1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse3 | FileCheck %s --check-prefix=SSE3
3; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE41
4; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX_ANY
5; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX_ANY
6; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX_X86_64
7
8define <3 x i16> @zext_i8(<3 x i8>) {
9; SSE3-LABEL: zext_i8:
10; SSE3:       # BB#0:
11; SSE3-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
12; SSE3-NEXT:    pxor %xmm0, %xmm0
13; SSE3-NEXT:    pxor %xmm1, %xmm1
14; SSE3-NEXT:    pinsrw $0, %eax, %xmm1
15; SSE3-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
16; SSE3-NEXT:    pinsrw $1, %eax, %xmm1
17; SSE3-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
18; SSE3-NEXT:    pinsrw $2, %eax, %xmm1
19; SSE3-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
20; SSE3-NEXT:    movd %xmm1, %eax
21; SSE3-NEXT:    pextrw $2, %xmm1, %edx
22; SSE3-NEXT:    pextrw $4, %xmm1, %ecx
23; SSE3-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
24; SSE3-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
25; SSE3-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
26; SSE3-NEXT:    retl
27;
28; SSE41-LABEL: zext_i8:
29; SSE41:       # BB#0:
30; SSE41-NEXT:    pxor %xmm0, %xmm0
31; SSE41-NEXT:    pinsrb $0, {{[0-9]+}}(%esp), %xmm0
32; SSE41-NEXT:    pinsrb $4, {{[0-9]+}}(%esp), %xmm0
33; SSE41-NEXT:    pinsrb $8, {{[0-9]+}}(%esp), %xmm0
34; SSE41-NEXT:    movd %xmm0, %eax
35; SSE41-NEXT:    pextrw $2, %xmm0, %edx
36; SSE41-NEXT:    pextrw $4, %xmm0, %ecx
37; SSE41-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
38; SSE41-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
39; SSE41-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
40; SSE41-NEXT:    retl
41;
42; AVX_ANY-LABEL: zext_i8:
43; AVX_ANY:       # BB#0:
44; AVX_ANY-NEXT:    vpxor %xmm0, %xmm0, %xmm0
45; AVX_ANY-NEXT:    vpinsrb $0, {{[0-9]+}}(%esp), %xmm0, %xmm0
46; AVX_ANY-NEXT:    vpinsrb $4, {{[0-9]+}}(%esp), %xmm0, %xmm0
47; AVX_ANY-NEXT:    vpinsrb $8, {{[0-9]+}}(%esp), %xmm0, %xmm0
48; AVX_ANY-NEXT:    vmovd %xmm0, %eax
49; AVX_ANY-NEXT:    vpextrw $2, %xmm0, %edx
50; AVX_ANY-NEXT:    vpextrw $4, %xmm0, %ecx
51; AVX_ANY-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
52; AVX_ANY-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
53; AVX_ANY-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
54; AVX_ANY-NEXT:    retl
55;
56; AVX_X86_64-LABEL: zext_i8:
57; AVX_X86_64:       # BB#0:
58; AVX_X86_64-NEXT:    vmovd %edi, %xmm0
59; AVX_X86_64-NEXT:    vpinsrd $1, %esi, %xmm0, %xmm0
60; AVX_X86_64-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0
61; AVX_X86_64-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
62; AVX_X86_64-NEXT:    vmovd %xmm0, %eax
63; AVX_X86_64-NEXT:    vpextrw $2, %xmm0, %edx
64; AVX_X86_64-NEXT:    vpextrw $4, %xmm0, %ecx
65; AVX_X86_64-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
66; AVX_X86_64-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
67; AVX_X86_64-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
