Skip to content

Commit 870f753

Browse files
fhahnfhahn
committed
[VPlan] Also materialize broadcasts for backedge-taken-counts (NFC).
Also include VPlan's BTC in the set of VPValues to materialize broadcasts for, if it is used.
1 parent 8a91f6b commit 870f753

11 files changed

+63
-61
lines changed

‎llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2195,6 +2195,8 @@ void VPlanTransforms::materializeBroadcasts(VPlan &Plan) {
21952195
#endif
21962196

21972197
SmallVector<VPValue *> VPValues;
2198+
if (Plan.getOrCreateBackedgeTakenCount()->getNumUsers() > 0)
2199+
VPValues.push_back(Plan.getOrCreateBackedgeTakenCount());
21982200
append_range(VPValues, Plan.getLiveIns());
21992201
for (VPRecipeBase &R : *Plan.getEntry())
22002202
append_range(VPValues, R.definedValues());

‎llvm/test/Transforms/LoopVectorize/AArch64/tail-folding-styles.ll‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -115,10 +115,10 @@ define void @simple_memset_tailfold(i32 %val, ptr %ptr, i64 %n) "target-features
115115
; DATA_NO_LANEMASK-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[UMAX]], 1
116116
; DATA_NO_LANEMASK-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
117117
; DATA_NO_LANEMASK-NEXT: [[TMP16:%.*]] = mul i64 [[TMP15]], 4
118-
; DATA_NO_LANEMASK-NEXT: [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
119-
; DATA_NO_LANEMASK-NEXT: [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT4]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
120118
; DATA_NO_LANEMASK-NEXT: [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
121119
; DATA_NO_LANEMASK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT5]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
120+
; DATA_NO_LANEMASK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
121+
; DATA_NO_LANEMASK-NEXT: [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
122122
; DATA_NO_LANEMASK-NEXT: br label [[VECTOR_BODY:%.*]]
123123
; DATA_NO_LANEMASK: vector.body:
124124
; DATA_NO_LANEMASK-NEXT: [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VECTOR_BODY]] ]

‎llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cond-reduction.ll‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -281,9 +281,9 @@ define i32 @cond_add_pred(ptr %a, i64 %n, i32 %start) {
281281
; IF-EVL-OUTLOOP-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
282282
; IF-EVL-OUTLOOP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
283283
; IF-EVL-OUTLOOP-NEXT: [[TMP8:%.*]] = mul i64 [[TMP7]], 4
284-
; IF-EVL-OUTLOOP-NEXT: [[TMP9:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START]], i32 0
285284
; IF-EVL-OUTLOOP-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
286285
; IF-EVL-OUTLOOP-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
286+
; IF-EVL-OUTLOOP-NEXT: [[TMP9:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START]], i32 0
287287
; IF-EVL-OUTLOOP-NEXT: br label [[VECTOR_BODY:%.*]]
288288
; IF-EVL-OUTLOOP: vector.body:
289289
; IF-EVL-OUTLOOP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]

‎llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-inloop-reduction.ll‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1553,10 +1553,10 @@ define float @fminimum(ptr %a, i64 %n, float %start) {
15531553
; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8
15541554
; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
15551555
; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
1556-
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1557-
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT]], <8 x float> poison, <8 x i32> zeroinitializer
15581556
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
15591557
; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer
1558+
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1559+
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer
15601560
; IF-EVL-NEXT: br label [[FOR_BODY:%.*]]
15611561
; IF-EVL: vector.body:
15621562
; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -1667,10 +1667,10 @@ define float @fmaximum(ptr %a, i64 %n, float %start) {
16671667
; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8
16681668
; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
16691669
; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
1670-
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1671-
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT]], <8 x float> poison, <8 x i32> zeroinitializer
16721670
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
16731671
; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer
1672+
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1673+
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer
16741674
; IF-EVL-NEXT: br label [[FOR_BODY:%.*]]
16751675
; IF-EVL: vector.body:
16761676
; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]

