1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+ssse3 | FileCheck %s --check-prefix=SSE --check-prefix=SSSE3
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE --check-prefix=SSE41
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
6 define <4 x i32> @trunc2x2i64(<2 x i64> %a, <2 x i64> %b) {
7 ; SSE-LABEL: trunc2x2i64:
8 ; SSE: # BB#0: # %entry
9 ; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
12 ; AVX-LABEL: trunc2x2i64:
13 ; AVX: # BB#0: # %entry
14 ; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
19 %0 = trunc <2 x i64> %a to <2 x i32>
20 %1 = trunc <2 x i64> %b to <2 x i32>
21 %2 = shufflevector <2 x i32> %0, <2 x i32> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
25 define i64 @trunc2i64(<2 x i64> %inval) {
26 ; SSE-LABEL: trunc2i64:
27 ; SSE: # BB#0: # %entry
28 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
29 ; SSE-NEXT: movd %xmm0, %rax
32 ; AVX-LABEL: trunc2i64:
33 ; AVX: # BB#0: # %entry
34 ; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
35 ; AVX-NEXT: vmovq %xmm0, %rax
40 %0 = trunc <2 x i64> %inval to <2 x i32>
41 %1 = bitcast <2 x i32> %0 to i64
45 define <8 x i16> @trunc2x4i32(<4 x i32> %a, <4 x i32> %b) {
46 ; SSE2-LABEL: trunc2x4i32:
47 ; SSE2: # BB#0: # %entry
48 ; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
49 ; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
50 ; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
51 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
52 ; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
53 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
54 ; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
57 ; SSSE3-LABEL: trunc2x4i32:
58 ; SSSE3: # BB#0: # %entry
59 ; SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
60 ; SSSE3-NEXT: pshufb %xmm2, %xmm1
61 ; SSSE3-NEXT: pshufb %xmm2, %xmm0
62 ; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
65 ; SSE41-LABEL: trunc2x4i32:
66 ; SSE41: # BB#0: # %entry
67 ; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
68 ; SSE41-NEXT: pshufb %xmm2, %xmm1
69 ; SSE41-NEXT: pshufb %xmm2, %xmm0
70 ; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
73 ; AVX-LABEL: trunc2x4i32:
74 ; AVX: # BB#0: # %entry
75 ; AVX-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
76 ; AVX-NEXT: vpshufb %xmm2, %xmm1, %xmm1
77 ; AVX-NEXT: vpshufb %xmm2, %xmm0, %xmm0
78 ; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
85 %0 = trunc <4 x i32> %a to <4 x i16>
86 %1 = trunc <4 x i32> %b to <4 x i16>
87 %2 = shufflevector <4 x i16> %0, <4 x i16> %1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
91 ; PR15524 http://llvm.org/bugs/show_bug.cgi?id=15524
92 define i64 @trunc4i32(<4 x i32> %inval) {
93 ; SSE2-LABEL: trunc4i32:
94 ; SSE2: # BB#0: # %entry
95 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
96 ; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
97 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
98 ; SSE2-NEXT: movd %xmm0, %rax
101 ; SSSE3-LABEL: trunc4i32:
102 ; SSSE3: # BB#0: # %entry
103 ; SSSE3-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
104 ; SSSE3-NEXT: movd %xmm0, %rax
107 ; SSE41-LABEL: trunc4i32:
108 ; SSE41: # BB#0: # %entry
109 ; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
110 ; SSE41-NEXT: movd %xmm0, %rax
113 ; AVX-LABEL: trunc4i32:
114 ; AVX: # BB#0: # %entry
115 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
116 ; AVX-NEXT: vmovq %xmm0, %rax
123 %0 = trunc <4 x i32> %inval to <4 x i16>
124 %1 = bitcast <4 x i16> %0 to i64
128 define <16 x i8> @trunc2x8i16(<8 x i16> %a, <8 x i16> %b) {
129 ; SSE2-LABEL: trunc2x8i16:
130 ; SSE2: # BB#0: # %entry
131 ; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
132 ; SSE2-NEXT: pand %xmm2, %xmm1
133 ; SSE2-NEXT: pand %xmm2, %xmm0
134 ; SSE2-NEXT: packuswb %xmm1, %xmm0
137 ; SSSE3-LABEL: trunc2x8i16:
138 ; SSSE3: # BB#0: # %entry
139 ; SSSE3-NEXT: pshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,2,4,6,8,10,12,14]
140 ; SSSE3-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero
141 ; SSSE3-NEXT: por %xmm1, %xmm0
144 ; SSE41-LABEL: trunc2x8i16:
145 ; SSE41: # BB#0: # %entry
146 ; SSE41-NEXT: pshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,2,4,6,8,10,12,14]
147 ; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero
148 ; SSE41-NEXT: por %xmm1, %xmm0
151 ; AVX-LABEL: trunc2x8i16:
152 ; AVX: # BB#0: # %entry
153 ; AVX-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,2,4,6,8,10,12,14]
154 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero
155 ; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
162 %0 = trunc <8 x i16> %a to <8 x i8>
163 %1 = trunc <8 x i16> %b to <8 x i8>
164 %2 = shufflevector <8 x i8> %0, <8 x i8> %1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
168 ; PR15524 http://llvm.org/bugs/show_bug.cgi?id=15524
169 define i64 @trunc8i16(<8 x i16> %inval) {
170 ; SSE2-LABEL: trunc8i16:
171 ; SSE2: # BB#0: # %entry
172 ; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
173 ; SSE2-NEXT: packuswb %xmm0, %xmm0
174 ; SSE2-NEXT: movd %xmm0, %rax
177 ; SSSE3-LABEL: trunc8i16:
178 ; SSSE3: # BB#0: # %entry
179 ; SSSE3-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
180 ; SSSE3-NEXT: movd %xmm0, %rax
183 ; SSE41-LABEL: trunc8i16:
184 ; SSE41: # BB#0: # %entry
185 ; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
186 ; SSE41-NEXT: movd %xmm0, %rax
189 ; AVX-LABEL: trunc8i16:
190 ; AVX: # BB#0: # %entry
191 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
192 ; AVX-NEXT: vmovq %xmm0, %rax
199 %0 = trunc <8 x i16> %inval to <8 x i8>
200 %1 = bitcast <8 x i8> %0 to i64