ruby_snowflake_client 1.0.0

data/ext/vendor/github.com/apache/arrow/go/v10/internal/utils/transpose_ints_sse4_amd64.s

@@ -0,0 +1,3074 @@
+//+build !noasm !appengine
+// AUTO-GENERATED BY C2GOASM -- DO NOT EDIT
+
+TEXT ·_transpose_uint8_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB0_1
+
+LBB0_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB0_5
+
+LBB0_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB0_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB0_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB0_3
+
+LBB0_4:
+	RET
+
+TEXT ·_transpose_int8_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB1_1
+
+LBB1_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688                 // mov    byte [rsi], dl
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01     // mov    byte [rsi + 1], dl
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02     // mov    byte [rsi + 2], dl
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03     // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x04c68348             // add    rsi, 4
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB1_5
+
+LBB1_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB1_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB1_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB1_3
+
+LBB1_4:
+	RET
+
+TEXT ·_transpose_uint16_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB2_1
+
+LBB2_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB2_5
+
+LBB2_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB2_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB2_3:
+	LONG $0x04b70f42; BYTE $0x47 // movzx    eax, word [rdi + 2*r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB2_3
+
+LBB2_4:
+	RET
+
+TEXT ·_transpose_int16_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB3_1
+
+LBB3_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688                 // mov    byte [rsi], dl
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01     // mov    byte [rsi + 1], dl
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02     // mov    byte [rsi + 2], dl
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03     // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x04c68348             // add    rsi, 4
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB3_5
+
+LBB3_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB3_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB3_3:
+	LONG $0x04bf0f4a; BYTE $0x47 // movsx    rax, word [rdi + 2*r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB3_3
+
+LBB3_4:
+	RET
+
+TEXT ·_transpose_uint32_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB4_1
+
+LBB4_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB4_5
+
+LBB4_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB4_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB4_3:
+	LONG $0x87048b42         // mov    eax, dword [rdi + 4*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB4_3
+
+LBB4_4:
+	RET
+
+TEXT ·_transpose_int32_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB5_1
+
+LBB5_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB5_5
+
+LBB5_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB5_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB5_3:
+	LONG $0x8704634a         // movsxd    rax, dword [rdi + 4*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB5_3
+
+LBB5_4:
+	RET
+
+TEXT ·_transpose_uint64_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB6_1
+
+LBB6_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB6_5
+
+LBB6_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB6_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB6_3:
+	LONG $0xc7048b4a         // mov    rax, qword [rdi + 8*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB6_3
+
+LBB6_4:
+	RET
+
+TEXT ·_transpose_int64_uint8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB7_1
+
+LBB7_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB7_5
+
+LBB7_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB7_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB7_3:
+	LONG $0xc7048b4a         // mov    rax, qword [rdi + 8*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB7_3
+
+LBB7_4:
+	RET
+
+TEXT ·_transpose_uint8_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB8_1
+
+LBB8_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB8_5
+
+LBB8_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB8_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB8_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB8_3
+
+LBB8_4:
+	RET
+
+TEXT ·_transpose_int8_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB9_1
+
+LBB9_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688                 // mov    byte [rsi], dl
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01     // mov    byte [rsi + 1], dl
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02     // mov    byte [rsi + 2], dl
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03     // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x04c68348             // add    rsi, 4
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB9_5
+
+LBB9_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB9_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB9_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB9_3
+
+LBB9_4:
+	RET
+
+TEXT ·_transpose_uint16_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB10_1
+
+LBB10_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB10_5
+
+LBB10_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB10_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB10_3:
+	LONG $0x04b70f42; BYTE $0x47 // movzx    eax, word [rdi + 2*r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB10_3
+
+LBB10_4:
+	RET
+
+TEXT ·_transpose_int16_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB11_1
+
+LBB11_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688                 // mov    byte [rsi], dl
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01     // mov    byte [rsi + 1], dl
