[LoopAccesses] Cache the result of canVectorizeMemory
LAA will be an on-demand analysis pass, so we need to cache the result
of the analysis. canVectorizeMemory is renamed to analyzeLoop which
computes the result. canVectorizeMemory becomes the query function for
the cached result.
This is part of the patchset that converts LoopAccessAnalysis into an
actual analysis pass.
llvm-svn: 229624
diff --git a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
index 93ee959b..0cdde4c 100644
--- a/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
+++ b/llvm/include/llvm/Analysis/LoopAccessAnalysis.h
@@ -134,11 +134,14 @@
const TargetLibraryInfo *TLI, AliasAnalysis *AA,
DominatorTree *DT) :
TheLoop(L), SE(SE), DL(DL), TLI(TLI), AA(AA), DT(DT), NumLoads(0),
- NumStores(0), MaxSafeDepDistBytes(-1U) {}
+ NumStores(0), MaxSafeDepDistBytes(-1U), CanVecMem(false) {}
+
+ /// \brief Analyze the loop. Replaces symbolic strides using Strides.
+ void analyzeLoop(ValueToValueMap &Strides);
/// Return true we can analyze the memory accesses in the loop and there are
- /// no memory dependence cycles. Replaces symbolic strides using Strides.
- bool canVectorizeMemory(ValueToValueMap &Strides);
+ /// no memory dependence cycles.
+ bool canVectorizeMemory() { return CanVecMem; }
RuntimePointerCheck *getRuntimePointerCheck() { return &PtrRtCheck; }
@@ -182,6 +185,9 @@
unsigned MaxSafeDepDistBytes;
+ /// \brief Cache the result of analyzeLoop.
+ bool CanVecMem;
+
/// \brief The diagnostics report generated for the analysis. E.g. why we
/// couldn't analyze the loop.
Optional<VectorizationReport> Report;
diff --git a/llvm/lib/Analysis/LoopAccessAnalysis.cpp b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
index 927ae49..6ebb0ef 100644
--- a/llvm/lib/Analysis/LoopAccessAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopAccessAnalysis.cpp
@@ -812,7 +812,7 @@
return true;
}
-bool LoopAccessInfo::canVectorizeMemory(ValueToValueMap &Strides) {
+void LoopAccessInfo::analyzeLoop(ValueToValueMap &Strides) {
typedef SmallVector<Value*, 16> ValueVector;
typedef SmallPtrSet<Value*, 16> ValueSet;
@@ -855,7 +855,8 @@
emitAnalysis(VectorizationReport(Ld)
<< "read with atomic ordering or volatile read");
DEBUG(dbgs() << "LV: Found a non-simple load.\n");
- return false;
+ CanVecMem = false;
+ return;
}
NumLoads++;
Loads.push_back(Ld);
@@ -869,13 +870,15 @@
if (!St) {
emitAnalysis(VectorizationReport(it) <<
"instruction cannot be vectorized");
- return false;
+ CanVecMem = false;
+ return;
}
if (!St->isSimple() && !IsAnnotatedParallel) {
emitAnalysis(VectorizationReport(St)
<< "write with atomic ordering or volatile write");
DEBUG(dbgs() << "LV: Found a non-simple store.\n");
- return false;
+ CanVecMem = false;
+ return;
}
NumStores++;
Stores.push_back(St);
@@ -891,7 +894,8 @@
// care if the pointers are *restrict*.
if (!Stores.size()) {
DEBUG(dbgs() << "LV: Found a read-only loop!\n");
- return true;
+ CanVecMem = true;
+ return;
}
AccessAnalysis::DepCandidates DependentAccesses;
@@ -914,7 +918,8 @@
VectorizationReport(ST)
<< "write to a loop invariant address could not be vectorized");
DEBUG(dbgs() << "LV: We don't allow storing to uniform addresses\n");
- return false;
+ CanVecMem = false;
+ return;
}
// If we did *not* see this pointer before, insert it to the read-write
@@ -937,7 +942,8 @@
DEBUG(dbgs()
<< "LV: A loop annotated parallel, ignore memory dependency "
<< "checks.\n");
- return true;
+ CanVecMem = true;
+ return;
}
for (I = Loads.begin(), IE = Loads.end(); I != IE; ++I) {
@@ -972,7 +978,8 @@
// other reads in this loop then is it safe to vectorize.
if (NumReadWrites == 1 && NumReads == 0) {
DEBUG(dbgs() << "LV: Found a write-only loop!\n");
- return true;
+ CanVecMem = true;
+ return;
}
// Build dependence sets and check whether we need a runtime pointer bounds
@@ -1013,12 +1020,13 @@
DEBUG(dbgs() << "LV: We can't vectorize because we can't find " <<
"the array bounds.\n");
PtrRtCheck.reset();
- return false;
+ CanVecMem = false;
+ return;
}
PtrRtCheck.Need = NeedRTCheck;
- bool CanVecMem = true;
+ CanVecMem = true;
if (Accesses.isDependencyCheckNeeded()) {
DEBUG(dbgs() << "LV: Checking memory dependencies\n");
CanVecMem = DepChecker.areDepsSafe(
@@ -1051,7 +1059,8 @@
<< " dependent memory operations checked at runtime");
DEBUG(dbgs() << "LV: Can't vectorize with memory checks\n");
PtrRtCheck.reset();
- return false;
+ CanVecMem = false;
+ return;
}
CanVecMem = true;
@@ -1064,8 +1073,6 @@
DEBUG(dbgs() << "LV: We" << (NeedRTCheck ? "" : " don't") <<
" need a runtime memory check.\n");
-
- return CanVecMem;
}
bool LoopAccessInfo::blockNeedsPredication(BasicBlock *BB) {
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index abb38d1..1def103d 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -3825,11 +3825,11 @@
}
bool LoopVectorizationLegality::canVectorizeMemory() {
- bool Success = LAI.canVectorizeMemory(Strides);
+ LAI.analyzeLoop(Strides);
auto &OptionalReport = LAI.getReport();
if (OptionalReport)
emitAnalysis(*OptionalReport);
- return Success;
+ return LAI.canVectorizeMemory();
}
static bool hasMultipleUsesOf(Instruction *I,