1; These are tests for SSE3 codegen.
2
3; RUN: llc < %s -march=x86-64 -mcpu=nocona -mtriple=i686-apple-darwin9 -O3 | FileCheck %s --check-prefix=X64
4
5; Test for v8xi16 lowering where we extract the first element of the vector and
6; placed it in the second element of the result.
7
8define void @t0(<8 x i16>* %dest, <8 x i16>* %old) nounwind {
9; X64-LABEL: t0:
10; X64:       ## BB#0: ## %entry
11; X64-NEXT:    movl $1, %eax
12; X64-NEXT:    movd %eax, %xmm0
13; X64-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
14; X64-NEXT:    movdqa %xmm0, (%rdi)
15; X64-NEXT:    retq
16entry:
17	%tmp3 = load <8 x i16>, <8 x i16>* %old
18	%tmp6 = shufflevector <8 x i16> %tmp3,
19                <8 x i16> < i16 1, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef >,
20                <8 x i32> < i32 8, i32 0, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef  >
21	store <8 x i16> %tmp6, <8 x i16>* %dest
22	ret void
23}
24
25define <8 x i16> @t1(<8 x i16>* %A, <8 x i16>* %B) nounwind {
26; X64-LABEL: t1:
27; X64:       ## BB#0:
28; X64-NEXT:    movaps {{.*#+}} xmm0 = [0,65535,65535,65535,65535,65535,65535,65535]
29; X64-NEXT:    movaps %xmm0, %xmm1
30; X64-NEXT:    andnps (%rsi), %xmm1
31; X64-NEXT:    andps (%rdi), %xmm0
32; X64-NEXT:    orps %xmm1, %xmm0
33; X64-NEXT:    retq
34	%tmp1 = load <8 x i16>, <8 x i16>* %A
35	%tmp2 = load <8 x i16>, <8 x i16>* %B
36	%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 >
37	ret <8 x i16> %tmp3
38
39}
40
41define <8 x i16> @t2(<8 x i16> %A, <8 x i16> %B) nounwind {
42; X64-LABEL: t2:
43; X64:       ## BB#0:
44; X64-NEXT:    movdqa {{.*#+}} xmm2 = [0,65535,65535,0,65535,65535,65535,65535]
45; X64-NEXT:    pand %xmm2, %xmm0
46; X64-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,1,2,1,4,5,6,7]
47; X64-NEXT:    pandn %xmm1, %xmm2
48; X64-NEXT:    por %xmm2, %xmm0
49; X64-NEXT:    retq
50	%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 >
51	ret <8 x i16> %tmp
52}
53
54define <8 x i16> @t3(<8 x i16> %A, <8 x i16> %B) nounwind {
55; X64-LABEL: t3:
56; X64:       ## BB#0:
57; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
58; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
59; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
60; X64-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
61; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
62; X64-NEXT:    retq
63	%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 >
64	ret <8 x i16> %tmp
65}
66
67define <8 x i16> @t4(<8 x i16> %A, <8 x i16> %B) nounwind {
68; X64-LABEL: t4:
69; X64:       ## BB#0:
70; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
71; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
72; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
73; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,7,4,7]
74; X64-NEXT:    retq
75	%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 >
76	ret <8 x i16> %tmp
77}
78
79define <8 x i16> @t5(<8 x i16> %A, <8 x i16> %B) nounwind {
80; X64-LABEL: t5:
81; X64:       ## BB#0:
82; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
83; X64-NEXT:    movdqa %xmm1, %xmm0
84; X64-NEXT:    retq
85	%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 >
86	ret <8 x i16> %tmp
87}
88
89define <8 x i16> @t6(<8 x i16> %A, <8 x i16> %B) nounwind {
