1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; These are tests for SSE3 codegen.
4 ; RUN: llc < %s -mtriple=x86_64-apple-darwin9 --mattr=+sse3 | FileCheck %s --check-prefix=X64
6 ; Test for v8xi16 lowering where we extract the first element of the vector and
7 ; placed it in the second element of the result.
9 define void @t0(<8 x i16>* %dest, <8 x i16>* %old) nounwind {
11 ; X64: ## BB#0: ## %entry
12 ; X64-NEXT: movl $1, %eax
13 ; X64-NEXT: movd %eax, %xmm0
14 ; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
15 ; X64-NEXT: movdqa %xmm0, (%rdi)
18 %tmp3 = load <8 x i16>, <8 x i16>* %old
19 %tmp6 = shufflevector <8 x i16> %tmp3,
20 <8 x i16> < i16 1, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef >,
21 <8 x i32> < i32 8, i32 0, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef >
22 store <8 x i16> %tmp6, <8 x i16>* %dest
26 define <8 x i16> @t1(<8 x i16>* %A, <8 x i16>* %B) nounwind {
29 ; X64-NEXT: movaps {{.*#+}} xmm0 = [0,65535,65535,65535,65535,65535,65535,65535]
30 ; X64-NEXT: movaps %xmm0, %xmm1
31 ; X64-NEXT: andnps (%rsi), %xmm1
32 ; X64-NEXT: andps (%rdi), %xmm0
33 ; X64-NEXT: orps %xmm1, %xmm0
35 %tmp1 = load <8 x i16>, <8 x i16>* %A
36 %tmp2 = load <8 x i16>, <8 x i16>* %B
37 %tmp3 = shufflevector <8 x i16> %tmp1, <8 x i16> %tmp2, <8 x i32> < i32 8, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 >
42 define <8 x i16> @t2(<8 x i16> %A, <8 x i16> %B) nounwind {
45 ; X64-NEXT: movdqa {{.*#+}} xmm2 = [0,65535,65535,0,65535,65535,65535,65535]
46 ; X64-NEXT: pand %xmm2, %xmm0
47 ; X64-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,1,2,1,4,5,6,7]
48 ; X64-NEXT: pandn %xmm1, %xmm2
49 ; X64-NEXT: por %xmm2, %xmm0
51 %tmp = shufflevector <8 x i16> %A, <8 x i16> %B, <8 x i32> < i32 9, i32 1, i32 2, i32 9, i32 4, i32 5, i32 6, i32 7 >
55 define <8 x i16> @t3(<8 x i16> %A, <8 x i16> %B) nounwind {
58 ; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
59 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
60 ; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
61 ; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
62 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
64 %tmp = shufflevector <8 x i16> %A, <8 x i16> %A, <8 x i32> < i32 8, i32 3, i32 2, i32 13, i32 7, i32 6, i32 5, i32 4 >
68 define <8 x i16> @t4(<8 x i16> %A, <8 x i16> %B) nounwind {
71 ; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
72 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
73 ; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
74 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,7,4,7]
76 %tmp = shufflevector <8 x i16> %A, <8 x i16> %B, <8 x i32> < i32 0, i32 7, i32 2, i32 3, i32 1, i32 5, i32 6, i32 5 >
80 define <8 x i16> @t5(<8 x i16> %A, <8 x i16> %B) nounwind {
83 ; X64-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
84 ; X64-NEXT: movdqa %xmm1, %xmm0
86 %tmp = shufflevector <8 x i16> %A, <8 x i16> %B, <8 x i32> < i32 8, i32 9, i32 0, i32 1, i32 10, i32 11, i32 2, i32 3 >
90 define <8 x i16> @t6(<8 x i16> %A, <8 x i16> %B) nounwind {
93 ; X64-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
95 %tmp = shufflevector <8 x i16> %A, <8 x i16> %B, <8 x i32> < i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 >
99 define <8 x i16> @t7(<8 x i16> %A, <8 x i16> %B) nounwind {
102 ; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
103 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