68; AVX_X86_64-NEXT:    retq
69  %2 = zext <3 x i8> %0 to <3 x i16>
70  ret <3 x i16> %2
71}
72
73define <3 x i16> @sext_i8(<3 x i8>) {
74; SSE3-LABEL: sext_i8:
75; SSE3:       # BB#0:
76; SSE3-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
77; SSE3-NEXT:    pinsrw $0, %eax, %xmm0
78; SSE3-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
79; SSE3-NEXT:    pinsrw $1, %eax, %xmm0
80; SSE3-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
81; SSE3-NEXT:    pinsrw $2, %eax, %xmm0
82; SSE3-NEXT:    psllw $8, %xmm0
83; SSE3-NEXT:    psraw $8, %xmm0
84; SSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
85; SSE3-NEXT:    psrad $16, %xmm0
86; SSE3-NEXT:    movd %xmm0, %eax
87; SSE3-NEXT:    pextrw $2, %xmm0, %edx
88; SSE3-NEXT:    pextrw $4, %xmm0, %ecx
89; SSE3-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
90; SSE3-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
91; SSE3-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
92; SSE3-NEXT:    retl
93;
94; SSE41-LABEL: sext_i8:
95; SSE41:       # BB#0:
96; SSE41-NEXT:    pinsrb $0, {{[0-9]+}}(%esp), %xmm0
97; SSE41-NEXT:    pinsrb $4, {{[0-9]+}}(%esp), %xmm0
98; SSE41-NEXT:    pinsrb $8, {{[0-9]+}}(%esp), %xmm0
99; SSE41-NEXT:    pslld $24, %xmm0
100; SSE41-NEXT:    psrad $24, %xmm0
101; SSE41-NEXT:    movd %xmm0, %eax
102; SSE41-NEXT:    pextrw $2, %xmm0, %edx
103; SSE41-NEXT:    pextrw $4, %xmm0, %ecx
104; SSE41-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
105; SSE41-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
106; SSE41-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
107; SSE41-NEXT:    retl
108;
109; AVX_ANY-LABEL: sext_i8:
110; AVX_ANY:       # BB#0:
111; AVX_ANY-NEXT:    vpinsrb $0, {{[0-9]+}}(%esp), %xmm0, %xmm0
112; AVX_ANY-NEXT:    vpinsrb $4, {{[0-9]+}}(%esp), %xmm0, %xmm0
113; AVX_ANY-NEXT:    vpinsrb $8, {{[0-9]+}}(%esp), %xmm0, %xmm0
114; AVX_ANY-NEXT:    vpslld $24, %xmm0, %xmm0
115; AVX_ANY-NEXT:    vpsrad $24, %xmm0, %xmm0
116; AVX_ANY-NEXT:    vmovd %xmm0, %eax
117; AVX_ANY-NEXT:    vpextrw $2, %xmm0, %edx
118; AVX_ANY-NEXT:    vpextrw $4, %xmm0, %ecx
119; AVX_ANY-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
120; AVX_ANY-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
121; AVX_ANY-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
122; AVX_ANY-NEXT:    retl
123;
124; AVX_X86_64-LABEL: sext_i8:
125; AVX_X86_64:       # BB#0:
126; AVX_X86_64-NEXT:    vmovd %edi, %xmm0
127; AVX_X86_64-NEXT:    vpinsrd $1, %esi, %xmm0, %xmm0
128; AVX_X86_64-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0
129; AVX_X86_64-NEXT:    vpslld $24, %xmm0, %xmm0
130; AVX_X86_64-NEXT:    vpsrad $24, %xmm0, %xmm0
131; AVX_X86_64-NEXT:    vmovd %xmm0, %eax
132; AVX_X86_64-NEXT:    vpextrw $2, %xmm0, %edx
133; AVX_X86_64-NEXT:    vpextrw $4, %xmm0, %ecx
134; AVX_X86_64-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
135; AVX_X86_64-NEXT:    # kill: %DX<def> %DX<kill> %EDX<kill>
136; AVX_X86_64-NEXT:    # kill: %CX<def> %CX<kill> %ECX<kill>
137; AVX_X86_64-NEXT:    retq
138  %2 = sext <3 x i8> %0 to <3 x i16>
139  ret <3 x i16> %2
140}
141