1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve --verify-machineinstrs %s -o - | FileCheck %s
; Tail predicated so we use DLSTP
define void @simple(i32* nocapture readonly %x, i32* nocapture readnone %y, i32* nocapture %z, i32 %m, i32 %n) {
; CHECK-LABEL: simple:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r7, lr}
; CHECK-NEXT: push {r7, lr}
; CHECK-NEXT: ldr r1, [sp, #8]
; CHECK-NEXT: mov r12, r3
; CHECK-NEXT: adds r3, r1, #3
; CHECK-NEXT: lsrs r3, r3, #2
; CHECK-NEXT: beq .LBB0_3
; CHECK-NEXT: @ %bb.1: @ %do.body.preheader
; CHECK-NEXT: dlstp.32 lr, r1
; CHECK-NEXT: .LBB0_2: @ %do.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: vldrw.u32 q0, [r0], #16
; CHECK-NEXT: vaddva.s32 r12, q0
; CHECK-NEXT: letp lr, .LBB0_2
; CHECK-NEXT: .LBB0_3: @ %if.end
; CHECK-NEXT: str.w r12, [r2]
; CHECK-NEXT: pop {r7, pc}
entry:
%add = add i32 %n, 3
%div = lshr i32 %add, 2
%cmp.not = icmp eq i32 %div, 0
br i1 %cmp.not, label %if.end, label %do.body
do.body: ; preds = %entry, %do.body
%n.addr.0 = phi i32 [ %sub, %do.body ], [ %n, %entry ]
%count.0 = phi i32 [ %sub3, %do.body ], [ %div, %entry ]
%s.0 = phi i32 [ %add2, %do.body ], [ %m, %entry ]
%x.addr.0 = phi i32* [ %add.ptr, %do.body ], [ %x, %entry ]
%0 = tail call <4 x i1> @llvm.arm.mve.vctp32(i32 %n.addr.0)
%1 = bitcast i32* %x.addr.0 to <4 x i32>*
%2 = load <4 x i32>, <4 x i32>* %1, align 4
%3 = tail call i32 @llvm.arm.mve.addv.predicated.v4i32.v4i1(<4 x i32> %2, i32 0, <4 x i1> %0)
%add2 = add nsw i32 %3, %s.0
%add.ptr = getelementptr inbounds i32, i32* %x.addr.0, i32 4
%sub = add i32 %n.addr.0, -4
%sub3 = add nsw i32 %count.0, -1
%cmp4 = icmp sgt i32 %count.0, 1
br i1 %cmp4, label %do.body, label %if.end
if.end: ; preds = %do.body, %entry
%s.1 = phi i32 [ %m, %entry ], [ %add2, %do.body ]
store i32 %s.1, i32* %z, align 4
ret void
}
; Tail predicated so we use DLSTP
define void @nested(i32* nocapture readonly %x, i32* nocapture readnone %y, i32* nocapture %z, i32 %m, i32 %n) {
; CHECK-LABEL: nested:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT: push.w {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT: cbz r3, .LBB1_8
; CHECK-NEXT: @ %bb.1: @ %for.body.preheader
; CHECK-NEXT: ldr.w r12, [sp, #24]
; CHECK-NEXT: movs r1, #0
; CHECK-NEXT: b .LBB1_4
; CHECK-NEXT: .LBB1_2: @ in Loop: Header=BB1_4 Depth=1
; CHECK-NEXT: mov r4, r3
; CHECK-NEXT: .LBB1_3: @ %if.end
; CHECK-NEXT: @ in Loop: Header=BB1_4 Depth=1
; CHECK-NEXT: str.w r4, [r2, r1, lsl #2]
; CHECK-NEXT: adds r1, #1
; CHECK-NEXT: cmp r1, r3
; CHECK-NEXT: beq .LBB1_8
