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

//go:build !math_big_pure_go && loong64

#include "textflag.h"

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

TEXT ·addVV(SB),NOSPLIT,$0
	JMP ·addVV_g(SB)

TEXT ·subVV(SB),NOSPLIT,$0
	JMP ·subVV_g(SB)

TEXT ·addVW(SB),NOSPLIT,$0
	JMP ·addVW_g(SB)

TEXT ·subVW(SB),NOSPLIT,$0

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

//go:build !math_big_pure_go && (mips64 || mips64le)

#include "textflag.h"

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

TEXT ·addVV(SB),NOSPLIT,$0
	JMP ·addVV_g(SB)

TEXT ·subVV(SB),NOSPLIT,$0
	JMP ·subVV_g(SB)

TEXT ·addVW(SB),NOSPLIT,$0
	JMP ·addVW_g(SB)

TEXT ·subVW(SB),NOSPLIT,$0

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

//go:build !math_big_pure_go && riscv64

#include "textflag.h"

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

TEXT ·addVV(SB),NOSPLIT,$0
	JMP ·addVV_g(SB)

TEXT ·subVV(SB),NOSPLIT,$0
	JMP ·subVV_g(SB)

TEXT ·addVW(SB),NOSPLIT,$0
	JMP ·addVW_g(SB)

TEXT ·subVW(SB),NOSPLIT,$0

	MOVD	$16, R10
	MOVD	$32, R14
	MOVD	$48, R15
	ANDCC	$63, R6, R9	// Check for tailing bytes for later
	PCALIGN $16
	// Case for >= 64 bytes
	// Process 64 bytes per iteration
	// Load 4 vectors of a and b
	// XOR the corresponding vectors
	// from a and b and store the result
loop64:
	LXVD2X	(R4)(R8), VS32
	LXVD2X	(R4)(R10), VS34
	LXVD2X	(R4)(R14), VS36
	LXVD2X	(R4)(R15), VS38
	LXVD2X	(R5)(R8), VS33
	LXVD2X	(R5)(R10), VS35

// 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.

// The carry bit is saved with SBBQ Rx, Rx: if the carry was set, Rx is -1, otherwise it is 0.
// It is restored with ADDQ Rx, Rx: if Rx was -1 the carry is set, otherwise it is cleared.

	MOVD	$32,R11		// set offsets to load into vector
	MOVD	$48,R12		// set offsets to load into vector

	PCALIGN	$16
cmp64_loop:
	LXVD2X	(R5)(R0),V3	// load bytes of A at offset 0 into vector
	LXVD2X	(R6)(R0),V4	// load bytes of B at offset 0 into vector
	VCMPEQUDCC	V3,V4,V1
	BGE	CR6,different	// jump out if its different

	LXVD2X	(R5)(R10),V3	// load bytes of A at offset 16 into vector

#ifndef GOPPC64_power9
#define ADJUST_FOR_CNTLZW -16
#else
#define ADJUST_FOR_CNTLZW 0
#endif

// Now, find the index of the 16B vector the match was discovered in. If CNTLZW is used
// to determine the offset into the 16B vector, it will overcount by 16. Account for it here.
foundat3:
	SUB	R3,R8,R3
	ADD	$48+ADJUST_FOR_CNTLZW,R3
	BR	vfound
foundat2:
	SUB	R3,R8,R3

	MOVD  plist_base+0(FP), R1  // r1 points to plist
	MOVD  bpx_offset+24(FP), R2 // r2 offset to BPX vector table
	MOVD  R14, R7               // save r14
	MOVD  R15, R8               // save r15
	MOVWZ 16(R0), R9
	MOVWZ 544(R9), R9
	MOVWZ 24(R9), R9            // call vector in r9
	ADD   R2, R9                // add offset to vector table
	MOVWZ (R9), R9              // r9 points to entry point