[llvm-branch-commits] [llvm] release/21.x: [SCEVExp] Fix early exit in ComputeEndCheck. (PR #157410)
via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Mon Sep 8 02:48:34 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-transforms
Author: Nikita Popov (nikic)
<details>
<summary>Changes</summary>
Backport of: https://github.com/llvm/llvm-project/commit/a6148071300302b451ae92fbd8f8b955a6974891 https://github.com/llvm/llvm-project/commit/ec581e460ae92ef29c1ea4f200b36b79188fdd21 https://github.com/llvm/llvm-project/commit/f8972c8280d28660aaff888c093a9e01b9ee71e6
---
Patch is 31.84 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/157410.diff
7 Files Affected:
- (modified) llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp (+9-2)
- (modified) llvm/test/Transforms/LoopDistribute/scev-inserted-runtime-check.ll (+14-4)
- (modified) llvm/test/Transforms/LoopVectorize/X86/cost-model.ll (+3-2)
- (removed) llvm/test/Transforms/LoopVectorize/interleaved-accesses-3.ll (-124)
- (added) llvm/test/Transforms/LoopVectorize/interleaved-accesses-requiring-scev-predicates.ll (+258)
- (modified) llvm/test/Transforms/LoopVectorize/runtime-check-small-clamped-bounds.ll (+3)
- (modified) llvm/test/Transforms/LoopVersioning/wrapping-pointer-versioning.ll (+7-2)
``````````diff
diff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
index 24fe08d6c3e4e..bf457194bfd8e 100644
--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
@@ -2133,8 +2133,15 @@ Value *SCEVExpander::generateOverflowCheck(const SCEVAddRecExpr *AR,
// negative. If Step is known to be positive or negative, only create
// either 1. or 2.
auto ComputeEndCheck = [&]() -> Value * {
- // Checking <u 0 is always false.
- if (!Signed && Start->isZero() && SE.isKnownPositive(Step))
+ // Checking <u 0 is always false, if (Step * trunc ExitCount) does not wrap.
+ // TODO: Predicates that can be proven true/false should be discarded when
+ // the predicates are created, not late during expansion.
+ if (!Signed && Start->isZero() && SE.isKnownPositive(Step) &&
+ DstBits < SrcBits &&
+ ExitCount == SE.getZeroExtendExpr(SE.getTruncateExpr(ExitCount, ARTy),
+ ExitCount->getType()) &&
+ SE.willNotOverflow(Instruction::Mul, Signed, Step,
+ SE.getTruncateExpr(ExitCount, ARTy)))
return ConstantInt::getFalse(Loc->getContext());
// Get the backedge taken count and truncate or extended to the AR type.
diff --git a/llvm/test/Transforms/LoopDistribute/scev-inserted-runtime-check.ll b/llvm/test/Transforms/LoopDistribute/scev-inserted-runtime-check.ll
index 98b89abfeafda..057a60918ecf3 100644
--- a/llvm/test/Transforms/LoopDistribute/scev-inserted-runtime-check.ll
+++ b/llvm/test/Transforms/LoopDistribute/scev-inserted-runtime-check.ll
@@ -10,8 +10,13 @@ define void @f(ptr noalias %a, ptr noalias %b, ptr noalias %c, ptr noalias %d, p
; CHECK-NEXT: br label [[FOR_BODY_LVER_CHECK:%.*]]
; CHECK: for.body.lver.check:
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N:%.*]], -1
+; CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP0]] to i32
+; CHECK-NEXT: [[MUL1:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 2, i32 [[TMP2]])
+; CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL1]], 0
+; CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL1]], 1
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 4294967295
-; CHECK-NEXT: br i1 [[TMP1]], label [[FOR_BODY_PH_LVER_ORIG:%.*]], label [[FOR_BODY_PH_LDIST1:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = or i1 [[MUL_OVERFLOW]], [[TMP1]]
+; CHECK-NEXT: br i1 [[TMP3]], label [[FOR_BODY_PH_LVER_ORIG:%.*]], label [[FOR_BODY_PH_LDIST1:%.*]]
; CHECK: for.body.ph.lver.orig:
; CHECK-NEXT: br label [[FOR_BODY_LVER_ORIG:%.*]]
; CHECK: for.body.lver.orig:
@@ -75,7 +80,7 @@ define void @f(ptr noalias %a, ptr noalias %b, ptr noalias %c, ptr noalias %d, p
