1; RUN: llc -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr8 < %s | FileCheck %s
2; RUN: llc -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr8 < %s | FileCheck %s
3; RUN: llc -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr7 -mattr=+crypto < %s | FileCheck %s
4; FIXME: llc -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr8 -mattr=-vsx < %s | FileCheck %s
5; FIXME: The original intent was to add a check-next for the blr after every check.
6; However, this currently fails since we don't eliminate stores of the unused
7; locals. These stores are sometimes scheduled after the crypto instruction
8
9; Function Attrs: nounwind
10define <16 x i8> @test_vpmsumb() #0 {
11entry:
12  %a = alloca <16 x i8>, align 16
13  %b = alloca <16 x i8>, align 16
14  store <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16>, <16 x i8>* %a, align 16
15  store <16 x i8> <i8 113, i8 114, i8 115, i8 116, i8 117, i8 118, i8 119, i8 120, i8 121, i8 122, i8 123, i8 124, i8 125, i8 126, i8 127, i8 112>, <16 x i8>* %b, align 16
16  %0 = load <16 x i8>,  <16 x i8>* %a, align 16
17  %1 = load <16 x i8>,  <16 x i8>* %b, align 16
18  %2 = call <16 x i8> @llvm.ppc.altivec.crypto.vpmsumb(<16 x i8> %0, <16 x i8> %1)
19  ret <16 x i8> %2
20; CHECK: vpmsumb 2,
21}
22
23; Function Attrs: nounwind readnone
24declare <16 x i8> @llvm.ppc.altivec.crypto.vpmsumb(<16 x i8>, <16 x i8>) #1
25
26; Function Attrs: nounwind
27define <8 x i16> @test_vpmsumh() #0 {
28entry:
29  %a = alloca <8 x i16>, align 16
30  %b = alloca <8 x i16>, align 16
31  store <8 x i16> <i16 258, i16 772, i16 1286, i16 1800, i16 2314, i16 2828, i16 3342, i16 3856>, <8 x i16>* %a, align 16
32  store <8 x i16> <i16 29042, i16 29556, i16 30070, i16 30584, i16 31098, i16 31612, i16 32126, i16 32624>, <8 x i16>* %b, align 16
33  %0 = load <8 x i16>,  <8 x i16>* %a, align 16
34  %1 = load <8 x i16>,  <8 x i16>* %b, align 16
35  %2 = call <8 x i16> @llvm.ppc.altivec.crypto.vpmsumh(<8 x i16> %0, <8 x i16> %1)
36  ret <8 x i16> %2
37; CHECK: vpmsumh 2,
38}
39
40; Function Attrs: nounwind readnone
41declare <8 x i16> @llvm.ppc.altivec.crypto.vpmsumh(<8 x i16>, <8 x i16>) #1
42
43; Function Attrs: nounwind
44define <4 x i32> @test_vpmsumw() #0 {
45entry:
46  %a = alloca <4 x i32>, align 16
47  %b = alloca <4 x i32>, align 16
48  store <4 x i32> <i32 16909060, i32 84281096, i32 151653132, i32 219025168>, <4 x i32>* %a, align 16
49  store <4 x i32> <i32 1903326068, i32 1970698104, i32 2038070140, i32 2105442160>, <4 x i32>* %b, align 16
50  %0 = load <4 x i32>,  <4 x i32>* %a, align 16
51  %1 = load <4 x i32>,  <4 x i32>* %b, align 16
52  %2 = call <4 x i32> @llvm.ppc.altivec.crypto.vpmsumw(<4 x i32> %0, <4 x i32> %1)
53  ret <4 x i32> %2
54; CHECK: vpmsumw 2,
55}
56
57; Function Attrs: nounwind readnone
58declare <4 x i32> @llvm.ppc.altivec.crypto.vpmsumw(<4 x i32>, <4 x i32>) #1
59
60; Function Attrs: nounwind
61define <2 x i64> @test_vpmsumd() #0 {
62entry:
63  %a = alloca <2 x i64>, align 16
64  %b = alloca <2 x i64>, align 16
