[llvm] [LAA] Keep pointer checks on partial analysis (PR #139719)
John Brawn via llvm-commits
llvm-commits at lists.llvm.org
Wed Jun 4 07:56:37 PDT 2025
https://github.com/john-brawn-arm updated https://github.com/llvm/llvm-project/pull/139719
>From babdff19cc33a5843de761b8f09ab3b65d89bcf4 Mon Sep 17 00:00:00 2001
From: John Brawn <john.brawn at arm.com>
Date: Wed, 26 Mar 2025 10:34:27 +0000
Subject: [PATCH 1/6] [LoopAccessAnalysis] Keep pointer checks on partial
analysis
Currently if there's any memory access that AccessAnalysis couldn't
analyze then all of the runtime pointer check results are discarded.
This patch changes that so they're not, and we generate the runtime
check information for those pointers that we could analyze, as
transformations may still be able to make use of the partial
information.
Of the transformations that use LoopAccessAnalysis, only
LoopVersioningLICM changes behaviour as a result of this change. This
is because the others either:
* Check canVectorizeMemory, which will return false when we have
partial pointer information as analyzeLoop() will return false.
* Examine the dependencies returned by getDepChecker(), which will be
empty as we exit analyzeLoop if we have partial pointer information
before calling areDepsSafe(), which is what fills in the dependency
information.
---
.../llvm/Analysis/LoopAccessAnalysis.h | 9 +-
llvm/lib/Analysis/LoopAccessAnalysis.cpp | 19 +-
.../LoopAccessAnalysis/forked-pointers.ll | 307 +++++++++++++++---
.../non-constant-strides-backward.ll | 12 +
.../non-constant-strides-forward.ll | 9 +
.../LoopAccessAnalysis/pointer-phis.ll | 21 +-
...ter-dependence-analysis-forked-pointers.ll | 14 +
.../underlying-object-loop-varying-phi.ll | 14 +
.../load-from-unknown-address.ll | 168 ++++++++--
9 files changed, 471 insertions(+), 102 deletions(-)
diff --git a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
index f715e0ec8dbb4..0abd7f1182108 100644
--- a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
+++ b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
@@ -192,7 +192,8 @@ class MemoryDepChecker {
/// of a write access.
void addAccess(LoadInst *LI);
- /// Check whether the dependencies between the accesses are safe.
+ /// Check whether the dependencies between the accesses are safe, and records
+ /// the dependence information in Dependences if so.
///
/// Only checks sets with elements in \p CheckDeps.
bool areDepsSafe(const DepCandidates &AccessSets,
@@ -779,10 +780,14 @@ class LoopAccessInfo {
/// We need to check that all of the pointers in this list are disjoint
/// at runtime. Using std::unique_ptr to make using move ctor simpler.
+ /// This list may contain only partial information when we've failed to
+ /// analyze all the memory accesses in the loop (i.e. CanVecMem is false).
std::unique_ptr<RuntimePointerChecking> PtrRtChecking;
- /// the Memory Dependence Checker which can determine the
+ /// The Memory Dependence Checker which can determine the
/// loop-independent and loop-carried dependences between memory accesses.
+ /// This will be empty if we've failed to analyze all the memory access in the
+ /// loop (i.e. CanVecMem is false).
std::unique_ptr<MemoryDepChecker> DepChecker;
Loop *TheLoop;
diff --git a/llvm/lib/Analysis/LoopAccessAnalysis.cpp b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
index af1a3c593c514..4d912abdf59a8 100644
--- a/llvm/lib/Analysis/LoopAccessAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
@@ -529,8 +529,10 @@ void RuntimePointerChecking::groupChecks(
// equivalence class, the iteration order is deterministic.
for (auto M : DepCands.members(Access)) {
auto PointerI = PositionMap.find(M.getPointer());
- assert(PointerI != PositionMap.end() &&
- "pointer in equivalence class not found in PositionMap");
+ // If we can't find the pointer in PositionMap that means we can't
+ // generate a memcheck for it.
+ if (PointerI == PositionMap.end())
+ continue;
for (unsigned Pointer : PointerI->second) {
bool Merged = false;
// Mark this pointer as seen.
@@ -682,7 +684,9 @@ class AccessAnalysis {
/// non-intersection.
///
/// Returns true if we need no check or if we do and we can generate them
- /// (i.e. the pointers have computable bounds).
+ /// (i.e. the pointers have computable bounds). A return value of false means
+ /// we couldn't analyze and generate runtime checks for all pointers in the
+ /// loop, but we will have checks for those pointers we could analyze.
bool canCheckPtrAtRT(RuntimePointerChecking &RtCheck, ScalarEvolution *SE,
Loop *TheLoop,
const DenseMap<Value *, const SCEV *> &Strides,
@@ -1273,7 +1277,6 @@ bool AccessAnalysis::canCheckPtrAtRT(
/*Assume=*/true)) {
CanDoAliasSetRT = false;
UncomputablePtr = Access.getPointer();
- break;
}
}
}
@@ -1313,7 +1316,7 @@ bool AccessAnalysis::canCheckPtrAtRT(
}
}
- if (MayNeedRTCheck && CanDoRT)
+ if (MayNeedRTCheck)
RtCheck.generateChecks(DepCands, IsDepCheckNeeded);
LLVM_DEBUG(dbgs() << "LAA: We need to do " << RtCheck.getNumberOfChecks()
@@ -1323,11 +1326,7 @@ bool AccessAnalysis::canCheckPtrAtRT(
// are needed. This can happen when all pointers point to the same underlying
// object for example.
RtCheck.Need = CanDoRT ? RtCheck.getNumberOfChecks() != 0 : MayNeedRTCheck;
-
- bool CanDoRTIfNeeded = !RtCheck.Need || CanDoRT;
- if (!CanDoRTIfNeeded)
- RtCheck.reset();
- return CanDoRTIfNeeded;
+ return !RtCheck.Need || CanDoRT;
}
void AccessAnalysis::processMemAccesses() {
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll b/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
index cd388b4ee87f2..f0885cf0a6043 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
@@ -633,7 +633,18 @@ define dso_local void @forked_ptrs_same_base_different_offset(ptr nocapture read
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP47:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; CHECK-NEXT: Against group ([[GRP48:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP47]]:
+; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; CHECK-NEXT: Group [[GRP48]]:
+; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -645,7 +656,18 @@ define dso_local void @forked_ptrs_same_base_different_offset(ptr nocapture read
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP49:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP50:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP49]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP50]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -684,24 +706,24 @@ define dso_local void @forked_ptrs_add_to_offset(ptr nocapture readonly %Base, p
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Check 0:
-; CHECK-NEXT: Comparing group ([[GRP47:0x[0-9a-f]+]]):
+; CHECK-NEXT: Comparing group ([[GRP51:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP48:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP52:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Check 1:
-; CHECK-NEXT: Comparing group ([[GRP47]]):
+; CHECK-NEXT: Comparing group ([[GRP51]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP49:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP53:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx3 = getelementptr inbounds float, ptr %Base, i64 %offset
; CHECK-NEXT: %arrayidx3 = getelementptr inbounds float, ptr %Base, i64 %offset
; CHECK-NEXT: Grouped accesses:
-; CHECK-NEXT: Group [[GRP47]]:
+; CHECK-NEXT: Group [[GRP51]]:
; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group [[GRP48]]:
+; CHECK-NEXT: Group [[GRP52]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group [[GRP49]]:
+; CHECK-NEXT: Group [[GRP53]]:
; CHECK-NEXT: (Low: ((4 * %extra_offset) + %Base) High: (404 + (4 * %extra_offset) + %Base))
; CHECK-NEXT: Member: {(4 + (4 * %extra_offset) + %Base),+,4}<%for.body>
; CHECK-NEXT: Member: {((4 * %extra_offset) + %Base),+,4}<%for.body>
@@ -716,7 +738,18 @@ define dso_local void @forked_ptrs_add_to_offset(ptr nocapture readonly %Base, p
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP54:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP55:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP54]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP55]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -752,24 +785,24 @@ define dso_local void @forked_ptrs_sub_from_offset(ptr nocapture readonly %Base,
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Check 0:
-; CHECK-NEXT: Comparing group ([[GRP50:0x[0-9a-f]+]]):
+; CHECK-NEXT: Comparing group ([[GRP56:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP51:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP57:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Check 1:
-; CHECK-NEXT: Comparing group ([[GRP50]]):
+; CHECK-NEXT: Comparing group ([[GRP56]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP52:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP58:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx3 = getelementptr inbounds float, ptr %Base, i64 %offset
; CHECK-NEXT: %arrayidx3 = getelementptr inbounds float, ptr %Base, i64 %offset
; CHECK-NEXT: Grouped accesses:
-; CHECK-NEXT: Group [[GRP50]]:
+; CHECK-NEXT: Group [[GRP56]]:
; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group [[GRP51]]:
+; CHECK-NEXT: Group [[GRP57]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group [[GRP52]]:
+; CHECK-NEXT: Group [[GRP58]]:
; CHECK-NEXT: (Low: ((-4 * %extra_offset) + %Base) High: (404 + (-4 * %extra_offset) + %Base))
; CHECK-NEXT: Member: {(4 + (-4 * %extra_offset) + %Base),+,4}<%for.body>
; CHECK-NEXT: Member: {((-4 * %extra_offset) + %Base),+,4}<%for.body>
@@ -784,7 +817,18 @@ define dso_local void @forked_ptrs_sub_from_offset(ptr nocapture readonly %Base,
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP59:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP60:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP59]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP60]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -820,24 +864,24 @@ define dso_local void @forked_ptrs_add_sub_offset(ptr nocapture readonly %Base,
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Check 0:
-; CHECK-NEXT: Comparing group ([[GRP53:0x[0-9a-f]+]]):
+; CHECK-NEXT: Comparing group ([[GRP61:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP54:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP62:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Check 1:
-; CHECK-NEXT: Comparing group ([[GRP53]]):
+; CHECK-NEXT: Comparing group ([[GRP61]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP55:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP63:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx3 = getelementptr inbounds float, ptr %Base, i64 %offset
; CHECK-NEXT: %arrayidx3 = getelementptr inbounds float, ptr %Base, i64 %offset
; CHECK-NEXT: Grouped accesses:
-; CHECK-NEXT: Group [[GRP53]]:
+; CHECK-NEXT: Group [[GRP61]]:
; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group [[GRP54]]:
+; CHECK-NEXT: Group [[GRP62]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group [[GRP55]]:
+; CHECK-NEXT: Group [[GRP63]]:
; CHECK-NEXT: (Low: ((4 * %to_add) + (-4 * %to_sub) + %Base) High: (404 + (4 * %to_add) + (-4 * %to_sub) + %Base))
; CHECK-NEXT: Member: {(4 + (4 * %to_add) + (-4 * %to_sub) + %Base),+,4}<%for.body>
; CHECK-NEXT: Member: {((4 * %to_add) + (-4 * %to_sub) + %Base),+,4}<%for.body>
@@ -852,7 +896,18 @@ define dso_local void @forked_ptrs_add_sub_offset(ptr nocapture readonly %Base,
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP64:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP65:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP64]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP65]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -890,7 +945,18 @@ define dso_local void @forked_ptrs_mul_by_offset(ptr nocapture readonly %Base, p