; CHECK-NEXT: .LBB1_4: @ %for.body
; CHECK-NEXT: @ =>This Loop Header: Depth=1
; CHECK-NEXT: @ Child Loop BB1_6 Depth 2
; CHECK-NEXT: add.w r6, r12, #3
; CHECK-NEXT: lsrs r7, r6, #2
; CHECK-NEXT: beq .LBB1_2
; CHECK-NEXT: @ %bb.5: @ %do.body.preheader
; CHECK-NEXT: @ in Loop: Header=BB1_4 Depth=1
; CHECK-NEXT: bic r5, r6, #3
; CHECK-NEXT: mov r4, r3
; CHECK-NEXT: add.w r8, r0, r5, lsl #2
; CHECK-NEXT: dlstp.32 lr, r12
; CHECK-NEXT: .LBB1_6: @ %do.body
; CHECK-NEXT: @ Parent Loop BB1_4 Depth=1
; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
; CHECK-NEXT: vldrw.u32 q0, [r0], #16
; CHECK-NEXT: vaddva.s32 r4, q0
; CHECK-NEXT: letp lr, .LBB1_6
; CHECK-NEXT: @ %bb.7: @ %if.end.loopexit
; CHECK-NEXT: @ in Loop: Header=BB1_4 Depth=1
; CHECK-NEXT: sub.w r12, r12, r5
; CHECK-NEXT: mov r0, r8
; CHECK-NEXT: b .LBB1_3
; CHECK-NEXT: .LBB1_8: @ %for.cond.cleanup
; CHECK-NEXT: pop.w {r4, r5, r6, r7, r8, pc}
entry:
%cmp20.not = icmp eq i32 %m, 0
br i1 %cmp20.not, label %for.cond.cleanup, label %for.body
for.cond.cleanup: ; preds = %if.end, %entry
ret void
for.body: ; preds = %entry, %if.end
%x.addr.023 = phi i32* [ %x.addr.2, %if.end ], [ %x, %entry ]
%a.022 = phi i32 [ %inc, %if.end ], [ 0, %entry ]
%n.addr.021 = phi i32 [ %n.addr.2, %if.end ], [ %n, %entry ]
%add = add i32 %n.addr.021, 3
%div = lshr i32 %add, 2
%cmp1.not = icmp eq i32 %div, 0
br i1 %cmp1.not, label %if.end, label %do.body.preheader
do.body.preheader: ; preds = %for.body
%0 = and i32 %add, -4
%scevgep = getelementptr i32, i32* %x.addr.023, i32 %0
br label %do.body
do.body: ; preds = %do.body.preheader, %do.body
%n.addr.1 = phi i32 [ %sub, %do.body ], [ %n.addr.021, %do.body.preheader ]
%count.0 = phi i32 [ %sub4, %do.body ], [ %div, %do.body.preheader ]
%s.0 = phi i32 [ %add3, %do.body ], [ %m, %do.body.preheader ]
%x.addr.1 = phi i32* [ %add.ptr, %do.body ], [ %x.addr.023, %do.body.preheader ]
%1 = tail call <4 x i1> @llvm.arm.mve.vctp32(i32 %n.addr.1)
%2 = bitcast i32* %x.addr.1 to <4 x i32>*
%3 = load <4 x i32>, <4 x i32>* %2, align 4
%4 = tail call i32 @llvm.arm.mve.addv.predicated.v4i32.v4i1(<4 x i32> %3, i32 0, <4 x i1> %1)
%add3 = add nsw i32 %4, %s.0
%add.ptr = getelementptr inbounds i32, i32* %x.addr.1, i32 4
%sub = add i32 %n.addr.1, -4
%sub4 = add nsw i32 %count.0, -1
%cmp5 = icmp sgt i32 %count.0, 1
br i1 %cmp5, label %do.body, label %if.end.loopexit
if.end.loopexit: ; preds = %do.body
%5 = sub i32 %n.addr.021, %0
br label %if.end
if.end: ; preds = %if.end.loopexit, %for.body
%n.addr.2 = phi i32 [ %n.addr.021, %for.body ], [ %5, %if.end.loopexit ]
