Skip to content

Commit 84b1489

Browse files
davemgreendavemgreen
committed
[ARM] Update VBIC tests. NFC
1 parent 8a6248b commit 84b1489

File tree

1 file changed

+49
-127
lines changed

1 file changed

+49
-127
lines changed

‎llvm/test/CodeGen/ARM/vbsl.ll

Lines changed: 49 additions & 127 deletions
Original file line numberDiff line numberDiff line change
@@ -1,17 +1,15 @@
11
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2-
; RUN: llc -mtriple=arm-eabi -mattr=+neon %s -o - | FileCheck %s
3-
4-
; rdar://12471808
2+
; RUN: llc -mtriple=armv7-eabihf -mattr=+neon %s -o - | FileCheck %s
53

64
define <8 x i8> @v_bsli8(ptr %A, ptr %B, ptr %C) nounwind {
75
; CHECK-LABEL: v_bsli8:
86
; CHECK: @ %bb.0:
97
; CHECK-NEXT: vldr d18, [r0]
108
; CHECK-NEXT: vldr d16, [r2]
9+
; CHECK-NEXT: vorr d0, d18, d18
1110
; CHECK-NEXT: vldr d17, [r1]
12-
; CHECK-NEXT: vbit d16, d17, d18
13-
; CHECK-NEXT: vmov r0, r1, d16
14-
; CHECK-NEXT: mov pc, lr
11+
; CHECK-NEXT: vbsl d0, d17, d16
12+
; CHECK-NEXT: bx lr
1513
%tmp1 = load <8 x i8>, ptr %A
1614
%tmp2 = load <8 x i8>, ptr %B
1715
%tmp3 = load <8 x i8>, ptr %C
@@ -27,10 +25,10 @@ define <4 x i16> @v_bsli16(ptr %A, ptr %B, ptr %C) nounwind {
2725
; CHECK: @ %bb.0:
2826
; CHECK-NEXT: vldr d18, [r0]
2927
; CHECK-NEXT: vldr d16, [r2]
28+
; CHECK-NEXT: vorr d0, d18, d18
3029
; CHECK-NEXT: vldr d17, [r1]
31-
; CHECK-NEXT: vbit d16, d17, d18
32-
; CHECK-NEXT: vmov r0, r1, d16
33-
; CHECK-NEXT: mov pc, lr
30+
; CHECK-NEXT: vbsl d0, d17, d16
31+
; CHECK-NEXT: bx lr
3432
%tmp1 = load <4 x i16>, ptr %A
3533
%tmp2 = load <4 x i16>, ptr %B
3634
%tmp3 = load <4 x i16>, ptr %C
@@ -46,10 +44,10 @@ define <2 x i32> @v_bsli32(ptr %A, ptr %B, ptr %C) nounwind {
4644
; CHECK: @ %bb.0:
4745
; CHECK-NEXT: vldr d18, [r0]
4846
; CHECK-NEXT: vldr d16, [r2]
47+
; CHECK-NEXT: vorr d0, d18, d18
4948
; CHECK-NEXT: vldr d17, [r1]
50-
; CHECK-NEXT: vbit d16, d17, d18
51-
; CHECK-NEXT: vmov r0, r1, d16
52-
; CHECK-NEXT: mov pc, lr
49+
; CHECK-NEXT: vbsl d0, d17, d16
50+
; CHECK-NEXT: bx lr
5351
%tmp1 = load <2 x i32>, ptr %A
5452
%tmp2 = load <2 x i32>, ptr %B
5553
%tmp3 = load <2 x i32>, ptr %C
@@ -65,10 +63,10 @@ define <1 x i64> @v_bsli64(ptr %A, ptr %B, ptr %C) nounwind {
6563
; CHECK: @ %bb.0:
6664
; CHECK-NEXT: vldr d18, [r0]
6765
; CHECK-NEXT: vldr d16, [r2]
66+
; CHECK-NEXT: vorr d0, d18, d18
6867
; CHECK-NEXT: vldr d17, [r1]
69-
; CHECK-NEXT: vbit d16, d17, d18
70-
; CHECK-NEXT: vmov r0, r1, d16
71-
; CHECK-NEXT: mov pc, lr
68+
; CHECK-NEXT: vbsl d0, d17, d16
69+
; CHECK-NEXT: bx lr
7270
%tmp1 = load <1 x i64>, ptr %A
7371
%tmp2 = load <1 x i64>, ptr %B
7472
%tmp3 = load <1 x i64>, ptr %C
@@ -83,12 +81,11 @@ define <16 x i8> @v_bslQi8(ptr %A, ptr %B, ptr %C) nounwind {
8381
; CHECK-LABEL: v_bslQi8:
8482
; CHECK: @ %bb.0:
8583
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
84+
; CHECK-NEXT: vorr q0, q10, q10
8685
; CHECK-NEXT: vld1.64 {d16, d17}, [r2]
8786
