"VSLD",
	"VSRD",
	"VSA",
	"VSRAB",
	"VSRAH",
	"VSRAW",
	"VSRAD",
	"VSOI",
	"VSLDOI",
	"VCLZ",
	"VCLZB",
	"VCLZH",
	"VCLZW",
	"VCLZD",
	"VPOPCNT",
	"VPOPCNTB",
	"VPOPCNTH",
	"VPOPCNTW",
	"VPOPCNTD",
	"VCMPEQ",
	"VCMPEQUB",
	"VCMPEQUBCC",
	"VCMPEQUH",
	"VCMPEQUHCC",
	"VCMPEQUW",
	"VCMPEQUWCC",
	"VCMPEQUD",
	"VCMPEQUDCC",
	"VCMPGT",
	"VCMPGTUB",

	VPOPCNTD X16, K2, X26                              // 62227d0a55d0
	VPOPCNTD X23, K2, X26                              // 62227d0a55d7
	VPOPCNTD (R14), K2, X26                            // 62427d0a5516
	VPOPCNTD -7(DI)(R8*8), K2, X26                     // 62227d0a5594c7f9ffffff
	VPOPCNTD X12, K2, X23                              // 62c27d0a55fc
	VPOPCNTD X16, K2, X23                              // 62a27d0a55f8

	PCALIGN $32
avx2_loop:
	VMOVDQU (DI), Y2
	VMOVDQU 32(DI), Y4
	VPCMPEQB Y1, Y2, Y3
	VPCMPEQB Y1, Y4, Y5
	VPMOVMSKB Y3, DX
	VPMOVMSKB Y5, CX
	POPCNTL DX, DX
	POPCNTL CX, CX
	ADDQ DX, R12
	ADDQ CX, R12
	ADDQ $64, DI
	CMPQ DI, R11
	JLE avx2_loop

	// If last block is already processed,
	// skip to the end.
	//
	// This check is NOT an optimization; if the input length is a

	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
	ADD	$8, R3, R3
	ADD	$-8, R4, R4