CLD

	// determine index into runtime·cbs table
	SUBL	$runtime·callbackasm(SB), AX
	MOVL	$0, DX
	MOVL	$5, BX	// divide by 5 because each call instruction in runtime·callbacks is 5 bytes long
	DIVL	BX
	SUBL	$1, AX	// subtract 1 because return PC is to the next slot

	// Create a struct callbackArgs on our stack.
	SUBL	$(12+callbackArgs__size), SP
	MOVL	AX, (12+callbackArgs_index)(SP)		// callback index

	MOVL $0, BX		// i = 0
	JMP E4

L4:	MOVL (SI)(BX*4), DX
	SUBL AX, DX
	MOVL DX, (DI)(BX*4)
	SBBL AX, AX		// save CF
	NEGL AX
	ADDL $1, BX		// i++

E4:	CMPL BX, BP		// i < n
	JL L4

	MOVL AX, c+28(FP)
	RET


// func shlVU(z, x []Word, s uint) (c Word)
TEXT ·shlVU(SB),NOSPLIT,$0
	MOVL z_len+4(FP), BX	// i = z
	SUBL $1, BX		// i--
	JL X8b			// i < 0	(n <= 0)

	// n > 0

	MOVL	(_SYSTEM_TIME+time_lo), AX
	MOVL	(_SYSTEM_TIME+time_hi2), DX
	CMPL	CX, DX
	JNE	wall

	// w = DX:AX
	// convert to Unix epoch (but still 100ns units)
	#define delta 116444736000000000
	SUBL	$(delta & 0xFFFFFFFF), AX
	SBBL $(delta >> 32), DX

	// nano/100 = DX:AX
	// split into two decimal halves by div 1e9.
	// (decimal point is two spots over from correct place,
	// but we avoid overflow in the high word.)

func SubMem(arr []int, b, c, d int) int {
	// 386:`SUBL\s[A-Z]+,\s8\([A-Z]+\)`
	// amd64:`SUBQ\s[A-Z]+,\s16\([A-Z]+\)`
	arr[2] -= b
	// 386:`SUBL\s[A-Z]+,\s12\([A-Z]+\)`
	// amd64:`SUBQ\s[A-Z]+,\s24\([A-Z]+\)`
	arr[3] -= b
	// 386:`DECL\s16\([A-Z]+\)`
	arr[4]--
	// 386:`ADDL\s[$]-20,\s20\([A-Z]+\)`
	arr[5] -= 20
	// 386:`SUBL\s\([A-Z]+\)\([A-Z]+\*4\),\s[A-Z]+`
	ef -= arr[b]
	// 386:`SUBL\s[A-Z]+,\s\([A-Z]+\)\([A-Z]+\*4\)`
	arr[c] -= b

	get_tls(CX)
	MOVL	SI, g(CX)
	MOVL	(g_sched+gobuf_sp)(SI), SP

noswitch:
	// Now on a scheduling stack (a pthread-created stack).
	SUBL	$32, SP
	ANDL	$~15, SP	// alignment, perhaps unnecessary
	MOVL	DI, 8(SP)	// save g
	MOVL	(g_stack+stack_hi)(DI), DI
	SUBL	DX, DI
	MOVL	DI, 4(SP)	// save depth in stack (can't just save SP, as stack might be copied during a callback)
	MOVL	BX, 0(SP)	// first argument in x86-32 ABI

	s := x[0]
	// 386: `ADDL\t4\([A-Z]+\)\([A-Z]+\*4\), [A-Z]+`
	// amd64: `ADDL\t4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\), [A-Z]+[0-9]*`
	s += x[i+1]
	// 386: `SUBL\t8\([A-Z]+\)\([A-Z]+\*4\), [A-Z]+`
	// amd64: `SUBL\t8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\), [A-Z]+[0-9]*`
	s -= x[i+2]
	// 386: `IMULL\t12\([A-Z]+\)\([A-Z]+\*4\), [A-Z]+`
	s *= x[i+3]
	// 386: `ANDL\t16\([A-Z]+\)\([A-Z]+\*4\), [A-Z]+`

end
"""
        } else {
            return """
.text

LC0:
.ascii "fooFromAsm\\12\\0"
.globl _sayFooFromAsm

_sayFooFromAsm:
        pushl   %ebp
        movl    %esp, %ebp
        subl    \$8, %esp
        andl    \$-16, %esp
        movl    \$0, %eax
        movl    %eax, -4(%ebp)
        movl    -4(%ebp), %eax
        movl    \$LC0, (%esp)

		{name: "SUBQ", argLength: 2, reg: gp21, asm: "SUBQ", resultInArg0: true, clobberFlags: true},
		{name: "SUBL", argLength: 2, reg: gp21, asm: "SUBL", resultInArg0: true, clobberFlags: true},
		{name: "SUBQconst", argLength: 1, reg: gp11, asm: "SUBQ", aux: "Int32", resultInArg0: true, clobberFlags: true},
		{name: "SUBLconst", argLength: 1, reg: gp11, asm: "SUBL", aux: "Int32", resultInArg0: true, clobberFlags: true},

(ADDL x (MOVLconst [c])) => (ADDLconst [c] x)

(SUBQ x (MOVQconst [c])) && is32Bit(c) => (SUBQconst x [int32(c)])
(SUBQ (MOVQconst [c]) x) && is32Bit(c) => (NEGQ (SUBQconst <v.Type> x [int32(c)]))
(SUBL x (MOVLconst [c])) => (SUBLconst x [c])
(SUBL (MOVLconst [c]) x) => (NEGL (SUBLconst <v.Type> x [c]))

(MULQ x (MOVQconst [c])) && is32Bit(c) => (MULQconst [int32(c)] x)
(MULL x (MOVLconst [c])) => (MULLconst [c] x)

CALLFN(·call134217728, 134217728)
CALLFN(·call268435456, 268435456)
CALLFN(·call536870912, 536870912)
CALLFN(·call1073741824, 1073741824)

TEXT runtime·procyield(SB),NOSPLIT,$0-0
	MOVL	cycles+0(FP), AX
again:
	PAUSE
	SUBL	$1, AX
	JNZ	again
	RET


TEXT ·publicationBarrier<ABIInternal>(SB),NOSPLIT,$0-0
	// Stores are already ordered on x86, so this is just a
	// compile barrier.
	RET