; CHECK-NEXT: vld1.64 {d18, d19}, [r1]
88-
; CHECK-NEXT: vbit q8, q9, q10
89-
; CHECK-NEXT: vmov r0, r1, d16
90-
; CHECK-NEXT: vmov r2, r3, d17
91-
; CHECK-NEXT: mov pc, lr
87+
; CHECK-NEXT: vbsl q0, q9, q8
88+
; CHECK-NEXT: bx lr
9289
%tmp1 = load <16 x i8>, ptr %A
9390
%tmp2 = load <16 x i8>, ptr %B
9491
%tmp3 = load <16 x i8>, ptr %C
@@ -103,12 +100,11 @@ define <8 x i16> @v_bslQi16(ptr %A, ptr %B, ptr %C) nounwind {
103100
; CHECK-LABEL: v_bslQi16:
104101
; CHECK: @ %bb.0:
105102
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
103+
; CHECK-NEXT: vorr q0, q10, q10
106104
; CHECK-NEXT: vld1.64 {d16, d17}, [r2]
107105
; CHECK-NEXT: vld1.64 {d18, d19}, [r1]
108-
; CHECK-NEXT: vbit q8, q9, q10
109-
; CHECK-NEXT: vmov r0, r1, d16
110-
; CHECK-NEXT: vmov r2, r3, d17
111-
; CHECK-NEXT: mov pc, lr
106+
; CHECK-NEXT: vbsl q0, q9, q8
107+
; CHECK-NEXT: bx lr
112108
%tmp1 = load <8 x i16>, ptr %A
113109
%tmp2 = load <8 x i16>, ptr %B
114110
%tmp3 = load <8 x i16>, ptr %C
@@ -123,12 +119,11 @@ define <4 x i32> @v_bslQi32(ptr %A, ptr %B, ptr %C) nounwind {
123119
; CHECK-LABEL: v_bslQi32:
124120
; CHECK: @ %bb.0:
125121
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
122+
; CHECK-NEXT: vorr q0, q10, q10
126123
; CHECK-NEXT: vld1.64 {d16, d17}, [r2]
127124
; CHECK-NEXT: vld1.64 {d18, d19}, [r1]
128-
; CHECK-NEXT: vbit q8, q9, q10
129-
; CHECK-NEXT: vmov r0, r1, d16
130-
; CHECK-NEXT: vmov r2, r3, d17
131-
; CHECK-NEXT: mov pc, lr
125+
; CHECK-NEXT: vbsl q0, q9, q8
126+
; CHECK-NEXT: bx lr
132127
%tmp1 = load <4 x i32>, ptr %A
133128
%tmp2 = load <4 x i32>, ptr %B
134129
%tmp3 = load <4 x i32>, ptr %C
@@ -143,12 +138,11 @@ define <2 x i64> @v_bslQi64(ptr %A, ptr %B, ptr %C) nounwind {
143138
; CHECK-LABEL: v_bslQi64:
144139
; CHECK: @ %bb.0:
145140
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
141+
; CHECK-NEXT: vorr q0, q10, q10
146142
; CHECK-NEXT: vld1.64 {d16, d17}, [r2]
147143
; CHECK-NEXT: vld1.64 {d18, d19}, [r1]
148-
; CHECK-NEXT: vbit q8, q9, q10
149-
; CHECK-NEXT: vmov r0, r1, d16
150-
; CHECK-NEXT: vmov r2, r3, d17
151-
; CHECK-NEXT: mov pc, lr
144+
; CHECK-NEXT: vbsl q0, q9, q8
145+
; CHECK-NEXT: bx lr
152146
%tmp1 = load <2 x i64>, ptr %A
153147
%tmp2 = load <2 x i64>, ptr %B
154148
%tmp3 = load <2 x i64>, ptr %C
@@ -162,179 +156,107 @@ define <2 x i64> @v_bslQi64(ptr %A, ptr %B, ptr %C) nounwind {
162156
define <8 x i8> @f1(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c) nounwind readnone optsize ssp {
163157
; CHECK-LABEL: f1:
164158
; CHECK: @ %bb.0:
165-
; CHECK-NEXT: vldr d16, [sp]
166-
; CHECK-NEXT: vmov d17, r2, r3
167-
; CHECK-NEXT: vmov d18, r0, r1
168-
; CHECK-NEXT: vbit d16, d17, d18
169-
; CHECK-NEXT: vmov r0, r1, d16
170-
; CHECK-NEXT: mov pc, lr
159+
; CHECK-NEXT: vbsl d0, d1, d2
160+
; CHECK-NEXT: bx lr
171161
%vbsl.i = tail call <8 x i8> @llvm.arm.neon.vbsl.v8i8(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c) nounwind
172162
ret <8 x i8> %vbsl.i
173163
}
174164

