[DependenceAnalysis] Dependecies for loads marked with "ivnariant.load" should not be shared with general accesses(PR42151).

[ebrevnov]

Summary

This is second attempt to fix the problem with incorrect dependencies reported in presence of invariant load. Initial fix (https://reviews.llvm.org/D64405) was reverted due to a regression reported in https://reviews.llvm.org/D70516.

The original fix changed caching behavior for invariant loads. Namely such loads are not put into the second level cache (NonLocalDepInfo). The problem with that fix is the first level cache (CachedNonLocalPointerInfo) still works as if invariant loads were in the second level cache. The solution is in addition to not putting dependence results into the second level cache avoid putting info about invariant loads into the first level cache as well.

[ebrevnov]

Added regression test.

[jdoerfert]

Just to make sure, this contains a test for the regression that was reported and lead to the initial revert, correct?

llvm/lib/Analysis/MemoryDependenceAnalysis.cpp

1395

Why is the foundBlock stuff gone?

[ebrevnov]

In D73027#1832510, @jdoerfert wrote:

Just to make sure, this contains a test for the regression that was reported and lead to the initial revert, correct?

Yes. It's @test4.

llvm/lib/Analysis/MemoryDependenceAnalysis.cpp

1395

This is change was done in the previous version of the patch. The reason for that was invariant loads which are not expected to be found in the cache. Now this whole cycle is guarded with !isInvariantLoad and we can restore previous code.

[jdoerfert]

llvm/lib/Analysis/MemoryDependenceAnalysis.cpp

1395

Let's do that then.

[ebrevnov]

Updated as agreed.

[ebrevnov]

ping

[ebrevnov]

ping @jdoerfert

[jdoerfert]

LGTM.

llvm/lib/Analysis/MemoryDependenceAnalysis.cpp

 /// cached information in Cache or by doing a lookup (which may use dirty cache
 /// info if available).
 ///
 /// If we do a lookup, add the result to the cache.
 MemDepResult MemoryDependenceResults::GetNonLocalInfoForBlock(
     Instruction *QueryInst, const MemoryLocation &Loc, bool isLoad,
     BasicBlock *BB, NonLocalDepInfo *Cache, unsigned NumSortedEntries) {
 
+  bool isInvariantLoad = false;
+
+  if (LoadInst *LI = dyn_cast_or_null<LoadInst>(QueryInst))
+    isInvariantLoad = LI->getMetadata(LLVMContext::MD_invariant_load);
+
   // Do a binary search to see if we already have an entry for this block in
   // the cache set.  If so, find it.
   NonLocalDepInfo::iterator Entry = std::upper_bound(
       Cache->begin(), Cache->begin() + NumSortedEntries, NonLocalDepEntry(BB));
   if (Entry != Cache->begin() && (Entry - 1)->getBB() == BB)
     --Entry;
 
   NonLocalDepEntry *ExistingResult = nullptr;
   if (Entry != Cache->begin() + NumSortedEntries && Entry->getBB() == BB)
     ExistingResult = &*Entry;
 
+  // Use cached result for invariant load only if there is no dependency for non
+  // invariant load. In this case invariant load can not have any dependency as
+  // well.
+  if (ExistingResult && isInvariantLoad &&
+      !ExistingResult->getResult().isNonFuncLocal())
+    ExistingResult = nullptr;
+
   // If we have a cached entry, and it is non-dirty, use it as the value for
   // this dependency.
   if (ExistingResult && !ExistingResult->getResult().isDirty()) {
     ++NumCacheNonLocalPtr;
     return ExistingResult->getResult();
   }
 
   // Otherwise, we have to scan for the value.  If we have a dirty cache
MemDepResult MemoryDependenceResults::GetNonLocalInfoForBlock(
   } else {
     ++NumUncacheNonLocalPtr;
   }
 
   // Scan the block for the dependency.
   MemDepResult Dep =
       getPointerDependencyFrom(Loc, isLoad, ScanPos, BB, QueryInst);
 
