// +build amd64

package sha1

// #cgo CFLAGS: -DHAS_AVX2
// #include <stdint.h>
// #include <stdlib.h>
// void sha1_transform(int32_t *hash, const char* input, size_t num_blocks);
// void sha1_update_intel(int32_t *hash, const char* input, size_t num_blocks );
import "C"
import "unsafe"

const (
	_K0 = 0x5A827999
	_K1 = 0x6ED9EBA1
	_K2 = 0x8F1BBCDC
	_K3 = 0xCA62C1D6
)

func blockAVX2(dig *digest, p []byte) {
	C.sha1_transform((*C.int32_t)(unsafe.Pointer(&dig.h[0])), (*C.char)(unsafe.Pointer(&p[0])), (C.size_t)(len(p)/chunk))
}

func blockSSE3(dig *digest, p []byte) {
	C.sha1_update_intel((*C.int32_t)(unsafe.Pointer(&dig.h[0])), (*C.char)(unsafe.Pointer(&p[0])), (C.size_t)(len(p)/chunk))
}

// blockGeneric is a portable, pure Go version of the SHA1 block step.
// It's used by sha1block_generic.go and tests.
func blockGeneric(dig *digest, p []byte) {
	var w [16]uint32

	h0, h1, h2, h3, h4 := dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4]
	for len(p) >= chunk {
		// Can interlace the computation of w with the
		// rounds below if needed for speed.
		for i := 0; i < 16; i++ {
			j := i * 4
			w[i] = uint32(p[j])<<24 | uint32(p[j+1])<<16 | uint32(p[j+2])<<8 | uint32(p[j+3])
		}

		a, b, c, d, e := h0, h1, h2, h3, h4

		// Each of the four 20-iteration rounds
		// differs only in the computation of f and
		// the choice of K (_K0, _K1, etc).
		i := 0
		for ; i < 16; i++ {
			f := b&c | (^b)&d
			a5 := a<<5 | a>>(32-5)
			b30 := b<<30 | b>>(32-30)
			t := a5 + f + e + w[i&0xf] + _K0
			a, b, c, d, e = t, a, b30, c, d
		}
		for ; i < 20; i++ {
			tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
			w[i&0xf] = tmp<<1 | tmp>>(32-1)

			f := b&c | (^b)&d
			a5 := a<<5 | a>>(32-5)
			b30 := b<<30 | b>>(32-30)
			t := a5 + f + e + w[i&0xf] + _K0
			a, b, c, d, e = t, a, b30, c, d
		}
		for ; i < 40; i++ {
			tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
			w[i&0xf] = tmp<<1 | tmp>>(32-1)
			f := b ^ c ^ d
			a5 := a<<5 | a>>(32-5)
			b30 := b<<30 | b>>(32-30)
			t := a5 + f + e + w[i&0xf] + _K1
			a, b, c, d, e = t, a, b30, c, d
		}
		for ; i < 60; i++ {
			tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
			w[i&0xf] = tmp<<1 | tmp>>(32-1)
			f := ((b | c) & d) | (b & c)

			a5 := a<<5 | a>>(32-5)
			b30 := b<<30 | b>>(32-30)
			t := a5 + f + e + w[i&0xf] + _K2
			a, b, c, d, e = t, a, b30, c, d
		}
		for ; i < 80; i++ {
			tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
			w[i&0xf] = tmp<<1 | tmp>>(32-1)
			f := b ^ c ^ d
			a5 := a<<5 | a>>(32-5)
			b30 := b<<30 | b>>(32-30)
			t := a5 + f + e + w[i&0xf] + _K3
			a, b, c, d, e = t, a, b30, c, d
		}

		h0 += a
		h1 += b
		h2 += c
		h3 += d
		h4 += e

		p = p[chunk:]
	}
	dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4] = h0, h1, h2, h3, h4
}