// 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.

// func addVV(z, x, y []Word) (c Word)
TEXT ·addVV(SB),NOSPLIT,$0
	ADD.S	$0, R0		// clear carry flag
	MOVW	z+0(FP), R1
	MOVW	z_len+4(FP), R4
	MOVW	x+12(FP), R2
	MOVW	y+24(FP), R3

	auxvp := (*[1 << 28]uintptr)(add(unsafe.Pointer(argv), uintptr(n)*goarch.PtrSize))
	pairs := sysauxv(auxvp[:])
	auxv = auxvp[: pairs*2 : pairs*2]
}

const (
	_AT_NULL   = 0 // Terminates the vector
	_AT_PAGESZ = 6 // Page size in bytes
)

func sysauxv(auxv []uintptr) (pairs int) {
	var i int
	for i = 0; auxv[i] != _AT_NULL; i += 2 {
		tag, val := auxv[i], auxv[i+1]
		switch tag {

#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

	BLT	tail		// Jump to the small string case
	SRD	$5, R4, R20
	MOVD	R20, CTR
	MOVD	$16, R21
	XXLXOR	V4, V4, V4
	XXLXOR	V5, V5, V5

	PCALIGN	$16
cmploop:
	LXVD2X	(R0)(R3), V0	// Count 32B per loop with two vector accumulators.
	LXVD2X	(R21)(R3), V2
	VCMPEQUB V2, V1, V2
	VCMPEQUB V0, V1, V0
	VPOPCNTD V2, V2		// A match is 0xFF or 0. Count the bits into doubleword buckets.
	VPOPCNTD V0, V0

	MOVL (AX)(X0*1), AX             // ERROR "invalid instruction"
	MOVL (AX)(Y0*1), AX             // ERROR "invalid instruction"
	// VSIB/VM is invalid without vector index.
	// TODO(quasilyte): improve error message (#21860).
	// "invalid VSIB address (missing vector index)"
	VPGATHERQQ Y2, (BP), Y1         // ERROR "invalid instruction"
	// AVX2GATHER mask/index/dest #UD cases.

		fe.reduce()
		r.reduce()
		return fe == r
	}
	if err := quick.Check(f2, nil); err != nil {
		t.Errorf("failed FE->bytes->FE round-trip: %v", err)
	}

	// Check some fixed vectors from dalek
	type feRTTest struct {
		fe Element
		b  []byte
	}
	var tests = []feRTTest{
		{
			fe: Element{358744748052810, 1691584618240980, 977650209285361, 1429865912637724, 560044844278676},

	CBitFieldMaskBit62 = 0x2
	CBitFieldMaskBit63 = 0x1
)

type SockaddrStorage struct {
	Family uint16
	Data   [122]byte
	_      uint32
}

type HDGeometry struct {
	Heads     uint8
	Sectors   uint8
	Cylinders uint16
	Start     uint32
}

type Statfs_t struct {
	Type    int32
	Bsize   int32
	Blocks  uint64
	Bfree   uint64
	Bavail  uint64
	Files   uint64
	Ffree   uint64

	CBitFieldMaskBit62 = 0x2
	CBitFieldMaskBit63 = 0x1
)

type SockaddrStorage struct {
	Family uint16
	Data   [118]byte
	_      uint64
}

type HDGeometry struct {
	Heads     uint8
	Sectors   uint8
	Cylinders uint16
	Start     uint64
}

type Statfs_t struct {
	Type    int64
	Bsize   int64
	Blocks  uint64
	Bfree   uint64
	Bavail  uint64
	Files   uint64
	Ffree   uint64

	CBitFieldMaskBit63 = 0x8000000000000000
)

type SockaddrStorage struct {
	Family uint16
	Data   [118]byte
	_      uint64
}

type HDGeometry struct {
	Heads     uint8
	Sectors   uint8
	Cylinders uint16
	Start     uint64
}

type Statfs_t struct {
	Type    int64
	Bsize   int64
	Blocks  uint64
	Bfree   uint64
	Bavail  uint64
	Files   uint64
	Ffree   uint64

	CBitFieldMaskBit63 = 0x8000000000000000
)

type SockaddrStorage struct {
	Family uint16
	Data   [122]byte
	_      uint32
}

type HDGeometry struct {
	Heads     uint8
	Sectors   uint8
	Cylinders uint16
	Start     uint32
}

type Statfs_t struct {
	Type    int32
	Bsize   int32
	Blocks  uint64
	Bfree   uint64
	Bavail  uint64
	Files   uint64
	Ffree   uint64