65  store <2 x i64> <i64 72623859790382856, i64 651345242494996240>, <2 x i64>* %a, align 16
66  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %b, align 16
67  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
68  %1 = load <2 x i64>,  <2 x i64>* %b, align 16
69  %2 = call <2 x i64> @llvm.ppc.altivec.crypto.vpmsumd(<2 x i64> %0, <2 x i64> %1)
70  ret <2 x i64> %2
71; CHECK: vpmsumd 2,
72}
73
74; Function Attrs: nounwind readnone
75declare <2 x i64> @llvm.ppc.altivec.crypto.vpmsumd(<2 x i64>, <2 x i64>) #1
76
77; Function Attrs: nounwind
78define <2 x i64> @test_vsbox() #0 {
79entry:
80  %a = alloca <2 x i64>, align 16
81  store <2 x i64> <i64 72623859790382856, i64 651345242494996240>, <2 x i64>* %a, align 16
82  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
83  %1 = call <2 x i64> @llvm.ppc.altivec.crypto.vsbox(<2 x i64> %0)
84  ret <2 x i64> %1
85; CHECK: vsbox 2,
86}
87
88; Function Attrs: nounwind readnone
89declare <2 x i64> @llvm.ppc.altivec.crypto.vsbox(<2 x i64>) #1
90
91; Function Attrs: nounwind
92define <16 x i8> @test_vpermxorb() #0 {
93entry:
94  %a = alloca <16 x i8>, align 16
95  %b = alloca <16 x i8>, align 16
96  %c = alloca <16 x i8>, align 16
97  store <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16>, <16 x i8>* %a, align 16
98  store <16 x i8> <i8 113, i8 114, i8 115, i8 116, i8 117, i8 118, i8 119, i8 120, i8 121, i8 122, i8 123, i8 124, i8 125, i8 126, i8 127, i8 112>, <16 x i8>* %b, align 16
99  store <16 x i8> <i8 113, i8 114, i8 115, i8 116, i8 117, i8 118, i8 119, i8 120, i8 121, i8 122, i8 123, i8 124, i8 125, i8 126, i8 127, i8 112>, <16 x i8>* %c, align 16
100  %0 = load <16 x i8>,  <16 x i8>* %a, align 16
101  %1 = load <16 x i8>,  <16 x i8>* %b, align 16
102  %2 = load <16 x i8>,  <16 x i8>* %c, align 16
103  %3 = call <16 x i8> @llvm.ppc.altivec.crypto.vpermxor(<16 x i8> %0, <16 x i8> %1, <16 x i8> %2)
104  ret <16 x i8> %3
105; CHECK: vpermxor 2,
106}
107
108; Function Attrs: nounwind readnone
109declare <16 x i8> @llvm.ppc.altivec.crypto.vpermxor(<16 x i8>, <16 x i8>, <16 x i8>) #1
110
111; Function Attrs: nounwind
112define <8 x i16> @test_vpermxorh() #0 {
113entry:
114  %a = alloca <8 x i16>, align 16
115  %b = alloca <8 x i16>, align 16
116  %c = alloca <8 x i16>, align 16
117  store <8 x i16> <i16 258, i16 772, i16 1286, i16 1800, i16 2314, i16 2828, i16 3342, i16 3856>, <8 x i16>* %a, align 16
118  store <8 x i16> <i16 29042, i16 29556, i16 30070, i16 30584, i16 31098, i16 31612, i16 32126, i16 32624>, <8 x i16>* %b, align 16
119  store <8 x i16> <i16 29042, i16 29556, i16 30070, i16 30584, i16 31098, i16 31612, i16 32126, i16 32624>, <8 x i16>* %c, align 16
120  %0 = load <8 x i16>,  <8 x i16>* %a, align 16
121  %1 = bitcast <8 x i16> %0 to <16 x i8>
122  %2 = load <8 x i16>,  <8 x i16>* %b, align 16
123  %3 = bitcast <8 x i16> %2 to <16 x i8>
124  %4 = load <8 x i16>,  <8 x i16>* %c, align 16
125  %5 = bitcast <8 x i16> %4 to <16 x i8>
126  %6 = call <16 x i8> @llvm.ppc.altivec.crypto.vpermxor(<16 x i8> %1, <16 x i8> %3, <16 x i8> %5)
127  %7 = bitcast <16 x i8> %6 to <8 x i16>