+  // Don't cache results for invariant load.
+  if (isInvariantLoad)
+    return Dep;
+
   // If we had a dirty entry for the block, update it.  Otherwise, just add
   // a new entry.
   if (ExistingResult)
     ExistingResult->setResult(Dep);
   else
     Cache->push_back(NonLocalDepEntry(BB, Dep));
 
   // If the block has a dependency (i.e. it isn't completely transparent to
bool MemoryDependenceResults::getNonLocalPointerDepFromBB(
   // Set up a temporary NLPI value. If the map doesn't yet have an entry for
   // CacheKey, this value will be inserted as the associated value. Otherwise,
   // it'll be ignored, and we'll have to check to see if the cached size and
   // aa tags are consistent with the current query.
   NonLocalPointerInfo InitialNLPI;
   InitialNLPI.Size = Loc.Size;
   InitialNLPI.AATags = Loc.AATags;
 
+  bool isInvariantLoad = false;
+  if (LoadInst *LI = dyn_cast_or_null<LoadInst>(QueryInst))
+    isInvariantLoad = LI->getMetadata(LLVMContext::MD_invariant_load);
+
   // Get the NLPI for CacheKey, inserting one into the map if it doesn't
   // already have one.
   std::pair<CachedNonLocalPointerInfo::iterator, bool> Pair =
       NonLocalPointerDeps.insert(std::make_pair(CacheKey, InitialNLPI));
   NonLocalPointerInfo *CacheInfo = &Pair.first->second;
 
   // If we already have a cache entry for this CacheKey, we may need to do some
   // work to reconcile the cache entry and the current query.
-  if (!Pair.second) {
+  // Invariant loads don't participate in caching. Thus no need to reconcile.
+  if (!isInvariantLoad && !Pair.second) {
     if (CacheInfo->Size != Loc.Size) {
       bool ThrowOutEverything;
       if (CacheInfo->Size.hasValue() && Loc.Size.hasValue()) {
         // FIXME: We may be able to do better in the face of results with mixed
         // precision. We don't appear to get them in practice, though, so just
         // be conservative.
         ThrowOutEverything =
             CacheInfo->Size.isPrecise() != Loc.Size.isPrecise() ||
if (CacheInfo->AATags != Loc.AATags) {
             Visited, SkipFirstBlock);
     }
   }
 
   NonLocalDepInfo *Cache = &CacheInfo->NonLocalDeps;
 
   // If we have valid cached information for exactly the block we are
   // investigating, just return it with no recomputation.
-  if (CacheInfo->Pair == BBSkipFirstBlockPair(StartBB, SkipFirstBlock)) {
+  // Don't use cached information for invariant loads since it is valid for
+  // non-invariant loads only.
+  if (!isInvariantLoad &&
+      CacheInfo->Pair == BBSkipFirstBlockPair(StartBB, SkipFirstBlock)) {
     // We have a fully cached result for this query then we can just return the
     // cached results and populate the visited set.  However, we have to verify
     // that we don't already have conflicting results for these blocks.  Check
     // to ensure that if a block in the results set is in the visited set that
     // it was for the same pointer query.
     if (!Visited.empty()) {
       for (auto &Entry : *Cache) {
         DenseMap<BasicBlock *, Value *>::iterator VI =
for (auto &Entry : *Cache) {
         Result.push_back(
             NonLocalDepResult(Entry.getBB(), Entry.getResult(), Addr));
       }
     }
     ++NumCacheCompleteNonLocalPtr;
     return true;
   }
 
+  // Invariant loads don't affect cache in any way thus no need to update
+  // CacheInfo as well.
+  if (!isInvariantLoad) {
     // Otherwise, either this is a new block, a block with an invalid cache
-  // pointer or one that we're about to invalidate by putting more info into it
-  // than its valid cache info.  If empty, the result will be valid cache info,
-  // otherwise it isn't.
+    // pointer or one that we're about to invalidate by putting more info into
+    // it than its valid cache info.  If empty, the result will be valid cache
+    // info, otherwise it isn't.
     if (Cache->empty())
       CacheInfo->Pair = BBSkipFirstBlockPair(StartBB, SkipFirstBlock);
     else
       CacheInfo->Pair = BBSkipFirstBlockPair();
+  }
 