105 %tmp = shufflevector <8 x i16> %A, <8 x i16> %B, <8 x i32> < i32 0, i32 0, i32 3, i32 2, i32 4, i32 6, i32 4, i32 7 >
109 define void @t8(<2 x i64>* %res, <2 x i64>* %A) nounwind {
112 ; X64-NEXT: pshuflw {{.*#+}} xmm0 = mem[2,1,0,3,4,5,6,7]
113 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
114 ; X64-NEXT: movdqa %xmm0, (%rdi)
116 %tmp = load <2 x i64>, <2 x i64>* %A
117 %tmp.upgrd.1 = bitcast <2 x i64> %tmp to <8 x i16>
118 %tmp0 = extractelement <8 x i16> %tmp.upgrd.1, i32 0
119 %tmp1 = extractelement <8 x i16> %tmp.upgrd.1, i32 1
120 %tmp2 = extractelement <8 x i16> %tmp.upgrd.1, i32 2
121 %tmp3 = extractelement <8 x i16> %tmp.upgrd.1, i32 3
122 %tmp4 = extractelement <8 x i16> %tmp.upgrd.1, i32 4
123 %tmp5 = extractelement <8 x i16> %tmp.upgrd.1, i32 5
124 %tmp6 = extractelement <8 x i16> %tmp.upgrd.1, i32 6
125 %tmp7 = extractelement <8 x i16> %tmp.upgrd.1, i32 7
126 %tmp8 = insertelement <8 x i16> undef, i16 %tmp2, i32 0
127 %tmp9 = insertelement <8 x i16> %tmp8, i16 %tmp1, i32 1
128 %tmp10 = insertelement <8 x i16> %tmp9, i16 %tmp0, i32 2
129 %tmp11 = insertelement <8 x i16> %tmp10, i16 %tmp3, i32 3
130 %tmp12 = insertelement <8 x i16> %tmp11, i16 %tmp6, i32 4
131 %tmp13 = insertelement <8 x i16> %tmp12, i16 %tmp5, i32 5
132 %tmp14 = insertelement <8 x i16> %tmp13, i16 %tmp4, i32 6
133 %tmp15 = insertelement <8 x i16> %tmp14, i16 %tmp7, i32 7
134 %tmp15.upgrd.2 = bitcast <8 x i16> %tmp15 to <2 x i64>
135 store <2 x i64> %tmp15.upgrd.2, <2 x i64>* %res
139 define void @t9(<4 x float>* %r, <2 x i32>* %A) nounwind {
142 ; X64-NEXT: movapd (%rdi), %xmm0
143 ; X64-NEXT: movhpd (%rsi), %xmm0
144 ; X64-NEXT: movapd %xmm0, (%rdi)
146 %tmp = load <4 x float>, <4 x float>* %r
147 %tmp.upgrd.3 = bitcast <2 x i32>* %A to double*
148 %tmp.upgrd.4 = load double, double* %tmp.upgrd.3
149 %tmp.upgrd.5 = insertelement <2 x double> undef, double %tmp.upgrd.4, i32 0
150 %tmp5 = insertelement <2 x double> %tmp.upgrd.5, double undef, i32 1
151 %tmp6 = bitcast <2 x double> %tmp5 to <4 x float>
152 %tmp.upgrd.6 = extractelement <4 x float> %tmp, i32 0
153 %tmp7 = extractelement <4 x float> %tmp, i32 1
154 %tmp8 = extractelement <4 x float> %tmp6, i32 0
155 %tmp9 = extractelement <4 x float> %tmp6, i32 1
156 %tmp10 = insertelement <4 x float> undef, float %tmp.upgrd.6, i32 0
157 %tmp11 = insertelement <4 x float> %tmp10, float %tmp7, i32 1
158 %tmp12 = insertelement <4 x float> %tmp11, float %tmp8, i32 2
159 %tmp13 = insertelement <4 x float> %tmp12, float %tmp9, i32 3
160 store <4 x float> %tmp13, <4 x float>* %r
166 ; FIXME: This testcase produces icky code. It can be made much better!
169 @g1 = external constant <4 x i32>
170 @g2 = external constant <4 x i16>
172 define void @t10() nounwind {
175 ; X64-NEXT: movq _g1@{{.*}}(%rip), %rax
176 ; X64-NEXT: pshuflw {{.*#+}} xmm0 = mem[0,2,2,3,4,5,6,7]
177 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
178 ; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
179 ; X64-NEXT: movq _g2@{{.*}}(%rip), %rax
180 ; X64-NEXT: movq %xmm0, (%rax)
182 load <4 x i32>, <4 x i32>* @g1, align 16
183 bitcast <4 x i32> %1 to <8 x i16>
184 shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> < i32 0, i32 2, i32 4, i32 6, i32 undef, i32 undef, i32 undef, i32 undef >
185 bitcast <8 x i16> %3 to <2 x i64>
186 extractelement <2 x i64> %4, i32 0
187 bitcast i64 %5 to <4 x i16>
188 store <4 x i16> %6, <4 x i16>* @g2, align 8
192 ; Pack various elements via shuffles.
