MyFavMirrors/minio
1
0
Fork 0
mirror of https://github.com/minio/minio.git synced 2025-04-21 11:04:20 -04:00
Code Issues Packages Projects Releases Wiki Activity

vendor: Bring new updates from blake2b-simd repo. (#2094)

Browse Source 
This vendorization is needed to bring in new improvements
and support for AVX2 and SSE.

Fixes #2081
This commit is contained in:
Harshavardhana 2016-07-06 18:24:31 -07:00 committed by Anand Babu (AB) Periasamy
parent 8c767218a4
commit 169c72cdab
13 changed files with 1974 additions and 124 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

136
vendor/github.com/minio/blake2b-simd/README.md generated vendored
View File

@ -6,15 +6,139 @@ Pure Go implementation of BLAKE2b using SIMD optimizations.
Introduction Introduction
------------ ------------
This package is based on the pure go [BLAKE2b](https://github.com/dchest/blake2b) implementation of Dmitry Chestnykh and merges it with the (`cgo` dependent) SSE optimized [BLAKE2](https://github.com/codahale/blake2) implementation (which in turn is based on [official implementation](https://github.com/BLAKE2/BLAKE2). It does so by using [Go's Assembler](https://golang.org/doc/asm) for amd64 architectures with a fallback for other architectures. This package was initially based on the pure go [BLAKE2b](https://github.com/dchest/blake2b) implementation of Dmitry Chestnykh and merged with the (`cgo` dependent) AVX optimized [BLAKE2](https://github.com/codahale/blake2) implementation (which in turn is based on the [official implementation](https://github.com/BLAKE2/BLAKE2). It does so by using [Go's Assembler](https://golang.org/doc/asm) for amd64 architectures with a golang only fallback for other architectures.
It gives roughly a 3x performance improvement over the non-optimized go version. In addition to AVX there is also support for AVX2 as well as SSE. Best performance is obtained with AVX2 which gives roughly a **4X** performance increase approaching hashing speeds of **1GB/sec** on a single core.
Benchmarks Benchmarks
---------- ----------
| Dura | 1 GB | This is a summary of the performance improvements. Full details are shown below.
| ------------- |:-----:|
| blake2b-SIMD | 1.59s |
| blake2b | 4.66s |
| Technology | 128K |
| ---------- |:-----:|
| AVX2 | 3.94x |
| AVX | 3.28x |
| SSE | 2.85x |
asm2plan9s
----------
In order to be able to work more easily with AVX2/AVX instructions, a separate tool was developed to convert AVX2/AVX instructions into the corresponding BYTE sequence as accepted by Go assembly. See [asm2plan9s](https://github.com/fwessels/asm2plan9s) for more information.
bt2sum
------
[bt2sum](https://github.com/s3git/bt2sum) is a utility that takes advantages of the BLAKE2b SIMD optimizations to compute check sums using the BLAKE2 Tree hashing mode in so called 'unlimited fanout' mode.
Technical details
-----------------
BLAKE2b is a hashing algorithm that operates on 64-bit integer values. The AVX2 version uses the 256-bit wide YMM registers in order to essentially process four operations in parallel. AVX and SSE operate on 128-bit values simultaneously (two operations in parallel). Below are excerpts from `compressAvx2_amd64.s`, `compressAvx_amd64.s`, and `compress_generic.go` respectively.
```
VPADDQ YMM0,YMM0,YMM1 /* v0 += v4, v1 += v5, v2 += v6, v3 += v7 */
```
```
VPADDQ XMM0,XMM0,XMM2 /* v0 += v4, v1 += v5 */
VPADDQ XMM1,XMM1,XMM3 /* v2 += v6, v3 += v7 */
```
```
v0 += v4
v1 += v5
v2 += v6
v3 += v7
```
Detailed benchmarks
-------------------
Example performance metrics were generated on Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz - 6 physical cores, 12 logical cores running Ubuntu GNU/Linux with kernel version 4.4.0-24-generic (vanilla with no optimizations).
### AVX2
```
$ benchcmp go.txt avx2.txt
benchmark old ns/op new ns/op delta
BenchmarkHash64-12 1481 849 -42.67%
BenchmarkHash128-12 1428 746 -47.76%
BenchmarkHash1K-12 6379 2227 -65.09%
BenchmarkHash8K-12 37219 11714 -68.53%
BenchmarkHash32K-12 140716 35935 -74.46%
BenchmarkHash128K-12 561656 142634 -74.60%
benchmark old MB/s new MB/s speedup
BenchmarkHash64-12 43.20 75.37 1.74x
BenchmarkHash128-12 89.64 171.35 1.91x
BenchmarkHash1K-12 160.52 459.69 2.86x
BenchmarkHash8K-12 220.10 699.32 3.18x
BenchmarkHash32K-12 232.87 911.85 3.92x
BenchmarkHash128K-12 233.37 918.93 3.94x
```
### AVX2: Comparison to other hashing techniques
```
$ go test -bench=Comparison
BenchmarkComparisonMD5-12 1000 1726121 ns/op 607.48 MB/s
BenchmarkComparisonSHA1-12 500 2005164 ns/op 522.94 MB/s
BenchmarkComparisonSHA256-12 300 5531036 ns/op 189.58 MB/s
BenchmarkComparisonSHA512-12 500 3423030 ns/op 306.33 MB/s
BenchmarkComparisonBlake2B-12 1000 1232690 ns/op 850.64 MB/s
```
Benchmarks below were generated on a MacBook Pro with a 2.7 GHz Intel Core i7.
### AVX
```
$ benchcmp go.txt avx.txt
benchmark old ns/op new ns/op delta
BenchmarkHash64-8 813 458 -43.67%
BenchmarkHash128-8 766 401 -47.65%
BenchmarkHash1K-8 4881 1763 -63.88%
BenchmarkHash8K-8 36127 12273 -66.03%
BenchmarkHash32K-8 140582 43155 -69.30%
BenchmarkHash128K-8 567850 173246 -69.49%
benchmark old MB/s new MB/s speedup
BenchmarkHash64-8 78.63 139.57 1.78x
BenchmarkHash128-8 166.98 318.73 1.91x
BenchmarkHash1K-8 209.76 580.68 2.77x
BenchmarkHash8K-8 226.76 667.46 2.94x
BenchmarkHash32K-8 233.09 759.29 3.26x
BenchmarkHash128K-8 230.82 756.56 3.28x
```
### SSE
```
$ benchcmp go.txt sse.txt
benchmark old ns/op new ns/op delta
BenchmarkHash64-8 813 478 -41.21%
BenchmarkHash128-8 766 411 -46.34%
BenchmarkHash1K-8 4881 1870 -61.69%
BenchmarkHash8K-8 36127 12427 -65.60%
BenchmarkHash32K-8 140582 49512 -64.78%
BenchmarkHash128K-8 567850 199040 -64.95%
benchmark old MB/s new MB/s speedup
BenchmarkHash64-8 78.63 133.78 1.70x
BenchmarkHash128-8 166.98 311.23 1.86x
BenchmarkHash1K-8 209.76 547.37 2.61x
BenchmarkHash8K-8 226.76 659.20 2.91x
BenchmarkHash32K-8 233.09 661.81 2.84x
BenchmarkHash128K-8 230.82 658.52 2.85x
```
License
-------
Released under the Apache License v2.0. You can find the complete text in the file LICENSE.
Contributing
------------
Contributions are welcome, please send PRs for any enhancements.

46
vendor/github.com/minio/blake2b-simd/compressAvx2_amd64.go generated vendored Normal file
View File

@ -0,0 +1,46 @@
//+build !noasm
//+build !appengine
/*
* Minio Cloud Storage, (C) 2016 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 blake2b
//go:noescape
func compressAVX2Loop(p []uint8, in, iv, t, f, shffle, out []uint64)
func compressAVX2(d *digest, p []uint8) {
in := make([]uint64, 8, 8)
out := make([]uint64, 8, 8)
shffle := make([]uint64, 8, 8)
// vector for PSHUFB instruction
shffle[0] = 0x0201000706050403
shffle[1] = 0x0a09080f0e0d0c0b
shffle[2] = 0x0201000706050403
shffle[3] = 0x0a09080f0e0d0c0b
shffle[4] = 0x0100070605040302
shffle[5] = 0x09080f0e0d0c0b0a
shffle[6] = 0x0100070605040302
shffle[7] = 0x09080f0e0d0c0b0a
in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7] = d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7]
compressAVX2Loop(p, in, iv[:], d.t[:], d.f[:], shffle, out)
d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7] = out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7]
}

671
vendor/github.com/minio/blake2b-simd/compressAvx2_amd64.s generated vendored Normal file
View File

