// CLWB instructions:
	CLWB BX                          // ERROR "invalid instruction"
	// CLDEMOTE instructions:
	CLDEMOTE BX                      // ERROR "invalid instruction"
	// WAITPKG instructions:
	TPAUSE (BX)                      // ERROR "invalid instruction"
	UMONITOR (BX)                    // ERROR "invalid instruction"
	UMWAIT (BX)                      // ERROR "invalid instruction"

	RET()
}

// mul64 sets r to i * aX * bX.
func mul64(r uint128, i int, aX, bX namedComponent) {
	switch i {
	case 1:
		Comment(fmt.Sprintf("%s = %s×%s", r, aX, bX))
		Load(aX, RAX)
	case 2:
		Comment(fmt.Sprintf("%s = 2×%s×%s", r, aX, bX))
		Load(aX, RAX)
		SHLQ(Imm(1), RAX)
	default:
		panic("unsupported i value")
	}
	MULQ(mustAddr(bX)) // RDX, RAX = RAX * bX
	MOVQ(RAX, r.lo)
	MOVQ(RDX, r.hi)
}

TEXT ·xorBytes(SB), NOSPLIT, $0
	MOVQ  dst+0(FP), BX
	MOVQ  a+8(FP), SI
	MOVQ  b+16(FP), CX
	MOVQ  n+24(FP), DX
	TESTQ $15, DX            // AND 15 & len, if not zero jump to not_aligned.
	JNZ   not_aligned

aligned:
	MOVQ $0, AX // position in slices

	PCALIGN $16
loop16b:
	MOVOU (SI)(AX*1), X0   // XOR 16byte forwards.
	MOVOU (CX)(AX*1), X1
	PXOR  X1, X0
	MOVOU X0, (BX)(AX*1)
	ADDQ  $16, AX
	CMPQ  DX, AX

	MOVQ	_cgo_callers(SB), AX
	JMP	AX

TEXT runtime·mmap_trampoline(SB),NOSPLIT,$0
	MOVQ	DI, BX
	MOVQ	0(BX), DI		// arg 1 addr
	MOVQ	8(BX), SI		// arg 2 len
	MOVL	16(BX), DX		// arg 3 prot
	MOVL	20(BX), CX		// arg 4 flags
	MOVL	24(BX), R8		// arg 5 fid
	MOVL	28(BX), R9		// arg 6 offset
	CALL	libc_mmap(SB)
	XORL	DX, DX
	CMPQ	AX, $-1
	JNE	ok
	CALL	libc_error(SB)

	VPDPBUSD -15(BX), X16, K2, X6                      // 62f27d0250b3f1ffffff
	VPDPBUSD X15, X28, K2, X6                          // 62d21d0250f7
	VPDPBUSD X11, X28, K2, X6                          // 62d21d0250f3
	VPDPBUSD X1, X28, K2, X6                           // 62f21d0250f1
	VPDPBUSD -15(R14)(R15*1), X28, K2, X6              // 62921d0250b43ef1ffffff
	VPDPBUSD -15(BX), X28, K2, X6                      // 62f21d0250b3f1ffffff

				{1, 239}, // AX CX DX BX BP SI DI
			},
			outputs: []outputInfo{
				{0, 239}, // AX CX DX BX BP SI DI
			},
		},
	},
	{
		name:         "ADCLconst",
		auxType:      auxInt32,
		argLen:       2,
		resultInArg0: true,
		clobberFlags: true,
		asm:          x86.AADCL,
		reg: regInfo{
			inputs: []inputInfo{
				{0, 239}, // AX CX DX BX BP SI DI
			},

	VGF2P8AFFINEINVQB $27, (BX), Y20, K3, Y0           // 62f3dd23cf031b
	VGF2P8AFFINEINVQB $27, -17(BP)(SI*1), Y20, K3, Y0  // 62f3dd23cf8435efffffff1b
	VGF2P8AFFINEINVQB $27, Y5, Y12, K3, Y0             // 62f39d2bcfc51b
	VGF2P8AFFINEINVQB $27, Y28, Y12, K3, Y0            // 62939d2bcfc41b
	VGF2P8AFFINEINVQB $27, Y7, Y12, K3, Y0             // 62f39d2bcfc71b
	VGF2P8AFFINEINVQB $27, (BX), Y12, K3, Y0           // 62f39d2bcf031b

//go:build gc

#include "textflag.h"

TEXT ·RewindAndSetgid(SB),NOSPLIT,$0-0
	MOVL	$·Baton(SB), BX
	// Rewind stack pointer so anything that happens on the stack
	// will clobber the test pattern created by the caller
	ADDL	$(1024 * 8), SP

	// Ask signaller to setgid
	MOVL	$1, (BX)

	// Wait for setgid completion
loop:
	PAUSE
	MOVL	(BX), AX
	CMPL	AX, $0
	JNE	loop

	// Restore stack
	SUBL	$(1024 * 8), SP

#define PUSH_REGS_HOST_TO_ABI0()	\
	ADJSP	$(REGS_HOST_TO_ABI0_STACK)	\
	MOVQ	BP, (5*8)(SP)	\
	LEAQ	(5*8)(SP), BP	\
	MOVQ	BX, (0*8)(SP)	\
	MOVQ	R12, (1*8)(SP)	\
	MOVQ	R13, (2*8)(SP)	\
	MOVQ	R14, (3*8)(SP)	\
	MOVQ	R15, (4*8)(SP)

#define POP_REGS_HOST_TO_ABI0()	\
	MOVQ	(0*8)(SP), BX	\
	MOVQ	(1*8)(SP), R12	\
	MOVQ	(2*8)(SP), R13	\
	MOVQ	(3*8)(SP), R14	\
	MOVQ	(4*8)(SP), R15	\
	MOVQ	(5*8)(SP), BP	\

	MOVQ	g_m(R14), BX // BX unchanged by C code.

	// Set vdsoPC and vdsoSP for SIGPROF traceback.
	// Save the old values on stack and restore them on exit,
	// so this function is reentrant.
	MOVQ	m_vdsoPC(BX), CX
	MOVQ	m_vdsoSP(BX), DX
	MOVQ	CX, 0(SP)
	MOVQ	DX, 8(SP)

	LEAQ	ret+0(FP), DX
	MOVQ	-8(DX), CX
	MOVQ	CX, m_vdsoPC(BX)
	MOVQ	DX, m_vdsoSP(BX)