‎llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-interleave.ll‎

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -29,21 +29,21 @@ define void @interleave(ptr noalias %a, ptr noalias %b, i64 %N) {
2929
; IF-EVL-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
3030
; IF-EVL-NEXT: [[TMP8:%.*]] = mul i64 [[TMP7]], 4
3131
; IF-EVL-NEXT: [[TMP9:%.*]] = mul i64 [[TMP8]], 2
32+
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
33+
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
3234
; IF-EVL-NEXT: [[TMP10:%.*]] = call <vscale x 4 x i64> @llvm.stepvector.nxv4i64()
3335
; IF-EVL-NEXT: [[TMP12:%.*]] = mul <vscale x 4 x i64> [[TMP10]], splat (i64 1)
3436
; IF-EVL-NEXT: [[INDUCTION:%.*]] = add <vscale x 4 x i64> zeroinitializer, [[TMP12]]
35-
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TMP8]], i64 0
36-
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
37-
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
37+
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TMP8]], i64 0
3838
; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
3939
; IF-EVL-NEXT: br label [[VECTOR_BODY:%.*]]
4040
; IF-EVL: vector.body:
4141
; IF-EVL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
4242
; IF-EVL-NEXT: [[VEC_IND:%.*]] = phi <vscale x 4 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
43-
; IF-EVL-NEXT: [[STEP_ADD:%.*]] = add <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
43+
; IF-EVL-NEXT: [[STEP_ADD:%.*]] = add <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
4444
; IF-EVL-NEXT: [[TMP13:%.*]] = add i64 [[INDEX]], 0
45-
; IF-EVL-NEXT: [[TMP19:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
46-
; IF-EVL-NEXT: [[TMP20:%.*]] = icmp ule <vscale x 4 x i64> [[STEP_ADD]], [[BROADCAST_SPLAT2]]
45+
; IF-EVL-NEXT: [[TMP19:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
46+
; IF-EVL-NEXT: [[TMP20:%.*]] = icmp ule <vscale x 4 x i64> [[STEP_ADD]], [[BROADCAST_SPLAT]]
4747
; IF-EVL-NEXT: [[TMP21:%.*]] = getelementptr inbounds [2 x i32], ptr [[B:%.*]], <vscale x 4 x i64> [[VEC_IND]], i32 0
4848
; IF-EVL-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i32], ptr [[B]], <vscale x 4 x i64> [[STEP_ADD]], i32 0
4949
; IF-EVL-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP21]], i32 4, <vscale x 4 x i1> [[TMP19]], <vscale x 4 x i32> poison)
@@ -62,7 +62,7 @@ define void @interleave(ptr noalias %a, ptr noalias %b, i64 %N) {
6262
; IF-EVL-NEXT: call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP25]], ptr [[TMP29]], i32 4, <vscale x 4 x i1> [[TMP19]])
6363
; IF-EVL-NEXT: call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP26]], ptr [[TMP32]], i32 4, <vscale x 4 x i1> [[TMP20]])
6464
; IF-EVL-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP9]]
65-
; IF-EVL-NEXT: [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[STEP_ADD]], [[BROADCAST_SPLAT]]
65+
; IF-EVL-NEXT: [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[STEP_ADD]], [[BROADCAST_SPLAT2]]
6666
; IF-EVL-NEXT: [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
6767
; IF-EVL-NEXT: br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
6868
; IF-EVL: middle.block:

‎llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll‎

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -139,9 +139,9 @@ define i32 @mul(ptr %a, i64 %n, i32 %start) {
139139
; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8
140140
; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
141141
; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
142-
; IF-EVL-NEXT: [[TMP9:%.*]] = insertelement <8 x i32> splat (i32 1), i32 [[START:%.*]], i32 0
143142
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
144143
; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer
144+
; IF-EVL-NEXT: [[TMP9:%.*]] = insertelement <8 x i32> splat (i32 1), i32 [[START:%.*]], i32 0
145145
; IF-EVL-NEXT: br label [[FOR_BODY:%.*]]
146146
; IF-EVL: vector.body:
147147
; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -1240,9 +1240,9 @@ define float @fmul(ptr %a, i64 %n, float %start) {
12401240
; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8
12411241
; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
12421242
; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
1243-
; IF-EVL-NEXT: [[TMP9:%.*]] = insertelement <8 x float> splat (float 1.000000e+00), float [[START:%.*]], i32 0
12441243
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
12451244
; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer
1245+
; IF-EVL-NEXT: [[TMP9:%.*]] = insertelement <8 x float> splat (float 1.000000e+00), float [[START:%.*]], i32 0
12461246
; IF-EVL-NEXT: br label [[FOR_BODY:%.*]]
12471247
; IF-EVL: vector.body:
12481248
; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -1606,10 +1606,10 @@ define float @fminimum(ptr %a, i64 %n, float %start) {
16061606
; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8
16071607
; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
16081608
; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
1609-
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1610-
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT]], <8 x float> poison, <8 x i32> zeroinitializer
16111609
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
16121610
; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer
1611+
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1612+
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer
16131613
; IF-EVL-NEXT: br label [[FOR_BODY:%.*]]
16141614
; IF-EVL: vector.body:
16151615
; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -1720,10 +1720,10 @@ define float @fmaximum(ptr %a, i64 %n, float %start) {
17201720
; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8
17211721
; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
17221722
; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
1723-
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1724-
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT]], <8 x float> poison, <8 x i32> zeroinitializer
17251723
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
17261724
; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer
1725+
; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
1726+
; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer
17271727
; IF-EVL-NEXT: br label [[FOR_BODY:%.*]]
17281728
; IF-EVL: vector.body:
17291729
; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]

‎llvm/test/Transforms/LoopVectorize/X86/divs-with-tail-folding.ll‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -108,10 +108,10 @@ define void @sdiv_feeding_gep_predicated(ptr %dst, i32 %x, i64 %M, i64 %conv6, i
108108
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4
109109
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
110110
; CHECK-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
111-
; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[M]], i64 0
112-
; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer
113111
; CHECK-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
114112
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT2]], <4 x i64> poison, <4 x i32> zeroinitializer
113+
; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[M]], i64 0
114+
; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer
115115
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
116116
; CHECK: [[VECTOR_BODY]]:
117117
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_SDIV_CONTINUE8:.*]] ]

‎llvm/test/Transforms/LoopVectorize/X86/scev-checks-unprofitable.ll‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -27,10 +27,10 @@ define void @value_defined_in_loop1_used_for_trip_counts(i32 %start, i1 %c, ptr
2727
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 16
2828
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
2929
; CHECK-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[IV_1_LCSSA2]], 1
30-
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
31-
; CHECK: [[VECTOR_BODY]]:
3230
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <16 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
3331
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <16 x i64> [[BROADCAST_SPLATINSERT]], <16 x i64> poison, <16 x i32> zeroinitializer
32+
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
33+
; CHECK: [[VECTOR_BODY]]:
3434
; CHECK-NEXT: [[TMP0:%.*]] = icmp ule <16 x i64> <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15>, [[BROADCAST_SPLAT]]
3535
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[DST]], i64 0
3636
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0

‎llvm/test/Transforms/LoopVectorize/X86/small-size.ll‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -142,7 +142,7 @@ define void @example2(i32 %n, i32 %x) optsize {
142142
; CHECK-NEXT: [[BROADCAST_SPLATINSERT19:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_114]], i64 0
143143
; CHECK-NEXT: [[BROADCAST_SPLAT20:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT19]], <4 x i64> poison, <4 x i32> zeroinitializer
144144
; CHECK-NEXT: br label [[VECTOR_BODY13:%.*]]
145-
; CHECK: vector.body13:
145+
; CHECK: vector.body15:
146146
; CHECK-NEXT: [[INDEX16:%.*]] = phi i64 [ 0, [[VECTOR_PH8]] ], [ [[INDEX_NEXT29:%.*]], [[PRED_STORE_CONTINUE26:%.*]] ]
147147
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 [[I_0_LCSSA]], [[INDEX16]]
148148
; CHECK-NEXT: [[BROADCAST_SPLATINSERT17:%.*]] = insertelement <4 x i64> poison, i64 [[INDEX16]], i64 0

0 commit comments

Comments
 (0)