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]
18 %0 = trunc <2 x i64> %a to <2 x i32>
19 %1 = trunc <2 x i64> %b to <2 x i32>
20 %2 = shufflevector <2 x i32> %0, <2 x i32> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
24 define i64 @trunc2i64(<2 x i64> %inval) {
25 ; SSE-LABEL: trunc2i64:
26 ; SSE: # BB#0: # %entry
27 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
28 ; SSE-NEXT: movd %xmm0, %rax
31 ; AVX-LABEL: trunc2i64:
32 ; AVX: # BB#0: # %entry
33 ; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
34 ; AVX-NEXT: vmovq %xmm0, %rax
38 %0 = trunc <2 x i64> %inval to <2 x i32>
39 %1 = bitcast <2 x i32> %0 to i64
43 define <8 x i16> @trunc2x4i32(<4 x i32> %a, <4 x i32> %b) {
44 ; SSE2-LABEL: trunc2x4i32:
45 ; SSE2: # BB#0: # %entry
46 ; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
47 ; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
48 ; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
49 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
50 ; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
51 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
52 ; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
55 ; SSSE3-LABEL: trunc2x4i32:
56 ; SSSE3: # BB#0: # %entry
57 ; SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
58 ; SSSE3-NEXT: pshufb %xmm2, %xmm1
59 ; SSSE3-NEXT: pshufb %xmm2, %xmm0
60 ; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
63 ; SSE41-LABEL: trunc2x4i32:
64 ; SSE41: # BB#0: # %entry
65 ; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
66 ; SSE41-NEXT: pshufb %xmm2, %xmm1
67 ; SSE41-NEXT: pshufb %xmm2, %xmm0
68 ; SSE41-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
71 ; AVX-LABEL: trunc2x4i32:
72 ; AVX: # BB#0: # %entry
73 ; AVX-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
74 ; AVX-NEXT: vpshufb %xmm2, %xmm1, %xmm1
75 ; AVX-NEXT: vpshufb %xmm2, %xmm0, %xmm0
76 ; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
80 %0 = trunc <4 x i32> %a to <4 x i16>
81 %1 = trunc <4 x i32> %b to <4 x i16>
82 %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>
86 ; PR15524 http://llvm.org/bugs/show_bug.cgi?id=15524
87 define i64 @trunc4i32(<4 x i32> %inval) {
88 ; SSE2-LABEL: trunc4i32:
89 ; SSE2: # BB#0: # %entry
90 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
91 ; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
92 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
93 ; SSE2-NEXT: movd %xmm0, %rax
96 ; SSSE3-LABEL: trunc4i32:
97 ; SSSE3: # BB#0: # %entry
98 ; SSSE3-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
99 ; SSSE3-NEXT: movd %xmm0, %rax
102 ; SSE41-LABEL: trunc4i32:
103 ; SSE41: # BB#0: # %entry
104 ; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
105 ; SSE41-NEXT: movd %xmm0, %rax
108 ; AVX-LABEL: trunc4i32:
109 ; AVX: # BB#0: # %entry
110 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
111 ; AVX-NEXT: vmovq %xmm0, %rax
115 %0 = trunc <4 x i32> %inval to <4 x i16>
116 %1 = bitcast <4 x i16> %0 to i64
120 define <16 x i8> @trunc2x8i16(<8 x i16> %a, <8 x i16> %b) {
121 ; SSE2-LABEL: trunc2x8i16:
122 ; SSE2: # BB#0: # %entry
123 ; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
124 ; SSE2-NEXT: pand %xmm2, %xmm1
125 ; SSE2-NEXT: pand %xmm2, %xmm0
126 ; SSE2-NEXT: packuswb %xmm1, %xmm0
129 ; SSSE3-LABEL: trunc2x8i16:
130 ; SSSE3: # BB#0: # %entry
131 ; SSSE3-NEXT: pshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,2,4,6,8,10,12,14]
132 ; SSSE3-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero
133 ; SSSE3-NEXT: por %xmm1, %xmm0
136 ; SSE41-LABEL: trunc2x8i16:
137 ; SSE41: # BB#0: # %entry
138 ; SSE41-NEXT: pshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,2,4,6,8,10,12,14]
139 ; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero
140 ; SSE41-NEXT: por %xmm1, %xmm0
143 ; AVX-LABEL: trunc2x8i16:
144 ; AVX: # BB#0: # %entry
145 ; AVX-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,2,4,6,8,10,12,14]
146 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14],zero,zero,zero,zero,zero,zero,zero,zero
147 ; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
151 %0 = trunc <8 x i16> %a to <8 x i8>
152 %1 = trunc <8 x i16> %b to <8 x i8>
153 %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>
157 ; PR15524 http://llvm.org/bugs/show_bug.cgi?id=15524
158 define i64 @trunc8i16(<8 x i16> %inval) {
159 ; SSE2-LABEL: trunc8i16:
160 ; SSE2: # BB#0: # %entry
161 ; SSE2-NEXT: pand .LCP{{.*}}(%rip), %xmm0
162 ; SSE2-NEXT: packuswb %xmm0, %xmm0
163 ; SSE2-NEXT: movd %xmm0, %rax
166 ; SSSE3-LABEL: trunc8i16:
167 ; SSSE3: # BB#0: # %entry
168 ; SSSE3-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
169 ; SSSE3-NEXT: movd %xmm0, %rax
172 ; SSE41-LABEL: trunc8i16:
173 ; SSE41: # BB#0: # %entry
174 ; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
175 ; SSE41-NEXT: movd %xmm0, %rax
178 ; AVX-LABEL: trunc8i16:
179 ; AVX: # BB#0: # %entry
180 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
181 ; AVX-NEXT: vmovq %xmm0, %rax
185 %0 = trunc <8 x i16> %inval to <8 x i8>
186 %1 = bitcast <8 x i8> %0 to i64