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP66:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; CHECK-NEXT: Against group ([[GRP67:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP66]]:
+; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; CHECK-NEXT: Group [[GRP67]]:
+; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -902,7 +968,18 @@ define dso_local void @forked_ptrs_mul_by_offset(ptr nocapture readonly %Base, p
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP68:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP69:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP68]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP69]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -940,7 +1017,18 @@ define dso_local void @forked_ptrs_uniform_and_strided_forks(ptr nocapture reado
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP70:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; CHECK-NEXT: Against group ([[GRP71:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP70]]:
+; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; CHECK-NEXT: Group [[GRP71]]:
+; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -952,7 +1040,18 @@ define dso_local void @forked_ptrs_uniform_and_strided_forks(ptr nocapture reado
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP72:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP73:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP72]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP73]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -995,7 +1094,18 @@ define dso_local void @forked_ptrs_gather_and_contiguous_forks(ptr nocapture rea
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP74:0x[0-9a-f]+]]):
+; CHECK-NEXT: %1 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; CHECK-NEXT: Against group ([[GRP75:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP74]]:
+; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; CHECK-NEXT: Group [[GRP75]]:
+; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1007,7 +1117,18 @@ define dso_local void @forked_ptrs_gather_and_contiguous_forks(ptr nocapture rea
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP76:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %1 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP77:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP76]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP77]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1046,7 +1167,18 @@ define dso_local void @forked_ptrs_two_forks_gep(ptr nocapture readonly %Base1,
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP78:0x[0-9a-f]+]]):
+; CHECK-NEXT: %1 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; CHECK-NEXT: Against group ([[GRP79:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP78]]:
+; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; CHECK-NEXT: Group [[GRP79]]:
+; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1058,7 +1190,18 @@ define dso_local void @forked_ptrs_two_forks_gep(ptr nocapture readonly %Base1,
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP80:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %1 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP81:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP80]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP81]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1096,6 +1239,10 @@ define void @forked_ptrs_two_select(ptr nocapture readonly %Base1, ptr nocapture
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP82:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%loop>
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1108,6 +1255,10 @@ define void @forked_ptrs_two_select(ptr nocapture readonly %Base1, ptr nocapture
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP83:0x[0-9a-f]+]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%loop>
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%loop>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1146,7 +1297,18 @@ define void @forked_ptrs_too_many_gep_ops(ptr nocapture readonly %Base1, ptr noc
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP84:0x[0-9a-f]+]]):
+; CHECK-NEXT: %1 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; CHECK-NEXT: Against group ([[GRP85:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP84]]:
+; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; CHECK-NEXT: Group [[GRP85]]:
+; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1158,7 +1320,18 @@ define void @forked_ptrs_too_many_gep_ops(ptr nocapture readonly %Base1, ptr noc
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP86:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %1 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP87:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP86]]:
+; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP87]]:
+; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1194,7 +1367,18 @@ define void @forked_ptrs_vector_gep(ptr nocapture readonly %Base1, ptr nocapture
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP88:0x[0-9a-f]+]]):
+; CHECK-NEXT: %1 = getelementptr inbounds <4 x float>, ptr %Dest, i64 %indvars.iv
+; CHECK-NEXT: Against group ([[GRP89:0x[0-9a-f]+]]):
+; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP88]]:
+; CHECK-NEXT: (Low: %Dest High: (1552 + %Dest))
+; CHECK-NEXT: Member: {%Dest,+,64}<nuw><%for.body>
+; CHECK-NEXT: Group [[GRP89]]:
+; CHECK-NEXT: (Low: %Preds High: (388 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,16}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1206,7 +1390,18 @@ define void @forked_ptrs_vector_gep(ptr nocapture readonly %Base1, ptr nocapture
; RECURSE-NEXT: Report: cannot identify array bounds
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
+; RECURSE-NEXT: Check 0:
+; RECURSE-NEXT: Comparing group ([[GRP90:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %1 = getelementptr inbounds <4 x float>, ptr %Dest, i64 %indvars.iv
+; RECURSE-NEXT: Against group ([[GRP91:0x[0-9a-f]+]]):
+; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP90]]:
+; RECURSE-NEXT: (Low: %Dest High: (1552 + %Dest))
+; RECURSE-NEXT: Member: {%Dest,+,64}<nuw><%for.body>
+; RECURSE-NEXT: Group [[GRP91]]:
+; RECURSE-NEXT: (Low: %Preds High: (388 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,16}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1250,15 +1445,15 @@ define void @sc_add_expr_ice(ptr %Base1, ptr %Base2, i64 %N) {
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Check 0:
-; CHECK-NEXT: Comparing group ([[GRP56:0x[0-9a-f]+]]):
+; CHECK-NEXT: Comparing group ([[GRP92:0x[0-9a-f]+]]):
; CHECK-NEXT: ptr %Base1
-; CHECK-NEXT: Against group ([[GRP57:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP93:0x[0-9a-f]+]]):
; CHECK-NEXT: %fptr = getelementptr inbounds double, ptr %Base2, i64 %sel
; CHECK-NEXT: Grouped accesses:
-; CHECK-NEXT: Group [[GRP56]]:
+; CHECK-NEXT: Group [[GRP92]]:
; CHECK-NEXT: (Low: %Base1 High: (8 + %Base1))
; CHECK-NEXT: Member: %Base1
-; CHECK-NEXT: Group [[GRP57]]:
+; CHECK-NEXT: Group [[GRP93]]:
; CHECK-NEXT: (Low: %Base2 High: ((8 * %N) + %Base2))
; CHECK-NEXT: Member: {%Base2,+,8}<%for.body>
; CHECK-EMPTY:
@@ -1277,15 +1472,15 @@ define void @sc_add_expr_ice(ptr %Base1, ptr %Base2, i64 %N) {
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
; RECURSE-NEXT: Check 0:
-; RECURSE-NEXT: Comparing group ([[GRP58:0x[0-9a-f]+]]):
+; RECURSE-NEXT: Comparing group ([[GRP94:0x[0-9a-f]+]]):
; RECURSE-NEXT: ptr %Base1
-; RECURSE-NEXT: Against group ([[GRP59:0x[0-9a-f]+]]):
+; RECURSE-NEXT: Against group ([[GRP95:0x[0-9a-f]+]]):
; RECURSE-NEXT: %fptr = getelementptr inbounds double, ptr %Base2, i64 %sel
; RECURSE-NEXT: Grouped accesses:
-; RECURSE-NEXT: Group [[GRP58]]:
+; RECURSE-NEXT: Group [[GRP94]]:
; RECURSE-NEXT: (Low: %Base1 High: (8 + %Base1))
; RECURSE-NEXT: Member: %Base1
-; RECURSE-NEXT: Group [[GRP59]]:
+; RECURSE-NEXT: Group [[GRP95]]:
; RECURSE-NEXT: (Low: %Base2 High: ((8 * %N) + %Base2))
; RECURSE-NEXT: Member: {%Base2,+,8}<%for.body>
; RECURSE-EMPTY:
@@ -1324,36 +1519,36 @@ define void @forked_ptrs_with_different_base(ptr nocapture readonly %Preds, ptr
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Check 0:
-; CHECK-NEXT: Comparing group ([[GRP60:0x[0-9a-f]+]]):
+; CHECK-NEXT: Comparing group ([[GRP96:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP61:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP97:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Check 1:
-; CHECK-NEXT: Comparing group ([[GRP60]]):
+; CHECK-NEXT: Comparing group ([[GRP96]]):
; CHECK-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP62:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP98:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds double, ptr %0, i64 %indvars.iv
; CHECK-NEXT: Check 2:
-; CHECK-NEXT: Comparing group ([[GRP63:0x[0-9a-f]+]]):
+; CHECK-NEXT: Comparing group ([[GRP99:0x[0-9a-f]+]]):
; CHECK-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP61]]):
+; CHECK-NEXT: Against group ([[GRP97]]):
; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Check 3:
-; CHECK-NEXT: Comparing group ([[GRP63]]):
+; CHECK-NEXT: Comparing group ([[GRP99]]):
; CHECK-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; CHECK-NEXT: Against group ([[GRP62]]):
+; CHECK-NEXT: Against group ([[GRP98]]):
; CHECK-NEXT: %arrayidx5 = getelementptr inbounds double, ptr %0, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
-; CHECK-NEXT: Group [[GRP60]]:
+; CHECK-NEXT: Group [[GRP96]]:
; CHECK-NEXT: (Low: %1 High: (63992 + %1))
; CHECK-NEXT: Member: {%1,+,8}<nw><%for.body>
-; CHECK-NEXT: Group [[GRP63]]:
+; CHECK-NEXT: Group [[GRP99]]:
; CHECK-NEXT: (Low: %2 High: (63992 + %2))
; CHECK-NEXT: Member: {%2,+,8}<nw><%for.body>
-; CHECK-NEXT: Group [[GRP61]]:
+; CHECK-NEXT: Group [[GRP97]]:
; CHECK-NEXT: (Low: %Preds High: (31996 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group [[GRP62]]:
+; CHECK-NEXT: Group [[GRP98]]:
; CHECK-NEXT: (Low: %0 High: (63992 + %0))
; CHECK-NEXT: Member: {%0,+,8}<nw><%for.body>
; CHECK-EMPTY:
@@ -1368,36 +1563,36 @@ define void @forked_ptrs_with_different_base(ptr nocapture readonly %Preds, ptr
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
; RECURSE-NEXT: Check 0:
-; RECURSE-NEXT: Comparing group ([[GRP64:0x[0-9a-f]+]]):
+; RECURSE-NEXT: Comparing group ([[GRP100:0x[0-9a-f]+]]):
; RECURSE-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; RECURSE-NEXT: Against group ([[GRP65:0x[0-9a-f]+]]):
+; RECURSE-NEXT: Against group ([[GRP101:0x[0-9a-f]+]]):
; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Check 1:
-; RECURSE-NEXT: Comparing group ([[GRP64]]):
+; RECURSE-NEXT: Comparing group ([[GRP100]]):
; RECURSE-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; RECURSE-NEXT: Against group ([[GRP66:0x[0-9a-f]+]]):
+; RECURSE-NEXT: Against group ([[GRP102:0x[0-9a-f]+]]):
; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds double, ptr %0, i64 %indvars.iv
; RECURSE-NEXT: Check 2:
-; RECURSE-NEXT: Comparing group ([[GRP67:0x[0-9a-f]+]]):
+; RECURSE-NEXT: Comparing group ([[GRP103:0x[0-9a-f]+]]):
; RECURSE-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; RECURSE-NEXT: Against group ([[GRP65]]):
+; RECURSE-NEXT: Against group ([[GRP101]]):
; RECURSE-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; RECURSE-NEXT: Check 3:
-; RECURSE-NEXT: Comparing group ([[GRP67]]):
+; RECURSE-NEXT: Comparing group ([[GRP103]]):
; RECURSE-NEXT: %arrayidx7 = getelementptr inbounds double, ptr %.sink, i64 %indvars.iv
-; RECURSE-NEXT: Against group ([[GRP66]]):
+; RECURSE-NEXT: Against group ([[GRP102]]):
; RECURSE-NEXT: %arrayidx5 = getelementptr inbounds double, ptr %0, i64 %indvars.iv
; RECURSE-NEXT: Grouped accesses:
-; RECURSE-NEXT: Group [[GRP64]]:
+; RECURSE-NEXT: Group [[GRP100]]:
; RECURSE-NEXT: (Low: %1 High: (63992 + %1))
; RECURSE-NEXT: Member: {%1,+,8}<nw><%for.body>
-; RECURSE-NEXT: Group [[GRP67]]:
+; RECURSE-NEXT: Group [[GRP103]]:
; RECURSE-NEXT: (Low: %2 High: (63992 + %2))
; RECURSE-NEXT: Member: {%2,+,8}<nw><%for.body>
-; RECURSE-NEXT: Group [[GRP65]]:
+; RECURSE-NEXT: Group [[GRP101]]:
; RECURSE-NEXT: (Low: %Preds High: (31996 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
-; RECURSE-NEXT: Group [[GRP66]]:
+; RECURSE-NEXT: Group [[GRP102]]:
; RECURSE-NEXT: (Low: %0 High: (63992 + %0))
; RECURSE-NEXT: Member: {%0,+,8}<nw><%for.body>
; RECURSE-EMPTY:
@@ -1452,6 +1647,9 @@ define void @forked_ptrs_with_different_base3(ptr nocapture readonly %Preds, ptr
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP104:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %Preds High: (31996 + %Preds))
+; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1464,6 +1662,9 @@ define void @forked_ptrs_with_different_base3(ptr nocapture readonly %Preds, ptr
; RECURSE-NEXT: Dependences:
; RECURSE-NEXT: Run-time memory checks:
; RECURSE-NEXT: Grouped accesses:
+; RECURSE-NEXT: Group [[GRP105:0x[0-9a-f]+]]:
+; RECURSE-NEXT: (Low: %Preds High: (31996 + %Preds))
+; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
index d263749ea1f46..1e696af9e2e4b 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
@@ -237,6 +237,9 @@ define void @different_non_constant_strides_known_backward_via_assume(ptr %A, i6
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP1:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (1024 + %A))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -271,6 +274,9 @@ define void @different_non_constant_strides_known_backward_via_assume_distance_l
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP2:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (1024 + %A)<nuw>)
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -306,6 +312,9 @@ define void @different_non_constant_strides_known_backward_via_assume_min_distan
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP3:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (1024 + %A))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -341,6 +350,9 @@ define void @different_non_constant_strides_not_known_backward(ptr %A, i64 %scal
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP4:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (1024 + %A))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
index 5f4c732dc19df..e33d20e795646 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
@@ -85,6 +85,9 @@ define void @different_non_constant_strides_known_forward_via_assume(ptr %A, i64
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP1:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (1024 + %A))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -119,6 +122,9 @@ define void @different_non_constant_strides_known_forward_via_assume_min_distanc
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP2:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (1024 + %A))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -154,6 +160,9 @@ define void @different_non_constant_strides_not_known_forward(ptr %A, i64 %scale
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP3:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (1024 + %A))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
index 48586ee9d9ed9..0f0635913c891 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
@@ -462,6 +462,9 @@ define i32 @store_with_pointer_phi_in_same_bb_use_other_phi(ptr %A, ptr %B, ptr
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP10:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %A High: (256000 + %A))
+; CHECK-NEXT: Member: {%A,+,8}<nuw><%loop.header>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -503,36 +506,36 @@ define void @phi_load_store_memdep_check(i1 %c, ptr %A, ptr %B, ptr %C) {
; CHECK-EMPTY:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Check 0:
-; CHECK-NEXT: Comparing group ([[GRP10:0x[0-9a-f]+]]):
+; CHECK-NEXT: Comparing group ([[GRP11:0x[0-9a-f]+]]):
; CHECK-NEXT: ptr %A
; CHECK-NEXT: ptr %A
-; CHECK-NEXT: Against group ([[GRP11:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP12:0x[0-9a-f]+]]):
; CHECK-NEXT: ptr %C
; CHECK-NEXT: ptr %C
; CHECK-NEXT: Check 1:
-; CHECK-NEXT: Comparing group ([[GRP10]]):
+; CHECK-NEXT: Comparing group ([[GRP11]]):
; CHECK-NEXT: ptr %A
; CHECK-NEXT: ptr %A
-; CHECK-NEXT: Against group ([[GRP12:0x[0-9a-f]+]]):
+; CHECK-NEXT: Against group ([[GRP13:0x[0-9a-f]+]]):
; CHECK-NEXT: ptr %B
; CHECK-NEXT: ptr %B
; CHECK-NEXT: Check 2:
-; CHECK-NEXT: Comparing group ([[GRP11]]):
+; CHECK-NEXT: Comparing group ([[GRP12]]):
; CHECK-NEXT: ptr %C
; CHECK-NEXT: ptr %C
-; CHECK-NEXT: Against group ([[GRP12]]):
+; CHECK-NEXT: Against group ([[GRP13]]):
; CHECK-NEXT: ptr %B
; CHECK-NEXT: ptr %B
; CHECK-NEXT: Grouped accesses:
-; CHECK-NEXT: Group [[GRP10]]:
+; CHECK-NEXT: Group [[GRP11]]:
; CHECK-NEXT: (Low: %A High: (2 + %A))
; CHECK-NEXT: Member: %A
; CHECK-NEXT: Member: %A
-; CHECK-NEXT: Group [[GRP11]]:
+; CHECK-NEXT: Group [[GRP12]]:
; CHECK-NEXT: (Low: %C High: (2 + %C))
; CHECK-NEXT: Member: %C
; CHECK-NEXT: Member: %C
-; CHECK-NEXT: Group [[GRP12]]:
+; CHECK-NEXT: Group [[GRP13]]:
; CHECK-NEXT: (Low: %B High: (2 + %B))
; CHECK-NEXT: Member: %B
; CHECK-NEXT: Member: %B
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll b/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
index 021447d53f943..35233de1cce73 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
@@ -167,6 +167,13 @@ define void @dependency_check_and_runtime_checks_needed_select_of_ptr_add_recs_m
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP9:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: ((4 * %offset) + %a) High: ((4 * %offset) + (4 * %n) + %a))
+; CHECK-NEXT: Member: {((4 * %offset) + %a),+,4}<%loop>
+; CHECK-NEXT: Group [[GRP10:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %a High: ((4 * %n) + %a))
+; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
+; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -208,6 +215,13 @@ define void @dependency_check_and_runtime_checks_needed_select_of_ptr_add_recs_m
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP11:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: ((4 * %offset) + %a) High: ((4 * %offset) + (4 * %n) + %a))
+; CHECK-NEXT: Member: {((4 * %offset) + %a),+,4}<%loop>
+; CHECK-NEXT: Group [[GRP12:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: %a High: ((4 * %n) + %a))
+; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
+; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll b/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
index 402081fb939f0..9e25baf31107d 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
@@ -80,6 +80,9 @@ define void @indirect_ptr_recurrences_read_write_may_alias_no_tbaa(ptr %A, ptr %
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP1:0x[0-9a-f]+]]:
+; CHECK-NEXT: (Low: (8 + %B)<nuw> High: (40 + %B))
+; CHECK-NEXT: Member: {(8 + %B)<nuw>,+,8}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -111,7 +114,18 @@ define void @indirect_ptr_recurrences_read_write_may_alias_different_obj(ptr %A,
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group ([[GRP2:0x[0-9a-f]+]]):
+; CHECK-NEXT: %gep.C = getelementptr inbounds ptr, ptr %C, i64 %iv
+; CHECK-NEXT: Against group ([[GRP3:0x[0-9a-f]+]]):
+; CHECK-NEXT: %gep.B = getelementptr inbounds ptr, ptr %B, i64 %iv
; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group [[GRP2]]:
+; CHECK-NEXT: (Low: (8 + %C)<nuw> High: (36 + %C))
+; CHECK-NEXT: Member: {(8 + %C)<nuw>,+,8}<nuw><%loop>
+; CHECK-NEXT: Group [[GRP3]]:
+; CHECK-NEXT: (Low: (8 + %B)<nuw> High: (40 + %B))
+; CHECK-NEXT: Member: {(8 + %B)<nuw>,+,8}<nuw><%loop>
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Transforms/LoopVersioningLICM/load-from-unknown-address.ll b/llvm/test/Transforms/LoopVersioningLICM/load-from-unknown-address.ll
index e9b2954039198..89576f1684c3d 100644
--- a/llvm/test/Transforms/LoopVersioningLICM/load-from-unknown-address.ll
+++ b/llvm/test/Transforms/LoopVersioningLICM/load-from-unknown-address.ll
@@ -7,15 +7,25 @@ target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i8:8:32-i16:16:32-i6
; accesses but not others.
; Load from a gep whose bounds can't be calculated as the offset is loaded from memory
-; FIXME: Not knowing the bounds of the gep shouldn't stop us from hoisting the load of rval
define void @gep_loaded_offset(ptr %p, ptr %q, ptr %r, i32 %n) {
; CHECK-LABEL: define void @gep_loaded_offset(
; CHECK-SAME: ptr [[P:%.*]], ptr [[Q:%.*]], ptr [[R:%.*]], i32 [[N:%.*]]) {
-; CHECK-NEXT: [[ENTRY:.*]]:
-; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
-; CHECK: [[WHILE_BODY]]:
-; CHECK-NEXT: [[N_ADDR:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[ENTRY]] ]
-; CHECK-NEXT: [[P_ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], %[[WHILE_BODY]] ], [ [[P]], %[[ENTRY]] ]
+; CHECK-NEXT: [[WHILE_BODY_LVER_CHECK:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = add nsw i32 [[N]], -1
+; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 2
+; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 4
+; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP3]]
+; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[R]], i64 8
+; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[P]], [[SCEVGEP1]]
+; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[R]], [[SCEVGEP]]
+; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label %[[WHILE_BODY_PH_LVER_ORIG:.*]], label %[[WHILE_BODY_PH:.*]]
+; CHECK: [[WHILE_BODY_PH_LVER_ORIG]]:
+; CHECK-NEXT: br label %[[WHILE_BODY_LVER_ORIG:.*]]
+; CHECK: [[WHILE_BODY_LVER_ORIG]]:
+; CHECK-NEXT: [[N_ADDR:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[N]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
+; CHECK-NEXT: [[P_ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[P]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
; CHECK-NEXT: [[DEC]] = add nsw i32 [[N_ADDR]], -1
; CHECK-NEXT: [[RVAL:%.*]] = load i64, ptr [[R]], align 4
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[Q]], i64 [[RVAL]]
@@ -23,7 +33,24 @@ define void @gep_loaded_offset(ptr %p, ptr %q, ptr %r, i32 %n) {
; CHECK-NEXT: [[INCDEC_PTR]] = getelementptr inbounds nuw i8, ptr [[P_ADDR]], i64 4
; CHECK-NEXT: store i32 [[VAL]], ptr [[P_ADDR]], align 4
; CHECK-NEXT: [[TOBOOL_NOT:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_BODY]]
+; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY_LVER_ORIG]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK: [[WHILE_BODY_PH]]:
+; CHECK-NEXT: [[RVAL1:%.*]] = load i64, ptr [[R]], align 4, !alias.scope [[META2:![0-9]+]]
+; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, ptr [[Q]], i64 [[RVAL1]]
+; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
+; CHECK: [[WHILE_BODY]]:
+; CHECK-NEXT: [[N_ADDR1:%.*]] = phi i32 [ [[DEC1:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[P_ADDR1:%.*]] = phi ptr [ [[INCDEC_PTR1:%.*]], %[[WHILE_BODY]] ], [ [[P]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[DEC1]] = add nsw i32 [[N_ADDR1]], -1
+; CHECK-NEXT: [[VAL1:%.*]] = load i32, ptr [[ARRAYIDX1]], align 4
+; CHECK-NEXT: [[INCDEC_PTR1]] = getelementptr inbounds nuw i8, ptr [[P_ADDR1]], i64 4
+; CHECK-NEXT: store i32 [[VAL1]], ptr [[P_ADDR1]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META2]]
+; CHECK-NEXT: [[TOBOOL_NOT1:%.*]] = icmp eq i32 [[DEC1]], 0
+; CHECK-NEXT: br i1 [[TOBOOL_NOT1]], label %[[WHILE_END_LOOPEXIT2:.*]], label %[[WHILE_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK: [[WHILE_END_LOOPEXIT]]:
+; CHECK-NEXT: br label %[[WHILE_END:.*]]
+; CHECK: [[WHILE_END_LOOPEXIT2]]:
+; CHECK-NEXT: br label %[[WHILE_END]]
; CHECK: [[WHILE_END]]:
; CHECK-NEXT: ret void
;
@@ -89,15 +116,25 @@ while.end:
}
; Load from a gep whose bounds can't be calculated as the pointer is loaded from memory
-; FIXME: Not knowing the bounds of the gep shouldn't stop us from hoisting the load of rval