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT2:%.*]], label [[FOR_BODY]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: br label [[FOR_END:%.*]]
-; CHECK: for.end.loopexit1:
+; CHECK: for.end.loopexit2:
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: ret void
@@ -135,8 +140,13 @@ define void @f_with_offset(ptr noalias %b, ptr noalias %c, ptr noalias %d, ptr n
; CHECK-NEXT: br label [[FOR_BODY_LVER_CHECK:%.*]]
; CHECK: for.body.lver.check:
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N:%.*]], -1
+; CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP0]] to i32
+; CHECK-NEXT: [[MUL1:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 2, i32 [[TMP2]])
+; CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL1]], 0
+; CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL1]], 1
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 4294967295
-; CHECK-NEXT: br i1 [[TMP1]], label [[FOR_BODY_PH_LVER_ORIG:%.*]], label [[FOR_BODY_PH_LDIST1:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = or i1 [[MUL_OVERFLOW]], [[TMP1]]
+; CHECK-NEXT: br i1 [[TMP3]], label [[FOR_BODY_PH_LVER_ORIG:%.*]], label [[FOR_BODY_PH_LDIST1:%.*]]
; CHECK: for.body.ph.lver.orig:
; CHECK-NEXT: br label [[FOR_BODY_LVER_ORIG:%.*]]
; CHECK: for.body.lver.orig:
@@ -200,7 +210,7 @@ define void @f_with_offset(ptr noalias %b, ptr noalias %c, ptr noalias %d, ptr n
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT2:%.*]], label [[FOR_BODY]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: br label [[FOR_END:%.*]]
-; CHECK: for.end.loopexit1:
+; CHECK: for.end.loopexit2:
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: ret void
diff --git a/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll b/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
index 147e949808b54..c1566c4dd32bf 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
@@ -1045,8 +1045,8 @@ define i64 @live_in_known_1_via_scev() {
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ <i64 3, i64 1, i64 1, i64 1>, [[VECTOR_PH]] ], [ [[VEC_PHI]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 8
-; CHECK-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 8
+; CHECK-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP3:%.*]] = call i64 @llvm.vector.reduce.mul.v4i64(<4 x i64> [[VEC_PHI]])
; CHECK-NEXT: br label [[EXIT:%.*]]
@@ -1213,6 +1213,7 @@ define i32 @g(i64 %n) {
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
+; CHECK-NEXT: [[TMP28:%.*]] = trunc i64 [[N]] to i32
; CHECK-NEXT: [[TMP2:%.*]] = icmp ugt i64 [[N]], 4294967295
; CHECK-NEXT: br i1 [[TMP2]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK: vector.main.loop.iter.check:
diff --git a/llvm/test/Transforms/LoopVectorize/interleaved-accesses-3.ll b/llvm/test/Transforms/LoopVectorize/interleaved-accesses-3.ll
deleted file mode 100644
index 661e8eb666d54..0000000000000
--- a/llvm/test/Transforms/LoopVectorize/interleaved-accesses-3.ll
+++ /dev/null
@@ -1,124 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
-; RUN: opt -S -passes=loop-vectorize,instcombine -force-vector-width=4 -force-vector-interleave=1 -enable-interleaved-mem-accesses=true < %s | FileCheck %s
-
-target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
-
-; Check that the interleaved-mem-access analysis currently does not create an
-; interleave group for access 'a' due to the possible pointer wrap-around.
-;
-; To begin with, in this test the candidate interleave group can be created
-; only when getPtrStride is called with Assume=true. Next, because
-; the interleave-group of the loads is not full (has gaps), we also need to check
-; for possible pointer wrapping. Here we currently use Assume=false and as a
-; result cannot prove the transformation is safe and therefore invalidate the
-; candidate interleave group.