@ -0,0 +1,671 @@
//+build !noasm !appengine
//
// Minio Cloud Storage, (C) 2016 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.
//
//
// Based on AVX2 implementation from https://github.com/sneves/blake2-avx2/blob/master/blake2b-common.h
//
// Use github.com/fwessels/asm2plan9s on this file to assemble instructions to their Plan9 equivalent
//
// Assembly code below essentially follows the ROUND macro (see blake2b-round.h) which is defined as:
// #define ROUND(r) \
// LOAD_MSG_ ##r ##_1(b0, b1); \
// G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// LOAD_MSG_ ##r ##_2(b0, b1); \
// G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
// LOAD_MSG_ ##r ##_3(b0, b1); \
// G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// LOAD_MSG_ ##r ##_4(b0, b1); \
// G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
//
// as well as the go equivalent in https://github.com/dchest/blake2b/blob/master/block.go
//
// As in the macro, G1/G2 in the 1st and 2nd half are identical (so literal copy of assembly)
//
// Rounds are also the same, except for the loading of the message (and rounds 1 & 11 and
// rounds 2 & 12 are identical)
//
#define G1 \
\ // G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1);
BYTE $0xc5; BYTE $0xfd; BYTE $0xd4; BYTE $0xc4 \ // VPADDQ YMM0,YMM0,YMM4 /* v0 += m[0], v1 += m[2], v2 += m[4], v3 += m[6] */
BYTE $0xc5; BYTE $0xfd; BYTE $0xd4; BYTE $0xc1 \ // VPADDQ YMM0,YMM0,YMM1 /* v0 += v4, v1 += v5, v2 += v6, v3 += v7 */
BYTE $0xc5; BYTE $0xe5; BYTE $0xef; BYTE $0xd8 \ // VPXOR YMM3,YMM3,YMM0 /* v12 ^= v0, v13 ^= v1, v14 ^= v2, v15 ^= v3 */
BYTE $0xc5; BYTE $0xfd; BYTE $0x70; BYTE $0xdb; BYTE $0xb1 \ // VPSHUFD YMM3,YMM3,0xb1 /* v12 = v12<<(64-32) | v12>>32, v13 = */
BYTE $0xc5; BYTE $0xed; BYTE $0xd4; BYTE $0xd3 \ // VPADDQ YMM2,YMM2,YMM3 /* v8 += v12, v9 += v13, v10 += v14, v11 += v15 */
BYTE $0xc5; BYTE $0xf5; BYTE $0xef; BYTE $0xca \ // VPXOR YMM1,YMM1,YMM2 /* v4 ^= v8, v5 ^= v9, v6 ^= v10, v7 ^= v11 */
BYTE $0xc4; BYTE $0xe2; BYTE $0x75; BYTE $0x00; BYTE $0xce // VPSHUFB YMM1,YMM1,YMM6 /* v4 = v4<<(64-24) | v4>>24, ..., ..., v7 = v7<<(64-24) | v7>>24 */
#define G2 \
BYTE $0xc5; BYTE $0xfd; BYTE $0xd4; BYTE $0xc5 \ // VPADDQ YMM0,YMM0,YMM5 /* v0 += m[1], v1 += m[3], v2 += m[5], v3 += m[7] */
BYTE $0xc5; BYTE $0xfd; BYTE $0xd4; BYTE $0xc1 \ // VPADDQ YMM0,YMM0,YMM1 /* v0 += v4, v1 += v5, v2 += v6, v3 += v7 */
BYTE $0xc5; BYTE $0xe5; BYTE $0xef; BYTE $0xd8 \ // VPXOR YMM3,YMM3,YMM0 /* v12 ^= v0, v13 ^= v1, v14 ^= v2, v15 ^= v3 */
BYTE $0xc4; BYTE $0xe2; BYTE $0x65; BYTE $0x00; BYTE $0xdf \ // VPSHUFB YMM3,YMM3,YMM7 /* v12 = v12<<(64-16) | v12>>16, ..., ..., v15 = v15<<(64-16) | v15>>16 */
BYTE $0xc5; BYTE $0xed; BYTE $0xd4; BYTE $0xd3 \ // VPADDQ YMM2,YMM2,YMM3 /* v8 += v12, v9 += v13, v10 += v14, v11 += v15 */
BYTE $0xc5; BYTE $0xf5; BYTE $0xef; BYTE $0xca \ // VPXOR YMM1,YMM1,YMM2 /* v4 ^= v8, v5 ^= v9, v6 ^= v10, v7 ^= v11 */
BYTE $0xc5; BYTE $0x75; BYTE $0xd4; BYTE $0xf9 \ // VPADDQ YMM15,YMM1,YMM1 /* temp reg = reg*2 */
BYTE $0xc5; BYTE $0xf5; BYTE $0x73; BYTE $0xd1; BYTE $0x3f \ // VPSRLQ YMM1,YMM1,0x3f /* reg = reg>>63 */
BYTE $0xc4; BYTE $0xc1; BYTE $0x75; BYTE $0xef; BYTE $0xcf // VPXOR YMM1,YMM1,YMM15 /* ORed together: v4 = v4<<(64-63) | v4>>63, v5 = v5<<(64-63) | v5>>63 */
#define DIAGONALIZE \
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb \ // VPERMQ YMM3, YMM3, 0x93
BYTE $0x93 \
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xd2 \ // VPERMQ YMM2, YMM2, 0x4e
BYTE $0x4e \
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 \ // VPERMQ YMM1, YMM1, 0x39
BYTE $0x39 \
// DO NOT DELETE -- macro delimiter (previous line extended)
#define UNDIAGONALIZE \
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb \ // VPERMQ YMM3, YMM3, 0x39
BYTE $0x39 \
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xd2 \ // VPERMQ YMM2, YMM2, 0x4e
BYTE $0x4e \
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 \ // VPERMQ YMM1, YMM1, 0x93
BYTE $0x93 \
// DO NOT DELETE -- macro delimiter (previous line extended)
#define LOAD_SHUFFLE \
MOVQ shffle+120(FP), SI \ // SI: &shuffle
BYTE $0xc5; BYTE $0xfe; BYTE $0x6f; BYTE $0x36 \ // VMOVDQU YMM6, [rsi]
BYTE $0xc5; BYTE $0xfe; BYTE $0x6f; BYTE $0x7e; BYTE $0x20 // VMOVDQU YMM7, 32[rsi]
// func compressAVX2Loop(compressSSE(p []uint8, in, iv, t, f, shffle, out []uint64)
TEXT ·compressAVX2Loop(SB), 7, $0
// REGISTER USE
// Y0 - Y3: v0 - v15
// Y4 - Y5: m[0] - m[7]
// Y6 - Y7: shuffle value
// Y8 - Y9: temp registers
// Y10 -Y13: copy of full message
// Y15: temp register
// Load digest
MOVQ in+24(FP), SI // SI: &in
BYTE $0xc5; BYTE $0xfe; BYTE $0x6f; BYTE $0x06 // VMOVDQU YMM0, [rsi]
BYTE $0xc5; BYTE $0xfe; BYTE $0x6f; BYTE $0x4e; BYTE $0x20 // VMOVDQU YMM1, 32[rsi]
// Already store digest into &out (so we can reload it later generically)
MOVQ out+144(FP), SI // SI: &out
BYTE $0xc5; BYTE $0xfe; BYTE $0x7f; BYTE $0x06 // VMOVDQU [rsi], YMM0
BYTE $0xc5; BYTE $0xfe; BYTE $0x7f; BYTE $0x4e; BYTE $0x20 // VMOVDQU 32[rsi], YMM1
// Initialize message pointer and loop counter
MOVQ message+0(FP), DX // DX: &p (message)
MOVQ message_len+8(FP), R8 // R8: len(message)
SHRQ $7, R8 // len(message) / 128
CMPQ R8, $0
JEQ complete
loop:
// Increment counter
MOVQ t+72(FP), SI // SI: &t
MOVQ 0(SI), R9 //
ADDQ $128, R9 // /* d.t[0] += BlockSize */
MOVQ R9, 0(SI) //
CMPQ R9, $128 // /* if d.t[0] < BlockSize { */
JGE noincr //
MOVQ 8(SI), R9 //
ADDQ $1, R9 // /* d.t[1]++ */
MOVQ R9, 8(SI) //
noincr: // /* } */
// Load initialization vector
MOVQ iv+48(FP), SI // SI: &iv
BYTE $0xc5; BYTE $0xfe; BYTE $0x6f; BYTE $0x16 // VMOVDQU YMM2, [rsi]
BYTE $0xc5; BYTE $0xfe; BYTE $0x6f; BYTE $0x5e; BYTE $0x20 // VMOVDQU YMM3, 32[rsi]
MOVQ t+72(FP), SI // SI: &t
BYTE $0xc4; BYTE $0x63; BYTE $0x3d; BYTE $0x38; BYTE $0x06 // VINSERTI128 YMM8, YMM8, [rsi], 0 /* Y8 = t[0]+t[1] */
BYTE $0x00
MOVQ t+96(FP), SI // SI: &f
BYTE $0xc4; BYTE $0x63; BYTE $0x3d; BYTE $0x38; BYTE $0x06 // VINSERTI128 YMM8, YMM8, [rsi], 1 /* Y8 = t[0]+t[1]+f[0]+f[1] */
BYTE $0x01
BYTE $0xc4; BYTE $0xc1; BYTE $0x65; BYTE $0xef; BYTE $0xd8 // VPXOR YMM3,YMM3,YMM8 /* Y3 = Y3 ^ Y8 */
BYTE $0xc5; BYTE $0x7e; BYTE $0x6f; BYTE $0x12 // VMOVDQU YMM10, [rdx] /* Y10 = m[0]+ m[1]+ m[2]+ m[3] */
BYTE $0xc5; BYTE $0x7e; BYTE $0x6f; BYTE $0x5a; BYTE $0x20 // VMOVDQU YMM11, 32[rdx] /* Y11 = m[4]+ m[5]+ m[6]+ m[7] */
BYTE $0xc5; BYTE $0x7e; BYTE $0x6f; BYTE $0x62; BYTE $0x40 // VMOVDQU YMM12, 64[rdx] /* Y12 = m[8]+ m[9]+m[10]+m[11] */
BYTE $0xc5; BYTE $0x7e; BYTE $0x6f; BYTE $0x6a; BYTE $0x60 // VMOVDQU YMM13, 96[rdx] /* Y13 = m[12]+m[13]+m[14]+m[15] */
LOAD_SHUFFLE
///////////////////////////////////////////////////////////////////////////
// R O U N D 1
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0xc1; BYTE $0x2d; BYTE $0x6c; BYTE $0xe3 // VPUNPCKLQDQ YMM4, YMM10, YMM11 /* m[0], m[4], m[2], m[6] */
BYTE $0xc4; BYTE $0xc1; BYTE $0x2d; BYTE $0x6d; BYTE $0xeb // VPUNPCKHQDQ YMM5, YMM10, YMM11 /* m[1], m[5], m[3], m[7] */
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xe4 // VPERMQ YMM4, YMM4, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xed // VPERMQ YMM5, YMM5, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0xc1; BYTE $0x1d; BYTE $0x6c; BYTE $0xe5 // VPUNPCKLQDQ YMM4, YMM12, YMM13 /* m[8], m[12], m[10], m[14] */
BYTE $0xc4; BYTE $0xc1; BYTE $0x1d; BYTE $0x6d; BYTE $0xed // VPUNPCKHQDQ YMM5, YMM12, YMM13 /* m[9], m[13], m[11], m[15] */
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xe4 // VPERMQ YMM4, YMM4, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xed // VPERMQ YMM5, YMM5, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 2
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xc5 // VPUNPCKLQDQ YMM8, YMM11, YMM13 /* m[4], ____, ____, m[14] */
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x03 /* m[14], m[4], ____, ____ */ /* xxxx 0011 = 0x03 */
BYTE $0x03
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ YMM9, YMM12, YMM13 /* m[9], m[13], ____, ____ */
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x20 /* m[9], m[13], ____, ____ */ /* 0010 0000 = 0x20 */
BYTE $0x20
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc4 // VPERMQ YMM8, YMM12, 0x02 /* m[10], m[8], ____, ____ */ /* xxxx 0010 = 0x02 */
BYTE $0x02
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcd // VPERMQ YMM9, YMM13, 0x30 /* ____, ____, m[15], ____ */ /* xx11 xxxx = 0x30 */
BYTE $0x30
BYTE $0xc4; BYTE $0x41; BYTE $0x35; BYTE $0x6c; BYTE $0xcb // VPUNPCKLQDQ YMM9, YMM9, YMM11 /* ____, ____, m[15], m[6] */
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30 /* m[9], m[13], m[15], m[6] */ /* 0011 0000 = 0x30 */
BYTE $0x30
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc2 // VPERMQ YMM8, YMM10, 0x01 /* m[1], m[0], ____, ____ */ /* xxxx 0001 = 0x01 */
BYTE $0x01
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xcc // VPUNPCKHQDQ YMM9, YMM11, YMM12 /* m[5], ____, ____, m[11] */
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x03 /* m[11], m[5], ____, ____ */ /* xxxx 0011 = 0x03 */
BYTE $0x03
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x20 /* m[1], m[0], m[11], m[5] */ /* 0010 0000 = 0x20 */
BYTE $0x20
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc5 // VPUNPCKLQDQ YMM8, YMM10, YMM13 /* ___, m[12], m[2], ____ */
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x09 /* m[12], m[2], ____, ____ */ /* xxxx 1001 = 0x09 */
BYTE $0x09
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xca // VPUNPCKHQDQ YMM9, YMM11, YMM10 /* ____, ____, m[7], m[3] */
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30 /* m[9], m[13], m[15], m[6] */ /* 0011 0000 = 0x30 */
BYTE $0x30
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 3
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc5 // VPERMQ YMM8, YMM13, 0x00
BYTE $0x00
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xc0 // VPUNPCKHQDQ YMM8, YMM12, YMM8
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ YMM9, YMM11, YMM13
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x21
BYTE $0x21
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6c; BYTE $0xc2 // VPUNPCKLQDQ YMM8, YMM12, YMM10
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcd // VPERMQ YMM9, YMM13, 0x55
BYTE $0x55
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM10, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30
BYTE $0x30
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc2 // VPERMQ YMM8, YMM10, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ YMM8, YMM12, YMM8
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xcc // VPUNPCKHQDQ YMM9, YMM11, YMM12
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x31
BYTE $0x31
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xc3 // VPUNPCKLQDQ YMM8, YMM13, YMM11
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcb // VPERMQ YMM9, YMM11, 0x00
BYTE $0x00
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6d; BYTE $0xc9 // VPUNPCKHQDQ YMM9, YMM10, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x21
BYTE $0x21
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 4
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xc2 // VPUNPCKHQDQ YMM8, YMM11, YMM10
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6d; BYTE $0xcc // VPUNPCKHQDQ YMM9, YMM13, YMM12
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x21
BYTE $0x21
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xc2 // VPUNPCKHQDQ YMM8, YMM12, YMM10
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcd // VPERMQ YMM9, YMM13, 0x08
BYTE $0x08
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x20
BYTE $0x20
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc3 // VPERMQ YMM8, YMM11, 0x55
BYTE $0x55
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ YMM8, YMM10, YMM8
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcd // VPERMQ YMM9, YMM13, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM11, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x21
BYTE $0x21
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xc4 // VPUNPCKLQDQ YMM8, YMM11, YMM12
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xcc // VPUNPCKLQDQ YMM9, YMM10, YMM12
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x21
BYTE $0x21