define void @gep_loaded_base(ptr %p, ptr %q, ptr %r, i32 %n) {
; CHECK-LABEL: define void @gep_loaded_base(
; CHECK-SAME: ptr [[P:%.*]], ptr [[Q:%.*]], ptr [[R:%.*]], i32 [[N:%.*]]) {
-; CHECK-NEXT: [[ENTRY:.*]]:
-; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
-; CHECK: [[WHILE_BODY]]:
-; CHECK-NEXT: [[N_ADDR:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[ENTRY]] ]
-; CHECK-NEXT: [[P_ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], %[[WHILE_BODY]] ], [ [[P]], %[[ENTRY]] ]
+; CHECK-NEXT: [[WHILE_BODY_LVER_CHECK:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = add nsw i32 [[N]], -1
+; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 2
+; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 4
+; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP3]]
+; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[R]], i64 8
+; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[P]], [[SCEVGEP1]]
+; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[R]], [[SCEVGEP]]
+; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label %[[WHILE_BODY_PH_LVER_ORIG:.*]], label %[[WHILE_BODY_PH:.*]]
+; CHECK: [[WHILE_BODY_PH_LVER_ORIG]]:
+; CHECK-NEXT: br label %[[WHILE_BODY_LVER_ORIG:.*]]
+; CHECK: [[WHILE_BODY_LVER_ORIG]]:
+; CHECK-NEXT: [[N_ADDR:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[N]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
+; CHECK-NEXT: [[P_ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[P]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
; CHECK-NEXT: [[DEC]] = add nsw i32 [[N_ADDR]], -1
; CHECK-NEXT: [[RVAL:%.*]] = load ptr, ptr [[R]], align 4
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[RVAL]], i64 0
@@ -105,7 +142,24 @@ define void @gep_loaded_base(ptr %p, ptr %q, ptr %r, i32 %n) {
; CHECK-NEXT: [[INCDEC_PTR]] = getelementptr inbounds nuw i8, ptr [[P_ADDR]], i64 4
; CHECK-NEXT: store i32 [[VAL]], ptr [[P_ADDR]], align 4
; CHECK-NEXT: [[TOBOOL_NOT:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_BODY]]
+; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY_LVER_ORIG]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK: [[WHILE_BODY_PH]]:
+; CHECK-NEXT: [[RVAL1:%.*]] = load ptr, ptr [[R]], align 4, !alias.scope [[META10:![0-9]+]]
+; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, ptr [[RVAL1]], i64 0
+; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
+; CHECK: [[WHILE_BODY]]:
+; CHECK-NEXT: [[N_ADDR1:%.*]] = phi i32 [ [[DEC1:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[P_ADDR1:%.*]] = phi ptr [ [[INCDEC_PTR1:%.*]], %[[WHILE_BODY]] ], [ [[P]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[DEC1]] = add nsw i32 [[N_ADDR1]], -1
+; CHECK-NEXT: [[VAL1:%.*]] = load i32, ptr [[ARRAYIDX1]], align 4
+; CHECK-NEXT: [[INCDEC_PTR1]] = getelementptr inbounds nuw i8, ptr [[P_ADDR1]], i64 4
+; CHECK-NEXT: store i32 [[VAL1]], ptr [[P_ADDR1]], align 4, !alias.scope [[META13:![0-9]+]], !noalias [[META10]]
+; CHECK-NEXT: [[TOBOOL_NOT1:%.*]] = icmp eq i32 [[DEC1]], 0
+; CHECK-NEXT: br i1 [[TOBOOL_NOT1]], label %[[WHILE_END_LOOPEXIT2:.*]], label %[[WHILE_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; CHECK: [[WHILE_END_LOOPEXIT]]:
+; CHECK-NEXT: br label %[[WHILE_END:.*]]
+; CHECK: [[WHILE_END_LOOPEXIT2]]:
+; CHECK-NEXT: br label %[[WHILE_END]]
; CHECK: [[WHILE_END]]:
; CHECK-NEXT: ret void
;
@@ -129,15 +183,25 @@ while.end:
}
; Load from a gep with an offset that scalar evolution can't describe
-; FIXME: Not knowing the bounds of the gep shouldn't stop us from hoisting the load of qval
define void @gep_strange_offset(ptr %p, ptr %q, ptr %r, i32 %n) {
; CHECK-LABEL: define void @gep_strange_offset(
; CHECK-SAME: ptr [[P:%.*]], ptr [[Q:%.*]], ptr [[R:%.*]], i32 [[N:%.*]]) {
-; CHECK-NEXT: [[ENTRY:.*]]:
-; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
-; CHECK: [[WHILE_BODY]]:
-; CHECK-NEXT: [[N_ADDR:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[ENTRY]] ]
-; CHECK-NEXT: [[P_ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], %[[WHILE_BODY]] ], [ [[P]], %[[ENTRY]] ]
+; CHECK-NEXT: [[WHILE_BODY_LVER_CHECK:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = add nsw i32 [[N]], -1
+; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 2
+; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 4
+; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP3]]
+; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[Q]], i64 4
+; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[P]], [[SCEVGEP1]]
+; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[Q]], [[SCEVGEP]]
+; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label %[[WHILE_BODY_PH_LVER_ORIG:.*]], label %[[WHILE_BODY_PH:.*]]
+; CHECK: [[WHILE_BODY_PH_LVER_ORIG]]:
+; CHECK-NEXT: br label %[[WHILE_BODY_LVER_ORIG:.*]]
+; CHECK: [[WHILE_BODY_LVER_ORIG]]:
+; CHECK-NEXT: [[N_ADDR:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[N]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
+; CHECK-NEXT: [[P_ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[P]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
; CHECK-NEXT: [[DEC]] = add nsw i32 [[N_ADDR]], -1
; CHECK-NEXT: [[QVAL:%.*]] = load i32, ptr [[Q]], align 4
; CHECK-NEXT: [[REM:%.*]] = srem i32 [[DEC]], 2
@@ -148,7 +212,27 @@ define void @gep_strange_offset(ptr %p, ptr %q, ptr %r, i32 %n) {
; CHECK-NEXT: [[INCDEC_PTR]] = getelementptr inbounds nuw i8, ptr [[P_ADDR]], i64 4
; CHECK-NEXT: store i32 [[ADD]], ptr [[P_ADDR]], align 4
; CHECK-NEXT: [[TOBOOL_NOT:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_BODY]]
+; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY_LVER_ORIG]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK: [[WHILE_BODY_PH]]:
+; CHECK-NEXT: [[QVAL1:%.*]] = load i32, ptr [[Q]], align 4, !alias.scope [[META17:![0-9]+]]
+; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
+; CHECK: [[WHILE_BODY]]:
+; CHECK-NEXT: [[N_ADDR1:%.*]] = phi i32 [ [[DEC1:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[P_ADDR1:%.*]] = phi ptr [ [[INCDEC_PTR1:%.*]], %[[WHILE_BODY]] ], [ [[P]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[DEC1]] = add nsw i32 [[N_ADDR1]], -1
+; CHECK-NEXT: [[REM1:%.*]] = srem i32 [[DEC1]], 2
+; CHECK-NEXT: [[IDXPROM1:%.*]] = sext i32 [[REM1]] to i64
+; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, ptr [[R]], i64 [[IDXPROM1]]
+; CHECK-NEXT: [[VAL1:%.*]] = load i32, ptr [[ARRAYIDX1]], align 4
+; CHECK-NEXT: [[ADD1:%.*]] = add nsw i32 [[VAL1]], [[QVAL1]]
+; CHECK-NEXT: [[INCDEC_PTR1]] = getelementptr inbounds nuw i8, ptr [[P_ADDR1]], i64 4
+; CHECK-NEXT: store i32 [[ADD1]], ptr [[P_ADDR1]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
+; CHECK-NEXT: [[TOBOOL_NOT1:%.*]] = icmp eq i32 [[DEC1]], 0
+; CHECK-NEXT: br i1 [[TOBOOL_NOT1]], label %[[WHILE_END_LOOPEXIT2:.*]], label %[[WHILE_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK: [[WHILE_END_LOOPEXIT]]:
+; CHECK-NEXT: br label %[[WHILE_END:.*]]
+; CHECK: [[WHILE_END_LOOPEXIT2]]:
+; CHECK-NEXT: br label %[[WHILE_END]]
; CHECK: [[WHILE_END]]:
; CHECK-NEXT: ret void
;
@@ -175,15 +259,24 @@ while.end:
}
; A memcpy-like loop where the source address is loaded from a pointer
-; FIXME: We should be able to hoist the load of the source address pointer
define void @memcpy_load_src(ptr %dst, ptr %src, i32 %n) {
; CHECK-LABEL: define void @memcpy_load_src(
; CHECK-SAME: ptr [[DST:%.*]], ptr [[SRC:%.*]], i32 [[N:%.*]]) {
-; CHECK-NEXT: [[ENTRY:.*]]:
-; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
-; CHECK: [[WHILE_BODY]]:
-; CHECK-NEXT: [[N_VAL:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[ENTRY]] ]
-; CHECK-NEXT: [[DST_VAL:%.*]] = phi ptr [ [[DST_VAL_NEXT:%.*]], %[[WHILE_BODY]] ], [ [[DST]], %[[ENTRY]] ]
+; CHECK-NEXT: [[WHILE_BODY_LVER_CHECK:.*:]]
+; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[SRC]], i64 8
+; CHECK-NEXT: [[TMP0:%.*]] = add nsw i32 [[N]], -1
+; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[DST]], i64 [[TMP2]]
+; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SRC]], [[SCEVGEP1]]
+; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[DST]], [[SCEVGEP]]
+; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label %[[WHILE_BODY_PH_LVER_ORIG:.*]], label %[[WHILE_BODY_PH:.*]]
+; CHECK: [[WHILE_BODY_PH_LVER_ORIG]]:
+; CHECK-NEXT: br label %[[WHILE_BODY_LVER_ORIG:.*]]
+; CHECK: [[WHILE_BODY_LVER_ORIG]]:
+; CHECK-NEXT: [[N_VAL:%.*]] = phi i32 [ [[DEC:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[N]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
+; CHECK-NEXT: [[DST_VAL:%.*]] = phi ptr [ [[DST_VAL_NEXT:%.*]], %[[WHILE_BODY_LVER_ORIG]] ], [ [[DST]], %[[WHILE_BODY_PH_LVER_ORIG]] ]
; CHECK-NEXT: [[DEC]] = add nsw i32 [[N_VAL]], -1
; CHECK-NEXT: [[SRC_VAL:%.*]] = load ptr, ptr [[SRC]], align 8
; CHECK-NEXT: [[SRC_VAL_NEXT:%.*]] = getelementptr inbounds nuw i8, ptr [[SRC_VAL]], i64 1
@@ -192,7 +285,26 @@ define void @memcpy_load_src(ptr %dst, ptr %src, i32 %n) {
; CHECK-NEXT: [[VAL:%.*]] = load i8, ptr [[SRC_VAL]], align 1
; CHECK-NEXT: store i8 [[VAL]], ptr [[DST_VAL]], align 1
; CHECK-NEXT: [[TOBOOL_NOT:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END:.*]], label %[[WHILE_BODY]]
+; CHECK-NEXT: br i1 [[TOBOOL_NOT]], label %[[WHILE_END_LOOPEXIT:.*]], label %[[WHILE_BODY_LVER_ORIG]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK: [[WHILE_BODY_PH]]:
+; CHECK-NEXT: [[SRC_PROMOTED:%.*]] = load ptr, ptr [[SRC]], align 8, !alias.scope [[META24:![0-9]+]], !noalias [[META27:![0-9]+]]
+; CHECK-NEXT: br label %[[WHILE_BODY:.*]]
+; CHECK: [[WHILE_BODY]]:
+; CHECK-NEXT: [[SRC_VAL_NEXT3:%.*]] = phi ptr [ [[SRC_VAL_NEXT1:%.*]], %[[WHILE_BODY]] ], [ [[SRC_PROMOTED]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[N_VAL1:%.*]] = phi i32 [ [[DEC1:%.*]], %[[WHILE_BODY]] ], [ [[N]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[DST_VAL1:%.*]] = phi ptr [ [[DST_VAL_NEXT1:%.*]], %[[WHILE_BODY]] ], [ [[DST]], %[[WHILE_BODY_PH]] ]
+; CHECK-NEXT: [[DEC1]] = add nsw i32 [[N_VAL1]], -1
+; CHECK-NEXT: [[SRC_VAL_NEXT1]] = getelementptr inbounds nuw i8, ptr [[SRC_VAL_NEXT3]], i64 1
+; CHECK-NEXT: [[DST_VAL_NEXT1]] = getelementptr inbounds nuw i8, ptr [[DST_VAL1]], i64 1
+; CHECK-NEXT: store ptr [[SRC_VAL_NEXT1]], ptr [[SRC]], align 8, !alias.scope [[META24]], !noalias [[META27]]
+; CHECK-NEXT: [[VAL1:%.*]] = load i8, ptr [[SRC_VAL_NEXT3]], align 1
+; CHECK-NEXT: store i8 [[VAL1]], ptr [[DST_VAL1]], align 1, !alias.scope [[META27]]
+; CHECK-NEXT: [[TOBOOL_NOT1:%.*]] = icmp eq i32 [[DEC1]], 0
+; CHECK-NEXT: br i1 [[TOBOOL_NOT1]], label %[[WHILE_END_LOOPEXIT2:.*]], label %[[WHILE_BODY]], !llvm.loop [[LOOP29:![0-9]+]]
+; CHECK: [[WHILE_END_LOOPEXIT]]:
+; CHECK-NEXT: br label %[[WHILE_END:.*]]
+; CHECK: [[WHILE_END_LOOPEXIT2]]:
+; CHECK-NEXT: br label %[[WHILE_END]]
; CHECK: [[WHILE_END]]:
; CHECK-NEXT: ret void
;
>From 6dc3239e79ff1076e7eb49558f18882f55cad34f Mon Sep 17 00:00:00 2001
From: John Brawn <john.brawn at arm.com>
Date: Thu, 15 May 2025 17:58:15 +0100
Subject: [PATCH 2/6] Add extra line to LoopAccessInfo::print when we generate
some but not all checks
---
llvm/lib/Analysis/LoopAccessAnalysis.cpp | 2 ++
.../LoopAccessAnalysis/forked-pointers.ll | 21 +++++++++++++++++++
.../non-constant-strides-backward.ll | 4 ++++
.../non-constant-strides-forward.ll | 3 +++
.../LoopAccessAnalysis/nusw-predicates.ll | 1 +
.../LoopAccessAnalysis/pointer-phis.ll | 4 ++++
...ter-dependence-analysis-forked-pointers.ll | 2 ++
.../underlying-object-loop-varying-phi.ll | 2 ++
.../unsafe-and-rt-checks-convergent.ll | 1 +
9 files changed, 40 insertions(+)
diff --git a/llvm/lib/Analysis/LoopAccessAnalysis.cpp b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
index 4d912abdf59a8..3117214cf8f61 100644
--- a/llvm/lib/Analysis/LoopAccessAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
@@ -3001,6 +3001,8 @@ void LoopAccessInfo::print(raw_ostream &OS, unsigned Depth) const {
// List the pair of accesses need run-time checks to prove independence.