-;
-
-; void func(unsigned * __restrict a, unsigned * __restrict b, unsigned char x, unsigned char y) {
-; int i = 0;
-; for (unsigned char index = x; i < y; index +=2, ++i)
-; b[i] = aptr 2;
-;
-; }
-
-define void @_Z4funcPjS_hh(ptr noalias nocapture readonly %a, ptr noalias nocapture %b, i8 zeroext %x, i8 zeroext %y) local_unnamed_addr {
-; CHECK-LABEL: define void @_Z4funcPjS_hh(
-; CHECK-SAME: ptr noalias readonly captures(none) [[A:%.*]], ptr noalias captures(none) [[B:%.*]], i8 zeroext [[X:%.*]], i8 zeroext [[Y:%.*]]) local_unnamed_addr {
-; CHECK-NEXT: [[ENTRY:.*:]]
-; CHECK-NEXT: [[CMP9:%.*]] = icmp eq i8 [[Y]], 0
-; CHECK-NEXT: br i1 [[CMP9]], label %[[FOR_COND_CLEANUP:.*]], label %[[FOR_BODY_PREHEADER:.*]]
-; CHECK: [[FOR_BODY_PREHEADER]]:
-; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i8 [[Y]] to i64
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i8 [[Y]], 5
-; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
-; CHECK: [[VECTOR_SCEVCHECK]]:
-; CHECK-NEXT: [[TMP0:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1
-; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[TMP0]] to i8
-; CHECK-NEXT: [[MUL_RESULT:%.*]] = shl i8 [[TMP1]], 1
-; CHECK-NEXT: [[TMP2:%.*]] = xor i8 [[X]], -1
-; CHECK-NEXT: [[TMP3:%.*]] = icmp ugt i8 [[MUL_RESULT]], [[TMP2]]
-; CHECK-NEXT: [[TMP4:%.*]] = icmp ugt i64 [[TMP0]], 127
-; CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
-; CHECK-NEXT: br i1 [[TMP5]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
-; CHECK: [[VECTOR_PH]]:
-; CHECK-NEXT: [[N_MOD_VF:%.*]] = and i64 [[WIDE_TRIP_COUNT]], 3
-; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
-; CHECK-NEXT: [[TMP7:%.*]] = select i1 [[TMP8]], i64 4, i64 [[N_MOD_VF]]
-; CHECK-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[WIDE_TRIP_COUNT]], [[TMP7]]
-; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i8
-; CHECK-NEXT: [[TMP6:%.*]] = shl i8 [[DOTCAST]], 1
-; CHECK-NEXT: [[IND_END:%.*]] = add i8 [[X]], [[TMP6]]
-; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
-; CHECK: [[VECTOR_BODY]]:
-; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT: [[DOTCAST3:%.*]] = trunc i64 [[INDEX]] to i8
-; CHECK-NEXT: [[TMP9:%.*]] = shl i8 [[DOTCAST3]], 1
-; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[X]], [[TMP9]]
-; CHECK-NEXT: [[TMP14:%.*]] = zext i8 [[OFFSET_IDX]] to i64
-; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[TMP14]]
-; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP15]], align 4
-; CHECK-NEXT: [[TMP23:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
-; CHECK-NEXT: [[TMP24:%.*]] = shl <4 x i32> [[TMP23]], splat (i32 1)
-; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT: store <4 x i32> [[TMP24]], ptr [[TMP25]], align 4
-; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT: [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP26]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
-; CHECK: [[MIDDLE_BLOCK]]:
-; CHECK-NEXT: br label %[[SCALAR_PH]]
-; CHECK: [[SCALAR_PH]]:
-; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[FOR_BODY_PREHEADER]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
-; CHECK-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i8 [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[X]], %[[FOR_BODY_PREHEADER]] ], [ [[X]], %[[VECTOR_SCEVCHECK]] ]