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 5
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xc3 // VPUNPCKHQDQ YMM8, YMM12, YMM11
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xcc // VPUNPCKLQDQ YMM9, YMM10, YMM12
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x30
BYTE $0x30
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc3 // VPERMQ YMM8, YMM11, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ YMM8, YMM10, YMM8
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcd // VPERMQ YMM9, YMM13, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM11, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x20
BYTE $0x20
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc4 // VPERMQ YMM8, YMM12, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ YMM8, YMM13, YMM8
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xca // VPERMQ YMM9, YMM10, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM11, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x31
BYTE $0x31
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc5 // VPERMQ YMM8, YMM13, 0x00
BYTE $0x00
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6d; BYTE $0xc0 // VPUNPCKHQDQ YMM8, YMM10, YMM8
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcd // VPERMQ YMM9, YMM13, 0x55
BYTE $0x55
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM12, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x20
BYTE $0x20
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 6
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc3 // VPUNPCKLQDQ YMM8, YMM10, YMM11
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xcc // VPUNPCKLQDQ YMM9, YMM10, YMM12
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x21
BYTE $0x21
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xc4 // VPUNPCKLQDQ YMM8, YMM13, YMM12
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xca // VPUNPCKHQDQ YMM9, YMM12, YMM10
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30
BYTE $0x30
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc3 // VPERMQ YMM8, YMM11, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6d; BYTE $0xca // VPUNPCKHQDQ YMM9, YMM13, YMM10
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x30
BYTE $0x30
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6d; BYTE $0xc3 // VPUNPCKHQDQ YMM8, YMM13, YMM11
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcc // VPERMQ YMM9, YMM12, 0x55
BYTE $0x55
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM13, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30
BYTE $0x30
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 7
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc2 // VPERMQ YMM8, YMM10, 0x55
BYTE $0x55
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ YMM8, YMM13, YMM8
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xcb // VPUNPCKLQDQ YMM9, YMM13, YMM11
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x30
BYTE $0x30
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xc5 // VPUNPCKHQDQ YMM8, YMM11, YMM13
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcc // VPERMQ YMM9, YMM12, 0xaa
BYTE $0xaa
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6d; BYTE $0xc9 // VPUNPCKHQDQ YMM9, YMM13, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x20
BYTE $0x20
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc3 // VPUNPCKLQDQ YMM8, YMM10, YMM11
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcc // VPERMQ YMM9, YMM12, 0x01
BYTE $0x01
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x20
BYTE $0x20
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xc2 // VPUNPCKHQDQ YMM8, YMM11, YMM10
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcc // VPERMQ YMM9, YMM12, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM10, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x31
BYTE $0x31
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 8
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6d; BYTE $0xc3 // VPUNPCKHQDQ YMM8, YMM13, YMM11
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xca // VPERMQ YMM9, YMM10, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xc9 // VPUNPCKLQDQ YMM9, YMM13, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x20
BYTE $0x20
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc5 // VPERMQ YMM8, YMM13, 0xaa
BYTE $0xaa
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xc0 // VPUNPCKHQDQ YMM8, YMM12, YMM8
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6d; BYTE $0xcc // VPUNPCKHQDQ YMM9, YMM10, YMM12
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x21
BYTE $0x21
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xc5 // VPUNPCKHQDQ YMM8, YMM11, YMM13
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6c; BYTE $0xca // VPUNPCKLQDQ YMM9, YMM12, YMM10
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x0c
BYTE $0x0c
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x20
BYTE $0x20
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc3 // VPUNPCKLQDQ YMM8, YMM10, YMM11
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xcc // VPUNPCKLQDQ YMM9, YMM11, YMM12
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30
BYTE $0x30
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 9
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xc5 // VPUNPCKLQDQ YMM8, YMM11, YMM13
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xca // VPERMQ YMM9, YMM10, 0x00
BYTE $0x00
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xc9 // VPUNPCKHQDQ YMM9, YMM12, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x31
BYTE $0x31
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6d; BYTE $0xc4 // VPUNPCKHQDQ YMM8, YMM13, YMM12
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcc // VPERMQ YMM9, YMM12, 0x00
BYTE $0x00
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6d; BYTE $0xc9 // VPUNPCKHQDQ YMM9, YMM10, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x31
BYTE $0x31
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcc // VPERMQ YMM9, YMM12, 0xaa
BYTE $0xaa
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6d; BYTE $0xc9 // VPUNPCKHQDQ YMM9, YMM10, YMM9
BYTE $0xc4; BYTE $0xc3; BYTE $0x15; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM13, YMM9, 0x20
BYTE $0x20
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc3 // VPERMQ YMM8, YMM11, 0xff
BYTE $0xff
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ YMM8, YMM10, YMM8
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcb // VPERMQ YMM9, YMM11, 0x04
BYTE $0x04
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x21
BYTE $0x21
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 10
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc4 // VPERMQ YMM8, YMM12, 0x20
BYTE $0x20
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xca // VPUNPCKHQDQ YMM9, YMM11, YMM10
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x31
BYTE $0x31
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc3 // VPUNPCKLQDQ YMM8, YMM10, YMM11
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcb // VPERMQ YMM9, YMM11, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x31
BYTE $0x31
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6d; BYTE $0xc4 // VPUNPCKHQDQ YMM8, YMM13, YMM12
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ YMM9, YMM10, YMM13
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x60
BYTE $0x60
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x31
BYTE $0x31
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc5 // VPERMQ YMM8, YMM13, 0xaa
BYTE $0xaa
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xc0 // VPUNPCKHQDQ YMM8, YMM12, YMM8
BYTE $0xc4; BYTE $0x41; BYTE $0x15; BYTE $0x6c; BYTE $0xca // VPUNPCKLQDQ YMM9, YMM13, YMM10
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x21
BYTE $0x21
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 1 1
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0xc1; BYTE $0x2d; BYTE $0x6c; BYTE $0xe3 // VPUNPCKLQDQ YMM4, YMM10, YMM11 /* m[0], m[4], m[2], m[6] */
BYTE $0xc4; BYTE $0xc1; BYTE $0x2d; BYTE $0x6d; BYTE $0xeb // VPUNPCKHQDQ YMM5, YMM10, YMM11 /* m[1], m[5], m[3], m[7] */
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xe4 // VPERMQ YMM4, YMM4, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xed // VPERMQ YMM5, YMM5, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0xc1; BYTE $0x1d; BYTE $0x6c; BYTE $0xe5 // VPUNPCKLQDQ YMM4, YMM12, YMM13 /* m[8], m[12], m[10], m[14] */
BYTE $0xc4; BYTE $0xc1; BYTE $0x1d; BYTE $0x6d; BYTE $0xed // VPUNPCKHQDQ YMM5, YMM12, YMM13 /* m[9], m[13], m[11], m[15] */
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xe4 // VPERMQ YMM4, YMM4, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xed // VPERMQ YMM5, YMM5, 0xd8 /* 0x1101 1000 = 0xd8 */
BYTE $0xd8
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 1 2
///////////////////////////////////////////////////////////////////////////
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6c; BYTE $0xc5 // VPUNPCKLQDQ YMM8, YMM11, YMM13 /* m[4], ____, ____, m[14] */
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x03 /* m[14], m[4], ____, ____ */ /* xxxx 0011 = 0x03 */
BYTE $0x03
BYTE $0xc4; BYTE $0x41; BYTE $0x1d; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ YMM9, YMM12, YMM13 /* m[9], m[13], ____, ____ */
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x20 /* m[9], m[13], ____, ____ */ /* 0010 0000 = 0x20 */
BYTE $0x20
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc4 // VPERMQ YMM8, YMM12, 0x02 /* m[10], m[8], ____, ____ */ /* xxxx 0010 = 0x02 */
BYTE $0x02
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xcd // VPERMQ YMM9, YMM13, 0x30 /* ____, ____, m[15], ____ */ /* xx11 xxxx = 0x30 */
BYTE $0x30
BYTE $0xc4; BYTE $0x41; BYTE $0x35; BYTE $0x6c; BYTE $0xcb // VPUNPCKLQDQ YMM9, YMM9, YMM11 /* ____, ____, m[15], m[6] */
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30 /* m[9], m[13], m[15], m[6] */ /* 0011 0000 = 0x30 */
BYTE $0x30
G1
G2
DIAGONALIZE
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc2 // VPERMQ YMM8, YMM10, 0x01 /* m[1], m[0], ____, ____ */ /* xxxx 0001 = 0x01 */
BYTE $0x01
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xcc // VPUNPCKHQDQ YMM9, YMM11, YMM12 /* m[5], ____, ____, m[11] */
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc9 // VPERMQ YMM9, YMM9, 0x03 /* m[11], m[5], ____, ____ */ /* xxxx 0011 = 0x03 */
BYTE $0x03
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe1 // VPERM2I128 YMM4, YMM8, YMM9, 0x20 /* m[1], m[0], m[11], m[5] */ /* 0010 0000 = 0x20 */
BYTE $0x20
BYTE $0xc4; BYTE $0x41; BYTE $0x2d; BYTE $0x6c; BYTE $0xc5 // VPUNPCKLQDQ YMM8, YMM10, YMM13 /* ___, m[12], m[2], ____ */
BYTE $0xc4; BYTE $0x43; BYTE $0xfd; BYTE $0x00; BYTE $0xc0 // VPERMQ YMM8, YMM8, 0x09 /* m[12], m[2], ____, ____ */ /* xxxx 1001 = 0x09 */
BYTE $0x09
BYTE $0xc4; BYTE $0x41; BYTE $0x25; BYTE $0x6d; BYTE $0xca // VPUNPCKHQDQ YMM9, YMM11, YMM10 /* ____, ____, m[7], m[3] */
BYTE $0xc4; BYTE $0xc3; BYTE $0x3d; BYTE $0x46; BYTE $0xe9 // VPERM2I128 YMM5, YMM8, YMM9, 0x30 /* m[9], m[13], m[15], m[6] */ /* 0011 0000 = 0x30 */
BYTE $0x30
G1
G2
UNDIAGONALIZE
// Reload digest (most current value store in &out)
MOVQ out+144(FP), SI // SI: &in
BYTE $0xc5; BYTE $0x7e; BYTE $0x6f; BYTE $0x26 // VMOVDQU YMM12, [rsi]
BYTE $0xc5; BYTE $0x7e; BYTE $0x6f; BYTE $0x6e; BYTE $0x20 // VMOVDQU YMM13, 32[rsi]
BYTE $0xc5; BYTE $0xfd; BYTE $0xef; BYTE $0xc2 // VPXOR YMM0,YMM0,YMM2 /* X0 = X0 ^ X4, X1 = X1 ^ X5 */
BYTE $0xc4; BYTE $0xc1; BYTE $0x7d; BYTE $0xef; BYTE $0xc4 // VPXOR YMM0,YMM0,YMM12 /* X0 = X0 ^ X12, X1 = X1 ^ X13 */
BYTE $0xc5; BYTE $0xf5; BYTE $0xef; BYTE $0xcb // VPXOR YMM1,YMM1,YMM3 /* X2 = X2 ^ X6, X3 = X3 ^ X7 */
BYTE $0xc4; BYTE $0xc1; BYTE $0x75; BYTE $0xef; BYTE $0xcd // VPXOR YMM1,YMM1,YMM13 /* X2 = X2 ^ X14, X3 = X3 ^ X15 */
// Store digest into &out
MOVQ out+144(FP), SI // SI: &out
BYTE $0xc5; BYTE $0xfe; BYTE $0x7f; BYTE $0x06 // VMOVDQU [rsi], YMM0
BYTE $0xc5; BYTE $0xfe; BYTE $0x7f; BYTE $0x4e; BYTE $0x20 // VMOVDQU 32[rsi], YMM1
// Increment message pointer and check if there's more to do
ADDQ $128, DX // message += 128
SUBQ $1, R8
JNZ loop
complete:
BYTE $0xc5; BYTE $0xf8; BYTE $0x77 // VZEROUPPER /* Prevent further context switches */
RET

