xgft5_hi, xtmph5
pxor xd5, xgft5_hi
XSTR [dest1+tmp], xd1 ;Store result into dest1
XSTR [dest2+tmp], xd2 ;Store result into dest2
XSTR [dest3+tmp], xd3 ;Store result into dest3
XSTR [dest4+tmp], xd4 ;Store result into dest4
XSTR [dest5+tmp], xd5 ;Store result into dest5
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f:
ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_5VECT_MAD_SSE, 00, 00, 020c

360
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_6vect_dot_prod_avx.asm generated vendored
View File

@ -1,360 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_6vect_dot_prod_avx(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_6VECT_DOT_PROD_AVX _gf_6vect_dot_prod_avx
%else
%define GF_6VECT_DOT_PROD_AVX gf_6vect_dot_prod_avx
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 10*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_xmm128 xmm12, 6*16
save_xmm128 xmm13, 7*16
save_xmm128 xmm14, 8*16
save_xmm128 xmm15, 9*16
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
save_reg rsi, 10*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
vmovdqa xmm15, [rsp + 9*16]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
mov rsi, [rsp + 10*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define ptr arg5
%define vec_i tmp2
%define dest1 tmp3
%define dest2 tmp4
%define vskip1 tmp5
%define vskip3 tmp6
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft1_lo xmm14
%define xgft1_hi xmm13
%define xgft2_lo xmm12
%define xgft2_hi xmm11
%define xgft3_lo xmm10
%define xgft3_hi xmm9
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%define xp3 xmm4
%define xp4 xmm5
%define xp5 xmm6
%define xp6 xmm7
align 16
global GF_6VECT_DOT_PROD_AVX:function
func(GF_6VECT_DOT_PROD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov vskip1, vec
imul vskip1, 32
mov vskip3, vec
imul vskip3, 96
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
mov dest1, [dest]
mov dest2, [dest+PS]
.loop16:
mov tmp, mul_array
xor vec_i, vec_i
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
vpxor xp4, xp4
vpxor xp5, xp5
vpxor xp6, xp6
.next_vect:
mov ptr, [src+vec_i]
add vec_i, PS
XLDR x0, [ptr+pos] ;Get next source vector
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vskip1*1] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vskip1*1+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+vskip1*2] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
vmovdqu xgft3_hi, [tmp+vskip1*2+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
lea ptr, [vskip1 + vskip1*4] ;ptr = vskip5
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
vmovdqu xgft1_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
vmovdqu xgft1_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
vmovdqu xgft2_lo, [tmp+vskip1*4] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
vmovdqu xgft2_hi, [tmp+vskip1*4+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
vmovdqu xgft3_lo, [tmp+ptr] ;Load array Fx{00}, Fx{01}, ..., Fx{0f}
vmovdqu xgft3_hi, [tmp+ptr+16] ; " Fx{00}, Fx{10}, ..., Fx{f0}
add tmp, 32
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp4, xgft1_hi ;xp4 += partial
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp5, xgft2_hi ;xp5 += partial
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp6, xgft3_hi ;xp6 += partial
cmp vec_i, vec
jl .next_vect
mov tmp, [dest+2*PS]
mov ptr, [dest+3*PS]
mov vec_i, [dest+4*PS]
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
XSTR [tmp+pos], xp3
mov tmp, [dest+5*PS]
XSTR [ptr+pos], xp4
XSTR [vec_i+pos], xp5
XSTR [tmp+pos], xp6
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_6VECT_DOT_PROD_AVX, 02, 03, 0195

373
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_6vect_dot_prod_avx2.asm generated vendored
View File

@ -1,373 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_6vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_6VECT_DOT_PROD_AVX2 _gf_6vect_dot_prod_avx2
%else
%define GF_6VECT_DOT_PROD_AVX2 gf_6vect_dot_prod_avx2
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 10*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
vmovdqa [rsp + 3*16], xmm9
vmovdqa [rsp + 4*16], xmm10
vmovdqa [rsp + 5*16], xmm11
vmovdqa [rsp + 6*16], xmm12
vmovdqa [rsp + 7*16], xmm13
vmovdqa [rsp + 8*16], xmm14
vmovdqa [rsp + 9*16], xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
save_reg rsi, 10*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
vmovdqa xmm15, [rsp + 9*16]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
mov rsi, [rsp + 10*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define ptr arg5
%define vec_i tmp2
%define dest1 tmp3
%define dest2 tmp4
%define vskip1 tmp5
%define vskip3 tmp6
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm15
%define xmask0fx xmm15
%define xgft1_lo ymm14
%define xgft1_hi ymm13
%define xgft2_lo ymm12
%define xgft2_hi ymm11
%define xgft3_lo ymm10
%define xgft3_hi ymm9
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%define xp3 ymm4
%define xp4 ymm5
%define xp5 ymm6
%define xp6 ymm7
align 16
global GF_6VECT_DOT_PROD_AVX2:function
func(GF_6VECT_DOT_PROD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
mov vskip1, vec
imul vskip1, 32
mov vskip3, vec
imul vskip3, 96
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
mov dest1, [dest]
mov dest2, [dest+PS]
.loop32:
mov tmp, mul_array
xor vec_i, vec_i
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
vpxor xp4, xp4
vpxor xp5, xp5
vpxor xp6, xp6
.next_vect:
mov ptr, [src+vec_i]
XLDR x0, [ptr+pos] ;Get next source vector
add vec_i, PS
vpand xgft3_lo, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xgft3_lo, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xgft3_lo, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vskip1*1] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+vskip1*2] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
lea ptr, [vskip1 + vskip1*4] ;ptr = vskip5
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
vmovdqu xgft1_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
; " Dx{00}, Dx{10}, ..., Dx{f0}
vmovdqu xgft2_lo, [tmp+vskip1*4] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
; " Ex{00}, Ex{10}, ..., Ex{f0}
vmovdqu xgft3_lo, [tmp+ptr] ;Load array Fx{00}, Fx{01}, ..., Fx{0f}
; " Fx{00}, Fx{10}, ..., Fx{f0}
add tmp, 32
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp4, xgft1_hi ;xp4 += partial
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp5, xgft2_hi ;xp5 += partial
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp6, xgft3_hi ;xp6 += partial
cmp vec_i, vec
jl .next_vect
mov tmp, [dest+2*PS]
mov ptr, [dest+3*PS]
mov vec_i, [dest+4*PS]
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
XSTR [tmp+pos], xp3
mov tmp, [dest+5*PS]
XSTR [ptr+pos], xp4
XSTR [vec_i+pos], xp5
XSTR [tmp+pos], xp6
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop32 ;Do one more overlap pass
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_6VECT_DOT_PROD_AVX2, 04, 03, 019a

360
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_6vect_dot_prod_sse.asm generated vendored
View File

@ -1,360 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_6vect_dot_prod_sse(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_6VECT_DOT_PROD_SSE _gf_6vect_dot_prod_sse
%else
%define GF_6VECT_DOT_PROD_SSE gf_6vect_dot_prod_sse
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define tmp5 r14 ; must be saved and restored
%define tmp6 r15 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define tmp5 rdi ; must be saved and restored
%define tmp6 rsi ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 10*16 + 7*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_xmm128 xmm9, 3*16
save_xmm128 xmm10, 4*16
save_xmm128 xmm11, 5*16
save_xmm128 xmm12, 6*16
save_xmm128 xmm13, 7*16
save_xmm128 xmm14, 8*16
save_xmm128 xmm15, 9*16
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
save_reg rsi, 10*16 + 5*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp + 0*16]
movdqa xmm7, [rsp + 1*16]
movdqa xmm8, [rsp + 2*16]
movdqa xmm9, [rsp + 3*16]
movdqa xmm10, [rsp + 4*16]
movdqa xmm11, [rsp + 5*16]
movdqa xmm12, [rsp + 6*16]
movdqa xmm13, [rsp + 7*16]
movdqa xmm14, [rsp + 8*16]
movdqa xmm15, [rsp + 9*16]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
mov rsi, [rsp + 10*16 + 5*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define ptr arg5
%define vec_i tmp2
%define dest1 tmp3
%define dest2 tmp4
%define vskip1 tmp5
%define vskip3 tmp6
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft1_lo xmm2
%define xgft1_hi xmm3
%define xgft2_lo xmm4
%define xgft2_hi xmm5
%define xgft3_lo xmm6
%define xgft3_hi xmm7
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm8
%define xp2 xmm9
%define xp3 xmm10
%define xp4 xmm11
%define xp5 xmm12
%define xp6 xmm13
align 16
global GF_6VECT_DOT_PROD_SSE:function
func(GF_6VECT_DOT_PROD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov vskip1, vec
imul vskip1, 32
mov vskip3, vec
imul vskip3, 96
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
mov dest1, [dest]
mov dest2, [dest+PS]
.loop16:
mov tmp, mul_array
xor vec_i, vec_i
pxor xp1, xp1
pxor xp2, xp2
pxor xp3, xp3
pxor xp4, xp4
pxor xp5, xp5
pxor xp6, xp6
.next_vect:
mov ptr, [src+vec_i]
add vec_i, PS
XLDR x0, [ptr+pos] ;Get next source vector
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
movdqu xgft2_lo, [tmp+vskip1*1] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vskip1*1+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
movdqu xgft3_lo, [tmp+vskip1*2] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
movdqu xgft3_hi, [tmp+vskip1*2+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
lea ptr, [vskip1 + vskip1*4] ;ptr = vskip5
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp1, xgft1_hi ;xp1 += partial
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pxor xp2, xgft2_hi ;xp2 += partial
pshufb xgft3_hi, x0 ;Lookup mul table of high nibble
pshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft3_hi, xgft3_lo ;GF add high and low partials
pxor xp3, xgft3_hi ;xp3 += partial
movdqu xgft1_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
movdqu xgft1_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0}
movdqu xgft2_lo, [tmp+vskip1*4] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
movdqu xgft2_hi, [tmp+vskip1*4+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
movdqu xgft3_lo, [tmp+ptr] ;Load array Fx{00}, Fx{01}, ..., Fx{0f}
movdqu xgft3_hi, [tmp+ptr+16] ; " Fx{00}, Fx{10}, ..., Fx{f0}
add tmp, 32
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp4, xgft1_hi ;xp4 += partial
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pxor xp5, xgft2_hi ;xp5 += partial
pshufb xgft3_hi, x0 ;Lookup mul table of high nibble
pshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft3_hi, xgft3_lo ;GF add high and low partials
pxor xp6, xgft3_hi ;xp6 += partial
cmp vec_i, vec
jl .next_vect
mov tmp, [dest+2*PS]
mov ptr, [dest+3*PS]
mov vec_i, [dest+4*PS]
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
XSTR [tmp+pos], xp3
mov tmp, [dest+5*PS]
XSTR [ptr+pos], xp4
XSTR [vec_i+pos], xp5
XSTR [tmp+pos], xp6
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_6VECT_DOT_PROD_SSE, 00, 04, 0066

