Skip to content

Commit e28fca2

Browse files
petrhosekpetrhosek
committed
Revert "[IRBuilder] Fold consistently for or/and whether constant is LHS or RHS"
This reverts commit r365260 which broke the following tests: Clang :: CodeGenCXX/cfi-mfcall.cpp Clang :: CodeGenObjC/ubsan-nullability.m LLVM :: Transforms/LoopVectorize/AArch64/pr36032.ll llvm-svn: 365284
1 parent 8d9d290 commit e28fca2

File tree

10 files changed

+88
-70
lines changed

10 files changed

+88
-70
lines changed

‎llvm/include/llvm/IR/IRBuilder.h

Lines changed: 0 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1197,9 +1197,6 @@ class IRBuilder : public IRBuilderBase, public Inserter {
11971197
}
11981198

11991199
Value *CreateAnd(Value *LHS, Value *RHS, const Twine &Name = "") {
1200-
if (auto *LC = dyn_cast<ConstantInt>(LHS))
1201-
if (LC->isMinusOne())
1202-
return RHS; // -1 & RHS = RHS
12031200
if (auto *RC = dyn_cast<Constant>(RHS)) {
12041201
if (isa<ConstantInt>(RC) && cast<ConstantInt>(RC)->isMinusOne())
12051202
return LHS; // LHS & -1 -> LHS
@@ -1226,9 +1223,6 @@ class IRBuilder : public IRBuilderBase, public Inserter {
12261223
}
12271224

12281225
Value *CreateOr(Value *LHS, Value *RHS, const Twine &Name = "") {
1229-
if (auto *LC = dyn_cast<Constant>(LHS))
1230-
if (LC->isNullValue())
1231-
return RHS; // 0 | RHS -> RHS
12321226
if (auto *RC = dyn_cast<Constant>(RHS)) {
12331227
if (RC->isNullValue())
12341228
return LHS; // LHS | 0 -> LHS

‎llvm/test/Analysis/LoopAccessAnalysis/wrapping-pointer-versioning.ll

Lines changed: 19 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -53,6 +53,8 @@ target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
5353
; LV-NEXT: [[CheckOr0:%[^ ]*]] = or i1 [[Cmp]], [[BECheck]]
5454
; LV-NEXT: [[PredCheck0:%[^ ]*]] = or i1 [[CheckOr0]], [[OFMulOverflow]]
5555

56+
; LV-NEXT: [[Or0:%[^ ]*]] = or i1 false, [[PredCheck0]]
57+
5658
; LV-NEXT: [[OFMul1:%[^ ]*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 4, i64 [[BE]])
5759
; LV-NEXT: [[OFMulResult1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 0
5860
; LV-NEXT: [[OFMulOverflow1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 1
@@ -63,7 +65,7 @@ target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
6365
; LV-NEXT: [[Cmp:%[^ ]*]] = select i1 false, i1 [[CmpNeg1]], i1 [[CmpPos1]]
6466
; LV-NEXT: [[PredCheck1:%[^ ]*]] = or i1 [[Cmp]], [[OFMulOverflow1]]
6567

66-
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[PredCheck0]], [[PredCheck1]]
68+
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[Or0]], [[PredCheck1]]
6769
; LV: br i1 [[FinalCheck]], label %for.body.ph.lver.orig, label %for.body.ph
6870
define void @f1(i16* noalias %a,
6971
i16* noalias %b, i64 %N) {
@@ -145,6 +147,9 @@ for.end: ; preds = %for.body
145147
; LV-NEXT: [[BECheck:%[^ ]*]] = icmp ugt i64 [[BE]], 4294967295
146148
; LV-NEXT: [[CheckOr0:%[^ ]*]] = or i1 [[Cmp]], [[BECheck]]
147149
; LV-NEXT: [[PredCheck0:%[^ ]*]] = or i1 [[CheckOr0]], [[OFMulOverflow]]
150+
151+
; LV-NEXT: [[Or0:%[^ ]*]] = or i1 false, [[PredCheck0]]
152+
148153
; LV: [[OFMul1:%[^ ]*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 4, i64 [[BE]])
149154
; LV-NEXT: [[OFMulResult1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 0
150155
; LV-NEXT: [[OFMulOverflow1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 1
@@ -155,7 +160,7 @@ for.end: ; preds = %for.body
155160
; LV-NEXT: [[Cmp:%[^ ]*]] = select i1 true, i1 [[CmpNeg1]], i1 [[CmpPos1]]
156161
; LV-NEXT: [[PredCheck1:%[^ ]*]] = or i1 [[Cmp]], [[OFMulOverflow1]]
157162

