MOVD	h_base+0(FP), R0                           // Hash value first address
	MOVD	p_base+24(FP), R1                          // message first address
	MOVD	k_base+48(FP), R2                          // k constants first address
	MOVD	p_len+32(FP), R3                           // message length
	VLD1	(R0), [V0.S4, V1.S4]                       // load h(a,b,c,d,e,f,g,h)
	VLD1.P	64(R2), [V16.S4, V17.S4, V18.S4, V19.S4]
	VLD1.P	64(R2), [V20.S4, V21.S4, V22.S4, V23.S4]

// in the public domain.

//go:build !purego

#include "textflag.h"

// func block(dig *digest, p []byte)
TEXT ·block(SB),NOSPLIT,$16-32
	MOVD	dig+0(FP), R1
	MOVD	p+8(FP), R6
	MOVD	p_len+16(FP), R5
	AND	$-64, R5
	LAY	(R6)(R5*1), R7

	LMY	0(R1), R2, R5
	CMPBEQ	R6, R7, end

loop:
	STMY	R2, R5, tmp-16(SP)

	MOVWBR	0(R6), R8
	MOVWZ	R5, R9

// Author: Marc Bevand <bevand_m (at) epita.fr>
// Licence: I hereby disclaim the copyright on this code and place it
// in the public domain.

TEXT	·block(SB),NOSPLIT,$8-32
	MOVQ	dig+0(FP),	BP
	MOVQ	p+8(FP),	SI
	MOVQ	p_len+16(FP), DX
	SHRQ	$6,		DX
	SHLQ	$6,		DX

	LEAQ	(SI)(DX*1),	DI
	MOVL	(0*4)(BP),	AX
	MOVL	(1*4)(BP),	BX
	MOVL	(2*4)(BP),	CX
	MOVL	(3*4)(BP),	DX
	MOVL	$0xffffffff,	R11

	CMPQ	SI,		DI
	JEQ	end

	SHA512H	V7.D2, V6, i3 \
	VADD	i3.D2, i1.D2, i4.D2 \
	SHA512H2	i0.D2, i1, i3

// func blockAsm(dig *digest, p []byte)
TEXT ·blockAsm(SB),NOSPLIT,$0
	MOVD	dig+0(FP), R0
	MOVD	p_base+8(FP), R1
	MOVD	p_len+16(FP), R2
	MOVD	·_K+0(SB), R3

	// long enough to prefetch
	PRFM	(R3), PLDL3KEEP
	// load digest
	VLD1	(R0), [V8.D2, V9.D2, V10.D2, V11.D2]
loop:
	// load digest in V0-V3 keeping original in V8-V11

	SHUFFLE(Re)	; \
	FUNC4(Ra, Rb, Rc, Rd, Re)	; \
	MIX(Ra, Rb, Rc, Rd, Re)


// func block(dig *digest, p []byte)
TEXT	·block(SB), 0, $352-16
	MOVW	p+4(FP), Rdata	// pointer to the data
	MOVW	p_len+8(FP), Rt0	// number of bytes
	ADD	Rdata, Rt0
	MOVW	Rt0, p_end	// pointer to end of data

	// Load up initial SHA-1 accumulator
	MOVW	dig+0(FP), Rt0
	MOVM.IA (Rt0), [Ra,Rb,Rc,Rd,Re]

loop:

#define MAX_SIZE	32*1024
#define REFLECT

TEXT ·ppc64SlicingUpdateBy8(SB), NOSPLIT|NOFRAME, $0-44
	MOVWZ	crc+0(FP), R3   // incoming crc
	MOVD    table8+8(FP), R4   // *Table
	MOVD    p+16(FP), R5
	MOVD    p_len+24(FP), R6 // p len

	CMP     $0,R6           // len == 0?
	BNE     start
	MOVW    R3,ret+40(FP)   // return crc
	RET

start:
	NOR     R3,R3,R7        // ^crc
	MOVWZ	R7,R7		// 32 bits
	CMP	R6,$16

#define SHA256ROUND1(index, const, a, b, c, d, e, f, g, h) \
	MSGSCHEDULE1(index); \
	SHA256ROUND(index, const, a, b, c, d, e, f, g, h)

TEXT ·block(SB),0,$296-16
	MOVL	p_base+4(FP), SI
	MOVL	p_len+8(FP), DX
	SHRL	$6, DX
	SHLL	$6, DX

	LEAL	(SI)(DX*1), DI
	MOVL	DI, 288(SP)
	CMPL	SI, DI
	JEQ	end

	LEAL	256(SP), DI		// variables

	MOVL	dig+0(FP), BP
	MOVL	(0*4)(BP), AX		// a = H0

	SHUFFLE(index, e); \
	FUNC4(a, b, c, d, e); \
	MIX(a, b, c, d, e, 0xCA62C1D6)

// func block(dig *digest, p []byte)
TEXT ·block(SB),NOSPLIT,$92-16
	MOVL	dig+0(FP),	BP
	MOVL	p+4(FP),	SI
	MOVL	p_len+8(FP),	DX
	SHRL	$6,		DX
	SHLL	$6,		DX

	LEAL	(SI)(DX*1),	DI
	MOVL	(0*4)(BP),	AX
	MOVL	(1*4)(BP),	BX
	MOVL	(2*4)(BP),	CX
	MOVL	(3*4)(BP),	DX
	MOVL	(4*4)(BP),	BP

	CMPL	SI,		DI
	JEQ	end

#define SHA256ROUND1(index, const, a, b, c, d, e, f, g, h) \
	MSGSCHEDULE1(index); \
	SHA256ROUND(index, const, a, b, c, d, e, f, g, h)

TEXT ·block(SB),0,$296-16
	MOVL	p_base+4(FP), SI
	MOVL	p_len+8(FP), DX
	SHRL	$6, DX
	SHLL	$6, DX

	LEAL	(SI)(DX*1), DI
	MOVL	DI, 288(SP)
	CMPL	SI, DI
	JEQ	end

	LEAL	256(SP), DI		// variables

	MOVL	dig+0(FP), BP
	MOVL	(0*4)(BP), AX		// a = H0

	case asmSlice:
		cc = append(cc, newComponent(suffix+"_base", asmKind(arch.ptrSize), "slice base", off, arch.ptrSize, suffix))
		cc = append(cc, newComponent(suffix+"_len", asmKind(arch.intSize), "slice len", off+arch.ptrSize, arch.intSize, suffix))
		cc = append(cc, newComponent(suffix+"_cap", asmKind(arch.intSize), "slice cap", off+arch.ptrSize+arch.intSize, arch.intSize, suffix))