433
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_6vect_mad_avx.asm generated vendored
View File

@ -1,433 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_6vect_mad_avx(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_6VECT_MAD_AVX _gf_6vect_mad_avx
%else
%define GF_6VECT_MAD_AVX gf_6vect_mad_avx
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define tmp3 r13
%define tmp4 r14
%define tmp5 rdi
%define return rax
%define return.w eax
%define stack_size 16*10 + 5*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define tmp5 r14
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
%endmacro
%macro FUNC_RESTORE 0
pop r14
pop r13
pop r12
%endmacro
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define tmp5 r14
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
%endmacro
%macro FUNC_RESTORE 0
pop r14
pop r13
pop r12
%endmacro
%endif
;;; gf_6vect_mad_avx(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp4
%define dest3 tmp2
%define dest4 mul_array
%define dest5 tmp5
%define dest6 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft4_lo xmm14
%define xgft4_hi xmm13
%define xgft5_lo xmm12
%define xgft5_hi xmm11
%define xgft6_lo xmm10
%define xgft6_hi xmm9
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xd1 xmm8
%define xd2 xtmpl1
%define xd3 xtmph1
align 16
global GF_6VECT_MAD_AVX:function
func(GF_6VECT_MAD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov tmp, vec
sal vec_i, 5 ;Multiply by 32
lea tmp3, [mul_array + vec_i]
sal tmp, 6 ;Multiply by 64
sal vec, 5 ;Multiply by 32
lea vec_i, [tmp + vec] ;vec_i = vec*96
lea mul_array, [tmp + vec_i] ;mul_array = vec*160
vmovdqu xgft5_lo, [tmp3+2*tmp] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
vmovdqu xgft5_hi, [tmp3+2*tmp+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
vmovdqu xgft4_lo, [tmp3+vec_i] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
vmovdqu xgft4_hi, [tmp3+vec_i+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
vmovdqu xgft6_lo, [tmp3+mul_array] ;Load array Fx{00}, Fx{01}, ..., Fx{0f}
vmovdqu xgft6_hi, [tmp3+mul_array+16] ; " Fx{00}, Fx{10}, ..., Fx{f0}
mov dest2, [dest1+PS]
mov dest3, [dest1+2*PS]
mov dest4, [dest1+3*PS] ; reuse mul_array
mov dest5, [dest1+4*PS]
mov dest6, [dest1+5*PS] ; reuse vec_i
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
vmovdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
vmovdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
XLDR xd1, [dest1+pos] ;Get next dest vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
;dest1
vpshufb xtmph1, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xtmph1
XLDR xd2, [dest2+pos] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+pos] ;reuse xtmph1. Get next dest vector
;dest2
vpshufb xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xtmph2
;dest3
vpshufb xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmpl3 ;GF add high and low partials
vpxor xd3, xtmph3
XSTR [dest1+pos], xd1 ;Store result into dest1
XSTR [dest2+pos], xd2 ;Store result into dest2
XSTR [dest3+pos], xd3 ;Store result into dest3
;dest4
XLDR xd1, [dest4+pos] ;Get next dest vector
vpshufb xtmph1, xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1
XLDR xd2, [dest5+pos] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest6+pos] ;reuse xtmph1. Get next dest vector
;dest5
vpshufb xtmph2, xgft5_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xgft5_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2
;dest6
vpshufb xtmph3, xgft6_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xgft6_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xtmpl3 ;GF add high and low partials
vpxor xd3, xd3, xtmph3
XSTR [dest4+pos], xd1 ;Store result into dest4
XSTR [dest5+pos], xd2 ;Store result into dest5
XSTR [dest6+pos], xd3 ;Store result into dest6
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
;; Overlapped offset length-16
mov tmp, len ;Backup len as len=rdi
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest4+tmp] ;Get next dest vector
XLDR xd2, [dest5+tmp] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest6+tmp] ;reuse xtmph1. Get next dest vector
sub len, pos
vmovdqa xtmph3, [constip16] ;Load const of i + 16
vpinsrb xtmpl3, len.w, 15
vpshufb xtmpl3, xmask0f ;Broadcast len to all bytes
vpcmpgtb xtmpl3, xtmpl3, xtmph3
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
;dest4
vpshufb xgft4_hi, xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_hi, xgft4_lo ;GF add high and low partials
vpand xgft4_hi, xgft4_hi, xtmpl3
vpxor xd1, xd1, xgft4_hi
;dest5
vpshufb xgft5_hi, xgft5_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft5_lo, xgft5_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft5_hi, xgft5_hi, xgft5_lo ;GF add high and low partials
vpand xgft5_hi, xgft5_hi, xtmpl3
vpxor xd2, xd2, xgft5_hi
;dest6
vpshufb xgft6_hi, xgft6_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft6_lo, xgft6_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft6_hi, xgft6_hi, xgft6_lo ;GF add high and low partials
vpand xgft6_hi, xgft6_hi, xtmpl3
vpxor xd3, xd3, xgft6_hi
XSTR [dest4+tmp], xd1 ;Store result into dest4
XSTR [dest5+tmp], xd2 ;Store result into dest5
XSTR [dest6+tmp], xd3 ;Store result into dest6
vmovdqu xgft4_lo, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xgft4_hi, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xgft5_lo, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xgft5_hi, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xgft6_lo, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
vmovdqu xgft6_hi, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+tmp] ;reuse xtmph1. Get next dest3 vector
;dest1
vpshufb xgft4_hi, xgft4_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft4_hi, xgft4_hi, xgft4_lo ;GF add high and low partials
vpand xgft4_hi, xgft4_hi, xtmpl3
vpxor xd1, xd1, xgft4_hi
;dest2
vpshufb xgft5_hi, xgft5_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft5_lo, xgft5_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft5_hi, xgft5_hi, xgft5_lo ;GF add high and low partials
vpand xgft5_hi, xgft5_hi, xtmpl3
vpxor xd2, xd2, xgft5_hi
;dest3
vpshufb xgft6_hi, xgft6_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft6_lo, xgft6_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft6_hi, xgft6_hi, xgft6_lo ;GF add high and low partials
vpand xgft6_hi, xgft6_hi, xtmpl3
vpxor xd3, xd3, xgft6_hi
XSTR [dest1+tmp], xd1 ;Store result into dest1
XSTR [dest2+tmp], xd2 ;Store result into dest2
XSTR [dest3+tmp], xd3 ;Store result into dest3
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_6VECT_MAD_AVX, 02, 00, 0210

435
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_6vect_mad_avx2.asm generated vendored
View File