-; CHECK-NEXT: br label %[[FOR_BODY:.*]]
-; CHECK: [[FOR_COND_CLEANUP_LOOPEXIT:.*]]:
-; CHECK-NEXT: br label %[[FOR_COND_CLEANUP]]
-; CHECK: [[FOR_COND_CLEANUP]]:
-; CHECK-NEXT: ret void
-; CHECK: [[FOR_BODY]]:
-; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
-; CHECK-NEXT: [[INDEX_011:%.*]] = phi i8 [ [[ADD:%.*]], %[[FOR_BODY]] ], [ [[BC_RESUME_VAL3]], %[[SCALAR_PH]] ]
-; CHECK-NEXT: [[IDXPROM:%.*]] = zext i8 [[INDEX_011]] to i64
-; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[IDXPROM]]
-; CHECK-NEXT: [[TMP27:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
-; CHECK-NEXT: [[MUL:%.*]] = shl i32 [[TMP27]], 1
-; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
-; CHECK-NEXT: store i32 [[MUL]], ptr [[ARRAYIDX2]], align 4
-; CHECK-NEXT: [[ADD]] = add i8 [[INDEX_011]], 2
-; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
-; CHECK-NEXT: br i1 [[EXITCOND]], label %[[FOR_COND_CLEANUP_LOOPEXIT]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
-;
-entry:
- %cmp9 = icmp eq i8 %y, 0
- br i1 %cmp9, label %for.cond.cleanup, label %for.body.preheader
-
-for.body.preheader:
- %wide.trip.count = zext i8 %y to i64
- br label %for.body
-
-for.cond.cleanup.loopexit:
- br label %for.cond.cleanup
-
-for.cond.cleanup:
- ret void
-
-for.body:
- %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
- %index.011 = phi i8 [ %add, %for.body ], [ %x, %for.body.preheader ]
- %idxprom = zext i8 %index.011 to i64
- %arrayidx = getelementptr inbounds i32, ptr %a, i64 %idxprom
- %0 = load i32, ptr %arrayidx, align 4
- %mul = shl i32 %0, 1
- %arrayidx2 = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
- store i32 %mul, ptr %arrayidx2, align 4
- %add = add i8 %index.011, 2
- %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
- %exitcond = icmp eq i64 %indvars.iv.next, %wide.trip.count
- br i1 %exitcond, label %for.cond.cleanup.loopexit, label %for.body
-}
-;.
-; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
-; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
-; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
-; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]]}
-;.
diff --git a/llvm/test/Transforms/LoopVectorize/interleaved-accesses-requiring-scev-predicates.ll b/llvm/test/Transforms/LoopVectorize/interleaved-accesses-requiring-scev-predicates.ll
new file mode 100644
index 0000000000000..38438941eb357
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/interleaved-accesses-requiring-scev-predicates.ll
@@ -0,0 +1,258 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
+; RUN: opt -S -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -enable-interleaved-mem-accesses=true %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+; Check that the interleaved-mem-access analysis currently does not create an
+; interleave group for access 'a' due to the possible pointer wrap-around.
+;
+; To begin with, in this test the candidate interleave group can be created
+; only when getPtrStride is called with Assume=true. Next, because
+; the interleave-group of the loads is not full (has gaps), we also need to check
+; for possible pointer wrapping. Here we currently use Assume=false and as a
+; result cannot prove the transformation is safe and therefore invalidate the
+; candidate interleave group.