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02     // mov    byte [rsi + 2], dl
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	LONG $0x9114b60f             // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03     // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x04c68348             // add    rsi, 4
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB11_5
+
+LBB11_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB11_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB11_3:
+	LONG $0x04bf0f4a; BYTE $0x47 // movsx    rax, word [rdi + 2*r8]
+	LONG $0x8104b60f             // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842             // mov    byte [rsi + r8], al
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB11_3
+
+LBB11_4:
+	RET
+
+TEXT ·_transpose_uint32_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB12_1
+
+LBB12_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB12_5
+
+LBB12_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB12_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB12_3:
+	LONG $0x87048b42         // mov    eax, dword [rdi + 4*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB12_3
+
+LBB12_4:
+	RET
+
+TEXT ·_transpose_int32_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB13_1
+
+LBB13_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB13_5
+
+LBB13_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB13_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB13_3:
+	LONG $0x8704634a         // movsxd    rax, dword [rdi + 4*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB13_3
+
+LBB13_4:
+	RET
+
+TEXT ·_transpose_uint64_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB14_1
+
+LBB14_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB14_5
+
+LBB14_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB14_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB14_3:
+	LONG $0xc7048b4a         // mov    rax, qword [rdi + 8*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB14_3
+
+LBB14_4:
+	RET
+
+TEXT ·_transpose_int64_int8_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB15_1
+
+LBB15_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x1688             // mov    byte [rsi], dl
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x01 // mov    byte [rsi + 1], dl
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x02 // mov    byte [rsi + 2], dl
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b60f         // movzx    edx, byte [rcx + 4*rdx]
+	WORD $0x5688; BYTE $0x03 // mov    byte [rsi + 3], dl
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x04c68348         // add    rsi, 4
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB15_5
+
+LBB15_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB15_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB15_3:
+	LONG $0xc7048b4a         // mov    rax, qword [rdi + 8*r8]
+	LONG $0x8104b60f         // movzx    eax, byte [rcx + 4*rax]
+	LONG $0x06048842         // mov    byte [rsi + r8], al
+	LONG $0x01c08349         // add    r8, 1
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB15_3
+
+LBB15_4:
+	RET
+
+TEXT ·_transpose_uint8_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB16_1
+
+LBB16_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB16_5
+
+LBB16_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB16_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB16_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x46 // mov    word [rsi + 2*r8], ax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB16_3
+
+LBB16_4:
+	RET
+
+TEXT ·_transpose_int8_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB17_1
+
+LBB17_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16     // mov    word [rsi], dx
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966             // mov    word [rsi + 2], dx
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966             // mov    word [rsi + 4], dx
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966             // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x08c68348             // add    rsi, 8
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB17_5
+
+LBB17_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB17_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB17_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x46 // mov    word [rsi + 2*r8], ax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB17_3
+
+LBB17_4:
+	RET
+
+TEXT ·_transpose_uint16_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB18_1
+
+LBB18_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB18_5
+
+LBB18_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB18_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB18_3:
+	LONG $0x04b70f42; BYTE $0x07 // movzx    eax, word [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB18_3
+
+LBB18_4:
+	RET
+
+TEXT ·_transpose_int16_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB19_1
+
+LBB19_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16     // mov    word [rsi], dx
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966             // mov    word [rsi + 2], dx
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966             // mov    word [rsi + 4], dx
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966             // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x08c68348             // add    rsi, 8
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB19_5
+
+LBB19_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB19_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB19_3:
+	LONG $0x04bf0f4a; BYTE $0x07 // movsx    rax, word [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB19_3
+
+LBB19_4:
+	RET
+
+TEXT ·_transpose_uint32_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB20_1
+
+LBB20_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB20_5
+
+LBB20_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB20_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB20_3:
