//go:build !purego

package sha1

import "internal/cpu"

//go:noescape
func blockAVX2(dig *digest, p []byte)

//go:noescape
func blockAMD64(dig *digest, p []byte)

var useAVX2 = cpu.X86.HasAVX2 && cpu.X86.HasBMI1 && cpu.X86.HasBMI2

func block(dig *digest, p []byte) {
	if useAVX2 && len(p) >= 256 {
		// blockAVX2 calculates sha1 for 2 block per iteration
		// it also interleaves precalculation for next block.

//go:build !purego

package sha512

import "internal/cpu"

//go:noescape
func blockAVX2(dig *digest, p []byte)

//go:noescape
func blockAMD64(dig *digest, p []byte)

var useAVX2 = cpu.X86.HasAVX2 && cpu.X86.HasBMI1 && cpu.X86.HasBMI2

func block(dig *digest, p []byte) {
	if useAVX2 {
		blockAVX2(dig, p)
	} else {
		blockAMD64(dig, p)
	}

	MOVL	SI, R12 \ //Reset state for  AVX2 reg permutation
	MOVL	DI, SI \
	MOVL	DX, DI \
	MOVL	BX, DX \
	MOVL	CX, AX \
	MOVL	R12, CX \
	XCHGQ   R15, R14 \
	JMP     loop



TEXT ·blockAVX2(SB),$1408-32

	MOVQ	dig+0(FP),	DI
	MOVQ	p_base+8(FP),	SI
	MOVQ	p_len+16(FP),	DX
	SHRQ	$6,		DX
	SHLQ	$6,		DX

	MOVQ	$K_XMM_AR<>(SB), R8

	MOVQ	DI, R9
	MOVQ	SI, R10

DATA MASK_YMM_LO<>+0x08(SB)/8, $0x0000000000000000
DATA MASK_YMM_LO<>+0x10(SB)/8, $0xFFFFFFFFFFFFFFFF
DATA MASK_YMM_LO<>+0x18(SB)/8, $0xFFFFFFFFFFFFFFFF

GLOBL MASK_YMM_LO<>(SB), (NOPTR+RODATA), $32

TEXT ·blockAVX2(SB), NOSPLIT, $56-32
	MOVQ dig+0(FP), SI
	MOVQ p_base+8(FP), DI
	MOVQ p_len+16(FP), DX

	SHRQ $7, DX
	SHLQ $7, DX

	JZ   done_hash
	ADDQ DI, DX
	MOVQ DX, frame_INPEND(SP)