@ -1,435 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_6vect_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_6VECT_MAD_AVX2 _gf_6vect_mad_avx2
%else
%define GF_6VECT_MAD_AVX2 gf_6vect_mad_avx2
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13
%define return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r15, 10*16 + 2*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r15, [rsp + 10*16 + 3*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r12
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r12
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
;;; gf_6vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp3
%define dest3 tmp2
%define dest4 mul_array
%define dest5 vec
%define dest6 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm15
%define xmask0fx xmm15
%define xgft1_lo ymm14
%define xgft2_lo ymm13
%define xgft3_lo ymm12
%define xgft4_lo ymm11
%define xgft5_lo ymm10
%define xgft6_lo ymm9
%define x0 ymm0
%define xtmpa ymm1
%define xtmpl ymm2
%define xtmplx xmm2
%define xtmph ymm3
%define xtmphx xmm3
%define xd1 ymm4
%define xd2 ymm5
%define xd3 ymm6
%define xd4 ymm7
%define xd5 ymm8
%define xd6 xd1
align 16
global GF_6VECT_MAD_AVX2:function
func(GF_6VECT_MAD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec_i, 5 ;Multiply by 32
sal vec, 5 ;Multiply by 32
lea tmp, [mul_array + vec_i]
mov vec_i, vec
mov mul_array, vec
sal vec_i, 1
sal mul_array, 1
add vec_i, vec ;vec_i=vec*96
add mul_array, vec_i ;vec_i=vec*160
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vmovdqu xgft3_lo, [tmp+2*vec] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vmovdqu xgft4_lo, [tmp+vec_i] ;Load array Fx{00}, Fx{01}, ..., Fx{0f}
; " Fx{00}, Fx{10}, ..., Fx{f0}
vmovdqu xgft5_lo, [tmp+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
; " Ex{00}, Ex{10}, ..., Ex{f0}
vmovdqu xgft6_lo, [tmp+mul_array] ;Load array Dx{00}, Dx{01}, ..., Dx{0f}
; " Dx{00}, Dx{10}, ..., Dx{f0}
mov dest2, [dest1+PS] ; reuse tmp3
mov dest3, [dest1+2*PS] ; reuse tmp2
mov dest4, [dest1+3*PS] ; reuse mul_array
mov dest5, [dest1+4*PS] ; reuse vec
mov dest6, [dest1+5*PS] ; reuse vec_i
mov dest1, [dest1]
.loop32:
XLDR x0, [src+pos] ;Get next source vector
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
XLDR xd3, [dest3+pos] ;Get next dest vector
XLDR xd4, [dest4+pos] ;Get next dest vector
XLDR xd5, [dest5+pos] ;Get next dest vector
vpand xtmpl, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xtmpl, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xtmpl, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
;dest1
vperm2i128 xtmph, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd1, xd1, xtmph ;xd1 += partial
XSTR [dest1+pos], xd1 ;Store result into dest1
;dest2
vperm2i128 xtmph, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd2, xd2, xtmph ;xd2 += partial
;dest3
vperm2i128 xtmph, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd3, xd3, xtmph ;xd3 += partial
XLDR xd6, [dest6+pos] ;reuse xd1. Get next dest vector
;dest4
vperm2i128 xtmph, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd4, xd4, xtmph ;xd4 += partial
;dest5
vperm2i128 xtmph, xgft5_lo, xgft5_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft5_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd5, xd5, xtmph ;xd5 += partial
;dest6
vperm2i128 xtmph, xgft6_lo, xgft6_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft6_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd6, xd6, xtmph ;xd6 += partial
XSTR [dest2+pos], xd2 ;Store result into dest2
XSTR [dest3+pos], xd3 ;Store result into dest3
XSTR [dest4+pos], xd4 ;Store result into dest4
XSTR [dest5+pos], xd5 ;Store result into dest5
XSTR [dest6+pos], xd6 ;Store result into dest6
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
.lessthan32:
;; Tail len
;; Do one more overlap pass
mov tmp.b, 0x1f
vpinsrb xtmphx, xtmphx, tmp.w, 0
vpbroadcastb xtmph, xtmphx ;Construct mask 0x1f1f1f...
mov tmp, len ;Overlapped offset length-32
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
XLDR xd3, [dest3+tmp] ;Get next dest vector
XLDR xd4, [dest4+tmp] ;Get next dest vector
XLDR xd5, [dest5+tmp] ;Get next dest vector
sub len, pos
vpinsrb xtmplx, xtmplx, len.w, 15
vinserti128 xtmpl, xtmpl, xtmplx, 1 ;swapped to xtmplx | xtmplx
vpshufb xtmpl, xtmpl, xtmph ;Broadcast len to all bytes. xtmph=0x1f1f1f...
vpcmpgtb xtmpl, xtmpl, [constip32]
vpand xtmph, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vperm2i128 xtmpa, xtmph, x0, 0x30 ;swap xtmpa from 1lo|2lo to 1lo|2hi
vperm2i128 x0, xtmph, x0, 0x12 ;swap x0 from 1hi|2hi to 1hi|2lo
;dest1
vperm2i128 xtmph, xgft1_lo, xgft1_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xgft1_lo ;GF add high and low partials
vpand xtmph, xtmph, xtmpl
vpxor xd1, xd1, xtmph ;xd1 += partial
XSTR [dest1+tmp], xd1 ;Store result into dest1
;dest2
vperm2i128 xtmph, xgft2_lo, xgft2_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xgft2_lo ;GF add high and low partials
vpand xtmph, xtmph, xtmpl
vpxor xd2, xd2, xtmph ;xd2 += partial
;dest3
vperm2i128 xtmph, xgft3_lo, xgft3_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xgft3_lo ;GF add high and low partials
vpand xtmph, xtmph, xtmpl
vpxor xd3, xd3, xtmph ;xd3 += partial
XLDR xd6, [dest6+tmp] ;reuse xd1. Get next dest vector
;dest4
vperm2i128 xtmph, xgft4_lo, xgft4_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xgft4_lo ;GF add high and low partials
vpand xtmph, xtmph, xtmpl
vpxor xd4, xd4, xtmph ;xd4 += partial
;dest5
vperm2i128 xtmph, xgft5_lo, xgft5_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xgft5_lo, xgft5_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xgft5_lo ;GF add high and low partials
vpand xtmph, xtmph, xtmpl
vpxor xd5, xd5, xtmph ;xd5 += partial
;dest6
vperm2i128 xtmph, xgft6_lo, xgft6_lo, 0x01 ; swapped to hi | lo
vpshufb xtmph, xtmph, x0 ;Lookup mul table of high nibble
vpshufb xgft6_lo, xgft6_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xgft6_lo ;GF add high and low partials
vpand xtmph, xtmph, xtmpl
vpxor xd6, xd6, xtmph ;xd6 += partial
XSTR [dest2+tmp], xd2 ;Store result into dest2
XSTR [dest3+tmp], xd3 ;Store result into dest3
XSTR [dest4+tmp], xd4 ;Store result into dest4
XSTR [dest5+tmp], xd5 ;Store result into dest5
XSTR [dest6+tmp], xd6 ;Store result into dest6
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 32
constip32:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
ddq 0xe0e1e2e3e4e5e6e7e8e9eaebecedeeef
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_6VECT_MAD_AVX2, 04, 00, 0211

446
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_6vect_mad_sse.asm generated vendored
View File