+	LONG $0x47048b42             // mov    eax, dword [rdi + 2*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB20_3
+
+LBB20_4:
+	RET
+
+TEXT ·_transpose_int32_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB21_1
+
+LBB21_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB21_5
+
+LBB21_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB21_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB21_3:
+	LONG $0x4704634a             // movsxd    rax, dword [rdi + 2*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB21_3
+
+LBB21_4:
+	RET
+
+TEXT ·_transpose_uint64_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB22_1
+
+LBB22_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB22_5
+
+LBB22_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB22_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB22_3:
+	LONG $0x87048b4a             // mov    rax, qword [rdi + 4*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB22_3
+
+LBB22_4:
+	RET
+
+TEXT ·_transpose_int64_uint16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB23_1
+
+LBB23_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB23_5
+
+LBB23_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB23_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB23_3:
+	LONG $0x87048b4a             // mov    rax, qword [rdi + 4*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB23_3
+
+LBB23_4:
+	RET
+
+TEXT ·_transpose_uint8_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB24_1
+
+LBB24_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB24_5
+
+LBB24_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB24_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB24_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x46 // mov    word [rsi + 2*r8], ax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB24_3
+
+LBB24_4:
+	RET
+
+TEXT ·_transpose_int8_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB25_1
+
+LBB25_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16     // mov    word [rsi], dx
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966             // mov    word [rsi + 2], dx
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966             // mov    word [rsi + 4], dx
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966             // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x08c68348             // add    rsi, 8
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB25_5
+
+LBB25_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB25_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB25_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x46 // mov    word [rsi + 2*r8], ax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB25_3
+
+LBB25_4:
+	RET
+
+TEXT ·_transpose_uint16_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB26_1
+
+LBB26_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB26_5
+
+LBB26_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB26_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB26_3:
+	LONG $0x04b70f42; BYTE $0x07 // movzx    eax, word [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB26_3
+
+LBB26_4:
+	RET
+
+TEXT ·_transpose_int16_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB27_1
+
+LBB27_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16     // mov    word [rsi], dx
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966             // mov    word [rsi + 2], dx
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966             // mov    word [rsi + 4], dx
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	LONG $0x9114b70f             // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966             // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x08c68348             // add    rsi, 8
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB27_5
+
+LBB27_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB27_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB27_3:
+	LONG $0x04bf0f4a; BYTE $0x07 // movsx    rax, word [rdi + r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB27_3
+
+LBB27_4:
+	RET
+
+TEXT ·_transpose_uint32_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB28_1
+
+LBB28_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB28_5
+
+LBB28_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB28_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB28_3:
+	LONG $0x47048b42             // mov    eax, dword [rdi + 2*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB28_3
+
+LBB28_4:
+	RET
+
+TEXT ·_transpose_int32_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB29_1
+
+LBB29_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB29_5
+
+LBB29_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB29_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB29_3:
+	LONG $0x4704634a             // movsxd    rax, dword [rdi + 2*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB29_3
+
+LBB29_4:
+	RET
+
+TEXT ·_transpose_uint64_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB30_1
+
+LBB30_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB30_5
+
+LBB30_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB30_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB30_3:
+	LONG $0x87048b4a             // mov    rax, qword [rdi + 4*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB30_3
+
+LBB30_4:
+	RET
+
+TEXT ·_transpose_int64_int16_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB31_1
+
+LBB31_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	WORD $0x8966; BYTE $0x16 // mov    word [rsi], dx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x02568966         // mov    word [rsi + 2], dx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x04568966         // mov    word [rsi + 4], dx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x9114b70f         // movzx    edx, word [rcx + 4*rdx]
+	LONG $0x06568966         // mov    word [rsi + 6], dx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x08c68348         // add    rsi, 8
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB31_5
+