   SmallVector<BasicBlock *, 32> Worklist;
   Worklist.push_back(StartBB);
 
   // PredList used inside loop.
   SmallVector<std::pair<BasicBlock *, PHITransAddr>, 16> PredList;
 
   // Keep track of the entries that we know are sorted.  Previously cached
PredTranslationFailure:
     // If *nothing* works, mark the pointer as unknown.
     //
     // If this is the magic first block, return this as a clobber of the whole
     // incoming value.  Since we can't phi translate to one of the predecessors,
     // we have to bail out.
     if (SkipFirstBlock)
       return false;
 
+    // Results of invariant loads are not cached thus no need to update cached
+    // information.
+    if (!isInvariantLoad) {
       bool foundBlock = false;
       for (NonLocalDepEntry &I : llvm::reverse(*Cache)) {
         if (I.getBB() != BB)
           continue;
 
         assert((GotWorklistLimit || I.getResult().isNonLocal() ||
                 !DT.isReachableFromEntry(BB)) &&
                "Should only be here with transparent block");
         foundBlock = true;
         I.setResult(MemDepResult::getUnknown());
         Result.push_back(
             NonLocalDepResult(I.getBB(), I.getResult(), Pointer.getAddr()));
         break;
       }
       (void)foundBlock; (void)GotWorklistLimit;
       assert((foundBlock || GotWorklistLimit) && "Current block not in cache?");
     }
+  }
 
   // Okay, we're done now.  If we added new values to the cache, re-sort it.
   SortNonLocalDepInfoCache(*Cache, NumSortedEntries);
   LLVM_DEBUG(AssertSorted(*Cache));
   return true;
 }
 
 /// If P exists in CachedNonLocalPointerInfo or NonLocalDefsCache, remove it.

