[llvm] f4e8f6d - [Reassociate] Use a reference to DataLayout instead of copying the underlying string data (NFC) (#128269)
via llvm-commits
llvm-commits at lists.llvm.org
Sat Feb 22 01:37:28 PST 2025
Author: cooperp
Date: 2025-02-22T10:37:24+01:00
New Revision: f4e8f6da41a5ca3e03808d86bce0bcde339b9414
URL: https://github.com/llvm/llvm-project/commit/f4e8f6da41a5ca3e03808d86bce0bcde339b9414
DIFF: https://github.com/llvm/llvm-project/commit/f4e8f6da41a5ca3e03808d86bce0bcde339b9414.diff
LOG: [Reassociate] Use a reference to DataLayout instead of copying the underlying string data (NFC) (#128269)
I noticed this when looking at all allocations by clang. For a medium
sized file this was around 6000 calls to operator new, although i
suspect there were more allocations in total as the SmallVectors in
DataLayout may have their own allocations in some cases.
In a follow-up i'm tempted to make the DataLayout copy constructor
private, to avoid this in future. There are a few tests which copy the
DataLayout, and perhaps need to (I didn't check yet), but we could
provide a clone() method for them if needed. Its only accidental copying
I think we should consider avoiding, not people who really do need to
copy it for reasons.
Added:
Modified:
llvm/lib/Transforms/Scalar/Reassociate.cpp
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/Scalar/Reassociate.cpp b/llvm/lib/Transforms/Scalar/Reassociate.cpp
index 7cb9bace47bf4..a7db9fd817883 100644
--- a/llvm/lib/Transforms/Scalar/Reassociate.cpp
+++ b/llvm/lib/Transforms/Scalar/Reassociate.cpp
@@ -420,7 +420,7 @@ static bool LinearizeExprTree(Instruction *I,
using LeafMap = DenseMap<Value *, uint64_t>;
LeafMap Leaves; // Leaf -> Total weight so far.
SmallVector<Value *, 8> LeafOrder; // Ensure deterministic leaf output order.
- const DataLayout DL = I->getDataLayout();
+ const DataLayout &DL = I->getDataLayout();
#ifndef NDEBUG
SmallPtrSet<Value *, 8> Visited; // For checking the iteration scheme.
More information about the llvm-commits
mailing list