[llvm] df3b95b - [SLP][NFC] PR45269 getVectorElementSize() is slow
Dinar Temirbulatov via llvm-commits
llvm-commits at lists.llvm.org
Thu May 21 08:27:26 PDT 2020
Author: Dinar Temirbulatov
Date: 2020-05-21T17:26:50+02:00
New Revision: df3b95bc0ae0f9de0a54db897ed37ffb6ba44226
URL: https://github.com/llvm/llvm-project/commit/df3b95bc0ae0f9de0a54db897ed37ffb6ba44226
DIFF: https://github.com/llvm/llvm-project/commit/df3b95bc0ae0f9de0a54db897ed37ffb6ba44226.diff
LOG: [SLP][NFC] PR45269 getVectorElementSize() is slow
The algorithm inside getVectorElementSize() is almost O(x^2) complexity and
when, for example, we compile MultiSource/Applications/ClamAV/shared_sha256.c
with 1k instructions inside sha256_transform() function that resulted in almost
~800k iterations. The following change improves the algorithm with the map to
a liner complexity.
Differential Revision: https://reviews.llvm.org/D80241
Added:
Modified:
llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index d800e0b11d55..26d9705fade5 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -629,7 +629,7 @@ class BoUpSLP {
/// the stored value. Otherwise, the size is the width of the largest loaded
/// value reaching V. This method is used by the vectorizer to calculate
/// vectorization factors.
- unsigned getVectorElementSize(Value *V) const;
+ unsigned getVectorElementSize(Value *V);
/// Compute the minimum type sizes required to represent the entries in a
/// vectorizable tree.
@@ -1715,6 +1715,9 @@ class BoUpSLP {
/// Maps a specific scalar to its tree entry.
SmallDenseMap<Value*, TreeEntry *> ScalarToTreeEntry;
+ /// Maps a value to the proposed vectorizable size.
+ SmallDenseMap<Value *, unsigned> InstrElementSize;
+
/// A list of scalars that we found that we need to keep as scalars.
ValueSet MustGather;
@@ -4785,6 +4788,7 @@ BoUpSLP::vectorizeTree(ExtraValueToDebugLocsMap &ExternallyUsedValues) {
}
Builder.ClearInsertionPoint();
+ InstrElementSize.clear();
return VectorizableTree[0]->VectorizedValue;
}
@@ -5321,12 +5325,16 @@ void BoUpSLP::scheduleBlock(BlockScheduling *BS) {
BS->ScheduleStart = nullptr;
}
-unsigned BoUpSLP::getVectorElementSize(Value *V) const {
+unsigned BoUpSLP::getVectorElementSize(Value *V) {
// If V is a store, just return the width of the stored value without
// traversing the expression tree. This is the common case.
if (auto *Store = dyn_cast<StoreInst>(V))
return DL->getTypeSizeInBits(Store->getValueOperand()->getType());
+ auto E = InstrElementSize.find(V);
+ if (E != InstrElementSize.end())
+ return E->second;
+
// If V is not a store, we can traverse the expression tree to find loads
// that feed it. The type of the loaded value may indicate a more suitable
// width than V's type. We want to base the vector element size on the width
@@ -5372,13 +5380,17 @@ unsigned BoUpSLP::getVectorElementSize(Value *V) const {
FoundUnknownInst = true;
}
+ int Width = MaxWidth;
// If we didn't encounter a memory access in the expression tree, or if we
- // gave up for some reason, just return the width of V.
+ // gave up for some reason, just return the width of V. Otherwise, return the
+ // maximum width we found.
if (!MaxWidth || FoundUnknownInst)
- return DL->getTypeSizeInBits(V->getType());
+ Width = DL->getTypeSizeInBits(V->getType());
+
+ for (Instruction *I : Visited)
+ InstrElementSize[I] = Width;
- // Otherwise, return the maximum width we found.
- return MaxWidth;
+ return Width;
}
// Determine if a value V in a vectorizable expression Expr can be demoted to a
More information about the llvm-commits
mailing list