158-
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[PredCheck0]], [[PredCheck1]]
163+
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[Or0]], [[PredCheck1]]
159164
; LV: br i1 [[FinalCheck]], label %for.body.ph.lver.orig, label %for.body.ph
160165
define void @f2(i16* noalias %a,
161166
i16* noalias %b, i64 %N) {
@@ -222,6 +227,9 @@ for.end: ; preds = %for.body
222227
; LV-NEXT: [[BECheck:%[^ ]*]] = icmp ugt i64 [[BE]], 4294967295
223228
; LV-NEXT: [[CheckOr0:%[^ ]*]] = or i1 [[Cmp]], [[BECheck]]
224229
; LV-NEXT: [[PredCheck0:%[^ ]*]] = or i1 [[CheckOr0]], [[OFMulOverflow]]
230+
231+
; LV-NEXT: [[Or0:%[^ ]*]] = or i1 false, [[PredCheck0]]
232+
225233
; LV: [[OFMul1:%[^ ]*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 4, i64 [[BE:%[^ ]*]])
226234
; LV-NEXT: [[OFMulResult1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 0
227235
; LV-NEXT: [[OFMulOverflow1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 1
@@ -232,7 +240,7 @@ for.end: ; preds = %for.body
232240
; LV-NEXT: [[Cmp:%[^ ]*]] = select i1 false, i1 [[CmpNeg1]], i1 [[CmpPos1]]
233241
; LV-NEXT: [[PredCheck1:%[^ ]*]] = or i1 [[Cmp]], [[OFMulOverflow1]]
234242

235-
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[PredCheck0]], [[PredCheck1]]
243+
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[Or0]], [[PredCheck1]]
236244
; LV: br i1 [[FinalCheck]], label %for.body.ph.lver.orig, label %for.body.ph
237245
define void @f3(i16* noalias %a,
238246
i16* noalias %b, i64 %N) {
@@ -295,6 +303,9 @@ for.end: ; preds = %for.body
295303
; LV-NEXT: [[BECheck:%[^ ]*]] = icmp ugt i64 [[BE]], 4294967295
296304
; LV-NEXT: [[CheckOr0:%[^ ]*]] = or i1 [[Cmp]], [[BECheck]]
297305
; LV-NEXT: [[PredCheck0:%[^ ]*]] = or i1 [[CheckOr0]], [[OFMulOverflow]]
306+
307+
; LV-NEXT: [[Or0:%[^ ]*]] = or i1 false, [[PredCheck0]]
308+
298309
; LV: [[OFMul1:%[^ ]*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 4, i64 [[BE:%[^ ]*]])
299310
; LV-NEXT: [[OFMulResult1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 0
300311
; LV-NEXT: [[OFMulOverflow1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 1
@@ -305,7 +316,7 @@ for.end: ; preds = %for.body
305316
; LV-NEXT: [[Cmp:%[^ ]*]] = select i1 true, i1 [[CmpNeg1]], i1 [[CmpPos1]]
306317
; LV-NEXT: [[PredCheck1:%[^ ]*]] = or i1 [[Cmp]], [[OFMulOverflow1]]
307318

308-
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[PredCheck0]], [[PredCheck1]]
319+
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[Or0]], [[PredCheck1]]
309320
; LV: br i1 [[FinalCheck]], label %for.body.ph.lver.orig, label %for.body.ph
310321
define void @f4(i16* noalias %a,
311322
i16* noalias %b, i64 %N) {
@@ -370,6 +381,9 @@ for.end: ; preds = %for.body
370381
; LV-NEXT: [[BECheck:%[^ ]*]] = icmp ugt i64 [[BE]], 4294967295
371382
; LV-NEXT: [[CheckOr0:%[^ ]*]] = or i1 [[Cmp]], [[BECheck]]
372383
; LV-NEXT: [[PredCheck0:%[^ ]*]] = or i1 [[CheckOr0]], [[OFMulOverflow]]
384+
385+
; LV-NEXT: [[Or0:%[^ ]*]] = or i1 false, [[PredCheck0]]
386+
373387
; LV: [[OFMul1:%[^ ]*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 4, i64 [[BE:%[^ ]*]])
374388
; LV-NEXT: [[OFMulResult1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 0
375389
; LV-NEXT: [[OFMulOverflow1:%[^ ]*]] = extractvalue { i64, i1 } [[OFMul1]], 1
@@ -380,7 +394,7 @@ for.end: ; preds = %for.body
380394
; LV-NEXT: [[Cmp:%[^ ]*]] = select i1 true, i1 [[CmpNeg1]], i1 [[CmpPos1]]
381395
; LV-NEXT: [[PredCheck1:%[^ ]*]] = or i1 [[Cmp]], [[OFMulOverflow1]]
382396