@ -1,446 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_6vect_mad_sse(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_6VECT_MAD_SSE _gf_6vect_mad_sse
%else
%define GF_6VECT_MAD_SSE gf_6vect_mad_sse
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp.w r11d
%define tmp2 r10
%define tmp3 r13
%define tmp4 r14
%define tmp5 rdi
%define return rax
%define return.w eax
%define stack_size 16*10 + 5*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r13, 10*16 + 1*8
save_reg r14, 10*16 + 2*8
save_reg r15, 10*16 + 3*8
save_reg rdi, 10*16 + 4*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r13, [rsp + 10*16 + 1*8]
mov r14, [rsp + 10*16 + 2*8]
mov r15, [rsp + 10*16 + 3*8]
mov rdi, [rsp + 10*16 + 4*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define tmp5 r14
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
%endmacro
%macro FUNC_RESTORE 0
pop r14
pop r13
pop r12
%endmacro
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp2 r10
%define tmp3 r12
%define tmp4 r13
%define tmp5 r14
%define return rax
%define return.w eax
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
push r14
%endmacro
%macro FUNC_RESTORE 0
pop r14
pop r13
pop r12
%endmacro
%endif
;;; gf_6vect_mad_sse(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 tmp2
%define dest4 tmp4
%define dest5 tmp5
%define dest6 vec_i
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft4_lo xmm14
%define xgft4_hi xmm13
%define xgft5_lo xmm12
%define xgft5_hi xmm11
%define xgft6_lo xmm10
%define xgft6_hi xmm9
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xtmph3 xmm6
%define xtmpl3 xmm7
%define xd1 xmm8
%define xd2 xtmpl1
%define xd3 xtmph1
align 16
global GF_6VECT_MAD_SSE:function
func(GF_6VECT_MAD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
mov tmp, vec
sal vec_i, 5 ;Multiply by 32
lea tmp3, [mul_array + vec_i]
sal tmp, 6 ;Multiply by 64
sal vec, 5 ;Multiply by 32
lea vec_i, [tmp + vec] ;vec_i = 96
lea mul_array, [tmp + vec_i] ;mul_array = 160
movdqu xgft5_lo, [tmp3+2*tmp] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
movdqu xgft5_hi, [tmp3+2*tmp+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
movdqu xgft4_lo, [tmp3+vec_i] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
movdqu xgft4_hi, [tmp3+vec_i+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
movdqu xgft6_lo, [tmp3+mul_array] ;Load array Fx{00}, Fx{01}, ..., Fx{0f}
movdqu xgft6_hi, [tmp3+mul_array+16] ; " Fx{00}, Fx{10}, ..., Fx{f0}
mov dest2, [dest1+PS]
mov dest3, [dest1+2*PS]
mov dest4, [dest1+3*PS] ; reuse mul_array
mov dest5, [dest1+4*PS]
mov dest6, [dest1+5*PS] ; reuse vec_i
mov dest1, [dest1]
.loop16:
XLDR x0, [src+pos] ;Get next source vector
movdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
movdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
XLDR xd1, [dest1+pos] ;Get next dest vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
;dest1
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pxor xd1, xtmph1
XLDR xd2, [dest2+pos] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+pos] ;reuse xtmph1. Get next dest3 vector
;dest2
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pxor xd2, xtmph2
;dest3
pshufb xtmph3, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xtmph3, xtmpl3 ;GF add high and low partials
pxor xd3, xtmph3
XSTR [dest1+pos], xd1 ;Store result into dest1
XSTR [dest2+pos], xd2 ;Store result into dest2
XSTR [dest3+pos], xd3 ;Store result into dest3
movdqa xtmph1, xgft4_hi ;Reload const array registers
movdqa xtmpl1, xgft4_lo ;Reload const array registers
movdqa xtmph2, xgft5_hi ;Reload const array registers
movdqa xtmpl2, xgft5_lo ;Reload const array registers
movdqa xtmph3, xgft6_hi ;Reload const array registers
movdqa xtmpl3, xgft6_lo ;Reload const array registers
;dest4
XLDR xd1, [dest4+pos] ;Get next dest vector
pshufb xtmph1, x0 ;Lookup mul table of high nibble
pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
pxor xtmph1, xtmpl1 ;GF add high and low partials
pxor xd1, xtmph1
XLDR xd2, [dest5+pos] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest6+pos] ;reuse xtmph1. Get next dest vector
;dest5
pshufb xtmph2, x0 ;Lookup mul table of high nibble
pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
pxor xtmph2, xtmpl2 ;GF add high and low partials
pxor xd2, xtmph2
;dest6
pshufb xtmph3, x0 ;Lookup mul table of high nibble
pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
pxor xtmph3, xtmpl3 ;GF add high and low partials
pxor xd3, xtmph3
XSTR [dest4+pos], xd1 ;Store result into dest4
XSTR [dest5+pos], xd2 ;Store result into dest5
XSTR [dest6+pos], xd3 ;Store result into dest6
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
.lessthan16:
;; Tail len
;; Do one more overlap pass
;; Overlapped offset length-16
mov tmp, len ;Backup len as len=rdi
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest4+tmp] ;Get next dest vector
XLDR xd2, [dest5+tmp] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest6+tmp] ;reuse xtmph1. Get next dest vector
sub len, pos
movdqa xtmph3, [constip16] ;Load const of i + 16
pinsrb xtmpl3, len.w, 15
pshufb xtmpl3, xmask0f ;Broadcast len to all bytes
pcmpgtb xtmpl3, xtmph3
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
;dest4
pshufb xgft4_hi, x0 ;Lookup mul table of high nibble
pshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft4_hi, xgft4_lo ;GF add high and low partials
pand xgft4_hi, xtmpl3
pxor xd1, xgft4_hi
;dest5
pshufb xgft5_hi, x0 ;Lookup mul table of high nibble
pshufb xgft5_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft5_hi, xgft5_lo ;GF add high and low partials
pand xgft5_hi, xtmpl3
pxor xd2, xgft5_hi
;dest6
pshufb xgft6_hi, x0 ;Lookup mul table of high nibble
pshufb xgft6_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft6_hi, xgft6_lo ;GF add high and low partials
pand xgft6_hi, xtmpl3
pxor xd3, xgft6_hi
XSTR [dest4+tmp], xd1 ;Store result into dest4
XSTR [dest5+tmp], xd2 ;Store result into dest5
XSTR [dest6+tmp], xd3 ;Store result into dest6
movdqu xgft4_lo, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
movdqu xgft4_hi, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
movdqu xgft5_lo, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
movdqu xgft5_hi, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
movdqu xgft6_lo, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqu xgft6_hi, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;reuse xtmpl1. Get next dest vector
XLDR xd3, [dest3+tmp] ;reuse xtmph1. Get next dest3 vector
;dest1
pshufb xgft4_hi, x0 ;Lookup mul table of high nibble
pshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft4_hi, xgft4_lo ;GF add high and low partials
pand xgft4_hi, xtmpl3
pxor xd1, xgft4_hi
;dest2
pshufb xgft5_hi, x0 ;Lookup mul table of high nibble
pshufb xgft5_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft5_hi, xgft5_lo ;GF add high and low partials
pand xgft5_hi, xtmpl3
pxor xd2, xgft5_hi
;dest3
pshufb xgft6_hi, x0 ;Lookup mul table of high nibble
pshufb xgft6_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft6_hi, xgft6_lo ;GF add high and low partials
pand xgft6_hi, xtmpl3
pxor xd3, xgft6_hi
XSTR [dest1+tmp], xd1 ;Store result into dest1
XSTR [dest2+tmp], xd2 ;Store result into dest2
XSTR [dest3+tmp], xd3 ;Store result into dest3
.return_pass:
FUNC_RESTORE
mov return, 0
ret
.return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
constip16:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_6VECT_MAD_SSE, 00, 00, 020f

303
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_dot_prod_avx.asm generated vendored
View File

@ -1,303 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_dot_prod_avx(len, vec, *g_tbls, **buffs, *dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_DOT_PROD_AVX _gf_vect_dot_prod_avx
%else
%define GF_VECT_DOT_PROD_AVX gf_vect_dot_prod_avx
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved and loaded
%define tmp r11
%define tmp2 r10
%define tmp3 rdi ; must be saved and loaded
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define frame_size 2*8
%define arg(x) [rsp + frame_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
rex_push_reg r12
push_reg rdi
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
pop rdi
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define trans ecx ;trans is for the variables in stack
%define arg0 trans
%define arg0_m arg(0)
%define arg1 trans
%define arg1_m arg(1)
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 ebx
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp2 edi
%define tmp3 esi
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
push esi
push edi
push ebx
mov arg3, arg(3)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
mov esp, ebp
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define vec_i tmp2
%define ptr tmp3
%define pos return
%ifidn PS,4 ;32-bit code
%define vec_m arg1_m
%define len_m arg0_m
%define dest_m arg4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ; 64-bit code
default rel
[bits 64]
%endif
section .text
%define xmask0f xmm5
%define xgft_lo xmm4
%define xgft_hi xmm3
%define x0 xmm0
%define xtmpa xmm1
%define xp xmm2
align 16
global GF_VECT_DOT_PROD_AVX:function
func(GF_VECT_DOT_PROD_AVX)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
.loop16:
vpxor xp, xp
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
mov ptr, [src+vec_i*PS]
vmovdqu xgft_lo, [tmp] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
vmovdqu xgft_hi, [tmp+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
XLDR x0, [ptr+pos] ;Get next source vector
add tmp, 32
add vec_i, 1
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft_hi, xgft_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft_lo, xgft_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft_hi, xgft_hi, xgft_lo ;GF add high and low partials
vpxor xp, xp, xgft_hi ;xp += partial
SLDR vec, vec_m
cmp vec_i, vec
jl .next_vect
SLDR dest, dest_m
XSTR [dest+pos], xp
add pos, 16 ;Loop on 16 bytes at a time
SLDR len, len_m
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f:
ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_DOT_PROD_AVX, 02, 04, 0061

315
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_dot_prod_avx2.asm generated vendored
View File

@ -1,315 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, *dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_DOT_PROD_AVX2 _gf_vect_dot_prod_avx2
%else
%define GF_VECT_DOT_PROD_AVX2 gf_vect_dot_prod_avx2
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r9
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r9
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved and loaded
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 rdi ; must be saved and loaded
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define frame_size 2*8
%define arg(x) [rsp + frame_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
rex_push_reg r12
push_reg rdi
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
pop rdi
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define trans ecx ;trans is for the variables in stack
%define arg0 trans
%define arg0_m arg(0)
%define arg1 trans
%define arg1_m arg(1)
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 ebx
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp.w edx
%define tmp.b dl
%define tmp2 edi
%define tmp3 esi
%define return eax
%macro SLDR 2 ;stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
push esi
push edi
push ebx
mov arg3, arg(3)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
mov esp, ebp
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define vec_i tmp2
%define ptr tmp3
%define pos return
%ifidn PS,4 ;32-bit code
%define vec_m arg1_m
%define len_m arg0_m
%define dest_m arg4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%define xmask0f ymm3
%define xmask0fx xmm3
%define xgft_lo ymm4
%define xgft_hi ymm5
%define x0 ymm0
%define xtmpa ymm1
%define xp ymm2
align 16
global GF_VECT_DOT_PROD_AVX2:function
func(GF_VECT_DOT_PROD_AVX2)
FUNC_SAVE
SLDR len, len_m
sub len, 32
SSTR len_m, len
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
.loop32:
vpxor xp, xp
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
mov ptr, [src+vec_i*PS]
vmovdqu xgft_lo, [tmp] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
vperm2i128 xgft_hi, xgft_lo, xgft_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft_lo, xgft_lo, xgft_lo, 0x00 ; swapped to lo | lo
XLDR x0, [ptr+pos] ;Get next source vector
add tmp, 32
add vec_i, 1
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft_hi, xgft_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft_lo, xgft_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft_hi, xgft_hi, xgft_lo ;GF add high and low partials
vpxor xp, xp, xgft_hi ;xp += partial
SLDR vec, vec_m
cmp vec_i, vec
jl .next_vect
SLDR dest, dest_m
XSTR [dest+pos], xp
add pos, 32 ;Loop on 32 bytes at a time
SLDR len, len_m
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-32
jmp .loop32 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_DOT_PROD_AVX2, 04, 04, 0190

303
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_dot_prod_sse.asm generated vendored
View File