40
vendor/github.com/minio/blake2b-simd/compressAvx_amd64.go generated vendored Normal file
View File

@ -0,0 +1,40 @@
//+build !noasm
//+build !appengine
/*
* Minio Cloud Storage, (C) 2016 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 blake2b
//go:noescape
func blockAVXLoop(p []uint8, in, iv, t, f, shffle, out []uint64)
func compressAVX(d *digest, p []uint8) {
in := make([]uint64, 8, 8)
out := make([]uint64, 8, 8)
shffle := make([]uint64, 2, 2)
// vector for PSHUFB instruction
shffle[0] = 0x0201000706050403
shffle[1] = 0x0a09080f0e0d0c0b
in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7] = d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7]
blockAVXLoop(p, in, iv[:], d.t[:], d.f[:], shffle, out)
d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7] = out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7]
}

140
vendor/github.com/minio/blake2b-simd/compress_amd64.s → vendor/github.com/minio/blake2b-simd/compressAvx_amd64.s generated vendored
View File

@ -119,10 +119,12 @@
MOVQ shffle+120(FP), SI \ // SI: &shuffle MOVQ shffle+120(FP), SI \ // SI: &shuffle
MOVOU 0(SI), X12 // X12 = 03040506 07000102 0b0c0d0e 0f08090a MOVOU 0(SI), X12 // X12 = 03040506 07000102 0b0c0d0e 0f08090a
// func blockAVXLoop(p []uint8, in, iv, t, f, shffle, out []uint64)
// func blockAVX(p []uint8, in, iv, t, f, shffle, out []uint64) TEXT ·blockAVXLoop(SB), 7, $0
TEXT ·blockAVX(SB), 7, $0
// REGISTER USE // REGISTER USE
// R8: loop counter
// DX: message pointer
// SI: temp pointer for loading
// X0 - X7: v0 - v15 // X0 - X7: v0 - v15
// X8 - X11: m[0] - m[7] // X8 - X11: m[0] - m[7]
// X12: shuffle value // X12: shuffle value
@ -135,16 +137,43 @@ TEXT ·blockAVX(SB), 7, $0
MOVOU 32(SI), X2 // X2 = in[4]+in[5] /* row2l = LOAD( &S->h[4] ); */ MOVOU 32(SI), X2 // X2 = in[4]+in[5] /* row2l = LOAD( &S->h[4] ); */
MOVOU 48(SI), X3 // X3 = in[6]+in[7] /* row2h = LOAD( &S->h[6] ); */ MOVOU 48(SI), X3 // X3 = in[6]+in[7] /* row2h = LOAD( &S->h[6] ); */
// Load initialization vector // Already store digest into &out (so we can reload it later generically)
MOVQ iv+48(FP), DX // DX: &iv MOVQ out+144(FP), SI // SI: &out
MOVOU 0(DX), X4 // X4 = iv[0]+iv[1] /* row3l = LOAD( &blake2b_IV[0] ); */ MOVOU X0, 0(SI) // out[0]+out[1] = X0
MOVOU 16(DX), X5 // X5 = iv[2]+iv[3] /* row3h = LOAD( &blake2b_IV[2] ); */ MOVOU X1, 16(SI) // out[2]+out[3] = X1
MOVOU X2, 32(SI) // out[4]+out[5] = X2
MOVOU X3, 48(SI) // out[6]+out[7] = X3
// Initialize message pointer and loop counter
MOVQ message+0(FP), DX // DX: &p (message)
MOVQ message_len+8(FP), R8 // R8: len(message)
SHRQ $7, R8 // len(message) / 128
CMPQ R8, $0
JEQ complete
loop:
// Increment counter
MOVQ t+72(FP), SI // SI: &t MOVQ t+72(FP), SI // SI: &t
MOVOU 32(DX), X6 // X6 = iv[4]+iv[5] /* LOAD( &blake2b_IV[4] ) */ MOVQ 0(SI), R9 //
MOVOU 0(SI), X7 // X7 = t[0]+t[1] /* LOAD( &S->t[0] ) */ ADDQ $128, R9 // /* d.t[0] += BlockSize */
PXOR X7, X6 // X6 = X6 ^ X7 /* row4l = _mm_xor_si128( , ); */ MOVQ R9, 0(SI) //
CMPQ R9, $128 // /* if d.t[0] < BlockSize { */
JGE noincr //
MOVQ 8(SI), R9 //
ADDQ $1, R9 // /* d.t[1]++ */
MOVQ R9, 8(SI) //
noincr: // /* } */
// Load initialization vector
MOVQ iv+48(FP), SI // SI: &iv
MOVOU 0(SI), X4 // X4 = iv[0]+iv[1] /* row3l = LOAD( &blake2b_IV[0] ); */
MOVOU 16(SI), X5 // X5 = iv[2]+iv[3] /* row3h = LOAD( &blake2b_IV[2] ); */
MOVOU 32(SI), X6 // X6 = iv[4]+iv[5] /* LOAD( &blake2b_IV[4] ) */
MOVOU 48(SI), X7 // X7 = iv[6]+iv[7] /* LOAD( &blake2b_IV[6] ) */
MOVQ t+72(FP), SI // SI: &t
MOVOU 0(SI), X8 // X8 = t[0]+t[1] /* LOAD( &S->t[0] ) */
PXOR X8, X6 // X6 = X6 ^ X8 /* row4l = _mm_xor_si128( , ); */
MOVQ t+96(FP), SI // SI: &f MOVQ t+96(FP), SI // SI: &f
MOVOU 48(DX), X7 // X7 = iv[6]+iv[7] /* LOAD( &blake2b_IV[6] ) */
MOVOU 0(SI), X8 // X8 = f[0]+f[1] /* LOAD( &S->f[0] ) */ MOVOU 0(SI), X8 // X8 = f[0]+f[1] /* LOAD( &S->f[0] ) */
PXOR X8, X7 // X7 = X7 ^ X8 /* row4h = _mm_xor_si128( , ); */ PXOR X8, X7 // X7 = X7 ^ X8 /* row4h = _mm_xor_si128( , ); */
@ -155,7 +184,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+m[1] MOVOU 0(DX), X12 // X12 = m[0]+m[1]
MOVOU 16(DX), X13 // X13 = m[2]+m[3] MOVOU 16(DX), X13 // X13 = m[2]+m[3]
MOVOU 32(DX), X14 // X14 = m[4]+m[5] MOVOU 32(DX), X14 // X14 = m[4]+m[5]
@ -175,7 +203,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 64(DX), X12 // X12 = m[8]+ m[9] MOVOU 64(DX), X12 // X12 = m[8]+ m[9]
MOVOU 80(DX), X13 // X13 = m[10]+m[11] MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 96(DX), X14 // X14 = m[12]+m[13] MOVOU 96(DX), X14 // X14 = m[12]+m[13]
@ -199,7 +226,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 112(DX), X12 // X12 = m[14]+m[15] MOVOU 112(DX), X12 // X12 = m[14]+m[15]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
@ -209,7 +235,8 @@ TEXT ·blockAVX(SB), 7, $0
MOVOU 80(DX), X13 // X13 = m[10]+m[11] MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 48(DX), X15 // X15 = m[6]+ m[7] MOVOU 48(DX), X15 // X15 = m[6]+ m[7]
BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xd6 // VPUNPCKLQDQ XMM10, XMM13, XMM14 /* m[10], m[8] */ BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xd6 // VPUNPCKLQDQ XMM10, XMM13, XMM14 /* m[10], m[8] */
BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xdc; BYTE $0x08// VPALIGNR XMM11, XMM15, XMM12, 0x8 /* m[15], m[6] */ BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xdc // VPALIGNR XMM11, XMM15, XMM12, 0x8 /* m[15], m[6] */
BYTE $0x08
LOAD_SHUFFLE LOAD_SHUFFLE
@ -221,11 +248,11 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xc4; BYTE $0x08 // VPALIGNR XMM8, XMM12, XMM12, 0x8 /* m[1], m[0] */ BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xc4 // VPALIGNR XMM8, XMM12, XMM12, 0x8 /* m[1], m[0] */
BYTE $0x08
BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ XMM9, XMM14, XMM13 /* m[11], m[5] */ BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ XMM9, XMM14, XMM13 /* m[11], m[5] */
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
@ -247,12 +274,12 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 32(DX), X12 // X12 = m[4]+ m[5] MOVOU 32(DX), X12 // X12 = m[4]+ m[5]
MOVOU 80(DX), X13 // X13 = m[10]+m[11] MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 96(DX), X14 // X14 = m[12]+m[13] MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU 112(DX), X15 // X15 = m[14]+m[15] MOVOU 112(DX), X15 // X15 = m[14]+m[15]
BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xc5; BYTE $0x08// VPALIGNR XMM8, XMM14, XMM13, 0x8 /* m[11], m[12] */ BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xc5 // VPALIGNR XMM8, XMM14, XMM13, 0x8 /* m[11], m[12] */
BYTE $0x08
BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6d; BYTE $0xcf // VPUNPCKHQDQ XMM9, XMM12, XMM15 /* m[5], m[15] */ BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6d; BYTE $0xcf // VPUNPCKHQDQ XMM9, XMM12, XMM15 /* m[5], m[15] */
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 16(DX), X13 // X13 = m[2]+ m[3] MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
@ -271,7 +298,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
@ -283,7 +309,8 @@ TEXT ·blockAVX(SB), 7, $0
MOVOU 32(DX), X14 // X14 = m[4]+ m[5] MOVOU 32(DX), X14 // X14 = m[4]+ m[5]
MOVOU 112(DX), X15 // X15 = m[14]+m[15] MOVOU 112(DX), X15 // X15 = m[14]+m[15]
BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6c; BYTE $0xd5 // VPUNPCKLQDQ XMM10, XMM15, XMM13 /* m[14], m[6] */ BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6c; BYTE $0xd5 // VPUNPCKLQDQ XMM10, XMM15, XMM13 /* m[14], m[6] */
BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xdc; BYTE $0x08// VPALIGNR XMM11, XMM14, XMM12, 0x8 /* m[1], m[4] */ BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xdc // VPALIGNR XMM11, XMM14, XMM12, 0x8 /* m[1], m[4] */
BYTE $0x08
LOAD_SHUFFLE LOAD_SHUFFLE
@ -299,7 +326,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
@ -322,7 +348,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
@ -351,7 +376,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
@ -376,7 +400,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
@ -388,7 +411,8 @@ TEXT ·blockAVX(SB), 7, $0
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 64(DX), X13 // X13 = m[8]+ m[9] MOVOU 64(DX), X13 // X13 = m[8]+ m[9]
MOVOU 96(DX), X14 // X14 = m[12]+m[13] MOVOU 96(DX), X14 // X14 = m[12]+m[13]
BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xd4; BYTE $0x08// VPALIGNR XMM10, XMM14, XMM12, 0x8 /* m[1], m[12] */ BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xd4 // VPALIGNR XMM10, XMM14, XMM12, 0x8 /* m[1], m[12] */
BYTE $0x08
BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6d; BYTE $0xde // VPUNPCKHQDQ XMM11, XMM14, XMM14 /* ___, m[13] */ BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6d; BYTE $0xde // VPUNPCKHQDQ XMM11, XMM14, XMM14 /* ___, m[13] */
BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xdb // VPUNPCKLQDQ XMM11, XMM13, XMM11 /* m[8], ____ */ BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xdb // VPUNPCKLQDQ XMM11, XMM13, XMM11 /* m[8], ____ */
@ -406,7 +430,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 16(DX), X13 // X13 = m[2]+ m[3] MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
MOVOU 48(DX), X14 // X14 = m[6]+ m[7] MOVOU 48(DX), X14 // X14 = m[6]+ m[7]
@ -428,7 +451,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 48(DX), X14 // X14 = m[6]+ m[7] MOVOU 48(DX), X14 // X14 = m[6]+ m[7]
@ -456,7 +478,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 96(DX), X14 // X14 = m[12]+m[13] MOVOU 96(DX), X14 // X14 = m[12]+m[13]
@ -466,7 +487,8 @@ TEXT ·blockAVX(SB), 7, $0
BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6c; BYTE $0xcd // VPUNPCKLQDQ XMM9, XMM15, XMM13 /* m[14], m[4] */ BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6c; BYTE $0xcd // VPUNPCKLQDQ XMM9, XMM15, XMM13 /* m[14], m[4] */
MOVOU 80(DX), X12 // X12 = m[10]+m[11] MOVOU 80(DX), X12 // X12 = m[10]+m[11]
BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6d; BYTE $0xd7 // VPUNPCKHQDQ XMM10, XMM13, XMM15 /* m[5], m[15] */ BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6d; BYTE $0xd7 // VPUNPCKHQDQ XMM10, XMM13, XMM15 /* m[5], m[15] */
BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xde; BYTE $0x08// VPALIGNR XMM11, XMM12, XMM14, 0x8 /* m[13], m[10] */ BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xde // VPALIGNR XMM11, XMM12, XMM14, 0x8 /* m[13], m[10] */