llvm/test/Analysis/MemoryDependenceAnalysis/InvariantLoad.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -gvn -S | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128-ni:1"
+target triple = "x86_64-unknown-linux-gnu"
+
+declare void @llvm.memset.p0i8.i8(i8*, i8, i32, i1)
+declare void @foo(i8*)
+
+define i8 @test(i1 %cmp) {
+; CHECK-LABEL: @test(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[P:%.*]] = alloca i8
+; CHECK-NEXT:    store i8 5, i8* [[P]]
+; CHECK-NEXT:    br label [[HEADER:%.*]]
+; CHECK:       header:
+; CHECK-NEXT:    [[V:%.*]] = phi i8 [ 5, [[ENTRY:%.*]] ], [ -5, [[ALIVE:%.*]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i8 [ 0, [[ENTRY]] ], [ [[I_INC:%.*]], [[ALIVE]] ]
+; CHECK-NEXT:    br i1 [[CMP:%.*]], label [[ALIVE]], label [[DEAD:%.*]]
+; CHECK:       dead:
+; CHECK-NEXT:    call void @foo(i8* [[P]])
+; CHECK-NEXT:    [[I_1:%.*]] = add i8 [[I]], [[V]]
+; CHECK-NEXT:    br label [[ALIVE]]
+; CHECK:       alive:
+; CHECK-NEXT:    [[I_2:%.*]] = phi i8 [ [[I]], [[HEADER]] ], [ [[I_1]], [[DEAD]] ]
+; CHECK-NEXT:    store i8 -5, i8* [[P]]
+; CHECK-NEXT:    call void @llvm.memset.p0i8.i32(i8* align 1 [[P]], i8 0, i32 1, i1 false)
+; CHECK-NEXT:    [[I_INC]] = add i8 [[I_2]], 1
+; CHECK-NEXT:    [[CMP_LOOP:%.*]] = icmp ugt i8 [[I_INC]], 100
+; CHECK-NEXT:    br i1 [[CMP_LOOP]], label [[EXIT:%.*]], label [[HEADER]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret i8 0
+;
+
+entry:
+  %p = alloca i8
+  %addr = getelementptr inbounds i8, i8* %p, i64 0
+  store i8 5, i8* %addr
+  br label %header
+header:
+  %i = phi i8 [0, %entry], [%i.inc, %backedge]
+  br i1 %cmp, label %alive, label %dead
+dead:
+  call void @foo(i8* %p)
+  %v = load i8, i8* %addr, !invariant.load !1
+  %i.1 = add i8 %i, %v
+  br label %alive
+alive:
+  %i.2 = phi i8 [%i, %header], [%i.1, %dead]
+  store i8 -5, i8* %addr
+  br label %backedge
+backedge:
+  call void @llvm.memset.p0i8.i8(i8 * align 1 %p, i8 0, i32 1, i1 false)
+  %i.inc = add i8 %i.2, 1
+  %cmp.loop = icmp ugt i8 %i.inc, 100
+  br i1 %cmp.loop, label %exit, label %header
+exit:
+  %res = load i8, i8* %addr
+  ret i8 %res
+}
+
+; Check that first two loads are not optimized out while the one marked with
+; invariant.load reuses %res1
+define i8 @test2(i1 %cmp, i8 *%p) {
+; CHECK-LABEL: @test2(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[RES1:%.*]] = load i8, i8* [[P:%.*]]
+; CHECK-NEXT:    call void @foo(i8* [[P]])
+; CHECK-NEXT:    br i1 [[CMP:%.*]], label [[B2:%.*]], label [[B1:%.*]]
+; CHECK:       b1:
+; CHECK-NEXT:    [[RES2:%.*]] = load i8, i8* [[P]]
+; CHECK-NEXT:    [[RES3:%.*]] = add i8 [[RES1]], [[RES2]]
+; CHECK-NEXT:    br label [[ALIVE:%.*]]
+; CHECK:       b2:
+; CHECK-NEXT:    [[RES_DEAD:%.*]] = add i8 [[RES1]], [[RES1]]
+; CHECK-NEXT:    br label [[ALIVE]]
+; CHECK:       alive:
+; CHECK-NEXT:    [[RES_PHI:%.*]] = phi i8 [ [[RES3]], [[B1]] ], [ [[RES_DEAD]], [[B2]] ]
+; CHECK-NEXT:    ret i8 [[RES_PHI]]
+;
+
+entry:
+  %res1 = load i8, i8* %p
+  call void @foo(i8 *%p)
+  br i1 %cmp, label %b2, label %b1
+b1:
+  %res2 = load i8, i8* %p
+  %res3 = add i8 %res1, %res2
+  br label %alive
+b2:
+  %v = load i8, i8* %p, !invariant.load !1
+  %res.dead = add i8 %v, %res1
+  br label %alive
+alive:
+  %res.phi = phi i8 [%res3, %b1], [%res.dead, %b2]
+  ret i8 %res.phi
+}
+
+; This is essentially the same test case as the above one but with %b1 and %b2
+; swapped in "br i1 %cmp, label %b1, label %b2" instruction. That helps us to
+; ensure that results doesn't depend on visiting order.