@ -1,303 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_dot_prod_sse(len, vec, *g_tbls, **buffs, *dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_DOT_PROD_SSE _gf_vect_dot_prod_sse
%else
%define GF_VECT_DOT_PROD_SSE gf_vect_dot_prod_sse
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved and loaded
%define tmp r11
%define tmp2 r10
%define tmp3 rdi ; must be saved and loaded
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define frame_size 2*8
%define arg(x) [rsp + frame_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
rex_push_reg r12
push_reg rdi
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
pop rdi
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define trans ecx ;trans is for the variables in stack
%define arg0 trans
%define arg0_m arg(0)
%define arg1 trans
%define arg1_m arg(1)
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 ebx
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp2 edi
%define tmp3 esi
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
push esi
push edi
push ebx
mov arg3, arg(3)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
mov esp, ebp
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest arg4
%define vec_i tmp2
%define ptr tmp3
%define pos return
%ifidn PS,4 ;32-bit code
%define vec_m arg1_m
%define len_m arg0_m
%define dest_m arg4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%define xmask0f xmm5
%define xgft_lo xmm4
%define xgft_hi xmm3
%define x0 xmm0
%define xtmpa xmm1
%define xp xmm2
align 16
global GF_VECT_DOT_PROD_SSE:function
func(GF_VECT_DOT_PROD_SSE)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
.loop16:
pxor xp, xp
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
mov ptr, [src+vec_i*PS]
movdqu xgft_lo, [tmp] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
movdqu xgft_hi, [tmp+16] ; " Cx{00}, Cx{10}, ..., Cx{f0}
XLDR x0, [ptr+pos] ;Get next source vector
add tmp, 32
add vec_i, 1
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xgft_hi, x0 ;Lookup mul table of high nibble
pshufb xgft_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft_hi, xgft_lo ;GF add high and low partials
pxor xp, xgft_hi ;xp += partial
SLDR vec, vec_m
cmp vec_i, vec
jl .next_vect
SLDR dest, dest_m
XSTR [dest+pos], xp
add pos, 16 ;Loop on 16 bytes at a time
SLDR len, len_m
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_DOT_PROD_SSE, 00, 04, 0060

223
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_mad_avx.asm generated vendored
View File

@ -1,223 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_mad_avx(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_MAD_AVX _gf_vect_mad_avx
%else
%define GF_VECT_MAD_AVX gf_vect_mad_avx
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define return rax
%define return.w eax
%define PS 8
%define stack_size 16*3 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
save_reg r12, 3*16 + 0*8
save_reg r15, 3*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
mov r12, [rsp + 3*16 + 0*8]
mov r15, [rsp + 3*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_vect_mad_avx(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest arg5
%define pos return
%define pos.w return.w
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm8
%define xgft_lo xmm7
%define xgft_hi xmm6
%define x0 xmm0
%define xtmpa xmm1
%define xtmph xmm2
%define xtmpl xmm3
%define xd xmm4
%define xtmpd xmm5
align 16
global GF_VECT_MAD_AVX:function
func(GF_VECT_MAD_AVX)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec_i, 5 ;Multiply by 32
vmovdqu xgft_lo, [vec_i+mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
vmovdqu xgft_hi, [vec_i+mul_array+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
XLDR xtmpd, [dest+len] ;backup the last 16 bytes in dest
.loop16:
XLDR xd, [dest+pos] ;Get next dest vector
.loop16_overlap:
XLDR x0, [src+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xtmph, xgft_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd, xd, xtmph ;xd += partial
XSTR [dest+pos], xd
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
vmovdqa xd, xtmpd ;Restore xd
jmp .loop16_overlap ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_MAD_AVX, 02, 00, 0201

233
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_mad_avx2.asm generated vendored
View File

@ -1,233 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_MAD_AVX2 _gf_vect_mad_avx2
%else
%define GF_VECT_MAD_AVX2 gf_vect_mad_avx2
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved and loaded
%define arg5 r15
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define PS 8
%define stack_size 16*3 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
save_reg r12, 3*16 + 0*8
save_reg r15, 3*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
mov r12, [rsp + 3*16 + 0*8]
mov r15, [rsp + 3*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest arg5
%define pos return
%define pos.w return.w
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm8
%define xmask0fx xmm8
%define xgft_lo ymm7
%define xgft_hi ymm6
%define x0 ymm0
%define xtmpa ymm1
%define xtmph ymm2
%define xtmpl ymm3
%define xd ymm4
%define xtmpd ymm5
align 16
global GF_VECT_MAD_AVX2:function
func(GF_VECT_MAD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec_i, 5 ;Multiply by 32
vmovdqu xgft_lo, [vec_i+mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
vperm2i128 xgft_hi, xgft_lo, xgft_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft_lo, xgft_lo, xgft_lo, 0x00 ; swapped to lo | lo
XLDR xtmpd, [dest+len] ;backup the last 32 bytes in dest
.loop32:
XLDR xd, [dest+pos] ;Get next dest vector
.loop32_overlap:
XLDR x0, [src+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xtmph, xgft_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl, xgft_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph, xtmph, xtmpl ;GF add high and low partials
vpxor xd, xd, xtmph ;xd += partial
XSTR [dest+pos], xd
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-32
vmovdqa xd, xtmpd ;Restore xd
jmp .loop32_overlap ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_MAD_AVX2, 04, 00, 0202

224
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_mad_sse.asm generated vendored
View File

@ -1,224 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_mad_sse(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_MAD_SSE _gf_vect_mad_sse
%else
%define GF_VECT_MAD_SSE gf_vect_mad_sse
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define return rax
%define return.w eax
%define PS 8
%define stack_size 16*3 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
save_reg r12, 3*16 + 0*8
save_reg r15, 3*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
mov r12, [rsp + 3*16 + 0*8]
mov r15, [rsp + 3*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_vect_mad_sse(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest arg5
%define pos return
%define pos.w return.w
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm8
%define xgft_lo xmm7
%define xgft_hi xmm6
%define x0 xmm0
%define xtmpa xmm1
%define xtmph xmm2
%define xtmpl xmm3
%define xd xmm4
%define xtmpd xmm5
align 16
global GF_VECT_MAD_SSE:function
func(GF_VECT_MAD_SSE)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec_i, 5 ;Multiply by 32
movdqu xgft_lo, [vec_i+mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqu xgft_hi, [vec_i+mul_array+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
XLDR xtmpd, [dest+len] ;backup the last 16 bytes in dest
.loop16:
XLDR xd, [dest+pos] ;Get next dest vector
.loop16_overlap:
XLDR x0, [src+pos] ;Get next source vector
movdqa xtmph, xgft_hi ;Reload const array registers
movdqa xtmpl, xgft_lo
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xtmph, x0 ;Lookup mul table of high nibble
pshufb xtmpl, xtmpa ;Lookup mul table of low nibble
pxor xtmph, xtmpl ;GF add high and low partials
pxor xd, xtmph
XSTR [dest+pos], xd ;Store result
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
movdqa xd, xtmpd ;Restore xd
jmp .loop16_overlap ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_MAD_SSE, 00, 00, 0200

148
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_mul.h generated vendored
View File

@ -1,148 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**********************************************************************/
#ifndef _GF_VECT_MUL_H
#define _GF_VECT_MUL_H
/**
* @file gf_vect_mul.h
* @brief Interface to functions for vector (block) multiplication in GF(2^8).
*
* This file defines the interface to routines used in fast RAID rebuild and
* erasure codes.
*/
#ifdef __cplusplus
extern "C" {
#endif
/**
* @brief GF(2^8) vector multiply by constant.
*
* Does a GF(2^8) vector multiply b = Ca where a and b are arrays and C
* is a single field element in GF(2^8). Can be used for RAID6 rebuild
* and partial write functions. Function requires pre-calculation of a
* 32-element constant array based on constant C. gftbl(C) = {C{00},
* C{01}, C{02}, ... , C{0f} }, {C{00}, C{10}, C{20}, ... , C{f0} }. Len
* and src must be aligned to 32B.
* @requires SSE4.1
*
* @param len Length of vector in bytes. Must be aligned to 32B.
* @param gftbl Pointer to 32-byte array of pre-calculated constants based on C.
* @param src Pointer to src data array. Must be aligned to 32B.
* @param dest Pointer to destination data array. Must be aligned to 32B.
* @returns 0 pass, other fail
*/
int gf_vect_mul_sse(int len, unsigned char *gftbl, void *src, void *dest);
/**
* @brief GF(2^8) vector multiply by constant.
*
* Does a GF(2^8) vector multiply b = Ca where a and b are arrays and C
* is a single field element in GF(2^8). Can be used for RAID6 rebuild
* and partial write functions. Function requires pre-calculation of a
* 32-element constant array based on constant C. gftbl(C) = {C{00},
* C{01}, C{02}, ... , C{0f} }, {C{00}, C{10}, C{20}, ... , C{f0} }. Len
* and src must be aligned to 32B.
* @requires AVX
*
* @param len Length of vector in bytes. Must be aligned to 32B.
* @param gftbl Pointer to 32-byte array of pre-calculated constants based on C.
* @param src Pointer to src data array. Must be aligned to 32B.
* @param dest Pointer to destination data array. Must be aligned to 32B.
* @returns 0 pass, other fail
*/
int gf_vect_mul_avx(int len, unsigned char *gftbl, void *src, void *dest);
/**
* @brief GF(2^8) vector multiply by constant, runs appropriate version.
*
* Does a GF(2^8) vector multiply b = Ca where a and b are arrays and C
* is a single field element in GF(2^8). Can be used for RAID6 rebuild
* and partial write functions. Function requires pre-calculation of a
* 32-element constant array based on constant C. gftbl(C) = {C{00},
* C{01}, C{02}, ... , C{0f} }, {C{00}, C{10}, C{20}, ... , C{f0} }.
* Len and src must be aligned to 32B.
*
* This function determines what instruction sets are enabled
* and selects the appropriate version at runtime.
*
* @param len Length of vector in bytes. Must be aligned to 32B.
* @param gftbl Pointer to 32-byte array of pre-calculated constants based on C.
* @param src Pointer to src data array. Must be aligned to 32B.
* @param dest Pointer to destination data array. Must be aligned to 32B.
* @returns 0 pass, other fail
*/
int gf_vect_mul(int len, unsigned char *gftbl, void *src, void *dest);
/**
* @brief Initialize 32-byte constant array for GF(2^8) vector multiply
*
* Calculates array {C{00}, C{01}, C{02}, ... , C{0f} }, {C{00}, C{10},
* C{20}, ... , C{f0} } as required by other fast vector multiply
* functions.
* @param c Constant input.
* @param gftbl Table output.
*/
void gf_vect_mul_init(unsigned char c, unsigned char* gftbl);
/**
* @brief GF(2^8) vector multiply by constant, runs baseline version.
*
* Does a GF(2^8) vector multiply b = Ca where a and b are arrays and C
* is a single field element in GF(2^8). Can be used for RAID6 rebuild
* and partial write functions. Function requires pre-calculation of a
* 32-element constant array based on constant C. gftbl(C) = {C{00},
* C{01}, C{02}, ... , C{0f} }, {C{00}, C{10}, C{20}, ... , C{f0} }. Len
* and src must be aligned to 32B.
*
* @param len Length of vector in bytes. Must be aligned to 32B.
* @param a Pointer to 32-byte array of pre-calculated constants based on C.
* only use 2nd element is used.
* @param src Pointer to src data array. Must be aligned to 32B.
* @param dest Pointer to destination data array. Must be aligned to 32B.
*/
void gf_vect_mul_base(int len, unsigned char *a, unsigned char *src,
unsigned char *dest);
#ifdef __cplusplus
}
#endif
#endif //_GF_VECT_MUL_H