383-
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[PredCheck0]], [[PredCheck1]]
397+
; LV: [[FinalCheck:%[^ ]*]] = or i1 [[Or0]], [[PredCheck1]]
384398
; LV: br i1 [[FinalCheck]], label %for.body.ph.lver.orig, label %for.body.ph
385399
define void @f5(i16* noalias %a,
386400
i16* noalias %b, i64 %N) {

‎llvm/test/Instrumentation/BoundsChecking/opt.ll

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -50,6 +50,7 @@ for.body.i: ; preds = %for.body.i, %entry
5050
; CHECK: mul i64 {{.*}}, 4
5151
; CHECK: sub i64 4000, %
5252
; CHECK-NEXT: icmp ult i64 {{.*}}, 4
53+
; CHECK-NEXT: or i1
5354
; CHECK: trap
5455
%1 = load i32, i32* %arrayidx.i, align 4
5556
%add.i = add nsw i32 %1, %sum.01.i
@@ -242,6 +243,7 @@ for.body4: ; preds = %for.body4, %for.con
242243
; CHECK: add i64
243244
; CHECK: sub i64 16, %
244245
; CHECK-NEXT: icmp ult i64 {{.*}}, 4
246+
; CHECK-NEXT: or i1
245247
; CHECK: trap
246248
%1 = load i32, i32* %arrayidx7, align 4
247249
%add = add nsw i32 %1, %sum.119

