#else
tail:	// Count the last 0 - 31 bytes.
	CMP	R4, $16
	BLT	tail_8
	MOVD	(R3), R12
	MOVD	8(R3), R14
	CMPB	R12, R5, R12
	CMPB	R14, R5, R14
	POPCNTD	R12, R12
	POPCNTD	R14, R14
	ADD	R12, R18, R18
	ADD	R14, R18, R18
	ADD	$16, R3, R3
	ADD	$-16, R4, R4

tail_8:	// Count the remaining 0 - 15 bytes.
	CMP	R4, $8
	BLT	tail_4
	MOVD	(R3), R12
	CMPB	R12, R5, R12
	POPCNTD	R12, R12
	ADD	R12, R18, R18

//   b in DI
//   count in BX
// Output:
//   result in AX
TEXT memeqbody<>(SB),NOSPLIT,$0-0
	CMPQ	BX, $8
	JB	small
	CMPQ	BX, $64
	JB	bigloop
#ifndef hasAVX2
	CMPB	internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1
	JE	hugeloop_avx2

	// 64 bytes at a time using xmm registers
	PCALIGN $16
hugeloop:
	CMPQ	BX, $64
	JB	bigloop
	MOVOU	(SI), X0
	MOVOU	(DI), X1
	MOVOU	16(SI), X2
	MOVOU	16(DI), X3

#include "textflag.h"

TEXT ·Count(SB),NOSPLIT,$0-40
#ifndef hasPOPCNT
	CMPB	internal∕cpu·X86+const_offsetX86HasPOPCNT(SB), $1
	JEQ	2(PC)
	JMP	·countGeneric(SB)
#endif
	MOVQ	b_base+0(FP), SI
	MOVQ	b_len+8(FP), BX
	MOVB	c+24(FP), AL
	LEAQ	ret+32(FP), R8
	JMP	countbody<>(SB)

TEXT ·CountString(SB),NOSPLIT,$0-32
#ifndef hasPOPCNT
	CMPB	internal∕cpu·X86+const_offsetX86HasPOPCNT(SB), $1
	JEQ	2(PC)

	MOVQ	DX, (R8)
	RET

avx2:
#ifndef hasAVX2
	CMPB   internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1
	JNE sse
#endif
	MOVD AX, X0
	LEAQ -32(SI)(BX*1), R11
	VPBROADCASTB  X0, Y1

	PCALIGN $32
avx2_loop:
	VMOVDQU (DI), Y2
	VPCMPEQB Y1, Y2, Y3
	VPTEST Y3, Y3
	JNZ avx2success
	ADDQ $32, DI
	CMPQ DI, R11
	JLT avx2_loop
	MOVQ R11, DI
	VMOVDQU (DI), Y2

	"FRIZ",
	"FRIZCC",
	"FRIN",
	"FRINCC",
	"FRSQRTE",
	"FRSQRTECC",
	"FSEL",
	"FSELCC",
	"FSQRT",
	"FSQRTCC",
	"FSQRTS",
	"FSQRTSCC",
	"CNTLZD",
	"CNTLZDCC",
	"CMPW",
	"CMPWU",
	"CMPB",
	"FTDIV",
	"FTSQRT",
	"DIVD",
	"DIVDCC",
	"DIVDE",
	"DIVDECC",
	"DIVDEU",
	"DIVDEUCC",
	"DIVDVCC",
	"DIVDV",
	"DIVDU",
	"DIVDUCC",
	"DIVDUVCC",
	"DIVDUV",

// See issue 53333.
func mimetype(ext string) string {
	// amd64: `CMPB\s1\(.*\), \$104$`,-`cmpstring`
	// arm64: `MOVB\s1\(R.*\), R.*$`, `CMPW\s\$104, R.*$`, -`cmpstring`
	switch ext {
	// amd64: `CMPL\s\(.*\), \$1836345390$`
	// arm64: `MOVD\s\$1836345390`, `CMPW\sR.*, R.*$`
	case ".htm":
		return "A"
	// amd64: `CMPL\s\(.*\), \$1953457454$`
	// arm64: `MOVD\s\$1953457454`, `CMPW\sR.*, R.*$`
	case ".eot":
		return "B"

		for x := 0; x <= size-n; x = x*7 + 1 { // offset in src
			for y := 0; y <= size-n; y = y*9 + 1 { // offset in dst
				copy(dst[y:y+n], src[x:x+n])
				copyref(ref[y:y+n], src[x:x+n])
				p := cmpb(dst, ref)
				if p >= 0 {
					t.Fatalf("Copy failed, copying from src[%d:%d] to dst[%d:%d].\nOffset %d is different, %v != %v", x, x+n, y, y+n, p, dst[p], ref[p])
				}
			}
		}
	}
}

	"jmp",
	"callq",
	"cmpb",
}

var i386GnuNeed = []string{
	"jmp",
	"call",
	"cmp",
}

var armNeed = []string{
	"B main.main(SB)",
	"BL main.Println(SB)",
	"RET",
}

var arm64Need = []string{
	"JMP main.main(SB)",
	"CALL main.Println(SB)",
	"RET",
}

var armGnuNeed = []string{ // for both ARM and AMR64
	"ldr",
	"bl",
	"cmp",
}

var ppcNeed = []string{

	MOVL	$1, 0(SP)
	CALL	runtime·exit(SB)
	CALL	runtime·abort(SB)

has_cpuid:
	MOVL	$0, AX
	CPUID
	MOVL	AX, SI
	CMPL	AX, $0
	JE	nocpuinfo

	CMPL	BX, $0x756E6547  // "Genu"
	JNE	notintel
	CMPL	DX, $0x49656E69  // "ineI"
	JNE	notintel
	CMPL	CX, $0x6C65746E  // "ntel"
	JNE	notintel
	MOVB	$1, runtime·isIntel(SB)
notintel:

	// Load EAX=1 cpuid flags
	MOVL	$1, AX
	CPUID

	MOVQ	SP, (g_stack+stack_hi)(DI)

	// find out information about the processor we're on
	MOVL	$0, AX
	CPUID
	CMPL	AX, $0
	JE	nocpuinfo

	CMPL	BX, $0x756E6547  // "Genu"
	JNE	notintel
	CMPL	DX, $0x49656E69  // "ineI"
	JNE	notintel
	CMPL	CX, $0x6C65746E  // "ntel"
	JNE	notintel
	MOVB	$1, runtime·isIntel(SB)

notintel:
	// Load EAX=1 cpuid flags
	MOVL	$1, AX
	CPUID