TEXT ·p256OrdBigToLittle(SB), NOSPLIT, $0-16
	MOVD	res+0(FP), R3
	MOVD	in+8(FP), R4
	BR	p256InternalEndianSwap<>(SB)

//func p256OrdLittleToBig(res *[32]byte, in *p256OrdElement)
TEXT ·p256OrdLittleToBig(SB), NOSPLIT, $0-16
	MOVD	res+0(FP), R3
	MOVD	in+8(FP), R4
	BR	p256InternalEndianSwap<>(SB)

TEXT p256InternalEndianSwap<>(SB), NOSPLIT, $0-0
	// Index registers needed for BR movs
	MOVD	$8, R9

TEXT ·p256OrdLittleToBig(SB), NOSPLIT, $0
	JMP ·p256BigToLittle(SB)

// func p256OrdBigToLittle(res *p256OrdElement, in *[32]byte)
TEXT ·p256OrdBigToLittle(SB), NOSPLIT, $0
	JMP ·p256BigToLittle(SB)

// ---------------------------------------
// func p256LittleToBig(res *[32]byte, in *p256Element)
TEXT ·p256LittleToBig(SB), NOSPLIT, $0
	JMP ·p256BigToLittle(SB)

//       doc: https://www.ibm.com/support/knowledgecenter/en/SSLTBW_3.1.0/com.ibm.zos.v3r1.bpxb100/bpx2cr_List_of_offsets.htm
//
// func bpxcall(plist []unsafe.Pointer, bpx_offset int64)

TEXT ·bpxcall(SB), NOSPLIT|NOFRAME, $0
	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

//
// Do not remove or change the type signature.
// See go.dev/issue/67401.
//
//go:linkname overrideWrite
var overrideWrite func(fd uintptr, p unsafe.Pointer, n int32) int32

// write must be nosplit on Windows (see write1)
//
//go:nosplit
func write(fd uintptr, p unsafe.Pointer, n int32) int32 {
	if overrideWrite != nil {
		return overrideWrite(fd, noescape(p), n)
	}
	return write1(fd, p, n)

	if state.stopped {
		// We stopped it, so we need to re-schedule it.
		ready(gp, 0, true)
	}
}

// canPreemptM reports whether mp is in a state that is safe to preempt.
//
// It is nosplit because it has nosplit callers.
//
//go:nosplit
func canPreemptM(mp *m) bool {
	return mp.locks == 0 && mp.mallocing == 0 && mp.preemptoff == "" && mp.p.ptr().status == _Prunning
}

//go:generate go run mkpreempt.go

DATA invSRows<>+0x00(SB)/8, $0x0b0e0104070a0d00
DATA invSRows<>+0x08(SB)/8, $0x0306090c0f020508
GLOBL invSRows<>(SB), (NOPTR+RODATA), $16
// func encryptBlockAsm(nr int, xk *uint32, dst, src *byte)
TEXT ·encryptBlockAsm(SB),NOSPLIT,$0
	MOVD	nr+0(FP), R9
	MOVD	xk+8(FP), R10
	MOVD	dst+16(FP), R11
	MOVD	src+24(FP), R12

	VLD1	(R12), [V0.B16]

	CMP	$12, R9
	BLT	enc128
	BEQ	enc196
enc256:
	VLD1.P	32(R10), [V1.B16, V2.B16]

//
// It returns an iterator that tiles typ.GCData starting from addr. It's the caller's
// responsibility to limit iteration.
//
// nosplit because its callers are nosplit and require all their callees to be nosplit.
//
//go:nosplit
func (span *mspan) typePointersOfType(typ *abi.Type, addr uintptr) typePointers {
	const doubleCheck = false

	// gp and deadline are logically local variables, but they are written
	// as parameters so that the stack space they require is charged
	// to the caller.
	// This reduces the nosplit footprint of notetsleep_internal.
	gp = getg()

	// Register for wakeup on n->waitm.
	if !atomic.Casuintptr(&n.key, 0, uintptr(unsafe.Pointer(gp.m))) {
		// Must be locked (got wakeup).
		if n.key != locked {

#include "textflag.h"

// See memmove Go doc for important implementation constraints.

// func memmove(to, from unsafe.Pointer, n uintptr)
TEXT runtime·memmove<ABIInternal>(SB), NOSPLIT|NOFRAME, $0-24
	BNE	R6, check
	RET

check:
	SGTU	R4, R5, R7
	BNE	R7, backward

	ADDV	R4, R6, R9 // end pointer

	// if the two pointers are not of same alignments, do byte copying
	SUBVU	R5, R4, R7

//go:build !purego

#include "textflag.h"

// func feMul(out *Element, a *Element, b *Element)
TEXT ·feMul(SB), NOSPLIT, $0-24
	MOVQ a+8(FP), CX
	MOVQ b+16(FP), BX

	// r0 = a0×b0
	MOVQ (CX), AX
	MULQ (BX)
	MOVQ AX, DI
	MOVQ DX, SI

	// r0 += 19×a1×b4
	MOVQ   8(CX), AX
	IMUL3Q $0x13, AX, AX
	MULQ   32(BX)