+LBB31_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB31_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB31_3:
+	LONG $0x87048b4a             // mov    rax, qword [rdi + 4*r8]
+	LONG $0x8104b70f             // movzx    eax, word [rcx + 4*rax]
+	LONG $0x04894266; BYTE $0x06 // mov    word [rsi + r8], ax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB31_3
+
+LBB31_4:
+	RET
+
+TEXT ·_transpose_uint8_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB32_1
+
+LBB32_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB32_5
+
+LBB32_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB32_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB32_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x86048942             // mov    dword [rsi + 4*r8], eax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB32_3
+
+LBB32_4:
+	RET
+
+TEXT ·_transpose_int8_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB33_1
+
+LBB33_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689                 // mov    dword [rsi], edx
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04     // mov    dword [rsi + 4], edx
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08     // mov    dword [rsi + 8], edx
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c     // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x10c68348             // add    rsi, 16
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB33_5
+
+LBB33_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB33_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB33_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x86048942             // mov    dword [rsi + 4*r8], eax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB33_3
+
+LBB33_4:
+	RET
+
+TEXT ·_transpose_uint16_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB34_1
+
+LBB34_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB34_5
+
+LBB34_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB34_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB34_3:
+	LONG $0x04b70f42; BYTE $0x07 // movzx    eax, word [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x46048942             // mov    dword [rsi + 2*r8], eax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB34_3
+
+LBB34_4:
+	RET
+
+TEXT ·_transpose_int16_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB35_1
+
+LBB35_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689                 // mov    dword [rsi], edx
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04     // mov    dword [rsi + 4], edx
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08     // mov    dword [rsi + 8], edx
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c     // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x10c68348             // add    rsi, 16
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB35_5
+
+LBB35_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB35_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB35_3:
+	LONG $0x04bf0f4a; BYTE $0x07 // movsx    rax, word [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x46048942             // mov    dword [rsi + 2*r8], eax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB35_3
+
+LBB35_4:
+	RET
+
+TEXT ·_transpose_uint32_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB36_1
+
+LBB36_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB36_5
+
+LBB36_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB36_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB36_3:
+	LONG $0x07048b42         // mov    eax, dword [rdi + r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB36_3
+
+LBB36_4:
+	RET
+
+TEXT ·_transpose_int32_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB37_1
+
+LBB37_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB37_5
+
+LBB37_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB37_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB37_3:
+	LONG $0x0704634a         // movsxd    rax, dword [rdi + r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB37_3
+
+LBB37_4:
+	RET
+
+TEXT ·_transpose_uint64_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB38_1
+
+LBB38_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB38_5
+
+LBB38_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB38_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB38_3:
+	LONG $0x47048b4a         // mov    rax, qword [rdi + 2*r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB38_3
+
+LBB38_4:
+	RET
+
+TEXT ·_transpose_int64_uint32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB39_1
+
+LBB39_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB39_5
+
+LBB39_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB39_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB39_3:
+	LONG $0x47048b4a         // mov    rax, qword [rdi + 2*r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB39_3
+
+LBB39_4:
+	RET
+
+TEXT ·_transpose_uint8_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB40_1
+
+LBB40_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB40_5
+
+LBB40_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB40_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB40_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x86048942             // mov    dword [rsi + 4*r8], eax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB40_3
+
+LBB40_4:
+	RET
+
+TEXT ·_transpose_int8_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB41_1
+
+LBB41_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689                 // mov    dword [rsi], edx
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04     // mov    dword [rsi + 4], edx
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08     // mov    dword [rsi + 8], edx
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c     // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x10c68348             // add    rsi, 16
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB41_5
+