175165
define <4 x i16> @f2(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) nounwind readnone optsize ssp {
176166
; CHECK-LABEL: f2:
177167
; CHECK: @ %bb.0:
178-
; CHECK-NEXT: vldr d16, [sp]
179-
; CHECK-NEXT: vmov d17, r2, r3
180-
; CHECK-NEXT: vmov d18, r0, r1
181-
; CHECK-NEXT: vbit d16, d17, d18
182-
; CHECK-NEXT: vmov r0, r1, d16
183-
; CHECK-NEXT: mov pc, lr
168+
; CHECK-NEXT: vbsl d0, d1, d2
169+
; CHECK-NEXT: bx lr
184170
%vbsl3.i = tail call <4 x i16> @llvm.arm.neon.vbsl.v4i16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) nounwind
185171
ret <4 x i16> %vbsl3.i
186172
}
187173

188174
define <2 x i32> @f3(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) nounwind readnone optsize ssp {
189175
; CHECK-LABEL: f3:
190176
; CHECK: @ %bb.0:
191-
; CHECK-NEXT: vldr d16, [sp]
192-
; CHECK-NEXT: vmov d17, r2, r3
193-
; CHECK-NEXT: vmov d18, r0, r1
194-
; CHECK-NEXT: vbit d16, d17, d18
195-
; CHECK-NEXT: vmov r0, r1, d16
196-
; CHECK-NEXT: mov pc, lr
177+
; CHECK-NEXT: vbsl d0, d1, d2
178+
; CHECK-NEXT: bx lr
197179
%vbsl3.i = tail call <2 x i32> @llvm.arm.neon.vbsl.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) nounwind
198180
ret <2 x i32> %vbsl3.i
199181
}
200182

201183
define <2 x float> @f4(<2 x float> %a, <2 x float> %b, <2 x float> %c) nounwind readnone optsize ssp {
202184
; CHECK-LABEL: f4:
203185
; CHECK: @ %bb.0:
204-
; CHECK-NEXT: vldr d16, [sp]
205-
; CHECK-NEXT: vmov d17, r2, r3
206-
; CHECK-NEXT: vmov d18, r0, r1
207-
; CHECK-NEXT: vbit d16, d17, d18
208-
; CHECK-NEXT: vmov r0, r1, d16
209-
; CHECK-NEXT: mov pc, lr
186+
; CHECK-NEXT: vbsl d0, d1, d2
187+
; CHECK-NEXT: bx lr
210188
%vbsl4.i = tail call <2 x float> @llvm.arm.neon.vbsl.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) nounwind
211189
ret <2 x float> %vbsl4.i
212190
}
213191

