// Atomically:
//	if(*val == old){
//		*val = new;
//		return 1;
//	} else
//		return 0;
TEXT ·Casp1(SB), NOSPLIT, $0-25
	JMP	·Cas64(SB)

// uint32 xadd(uint32 volatile *ptr, int32 delta)
// Atomically:
//	*val += delta;
//	return *val;
TEXT ·Xadd(SB), NOSPLIT, $0-20
	MOVV	ptr+0(FP), R4
	MOVW	delta+8(FP), R5
	DBAR
	LL	(R4), R6
	ADDU	R6, R5, R7
	MOVV	R7, R6
	SC	R7, (R4)
	BEQ	R7, -4(PC)

		}
	}
}

//go:nosplit
func And64(addr *uint64, val uint64) (old uint64) {
	for {
		old = *addr
		if Cas64(addr, old, old&val) {
			return old
		}
	}
}

//go:noescape
func Xadd(ptr *uint32, delta int32) uint32

//go:noescape
func Xadduintptr(ptr *uintptr, delta uintptr) uintptr

//go:noescape
func Xchg(ptr *uint32, new uint32) uint32

//go:noescape

		for pb.Next() {
			atomic.Or(&x[63], i)
			i++
		}
	})
}

func BenchmarkXadd(b *testing.B) {
	var x uint32
	ptr := &x
	b.RunParallel(func(pb *testing.PB) {
		for pb.Next() {
			atomic.Xadd(ptr, 1)
		}
	})
}

func BenchmarkXadd64(b *testing.B) {
	var x uint64
	ptr := &x
	b.RunParallel(func(pb *testing.PB) {
		for pb.Next() {
			atomic.Xadd64(ptr, 1)
		}
	})
}

		if ret == _EWOULDBLOCK {
			return -1
		}
	}
}

//go:nosplit
func semawakeup(mp *m) {
	atomic.Xadd(&mp.waitsemacount, 1)
	ret := thrwakeup(uintptr(unsafe.Pointer(&mp.waitsemacount)), 1)
	if ret != 0 && ret != _ESRCH {
		// semawakeup can be called on signal stack.
		systemstack(func() {

	VADD	V18.S4, V4.S4, V16.S4
	HASHUPDATEPARITY

	VADD	V18.S4, V5.S4, V16.S4
	HASHUPDATEPARITY

	VADD	V18.S4, V6.S4, V16.S4
	HASHUPDATEPARITY

	VADD	V18.S4, V7.S4, V16.S4
	HASHUPDATEPARITY

	SUB	$64, R3, R3                                  // message length - 64bytes, then compare with 64bytes
	VADD	V2.S4, V0.S4, V0.S4
	VADD	V1.S4, V20.S4, V20.S4
	CBNZ	R3, blockloop

sha1ret:

	SHA512ROUND_LAST(V2, V3, V1, V4, V0, V25, V17)
	SHA512ROUND_LAST(V4, V2, V0, V1, V3, V26, V18)
	SHA512ROUND_LAST(V1, V4, V3, V0, V2, V27, V19)

	// add result to digest
	VADD	V0.D2, V8.D2, V8.D2
	VADD	V1.D2, V9.D2, V9.D2
	VADD	V2.D2, V10.D2, V10.D2
	VADD	V3.D2, V11.D2, V11.D2
	SUB	$128, R2
	CBNZ	R2, loop

	VST1	[V8.D2, V9.D2, V10.D2, V11.D2], (R0)

	AND $0x7ffffffffffff, R1, R11
	AND $0x7ffffffffffff, R2, R12
	AND $0x7ffffffffffff, R3, R13
	AND $0x7ffffffffffff, R4, R14

	ADD R0>>51, R11, R11
	ADD R1>>51, R12, R12
	ADD R2>>51, R13, R13
	ADD R3>>51, R14, R14
	// R4>>51 * 19 + R10 -> R10
	LSR $51, R4, R21
	MOVD $19, R22
	MADD R22, R10, R21, R10

	STP (R10, R11), 0(R20)
	STP (R12, R13), 16(R20)
	MOVD R14, 32(R20)

Jeremy L. Morris <******@****.***> 1717516806 -0400

my-git9 <******@****.***> 1708354813 +0800

//    Strength-reduce    //
// --------------------- //

func Mul2(f float64) float64 {
	// 386/sse2:"ADDSD",-"MULSD"
	// amd64:"ADDSD",-"MULSD"
	// arm/7:"ADDD",-"MULD"
	// arm64:"FADDD",-"FMULD"
	// ppc64x:"FADD",-"FMUL"
	// riscv64:"FADDD",-"FMULD"
	return f * 2.0
}

func DivPow2(f1, f2, f3 float64) (float64, float64, float64) {
	// 386/sse2:"MULSD",-"DIVSD"
	// amd64:"MULSD",-"DIVSD"
	// arm/7:"MULD",-"DIVD"