BYTE $0x08
LOAD_SHUFFLE LOAD_SHUFFLE
@ -478,13 +500,13 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 80(DX), X15 // X15 = m[10]+m[11] MOVOU 80(DX), X15 // X15 = m[10]+m[11]
BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6c; BYTE $0xc5 // VPUNPCKLQDQ XMM8, XMM12, XMM13 /* m[0], m[6] */ BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6c; BYTE $0xc5 // VPUNPCKLQDQ XMM8, XMM12, XMM13 /* m[0], m[6] */
BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xce; BYTE $0x08// VPALIGNR XMM9, XMM14, XMM14, 0x8 /* m[9], m[8] */ BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xce // VPALIGNR XMM9, XMM14, XMM14, 0x8 /* m[9], m[8] */
BYTE $0x08
MOVOU 16(DX), X14 // X14 = m[2]+ m[3] MOVOU 16(DX), X14 // X14 = m[2]+ m[3]
BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6d; BYTE $0xd6 // VPUNPCKHQDQ XMM10, XMM13, XMM14 /* m[7], m[3] */ BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6d; BYTE $0xd6 // VPUNPCKHQDQ XMM10, XMM13, XMM14 /* m[7], m[3] */
BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xdf // VPUNPCKHQDQ XMM11, XMM15, XMM15 /* ___, m[11] */ BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xdf // VPUNPCKHQDQ XMM11, XMM15, XMM15 /* ___, m[11] */
@ -504,7 +526,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 96(DX), X14 // X14 = m[12]+m[13] MOVOU 96(DX), X14 // X14 = m[12]+m[13]
@ -515,7 +536,8 @@ TEXT ·blockAVX(SB), 7, $0
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 64(DX), X13 // X13 = m[8]+ m[9] MOVOU 64(DX), X13 // X13 = m[8]+ m[9]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xd6; BYTE $0x08// VPALIGNR XMM10, XMM15, XMM14, 0x8 /* m[11], m[14] */ BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xd6 // VPALIGNR XMM10, XMM15, XMM14, 0x8 /* m[11], m[14] */
BYTE $0x08
BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6d; BYTE $0xdd // VPUNPCKHQDQ XMM11, XMM12, XMM13 /* m[1], m[9] */ BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6d; BYTE $0xdd // VPUNPCKHQDQ XMM11, XMM12, XMM13 /* m[1], m[9] */
LOAD_SHUFFLE LOAD_SHUFFLE
@ -528,7 +550,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
@ -555,17 +576,18 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU 112(DX), X15 // X15 = m[14]+m[15] MOVOU 112(DX), X15 // X15 = m[14]+m[15]
BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xc7 // VPUNPCKLQDQ XMM8, XMM13, XMM15 /* m[6], m[14] */ BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xc7 // VPUNPCKLQDQ XMM8, XMM13, XMM15 /* m[6], m[14] */
BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xce; BYTE $0x08// VPALIGNR XMM9, XMM12, XMM14, 0x8 /* m[11], m[0] */ BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xce // VPALIGNR XMM9, XMM12, XMM14, 0x8 /* m[11], m[0] */
BYTE $0x08
MOVOU 16(DX), X13 // X13 = m[2]+ m[3] MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xd6 // VPUNPCKHQDQ XMM10, XMM15, XMM14 /* m[15], m[9] */ BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xd6 // VPUNPCKHQDQ XMM10, XMM15, XMM14 /* m[15], m[9] */
BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xdd; BYTE $0x08// VPALIGNR XMM11, XMM14, XMM13, 0x8 /* m[3], m[8] */ BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xdd // VPALIGNR XMM11, XMM14, XMM13, 0x8 /* m[3], m[8] */
BYTE $0x08
LOAD_SHUFFLE LOAD_SHUFFLE
@ -577,14 +599,14 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 16(DX), X13 // X13 = m[2]+ m[3] MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU 96(DX), X15 // X15 = m[12]+m[13] MOVOU 96(DX), X15 // X15 = m[12]+m[13]
BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xc7 // VPUNPCKHQDQ XMM8, XMM15, XMM15 /* ___, m[13] */ BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xc7 // VPUNPCKHQDQ XMM8, XMM15, XMM15 /* ___, m[13] */
BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ XMM8, XMM15, XMM8 /* m[12], ____ */ BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6c; BYTE $0xc0 // VPUNPCKLQDQ XMM8, XMM15, XMM8 /* m[12], ____ */
BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xcc; BYTE $0x08// VPALIGNR XMM9, XMM14, XMM12, 0x8 /* m[1], m[10] */ BYTE $0xc4; BYTE $0x43; BYTE $0x09; BYTE $0x0f; BYTE $0xcc // VPALIGNR XMM9, XMM14, XMM12, 0x8 /* m[1], m[10] */
BYTE $0x08
MOVOU 32(DX), X12 // X12 = m[4]+ m[5] MOVOU 32(DX), X12 // X12 = m[4]+ m[5]
MOVOU 48(DX), X15 // X15 = m[6]+ m[7] MOVOU 48(DX), X15 // X15 = m[6]+ m[7]
BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xd7 // VPUNPCKHQDQ XMM10, XMM15, XMM15 /* ___, m[7] */ BYTE $0xc4; BYTE $0x41; BYTE $0x01; BYTE $0x6d; BYTE $0xd7 // VPUNPCKHQDQ XMM10, XMM15, XMM15 /* ___, m[7] */
@ -606,7 +628,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
@ -629,7 +650,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 64(DX), X13 // X13 = m[8]+ m[9] MOVOU 64(DX), X13 // X13 = m[8]+ m[9]
MOVOU 96(DX), X14 // X14 = m[12]+m[13] MOVOU 96(DX), X14 // X14 = m[12]+m[13]
@ -638,7 +658,8 @@ TEXT ·blockAVX(SB), 7, $0
BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6d; BYTE $0xce // VPUNPCKHQDQ XMM9, XMM12, XMM14 /* m[3], m[13] */ BYTE $0xc4; BYTE $0x41; BYTE $0x19; BYTE $0x6d; BYTE $0xce // VPUNPCKHQDQ XMM9, XMM12, XMM14 /* m[3], m[13] */
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 80(DX), X13 // X13 = m[10]+m[11] MOVOU 80(DX), X13 // X13 = m[10]+m[11]
BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xd5; BYTE $0x08// VPALIGNR XMM10, XMM15, XMM13, 0x8 /* m[11], m[14] */ BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xd5 // VPALIGNR XMM10, XMM15, XMM13, 0x8 /* m[11], m[14] */
BYTE $0x08
BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6c; BYTE $0xdc // VPUNPCKLQDQ XMM11, XMM14, XMM12 /* m[12], m[0] */ BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6c; BYTE $0xdc // VPUNPCKLQDQ XMM11, XMM14, XMM12 /* m[12], m[0] */
LOAD_SHUFFLE LOAD_SHUFFLE
@ -655,7 +676,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+m[1] MOVOU 0(DX), X12 // X12 = m[0]+m[1]
MOVOU 16(DX), X13 // X13 = m[2]+m[3] MOVOU 16(DX), X13 // X13 = m[2]+m[3]
MOVOU 32(DX), X14 // X14 = m[4]+m[5] MOVOU 32(DX), X14 // X14 = m[4]+m[5]
@ -675,7 +695,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 64(DX), X12 // X12 = m[8]+ m[9] MOVOU 64(DX), X12 // X12 = m[8]+ m[9]
MOVOU 80(DX), X13 // X13 = m[10]+m[11] MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 96(DX), X14 // X14 = m[12]+m[13] MOVOU 96(DX), X14 // X14 = m[12]+m[13]
@ -699,7 +718,6 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_1(b0, b1); // LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1); // LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 112(DX), X12 // X12 = m[14]+m[15] MOVOU 112(DX), X12 // X12 = m[14]+m[15]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9] MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
@ -709,7 +727,8 @@ TEXT ·blockAVX(SB), 7, $0
MOVOU 80(DX), X13 // X13 = m[10]+m[11] MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 48(DX), X15 // X15 = m[6]+ m[7] MOVOU 48(DX), X15 // X15 = m[6]+ m[7]
BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xd6 // VPUNPCKLQDQ XMM10, XMM13, XMM14 /* m[10], m[8] */ BYTE $0xc4; BYTE $0x41; BYTE $0x11; BYTE $0x6c; BYTE $0xd6 // VPUNPCKLQDQ XMM10, XMM13, XMM14 /* m[10], m[8] */
BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xdc; BYTE $0x08// VPALIGNR XMM11, XMM15, XMM12, 0x8 /* m[15], m[6] */ BYTE $0xc4; BYTE $0x43; BYTE $0x01; BYTE $0x0f; BYTE $0xdc // VPALIGNR XMM11, XMM15, XMM12, 0x8 /* m[15], m[6] */
BYTE $0x08
LOAD_SHUFFLE LOAD_SHUFFLE
@ -721,11 +740,11 @@ TEXT ·blockAVX(SB), 7, $0
// LOAD_MSG_ ##r ##_3(b0, b1); // LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1); // LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register) // (X12 used as additional temp register)
MOVQ message+0(FP), DX // DX: &p (message)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1] MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5] MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 80(DX), X14 // X14 = m[10]+m[11] MOVOU 80(DX), X14 // X14 = m[10]+m[11]
BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xc4; BYTE $0x08 // VPALIGNR XMM8, XMM12, XMM12, 0x8 /* m[1], m[0] */ BYTE $0xc4; BYTE $0x43; BYTE $0x19; BYTE $0x0f; BYTE $0xc4 // VPALIGNR XMM8, XMM12, XMM12, 0x8 /* m[1], m[0] */
BYTE $0x08
BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ XMM9, XMM14, XMM13 /* m[11], m[5] */ BYTE $0xc4; BYTE $0x41; BYTE $0x09; BYTE $0x6d; BYTE $0xcd // VPUNPCKHQDQ XMM9, XMM14, XMM13 /* m[11], m[5] */
MOVOU 16(DX), X12 // X12 = m[2]+ m[3] MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7] MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
@ -740,8 +759,8 @@ TEXT ·blockAVX(SB), 7, $0
UNDIAGONALIZE UNDIAGONALIZE
// Reload digest // Reload digest (most current value store in &out)
MOVQ in+24(FP), SI // SI: &in MOVQ out+144(FP), SI // SI: &in
MOVOU 0(SI), X12 // X12 = in[0]+in[1] /* row1l = LOAD( &S->h[0] ); */ MOVOU 0(SI), X12 // X12 = in[0]+in[1] /* row1l = LOAD( &S->h[0] ); */
MOVOU 16(SI), X13 // X13 = in[2]+in[3] /* row1h = LOAD( &S->h[2] ); */ MOVOU 16(SI), X13 // X13 = in[2]+in[3] /* row1h = LOAD( &S->h[2] ); */
MOVOU 32(SI), X14 // X14 = in[4]+in[5] /* row2l = LOAD( &S->h[4] ); */ MOVOU 32(SI), X14 // X14 = in[4]+in[5] /* row2l = LOAD( &S->h[4] ); */
@ -757,12 +776,17 @@ TEXT ·blockAVX(SB), 7, $0
PXOR X14, X2 // X2 = X2 ^ X14 /* STORE( &S->h[4], _mm_xor_si128( LOAD( &S->h[4] ), row2l ) ); */ PXOR X14, X2 // X2 = X2 ^ X14 /* STORE( &S->h[4], _mm_xor_si128( LOAD( &S->h[4] ), row2l ) ); */
PXOR X15, X3 // X3 = X3 ^ X15 /* STORE( &S->h[6], _mm_xor_si128( LOAD( &S->h[6] ), row2h ) ); */ PXOR X15, X3 // X3 = X3 ^ X15 /* STORE( &S->h[6], _mm_xor_si128( LOAD( &S->h[6] ), row2h ) ); */
// Store digest // Store digest into &out
MOVQ out+144(FP), DX // DX: &out MOVQ out+144(FP), SI // SI: &out
MOVOU X0, 0(DX) // out[0]+out[1] = X0 MOVOU X0, 0(SI) // out[0]+out[1] = X0
MOVOU X1, 16(DX) // out[2]+out[3] = X1 MOVOU X1, 16(SI) // out[2]+out[3] = X1
MOVOU X2, 32(DX) // out[4]+out[5] = X2 MOVOU X2, 32(SI) // out[4]+out[5] = X2
MOVOU X3, 48(DX) // out[6]+out[7] = X3 MOVOU X3, 48(SI) // out[6]+out[7] = X3
// Increment message pointer and check if there's more to do
ADDQ $128, DX // message += 128
SUBQ $1, R8
JNZ loop
complete:
RET RET