128  ret <8 x i16> %7
129; CHECK: vpermxor 2,
130}
131
132; Function Attrs: nounwind
133define <4 x i32> @test_vpermxorw() #0 {
134entry:
135  %a = alloca <4 x i32>, align 16
136  %b = alloca <4 x i32>, align 16
137  %c = alloca <4 x i32>, align 16
138  store <4 x i32> <i32 16909060, i32 84281096, i32 151653132, i32 219025168>, <4 x i32>* %a, align 16
139  store <4 x i32> <i32 1903326068, i32 1970698104, i32 2038070140, i32 2105442160>, <4 x i32>* %b, align 16
140  store <4 x i32> <i32 1903326068, i32 1970698104, i32 2038070140, i32 2105442160>, <4 x i32>* %c, align 16
141  %0 = load <4 x i32>,  <4 x i32>* %a, align 16
142  %1 = bitcast <4 x i32> %0 to <16 x i8>
143  %2 = load <4 x i32>,  <4 x i32>* %b, align 16
144  %3 = bitcast <4 x i32> %2 to <16 x i8>
145  %4 = load <4 x i32>,  <4 x i32>* %c, align 16
146  %5 = bitcast <4 x i32> %4 to <16 x i8>
147  %6 = call <16 x i8> @llvm.ppc.altivec.crypto.vpermxor(<16 x i8> %1, <16 x i8> %3, <16 x i8> %5)
148  %7 = bitcast <16 x i8> %6 to <4 x i32>
149  ret <4 x i32> %7
150; CHECK: vpermxor 2,
151}
152
153; Function Attrs: nounwind
154define <2 x i64> @test_vpermxord() #0 {
155entry:
156  %a = alloca <2 x i64>, align 16
157  %b = alloca <2 x i64>, align 16
158  %c = alloca <2 x i64>, align 16
159  store <2 x i64> <i64 72623859790382856, i64 651345242494996240>, <2 x i64>* %a, align 16
160  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %b, align 16
161  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %c, align 16
162  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
163  %1 = bitcast <2 x i64> %0 to <16 x i8>
164  %2 = load <2 x i64>,  <2 x i64>* %b, align 16
165  %3 = bitcast <2 x i64> %2 to <16 x i8>
166  %4 = load <2 x i64>,  <2 x i64>* %c, align 16
167  %5 = bitcast <2 x i64> %4 to <16 x i8>
168  %6 = call <16 x i8> @llvm.ppc.altivec.crypto.vpermxor(<16 x i8> %1, <16 x i8> %3, <16 x i8> %5)
169  %7 = bitcast <16 x i8> %6 to <2 x i64>
170  ret <2 x i64> %7
171; CHECK: vpermxor 2,
172}
173
174; Function Attrs: nounwind
175define <2 x i64> @test_vcipher() #0 {
176entry:
177  %a = alloca <2 x i64>, align 16
178  %b = alloca <2 x i64>, align 16
179  store <2 x i64> <i64 72623859790382856, i64 651345242494996240>, <2 x i64>* %a, align 16
180  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %b, align 16
181  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
182  %1 = load <2 x i64>,  <2 x i64>* %b, align 16
183  %2 = call <2 x i64> @llvm.ppc.altivec.crypto.vcipher(<2 x i64> %0, <2 x i64> %1)
184  ret <2 x i64> %2
185; CHECK: vcipher 2,
186}
187
188; Function Attrs: nounwind readnone
189declare <2 x i64> @llvm.ppc.altivec.crypto.vcipher(<2 x i64>, <2 x i64>) #1
190
191; Function Attrs: nounwind
192define <2 x i64> @test_vcipherlast() #0 {
193entry:
194  %a = alloca <2 x i64>, align 16
195  %b = alloca <2 x i64>, align 16
196  store <2 x i64> <i64 72623859790382856, i64 651345242494996240>, <2 x i64>* %a, align 16
197  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %b, align 16
198  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