PtrRtChecking->print(OS, Depth);
+ if (PtrRtChecking->Need && !CanVecMem)
+ OS.indent(Depth) << "Generated run-time checks are incomplete\n";
OS << "\n";
OS.indent(Depth)
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll b/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
index f0885cf0a6043..3bf3bf1353bcf 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
@@ -645,6 +645,7 @@ define dso_local void @forked_ptrs_same_base_different_offset(ptr nocapture read
; CHECK-NEXT: Group [[GRP48]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -668,6 +669,7 @@ define dso_local void @forked_ptrs_same_base_different_offset(ptr nocapture read
; RECURSE-NEXT: Group [[GRP50]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -750,6 +752,7 @@ define dso_local void @forked_ptrs_add_to_offset(ptr nocapture readonly %Base, p
; RECURSE-NEXT: Group [[GRP55]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -829,6 +832,7 @@ define dso_local void @forked_ptrs_sub_from_offset(ptr nocapture readonly %Base,
; RECURSE-NEXT: Group [[GRP60]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -908,6 +912,7 @@ define dso_local void @forked_ptrs_add_sub_offset(ptr nocapture readonly %Base,
; RECURSE-NEXT: Group [[GRP65]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -957,6 +962,7 @@ define dso_local void @forked_ptrs_mul_by_offset(ptr nocapture readonly %Base, p
; CHECK-NEXT: Group [[GRP67]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -980,6 +986,7 @@ define dso_local void @forked_ptrs_mul_by_offset(ptr nocapture readonly %Base, p
; RECURSE-NEXT: Group [[GRP69]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1029,6 +1036,7 @@ define dso_local void @forked_ptrs_uniform_and_strided_forks(ptr nocapture reado
; CHECK-NEXT: Group [[GRP71]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1052,6 +1060,7 @@ define dso_local void @forked_ptrs_uniform_and_strided_forks(ptr nocapture reado
; RECURSE-NEXT: Group [[GRP73]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1106,6 +1115,7 @@ define dso_local void @forked_ptrs_gather_and_contiguous_forks(ptr nocapture rea
; CHECK-NEXT: Group [[GRP75]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1129,6 +1139,7 @@ define dso_local void @forked_ptrs_gather_and_contiguous_forks(ptr nocapture rea
; RECURSE-NEXT: Group [[GRP77]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1179,6 +1190,7 @@ define dso_local void @forked_ptrs_two_forks_gep(ptr nocapture readonly %Base1,
; CHECK-NEXT: Group [[GRP79]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1202,6 +1214,7 @@ define dso_local void @forked_ptrs_two_forks_gep(ptr nocapture readonly %Base1,
; RECURSE-NEXT: Group [[GRP81]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1243,6 +1256,7 @@ define void @forked_ptrs_two_select(ptr nocapture readonly %Base1, ptr nocapture
; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%loop>
; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1259,6 +1273,7 @@ define void @forked_ptrs_two_select(ptr nocapture readonly %Base1, ptr nocapture
; RECURSE-NEXT: (Low: %Dest High: (400 + %Dest))
; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%loop>
; RECURSE-NEXT: Member: {%Dest,+,4}<nuw><%loop>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1309,6 +1324,7 @@ define void @forked_ptrs_too_many_gep_ops(ptr nocapture readonly %Base1, ptr noc
; CHECK-NEXT: Group [[GRP85]]:
; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1332,6 +1348,7 @@ define void @forked_ptrs_too_many_gep_ops(ptr nocapture readonly %Base1, ptr noc
; RECURSE-NEXT: Group [[GRP87]]:
; RECURSE-NEXT: (Low: %Preds High: (400 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1379,6 +1396,7 @@ define void @forked_ptrs_vector_gep(ptr nocapture readonly %Base1, ptr nocapture
; CHECK-NEXT: Group [[GRP89]]:
; CHECK-NEXT: (Low: %Preds High: (388 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,16}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1402,6 +1420,7 @@ define void @forked_ptrs_vector_gep(ptr nocapture readonly %Base1, ptr nocapture
; RECURSE-NEXT: Group [[GRP91]]:
; RECURSE-NEXT: (Low: %Preds High: (388 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,16}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
@@ -1650,6 +1669,7 @@ define void @forked_ptrs_with_different_base3(ptr nocapture readonly %Preds, ptr
; CHECK-NEXT: Group [[GRP104:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %Preds High: (31996 + %Preds))
; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -1665,6 +1685,7 @@ define void @forked_ptrs_with_different_base3(ptr nocapture readonly %Preds, ptr
; RECURSE-NEXT: Group [[GRP105:0x[0-9a-f]+]]:
; RECURSE-NEXT: (Low: %Preds High: (31996 + %Preds))
; RECURSE-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
+; RECURSE-NEXT: Generated run-time checks are incomplete
; RECURSE-EMPTY:
; RECURSE-NEXT: Non vectorizable stores to invariant address were not found in loop.
; RECURSE-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
index 1e696af9e2e4b..7b97465a9f844 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
@@ -240,6 +240,7 @@ define void @different_non_constant_strides_known_backward_via_assume(ptr %A, i6
; CHECK-NEXT: Group [[GRP1:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (1024 + %A))
; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -277,6 +278,7 @@ define void @different_non_constant_strides_known_backward_via_assume_distance_l
; CHECK-NEXT: Group [[GRP2:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (1024 + %A)<nuw>)
; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -315,6 +317,7 @@ define void @different_non_constant_strides_known_backward_via_assume_min_distan
; CHECK-NEXT: Group [[GRP3:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (1024 + %A))
; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -353,6 +356,7 @@ define void @different_non_constant_strides_not_known_backward(ptr %A, i64 %scal
; CHECK-NEXT: Group [[GRP4:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (1024 + %A))
; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
index e33d20e795646..1f3fcea69a636 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
@@ -88,6 +88,7 @@ define void @different_non_constant_strides_known_forward_via_assume(ptr %A, i64
; CHECK-NEXT: Group [[GRP1:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (1024 + %A))
; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -125,6 +126,7 @@ define void @different_non_constant_strides_known_forward_via_assume_min_distanc
; CHECK-NEXT: Group [[GRP2:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (1024 + %A))
; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -163,6 +165,7 @@ define void @different_non_constant_strides_not_known_forward(ptr %A, i64 %scale
; CHECK-NEXT: Group [[GRP3:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (1024 + %A))
; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll b/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll
index ab40a22a3274d..8293d4edda1c8 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll
@@ -95,6 +95,7 @@ define void @int_and_multiple_pointer_predicates(ptr %v, ptr %w, i32 %N) {
; CHECK-NEXT: Group [[GRP5]]:
; CHECK-NEXT: (Low: %w High: (6 + (4 * (trunc i32 %N to i16)) + %w))
; CHECK-NEXT: Member: {%w,+,4}<%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
index 0f0635913c891..e4fdb43bafb91 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
@@ -287,6 +287,7 @@ define i32 @store_with_pointer_phi_incoming_phi(ptr %A, ptr %B, ptr %C, i1 %c.0,
; CHECK-NEXT: (Low: %A High: (256000 + %A))
; CHECK-NEXT: Member: {%A,+,8}<nuw><%loop.header>
; CHECK-NEXT: Member: %A
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -370,6 +371,7 @@ define i32 @store_with_pointer_phi_incoming_phi_irreducible_cycle(ptr %A, ptr %B
; CHECK-NEXT: (Low: %A High: (256000 + %A))
; CHECK-NEXT: Member: {%A,+,8}<nuw><%loop.header>
; CHECK-NEXT: Member: %A
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -465,6 +467,7 @@ define i32 @store_with_pointer_phi_in_same_bb_use_other_phi(ptr %A, ptr %B, ptr
; CHECK-NEXT: Group [[GRP10:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: %A High: (256000 + %A))
; CHECK-NEXT: Member: {%A,+,8}<nuw><%loop.header>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -539,6 +542,7 @@ define void @phi_load_store_memdep_check(i1 %c, ptr %A, ptr %B, ptr %C) {
; CHECK-NEXT: (Low: %B High: (2 + %B))
; CHECK-NEXT: Member: %B
; CHECK-NEXT: Member: %B
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll b/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
index 35233de1cce73..8ea6bf7677005 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
@@ -174,6 +174,7 @@ define void @dependency_check_and_runtime_checks_needed_select_of_ptr_add_recs_m
; CHECK-NEXT: (Low: %a High: ((4 * %n) + %a))
; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -222,6 +223,7 @@ define void @dependency_check_and_runtime_checks_needed_select_of_ptr_add_recs_m
; CHECK-NEXT: (Low: %a High: ((4 * %n) + %a))
; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
; CHECK-NEXT: Member: {%a,+,4}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll b/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
index 9e25baf31107d..e4549338653e0 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
@@ -83,6 +83,7 @@ define void @indirect_ptr_recurrences_read_write_may_alias_no_tbaa(ptr %A, ptr %
; CHECK-NEXT: Group [[GRP1:0x[0-9a-f]+]]:
; CHECK-NEXT: (Low: (8 + %B)<nuw> High: (40 + %B))
; CHECK-NEXT: Member: {(8 + %B)<nuw>,+,8}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -126,6 +127,7 @@ define void @indirect_ptr_recurrences_read_write_may_alias_different_obj(ptr %A,
; CHECK-NEXT: Group [[GRP3]]:
; CHECK-NEXT: (Low: (8 + %B)<nuw> High: (40 + %B))
; CHECK-NEXT: Member: {(8 + %B)<nuw>,+,8}<nuw><%loop>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll b/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll
index d0b101079efbf..dd52cb059ce02 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll
@@ -40,6 +40,7 @@ define void @rtchecks_needed(ptr %a, ptr %b, ptr %c) {
; CHECK-NEXT: Group [[GRP3]]:
; CHECK-NEXT: (Low: %c High: (40 + %c))
; CHECK-NEXT: Member: {%c,+,2}<nuw><%for.body>
+; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
>From d78421a2efd0bbc94b40122f217a3c00f2686d8f Mon Sep 17 00:00:00 2001
From: John Brawn <john.brawn at arm.com>
Date: Tue, 20 May 2025 17:37:21 +0100
Subject: [PATCH 3/6] Add HasCompletePtrRtChecking
This records whether canCheckPtrAtRT succeeded, and is used to decide
if we print the "checks are incomplete" message.