214192
define <16 x i8> @g1(<16 x i8> %a, <16 x i8> %b, <16 x i8> %c) nounwind readnone optsize ssp {
215193
; CHECK-LABEL: g1:
216194
; CHECK: @ %bb.0:
217-
; CHECK-NEXT: vmov d19, r2, r3
218-
; CHECK-NEXT: add r12, sp, #16
219-
; CHECK-NEXT: vmov d18, r0, r1
220-
; CHECK-NEXT: mov r0, sp
221-
; CHECK-NEXT: vld1.64 {d16, d17}, [r12]
222-
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
223-
; CHECK-NEXT: vbit q8, q10, q9
224-
; CHECK-NEXT: vmov r0, r1, d16
225-
; CHECK-NEXT: vmov r2, r3, d17
226-
; CHECK-NEXT: mov pc, lr
195+
; CHECK-NEXT: vbsl q0, q1, q2
196+
; CHECK-NEXT: bx lr
227197
%vbsl.i = tail call <16 x i8> @llvm.arm.neon.vbsl.v16i8(<16 x i8> %a, <16 x i8> %b, <16 x i8> %c) nounwind
228198
ret <16 x i8> %vbsl.i
229199
}
230200

231201
define <8 x i16> @g2(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) nounwind readnone optsize ssp {
232202
; CHECK-LABEL: g2:
233203
; CHECK: @ %bb.0:
234-
; CHECK-NEXT: vmov d19, r2, r3
235-
; CHECK-NEXT: add r12, sp, #16
236-
; CHECK-NEXT: vmov d18, r0, r1
237-
; CHECK-NEXT: mov r0, sp
238-
; CHECK-NEXT: vld1.64 {d16, d17}, [r12]
239-
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
240-
; CHECK-NEXT: vbit q8, q10, q9
241-
; CHECK-NEXT: vmov r0, r1, d16
242-
; CHECK-NEXT: vmov r2, r3, d17
243-
; CHECK-NEXT: mov pc, lr
204+
; CHECK-NEXT: vbsl q0, q1, q2
205+
; CHECK-NEXT: bx lr
244206
%vbsl3.i = tail call <8 x i16> @llvm.arm.neon.vbsl.v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) nounwind
245207
ret <8 x i16> %vbsl3.i
246208
}
247209

248210
define <4 x i32> @g3(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) nounwind readnone optsize ssp {
249211
; CHECK-LABEL: g3:
250212
; CHECK: @ %bb.0:
251-
; CHECK-NEXT: vmov d19, r2, r3
252-
; CHECK-NEXT: add r12, sp, #16
253-
; CHECK-NEXT: vmov d18, r0, r1
254-
; CHECK-NEXT: mov r0, sp
255-
; CHECK-NEXT: vld1.64 {d16, d17}, [r12]
256-
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
257-
; CHECK-NEXT: vbit q8, q10, q9
258-
; CHECK-NEXT: vmov r0, r1, d16
259-
; CHECK-NEXT: vmov r2, r3, d17
260-
; CHECK-NEXT: mov pc, lr
213+
; CHECK-NEXT: vbsl q0, q1, q2
214+
; CHECK-NEXT: bx lr
261215
%vbsl3.i = tail call <4 x i32> @llvm.arm.neon.vbsl.v4i32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) nounwind
262216
ret <4 x i32> %vbsl3.i
263217
}
264218

265219
define <4 x float> @g4(<4 x float> %a, <4 x float> %b, <4 x float> %c) nounwind readnone optsize ssp {
266220
; CHECK-LABEL: g4:
267221
; CHECK: @ %bb.0:
268-
; CHECK-NEXT: vmov d19, r2, r3
269-
; CHECK-NEXT: add r12, sp, #16
270-
; CHECK-NEXT: vmov d18, r0, r1
271-
; CHECK-NEXT: mov r0, sp
272-
; CHECK-NEXT: vld1.64 {d16, d17}, [r12]
273-
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
274-
; CHECK-NEXT: vbit q8, q10, q9
275-
; CHECK-NEXT: vmov r0, r1, d16
276-
; CHECK-NEXT: vmov r2, r3, d17
277-
; CHECK-NEXT: mov pc, lr
222+
; CHECK-NEXT: vbsl q0, q1, q2
223+
; CHECK-NEXT: bx lr
278224
%vbsl4.i = tail call <4 x float> @llvm.arm.neon.vbsl.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) nounwind
279225
ret <4 x float> %vbsl4.i
280226
}
281227

