Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / dc957d49f9c0a2fcf9aebbc59a934a7bcccc91a4 / . / llvm / test / CodeGen / X86 / widen_mul.ll
blob: dad9d58d3c99d0f00cb32d64e9a1ffa9ed4633e1 [file] [log] [blame]
Craig Topperdc957d42018-11-16 06:15:20 +0000[diff] [blame^]1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -x86-experimental-vector-widening-legalization -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
3; RUN: llc < %s -x86-experimental-vector-widening-legalization -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE --check-prefix=SSE41
4; RUN: llc < %s -x86-experimental-vector-widening-legalization -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2
5; RUN: llc < %s -x86-experimental-vector-widening-legalization -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F
6; RUN: llc < %s -x86-experimental-vector-widening-legalization -mtriple=x86_64-unknown-unknown -mattr=+avx512bw | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512BW
7
8; Test multiplies of various narrow types.
9
10define <2 x i8> @mul_v2i8(<2 x i8> %x, <2 x i8> %y) {
11; SSE2-LABEL: mul_v2i8:
12; SSE2: # %bb.0:
13; SSE2-NEXT: movdqa %xmm1, %xmm2
14; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
15; SSE2-NEXT: movdqa %xmm0, %xmm3
16; SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
17; SSE2-NEXT: pmullw %xmm2, %xmm3
18; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
19; SSE2-NEXT: pand %xmm2, %xmm3
20; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
21; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
22; SSE2-NEXT: pmullw %xmm1, %xmm0
23; SSE2-NEXT: pand %xmm2, %xmm0
24; SSE2-NEXT: packuswb %xmm3, %xmm0
25; SSE2-NEXT: retq
26;
27; SSE41-LABEL: mul_v2i8:
28; SSE41: # %bb.0:
29; SSE41-NEXT: pmovzxbw {{.*#+}} xmm3 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
30; SSE41-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
31; SSE41-NEXT: pmovzxbw {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
32; SSE41-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
33; SSE41-NEXT: pmullw %xmm1, %xmm0
34; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
35; SSE41-NEXT: pand %xmm1, %xmm0
36; SSE41-NEXT: pmullw %xmm3, %xmm2
37; SSE41-NEXT: pand %xmm1, %xmm2
38; SSE41-NEXT: packuswb %xmm0, %xmm2
39; SSE41-NEXT: movdqa %xmm2, %xmm0
40; SSE41-NEXT: retq
41;
42; AVX2-LABEL: mul_v2i8:
43; AVX2: # %bb.0:
44; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
45; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
46; AVX2-NEXT: vpmullw %ymm1, %ymm0, %ymm0
47; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
48; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
49; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1
50; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0
51; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
52; AVX2-NEXT: vzeroupper
53; AVX2-NEXT: retq
54;
55; AVX512F-LABEL: mul_v2i8:
56; AVX512F: # %bb.0:
57; AVX512F-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
58; AVX512F-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
59; AVX512F-NEXT: vpmullw %ymm1, %ymm0, %ymm0
60; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
61; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
62; AVX512F-NEXT: vzeroupper
63; AVX512F-NEXT: retq
64;
65; AVX512BW-LABEL: mul_v2i8:
66; AVX512BW: # %bb.0:
67; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
68; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
69; AVX512BW-NEXT: vpmullw %ymm1, %ymm0, %ymm0
70; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0
71; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
72; AVX512BW-NEXT: vzeroupper
73; AVX512BW-NEXT: retq
74 %res = mul <2 x i8> %x, %y
75 ret <2 x i8> %res
76}
77
78define <4 x i8> @mul_v4i8(<4 x i8> %x, <4 x i8> %y) {
79; SSE2-LABEL: mul_v4i8:
80; SSE2: # %bb.0:
81; SSE2-NEXT: movdqa %xmm1, %xmm2
82; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
83; SSE2-NEXT: movdqa %xmm0, %xmm3
84; SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
85; SSE2-NEXT: pmullw %xmm2, %xmm3
86; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
87; SSE2-NEXT: pand %xmm2, %xmm3
88; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
89; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
90; SSE2-NEXT: pmullw %xmm1, %xmm0
91; SSE2-NEXT: pand %xmm2, %xmm0
92; SSE2-NEXT: packuswb %xmm3, %xmm0
93; SSE2-NEXT: retq
94;
95; SSE41-LABEL: mul_v4i8:
96; SSE41: # %bb.0:
97; SSE41-NEXT: pmovzxbw {{.*#+}} xmm3 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
98; SSE41-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
99; SSE41-NEXT: pmovzxbw {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
100; SSE41-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
101; SSE41-NEXT: pmullw %xmm1, %xmm0
102; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
103; SSE41-NEXT: pand %xmm1, %xmm0
104; SSE41-NEXT: pmullw %xmm3, %xmm2
105; SSE41-NEXT: pand %xmm1, %xmm2
106; SSE41-NEXT: packuswb %xmm0, %xmm2
107; SSE41-NEXT: movdqa %xmm2, %xmm0
108; SSE41-NEXT: retq
109;
110; AVX2-LABEL: mul_v4i8:
111; AVX2: # %bb.0:
112; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
113; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
114; AVX2-NEXT: vpmullw %ymm1, %ymm0, %ymm0
115; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
116; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
117; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1
118; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0
119; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
120; AVX2-NEXT: vzeroupper
121; AVX2-NEXT: retq
122;
123; AVX512F-LABEL: mul_v4i8:
124; AVX512F: # %bb.0:
125; AVX512F-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
126; AVX512F-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
127; AVX512F-NEXT: vpmullw %ymm1, %ymm0, %ymm0
128; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
129; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
130; AVX512F-NEXT: vzeroupper
131; AVX512F-NEXT: retq
132;
133; AVX512BW-LABEL: mul_v4i8:
134; AVX512BW: # %bb.0:
135; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
136; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
137; AVX512BW-NEXT: vpmullw %ymm1, %ymm0, %ymm0
138; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0
139; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
140; AVX512BW-NEXT: vzeroupper
141; AVX512BW-NEXT: retq
142 %res = mul <4 x i8> %x, %y
143 ret <4 x i8> %res
144}
145
146define <8 x i8> @mul_v8i8(<8 x i8> %x, <8 x i8> %y) {
147; SSE2-LABEL: mul_v8i8:
148; SSE2: # %bb.0:
149; SSE2-NEXT: movdqa %xmm1, %xmm2
150; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
151; SSE2-NEXT: movdqa %xmm0, %xmm3
152; SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
153; SSE2-NEXT: pmullw %xmm2, %xmm3
154; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
155; SSE2-NEXT: pand %xmm2, %xmm3
156; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
157; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
158; SSE2-NEXT: pmullw %xmm1, %xmm0
159; SSE2-NEXT: pand %xmm2, %xmm0
160; SSE2-NEXT: packuswb %xmm3, %xmm0
161; SSE2-NEXT: retq
162;
163; SSE41-LABEL: mul_v8i8:
164; SSE41: # %bb.0:
165; SSE41-NEXT: pmovzxbw {{.*#+}} xmm3 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
166; SSE41-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
167; SSE41-NEXT: pmovzxbw {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
168; SSE41-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
169; SSE41-NEXT: pmullw %xmm1, %xmm0
170; SSE41-NEXT: movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
171; SSE41-NEXT: pand %xmm1, %xmm0
172; SSE41-NEXT: pmullw %xmm3, %xmm2
173; SSE41-NEXT: pand %xmm1, %xmm2
174; SSE41-NEXT: packuswb %xmm0, %xmm2
175; SSE41-NEXT: movdqa %xmm2, %xmm0
176; SSE41-NEXT: retq
177;
178; AVX2-LABEL: mul_v8i8:
179; AVX2: # %bb.0:
180; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
181; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
182; AVX2-NEXT: vpmullw %ymm1, %ymm0, %ymm0
183; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
184; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
185; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1
186; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0
187; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
188; AVX2-NEXT: vzeroupper
189; AVX2-NEXT: retq
190;
191; AVX512F-LABEL: mul_v8i8:
192; AVX512F: # %bb.0:
193; AVX512F-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
194; AVX512F-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
195; AVX512F-NEXT: vpmullw %ymm1, %ymm0, %ymm0
196; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
197; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
198; AVX512F-NEXT: vzeroupper
199; AVX512F-NEXT: retq
200;
201; AVX512BW-LABEL: mul_v8i8:
202; AVX512BW: # %bb.0:
203; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
204; AVX512BW-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
205; AVX512BW-NEXT: vpmullw %ymm1, %ymm0, %ymm0
206; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0
207; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
208; AVX512BW-NEXT: vzeroupper
209; AVX512BW-NEXT: retq
210 %res = mul <8 x i8> %x, %y
211 ret <8 x i8> %res
212}
213
214define <2 x i16> @mul_v2i16(<2 x i16> %x, <2 x i16> %y) {
215; SSE-LABEL: mul_v2i16:
216; SSE: # %bb.0:
217; SSE-NEXT: pmullw %xmm1, %xmm0
218; SSE-NEXT: retq
219;
220; AVX-LABEL: mul_v2i16:
221; AVX: # %bb.0:
222; AVX-NEXT: vpmullw %xmm1, %xmm0, %xmm0
223; AVX-NEXT: retq
224 %res = mul <2 x i16> %x, %y
225 ret <2 x i16> %res
226}
227
228define <4 x i16> @mul_v4i16(<4 x i16> %x, <4 x i16> %y) {
229; SSE-LABEL: mul_v4i16:
230; SSE: # %bb.0:
231; SSE-NEXT: pmullw %xmm1, %xmm0
232; SSE-NEXT: retq
233;
234; AVX-LABEL: mul_v4i16:
235; AVX: # %bb.0:
236; AVX-NEXT: vpmullw %xmm1, %xmm0, %xmm0
237; AVX-NEXT: retq
238 %res = mul <4 x i16> %x, %y
239 ret <4 x i16> %res
240}
241
242define <2 x i32> @mul_v2i32(<2 x i32> %x, <2 x i32> %y) {
243; SSE2-LABEL: mul_v2i32:
244; SSE2: # %bb.0:
245; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
246; SSE2-NEXT: pmuludq %xmm1, %xmm0
247; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
248; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
249; SSE2-NEXT: pmuludq %xmm2, %xmm1
250; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
251; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
252; SSE2-NEXT: retq
253;
254; SSE41-LABEL: mul_v2i32:
255; SSE41: # %bb.0:
256; SSE41-NEXT: pmulld %xmm1, %xmm0
257; SSE41-NEXT: retq
258;
259; AVX-LABEL: mul_v2i32:
260; AVX: # %bb.0:
261; AVX-NEXT: vpmulld %xmm1, %xmm0, %xmm0
262; AVX-NEXT: retq
263 %res = mul <2 x i32> %x, %y
264 ret <2 x i32> %res
265}