189
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_mul_avx.asm generated vendored
View File

@ -1,189 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_mul_avx(len, mul_array, src, dest)
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_MUL_AVX _gf_vect_mul_avx
%else
%define GF_VECT_MUL_AVX gf_vect_mul_avx
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define return rax
%define stack_size 5*16 + 8 ; must be an odd multiple of 8
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm13, 2*16
save_xmm128 xmm14, 3*16
save_xmm128 xmm15, 4*16
end_prolog
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm13, [rsp + 2*16]
vmovdqa xmm14, [rsp + 3*16]
vmovdqa xmm15, [rsp + 4*16]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define mul_array arg1
%define src arg2
%define dest arg3
%define pos return
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft_lo xmm14
%define xgft_hi xmm13
%define x0 xmm0
%define xtmp1a xmm1
%define xtmp1b xmm2
%define xtmp1c xmm3
%define x1 xmm4
%define xtmp2a xmm5
%define xtmp2b xmm6
%define xtmp2c xmm7
align 16
global GF_VECT_MUL_AVX:function
func(GF_VECT_MUL_AVX)
FUNC_SAVE
mov pos, 0
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
vmovdqu xgft_lo, [mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
vmovdqu xgft_hi, [mul_array+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
loop32:
XLDR x0, [src+pos] ;Get next source vector
XLDR x1, [src+pos+16] ;Get next source vector + 16B ahead
add pos, 32 ;Loop on 16 bytes at a time
cmp pos, len
vpand xtmp1a, x0, xmask0f ;Mask low src nibble in bits 4-0
vpand xtmp2a, x1, xmask0f
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpsraw x1, x1, 4
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpand x1, x1, xmask0f
vpshufb xtmp1b, xgft_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmp1c, xgft_lo, xtmp1a ;Lookup mul table of low nibble
vpshufb xtmp2b, xgft_hi, x1 ;Lookup mul table of high nibble
vpshufb xtmp2c, xgft_lo, xtmp2a ;Lookup mul table of low nibble
vpxor xtmp1b, xtmp1b, xtmp1c ;GF add high and low partials
vpxor xtmp2b, xtmp2b, xtmp2c
XSTR [dest+pos-32], xtmp1b ;Store result
XSTR [dest+pos-16], xtmp2b ;Store +16B result
jl loop32
return_pass:
FUNC_RESTORE
sub pos, len
ret
return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
align 16
mask0f:
ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_MUL_AVX, 01, 02, 0036

195
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/gf_vect_mul_sse.asm generated vendored
View File

@ -1,195 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_mul_sse(len, mul_array, src, dest)
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_VECT_MUL_SSE _gf_vect_mul_sse
%else
%define GF_VECT_MUL_SSE gf_vect_mul_sse
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define return rax
%define stack_size 5*16 + 8 ; must be an odd multiple of 8
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm13, 2*16
save_xmm128 xmm14, 3*16
save_xmm128 xmm15, 4*16
end_prolog
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp + 0*16]
movdqa xmm7, [rsp + 1*16]
movdqa xmm13, [rsp + 2*16]
movdqa xmm14, [rsp + 3*16]
movdqa xmm15, [rsp + 4*16]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define mul_array arg1
%define src arg2
%define dest arg3
%define pos return
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm15
%define xgft_lo xmm14
%define xgft_hi xmm13
%define x0 xmm0
%define xtmp1a xmm1
%define xtmp1b xmm2
%define xtmp1c xmm3
%define x1 xmm4
%define xtmp2a xmm5
%define xtmp2b xmm6
%define xtmp2c xmm7
align 16
global GF_VECT_MUL_SSE:function
func(GF_VECT_MUL_SSE)
FUNC_SAVE
mov pos, 0
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
movdqu xgft_lo, [mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
movdqu xgft_hi, [mul_array+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
loop32:
XLDR x0, [src+pos] ;Get next source vector
XLDR x1, [src+pos+16] ;Get next source vector + 16B ahead
movdqa xtmp1b, xgft_hi ;Reload const array registers
movdqa xtmp1c, xgft_lo
movdqa xtmp2b, xgft_hi
movdqa xtmp2c, xgft_lo
movdqa xtmp1a, x0 ;Keep unshifted copy of src
movdqa xtmp2a, x1
psraw x0, 4 ;Shift to put high nibble into bits 4-0
psraw x1, 4
pand xtmp1a, xmask0f ;Mask low src nibble in bits 4-0
pand xtmp2a, xmask0f
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand x1, xmask0f
pshufb xtmp1b, x0 ;Lookup mul table of high nibble
pshufb xtmp1c, xtmp1a ;Lookup mul table of low nibble
pshufb xtmp2b, x1
pshufb xtmp2c, xtmp2a
pxor xtmp1b, xtmp1c ;GF add high and low partials
pxor xtmp2b, xtmp2c
XSTR [dest+pos], xtmp1b ;Store result
XSTR [dest+pos+16], xtmp2b ;Store +16B result
add pos, 32 ;Loop on 32 bytes at at time
cmp pos, len
jl loop32
return_pass:
sub pos, len
FUNC_RESTORE
ret
return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f:
ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_VECT_MUL_SSE, 00, 02, 0034

38
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/stdint.go generated vendored
View File

@ -1,38 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
//
// int sizeInt()
// {
// return sizeof(int);
// }
import "C"
import "unsafe"
var (
// See http://golang.org/ref/spec#Numeric_types
sizeInt = int(C.sizeInt())
// SizeInt8 is the byte size of a int8.
sizeInt8 = int(unsafe.Sizeof(int8(0)))
// SizeInt16 is the byte size of a int16.
sizeInt16 = int(unsafe.Sizeof(int16(0)))
// SizeInt32 is the byte size of a int32.
sizeInt32 = int(unsafe.Sizeof(int32(0)))
// SizeInt64 is the byte size of a int64.
sizeInt64 = int(unsafe.Sizeof(int64(0)))
)

66
pkg/api/Godeps/_workspace/src/github.com/minio-io/erasure/vandermonde_test.go generated vendored
View File

@ -1,66 +0,0 @@
/*
* Minimalist Object Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
import (
"bytes"
. "gopkg.in/check.v1"
)
func corruptChunks(chunks [][]byte, errorIndex []int) [][]byte {
for _, err := range errorIndex {
chunks[err] = nil
}
return chunks
}
func (s *MySuite) TestVanderMondeEncodeDecodeFailure(c *C) {
ep, _ := ValidateParams(k, m, Vandermonde)
data := []byte("Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry's standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum.")
e := NewErasure(ep)
chunks, err := e.Encode(data)
c.Assert(err, IsNil)
errorIndex := []int{0, 3, 5, 9, 11, 13}
chunks = corruptChunks(chunks, errorIndex)
_, err = e.Decode(chunks, len(data))
c.Assert(err, Not(IsNil))
}
func (s *MySuite) TestVanderMondeEncodeDecodeSuccess(c *C) {
ep, _ := ValidateParams(k, m, Vandermonde)
data := []byte("Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry's standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum.")
e := NewErasure(ep)
chunks, err := e.Encode(data)
c.Assert(err, IsNil)
errorIndex := []int{0, 3, 5, 9, 13}
chunks = corruptChunks(chunks, errorIndex)
recoveredData, err := e.Decode(chunks, len(data))
c.Assert(err, IsNil)
if !bytes.Equal(recoveredData, data) {
c.Fatalf("Recovered data mismatches with original data")
}
}

24
pkg/api/Godeps/_workspace/src/github.com/minio-io/iodine/.gitignore generated vendored
View File

@ -1,24 +0,0 @@
# Compiled Object files, Static and Dynamic libs (Shared Objects)
*.o
*.a
*.so
# Folders
_obj
_test
# Architecture specific extensions/prefixes
*.[568vq]
[568vq].out
*.cgo1.go
*.cgo2.c
_cgo_defun.c
_cgo_gotypes.go
_cgo_export.*
_testmain.go
*.exe
*.test
*.prof

202
pkg/api/Godeps/_workspace/src/github.com/minio-io/iodine/LICENSE generated vendored
View File

@ -1,202 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "{}"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright {yyyy} {name of copyright owner}
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

2
pkg/api/Godeps/_workspace/src/github.com/minio-io/iodine/README.md generated vendored
View File

@ -1,2 +0,0 @@
# iodine
Iodine is an error logging framework for tracing errors.