193 define <8 x i16> @t11(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
195 ; X64: ## BB#0: ## %entry
196 ; X64-NEXT: psrld $16, %xmm0
197 ; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
200 %tmp7 = shufflevector <8 x i16> %T0, <8 x i16> %T1, <8 x i32> < i32 1, i32 8, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef , i32 undef >
205 define <8 x i16> @t12(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
207 ; X64: ## BB#0: ## %entry
208 ; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
209 ; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
210 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,6,7]
213 %tmp9 = shufflevector <8 x i16> %T0, <8 x i16> %T1, <8 x i32> < i32 0, i32 1, i32 undef, i32 undef, i32 3, i32 11, i32 undef , i32 undef >
218 define <8 x i16> @t13(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
220 ; X64: ## BB#0: ## %entry
221 ; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
222 ; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
223 ; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,6,7]
226 %tmp9 = shufflevector <8 x i16> %T0, <8 x i16> %T1, <8 x i32> < i32 8, i32 9, i32 undef, i32 undef, i32 11, i32 3, i32 undef , i32 undef >
230 define <8 x i16> @t14(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
232 ; X64: ## BB#0: ## %entry
233 ; X64-NEXT: psrlq $16, %xmm0
234 ; X64-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
235 ; X64-NEXT: movdqa %xmm1, %xmm0
238 %tmp9 = shufflevector <8 x i16> %T0, <8 x i16> %T1, <8 x i32> < i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 2, i32 undef , i32 undef >
242 ; FIXME: t15 is worse off from disabling of scheduler 2-address hack.
243 define <8 x i16> @t15(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
245 ; X64: ## BB#0: ## %entry
246 ; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
247 ; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
248 ; X64-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
251 %tmp8 = shufflevector <8 x i16> %T0, <8 x i16> %T1, <8 x i32> < i32 undef, i32 undef, i32 7, i32 2, i32 8, i32 undef, i32 undef , i32 undef >
255 ; Test yonah where we convert a shuffle to pextrw and pinrsw
256 define <16 x i8> @t16(<16 x i8> %T0) nounwind readnone {
258 ; X64: ## BB#0: ## %entry
259 ; X64-NEXT: movdqa {{.*#+}} xmm1 = [0,0,0,0,1,1,1,1,0,0,0,0,0,0,0,0]
260 ; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
261 ; X64-NEXT: movdqa %xmm1, %xmm0
264 %tmp8 = shufflevector <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 1, i8 1, i8 1, i8 1, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, <16 x i8> %T0, <16 x i32> < i32 0, i32 1, i32 16, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef , i32 undef >
265 %tmp9 = shufflevector <16 x i8> %tmp8, <16 x i8> %T0, <16 x i32> < i32 0, i32 1, i32 2, i32 17, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef , i32 undef >
270 define <4 x i32> @t17() nounwind {
272 ; X64: ## BB#0: ## %entry
273 ; X64-NEXT: movaps (%rax), %xmm0
274 ; X64-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0,0,1,1]
275 ; X64-NEXT: pxor %xmm1, %xmm1
276 ; X64-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
279 %tmp1 = load <4 x float>, <4 x float>* undef, align 16
280 %tmp2 = shufflevector <4 x float> %tmp1, <4 x float> undef, <4 x i32> <i32 4, i32 1, i32 2, i32 3>
281 %tmp3 = load <4 x float>, <4 x float>* undef, align 16
282 %tmp4 = shufflevector <4 x float> %tmp2, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 1>
283 %tmp5 = bitcast <4 x float> %tmp3 to <4 x i32>
284 %tmp6 = shufflevector <4 x i32> %tmp5, <4 x i32> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 1>
285 %tmp7 = and <4 x i32> %tmp6, <i32 undef, i32 undef, i32 -1, i32 0>