%s.1 = phi i32 [ %m, %for.body ], [ %add3, %if.end.loopexit ]
%x.addr.2 = phi i32* [ %x.addr.023, %for.body ], [ %scevgep, %if.end.loopexit ]
%arrayidx = getelementptr inbounds i32, i32* %z, i32 %a.022
store i32 %s.1, i32* %arrayidx, align 4
%inc = add nuw nsw i32 %a.022, 1
%exitcond.not = icmp eq i32 %inc, %m
br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
}
declare <4 x i1> @llvm.arm.mve.vctp32(i32)
declare i32 @llvm.arm.mve.addv.predicated.v4i32.v4i1(<4 x i32>, i32, <4 x i1>)
; Long test that was spilling lr between t2LoopDec and End
define dso_local i32 @b(i32* %c, i32 %d, i32 %e, i32* %n) "frame-pointer"="all" {
; CHECK-LABEL: b:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r4, r5, r6, r7, lr}
; CHECK-NEXT: push {r4, r5, r6, r7, lr}
; CHECK-NEXT: .setfp r7, sp, #12
; CHECK-NEXT: add r7, sp, #12
; CHECK-NEXT: .save {r8, r9, r10, r11}
; CHECK-NEXT: push.w {r8, r9, r10, r11}
; CHECK-NEXT: .pad #12
; CHECK-NEXT: sub sp, #12
; CHECK-NEXT: wls lr, r1, .LBB2_3
; CHECK-NEXT: @ %bb.1: @ %while.body.preheader
; CHECK-NEXT: mov r4, r2
; CHECK-NEXT: adds r2, r3, #4
; CHECK-NEXT: add.w r9, r0, #4
; CHECK-NEXT: mvn r11, #1
; CHECK-NEXT: @ implicit-def: $r6
; CHECK-NEXT: @ implicit-def: $r12
; CHECK-NEXT: str r4, [sp] @ 4-byte Spill
; CHECK-NEXT: .LBB2_2: @ %while.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr r1, [r9, #-4]
; CHECK-NEXT: ldr.w r10, [r2]
; CHECK-NEXT: str r0, [sp, #4] @ 4-byte Spill
; CHECK-NEXT: muls r1, r3, r1
; CHECK-NEXT: adds.w r8, r1, #-2147483648
; CHECK-NEXT: asr.w r5, r1, #31
; CHECK-NEXT: adc r1, r5, #0
; CHECK-NEXT: mul r5, r10, r0
; CHECK-NEXT: mov r0, r2
; CHECK-NEXT: ldr.w r2, [r11, #4]
; CHECK-NEXT: str r2, [sp, #8] @ 4-byte Spill
; CHECK-NEXT: ldr r2, [sp, #8] @ 4-byte Reload
; CHECK-NEXT: add.w r5, r5, #-2147483648
; CHECK-NEXT: asrl r8, r1, r5
; CHECK-NEXT: smull r4, r5, r10, r8
; CHECK-NEXT: lsll r4, r5, #30
; CHECK-NEXT: asrs r1, r5, #31
; CHECK-NEXT: mov r4, r5
; CHECK-NEXT: lsll r4, r1, r10
; CHECK-NEXT: lsll r4, r1, #30
; CHECK-NEXT: ldr.w r4, [r11]
; CHECK-NEXT: asrs r5, r1, #31
; CHECK-NEXT: mov r8, r1
; CHECK-NEXT: muls r4, r6, r4
; CHECK-NEXT: adds r4, #2
; CHECK-NEXT: lsll r8, r5, r4
; CHECK-NEXT: ldr r4, [r9], #4
; CHECK-NEXT: asr.w r5, r12, #31
; CHECK-NEXT: add.w r8, r8, #-2147483648
; CHECK-NEXT: muls r4, r3, r4
; CHECK-NEXT: adds r3, #4
; CHECK-NEXT: adds.w r1, r12, r4
; CHECK-NEXT: adc.w r5, r5, r4, asr #31
; CHECK-NEXT: smull r6, r4, r2, r6
; CHECK-NEXT: adds.w r1, r1, #-2147483648
; CHECK-NEXT: adc r1, r5, #0
; CHECK-NEXT: mov r2, r0