201
pkg/api/Godeps/_workspace/src/github.com/minio-io/iodine/iodine.go generated vendored
View File

@ -1,201 +0,0 @@
/*
* Iodine, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package iodine
import (
"bytes"
"encoding/json"
"fmt"
"os"
"path"
"reflect"
"runtime"
"strconv"
"strings"
"sync"
)
// WrappedError is the iodine error which contains a pointer to the original error
// and stack traces.
type Error struct {
EmbeddedError error `json:"-"`
ErrorMessage string
ErrorType string
Stack []StackEntry
}
// StackEntry contains the entry in the stack trace
type StackEntry struct {
Host string
File string
Line int
Data map[string]string
}
var gopath string
var globalState = struct {
sync.RWMutex
m map[string]string
}{m: make(map[string]string)}
// SetGlobalState - set global state
func SetGlobalState(key, value string) {
globalState.Lock()
globalState.m[key] = value
globalState.Unlock()
}
// ClearGlobalState - clear info in globalState struct
func ClearGlobalState() {
globalState.Lock()
for k := range globalState.m {
delete(globalState.m, k)
}
globalState.Unlock()
}
// GetGlobalState - get map from globalState struct
func GetGlobalState() map[string]string {
result := make(map[string]string)
globalState.RLock()
for k, v := range globalState.m {
result[k] = v
}
globalState.RUnlock()
return result
}
// GetGlobalStateKey - get value for key from globalState struct
func GetGlobalStateKey(k string) string {
result, ok := globalState.m[k]
if !ok {
return ""
}
return result
}
// Error - instantiate an error, turning it into an iodine error.
// Adds an initial stack trace.
func New(err error, data map[string]string) error {
if err != nil {
entry := createStackEntry()
var newErr Error
// check if error is wrapped
switch typedError := err.(type) {
case Error:
{
newErr = typedError
}
default:
{
newErr = Error{
EmbeddedError: err,
ErrorMessage: err.Error(),
ErrorType: reflect.TypeOf(err).String(),
Stack: []StackEntry{},
}
}
}
for k, v := range data {
entry.Data[k] = v
}
newErr.Stack = append(newErr.Stack, entry)
return newErr
}
return nil
}
// createStackEntry - create stack entries
func createStackEntry() StackEntry {
host, _ := os.Hostname()
_, file, line, _ := runtime.Caller(2)
file = strings.TrimPrefix(file, gopath) // trim gopath from file
data := GetGlobalState()
for k, v := range getSystemData() {
data[k] = v
}
entry := StackEntry{
Host: host,
File: file,
Line: line,
Data: data,
}
return entry
}
func getSystemData() map[string]string {
host, err := os.Hostname()
if err != nil {
host = ""
}
memstats := &runtime.MemStats{}
runtime.ReadMemStats(memstats)
return map[string]string{
"sys.host": host,
"sys.os": runtime.GOOS,
"sys.arch": runtime.GOARCH,
"sys.go": runtime.Version(),
"sys.cpus": strconv.Itoa(runtime.NumCPU()),
"sys.mem.used": strconv.FormatUint(memstats.Alloc, 10),
"sys.mem.allocated": strconv.FormatUint(memstats.TotalAlloc, 10),
"sys.mem.heap.used": strconv.FormatUint(memstats.HeapAlloc, 10),
"sys.mem.heap.allocated": strconv.FormatUint(memstats.HeapSys, 10),
}
}
// Annotate an error with a stack entry and returns itself
//func (err *WrappedError) Annotate(info map[string]string) *WrappedError {
// entry := createStackEntry()
// for k, v := range info {
// entry.Data[k] = v
// }
// err.Stack = append(err.Stack, entry)
// return err
//}
// EmitJSON writes JSON output for the error
func (err Error) EmitJSON() ([]byte, error) {
return json.Marshal(err)
}
// EmitHumanReadable returns a human readable error message
func (err Error) EmitHumanReadable() string {
var errorBuffer bytes.Buffer
fmt.Fprintln(&errorBuffer, err.ErrorMessage)
for i, entry := range err.Stack {
fmt.Fprintln(&errorBuffer, "-", i, entry.Host+":"+entry.File+":"+strconv.Itoa(entry.Line), entry.Data)
}
return string(errorBuffer.Bytes())
}
// Emits the original error message
func (err Error) Error() string {
return err.EmitHumanReadable()
}
func init() {
_, iodineFile, _, _ := runtime.Caller(0)
iodineFile = path.Dir(iodineFile) // trim iodine.go
iodineFile = path.Dir(iodineFile) // trim iodine
iodineFile = path.Dir(iodineFile) // trim minio-io
gopath = path.Dir(iodineFile) + "/" // trim github.com
}

97
pkg/api/Godeps/_workspace/src/github.com/minio-io/iodine/iodine_test.go generated vendored
View File

@ -1,97 +0,0 @@
/*
* Iodine, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package iodine
import (
"encoding/json"
"errors"
"testing"
)
func TestIodine(t *testing.T) {
iodineError := New(errors.New("Hello"), nil)
iodineError = New(iodineError, nil)
iodineError = New(iodineError, nil)
iodineError = New(iodineError, nil)
switch typedError := iodineError.(type) {
case Error:
{
if len(typedError.Stack) != 4 {
t.Fail()
}
_, err := json.MarshalIndent(typedError, "", " ")
if err != nil {
t.Fail()
}
}
default:
{
t.Fail()
}
}
}
func TestState(t *testing.T) {
SetGlobalState("hello", "world")
result := GetGlobalStateKey("hello")
if result != "world" {
t.Error("global state not set: hello->world")
t.Fail()
}
ClearGlobalState()
if len(GetGlobalState()) != 0 {
t.Fail()
}
SetGlobalState("foo", "bar")
err := New(errors.New("a simple error"), nil)
switch typedError := err.(type) {
case Error:
{
if res, ok := typedError.Stack[0].Data["foo"]; ok {
if res != "bar" {
t.Error("global state not set: foo->bar")
}
} else {
t.Fail()
}
typedError = New(typedError, map[string]string{"foo2": "bar2"}).(Error)
if res, ok := typedError.Stack[0].Data["foo"]; ok {
if res != "bar" {
t.Error("annotate should not modify previous data entries")
}
} else {
t.Error("annotate should not remove previous data entries")
}
if res, ok := typedError.Stack[1].Data["foo"]; ok {
if res != "bar" {
t.Error("global state should set value properly in annotate")
}
} else {
t.Error("global state should set key properly in annotate")
}
if res, ok := typedError.Stack[1].Data["foo2"]; ok {
if res != "bar2" {
// typedError = Error(typedError, nil).(WrappedError)
t.Error("foo2 -> bar should be set")
}
} else {
// typedError = Error(typedError, nil).(WrappedError)
t.Error("foo2 should be set")
}
}
}
}

22
pkg/api/Godeps/_workspace/src/github.com/stretchr/objx/.gitignore generated vendored
View File

@ -1,22 +0,0 @@
# Compiled Object files, Static and Dynamic libs (Shared Objects)
*.o
*.a
*.so
# Folders
_obj
_test
# Architecture specific extensions/prefixes
*.[568vq]
[568vq].out
*.cgo1.go
*.cgo2.c
_cgo_defun.c
_cgo_gotypes.go
_cgo_export.*
_testmain.go
*.exe

23
pkg/api/Godeps/_workspace/src/github.com/stretchr/objx/LICENSE.md generated vendored
View File

@ -1,23 +0,0 @@
objx - by Mat Ryer and Tyler Bunnell
The MIT License (MIT)
Copyright (c) 2014 Stretchr, Inc.
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

3
pkg/api/Godeps/_workspace/src/github.com/stretchr/objx/README.md generated vendored
View File

@ -1,3 +0,0 @@
# objx
* Jump into the [API Documentation](http://godoc.org/github.com/stretchr/objx)

179
pkg/api/Godeps/_workspace/src/github.com/stretchr/objx/accessors.go generated vendored
View File