282228
define <1 x i64> @test_vbsl_s64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind readnone optsize ssp {
283229
; CHECK-LABEL: test_vbsl_s64:
284230
; CHECK: @ %bb.0:
285-
; CHECK-NEXT: vldr d16, [sp]
286-
; CHECK-NEXT: vmov d17, r2, r3
287-
; CHECK-NEXT: vmov d18, r0, r1
288-
; CHECK-NEXT: vbit d16, d17, d18
289-
; CHECK-NEXT: vmov r0, r1, d16
290-
; CHECK-NEXT: mov pc, lr
231+
; CHECK-NEXT: vbsl d0, d1, d2
232+
; CHECK-NEXT: bx lr
291233
%vbsl3.i = tail call <1 x i64> @llvm.arm.neon.vbsl.v1i64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind
292234
ret <1 x i64> %vbsl3.i
293235
}
294236

295237
define <1 x i64> @test_vbsl_u64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind readnone optsize ssp {
296238
; CHECK-LABEL: test_vbsl_u64:
297239
; CHECK: @ %bb.0:
298-
; CHECK-NEXT: vldr d16, [sp]
299-
; CHECK-NEXT: vmov d17, r2, r3
300-
; CHECK-NEXT: vmov d18, r0, r1
301-
; CHECK-NEXT: vbit d16, d17, d18
302-
; CHECK-NEXT: vmov r0, r1, d16
303-
; CHECK-NEXT: mov pc, lr
240+
; CHECK-NEXT: vbsl d0, d1, d2
241+
; CHECK-NEXT: bx lr
304242
%vbsl3.i = tail call <1 x i64> @llvm.arm.neon.vbsl.v1i64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind
305243
ret <1 x i64> %vbsl3.i
306244
}
307245

308246
define <2 x i64> @test_vbslq_s64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
309247
; CHECK-LABEL: test_vbslq_s64:
310248
; CHECK: @ %bb.0:
311-
; CHECK-NEXT: vmov d19, r2, r3
312-
; CHECK-NEXT: add r12, sp, #16
313-
; CHECK-NEXT: vmov d18, r0, r1
314-
; CHECK-NEXT: mov r0, sp
315-
; CHECK-NEXT: vld1.64 {d16, d17}, [r12]
316-
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
317-
; CHECK-NEXT: vbit q8, q10, q9
318-
; CHECK-NEXT: vmov r0, r1, d16
319-
; CHECK-NEXT: vmov r2, r3, d17
320-
; CHECK-NEXT: mov pc, lr
249+
; CHECK-NEXT: vbsl q0, q1, q2
250+
; CHECK-NEXT: bx lr
321251
%vbsl3.i = tail call <2 x i64> @llvm.arm.neon.vbsl.v2i64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind
322252
ret <2 x i64> %vbsl3.i
323253
}
324254

325255
define <2 x i64> @test_vbslq_u64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
326256
; CHECK-LABEL: test_vbslq_u64:
327257
; CHECK: @ %bb.0:
328-
; CHECK-NEXT: vmov d19, r2, r3
329-
; CHECK-NEXT: add r12, sp, #16
330-
; CHECK-NEXT: vmov d18, r0, r1
331-
; CHECK-NEXT: mov r0, sp
332-
; CHECK-NEXT: vld1.64 {d16, d17}, [r12]
333-
; CHECK-NEXT: vld1.64 {d20, d21}, [r0]
334-
; CHECK-NEXT: vbit q8, q10, q9
335-
; CHECK-NEXT: vmov r0, r1, d16
336-
; CHECK-NEXT: vmov r2, r3, d17
337-
; CHECK-NEXT: mov pc, lr
258+
; CHECK-NEXT: vbsl q0, q1, q2
259+
; CHECK-NEXT: bx lr
338260
%vbsl3.i = tail call <2 x i64> @llvm.arm.neon.vbsl.v2i64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind
339261
ret <2 x i64> %vbsl3.i
340262
}

0 commit comments

Comments
 (0)