+define i8 @test3(i1 %cmp, i8 *%p) {
+; CHECK-LABEL: @test3(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[RES1:%.*]] = load i8, i8* [[P:%.*]]
+; CHECK-NEXT:    call void @foo(i8* [[P]])
+; CHECK-NEXT:    br i1 [[CMP:%.*]], label [[B1:%.*]], label [[B2:%.*]]
+; CHECK:       b1:
+; CHECK-NEXT:    [[RES2:%.*]] = load i8, i8* [[P]]
+; CHECK-NEXT:    [[RES3:%.*]] = add i8 [[RES1]], [[RES2]]
+; CHECK-NEXT:    br label [[ALIVE:%.*]]
+; CHECK:       b2:
+; CHECK-NEXT:    [[RES_DEAD:%.*]] = add i8 [[RES1]], [[RES1]]
+; CHECK-NEXT:    br label [[ALIVE]]
+; CHECK:       alive:
+; CHECK-NEXT:    [[RES_PHI:%.*]] = phi i8 [ [[RES3]], [[B1]] ], [ [[RES_DEAD]], [[B2]] ]
+; CHECK-NEXT:    ret i8 [[RES_PHI]]
+;
+entry:
+  %res1 = load i8, i8* %p
+  call void @foo(i8 *%p)
+  br i1 %cmp, label %b1, label %b2
+b1:
+  %res2 = load i8, i8* %p
+  %res3 = add i8 %res1, %res2
+  br label %alive
+b2:
+  %v = load i8, i8* %p, !invariant.load !1
+  %res.dead = add i8 %v, %res1
+  br label %alive
+alive:
+  %res.phi = phi i8 [%res3, %b1], [%res.dead, %b2]
+  ret i8 %res.phi
+}
+
+
+; This is reduced test case catching regression in the first version of the
+; fix for invariant loads (https://reviews.llvm.org/D64405).
+define void @test4() {
+; CHECK-LABEL: @test4(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* inttoptr (i64 8 to float*), align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul float [[TMP0]], [[TMP0]]
+; CHECK-NEXT:    br label [[FUSION_LOOP_HEADER_DIM_1_PREHEADER:%.*]]
+; CHECK:       fusion.loop_header.dim.1.preheader:
+; CHECK-NEXT:    [[TMP2:%.*]] = phi float [ [[TMP0]], [[ENTRY:%.*]] ], [ [[DOTPRE:%.*]], [[FUSION_LOOP_HEADER_DIM_1_PREHEADER]] ]
+; CHECK-NEXT:    [[FUSION_INVAR_ADDRESS_DIM_0_03:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INVAR_INC3:%.*]], [[FUSION_LOOP_HEADER_DIM_1_PREHEADER]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [2 x [1 x [4 x float]]], [2 x [1 x [4 x float]]]* null, i64 0, i64 [[FUSION_INVAR_ADDRESS_DIM_0_03]], i64 0, i64 2
+; CHECK-NEXT:    [[TMP4:%.*]] = fmul float [[TMP2]], [[TMP2]]
+; CHECK-NEXT:    [[INVAR_INC3]] = add nuw nsw i64 [[FUSION_INVAR_ADDRESS_DIM_0_03]], 1
+; CHECK-NEXT:    [[DOTPHI_TRANS_INSERT:%.*]] = getelementptr inbounds [2 x [1 x [4 x float]]], [2 x [1 x [4 x float]]]* null, i64 0, i64 [[INVAR_INC3]], i64 0, i64 2
+; CHECK-NEXT:    [[DOTPRE]] = load float, float* [[DOTPHI_TRANS_INSERT]], align 4, !invariant.load !0
+; CHECK-NEXT:    br label [[FUSION_LOOP_HEADER_DIM_1_PREHEADER]]
+;
+entry:
+  %0 = getelementptr inbounds [2 x [1 x [4 x float]]], [2 x [1 x [4 x float]]]* null, i64 0, i64 0, i64 0, i64 2
+  %1 = load float, float* %0, align 4
+  %2 = fmul float %1, %1
+  br label %fusion.loop_header.dim.1.preheader
+
+fusion.loop_header.dim.1.preheader:               ; preds = %fusion.loop_header.dim.1.preheader, %entry
+  %fusion.invar_address.dim.0.03 = phi i64 [ 0, %entry ], [ %invar.inc3, %fusion.loop_header.dim.1.preheader ]
+  %3 = getelementptr inbounds [2 x [1 x [4 x float]]], [2 x [1 x [4 x float]]]* null, i64 0, i64 %fusion.invar_address.dim.0.03, i64 0, i64 2
+  %4 = load float, float* %3, align 4, !invariant.load !1
+  %5 = fmul float %4, %4
+  %6 = getelementptr inbounds [2 x [1 x [4 x float]]], [2 x [1 x [4 x float]]]* null, i64 0, i64 %fusion.invar_address.dim.0.03, i64 0, i64 2
+  %7 = load float, float* %6, align 4, !invariant.load !1
+  %8 = fmul float %7, %7
+  %invar.inc3 = add nuw nsw i64 %fusion.invar_address.dim.0.03, 1
+  br label %fusion.loop_header.dim.1.preheader
+}
+
+!1 = !{}