@ -1,179 +0,0 @@
package objx
import (
"fmt"
"regexp"
"strconv"
"strings"
)
// arrayAccesRegexString is the regex used to extract the array number
// from the access path
const arrayAccesRegexString = `^(.+)\[([0-9]+)\]$`
// arrayAccesRegex is the compiled arrayAccesRegexString
var arrayAccesRegex = regexp.MustCompile(arrayAccesRegexString)
// Get gets the value using the specified selector and
// returns it inside a new Obj object.
//
// If it cannot find the value, Get will return a nil
// value inside an instance of Obj.
//
// Get can only operate directly on map[string]interface{} and []interface.
//
// Example
//
// To access the title of the third chapter of the second book, do:
//
// o.Get("books[1].chapters[2].title")
func (m Map) Get(selector string) *Value {
rawObj := access(m, selector, nil, false, false)
return &Value{data: rawObj}
}
// Set sets the value using the specified selector and
// returns the object on which Set was called.
//
// Set can only operate directly on map[string]interface{} and []interface
//
// Example
//
// To set the title of the third chapter of the second book, do:
//
// o.Set("books[1].chapters[2].title","Time to Go")
func (m Map) Set(selector string, value interface{}) Map {
access(m, selector, value, true, false)
return m
}
// access accesses the object using the selector and performs the
// appropriate action.
func access(current, selector, value interface{}, isSet, panics bool) interface{} {
switch selector.(type) {
case int, int8, int16, int32, int64, uint, uint8, uint16, uint32, uint64:
if array, ok := current.([]interface{}); ok {
index := intFromInterface(selector)
if index >= len(array) {
if panics {
panic(fmt.Sprintf("objx: Index %d is out of range. Slice only contains %d items.", index, len(array)))
}
return nil
}
return array[index]
}
return nil
case string:
selStr := selector.(string)
selSegs := strings.SplitN(selStr, PathSeparator, 2)
thisSel := selSegs[0]
index := -1
var err error
// https://github.com/stretchr/objx/issues/12
if strings.Contains(thisSel, "[") {
arrayMatches := arrayAccesRegex.FindStringSubmatch(thisSel)
if len(arrayMatches) > 0 {
// Get the key into the map
thisSel = arrayMatches[1]
// Get the index into the array at the key
index, err = strconv.Atoi(arrayMatches[2])
if err != nil {
// This should never happen. If it does, something has gone
// seriously wrong. Panic.
panic("objx: Array index is not an integer. Must use array[int].")
}
}
}
if curMap, ok := current.(Map); ok {
current = map[string]interface{}(curMap)
}
// get the object in question
switch current.(type) {
case map[string]interface{}:
curMSI := current.(map[string]interface{})
if len(selSegs) <= 1 && isSet {
curMSI[thisSel] = value
return nil
} else {
current = curMSI[thisSel]
}
default:
current = nil
}
if current == nil && panics {
panic(fmt.Sprintf("objx: '%v' invalid on object.", selector))
}
// do we need to access the item of an array?
if index > -1 {
if array, ok := current.([]interface{}); ok {
if index < len(array) {
current = array[index]
} else {
if panics {
panic(fmt.Sprintf("objx: Index %d is out of range. Slice only contains %d items.", index, len(array)))
}
current = nil
}
}
}
if len(selSegs) > 1 {
current = access(current, selSegs[1], value, isSet, panics)
}
}
return current
}
// intFromInterface converts an interface object to the largest
// representation of an unsigned integer using a type switch and
// assertions
func intFromInterface(selector interface{}) int {
var value int
switch selector.(type) {
case int:
value = selector.(int)
case int8:
value = int(selector.(int8))
case int16:
value = int(selector.(int16))
case int32:
value = int(selector.(int32))
case int64:
value = int(selector.(int64))
case uint:
value = int(selector.(uint))
case uint8:
value = int(selector.(uint8))
case uint16:
value = int(selector.(uint16))
case uint32:
value = int(selector.(uint32))
case uint64:
value = int(selector.(uint64))
default:
panic("objx: array access argument is not an integer type (this should never happen)")
}
return value
}

145
pkg/api/Godeps/_workspace/src/github.com/stretchr/objx/accessors_test.go generated vendored
View File

@ -1,145 +0,0 @@
package objx
import (
"github.com/stretchr/testify/assert"
"testing"
)
func TestAccessorsAccessGetSingleField(t *testing.T) {
current := map[string]interface{}{"name": "Tyler"}
assert.Equal(t, "Tyler", access(current, "name", nil, false, true))
}
func TestAccessorsAccessGetDeep(t *testing.T) {
current := map[string]interface{}{"name": map[string]interface{}{"first": "Tyler", "last": "Bunnell"}}
assert.Equal(t, "Tyler", access(current, "name.first", nil, false, true))
assert.Equal(t, "Bunnell", access(current, "name.last", nil, false, true))
}
func TestAccessorsAccessGetDeepDeep(t *testing.T) {
current := map[string]interface{}{"one": map[string]interface{}{"two": map[string]interface{}{"three": map[string]interface{}{"four": 4}}}}
assert.Equal(t, 4, access(current, "one.two.three.four", nil, false, true))
}
func TestAccessorsAccessGetInsideArray(t *testing.T) {
current := map[string]interface{}{"names": []interface{}{map[string]interface{}{"first": "Tyler", "last": "Bunnell"}, map[string]interface{}{"first": "Capitol", "last": "Bollocks"}}}
assert.Equal(t, "Tyler", access(current, "names[0].first", nil, false, true))
assert.Equal(t, "Bunnell", access(current, "names[0].last", nil, false, true))
assert.Equal(t, "Capitol", access(current, "names[1].first", nil, false, true))
assert.Equal(t, "Bollocks", access(current, "names[1].last", nil, false, true))
assert.Panics(t, func() {
access(current, "names[2]", nil, false, true)
})
assert.Nil(t, access(current, "names[2]", nil, false, false))
}
func TestAccessorsAccessGetFromArrayWithInt(t *testing.T) {
current := []interface{}{map[string]interface{}{"first": "Tyler", "last": "Bunnell"}, map[string]interface{}{"first": "Capitol", "last": "Bollocks"}}
one := access(current, 0, nil, false, false)
two := access(current, 1, nil, false, false)
three := access(current, 2, nil, false, false)
assert.Equal(t, "Tyler", one.(map[string]interface{})["first"])
assert.Equal(t, "Capitol", two.(map[string]interface{})["first"])
assert.Nil(t, three)
}
func TestAccessorsGet(t *testing.T) {
current := New(map[string]interface{}{"name": "Tyler"})
assert.Equal(t, "Tyler", current.Get("name").data)
}
func TestAccessorsAccessSetSingleField(t *testing.T) {
current := map[string]interface{}{"name": "Tyler"}
access(current, "name", "Mat", true, false)
assert.Equal(t, current["name"], "Mat")
access(current, "age", 29, true, true)
assert.Equal(t, current["age"], 29)
}
func TestAccessorsAccessSetSingleFieldNotExisting(t *testing.T) {
current := map[string]interface{}{}
access(current, "name", "Mat", true, false)
assert.Equal(t, current["name"], "Mat")
}
func TestAccessorsAccessSetDeep(t *testing.T) {
current := map[string]interface{}{"name": map[string]interface{}{"first": "Tyler", "last": "Bunnell"}}
access(current, "name.first", "Mat", true, true)
access(current, "name.last", "Ryer", true, true)
assert.Equal(t, "Mat", access(current, "name.first", nil, false, true))
assert.Equal(t, "Ryer", access(current, "name.last", nil, false, true))
}
func TestAccessorsAccessSetDeepDeep(t *testing.T) {
current := map[string]interface{}{"one": map[string]interface{}{"two": map[string]interface{}{"three": map[string]interface{}{"four": 4}}}}
access(current, "one.two.three.four", 5, true, true)
assert.Equal(t, 5, access(current, "one.two.three.four", nil, false, true))
}
func TestAccessorsAccessSetArray(t *testing.T) {
current := map[string]interface{}{"names": []interface{}{"Tyler"}}
access(current, "names[0]", "Mat", true, true)
assert.Equal(t, "Mat", access(current, "names[0]", nil, false, true))
}
func TestAccessorsAccessSetInsideArray(t *testing.T) {
current := map[string]interface{}{"names": []interface{}{map[string]interface{}{"first": "Tyler", "last": "Bunnell"}, map[string]interface{}{"first": "Capitol", "last": "Bollocks"}}}
access(current, "names[0].first", "Mat", true, true)
access(current, "names[0].last", "Ryer", true, true)
access(current, "names[1].first", "Captain", true, true)
access(current, "names[1].last", "Underpants", true, true)
assert.Equal(t, "Mat", access(current, "names[0].first", nil, false, true))
assert.Equal(t, "Ryer", access(current, "names[0].last", nil, false, true))
assert.Equal(t, "Captain", access(current, "names[1].first", nil, false, true))
assert.Equal(t, "Underpants", access(current, "names[1].last", nil, false, true))
}
func TestAccessorsAccessSetFromArrayWithInt(t *testing.T) {
current := []interface{}{map[string]interface{}{"first": "Tyler", "last": "Bunnell"}, map[string]interface{}{"first": "Capitol", "last": "Bollocks"}}
one := access(current, 0, nil, false, false)
two := access(current, 1, nil, false, false)
three := access(current, 2, nil, false, false)
assert.Equal(t, "Tyler", one.(map[string]interface{})["first"])
assert.Equal(t, "Capitol", two.(map[string]interface{})["first"])
assert.Nil(t, three)
}
func TestAccessorsSet(t *testing.T) {
current := New(map[string]interface{}{"name": "Tyler"})
current.Set("name", "Mat")
assert.Equal(t, "Mat", current.Get("name").data)
}

14
pkg/api/Godeps/_workspace/src/github.com/stretchr/objx/codegen/array-access.txt generated vendored
View File

@ -1,14 +0,0 @@
case []{1}:
a := object.([]{1})
if isSet {
a[index] = value.({1})
} else {
if index >= len(a) {
if panics {
panic(fmt.Sprintf("objx: Index %d is out of range because the []{1} only contains %d items.", index, len(a)))
}
return nil
} else {
return a[index]
}
}

Some files were not shown because too many files have changed in this diff Show More