; CHECK-NEXT: asrs r5, r1, #31
; CHECK-NEXT: subs r6, r1, r6
; CHECK-NEXT: sbcs r5, r4
; CHECK-NEXT: adds.w r6, r6, #-2147483648
; CHECK-NEXT: adc r5, r5, #0
; CHECK-NEXT: asrl r6, r5, r8
; CHECK-NEXT: lsrl r6, r5, #2
; CHECK-NEXT: movs r5, #2
; CHECK-NEXT: str r6, [r5]
; CHECK-NEXT: ldr r5, [r11], #-4
; CHECK-NEXT: mls r1, r5, r10, r1
; CHECK-NEXT: adds.w r12, r1, #-2147483648
; CHECK-NEXT: asr.w r4, r1, #31
; CHECK-NEXT: adc r1, r4, #0
; CHECK-NEXT: ldrd r4, r0, [sp] @ 8-byte Folded Reload
; CHECK-NEXT: lsrl r12, r1, #2
; CHECK-NEXT: rsb.w r1, r12, #0
; CHECK-NEXT: adds r0, #4
; CHECK-NEXT: str r1, [r4]
; CHECK-NEXT: str r1, [r2, #-4]
; CHECK-NEXT: adds r2, #4
; CHECK-NEXT: le lr, .LBB2_2
; CHECK-NEXT: .LBB2_3: @ %while.end
; CHECK-NEXT: add sp, #12
; CHECK-NEXT: pop.w {r8, r9, r10, r11}
; CHECK-NEXT: pop {r4, r5, r6, r7, pc}
entry:
%0 = inttoptr i32 %e to i32*
%tobool.not70 = icmp eq i32 %d, 0
br i1 %tobool.not70, label %while.end, label %while.body
while.body: ; preds = %entry, %while.body
%p.077 = phi i32* [ %incdec.ptr22, %while.body ], [ inttoptr (i32 2 to i32*), %entry ]
%c.addr.076 = phi i32* [ %incdec.ptr1, %while.body ], [ %c, %entry ]
%n.075 = phi i32* [ %incdec.ptr43, %while.body ], [ %n, %entry ]
%m.074 = phi i32 [ %conv35, %while.body ], [ undef, %entry ]
%d.addr.073 = phi i32 [ %dec, %while.body ], [ %d, %entry ]
%h.072 = phi i32 [ %conv41, %while.body ], [ undef, %entry ]
%incdec.ptr43 = getelementptr inbounds i32, i32* %n.075, i32 1
%1 = ptrtoint i32* %n.075 to i32
%2 = load i32, i32* %incdec.ptr43, align 4
%3 = load i32, i32* %c.addr.076, align 4
%mul = mul nsw i32 %3, %1
%conv = sext i32 %mul to i64
%add = add nsw i64 %conv, 2147483648
%incdec.ptr1 = getelementptr inbounds i32, i32* %c.addr.076, i32 1
%4 = ptrtoint i32* %c.addr.076 to i32
%mul2 = mul nsw i32 %2, %4
%conv3 = sext i32 %mul2 to i64
%add4 = add nsw i64 %conv3, 2147483648
%shr = ashr i64 %add, %add4
%5 = shl nuw i64 %shr, 32
%conv6 = ashr exact i64 %5, 32
%conv7 = sext i32 %2 to i64
%conv11 = sext i32 %h.072 to i64
%6 = load i32, i32* %incdec.ptr1, align 4
%mul12 = mul nsw i32 %6, %1
%conv13 = sext i32 %mul12 to i64
%add14 = add nuw nsw i64 %conv11, 2147483648
%add15 = add nsw i64 %add14, %conv13
%shr16 = ashr i64 %add15, 32
%conv17 = trunc i64 %shr16 to i32
%mul8 = shl nsw i64 %conv7, 30
%7 = mul i64 %mul8, %conv6
%conv18 = ashr i64 %7, 32
%sh_prom = zext i32 %2 to i64
%shl = shl i64 %conv18, %sh_prom
%conv21 = sext i32 %conv17 to i64
%incdec.ptr22 = getelementptr inbounds i32, i32* %p.077, i32 -1
%8 = load i32, i32* %p.077, align 4