40
vendor/github.com/minio/blake2b-simd/compressSse_amd64.go generated vendored Normal file
View File

@ -0,0 +1,40 @@
//+build !noasm
//+build !appengine
/*
* Minio Cloud Storage, (C) 2016 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 blake2b
//go:noescape
func blockSSELoop(p []uint8, in, iv, t, f, shffle, out []uint64)
func compressSSE(d *digest, p []uint8) {
in := make([]uint64, 8, 8)
out := make([]uint64, 8, 8)
shffle := make([]uint64, 2, 2)
// vector for PSHUFB instruction
shffle[0] = 0x0201000706050403
shffle[1] = 0x0a09080f0e0d0c0b
in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7] = d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7]
blockSSELoop(p, in, iv[:], d.t[:], d.f[:], shffle, out)
d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7] = out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7]
}

882
vendor/github.com/minio/blake2b-simd/compressSse_amd64.s generated vendored Normal file
View File

@ -0,0 +1,882 @@
//+build !noasm !appengine
//
// Minio Cloud Storage, (C) 2016 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.
//
//
// Based on SSE implementation from https://github.com/BLAKE2/BLAKE2/blob/master/sse/blake2b.c
//
// Use github.com/fwessels/asm2plan9s on this file to assemble instructions to their Plan9 equivalent
//
// Assembly code below essentially follows the ROUND macro (see blake2b-round.h) which is defined as:
// #define ROUND(r) \
// LOAD_MSG_ ##r ##_1(b0, b1); \
// G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// LOAD_MSG_ ##r ##_2(b0, b1); \
// G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
// LOAD_MSG_ ##r ##_3(b0, b1); \
// G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// LOAD_MSG_ ##r ##_4(b0, b1); \
// G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
// UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
//
// as well as the go equivalent in https://github.com/dchest/blake2b/blob/master/block.go
//
// As in the macro, G1/G2 in the 1st and 2nd half are identical (so literal copy of assembly)
//
// Rounds are also the same, except for the loading of the message (and rounds 1 & 11 and
// rounds 2 & 12 are identical)
//
#define G1 \
\ // G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1);
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0xd4; BYTE $0xc0 \ // PADDQ XMM0,XMM8 /* v0 += m[0], v1 += m[2] */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0xd4; BYTE $0xc9 \ // PADDQ XMM1,XMM9 /* v2 += m[4], v3 += m[6] */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xc2 \ // PADDQ XMM0,XMM2 /* v0 += v4, v1 += v5 */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xcb \ // PADDQ XMM1,XMM3 /* v2 += v6, v3 += v7 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xf0 \ // PXOR XMM6,XMM0 /* v12 ^= v0, v13 ^= v1 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xf9 \ // PXOR XMM7,XMM1 /* v14 ^= v2, v15 ^= v3 */
BYTE $0x66; BYTE $0x0f; BYTE $0x70; BYTE $0xf6; BYTE $0xb1 \ // PSHUFD XMM6,XMM6,0xb1 /* v12 = v12<<(64-32) | v12>>32, v13 = v13<<(64-32) | v13>>32 */
BYTE $0x66; BYTE $0x0f; BYTE $0x70; BYTE $0xff; BYTE $0xb1 \ // PSHUFD XMM7,XMM7,0xb1 /* v14 = v14<<(64-32) | v14>>32, v15 = v15<<(64-32) | v15>>32 */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xe6 \ // PADDQ XMM4,XMM6 /* v8 += v12, v9 += v13 */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xef \ // PADDQ XMM5,XMM7 /* v10 += v14, v11 += v15 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xd4 \ // PXOR XMM2,XMM4 /* v4 ^= v8, v5 ^= v9 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xdd \ // PXOR XMM3,XMM5 /* v6 ^= v10, v7 ^= v11 */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x38; BYTE $0x00 \ // PSHUFB XMM2,XMM12 /* v4 = v4<<(64-24) | v4>>24, v5 = v5<<(64-24) | v5>>24 */
BYTE $0xd4 \
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x38; BYTE $0x00 \ // PSHUFB XMM3,XMM12 /* v6 = v6<<(64-24) | v6>>24, v7 = v7<<(64-24) | v7>>24 */
BYTE $0xdc \
// DO NOT DELETE -- macro delimiter (previous line extended)
#define G2 \
\ // G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1);
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0xd4; BYTE $0xc2 \ // PADDQ XMM0,XMM10 /* v0 += m[1], v1 += m[3] */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0xd4; BYTE $0xcb \ // PADDQ XMM1,XMM11 /* v2 += m[5], v3 += m[7] */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xc2 \ // PADDQ XMM0,XMM2 /* v0 += v4, v1 += v5 */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xcb \ // PADDQ XMM1,XMM3 /* v2 += v6, v3 += v7 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xf0 \ // PXOR XMM6,XMM0 /* v12 ^= v0, v13 ^= v1 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xf9 \ // PXOR XMM7,XMM1 /* v14 ^= v2, v15 ^= v3 */
BYTE $0xf2; BYTE $0x0f; BYTE $0x70; BYTE $0xf6; BYTE $0x39 \ // PSHUFLW XMM6,XMM6,0x39 /* combined with next ... */
BYTE $0xf3; BYTE $0x0f; BYTE $0x70; BYTE $0xf6; BYTE $0x39 \ // PSHUFHW XMM6,XMM6,0x39 /* v12 = v12<<(64-16) | v12>>16, v13 = v13<<(64-16) | v13>>16 */
BYTE $0xf2; BYTE $0x0f; BYTE $0x70; BYTE $0xff; BYTE $0x39 \ // PSHUFLW XMM7,XMM7,0x39 /* combined with next ... */
BYTE $0xf3; BYTE $0x0f; BYTE $0x70; BYTE $0xff; BYTE $0x39 \ // PSHUFHW XMM7,XMM7,0x39 /* v14 = v14<<(64-16) | v14>>16, v15 = v15<<(64-16) | v15>>16 */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xe6 \ // PADDQ XMM4,XMM6 /* v8 += v12, v9 += v13 */
BYTE $0x66; BYTE $0x0f; BYTE $0xd4; BYTE $0xef \ // PADDQ XMM5,XMM7 /* v10 += v14, v11 += v15 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xd4 \ // PXOR XMM2,XMM4 /* v4 ^= v8, v5 ^= v9 */
BYTE $0x66; BYTE $0x0f; BYTE $0xef; BYTE $0xdd \ // PXOR XMM3,XMM5 /* v6 ^= v10, v7 ^= v11 */
MOVOU X2, X15 \
BYTE $0x66; BYTE $0x44; BYTE $0x0f; BYTE $0xd4; BYTE $0xfa \ // PADDQ XMM15,XMM2 /* temp reg = reg*2 */
BYTE $0x66; BYTE $0x0f; BYTE $0x73; BYTE $0xd2; BYTE $0x3f \ // PSRLQ XMM2,0x3f /* reg = reg>>63 */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0xef; BYTE $0xd7 \ // PXOR XMM2,XMM15 /* ORed together: v4 = v4<<(64-63) | v4>>63, v5 = v5<<(64-63) | v5>>63 */
MOVOU X3, X15 \
BYTE $0x66; BYTE $0x44; BYTE $0x0f; BYTE $0xd4; BYTE $0xfb \ // PADDQ XMM15,XMM3 /* temp reg = reg*2 */
BYTE $0x66; BYTE $0x0f; BYTE $0x73; BYTE $0xd3; BYTE $0x3f \ // PSRLQ XMM3,0x3f /* reg = reg>>63 */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0xef; BYTE $0xdf // PXOR XMM3,XMM15 /* ORed together: v6 = v6<<(64-63) | v6>>63, v7 = v7<<(64-63) | v7>>63 */
#define DIAGONALIZE \
\ // DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
MOVOU X6, X13 \ /* t0 = row4l;\ */
MOVOU X2, X14 \ /* t1 = row2l;\ */
MOVOU X4, X6 \ /* row4l = row3l;\ */
MOVOU X5, X4 \ /* row3l = row3h;\ */
MOVOU X6, X5 \ /* row3h = row4l;\ */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xfd \ // PUNPCKLQDQ XMM15, XMM13 /* _mm_unpacklo_epi64(t0, t0) */
MOVOU X7, X6 \
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xf7 \ // PUNPCKHQDQ XMM6, XMM15 /* row4l = _mm_unpackhi_epi64(row4h, ); \ */
BYTE $0x66; BYTE $0x44; BYTE $0x0f; BYTE $0x6c; BYTE $0xff \ // PUNPCKLQDQ XMM15, XMM7 /* _mm_unpacklo_epi64(row4h, row4h) */
MOVOU X13, X7 \
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xff \ // PUNPCKHQDQ XMM7, XMM15 /* row4h = _mm_unpackhi_epi64(t0, ); \ */
BYTE $0x66; BYTE $0x44; BYTE $0x0f; BYTE $0x6c; BYTE $0xfb \ // PUNPCKLQDQ XMM15, XMM3 /* _mm_unpacklo_epi64(row2h, row2h) */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xd7 \ // PUNPCKHQDQ XMM2, XMM15 /* row2l = _mm_unpackhi_epi64(row2l, ); \ */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xfe \ // PUNPCKLQDQ XMM15, XMM14 /* _mm_unpacklo_epi64(t1, t1) */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xdf // PUNPCKHQDQ XMM3, XMM15 /* row2h = _mm_unpackhi_epi64(row2h, ) */
#define UNDIAGONALIZE \
\ // UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
MOVOU X4, X13 \ /* t0 = row3l;\ */
MOVOU X5, X4 \ /* row3l = row3h;\ */
MOVOU X13, X5 \ /* row3h = t0;\ */
MOVOU X2, X13 \ /* t0 = row2l;\ */
MOVOU X6, X14 \ /* t1 = row4l;\ */
BYTE $0x66; BYTE $0x44; BYTE $0x0f; BYTE $0x6c; BYTE $0xfa \ // PUNPCKLQDQ XMM15, XMM2 /* _mm_unpacklo_epi64(row2l, row2l) */
MOVOU X3, X2 \
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xd7 \ // PUNPCKHQDQ XMM2, XMM15 /* row2l = _mm_unpackhi_epi64(row2h, ); \ */
BYTE $0x66; BYTE $0x44; BYTE $0x0f; BYTE $0x6c; BYTE $0xfb \ // PUNPCKLQDQ XMM15, XMM3 /* _mm_unpacklo_epi64(row2h, row2h) */
MOVOU X13, X3 \
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xdf \ // PUNPCKHQDQ XMM3, XMM15 /* row2h = _mm_unpackhi_epi64(t0, ); \ */
BYTE $0x66; BYTE $0x44; BYTE $0x0f; BYTE $0x6c; BYTE $0xff \ // PUNPCKLQDQ XMM15, XMM7 /* _mm_unpacklo_epi64(row4h, row4h) */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xf7 \ // PUNPCKHQDQ XMM6, XMM15 /* row4l = _mm_unpackhi_epi64(row4l, ); \ */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xfe \ // PUNPCKLQDQ XMM15, XMM14 /* _mm_unpacklo_epi64(t1, t1) */
BYTE $0x66; BYTE $0x41; BYTE $0x0f; BYTE $0x6d; BYTE $0xff // PUNPCKHQDQ XMM7, XMM15 /* row4h = _mm_unpackhi_epi64(row4h, ) */
#define LOAD_SHUFFLE \
\ // Load shuffle value
MOVQ shffle+120(FP), SI \ // SI: &shuffle
MOVOU 0(SI), X12 // X12 = 03040506 07000102 0b0c0d0e 0f08090a
// func blockSSELoop(p []uint8, in, iv, t, f, shffle, out []uint64)
TEXT ·blockSSELoop(SB), 7, $0
// REGISTER USE
// R8: loop counter
// DX: message pointer
// SI: temp pointer for loading
// X0 - X7: v0 - v15
// X8 - X11: m[0] - m[7]
// X12: shuffle value
// X13 - X15: temp registers
// Load digest
MOVQ in+24(FP), SI // SI: &in
MOVOU 0(SI), X0 // X0 = in[0]+in[1] /* row1l = LOAD( &S->h[0] ); */
MOVOU 16(SI), X1 // X1 = in[2]+in[3] /* row1h = LOAD( &S->h[2] ); */
MOVOU 32(SI), X2 // X2 = in[4]+in[5] /* row2l = LOAD( &S->h[4] ); */
MOVOU 48(SI), X3 // X3 = in[6]+in[7] /* row2h = LOAD( &S->h[6] ); */
// Already store digest into &out (so we can reload it later generically)
MOVQ out+144(FP), SI // SI: &out
MOVOU X0, 0(SI) // out[0]+out[1] = X0
MOVOU X1, 16(SI) // out[2]+out[3] = X1
MOVOU X2, 32(SI) // out[4]+out[5] = X2
MOVOU X3, 48(SI) // out[6]+out[7] = X3
// Initialize message pointer and loop counter
MOVQ message+0(FP), DX // DX: &p (message)
MOVQ message_len+8(FP), R8 // R8: len(message)
SHRQ $7, R8 // len(message) / 128
CMPQ R8, $0
JEQ complete
loop:
// Increment counter
MOVQ t+72(FP), SI // SI: &t
MOVQ 0(SI), R9 //
ADDQ $128, R9 // /* d.t[0] += BlockSize */
MOVQ R9, 0(SI) //
CMPQ R9, $128 // /* if d.t[0] < BlockSize { */
JGE noincr //
MOVQ 8(SI), R9 //
ADDQ $1, R9 // /* d.t[1]++ */
MOVQ R9, 8(SI) //
noincr: // /* } */
// Load initialization vector
MOVQ iv+48(FP), SI // SI: &iv
MOVOU 0(SI), X4 // X4 = iv[0]+iv[1] /* row3l = LOAD( &blake2b_IV[0] ); */
MOVOU 16(SI), X5 // X5 = iv[2]+iv[3] /* row3h = LOAD( &blake2b_IV[2] ); */
MOVOU 32(SI), X6 // X6 = iv[4]+iv[5] /* LOAD( &blake2b_IV[4] ) */
MOVOU 48(SI), X7 // X7 = iv[6]+iv[7] /* LOAD( &blake2b_IV[6] ) */
MOVQ t+72(FP), SI // SI: &t
MOVOU 0(SI), X8 // X8 = t[0]+t[1] /* LOAD( &S->t[0] ) */
PXOR X8, X6 // X6 = X6 ^ X8 /* row4l = _mm_xor_si128( , ); */
MOVQ t+96(FP), SI // SI: &f
MOVOU 0(SI), X8 // X8 = f[0]+f[1] /* LOAD( &S->f[0] ) */
PXOR X8, X7 // X7 = X7 ^ X8 /* row4h = _mm_xor_si128( , ); */
///////////////////////////////////////////////////////////////////////////
// R O U N D 1
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+m[1]
MOVOU 16(DX), X13 // X13 = m[2]+m[3]
MOVOU 32(DX), X14 // X14 = m[4]+m[5]
MOVOU 48(DX), X15 // X15 = m[6]+m[7]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc5 // PUNPCKLQDQ XMM8, XMM13 /* m[0], m[2] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcf // PUNPCKLQDQ XMM9, XMM15 /* m[4], m[6] */
MOVOU X12, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd5 // PUNPCKHQDQ XMM10, XMM13 /* m[1], m[3] */
MOVOU X14, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdf // PUNPCKHQDQ XMM11, XMM15 /* m[5], m[7] */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 64(DX), X12 // X12 = m[8]+ m[9]
MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc5 // PUNPCKLQDQ XMM8, XMM13 /* m[8],m[10] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcf // PUNPCKLQDQ XMM9, XMM15 /* m[12],m[14] */
MOVOU X12, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd5 // PUNPCKHQDQ XMM10, XMM13 /* m[9],m[11] */
MOVOU X14, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdf // PUNPCKHQDQ XMM11, XMM15 /* m[13],m[15] */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 2
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 112(DX), X12 // X12 = m[14]+m[15]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 96(DX), X15 // X15 = m[12]+m[13]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc5 // PUNPCKLQDQ XMM8, XMM13 /* m[14], m[4] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcf // PUNPCKHQDQ XMM9, XMM15 /* m[9], m[13] */
MOVOU 80(DX), X10 // X10 = m[10]+m[11]
MOVOU 48(DX), X11 // X11 = m[6]+ m[7]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd6 // PUNPCKLQDQ XMM10, XMM14 /* m[10], m[8] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM11, XMM12, 0x8 /* m[15], m[6] */; ; ; ; ;
BYTE $0xdc; BYTE $0x08
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM8, XMM12, 0x8 /* m[1], m[0] */
BYTE $0xc4; BYTE $0x08
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcd // PUNPCKHQDQ XMM9, XMM13 /* m[11], m[5] */
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X11 // X11 = m[6]+ m[7]
MOVOU 96(DX), X10 // X10 = m[12]+m[13]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd4 // PUNPCKLQDQ XMM10, XMM12 /* m[12], m[2] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdc // PUNPCKHQDQ XMM11, XMM12 /* m[7], m[3] */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 3
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 32(DX), X12 // X12 = m[4]+ m[5]
MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X14, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM8, XMM13, 0x8 /* m[11], m[12] */
BYTE $0xc5; BYTE $0x08
MOVOU X12, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcf // PUNPCKHQDQ XMM9, XMM15 /* m[5], m[15] */
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
MOVOU 64(DX), X10 // X10 = m[8]+ m[9]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd4 // PUNPCKLQDQ XMM10, XMM12 /* m[8], m[0] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xf6 // PUNPCKHQDQ XMM14, XMM14 /* ___, m[13] */
MOVOU X13, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xde // PUNPCKLQDQ XMM11, XMM14 /* m[2], ___ */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 80(DX), X15 // X15 = m[10]+m[11]
MOVOU X12, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcc // PUNPCKHQDQ XMM9, XMM12 /* ___, m[3] */
MOVOU X15, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc1 // PUNPCKLQDQ XMM8, XMM9 /* m[10], ___ */
MOVOU X13, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xce // PUNPCKHQDQ XMM9, XMM14 /* m[7], m[9] */
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X11 // X11 = m[4]+ m[5]
MOVOU 112(DX), X10 // X10 = m[14]+m[15]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd5 // PUNPCKLQDQ XMM10, XMM13 /* m[14], m[6] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM11, XMM12, 0x8 /* m[1], m[4] */