90; X64-LABEL: t6:
91; X64:       ## BB#0:
92; X64-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
93; X64-NEXT:    retq
94	%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 >
95	ret <8 x i16> %tmp
96}
97
98define <8 x i16> @t7(<8 x i16> %A, <8 x i16> %B) nounwind {
99; X64-LABEL: t7:
100; X64:       ## BB#0:
101; X64-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
102; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
103; X64-NEXT:    retq
104	%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 >
105	ret <8 x i16> %tmp
106}
107
108define void @t8(<2 x i64>* %res, <2 x i64>* %A) nounwind {
109; X64-LABEL: t8:
110; X64:       ## BB#0:
111; X64-NEXT:    pshuflw {{.*#+}} xmm0 = mem[2,1,0,3,4,5,6,7]
112; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
113; X64-NEXT:    movdqa %xmm0, (%rdi)
114; X64-NEXT:    retq
115	%tmp = load <2 x i64>, <2 x i64>* %A
116	%tmp.upgrd.1 = bitcast <2 x i64> %tmp to <8 x i16>
117	%tmp0 = extractelement <8 x i16> %tmp.upgrd.1, i32 0
118	%tmp1 = extractelement <8 x i16> %tmp.upgrd.1, i32 1
119	%tmp2 = extractelement <8 x i16> %tmp.upgrd.1, i32 2
120	%tmp3 = extractelement <8 x i16> %tmp.upgrd.1, i32 3
121	%tmp4 = extractelement <8 x i16> %tmp.upgrd.1, i32 4
122	%tmp5 = extractelement <8 x i16> %tmp.upgrd.1, i32 5
123	%tmp6 = extractelement <8 x i16> %tmp.upgrd.1, i32 6
124	%tmp7 = extractelement <8 x i16> %tmp.upgrd.1, i32 7
125	%tmp8 = insertelement <8 x i16> undef, i16 %tmp2, i32 0
126	%tmp9 = insertelement <8 x i16> %tmp8, i16 %tmp1, i32 1
127	%tmp10 = insertelement <8 x i16> %tmp9, i16 %tmp0, i32 2
128	%tmp11 = insertelement <8 x i16> %tmp10, i16 %tmp3, i32 3
129	%tmp12 = insertelement <8 x i16> %tmp11, i16 %tmp6, i32 4
130	%tmp13 = insertelement <8 x i16> %tmp12, i16 %tmp5, i32 5
131	%tmp14 = insertelement <8 x i16> %tmp13, i16 %tmp4, i32 6
132	%tmp15 = insertelement <8 x i16> %tmp14, i16 %tmp7, i32 7
133	%tmp15.upgrd.2 = bitcast <8 x i16> %tmp15 to <2 x i64>
134	store <2 x i64> %tmp15.upgrd.2, <2 x i64>* %res
135	ret void
136}
137
138define void @t9(<4 x float>* %r, <2 x i32>* %A) nounwind {
139; X64-LABEL: t9:
140; X64:       ## BB#0:
141; X64-NEXT:    movapd (%rdi), %xmm0
142; X64-NEXT:    movhpd (%rsi), %xmm0
143; X64-NEXT:    movapd %xmm0, (%rdi)
144; X64-NEXT:    retq
145	%tmp = load <4 x float>, <4 x float>* %r
146	%tmp.upgrd.3 = bitcast <2 x i32>* %A to double*
147	%tmp.upgrd.4 = load double, double* %tmp.upgrd.3
148	%tmp.upgrd.5 = insertelement <2 x double> undef, double %tmp.upgrd.4, i32 0
149	%tmp5 = insertelement <2 x double> %tmp.upgrd.5, double undef, i32 1
150	%tmp6 = bitcast <2 x double> %tmp5 to <4 x float>
151	%tmp.upgrd.6 = extractelement <4 x float> %tmp, i32 0
152	%tmp7 = extractelement <4 x float> %tmp, i32 1
153	%tmp8 = extractelement <4 x float> %tmp6, i32 0
154	%tmp9 = extractelement <4 x float> %tmp6, i32 1
155	%tmp10 = insertelement <4 x float> undef, float %tmp.upgrd.6, i32 0
156	%tmp11 = insertelement <4 x float> %tmp10, float %tmp7, i32 1
157	%tmp12 = insertelement <4 x float> %tmp11, float %tmp8, i32 2
158	%tmp13 = insertelement <4 x float> %tmp12, float %tmp9, i32 3
159	store <4 x float> %tmp13, <4 x float>* %r
160	ret void
161}
162
163
164
165; FIXME: This testcase produces icky code. It can be made much better!