---
llvm/include/llvm/Analysis/LoopAccessAnalysis.h | 4 +++-
llvm/lib/Analysis/LoopAccessAnalysis.cpp | 10 +++++-----
.../Analysis/LoopAccessAnalysis/nusw-predicates.ll | 1 -
llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll | 3 ---
.../unsafe-and-rt-checks-convergent.ll | 1 -
5 files changed, 8 insertions(+), 11 deletions(-)
diff --git a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
index 5b58ba20a7bb4..8acbed4f5e499 100644
--- a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
+++ b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
@@ -781,7 +781,8 @@ class LoopAccessInfo {
/// We need to check that all of the pointers in this list are disjoint
/// at runtime. Using std::unique_ptr to make using move ctor simpler.
/// This list may contain only partial information when we've failed to
- /// analyze all the memory accesses in the loop (i.e. CanVecMem is false).
+ /// analyze all the memory accesses in the loop, in which case
+ /// HasCompletePtrRtChecking will be false.
std::unique_ptr<RuntimePointerChecking> PtrRtChecking;
/// The Memory Dependence Checker which can determine the
@@ -798,6 +799,7 @@ class LoopAccessInfo {
/// Cache the result of analyzeLoop.
bool CanVecMem = false;
bool HasConvergentOp = false;
+ bool HasCompletePtrRtChecking = false;
/// Indicator that there are two non vectorizable stores to the same uniform
/// address.
diff --git a/llvm/lib/Analysis/LoopAccessAnalysis.cpp b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
index 5436bc0f0cd8e..83ac060771636 100644
--- a/llvm/lib/Analysis/LoopAccessAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
@@ -2640,9 +2640,9 @@ bool LoopAccessInfo::analyzeLoop(AAResults *AA, const LoopInfo *LI,
// Find pointers with computable bounds. We are going to use this information
// to place a runtime bound check.
Value *UncomputablePtr = nullptr;
- bool CanDoRTIfNeeded = Accesses.canCheckPtrAtRT(
+ HasCompletePtrRtChecking = Accesses.canCheckPtrAtRT(
*PtrRtChecking, PSE->getSE(), TheLoop, SymbolicStrides, UncomputablePtr);
- if (!CanDoRTIfNeeded) {
+ if (!HasCompletePtrRtChecking) {
const auto *I = dyn_cast_or_null<Instruction>(UncomputablePtr);
recordAnalysis("CantIdentifyArrayBounds", I)
<< "cannot identify array bounds";
@@ -2671,11 +2671,11 @@ bool LoopAccessInfo::analyzeLoop(AAResults *AA, const LoopInfo *LI,
auto *SE = PSE->getSE();
UncomputablePtr = nullptr;
- CanDoRTIfNeeded = Accesses.canCheckPtrAtRT(
+ HasCompletePtrRtChecking = Accesses.canCheckPtrAtRT(
*PtrRtChecking, SE, TheLoop, SymbolicStrides, UncomputablePtr);
// Check that we found the bounds for the pointer.
- if (!CanDoRTIfNeeded) {
+ if (!HasCompletePtrRtChecking) {
auto *I = dyn_cast_or_null<Instruction>(UncomputablePtr);
recordAnalysis("CantCheckMemDepsAtRunTime", I)
<< "cannot check memory dependencies at runtime";
@@ -3009,7 +3009,7 @@ void LoopAccessInfo::print(raw_ostream &OS, unsigned Depth) const {
// List the pair of accesses need run-time checks to prove independence.
PtrRtChecking->print(OS, Depth);
- if (PtrRtChecking->Need && !CanVecMem)
+ if (PtrRtChecking->Need && !HasCompletePtrRtChecking)
OS.indent(Depth) << "Generated run-time checks are incomplete\n";
OS << "\n";
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll b/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll
index 8293d4edda1c8..ab40a22a3274d 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/nusw-predicates.ll
@@ -95,7 +95,6 @@ define void @int_and_multiple_pointer_predicates(ptr %v, ptr %w, i32 %N) {
; CHECK-NEXT: Group [[GRP5]]:
; CHECK-NEXT: (Low: %w High: (6 + (4 * (trunc i32 %N to i16)) + %w))
; CHECK-NEXT: Member: {%w,+,4}<%loop>
-; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
index e4fdb43bafb91..6479021005eaa 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
@@ -287,7 +287,6 @@ define i32 @store_with_pointer_phi_incoming_phi(ptr %A, ptr %B, ptr %C, i1 %c.0,
; CHECK-NEXT: (Low: %A High: (256000 + %A))
; CHECK-NEXT: Member: {%A,+,8}<nuw><%loop.header>
; CHECK-NEXT: Member: %A
-; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -371,7 +370,6 @@ define i32 @store_with_pointer_phi_incoming_phi_irreducible_cycle(ptr %A, ptr %B
; CHECK-NEXT: (Low: %A High: (256000 + %A))
; CHECK-NEXT: Member: {%A,+,8}<nuw><%loop.header>
; CHECK-NEXT: Member: %A
-; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -542,7 +540,6 @@ define void @phi_load_store_memdep_check(i1 %c, ptr %A, ptr %B, ptr %C) {
; CHECK-NEXT: (Low: %B High: (2 + %B))
; CHECK-NEXT: Member: %B
; CHECK-NEXT: Member: %B
-; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were found in loop.
; CHECK-NEXT: SCEV assumptions:
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll b/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll
index dd52cb059ce02..d0b101079efbf 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/unsafe-and-rt-checks-convergent.ll
@@ -40,7 +40,6 @@ define void @rtchecks_needed(ptr %a, ptr %b, ptr %c) {
; CHECK-NEXT: Group [[GRP3]]:
; CHECK-NEXT: (Low: %c High: (40 + %c))
; CHECK-NEXT: Member: {%c,+,2}<nuw><%for.body>
-; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
>From a77c6f94eb8c5f93763268ae2c371ad87297c11e Mon Sep 17 00:00:00 2001
From: John Brawn <john.brawn at arm.com>
Date: Wed, 21 May 2025 17:20:36 +0100
Subject: [PATCH 4/6] Add an option to control whether partial results are
allowed
---
.../llvm/Analysis/LoopAccessAnalysis.h | 21 ++--
.../Scalar/LoopAccessAnalysisPrinter.h | 4 +-
llvm/lib/Analysis/LoopAccessAnalysis.cpp | 35 ++++---
llvm/lib/Passes/PassRegistry.def | 11 ++-
.../Scalar/LoopAccessAnalysisPrinter.cpp | 2 +-
.../Transforms/Scalar/LoopVersioningLICM.cpp | 2 +-
.../LoopAccessAnalysis/allow-partial.ll | 99 +++++++++++++++++++
.../LoopAccessAnalysis/forked-pointers.ll | 4 +-
.../non-constant-strides-backward.ll | 2 +-
.../non-constant-strides-forward.ll | 2 +-
.../LoopAccessAnalysis/pointer-phis.ll | 2 +-
...ter-dependence-analysis-forked-pointers.ll | 2 +-
.../underlying-object-loop-varying-phi.ll | 2 +-
...-access-analysis-common-output.ll.expected | 24 -----
14 files changed, 158 insertions(+), 54 deletions(-)
create mode 100644 llvm/test/Analysis/LoopAccessAnalysis/allow-partial.ll
diff --git a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
index 1097953ab5bee..26b4bb607dfbe 100644
--- a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
+++ b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
@@ -655,7 +655,8 @@ class RuntimePointerChecking {
/// For memory dependences that cannot be determined at compile time, it
/// generates run-time checks to prove independence. This is done by
/// AccessAnalysis::canCheckPtrAtRT and the checks are maintained by the
-/// RuntimePointerCheck class.
+/// RuntimePointerCheck class. AllowPartial determines whether partial checks
+/// are generated when not all pointers could be analyzed.
///
/// If pointers can wrap or can't be expressed as affine AddRec expressions by
/// ScalarEvolution, we will generate run-time checks by emitting a
@@ -668,7 +669,8 @@ class LoopAccessInfo {
LLVM_ABI LoopAccessInfo(Loop *L, ScalarEvolution *SE,
const TargetTransformInfo *TTI,
const TargetLibraryInfo *TLI, AAResults *AA,
- DominatorTree *DT, LoopInfo *LI);
+ DominatorTree *DT, LoopInfo *LI,
+ bool AllowPartial = false);
/// Return true we can analyze the memory accesses in the loop and there are
/// no memory dependence cycles. Note that for dependences between loads &
@@ -683,6 +685,11 @@ class LoopAccessInfo {
/// not legal to insert them.
bool hasConvergentOp() const { return HasConvergentOp; }
+ /// Return true if, when runtime pointer checking does not have complete
+ /// results, it instead has partial results for those memory accesses that
+ /// could be analyzed.
+ bool hasAllowPartial() const { return AllowPartial; }
+
const RuntimePointerChecking *getRuntimePointerChecking() const {
return PtrRtChecking.get();
}
@@ -785,9 +792,9 @@ class LoopAccessInfo {
/// We need to check that all of the pointers in this list are disjoint
/// at runtime. Using std::unique_ptr to make using move ctor simpler.
- /// This list may contain only partial information when we've failed to
- /// analyze all the memory accesses in the loop, in which case
- /// HasCompletePtrRtChecking will be false.
+ /// If AllowPartial is true then this list may contain only partial
+ /// information when we've failed to analyze all the memory accesses in the
+ /// loop, in which case HasCompletePtrRtChecking will be false.
std::unique_ptr<RuntimePointerChecking> PtrRtChecking;
/// The Memory Dependence Checker which can determine the
@@ -798,6 +805,8 @@ class LoopAccessInfo {
Loop *TheLoop;
+ bool AllowPartial;
+
unsigned NumLoads = 0;
unsigned NumStores = 0;
@@ -927,7 +936,7 @@ class LoopAccessInfoManager {
const TargetLibraryInfo *TLI)
: SE(SE), AA(AA), DT(DT), LI(LI), TTI(TTI), TLI(TLI) {}
- LLVM_ABI const LoopAccessInfo &getInfo(Loop &L);
+ LLVM_ABI const LoopAccessInfo &getInfo(Loop &L, bool AllowPartial = false);
LLVM_ABI void clear();
diff --git a/llvm/include/llvm/Transforms/Scalar/LoopAccessAnalysisPrinter.h b/llvm/include/llvm/Transforms/Scalar/LoopAccessAnalysisPrinter.h
index f445e0696b5f8..44ba3c24f6afb 100644
--- a/llvm/include/llvm/Transforms/Scalar/LoopAccessAnalysisPrinter.h
+++ b/llvm/include/llvm/Transforms/Scalar/LoopAccessAnalysisPrinter.h
@@ -20,9 +20,11 @@ class raw_ostream;
class LoopAccessInfoPrinterPass
: public PassInfoMixin<LoopAccessInfoPrinterPass> {
raw_ostream &OS;
+ bool AllowPartial;
public:
- explicit LoopAccessInfoPrinterPass(raw_ostream &OS) : OS(OS) {}
+ explicit LoopAccessInfoPrinterPass(raw_ostream &OS, bool AllowPartial)
+ : OS(OS), AllowPartial(AllowPartial) {}
PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
static bool isRequired() { return true; }
};
diff --git a/llvm/lib/Analysis/LoopAccessAnalysis.cpp b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
index 3c50a06100ba2..ac1a8844d6cd5 100644
--- a/llvm/lib/Analysis/LoopAccessAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
@@ -700,7 +700,7 @@ class AccessAnalysis {
/// loop, but we will have checks for those pointers we could analyze.
bool canCheckPtrAtRT(RuntimePointerChecking &RtCheck, Loop *TheLoop,
const DenseMap<Value *, const SCEV *> &Strides,
- Value *&UncomputablePtr);
+ Value *&UncomputablePtr, bool AllowPartial);
/// Goes over all memory accesses, checks whether a RT check is needed
/// and builds sets of dependent accesses.