‎llvm/test/Transforms/LoopDistribute/scev-inserted-runtime-check.ll

Lines changed: 15 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -25,17 +25,18 @@ define void @f(i32* noalias %a, i32* noalias %b, i32* noalias %c, i32* noalias %
2525
; CHECK-NEXT: [[TMP7:%.*]] = icmp ugt i64 [[TMP0]], 4294967295
2626
; CHECK-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
2727
; CHECK-NEXT: [[TMP9:%.*]] = or i1 [[TMP8]], [[MUL_OVERFLOW]]
28+
; CHECK-NEXT: [[TMP10:%.*]] = or i1 false, [[TMP9]]
2829
; CHECK-NEXT: [[MUL3:%.*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 8, i64 [[TMP0]])
2930
; CHECK-NEXT: [[MUL_RESULT4:%.*]] = extractvalue { i64, i1 } [[MUL3]], 0
3031
; CHECK-NEXT: [[MUL_OVERFLOW5:%.*]] = extractvalue { i64, i1 } [[MUL3]], 1
31-
; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[A2]], [[MUL_RESULT4]]
32-
; CHECK-NEXT: [[TMP11:%.*]] = sub i64 [[A2]], [[MUL_RESULT4]]
33-
; CHECK-NEXT: [[TMP12:%.*]] = icmp ugt i64 [[TMP11]], [[A2]]
34-
; CHECK-NEXT: [[TMP13:%.*]] = icmp ult i64 [[TMP10]], [[A2]]
35-
; CHECK-NEXT: [[TMP14:%.*]] = select i1 false, i1 [[TMP12]], i1 [[TMP13]]
36-
; CHECK-NEXT: [[TMP15:%.*]] = or i1 [[TMP14]], [[MUL_OVERFLOW5]]
37-
; CHECK-NEXT: [[TMP16:%.*]] = or i1 [[TMP9]], [[TMP15]]
38-
; CHECK-NEXT: br i1 [[TMP16]], label [[FOR_BODY_PH_LVER_ORIG:%.*]], label [[FOR_BODY_PH_LDIST1:%.*]]
32+
; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[A2]], [[MUL_RESULT4]]
33+
; CHECK-NEXT: [[TMP12:%.*]] = sub i64 [[A2]], [[MUL_RESULT4]]
34+
; CHECK-NEXT: [[TMP13:%.*]] = icmp ugt i64 [[TMP12]], [[A2]]
35+
; CHECK-NEXT: [[TMP14:%.*]] = icmp ult i64 [[TMP11]], [[A2]]
36+
; CHECK-NEXT: [[TMP15:%.*]] = select i1 false, i1 [[TMP13]], i1 [[TMP14]]
37+
; CHECK-NEXT: [[TMP16:%.*]] = or i1 [[TMP15]], [[MUL_OVERFLOW5]]
38+
; CHECK-NEXT: [[TMP17:%.*]] = or i1 [[TMP10]], [[TMP16]]
39+
; CHECK-NEXT: br i1 [[TMP17]], label [[FOR_BODY_PH_LVER_ORIG:%.*]], label [[FOR_BODY_PH_LDIST1:%.*]]
3940
; CHECK: for.body.ph.lver.orig:
4041
; CHECK-NEXT: br label [[FOR_BODY_LVER_ORIG:%.*]]
4142
; CHECK: for.body.lver.orig:
@@ -69,14 +70,14 @@ define void @f(i32* noalias %a, i32* noalias %b, i32* noalias %c, i32* noalias %
6970
; CHECK-NEXT: [[MUL_LDIST1:%.*]] = mul i32 [[IND1_LDIST1]], 2
7071
; CHECK-NEXT: [[MUL_EXT_LDIST1:%.*]] = zext i32 [[MUL_LDIST1]] to i64
7172
; CHECK-NEXT: [[ARRAYIDXA_LDIST1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[MUL_EXT_LDIST1]]
72-
; CHECK-NEXT: [[LOADA_LDIST1:%.*]] = load i32, i32* [[ARRAYIDXA_LDIST1]], align 4, !alias.scope !0
73+
; CHECK-NEXT: [[LOADA_LDIST1:%.*]] = load i32, i32* [[ARRAYIDXA_LDIST1]], align 4
7374
; CHECK-NEXT: [[ARRAYIDXB_LDIST1:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[MUL_EXT_LDIST1]]
74-
; CHECK-NEXT: [[LOADB_LDIST1:%.*]] = load i32, i32* [[ARRAYIDXB_LDIST1]], align 4, !alias.scope !3
75+
; CHECK-NEXT: [[LOADB_LDIST1:%.*]] = load i32, i32* [[ARRAYIDXB_LDIST1]], align 4
7576
; CHECK-NEXT: [[MULA_LDIST1:%.*]] = mul i32 [[LOADB_LDIST1]], [[LOADA_LDIST1]]
7677
; CHECK-NEXT: [[ADD_LDIST1]] = add nuw nsw i64 [[IND_LDIST1]], 1
7778
; CHECK-NEXT: [[INC1_LDIST1]] = add i32 [[IND1_LDIST1]], 1
7879
; CHECK-NEXT: [[ARRAYIDXA_PLUS_4_LDIST1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[ADD_LDIST1]]
79-
; CHECK-NEXT: store i32 [[MULA_LDIST1]], i32* [[ARRAYIDXA_PLUS_4_LDIST1]], align 4, !alias.scope !5
80+
; CHECK-NEXT: store i32 [[MULA_LDIST1]], i32* [[ARRAYIDXA_PLUS_4_LDIST1]], align 4
8081
; CHECK-NEXT: [[EXITCOND_LDIST1:%.*]] = icmp eq i64 [[ADD_LDIST1]], [[N]]
8182
; CHECK-NEXT: br i1 [[EXITCOND_LDIST1]], label [[FOR_BODY_PH:%.*]], label [[FOR_BODY_LDIST1]]
8283
; CHECK: for.body.ph:
@@ -89,12 +90,12 @@ define void @f(i32* noalias %a, i32* noalias %b, i32* noalias %c, i32* noalias %
8990
; CHECK-NEXT: [[ADD]] = add nuw nsw i64 [[IND]], 1
9091
; CHECK-NEXT: [[INC1]] = add i32 [[IND1]], 1
9192
; CHECK-NEXT: [[ARRAYIDXD:%.*]] = getelementptr inbounds i32, i32* [[D]], i64 [[MUL_EXT]]
92-
; CHECK-NEXT: [[LOADD:%.*]] = load i32, i32* [[ARRAYIDXD]], align 4, !alias.scope !7
93+
; CHECK-NEXT: [[LOADD:%.*]] = load i32, i32* [[ARRAYIDXD]], align 4
9394
; CHECK-NEXT: [[ARRAYIDXE:%.*]] = getelementptr inbounds i32, i32* [[E]], i64 [[MUL_EXT]]
94-
; CHECK-NEXT: [[LOADE:%.*]] = load i32, i32* [[ARRAYIDXE]], align 4, !alias.scope !9
95+
; CHECK-NEXT: [[LOADE:%.*]] = load i32, i32* [[ARRAYIDXE]], align 4
9596
; CHECK-NEXT: [[MULC:%.*]] = mul i32 [[LOADD]], [[LOADE]]
9697
; CHECK-NEXT: [[ARRAYIDXC:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[MUL_EXT]]
97-
; CHECK-NEXT: store i32 [[MULC]], i32* [[ARRAYIDXC]], align 4, !alias.scope !11
98+
; CHECK-NEXT: store i32 [[MULC]], i32* [[ARRAYIDXC]], align 4
9899
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[ADD]], [[N]]
99100
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]]
100101
; CHECK: for.end:

‎llvm/test/Transforms/LoopPredication/basic.ll

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -232,11 +232,12 @@ define i32 @signed_loop_0_to_n_ult_check_length_range_known(i32* %array, i32* %l
232232
; CHECK-NEXT: br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
233233
; CHECK: loop.preheader:
234234
; CHECK-NEXT: [[TMP0:%.*]] = icmp sle i32 [[N]], [[LENGTH]]
235+
; CHECK-NEXT: [[TMP1:%.*]] = and i1 true, [[TMP0]]
235236
; CHECK-NEXT: br label [[LOOP:%.*]]
236237
; CHECK: loop:
237238
; CHECK-NEXT: [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
238239
; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
239-
; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[TMP0]], i32 9) [ "deopt"() ]
240+
; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[TMP1]], i32 9) [ "deopt"() ]
240241
; CHECK-NEXT: [[I_I64:%.*]] = zext i32 [[I]] to i64
241242
; CHECK-NEXT: [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
242243
; CHECK-NEXT: [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4

‎llvm/test/Transforms/LoopPredication/basic_widenable_branch_guards.ll

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -282,13 +282,14 @@ define i32 @signed_loop_0_to_n_ult_check_length_range_known(i32* %array, i32* %l
282282
; CHECK-NEXT: br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
283283
; CHECK: loop.preheader:
284284
; CHECK-NEXT: [[TMP0:%.*]] = icmp sle i32 [[N]], [[LENGTH]]
285+
; CHECK-NEXT: [[TMP1:%.*]] = and i1 true, [[TMP0]]
285286
; CHECK-NEXT: br label [[LOOP:%.*]]
286287
; CHECK: loop:
287288
; CHECK-NEXT: [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
288289
; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
289290
; CHECK-NEXT: [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
290-
; CHECK-NEXT: [[TMP1:%.*]] = and i1 [[TMP0]], [[WIDENABLE_COND]]
291-
; CHECK-NEXT: br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
291+
; CHECK-NEXT: [[TMP2:%.*]] = and i1 [[TMP1]], [[WIDENABLE_COND]]
292+
; CHECK-NEXT: br i1 [[TMP2]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
292293
; CHECK: deopt:
293294
; CHECK-NEXT: [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
294295
; CHECK-NEXT: ret i32 [[DEOPTCALL]]

‎llvm/test/Transforms/LoopPredication/invariant_load.ll

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -385,13 +385,14 @@ define i32 @constant_length(i32* %array, i32 %n) {
385385
; CHECK-NEXT: br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
386386
; CHECK: loop.preheader:
387387
; CHECK-NEXT: [[TMP0:%.*]] = icmp ule i32 [[N]], 20
388+
; CHECK-NEXT: [[TMP1:%.*]] = and i1 true, [[TMP0]]
388389
; CHECK-NEXT: br label [[LOOP:%.*]]
389390
; CHECK: loop:
390391
; CHECK-NEXT: [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
391392
; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
392393
; CHECK-NEXT: [[UNKNOWN:%.*]] = load volatile i1, i1* @UNKNOWN
393394
; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[UNKNOWN]]) [ "deopt"() ]
394-
; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[TMP0]], i32 9) [ "deopt"() ]
395+
; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[TMP1]], i32 9) [ "deopt"() ]
395396
; CHECK-NEXT: [[I_I64:%.*]] = zext i32 [[I]] to i64
396397
; CHECK-NEXT: [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
397398
; CHECK-NEXT: [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4