BYTE $0xdc; BYTE $0x08
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 4
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU 96(DX), X15 // X15 = m[12]+m[13]
MOVOU X13, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xc4 // PUNPCKHQDQ XMM8, XMM12 /* m[7], m[3] */
MOVOU X15, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xce // PUNPCKHQDQ XMM9, XMM14 /* m[13], m[11] */
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 64(DX), X10 // X10 = m[8]+ m[9]
MOVOU 112(DX), X14 // X14 = m[14]+m[15]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd4 // PUNPCKHQDQ XMM10, XMM12 /* m[9], m[1] */
MOVOU X15, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xde // PUNPCKLQDQ XMM11, XMM14 /* m[12], m[14] */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X13, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcd // PUNPCKHQDQ XMM9, XMM13 /* ___, m[5] */
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc1 // PUNPCKLQDQ XMM8, XMM9 /* m[2], ____ */
MOVOU X15, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd7 // PUNPCKHQDQ XMM10, XMM15 /* ___, m[15] */
MOVOU X13, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xca // PUNPCKLQDQ XMM9, XMM10 /* m[4], ____ */
MOVOU 0(DX), X11 // X11 = m[0]+ m[1]
MOVOU 48(DX), X10 // X10 = m[6]+ m[7]
MOVOU 64(DX), X15 // X15 = m[8]+ m[9]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd6 // PUNPCKLQDQ XMM10, XMM14 /* m[6], m[10] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdf // PUNPCKLQDQ XMM11, XMM15 /* m[0], m[8] */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 5
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 80(DX), X15 // X15 = m[10]+m[11]
MOVOU X14, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xc5 // PUNPCKHQDQ XMM8, XMM13 /* m[9], m[5] */
MOVOU X12, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcf // PUNPCKLQDQ XMM9, XMM15 /* m[2], m[10] */
MOVOU 0(DX), X10 // X10 = m[0]+ m[1]
MOVOU 48(DX), X14 // X14 = m[6]+ m[7]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xf6 // PUNPCKHQDQ XMM14, XMM14 /* ___, m[7] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd6 // PUNPCKLQDQ XMM10, XMM14 /* m[0], ____ */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xff // PUNPCKHQDQ XMM15, XMM15 /* ___, m[15] */
MOVOU X13, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdf // PUNPCKLQDQ XMM11, XMM15 /* m[4], ____ */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xf6 // PUNPCKHQDQ XMM14, XMM14 /* ___, m[11] */
MOVOU X15, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc6 // PUNPCKLQDQ XMM8, XMM14 /* m[14], ____ */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xe4 // PUNPCKHQDQ XMM12, XMM12 /* ___, m[3] */
MOVOU X13, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcc // PUNPCKLQDQ XMM9, XMM12 /* m[6], ____ */
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 64(DX), X11 // X11 = m[8]+ m[9]
MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU X14, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM10, XMM12, 0x8 /* m[1], m[12] */
BYTE $0xd4; BYTE $0x08
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xf6 // PUNPCKHQDQ XMM14, XMM14 /* ___, m[13] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xde // PUNPCKLQDQ XMM11, XMM14 /* m[8], ____ */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 6
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
MOVOU 48(DX), X14 // X14 = m[6]+ m[7]
MOVOU 64(DX), X15 // X15 = m[8]+ m[9]
MOVOU X13, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc6 // PUNPCKLQDQ XMM8, XMM14 /* m[2], m[6] */
MOVOU X12, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcf // PUNPCKLQDQ XMM9, XMM15 /* m[0], m[8] */
MOVOU 80(DX), X12 // X12 = m[10]+m[11]
MOVOU 96(DX), X10 // X10 = m[12]+m[13]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd4 // PUNPCKLQDQ XMM10, XMM12 /* m[12], m[10] */
MOVOU X12, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdd // PUNPCKHQDQ XMM11, XMM13 /* m[11], m[3] */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 48(DX), X14 // X14 = m[6]+ m[7]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xce // PUNPCKHQDQ XMM9, XMM14 /* ___, m[7] */
MOVOU X13, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc1 // PUNPCKLQDQ XMM8, XMM9 /* m[4], ____ */
MOVOU X15, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcc // PUNPCKHQDQ XMM9, XMM12 /* m[15], m[1] */
MOVOU 64(DX), X12 // X12 = m[8]+ m[9]
MOVOU 96(DX), X10 // X10 = m[12]+m[13]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd5 // PUNPCKHQDQ XMM10, XMM13 /* m[13], m[5] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xe4 // PUNPCKHQDQ XMM12, XMM12 /* ___, m[9] */
MOVOU X15, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdc // PUNPCKLQDQ XMM11, XMM12 /* m[14], ____ */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 7
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X12, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcc // PUNPCKHQDQ XMM9, XMM12 /* ___, m[1] */
MOVOU X14, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc1 // PUNPCKLQDQ XMM8, XMM9 /* m[12], ____ */
MOVOU X15, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcd // PUNPCKLQDQ XMM9, XMM13 /* m[14], m[4] */
MOVOU 80(DX), X11 // X11 = m[10]+m[11]
MOVOU X13, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd7 // PUNPCKHQDQ XMM10, XMM15 /* m[5], m[15] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM11, XMM14, 0x8 /* m[13], m[10] */
BYTE $0xde; BYTE $0x08
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 80(DX), X15 // X15 = m[10]+m[11]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc5 // PUNPCKLQDQ XMM8, XMM13 /* m[0], m[6] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM9, XMM14, 0x8 /* m[9], m[8] */
BYTE $0xce; BYTE $0x08
MOVOU 16(DX), X11 // X14 = m[2]+ m[3]
MOVOU X13, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd3 // PUNPCKHQDQ XMM10, XMM11 /* m[7], m[3] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xff // PUNPCKHQDQ XMM15, XMM15 /* ___, m[11] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdf // PUNPCKLQDQ XMM11, XMM15 /* m[2], ____ */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 8
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X14, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xc5 // PUNPCKHQDQ XMM8, XMM13 /* m[13], m[7] */
MOVOU X12, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd4 // PUNPCKHQDQ XMM10, XMM12 /* ___, m[3] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xca // PUNPCKLQDQ XMM9, XMM10 /* m[12], ____ */
MOVOU 0(DX), X11 // X11 = m[0]+ m[1]
MOVOU 64(DX), X13 // X13 = m[8]+ m[9]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU X15, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM10, XMM14, 0x8 /* m[11], m[14] */
BYTE $0xd6; BYTE $0x08
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdd // PUNPCKHQDQ XMM11, XMM13 /* m[1], m[9] */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X13, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xc7 // PUNPCKHQDQ XMM8, XMM15 /* m[5], m[15] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcc // PUNPCKLQDQ XMM9, XMM12 /* m[8], m[2] */
MOVOU 0(DX), X10 // X10 = m[0]+ m[1]
MOVOU 48(DX), X11 // X11 = m[6]+ m[7]
MOVOU 80(DX), X15 // X15 = m[10]+m[11]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd5 // PUNPCKLQDQ XMM10, XMM13 /* m[0], m[4] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdf // PUNPCKLQDQ XMM11, XMM15 /* m[6], m[10] */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 9
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X13, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc7 // PUNPCKLQDQ XMM8, XMM15 /* m[6], m[14] */
MOVOU X12, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM9, XMM14, 0x8 /* m[11], m[0] */
BYTE $0xce; BYTE $0x08
MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
MOVOU 64(DX), X11 // X11 = m[8]+ m[9]
MOVOU X15, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd3 // PUNPCKHQDQ XMM10, XMM11 /* m[15], m[9] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM11, XMM13, 0x8 /* m[3], m[8] */
BYTE $0xdd; BYTE $0x08
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 16(DX), X13 // X13 = m[2]+ m[3]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU 96(DX), X15 // X15 = m[12]+m[13]
MOVOU X15, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcf // PUNPCKHQDQ XMM9, XMM15 /* ___, m[13] */
MOVOU X15, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc1 // PUNPCKLQDQ XMM8, XMM9 /* m[12], ____ */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM9, XMM12, 0x8 /* m[1], m[10] */
BYTE $0xcc; BYTE $0x08
MOVOU 32(DX), X12 // X12 = m[4]+ m[5]
MOVOU 48(DX), X15 // X15 = m[6]+ m[7]
MOVOU X15, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdf // PUNPCKHQDQ XMM11, XMM15 /* ___, m[7] */
MOVOU X13, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd3 // PUNPCKLQDQ XMM10, XMM11 /* m[2], ____ */
MOVOU X12, X15
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xfc // PUNPCKHQDQ XMM15, XMM12 /* ___, m[5] */
MOVOU X12, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdf // PUNPCKLQDQ XMM11, XMM15 /* m[4], ____ */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 1 0
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 48(DX), X13 // X13 = m[6]+ m[7]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 80(DX), X15 // X15 = m[10]+m[11]
MOVOU X15, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc6 // PUNPCKLQDQ XMM8, XMM14 /* m[10], m[8] */
MOVOU X13, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcc // PUNPCKHQDQ XMM9, XMM12 /* m[7], m[1] */
MOVOU 16(DX), X10 // X10 = m[2]+ m[3]
MOVOU 32(DX), X14 // X14 = m[4]+ m[5]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd6 // PUNPCKLQDQ XMM10, XMM14 /* m[2], m[4] */
MOVOU X14, X15
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xfe // PUNPCKHQDQ XMM15, XMM14 /* ___, m[5] */
MOVOU X13, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdf // PUNPCKLQDQ XMM11, XMM15 /* m[6], ____ */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 64(DX), X13 // X13 = m[8]+ m[9]
MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X15, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xc5 // PUNPCKHQDQ XMM8, XMM13 /* m[15], m[9] */
MOVOU X12, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xce // PUNPCKHQDQ XMM9, XMM14 /* m[3], m[13] */
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU X15, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM10, XMM13, 0x8 /* m[11], m[14] */
BYTE $0xd5; BYTE $0x08
MOVOU X14, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xdc // PUNPCKLQDQ XMM11, XMM12 /* m[12], m[0] */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 1 1
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+m[1]
MOVOU 16(DX), X13 // X13 = m[2]+m[3]
MOVOU 32(DX), X14 // X14 = m[4]+m[5]
MOVOU 48(DX), X15 // X15 = m[6]+m[7]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc5 // PUNPCKLQDQ XMM8, XMM13 /* m[0], m[2] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcf // PUNPCKLQDQ XMM9, XMM15 /* m[4], m[6] */
MOVOU X12, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd5 // PUNPCKHQDQ XMM10, XMM13 /* m[1], m[3] */
MOVOU X14, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdf // PUNPCKHQDQ XMM11, XMM15 /* m[5], m[7] */
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 64(DX), X12 // X12 = m[8]+ m[9]
MOVOU 80(DX), X13 // X13 = m[10]+m[11]
MOVOU 96(DX), X14 // X14 = m[12]+m[13]
MOVOU 112(DX), X15 // X15 = m[14]+m[15]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc5 // PUNPCKLQDQ XMM8, XMM13 /* m[8],m[10] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xcf // PUNPCKLQDQ XMM9, XMM15 /* m[12],m[14] */
MOVOU X12, X10
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xd5 // PUNPCKHQDQ XMM10, XMM13 /* m[9],m[11] */
MOVOU X14, X11
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdf // PUNPCKHQDQ XMM11, XMM15 /* m[13],m[15] */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
///////////////////////////////////////////////////////////////////////////
// R O U N D 1 2
///////////////////////////////////////////////////////////////////////////
// LOAD_MSG_ ##r ##_1(b0, b1);
// LOAD_MSG_ ##r ##_2(b0, b1);
// (X12 used as additional temp register)
MOVOU 112(DX), X12 // X12 = m[14]+m[15]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 64(DX), X14 // X14 = m[8]+ m[9]
MOVOU 96(DX), X15 // X15 = m[12]+m[13]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xc5 // PUNPCKLQDQ XMM8, XMM13 /* m[14], m[4] */
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcf // PUNPCKHQDQ XMM9, XMM15 /* m[9], m[13] */
MOVOU 80(DX), X10 // X10 = m[10]+m[11]
MOVOU 48(DX), X11 // X11 = m[6]+ m[7]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd6 // PUNPCKLQDQ XMM10, XMM14 /* m[10], m[8] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM11, XMM12, 0x8 /* m[15], m[6] */; ; ; ; ;
BYTE $0xdc; BYTE $0x08
LOAD_SHUFFLE
G1
G2
DIAGONALIZE
// LOAD_MSG_ ##r ##_3(b0, b1);
// LOAD_MSG_ ##r ##_4(b0, b1);
// (X12 used as additional temp register)
MOVOU 0(DX), X12 // X12 = m[0]+ m[1]
MOVOU 32(DX), X13 // X13 = m[4]+ m[5]
MOVOU 80(DX), X14 // X14 = m[10]+m[11]
MOVOU X12, X8
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x3a; BYTE $0x0f // PALIGNR XMM8, XMM12, 0x8 /* m[1], m[0] */
BYTE $0xc4; BYTE $0x08
MOVOU X14, X9
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xcd // PUNPCKHQDQ XMM9, XMM13 /* m[11], m[5] */
MOVOU 16(DX), X12 // X12 = m[2]+ m[3]
MOVOU 48(DX), X11 // X11 = m[6]+ m[7]
MOVOU 96(DX), X10 // X10 = m[12]+m[13]
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6c; BYTE $0xd4 // PUNPCKLQDQ XMM10, XMM12 /* m[12], m[2] */
BYTE $0x66; BYTE $0x45; BYTE $0x0f; BYTE $0x6d; BYTE $0xdc // PUNPCKHQDQ XMM11, XMM12 /* m[7], m[3] */
LOAD_SHUFFLE
G1
G2
UNDIAGONALIZE
// Reload digest (most current value store in &out)
MOVQ out+144(FP), SI // SI: &in
MOVOU 0(SI), X12 // X12 = in[0]+in[1] /* row1l = LOAD( &S->h[0] ); */
MOVOU 16(SI), X13 // X13 = in[2]+in[3] /* row1h = LOAD( &S->h[2] ); */
MOVOU 32(SI), X14 // X14 = in[4]+in[5] /* row2l = LOAD( &S->h[4] ); */
MOVOU 48(SI), X15 // X15 = in[6]+in[7] /* row2h = LOAD( &S->h[6] ); */
// Final computations and prepare for storing
PXOR X4, X0 // X0 = X0 ^ X4 /* row1l = _mm_xor_si128( row3l, row1l ); */
PXOR X5, X1 // X1 = X1 ^ X5 /* row1h = _mm_xor_si128( row3h, row1h ); */
PXOR X12, X0 // X0 = X0 ^ X12 /* STORE( &S->h[0], _mm_xor_si128( LOAD( &S->h[0] ), row1l ) ); */
PXOR X13, X1 // X1 = X1 ^ X13 /* STORE( &S->h[2], _mm_xor_si128( LOAD( &S->h[2] ), row1h ) ); */
PXOR X6, X2 // X2 = X2 ^ X6 /* row2l = _mm_xor_si128( row4l, row2l ); */
PXOR X7, X3 // X3 = X3 ^ X7 /* row2h = _mm_xor_si128( row4h, row2h ); */
PXOR X14, X2 // X2 = X2 ^ X14 /* STORE( &S->h[4], _mm_xor_si128( LOAD( &S->h[4] ), row2l ) ); */
PXOR X15, X3 // X3 = X3 ^ X15 /* STORE( &S->h[6], _mm_xor_si128( LOAD( &S->h[6] ), row2h ) ); */
// Store digest into &out
MOVQ out+144(FP), SI // SI: &out
MOVOU X0, 0(SI) // out[0]+out[1] = X0
MOVOU X1, 16(SI) // out[2]+out[3] = X1
MOVOU X2, 32(SI) // out[4]+out[5] = X2
MOVOU X3, 48(SI) // out[6]+out[7] = X3
// Increment message pointer and check if there's more to do
ADDQ $128, DX // message += 128
SUBQ $1, R8
JNZ loop
complete:
RET

