// 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

	op_VCKSM  uint32 = 0xE766 // 	VRR-c	VECTOR CHECKSUM
	op_VCEQ   uint32 = 0xE7F8 // 	VRR-b	VECTOR COMPARE EQUAL
	op_VCH    uint32 = 0xE7FB // 	VRR-b	VECTOR COMPARE HIGH
	op_VCHL   uint32 = 0xE7F9 // 	VRR-b	VECTOR COMPARE HIGH LOGICAL
	op_VCLZ   uint32 = 0xE753 // 	VRR-a	VECTOR COUNT LEADING ZEROS
	op_VCTZ   uint32 = 0xE752 // 	VRR-a	VECTOR COUNT TRAILING ZEROS
	op_VEC    uint32 = 0xE7DB // 	VRR-a	VECTOR ELEMENT COMPARE

	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

	/* Vector instructions */

	/* Vector load */
	{as: ALVEBX, a1: C_XOREG, a6: C_VREG, type_: 45, size: 4}, /* vector load, x-form */

	/* Vector store */
	{as: ASTVEBX, a1: C_VREG, a6: C_XOREG, type_: 44, size: 4}, /* vector store, x-form */

	/* Vector logical */
	{as: AVAND, a1: C_VREG, a2: C_VREG, a6: C_VREG, type_: 82, size: 4}, /* vector and, vx-form */

package chacha20

import "golang.org/x/sys/cpu"

var haveAsm = cpu.S390X.HasVX

const bufSize = 256

// xorKeyStreamVX is an assembly implementation of XORKeyStream. It must only
// be called when the vector facility is available. Implementation in asm_s390x.s.
//
//go:noescape
func xorKeyStreamVX(dst, src []byte, key *[8]uint32, nonce *[3]uint32, counter *uint32)

func (c *Cipher) xorKeyStreamBlocks(dst, src []byte) {

	hwcap_SHA2    = 1 << 6
	hwcap_CRC32   = 1 << 7
	hwcap_ATOMICS = 1 << 8
	hwcap_CPUID   = 1 << 11
	hwcap_SHA512  = 1 << 21
)

func hwcapInit(os string) {
	// HWCap was populated by the runtime from the auxiliary vector.
	// Use HWCap information since reading aarch64 system registers
	// is not supported in user space on older linux kernels.
	ARM64.HasAES = isSet(HWCap, hwcap_AES)
	ARM64.HasPMULL = isSet(HWCap, hwcap_PMULL)

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package runtime

import "internal/cpu"

const (
	_HWCAP_VX = 1 << 11 // vector facility
)

func archauxv(tag, val uintptr) {
	switch tag {
	case _AT_HWCAP:
		cpu.HWCap = uint(val)
	}
}

func osArchInit() {}

func checkS390xCPU() {

	MOVQ	R9, 24(SP)
	CALL	runtime·sehhandler(SB)
	MOVL	32(SP), AX

	ADJSP	$-40

	POP_REGS_HOST_TO_ABI0()
	RET

TEXT runtime·callbackasm1(SB),NOSPLIT|NOFRAME,$0
	// Construct args vector for cgocallback().
	// By windows/amd64 calling convention first 4 args are in CX, DX, R8, R9
	// args from the 5th on are on the stack.
	// In any case, even if function has 0,1,2,3,4 args, there is reserved

	ROUND4(X0, X5, X15, X10, X1, X6, X12, X11, X2, X7, X13, X8,  X3, X4, X14, X9)

	ADD $-1, R1
	BNE loop

	// decrement length
	ADD $-256, R4

	// rearrange vectors
	SHUFFLE(X0, X1, X2, X3, M0, M1, M2, M3)
	ADDV(J0, X0, X1, X2, X3)
	SHUFFLE(X4, X5, X6, X7, M0, M1, M2, M3)
	ADDV(KEY0, X4, X5, X6, X7)
	SHUFFLE(X8, X9, X10, X11, M0, M1, M2, M3)
	ADDV(KEY1, X8, X9, X10, X11)

	return dk.EncapsulationKey(), dk
}

var millionFlag = flag.Bool("million", false, "run the million vector test")

// TestPQCrystalsAccumulated accumulates the 10k vectors generated by the
// reference implementation and checks the hash of the result, to avoid checking
// in 150MB of test vectors.
func TestPQCrystalsAccumulated(t *testing.T) {
	n := 10000