// license that can be found in the LICENSE file.

//go:build !math_big_pure_go

#include "textflag.h"

// This file provides fast assembly versions for the elementary
// arithmetic operations on vectors implemented in arith.go.

// DI = R3, CX = R4, SI = r10, r8 = r8, r9=r9, r10 = r2, r11 = r5, r12 = r6, r13 = r7, r14 = r1 (R0 set to 0) + use R11
// func addVV(z, x, y []Word) (c Word)

TEXT ·addVV(SB), NOSPLIT, $0

#ifdef GOARCH_ppc64le
#define VPERMLE(va,vb,vc,vt) VPERM va, vb, vc, vt
#else
#define VPERMLE(va,vb,vc,vt)
#endif

// 2 copies of each Kt, to fill both doublewords of a vector register
DATA  ·kcon+0x000(SB)/8, $0x428a2f98d728ae22
DATA  ·kcon+0x008(SB)/8, $0x428a2f98d728ae22
DATA  ·kcon+0x010(SB)/8, $0x7137449123ef65cd
DATA  ·kcon+0x018(SB)/8, $0x7137449123ef65cd

DATA  ·kcon+0x418(SB)/8, $0x1011121300010203
DATA  ·kcon+0x420(SB)/8, $0x1011121310111213
DATA  ·kcon+0x428(SB)/8, $0x0405060700010203
DATA  ·kcon+0x430(SB)/8, $0x1011121308090a0b
DATA  ·kcon+0x438(SB)/8, $0x0405060700010203
#else
DATA  ·kcon+0x410(SB)/8, $0x1011121300010203
DATA  ·kcon+0x418(SB)/8, $0x1011121310111213 // permutation control vectors
DATA  ·kcon+0x420(SB)/8, $0x0405060700010203

// license that can be found in the LICENSE file.

//go:build !math_big_pure_go

#include "textflag.h"

// This file provides fast assembly versions for the elementary
// arithmetic operations on vectors implemented in arith.go.

// TODO: Consider re-implementing using Advanced SIMD
// once the assembler supports those instructions.

// func addVV(z, x, y []Word) (c Word)
TEXT ·addVV(SB),NOSPLIT,$0

#define m1		X4
#define m2		X5
#define m3		X6
#define m4		X7
#define shufMask	X8	// input data endian conversion control mask
#define abefSave	X9	// digest hash vector inter-block buffer abef
#define cdghSave	X10	// digest hash vector inter-block buffer cdgh

#define nop(m,a)		// nop instead of final SHA256MSG1 for first and last few rounds

#define sha256msg1(m,a) \	// final SHA256MSG1 for middle rounds that require it

	KLMD	R2, R8                 // b93f0028
	KIMD	R0, R4                 // b93e0004
	KDSA	R0, R8                 // b93a0008
	KMA	R2, R6, R4              // b9296024
	KMCTR   R2, R6, R4              // b92d6024

	// vector add and sub instructions
	VAB	V3, V4, V4              // e743400000f3
	VAH	V3, V4, V4              // e743400010f3
	VAF	V3, V4, V4              // e743400020f3
	VAG	V3, V4, V4              // e743400030f3

// From http://software.intel.com/en-us/articles
// (look for improving-the-performance-of-the-secure-hash-algorithm-1)
// This implementation is 2x unrolled, and interleaves vector instructions,
// used to precompute W, with scalar computation of current round
// for optimal scheduling.

// Trivial helper macros.
#define UPDATE_HASH(A,TB,C,D,E) \
	ADDL	(R9), A \
	MOVL	A, (R9) \

DATA  ·kcon+0x418(SB)/8, $0x1011121300010203
DATA  ·kcon+0x420(SB)/8, $0x1011121310111213
DATA  ·kcon+0x428(SB)/8, $0x0405060700010203
DATA  ·kcon+0x430(SB)/8, $0x1011121308090a0b
DATA  ·kcon+0x438(SB)/8, $0x0405060700010203
#else
DATA  ·kcon+0x410(SB)/8, $0x1011121300010203
DATA  ·kcon+0x418(SB)/8, $0x1011121310111213 // permutation control vectors
DATA  ·kcon+0x420(SB)/8, $0x0405060700010203

	VSPLTB $2, V0, V8   // Splat 3rd byte of sep

	// Loop to process 3 byte separator.
	// string[0:16] is in V2
	// string[2:18] is in V3
	// sep[0:2] splatted in V1
	// sec[3] splatted in v8
	// Load vectors at string, string+1
	// and string+2. Compare string, string+1
	// against first 2 bytes of separator
	// splatted, and string+2 against 3rd
	// byte splatted. Merge the results with

// For {en,de}cryptBlockAsm
#define BLK_INP    R3
#define BLK_OUT    R4
#define BLK_KEY    R5
#define BLK_ROUNDS R6
#define BLK_IDX    R7

DATA ·rcon+0x00(SB)/8, $0x0f0e0d0c0b0a0908 // Permute for vector doubleword endian swap
DATA ·rcon+0x08(SB)/8, $0x0706050403020100
DATA ·rcon+0x10(SB)/8, $0x0100000001000000 // RCON
DATA ·rcon+0x18(SB)/8, $0x0100000001000000 // RCON
DATA ·rcon+0x20(SB)/8, $0x1b0000001b000000