[PATCH] D53245: Teach the DominatorTree fallback to recalculation when applying updates to speedup JT (PR37929)
Chijun Sima via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu Oct 25 08:53:20 PDT 2018
NutshellySima updated this revision to Diff 171112.
NutshellySima set the repository for this revision to rL LLVM.
NutshellySima added a comment.
1. Adjust the constant to be `40`.
2. Disable recalculation when `n<=100` && `k<=n` making the unittest of the incremental algorithm works as before.
Repository:
rL LLVM
https://reviews.llvm.org/D53245
Files:
include/llvm/Support/GenericDomTreeConstruction.h
Index: include/llvm/Support/GenericDomTreeConstruction.h
===================================================================
--- include/llvm/Support/GenericDomTreeConstruction.h
+++ include/llvm/Support/GenericDomTreeConstruction.h
@@ -1191,6 +1191,20 @@
});
LLVM_DEBUG(dbgs() << "\n");
+ // Recalculate the DominatorTree when the number of updates
+ // exceeds a threshold, which usually makes direct updating slower than
+ // recalculation. We select this threshold proportional to the
+ // size of the DominatorTree. The constant is selected
+ // by choosing the one with an acceptable performance on some real-world
+ // inputs.
+
+ // Make unittests of the incremental algorithm work
+ if (DT.DomTreeNodes.size() <= 100) {
+ if (NumLegalized > DT.DomTreeNodes.size())
+ CalculateFromScratch(DT, &BUI);
+ } else if (NumLegalized > DT.DomTreeNodes.size() / 40)
+ CalculateFromScratch(DT, &BUI);
+
// If the DominatorTree was recalculated at some point, stop the batch
// updates. Full recalculations ignore batch updates and look at the actual
// CFG.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D53245.171112.patch
Type: text/x-patch
Size: 1127 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20181025/3846b975/attachment.bin>
More information about the llvm-commits
mailing list