48
vendor/github.com/minio/blake2b-simd/compress_amd64.go generated vendored
View File

@ -1,6 +1,3 @@
//+build !noasm
//+build !appengine
/* /*
* Minio Cloud Storage, (C) 2016 Minio, Inc. * Minio Cloud Storage, (C) 2016 Minio, Inc.
* *
@ -19,44 +16,15 @@
package blake2b package blake2b
//go:noescape
func blockAVX(p []uint8, in, iv, t, f, shffle, out []uint64)
func compressAVX(d *digest, p []uint8) {
h0, h1, h2, h3, h4, h5, h6, h7 := d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7]
in := make([]uint64, 8, 8)
out := make([]uint64, 8, 8)
shffle := make([]uint64, 2, 2)
// vector for PSHUFB instruction
shffle[0] = 0x0201000706050403
shffle[1] = 0x0a09080f0e0d0c0b
for len(p) >= BlockSize {
// Increment counter.
d.t[0] += BlockSize
if d.t[0] < BlockSize {
d.t[1]++
}
in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7] = h0, h1, h2, h3, h4, h5, h6, h7
blockAVX(p, in, iv[:], d.t[:], d.f[:], shffle, out)
h0, h1, h2, h3, h4, h5, h6, h7 = out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7]
p = p[BlockSize:]
}
d.h[0], d.h[1], d.h[2], d.h[3], d.h[4], d.h[5], d.h[6], d.h[7] = h0, h1, h2, h3, h4, h5, h6, h7
}
func compress(d *digest, p []uint8) { func compress(d *digest, p []uint8) {
// Verifies if AVX is available, use optimized code path. // Verifies if AVX2 or AVX is available, use optimized code path.
if avx { if avx2 {
compressAVX2(d, p)
} else if avx {
compressAVX(d, p) compressAVX(d, p)
return } else if ssse3 {
} // else { fallback to generic approach. compressSSE(d, p)
} else {
compressGeneric(d, p) compressGeneric(d, p)
}
} }

21
vendor/github.com/minio/blake2b-simd/compress_noasm.go generated vendored
View File

@ -1,11 +1,20 @@
//+build !amd64 noasm appengine //+build !amd64 noasm appengine
// Written in 2012 by Dmitry Chestnykh. /*
// * Minio Cloud Storage, (C) 2016 Minio, Inc.
// To the extent possible under law, the author have dedicated all copyright *
// and related and neighboring rights to this software to the public domain * Licensed under the Apache License, Version 2.0 (the "License");
// worldwide. This software is distributed without any warranty. * you may not use this file except in compliance with the License.
// http://creativecommons.org/publicdomain/zero/1.0/ * 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 blake2b package blake2b

25
vendor/github.com/minio/blake2b-simd/cpuid.go generated vendored
View File

@ -18,12 +18,15 @@
package blake2b package blake2b
func cpuid(op uint32) (eax, ebx, ecx, edx uint32) func cpuid(op uint32) (eax, ebx, ecx, edx uint32)
func cpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32)
func xgetbv(index uint32) (eax, edx uint32) func xgetbv(index uint32) (eax, edx uint32)
// True when SIMD instructions are available. // True when SIMD instructions are available.
var avx2 = haveAVX2()
var avx = haveAVX() var avx = haveAVX()
var ssse3 = haveSSSE3()
// haveSSE returns true if we have streaming SIMD instructions. // haveAVX returns true when there is AVX support
func haveAVX() bool { func haveAVX() bool {
_, _, c, _ := cpuid(1) _, _, c, _ := cpuid(1)
@ -35,3 +38,23 @@ func haveAVX() bool {
} }
return false return false
} }
// haveAVX2 returns true when there is AVX2 support
func haveAVX2() bool {
mfi, _, _, _ := cpuid(0)
// Check AVX2, AVX2 requires OS support, but BMI1/2 don't.
if mfi >= 7 && haveAVX() {
_, ebx, _, _ := cpuidex(7, 0)
return (ebx & 0x00000020) != 0
}
return false
}
// haveSSSE3 returns true when there is SSSE3 support
func haveSSSE3() bool {
_, _, c, _ := cpuid(1)
return (c & 0x00000200) != 0
}

11
vendor/github.com/minio/blake2b-simd/cpuid_386.s generated vendored
View File

@ -13,6 +13,17 @@ TEXT ·cpuid(SB), 7, $0
MOVL DX, edx+16(FP) MOVL DX, edx+16(FP)
RET RET
// func cpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32)
TEXT ·cpuidex(SB), 7, $0
MOVL op+0(FP), AX
MOVL op2+4(FP), CX
CPUID
MOVL AX, eax+8(FP)
MOVL BX, ebx+12(FP)
MOVL CX, ecx+16(FP)
MOVL DX, edx+20(FP)
RET
// func xgetbv(index uint32) (eax, edx uint32) // func xgetbv(index uint32) (eax, edx uint32)
TEXT ·xgetbv(SB), 7, $0 TEXT ·xgetbv(SB), 7, $0
MOVL index+0(FP), CX MOVL index+0(FP), CX

12
vendor/github.com/minio/blake2b-simd/cpuid_amd64.s generated vendored
View File

@ -13,6 +13,18 @@ TEXT ·cpuid(SB), 7, $0
MOVL DX, edx+20(FP) MOVL DX, edx+20(FP)
RET RET
// func cpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32)
TEXT ·cpuidex(SB), 7, $0
MOVL op+0(FP), AX
MOVL op2+4(FP), CX
CPUID
MOVL AX, eax+8(FP)
MOVL BX, ebx+12(FP)
MOVL CX, ecx+16(FP)
MOVL DX, edx+20(FP)
RET
// func xgetbv(index uint32) (eax, edx uint32) // func xgetbv(index uint32) (eax, edx uint32)
TEXT ·xgetbv(SB), 7, $0 TEXT ·xgetbv(SB), 7, $0
MOVL index+0(FP), CX MOVL index+0(FP), CX

4
vendor/vendor.json vendored
View File

@ -84,8 +84,8 @@
}, },
{ {
"path": "github.com/minio/blake2b-simd", "path": "github.com/minio/blake2b-simd",
"revision": "0b3e695ecc77a334fafe30ee36de504c41ec4d6a", "revision": "25efc542f2c5064cf312cdca043790a7af861c4c",
"revisionTime": "2016-06-28T02:55:56-07:00" "revisionTime": "2016-07-06T10:29:24+02:00"
}, },
{ {
"path": "github.com/minio/cli", "path": "github.com/minio/cli",