+LBB41_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB41_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB41_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x86048942             // mov    dword [rsi + 4*r8], eax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB41_3
+
+LBB41_4:
+	RET
+
+TEXT ·_transpose_uint16_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB42_1
+
+LBB42_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB42_5
+
+LBB42_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB42_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB42_3:
+	LONG $0x04b70f42; BYTE $0x07 // movzx    eax, word [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x46048942             // mov    dword [rsi + 2*r8], eax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB42_3
+
+LBB42_4:
+	RET
+
+TEXT ·_transpose_int16_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB43_1
+
+LBB43_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689                 // mov    dword [rsi], edx
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04     // mov    dword [rsi + 4], edx
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08     // mov    dword [rsi + 8], edx
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	WORD $0x148b; BYTE $0x91     // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c     // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x10c68348             // add    rsi, 16
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB43_5
+
+LBB43_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB43_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB43_3:
+	LONG $0x04bf0f4a; BYTE $0x07 // movsx    rax, word [rdi + r8]
+	WORD $0x048b; BYTE $0x81     // mov    eax, dword [rcx + 4*rax]
+	LONG $0x46048942             // mov    dword [rsi + 2*r8], eax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB43_3
+
+LBB43_4:
+	RET
+
+TEXT ·_transpose_uint32_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB44_1
+
+LBB44_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB44_5
+
+LBB44_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB44_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB44_3:
+	LONG $0x07048b42         // mov    eax, dword [rdi + r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB44_3
+
+LBB44_4:
+	RET
+
+TEXT ·_transpose_int32_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB45_1
+
+LBB45_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB45_5
+
+LBB45_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB45_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB45_3:
+	LONG $0x0704634a         // movsxd    rax, dword [rdi + r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB45_3
+
+LBB45_4:
+	RET
+
+TEXT ·_transpose_uint64_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB46_1
+
+LBB46_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB46_5
+
+LBB46_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB46_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB46_3:
+	LONG $0x47048b4a         // mov    rax, qword [rdi + 2*r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB46_3
+
+LBB46_4:
+	RET
+
+TEXT ·_transpose_int64_int32_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB47_1
+
+LBB47_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x1689             // mov    dword [rsi], edx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x04 // mov    dword [rsi + 4], edx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x08 // mov    dword [rsi + 8], edx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	WORD $0x148b; BYTE $0x91 // mov    edx, dword [rcx + 4*rdx]
+	WORD $0x5689; BYTE $0x0c // mov    dword [rsi + 12], edx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x10c68348         // add    rsi, 16
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB47_5
+
+LBB47_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB47_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB47_3:
+	LONG $0x47048b4a         // mov    rax, qword [rdi + 2*r8]
+	WORD $0x048b; BYTE $0x81 // mov    eax, dword [rcx + 4*rax]
+	LONG $0x06048942         // mov    dword [rsi + r8], eax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB47_3
+
+LBB47_4:
+	RET
+
+TEXT ·_transpose_uint8_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB48_1
+
+LBB48_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB48_5
+
+LBB48_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB48_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB48_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0xc604894a             // mov    qword [rsi + 8*r8], rax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB48_3
+
+LBB48_4:
+	RET
+
+TEXT ·_transpose_int8_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB49_1
+
+LBB49_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16     // mov    qword [rsi], rdx
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948             // mov    qword [rsi + 8], rdx
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948             // mov    qword [rsi + 16], rdx
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948             // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x20c68348             // add    rsi, 32
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB49_5
+
+LBB49_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB49_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB49_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0xc604894a             // mov    qword [rsi + 8*r8], rax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB49_3
+
+LBB49_4:
+	RET
+
+TEXT ·_transpose_uint16_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB50_1
+
+LBB50_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB50_5
+
+LBB50_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB50_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB50_3:
+	LONG $0x04b70f42; BYTE $0x07 // movzx    eax, word [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x8604894a             // mov    qword [rsi + 4*r8], rax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB50_3
+
+LBB50_4:
+	RET
+
+TEXT ·_transpose_int16_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB51_1