@@ -1185,8 +1185,8 @@ bool AccessAnalysis::createCheckForAccess(
bool AccessAnalysis::canCheckPtrAtRT(
RuntimePointerChecking &RtCheck, Loop *TheLoop,
- const DenseMap<Value *, const SCEV *> &StridesMap,
- Value *&UncomputablePtr) {
+ const DenseMap<Value *, const SCEV *> &StridesMap, Value *&UncomputablePtr,
+ bool AllowPartial) {
// Find pointers with computable bounds. We are going to use this information
// to place a runtime bound check.
bool CanDoRT = true;
@@ -1279,6 +1279,8 @@ bool AccessAnalysis::canCheckPtrAtRT(
/*Assume=*/true)) {
CanDoAliasSetRT = false;
UncomputablePtr = Access.getPointer();
+ if (!AllowPartial)
+ break;
}
}
}
@@ -1318,7 +1320,7 @@ bool AccessAnalysis::canCheckPtrAtRT(
}
}
- if (MayNeedRTCheck)
+ if (MayNeedRTCheck && (CanDoRT || AllowPartial))
RtCheck.generateChecks(DepCands, IsDepCheckNeeded);
LLVM_DEBUG(dbgs() << "LAA: We need to do " << RtCheck.getNumberOfChecks()
@@ -1331,6 +1333,8 @@ bool AccessAnalysis::canCheckPtrAtRT(
bool CanDoRTIfNeeded = !RtCheck.Need || CanDoRT;
assert(CanDoRTIfNeeded == (CanDoRT || !MayNeedRTCheck) &&
"CanDoRTIfNeeded depends on RtCheck.Need");
+ if (!CanDoRTIfNeeded && !AllowPartial)
+ RtCheck.reset();
return CanDoRTIfNeeded;
}
@@ -2600,7 +2604,7 @@ bool LoopAccessInfo::analyzeLoop(AAResults *AA, const LoopInfo *LI,
// to place a runtime bound check.
Value *UncomputablePtr = nullptr;
HasCompletePtrRtChecking = Accesses.canCheckPtrAtRT(
- *PtrRtChecking, TheLoop, SymbolicStrides, UncomputablePtr);
+ *PtrRtChecking, TheLoop, SymbolicStrides, UncomputablePtr, AllowPartial);
if (!HasCompletePtrRtChecking) {
const auto *I = dyn_cast_or_null<Instruction>(UncomputablePtr);
recordAnalysis("CantIdentifyArrayBounds", I)
@@ -2629,8 +2633,9 @@ bool LoopAccessInfo::analyzeLoop(AAResults *AA, const LoopInfo *LI,
PtrRtChecking->Need = true;
UncomputablePtr = nullptr;
- HasCompletePtrRtChecking = Accesses.canCheckPtrAtRT(
- *PtrRtChecking, TheLoop, SymbolicStrides, UncomputablePtr);
+ HasCompletePtrRtChecking =
+ Accesses.canCheckPtrAtRT(*PtrRtChecking, TheLoop, SymbolicStrides,
+ UncomputablePtr, AllowPartial);
// Check that we found the bounds for the pointer.
if (!HasCompletePtrRtChecking) {
@@ -2908,9 +2913,10 @@ void LoopAccessInfo::collectStridedAccess(Value *MemAccess) {
LoopAccessInfo::LoopAccessInfo(Loop *L, ScalarEvolution *SE,
const TargetTransformInfo *TTI,
const TargetLibraryInfo *TLI, AAResults *AA,
- DominatorTree *DT, LoopInfo *LI)
+ DominatorTree *DT, LoopInfo *LI,
+ bool AllowPartial)
: PSE(std::make_unique<PredicatedScalarEvolution>(*SE, *L)),
- PtrRtChecking(nullptr), TheLoop(L) {
+ PtrRtChecking(nullptr), TheLoop(L), AllowPartial(AllowPartial) {
unsigned MaxTargetVectorWidthInBits = std::numeric_limits<unsigned>::max();
if (TTI && !TTI->enableScalableVectorization())
// Scale the vector width by 2 as rough estimate to also consider
@@ -2980,12 +2986,15 @@ void LoopAccessInfo::print(raw_ostream &OS, unsigned Depth) const {
PSE->print(OS, Depth);
}
-const LoopAccessInfo &LoopAccessInfoManager::getInfo(Loop &L) {
+const LoopAccessInfo &LoopAccessInfoManager::getInfo(Loop &L,
+ bool AllowPartial) {
const auto &[It, Inserted] = LoopAccessInfoMap.try_emplace(&L);
- if (Inserted)
- It->second =
- std::make_unique<LoopAccessInfo>(&L, &SE, TTI, TLI, &AA, &DT, &LI);
+ // We need to create the LoopAccessInfo if either we don't already have one,
+ // or if it was created with a different value of AllowPartial.
+ if (Inserted || It->second->hasAllowPartial() != AllowPartial)
+ It->second = std::make_unique<LoopAccessInfo>(&L, &SE, TTI, TLI, &AA, &DT,
+ &LI, AllowPartial);
return *It->second;
}
diff --git a/llvm/lib/Passes/PassRegistry.def b/llvm/lib/Passes/PassRegistry.def
index be93a7b1f5ba6..736b0f9c50bca 100644
--- a/llvm/lib/Passes/PassRegistry.def
+++ b/llvm/lib/Passes/PassRegistry.def
@@ -441,7 +441,6 @@ FUNCTION_PASS("print-cfg-sccs", CFGSCCPrinterPass(errs()))
FUNCTION_PASS("print-memderefs", MemDerefPrinterPass(errs()))
FUNCTION_PASS("print-mustexecute", MustExecutePrinterPass(errs()))
FUNCTION_PASS("print-predicateinfo", PredicateInfoPrinterPass(errs()))
-FUNCTION_PASS("print<access-info>", LoopAccessInfoPrinterPass(errs()))
FUNCTION_PASS("print<assumptions>", AssumptionPrinterPass(errs()))
FUNCTION_PASS("print<block-freq>", BlockFrequencyPrinterPass(errs()))
FUNCTION_PASS("print<branch-prob>", BranchProbabilityPrinterPass(errs()))
@@ -583,6 +582,16 @@ FUNCTION_PASS_WITH_PARAMS(
return MergedLoadStoreMotionPass(Opts);
},
parseMergedLoadStoreMotionOptions, "no-split-footer-bb;split-footer-bb")
+FUNCTION_PASS_WITH_PARAMS(
+ "print<access-info>", "LoopAccessInfoPrinterPass",
+ [](bool AllowPartial) {
+ return LoopAccessInfoPrinterPass(errs(), AllowPartial);
+ },
+ [](StringRef Params) {
+ return PassBuilder::parseSinglePassOption(Params, "allow-partial",
+ "LoopAccessInfoPrinterPass");
+ },
+ "allow-partial")
FUNCTION_PASS_WITH_PARAMS(
"print<da>", "DependenceAnalysisPrinterPass",
[](bool NormalizeResults) {
diff --git a/llvm/lib/Transforms/Scalar/LoopAccessAnalysisPrinter.cpp b/llvm/lib/Transforms/Scalar/LoopAccessAnalysisPrinter.cpp
index 3d3f22d686e32..7253c7f222509 100644
--- a/llvm/lib/Transforms/Scalar/LoopAccessAnalysisPrinter.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopAccessAnalysisPrinter.cpp
@@ -28,7 +28,7 @@ PreservedAnalyses LoopAccessInfoPrinterPass::run(Function &F,
while (!Worklist.empty()) {
Loop *L = Worklist.pop_back_val();
OS.indent(2) << L->getHeader()->getName() << ":\n";
- LAIs.getInfo(*L).print(OS, 4);
+ LAIs.getInfo(*L, AllowPartial).print(OS, 4);
}
return PreservedAnalyses::all();
}
diff --git a/llvm/lib/Transforms/Scalar/LoopVersioningLICM.cpp b/llvm/lib/Transforms/Scalar/LoopVersioningLICM.cpp
index 6b4fc88cd3a72..4f2bfb073bafa 100644
--- a/llvm/lib/Transforms/Scalar/LoopVersioningLICM.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopVersioningLICM.cpp
@@ -368,7 +368,7 @@ bool LoopVersioningLICM::legalLoopInstructions() {
IsReadOnlyLoop = true;
using namespace ore;
// Get LoopAccessInfo from current loop via the proxy.
- LAI = &LAIs.getInfo(*CurLoop);
+ LAI = &LAIs.getInfo(*CurLoop, /*AllowPartial=*/true);
// Check LoopAccessInfo for need of runtime check.
if (LAI->getRuntimePointerChecking()->getChecks().empty()) {
LLVM_DEBUG(dbgs() << " LAA: Runtime check not found !!\n");
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/allow-partial.ll b/llvm/test/Analysis/LoopAccessAnalysis/allow-partial.ll
new file mode 100644
index 0000000000000..3096d545eb6d5
--- /dev/null
+++ b/llvm/test/Analysis/LoopAccessAnalysis/allow-partial.ll
@@ -0,0 +1,99 @@
+; RUN: opt -disable-output -passes='print<access-info><allow-partial>,print<access-info>' %s 2>&1 | FileCheck %s --check-prefixes=ALLOW-BEFORE
+; RUN: opt -disable-output -passes='print<access-info>,print<access-info><allow-partial>' %s 2>&1 | FileCheck %s --check-prefixes=ALLOW-AFTER
+
+; Check that we get the right results when loop access analysis is run twice,
+; once without partial results and once with.
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32"
+
+define void @gep_loaded_offset(ptr %p, ptr %q, ptr %r, i32 %n) {
+; ALLOW-BEFORE-LABEL: 'gep_loaded_offset'
+; ALLOW-BEFORE-NEXT: while.body:
+; ALLOW-BEFORE-NEXT: Report: cannot identify array bounds
+; ALLOW-BEFORE-NEXT: Dependences:
+; ALLOW-BEFORE-NEXT: Run-time memory checks:
+; ALLOW-BEFORE-NEXT: Check 0:
+; ALLOW-BEFORE-NEXT: Comparing group GRP0:
+; ALLOW-BEFORE-NEXT: %p.addr = phi ptr [ %incdec.ptr, %while.body ], [ %p, %entry ]
+; ALLOW-BEFORE-NEXT: Against group GRP1:
+; ALLOW-BEFORE-NEXT: ptr %r
+; ALLOW-BEFORE-NEXT: Grouped accesses:
+; ALLOW-BEFORE-NEXT: Group GRP0:
+; ALLOW-BEFORE-NEXT: (Low: %p High: (4 + (4 * (zext i32 (-1 + %n)<nsw> to i64))<nuw><nsw> + %p))
+; ALLOW-BEFORE-NEXT: Member: {%p,+,4}<nuw><%while.body>
+; ALLOW-BEFORE-NEXT: Group GRP1:
+; ALLOW-BEFORE-NEXT: (Low: %r High: (8 + %r))
+; ALLOW-BEFORE-NEXT: Member: %r
+; ALLOW-BEFORE-NEXT: Generated run-time checks are incomplete
+; ALLOW-BEFORE-EMPTY:
+; ALLOW-BEFORE-NEXT: Non vectorizable stores to invariant address were not found in loop.
+; ALLOW-BEFORE-NEXT: SCEV assumptions:
+; ALLOW-BEFORE-EMPTY:
+; ALLOW-BEFORE-NEXT: Expressions re-written:
+;
+; ALLOW-BEFORE-LABEL: 'gep_loaded_offset'
+; ALLOW-BEFORE-NEXT: while.body:
+; ALLOW-BEFORE-NEXT: Report: cannot identify array bounds
+; ALLOW-BEFORE-NEXT: Dependences:
+; ALLOW-BEFORE-NEXT: Run-time memory checks:
+; ALLOW-BEFORE-NEXT: Grouped accesses:
+; ALLOW-BEFORE-EMPTY:
+; ALLOW-BEFORE-NEXT: Non vectorizable stores to invariant address were not found in loop.
+; ALLOW-BEFORE-NEXT: SCEV assumptions:
+; ALLOW-BEFORE-EMPTY:
+; ALLOW-BEFORE-NEXT: Expressions re-written:
+;
+; ALLOW-AFTER-LABEL: 'gep_loaded_offset'
+; ALLOW-AFTER-NEXT: while.body:
+; ALLOW-AFTER-NEXT: Report: cannot identify array bounds
+; ALLOW-AFTER-NEXT: Dependences:
+; ALLOW-AFTER-NEXT: Run-time memory checks:
+; ALLOW-AFTER-NEXT: Grouped accesses:
+; ALLOW-AFTER-EMPTY:
+; ALLOW-AFTER-NEXT: Non vectorizable stores to invariant address were not found in loop.