%conv23 = sext i32 %8 to i64
%conv24 = sext i32 %m.074 to i64
%mul25 = mul nsw i64 %conv23, %conv24
%sub = sub nsw i64 2147483648, %mul25
%add26 = add nsw i64 %sub, %conv21
%9 = shl i64 %shl, 30
%conv27 = ashr i64 %9, 32
%10 = load i32, i32* %incdec.ptr22, align 4
%mul28 = mul nsw i32 %10, %m.074
%add29 = add nsw i32 %mul28, 2
%sh_prom30 = zext i32 %add29 to i64
%shl31 = shl i64 %conv27, %sh_prom30
%add32 = add nsw i64 %shl31, 2147483648
%shr33 = ashr i64 %add26, %add32
%11 = lshr i64 %shr33, 2
%conv35 = trunc i64 %11 to i32
store i32 %conv35, i32* inttoptr (i32 2 to i32*), align 4
%12 = load i32, i32* %incdec.ptr22, align 4
%mul36 = mul nsw i32 %12, %2
%sub37 = sub nsw i32 %conv17, %mul36
%conv38 = sext i32 %sub37 to i64
%add39 = add nsw i64 %conv38, 2147483648
%13 = lshr i64 %add39, 2
%conv41 = trunc i64 %13 to i32
%sub42 = sub nsw i32 0, %conv41
store i32 %sub42, i32* %0, align 4
store i32 %sub42, i32* %n.075, align 4
%dec = add nsw i32 %d.addr.073, -1
%tobool.not = icmp eq i32 %dec, 0
br i1 %tobool.not, label %while.end, label %while.body
while.end: ; preds = %while.body, %entry
ret i32 undef
}
declare void @callee()
define void @callinpreheader(i32* noalias nocapture readonly %pAngle, i32* nocapture %pDst, i32 %size) {
; CHECK-LABEL: callinpreheader:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r4, r5, r6, lr}
; CHECK-NEXT: push {r4, r5, r6, lr}
; CHECK-NEXT: mov r5, r0
; CHECK-NEXT: mov r4, r1
; CHECK-NEXT: movs r0, #0
; CHECK-NEXT: cbz r2, .LBB3_3
; CHECK-NEXT: @ %bb.1: @ %for.body.ph
; CHECK-NEXT: mov r6, r2
; CHECK-NEXT: bl callee
; CHECK-NEXT: dls lr, r6
; CHECK-NEXT: movs r0, #0
; CHECK-NEXT: .LBB3_2: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr r1, [r5], #4
; CHECK-NEXT: add r0, r1
; CHECK-NEXT: le lr, .LBB3_2
; CHECK-NEXT: .LBB3_3: @ %for.cond.cleanup
; CHECK-NEXT: str r0, [r4]
; CHECK-NEXT: pop {r4, r5, r6, pc}
entry:
%cmp7.not = icmp eq i32 %size, 0
br i1 %cmp7.not, label %for.cond.cleanup, label %for.body.ph
for.body.ph:
call void @callee()
br label %for.body
for.body:
%i.09 = phi i32 [ %inc, %for.body ], [ 0, %for.body.ph ]
%s.08 = phi i32 [ %add, %for.body ], [ 0, %for.body.ph ]
%arrayidx = getelementptr inbounds i32, i32* %pAngle, i32 %i.09
%0 = load i32, i32* %arrayidx, align 4
%add = add nsw i32 %0, %s.08
%inc = add nuw nsw i32 %i.09, 1
%exitcond.not = icmp eq i32 %inc, %size
br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
for.cond.cleanup:
%s.0.lcssa = phi i32 [ 0, %entry ], [ %add, %for.body ]
store i32 %s.0.lcssa, i32* %pDst, align 4
ret void
}
|