166; PR2585
167
168@g1 = external constant <4 x i32>
169@g2 = external constant <4 x i16>
170
171define void @t10() nounwind {
172; X64-LABEL: t10:
173; X64:       ## BB#0:
174; X64-NEXT:    movq _g1@{{.*}}(%rip), %rax
175; X64-NEXT:    pshuflw {{.*#+}} xmm0 = mem[0,2,2,3,4,5,6,7]
176; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
177; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
178; X64-NEXT:    movq _g2@{{.*}}(%rip), %rax
179; X64-NEXT:    movq %xmm0, (%rax)
180; X64-NEXT:    retq
181  load <4 x i32>, <4 x i32>* @g1, align 16
182  bitcast <4 x i32> %1 to <8 x i16>
183  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 >
184  bitcast <8 x i16> %3 to <2 x i64>
185  extractelement <2 x i64> %4, i32 0
186  bitcast i64 %5 to <4 x i16>
187  store <4 x i16> %6, <4 x i16>* @g2, align 8
188  ret void
189}
190
191; Pack various elements via shuffles.
192define <8 x i16> @t11(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
193; X64-LABEL: t11:
194; X64:       ## BB#0: ## %entry
195; X64-NEXT:    psrld $16, %xmm0
196; X64-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
197; X64-NEXT:    retq
198entry:
199	%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 >
200	ret <8 x i16> %tmp7
201
202}
203
204define <8 x i16> @t12(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
205; X64-LABEL: t12:
206; X64:       ## BB#0: ## %entry
207; X64-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
208; X64-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
209; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,6,7]
210; X64-NEXT:    retq
211entry:
212	%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 >
213	ret <8 x i16> %tmp9
214
215}
216
217define <8 x i16> @t13(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
218; X64-LABEL: t13:
219; X64:       ## BB#0: ## %entry
220; X64-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
221; X64-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
222; X64-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,6,7]
223; X64-NEXT:    retq
224entry:
225	%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 >
226	ret <8 x i16> %tmp9
227}
228
229define <8 x i16> @t14(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
230; X64-LABEL: t14:
231; X64:       ## BB#0: ## %entry
232; X64-NEXT:    psrlq $16, %xmm0
233; X64-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
234; X64-NEXT:    movdqa %xmm1, %xmm0
235; X64-NEXT:    retq
236entry:
237	%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 >
238	ret <8 x i16> %tmp9
239}
240
241; FIXME: t15 is worse off from disabling of scheduler 2-address hack.
242define <8 x i16> @t15(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone {
243; X64-LABEL: t15:
244; X64:       ## BB#0: ## %entry
245; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
246; X64-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
247; X64-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
248; X64-NEXT:    retq
249entry:
250  %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 >
251  ret <8 x i16> %tmp8
252}
253
254; Test yonah where we convert a shuffle to pextrw and pinrsw
255define <16 x i8> @t16(<16 x i8> %T0) nounwind readnone {
256; X64-LABEL: t16:
257; X64:       ## BB#0: ## %entry
258; X64-NEXT:    movdqa {{.*#+}} xmm1 = [0,0,0,0,1,1,1,1,0,0,0,0,0,0,0,0]
259; X64-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
260; X64-NEXT:    movdqa %xmm1, %xmm0
261; X64-NEXT:    retq
262entry:
263  %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 >
264  %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 >
265  ret <16 x i8> %tmp9
266}
267
268; rdar://8520311
269define <4 x i32> @t17() nounwind {
270; X64-LABEL: t17:
271; X64:       ## BB#0: ## %entry
272; X64-NEXT:    movddup {{.*#+}} xmm0 = mem[0,0]
273; X64-NEXT:    andpd {{.*}}(%rip), %xmm0
274; X64-NEXT:    retq
275entry:
276  %tmp1 = load <4 x float>, <4 x float>* undef, align 16
277  %tmp2 = shufflevector <4 x float> %tmp1, <4 x float> undef, <4 x i32> <i32 4, i32 1, i32 2, i32 3>
278  %tmp3 = load <4 x float>, <4 x float>* undef, align 16
279  %tmp4 = shufflevector <4 x float> %tmp2, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 1>
280  %tmp5 = bitcast <4 x float> %tmp3 to <4 x i32>
281  %tmp6 = shufflevector <4 x i32> %tmp5, <4 x i32> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 1>
282  %tmp7 = and <4 x i32> %tmp6, <i32 undef, i32 undef, i32 -1, i32 0>
283  ret <4 x i32> %tmp7
284}
285