+
+LBB51_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16     // mov    qword [rsi], rdx
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948             // mov    qword [rsi + 8], rdx
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948             // mov    qword [rsi + 16], rdx
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948             // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x20c68348             // add    rsi, 32
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB51_5
+
+LBB51_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB51_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB51_3:
+	LONG $0x04bf0f4a; BYTE $0x07 // movsx    rax, word [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x8604894a             // mov    qword [rsi + 4*r8], rax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB51_3
+
+LBB51_4:
+	RET
+
+TEXT ·_transpose_uint32_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB52_1
+
+LBB52_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB52_5
+
+LBB52_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB52_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB52_3:
+	LONG $0x07048b42         // mov    eax, dword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x4604894a         // mov    qword [rsi + 2*r8], rax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB52_3
+
+LBB52_4:
+	RET
+
+TEXT ·_transpose_int32_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB53_1
+
+LBB53_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB53_5
+
+LBB53_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB53_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB53_3:
+	LONG $0x0704634a         // movsxd    rax, dword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x4604894a         // mov    qword [rsi + 2*r8], rax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB53_3
+
+LBB53_4:
+	RET
+
+TEXT ·_transpose_uint64_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB54_1
+
+LBB54_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB54_5
+
+LBB54_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB54_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB54_3:
+	LONG $0x07048b4a         // mov    rax, qword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x0604894a         // mov    qword [rsi + r8], rax
+	LONG $0x08c08349         // add    r8, 8
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB54_3
+
+LBB54_4:
+	RET
+
+TEXT ·_transpose_int64_uint64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB55_1
+
+LBB55_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB55_5
+
+LBB55_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB55_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB55_3:
+	LONG $0x07048b4a         // mov    rax, qword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x0604894a         // mov    qword [rsi + r8], rax
+	LONG $0x08c08349         // add    r8, 8
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB55_3
+
+LBB55_4:
+	RET
+
+TEXT ·_transpose_uint8_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB56_1
+
+LBB56_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb60f; BYTE $0x17 // movzx    edx, byte [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x0157b60f         // movzx    edx, byte [rdi + 1]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x0257b60f         // movzx    edx, byte [rdi + 2]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x0357b60f         // movzx    edx, byte [rdi + 3]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x04c78348         // add    rdi, 4
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB56_5
+
+LBB56_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB56_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB56_3:
+	LONG $0x04b60f42; BYTE $0x07 // movzx    eax, byte [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0xc604894a             // mov    qword [rsi + 8*r8], rax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB56_3
+
+LBB56_4:
+	RET
+
+TEXT ·_transpose_int8_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB57_1
+
+LBB57_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17be0f48             // movsx    rdx, byte [rdi]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16     // mov    qword [rsi], rdx
+	LONG $0x57be0f48; BYTE $0x01 // movsx    rdx, byte [rdi + 1]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948             // mov    qword [rsi + 8], rdx
+	LONG $0x57be0f48; BYTE $0x02 // movsx    rdx, byte [rdi + 2]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948             // mov    qword [rsi + 16], rdx
+	LONG $0x57be0f48; BYTE $0x03 // movsx    rdx, byte [rdi + 3]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948             // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x04c78348             // add    rdi, 4
+	LONG $0x20c68348             // add    rsi, 32
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB57_5
+
+LBB57_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB57_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB57_3:
+	LONG $0x04be0f4a; BYTE $0x07 // movsx    rax, byte [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0xc604894a             // mov    qword [rsi + 8*r8], rax
+	LONG $0x01c08349             // add    r8, 1
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB57_3
+
+LBB57_4:
+	RET
+
+TEXT ·_transpose_uint16_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB58_1
+
+LBB58_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0xb70f; BYTE $0x17 // movzx    edx, word [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x0257b70f         // movzx    edx, word [rdi + 2]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x0457b70f         // movzx    edx, word [rdi + 4]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x0657b70f         // movzx    edx, word [rdi + 6]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x08c78348         // add    rdi, 8
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB58_5
+