+; ALLOW-AFTER-NEXT: SCEV assumptions:
+; ALLOW-AFTER-EMPTY:
+; ALLOW-AFTER-NEXT: Expressions re-written:
+;
+; ALLOW-AFTER-LABEL: 'gep_loaded_offset'
+; ALLOW-AFTER-NEXT: while.body:
+; ALLOW-AFTER-NEXT: Report: cannot identify array bounds
+; ALLOW-AFTER-NEXT: Dependences:
+; ALLOW-AFTER-NEXT: Run-time memory checks:
+; ALLOW-AFTER-NEXT: Check 0:
+; ALLOW-AFTER-NEXT: Comparing group GRP0:
+; ALLOW-AFTER-NEXT: %p.addr = phi ptr [ %incdec.ptr, %while.body ], [ %p, %entry ]
+; ALLOW-AFTER-NEXT: Against group GRP1:
+; ALLOW-AFTER-NEXT: ptr %r
+; ALLOW-AFTER-NEXT: Grouped accesses:
+; ALLOW-AFTER-NEXT: Group GRP0:
+; ALLOW-AFTER-NEXT: (Low: %p High: (4 + (4 * (zext i32 (-1 + %n)<nsw> to i64))<nuw><nsw> + %p))
+; ALLOW-AFTER-NEXT: Member: {%p,+,4}<nuw><%while.body>
+; ALLOW-AFTER-NEXT: Group GRP1:
+; ALLOW-AFTER-NEXT: (Low: %r High: (8 + %r))
+; ALLOW-AFTER-NEXT: Member: %r
+; ALLOW-AFTER-NEXT: Generated run-time checks are incomplete
+; ALLOW-AFTER-EMPTY:
+; ALLOW-AFTER-NEXT: Non vectorizable stores to invariant address were not found in loop.
+; ALLOW-AFTER-NEXT: SCEV assumptions:
+; ALLOW-AFTER-EMPTY:
+; ALLOW-AFTER-NEXT: Expressions re-written:
+;
+entry:
+ br label %while.body
+
+while.body:
+ %n.addr = phi i32 [ %dec, %while.body ], [ %n, %entry ]
+ %p.addr = phi ptr [ %incdec.ptr, %while.body ], [ %p, %entry ]
+ %dec = add nsw i32 %n.addr, -1
+ %rval = load i64, ptr %r, align 4
+ %arrayidx = getelementptr inbounds i32, ptr %q, i64 %rval
+ %val = load i32, ptr %arrayidx, align 4
+ %incdec.ptr = getelementptr inbounds nuw i8, ptr %p.addr, i64 4
+ store i32 %val, ptr %p.addr, align 4
+ %tobool.not = icmp eq i32 %dec, 0
+ br i1 %tobool.not, label %while.end, label %while.body
+
+while.end:
+ ret void
+}
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll b/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
index 4b8a89b92c23f..fa72ae910cbbb 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/forked-pointers.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 3
-; RUN: opt -disable-output -passes='print<access-info>' %s 2>&1 | FileCheck %s --check-prefixes=CHECK,FULLDEPTH
-; RUN: opt -disable-output -passes='print<access-info>' -max-forked-scev-depth=2 %s 2>&1 | FileCheck %s --check-prefixes=CHECK,DEPTH2
+; RUN: opt -disable-output -passes='print<access-info><allow-partial>' %s 2>&1 | FileCheck %s --check-prefixes=CHECK,FULLDEPTH
+; RUN: opt -disable-output -passes='print<access-info><allow-partial>' -max-forked-scev-depth=2 %s 2>&1 | FileCheck %s --check-prefixes=CHECK,DEPTH2
target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
index 369abec6c2f25..3160a77e0ba5e 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-backward.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes='print<access-info>' -disable-output %s 2>&1 | FileCheck %s
+; RUN: opt -passes='print<access-info><allow-partial>' -disable-output %s 2>&1 | FileCheck %s
target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
index 016cbf6d59832..9e0a1e3a03dde 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/non-constant-strides-forward.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes='print<access-info>' -disable-output %s 2>&1 | FileCheck %s
+; RUN: opt -passes='print<access-info><allow-partial>' -disable-output %s 2>&1 | FileCheck %s
target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
index 9605c07ed4cc9..6fbe0e45976b6 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/pointer-phis.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 3
-; RUN: opt -passes='print<access-info>' -disable-output < %s 2>&1 | FileCheck %s
+; RUN: opt -passes='print<access-info><allow-partial>' -disable-output < %s 2>&1 | FileCheck %s
%s1 = type { [32000 x double], [32000 x double], [32000 x double] }
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll b/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
index d5b3729584ad3..d1d1ecb2af888 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/retry-runtime-checks-after-dependence-analysis-forked-pointers.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 5
-; RUN: opt -passes='print<access-info>' -disable-output %s 2>&1 | FileCheck %s
+; RUN: opt -passes='print<access-info><allow-partial>' -disable-output %s 2>&1 | FileCheck %s
define void @dependency_check_and_runtime_checks_needed_select_of_invariant_ptrs(ptr %a, ptr %b, ptr %c, i64 %offset, i64 %n) {
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll b/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
index de5cb8dd12596..0708f908211ef 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/underlying-object-loop-varying-phi.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes='print<access-info>' -disable-output %s 2>&1 | FileCheck %s
+; RUN: opt -passes='print<access-info><allow-partial>' -disable-output %s 2>&1 | FileCheck %s
target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
diff --git a/llvm/test/tools/UpdateTestChecks/update_analyze_test_checks/Inputs/loop-access-analysis-common-output.ll.expected b/llvm/test/tools/UpdateTestChecks/update_analyze_test_checks/Inputs/loop-access-analysis-common-output.ll.expected
index bca7a4331ff00..f2703e856ca22 100644
--- a/llvm/test/tools/UpdateTestChecks/update_analyze_test_checks/Inputs/loop-access-analysis-common-output.ll.expected
+++ b/llvm/test/tools/UpdateTestChecks/update_analyze_test_checks/Inputs/loop-access-analysis-common-output.ll.expected
@@ -8,19 +8,7 @@ define void @laa_common_output(ptr %Base, ptr %Dest, ptr %Preds) {
; CHECK-NEXT: Report: cannot identify array bounds
; CHECK-NEXT: Dependences:
; CHECK-NEXT: Run-time memory checks:
-; CHECK-NEXT: Check 0:
-; CHECK-NEXT: Comparing group GRP0:
-; CHECK-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; CHECK-NEXT: Against group GRP1:
-; CHECK-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; CHECK-NEXT: Grouped accesses:
-; CHECK-NEXT: Group GRP0:
-; CHECK-NEXT: (Low: %Dest High: (400 + %Dest))
-; CHECK-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
-; CHECK-NEXT: Group GRP1:
-; CHECK-NEXT: (Low: %Preds High: (400 + %Preds))
-; CHECK-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
-; CHECK-NEXT: Generated run-time checks are incomplete
; CHECK-EMPTY:
; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
; CHECK-NEXT: SCEV assumptions:
@@ -91,19 +79,7 @@ define void @laa_divergent_output(ptr %Base, ptr %Dest, ptr %Preds, i64 %extra_o
; DEPTH2-NEXT: Report: cannot identify array bounds
; DEPTH2-NEXT: Dependences:
; DEPTH2-NEXT: Run-time memory checks:
-; DEPTH2-NEXT: Check 0:
-; DEPTH2-NEXT: Comparing group GRP0:
-; DEPTH2-NEXT: %arrayidx5 = getelementptr inbounds float, ptr %Dest, i64 %indvars.iv
-; DEPTH2-NEXT: Against group GRP1:
-; DEPTH2-NEXT: %arrayidx = getelementptr inbounds i32, ptr %Preds, i64 %indvars.iv
; DEPTH2-NEXT: Grouped accesses:
-; DEPTH2-NEXT: Group GRP0:
-; DEPTH2-NEXT: (Low: %Dest High: (400 + %Dest))
-; DEPTH2-NEXT: Member: {%Dest,+,4}<nuw><%for.body>
-; DEPTH2-NEXT: Group GRP1:
-; DEPTH2-NEXT: (Low: %Preds High: (400 + %Preds))
-; DEPTH2-NEXT: Member: {%Preds,+,4}<nuw><%for.body>
-; DEPTH2-NEXT: Generated run-time checks are incomplete
; DEPTH2-EMPTY:
; DEPTH2-NEXT: Non vectorizable stores to invariant address were not found in loop.
; DEPTH2-NEXT: SCEV assumptions:
>From 32dcdac9d0ec19a71c86581c3b7e316ad7aa4d1b Mon Sep 17 00:00:00 2001
From: John Brawn <john.brawn at arm.com>
Date: Wed, 4 Jun 2025 11:20:20 +0100
Subject: [PATCH 5/6] Adjust based on review comments.
---
llvm/include/llvm/Analysis/LoopAccessAnalysis.h | 2 +-
llvm/lib/Analysis/LoopAccessAnalysis.cpp | 4 +++-
2 files changed, 4 insertions(+), 2 deletions(-)
diff --git a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
index 26b4bb607dfbe..ba0bf3f860967 100644
--- a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
+++ b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
@@ -655,7 +655,7 @@ class RuntimePointerChecking {
/// For memory dependences that cannot be determined at compile time, it
/// generates run-time checks to prove independence. This is done by
/// AccessAnalysis::canCheckPtrAtRT and the checks are maintained by the
-/// RuntimePointerCheck class. AllowPartial determines whether partial checks
+/// RuntimePointerCheck class. \p AllowPartial determines whether partial checks
/// are generated when not all pointers could be analyzed.
///
/// If pointers can wrap or can't be expressed as affine AddRec expressions by
diff --git a/llvm/lib/Analysis/LoopAccessAnalysis.cpp b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
index ac1a8844d6cd5..24661d72149d2 100644
--- a/llvm/lib/Analysis/LoopAccessAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
@@ -697,7 +697,8 @@ class AccessAnalysis {
/// Returns true if we need no check or if we do and we can generate them
/// (i.e. the pointers have computable bounds). A return value of false means
/// we couldn't analyze and generate runtime checks for all pointers in the
- /// loop, but we will have checks for those pointers we could analyze.
+ /// loop, but if \p AllowPartial is set then we will have checks for those
+ /// pointers we could analyze.
bool canCheckPtrAtRT(RuntimePointerChecking &RtCheck, Loop *TheLoop,
const DenseMap<Value *, const SCEV *> &Strides,
Value *&UncomputablePtr, bool AllowPartial);
@@ -1330,6 +1331,7 @@ bool AccessAnalysis::canCheckPtrAtRT(
// are needed. This can happen when all pointers point to the same underlying
// object for example.
RtCheck.Need = CanDoRT ? RtCheck.getNumberOfChecks() != 0 : MayNeedRTCheck;
+
bool CanDoRTIfNeeded = !RtCheck.Need || CanDoRT;
assert(CanDoRTIfNeeded == (CanDoRT || !MayNeedRTCheck) &&
"CanDoRTIfNeeded depends on RtCheck.Need");
>From 2b02a5bf6671d1feef49c49c947b7083e98e393b Mon Sep 17 00:00:00 2001
From: John Brawn <john.brawn at arm.com>
Date: Wed, 4 Jun 2025 15:56:09 +0100
Subject: [PATCH 6/6] Add an extra comment
---
llvm/include/llvm/Analysis/LoopAccessAnalysis.h | 2 ++
1 file changed, 2 insertions(+)
diff --git a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
index ba0bf3f860967..2a44ec8032236 100644
--- a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
+++ b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
@@ -805,6 +805,8 @@ class LoopAccessInfo {
Loop *TheLoop;
+ /// Determines whether we should generate partial runtime checks when not all
+ /// memory accesses could be analyzed.
bool AllowPartial;
unsigned NumLoads = 0;
More information about the llvm-commits
mailing list