+;
+
+; void func(unsigned * __restrict a, unsigned * __restrict b, unsigned char x, unsigned char y) {
+; int i = 0;
+; for (unsigned char index = x; i < y; index +=2, ++i)
+; b[i] = aptr 2;
+;
+; }
+
+define void @wrap_around_scev_check(ptr noalias %a, ptr noalias %b, i8 %x, i8 %y) {
+; CHECK-LABEL: define void @wrap_around_scev_check(
+; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i8 [[X:%.*]], i8 [[Y:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[CMP9:%.*]] = icmp eq i8 [[Y]], 0
+; CHECK-NEXT: br i1 [[CMP9]], label %[[EXIT:.*]], label %[[LOOP_PREHEADER:.*]]
+; CHECK: [[LOOP_PREHEADER]]:
+; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i8 [[Y]] to i64
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[WIDE_TRIP_COUNT]], 4
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
+; CHECK: [[VECTOR_SCEVCHECK]]:
+; CHECK-NEXT: [[TMP0:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1
+; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[TMP0]] to i8
+; CHECK-NEXT: [[MUL1:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 2, i8 [[TMP1]])
+; CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL1]], 0
+; CHECK-NEXT: [[TMP4:%.*]] = extractvalue { i8, i1 } [[MUL1]], 1
+; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[X]], [[MUL_RESULT]]
+; CHECK-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[X]]
+; CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ugt i64 [[TMP0]], 255
+; CHECK-NEXT: [[TMP18:%.*]] = or i1 [[TMP5]], [[TMP17]]
+; CHECK-NEXT: br i1 [[TMP18]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; CHECK: [[VECTOR_PH]]:
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
+; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = select i1 [[TMP6]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[TMP7]]
+; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i8
+; CHECK-NEXT: [[TMP8:%.*]] = mul i8 [[DOTCAST]], 2
+; CHECK-NEXT: [[TMP9:%.*]] = add i8 [[X]], [[TMP8]]
+; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
+; CHECK: [[VECTOR_BODY]]:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[DOTCAST2:%.*]] = trunc i64 [[INDEX]] to i8
+; CHECK-NEXT: [[TMP10:%.*]] = mul i8 [[DOTCAST2]], 2
+; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[X]], [[TMP10]]
+; CHECK-NEXT: [[TMP11:%.*]] = zext i8 [[OFFSET_IDX]] to i64
+; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP11]]
+; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP12]], align 4
+; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT: [[TMP13:%.*]] = shl <4 x i32> [[STRIDED_VEC]], splat (i32 1)
+; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
+; CHECK-NEXT: store <4 x i32> [[TMP13]], ptr [[TMP19]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP15]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
+; CHECK-NEXT: br label %[[SCALAR_PH]]
+; CHECK: [[SCALAR_PH]]:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i8 [ [[TMP9]], %[[MIDDLE_BLOCK]] ], [ [[X]], %[[LOOP_PREHEADER]] ], [ [[X]], %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT: br label %[[LOOP:.*]]
+; CHECK: [[LOOP]]:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT: [[INDEX_011:%.*]] = phi i8 [ [[BC_RESUME_VAL3]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[LOOP]] ]
+; CHECK-NEXT: [[IDXPROM:%.*]] = zext i8 [[INDEX_011]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IDXPROM]]
+; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT: [[MUL:%.*]] = shl i32 [[TMP16]], 1
+; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; CHECK-NEXT: store i32 [[MUL]], ptr [[ARRAYIDX2]], align 4
+; CHECK-NEXT: [[ADD]] = add i8 [[INDEX_011]], 2
+; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[WIDE_TRIP_COUNT]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label %[[EXIT_LOOPEXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK: [[EXIT_LOOPEXIT]]:
+; CHECK-NEXT: br label %[[EXIT]]
+; CHECK: [[EXIT]]:
+; CHECK-NEXT: ret void
+;
+entry:
+ %cmp9 = icmp eq i8 %y, 0
+ br i1 %cmp9, label %exit, label %loop.preheader
+
+loop.preheader:
+ %wide.trip.count = zext i8 %y to i64
+ br label %loop
+
+loop:
+ %iv = phi i64 [ 0, %loop.preheader ], [ %iv.next, %loop ]
+ %index.011 = phi i8 [ %x, %loop.preheader ], [ %add, %loop ]
+ %idxprom = zext i8 %index.011 to i64
+ %arrayidx = getelementptr inbounds i32, ptr %a, i64 %idxprom
+ %0 = load i32, ptr %arrayidx, align 4
+ %mul = shl i32 %0, 1
+ %arrayidx2 = getelementptr inbounds i32, ptr %b, i64 %iv
+ store i32 %mul, ptr %arrayidx2, align 4
+ %add = add i8 %i...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/157410
More information about the llvm-branch-commits
mailing list