MOVL	DX, BP
	SUBL	BX, DX // DX = blen-alen
	JLE	2(PC)
	MOVL	BX, BP // BP = min(alen, blen)
	CMPL	SI, DI
	JEQ	allsame
	CMPL	BP, $4
	JB	small
#ifdef GO386_softfloat
	JMP	mediumloop
#endif
largeloop:
	CMPL	BP, $16
	JB	mediumloop
	MOVOU	(SI), X0
	MOVOU	(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, BX
	XORL	$0xffff, BX	// convert EQ to NE
	JNE	diff16	// branch if at least one byte is not equal
	ADDL	$16, SI

	// return fr * 2**exponent
ldexp:
	ADDL    $0x3FF, BX // add bias
	JLE     denormal
	CMPL    BX, $0x7FF
	JGE     overflow
lastStep:
	SHLQ    $52, BX
	MOVQ    BX, X1
	MULSD   X1, X0
	MOVSD   X0, ret+8(FP)
	RET
notFinite:
	// test bits for -Inf
	MOVQ    $NegInf, AX
	CMPQ    AX, BX
	JNE     notNegInf
	// -Inf, return 0
underflow: // return 0
	MOVQ    $0, ret+8(FP)
	RET

	// Directly return DX, we don't need to accumulate
	// since we have <16 bytes.
	POPCNTL DX, DX
	MOVQ	DX, (R8)
	RET

avx2:
#ifndef hasAVX2
	CMPB   internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1
	JNE sse
#endif
	MOVD AX, X0
	LEAQ -64(SI)(BX*1), R11
	LEAQ (SI)(BX*1), R13
	VPBROADCASTB  X0, Y1
	PCALIGN $32
avx2_loop:
	VMOVDQU (DI), Y2
	VMOVDQU 32(DI), Y4
	VPCMPEQB Y1, Y2, Y3

	MOVOU	(SI), X0
	MOVOU	(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX
	XORQ	$0xffff, AX
	JNE	diff16

	MOVOU	16(SI), X0
	MOVOU	16(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX
	XORQ	$0xffff, AX
	JNE	diff32

	MOVOU	32(SI), X0
	MOVOU	32(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX
	XORQ	$0xffff, AX
	JNE	diff48

	MOVOU	48(SI), X0
	MOVOU	48(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX

	MOVL	trap+0(FP), AX	// syscall entry
	// copy args down
	LEAL	a1+4(FP), SI
	LEAL	sysargs-144(SP), DI
	CLD
	MOVSL
	MOVSL
	MOVSL
	INT	$64
	MOVL	AX, r1+16(FP)
	MOVL	$0, r2+20(FP)
	CMPL	AX, $-1
	JNE	ok3

	LEAL	errbuf-128(SP), AX
	MOVL	AX, sysargs-144(SP)
	MOVL	$128, sysargs1-140(SP)
	MOVL	$SYS_ERRSTR, AX
	INT	$64
	CALL	runtime·exitsyscall(SB)
	MOVL	sysargs-144(SP), AX
	MOVL	AX, errbuf-148(SP)

	// test bits for special cases
	MOVQ    p+0(FP), BX
	MOVQ    $~(1<<63), AX
	ANDQ    AX, BX // p = |p|
	MOVQ    q+8(FP), CX
	ANDQ    AX, CX // q = |q|
	MOVQ    $PosInf, AX
	CMPQ    AX, BX
	JLE     isInfOrNaN
	CMPQ    AX, CX
	JLE     isInfOrNaN
	// hypot = max * sqrt(1 + (min/max)**2)
	MOVQ    BX, X0
	MOVQ    CX, X1
	ORQ     CX, BX
	JEQ     isZero
	MOVAPD  X0, X2
	MAXSD   X1, X0
	MINSD   X2, X1
	DIVSD   X0, X1

func ptrEqual() {
	// amd64:-"JEQ",-"JNE"
	// ppc64x:-"BEQ",-"BNE"
	// s390x:-"BEQ",-"BNE"
	copy(x[:], x[:])
}

func ptrOneOffset() {
	// amd64:-"JEQ",-"JNE"
	// ppc64x:-"BEQ",-"BNE"
	// s390x:-"BEQ",-"BNE"
	copy(x[1:], x[:])
}

func ptrBothOffset() {
	// amd64:-"JEQ",-"JNE"
	// ppc64x:-"BEQ",-"BNE"
	// s390x:-"BEQ",-"BNE"
	copy(x[1:], x[2:])
}

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
	JNE   loop16b
	RET

	PCALIGN $16
loop_1b:
	SUBQ  $1, DX           // XOR 1byte backwards.
	MOVB  (SI)(DX*1), DI
	MOVB  (CX)(DX*1), AX
	XORB  AX, DI
	MOVB  DI, (BX)(DX*1)

	// 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

	MOVQ	trap+0(FP), BP	// syscall entry
	// copy args down
	LEAQ	a1+8(FP), SI
	LEAQ	sysargs-160(SP), DI
	CLD
	MOVSQ
	MOVSQ
	MOVSQ
	SYSCALL
	MOVQ	AX, r1+32(FP)
	MOVQ	$0, r2+40(FP)
	CMPL	AX, $-1
	JNE	ok3

	LEAQ	errbuf-128(SP), AX
	MOVQ	AX, sysargs-160(SP)
	MOVQ	$128, sysargs1-152(SP)
	MOVQ	$SYS_ERRSTR, BP
	SYSCALL
	CALL	runtime·exitsyscall(SB) // call via ABI wrapper, ensuring ABIInternal fixed registers are set