‎llvm/test/Transforms/LoopVectorize/X86/illegal-parallel-loop-uniform-write.ll

Lines changed: 18 additions & 17 deletions
Original file line numberDiff line numberDiff line change
@@ -71,33 +71,34 @@ define void @foo(i32* nocapture %a, i32* nocapture %b, i32 %k, i32 %m) #0 {
7171
; CHECK-NEXT: [[TMP13:%.*]] = icmp slt i32 [[TMP10]], [[TMP8]]
7272
; CHECK-NEXT: [[TMP14:%.*]] = select i1 false, i1 [[TMP12]], i1 [[TMP13]]
7373
; CHECK-NEXT: [[TMP15:%.*]] = or i1 [[TMP14]], [[MUL_OVERFLOW]]
74-
; CHECK-NEXT: br i1 [[TMP15]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
74+
; CHECK-NEXT: [[TMP16:%.*]] = or i1 false, [[TMP15]]
75+
; CHECK-NEXT: br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
7576
; CHECK: vector.ph:
7677
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
7778
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
7879
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
7980
; CHECK: vector.body:
8081
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
81-
; CHECK-NEXT: [[TMP16:%.*]] = trunc i64 [[INDEX]] to i32
82-
; CHECK-NEXT: [[TMP17:%.*]] = add i32 [[TMP16]], 0
83-
; CHECK-NEXT: [[TMP18:%.*]] = add i32 [[ADD_US]], [[TMP17]]
84-
; CHECK-NEXT: [[TMP19:%.*]] = sext i32 [[TMP18]] to i64
85-
; CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP19]]
86-
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP20]], i32 0
87-
; CHECK-NEXT: [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <4 x i32>*
88-
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP22]], align 4
89-
; CHECK-NEXT: [[TMP23:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], <i32 1, i32 1, i32 1, i32 1>
90-
; CHECK-NEXT: [[TMP24:%.*]] = extractelement <4 x i32> [[TMP23]], i32 0
91-
; CHECK-NEXT: store i32 [[TMP24]], i32* [[ARRAYIDX7_US]], align 4, !llvm.mem.parallel_loop_access !0
92-
; CHECK-NEXT: [[TMP25:%.*]] = extractelement <4 x i32> [[TMP23]], i32 1
82+
; CHECK-NEXT: [[TMP17:%.*]] = trunc i64 [[INDEX]] to i32
83+
; CHECK-NEXT: [[TMP18:%.*]] = add i32 [[TMP17]], 0
84+
; CHECK-NEXT: [[TMP19:%.*]] = add i32 [[ADD_US]], [[TMP18]]
85+
; CHECK-NEXT: [[TMP20:%.*]] = sext i32 [[TMP19]] to i64
86+
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
87+
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i32 0
88+
; CHECK-NEXT: [[TMP23:%.*]] = bitcast i32* [[TMP22]] to <4 x i32>*
89+
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP23]], align 4
90+
; CHECK-NEXT: [[TMP24:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], <i32 1, i32 1, i32 1, i32 1>
91+
; CHECK-NEXT: [[TMP25:%.*]] = extractelement <4 x i32> [[TMP24]], i32 0
9392
; CHECK-NEXT: store i32 [[TMP25]], i32* [[ARRAYIDX7_US]], align 4, !llvm.mem.parallel_loop_access !0
94-
; CHECK-NEXT: [[TMP26:%.*]] = extractelement <4 x i32> [[TMP23]], i32 2
93+
; CHECK-NEXT: [[TMP26:%.*]] = extractelement <4 x i32> [[TMP24]], i32 1
9594
; CHECK-NEXT: store i32 [[TMP26]], i32* [[ARRAYIDX7_US]], align 4, !llvm.mem.parallel_loop_access !0
96-
; CHECK-NEXT: [[TMP27:%.*]] = extractelement <4 x i32> [[TMP23]], i32 3
95+
; CHECK-NEXT: [[TMP27:%.*]] = extractelement <4 x i32> [[TMP24]], i32 2
9796
; CHECK-NEXT: store i32 [[TMP27]], i32* [[ARRAYIDX7_US]], align 4, !llvm.mem.parallel_loop_access !0
97+
; CHECK-NEXT: [[TMP28:%.*]] = extractelement <4 x i32> [[TMP24]], i32 3
98+
; CHECK-NEXT: store i32 [[TMP28]], i32* [[ARRAYIDX7_US]], align 4, !llvm.mem.parallel_loop_access !0
9899
; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4
99-
; CHECK-NEXT: [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
100-
; CHECK-NEXT: br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !5
100+
; CHECK-NEXT: [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
101+
; CHECK-NEXT: br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !5
101102
; CHECK: middle.block:
102103
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
103104
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_US]], label [[SCALAR_PH]]

0 commit comments

Comments
 (0)