+LBB58_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB58_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB58_3:
+	LONG $0x04b70f42; BYTE $0x07 // movzx    eax, word [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x8604894a             // mov    qword [rsi + 4*r8], rax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB58_3
+
+LBB58_4:
+	RET
+
+TEXT ·_transpose_int16_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB59_1
+
+LBB59_5:
+	WORD $0xd089                 // mov    eax, edx
+	LONG $0x17bf0f48             // movsx    rdx, word [rdi]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16     // mov    qword [rsi], rdx
+	LONG $0x57bf0f48; BYTE $0x02 // movsx    rdx, word [rdi + 2]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948             // mov    qword [rsi + 8], rdx
+	LONG $0x57bf0f48; BYTE $0x04 // movsx    rdx, word [rdi + 4]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948             // mov    qword [rsi + 16], rdx
+	LONG $0x57bf0f48; BYTE $0x06 // movsx    rdx, word [rdi + 6]
+	LONG $0x91146348             // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948             // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc     // lea    edx, [rax - 4]
+	LONG $0x08c78348             // add    rdi, 8
+	LONG $0x20c68348             // add    rsi, 32
+	WORD $0xf883; BYTE $0x07     // cmp    eax, 7
+	JG   LBB59_5
+
+LBB59_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB59_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB59_3:
+	LONG $0x04bf0f4a; BYTE $0x07 // movsx    rax, word [rdi + r8]
+	LONG $0x81046348             // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x8604894a             // mov    qword [rsi + 4*r8], rax
+	LONG $0x02c08349             // add    r8, 2
+	WORD $0xc283; BYTE $0xff     // add    edx, -1
+	WORD $0xfa83; BYTE $0x01     // cmp    edx, 1
+	JG   LBB59_3
+
+LBB59_4:
+	RET
+
+TEXT ·_transpose_uint32_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB60_1
+
+LBB60_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x178b             // mov    edx, dword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	WORD $0x578b; BYTE $0x04 // mov    edx, dword [rdi + 4]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	WORD $0x578b; BYTE $0x08 // mov    edx, dword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	WORD $0x578b; BYTE $0x0c // mov    edx, dword [rdi + 12]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB60_5
+
+LBB60_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB60_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB60_3:
+	LONG $0x07048b42         // mov    eax, dword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x4604894a         // mov    qword [rsi + 2*r8], rax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB60_3
+
+LBB60_4:
+	RET
+
+TEXT ·_transpose_int32_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB61_1
+
+LBB61_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x6348; BYTE $0x17 // movsxd    rdx, dword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x04576348         // movsxd    rdx, dword [rdi + 4]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x08576348         // movsxd    rdx, dword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x0c576348         // movsxd    rdx, dword [rdi + 12]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x10c78348         // add    rdi, 16
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB61_5
+
+LBB61_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB61_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB61_3:
+	LONG $0x0704634a         // movsxd    rax, dword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x4604894a         // mov    qword [rsi + 2*r8], rax
+	LONG $0x04c08349         // add    r8, 4
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB61_3
+
+LBB61_4:
+	RET
+
+TEXT ·_transpose_uint64_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB62_1
+
+LBB62_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB62_5
+
+LBB62_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB62_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB62_3:
+	LONG $0x07048b4a         // mov    rax, qword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x0604894a         // mov    qword [rsi + r8], rax
+	LONG $0x08c08349         // add    r8, 8
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB62_3
+
+LBB62_4:
+	RET
+
+TEXT ·_transpose_int64_int64_sse4(SB), $0-32
+
+	MOVQ src+0(FP), DI
+	MOVQ dest+8(FP), SI
+	MOVQ length+16(FP), DX
+	MOVQ transposeMap+24(FP), CX
+
+	WORD $0xfa83; BYTE $0x04 // cmp    edx, 4
+	JL   LBB63_1
+
+LBB63_5:
+	WORD $0xd089             // mov    eax, edx
+	WORD $0x8b48; BYTE $0x17 // mov    rdx, qword [rdi]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	WORD $0x8948; BYTE $0x16 // mov    qword [rsi], rdx
+	LONG $0x08578b48         // mov    rdx, qword [rdi + 8]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x08568948         // mov    qword [rsi + 8], rdx
+	LONG $0x10578b48         // mov    rdx, qword [rdi + 16]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x10568948         // mov    qword [rsi + 16], rdx
+	LONG $0x18578b48         // mov    rdx, qword [rdi + 24]
+	LONG $0x91146348         // movsxd    rdx, dword [rcx + 4*rdx]
+	LONG $0x18568948         // mov    qword [rsi + 24], rdx
+	WORD $0x508d; BYTE $0xfc // lea    edx, [rax - 4]
+	LONG $0x20c78348         // add    rdi, 32
+	LONG $0x20c68348         // add    rsi, 32
+	WORD $0xf883; BYTE $0x07 // cmp    eax, 7
+	JG   LBB63_5
+
+LBB63_1:
+	WORD $0xd285             // test    edx, edx
+	JLE  LBB63_4
+	WORD $0xc283; BYTE $0x01 // add    edx, 1
+	WORD $0x3145; BYTE $0xc0 // xor    r8d, r8d
+
+LBB63_3:
+	LONG $0x07048b4a         // mov    rax, qword [rdi + r8]
+	LONG $0x81046348         // movsxd    rax, dword [rcx + 4*rax]
+	LONG $0x0604894a         // mov    qword [rsi + r8], rax
+	LONG $0x08c08349         // add    r8, 8
+	WORD $0xc283; BYTE $0xff // add    edx, -1
+	WORD $0xfa83; BYTE $0x01 // cmp    edx, 1
+	JG   LBB63_3
+
+LBB63_4:
+	RET