199  %1 = load <2 x i64>,  <2 x i64>* %b, align 16
200  %2 = call <2 x i64> @llvm.ppc.altivec.crypto.vcipherlast(<2 x i64> %0, <2 x i64> %1)
201  ret <2 x i64> %2
202; CHECK: vcipherlast 2,
203}
204
205; Function Attrs: nounwind readnone
206declare <2 x i64> @llvm.ppc.altivec.crypto.vcipherlast(<2 x i64>, <2 x i64>) #1
207
208; Function Attrs: nounwind
209define <2 x i64> @test_vncipher() #0 {
210entry:
211  %a = alloca <2 x i64>, align 16
212  %b = alloca <2 x i64>, align 16
213  store <2 x i64> <i64 72623859790382856, i64 651345242494996240>, <2 x i64>* %a, align 16
214  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %b, align 16
215  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
216  %1 = load <2 x i64>,  <2 x i64>* %b, align 16
217  %2 = call <2 x i64> @llvm.ppc.altivec.crypto.vncipher(<2 x i64> %0, <2 x i64> %1)
218  ret <2 x i64> %2
219; CHECK: vncipher 2,
220}
221
222; Function Attrs: nounwind readnone
223declare <2 x i64> @llvm.ppc.altivec.crypto.vncipher(<2 x i64>, <2 x i64>) #1
224
225; Function Attrs: nounwind
226define <2 x i64> @test_vncipherlast() #0 {
227entry:
228  %a = alloca <2 x i64>, align 16
229  %b = alloca <2 x i64>, align 16
230  store <2 x i64> <i64 72623859790382856, i64 651345242494996240>, <2 x i64>* %a, align 16
231  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %b, align 16
232  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
233  %1 = load <2 x i64>,  <2 x i64>* %b, align 16
234  %2 = call <2 x i64> @llvm.ppc.altivec.crypto.vncipherlast(<2 x i64> %0, <2 x i64> %1)
235  ret <2 x i64> %2
236; CHECK: vncipherlast 2,
237}
238
239; Function Attrs: nounwind readnone
240declare <2 x i64> @llvm.ppc.altivec.crypto.vncipherlast(<2 x i64>, <2 x i64>) #1
241
242; Function Attrs: nounwind
243define <4 x i32> @test_vshasigmaw() #0 {
244entry:
245  %a = alloca <4 x i32>, align 16
246  store <4 x i32> <i32 16909060, i32 84281096, i32 151653132, i32 219025168>, <4 x i32>* %a, align 16
247  %0 = load <4 x i32>,  <4 x i32>* %a, align 16
248  %1 = call <4 x i32> @llvm.ppc.altivec.crypto.vshasigmaw(<4 x i32> %0, i32 1, i32 15)
249  ret <4 x i32> %1
250; CHECK: vshasigmaw 2,
251}
252
253; Function Attrs: nounwind readnone
254declare <4 x i32> @llvm.ppc.altivec.crypto.vshasigmaw(<4 x i32>, i32, i32) #1
255
256; Function Attrs: nounwind
257define <2 x i64> @test_vshasigmad() #0 {
258entry:
259  %a = alloca <2 x i64>, align 16
260  store <2 x i64> <i64 8174723217654970232, i64 8753444600359583600>, <2 x i64>* %a, align 16
261  %0 = load <2 x i64>,  <2 x i64>* %a, align 16
262  %1 = call <2 x i64> @llvm.ppc.altivec.crypto.vshasigmad(<2 x i64> %0, i32 1, i32 15)
263  ret <2 x i64> %1
264; CHECK: vshasigmad 2,
265}
266
267; Function Attrs: nounwind readnone
268declare <2 x i64> @llvm.ppc.altivec.crypto.vshasigmad(<2 x i64>, i32, i32) #1
269
270attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-realign-stack" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
271attributes #1 = { nounwind readnone }
272
273!llvm.ident = !{!0}
274
275!0 = !{!"clang version 3.7.0 (trunk 230949) (llvm/trunk 230946)"}
276