By default, use Early-CSE instead of GVN for vectorization cleanup.
As has been suggested by Duncan and others, Early-CSE and GVN should
do similar redundancy elimination, but Early-CSE is much less expensive.
Most of my autovectorization benchmarks show a performance regresion, but
all of these are < 0.1%, and so I think that it is still worth using
the less expensive pass.
llvm-svn: 154673
diff --git a/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp b/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
index a1b0a45..43b4ab5 100644
--- a/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
+++ b/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
@@ -35,6 +35,11 @@
static cl::opt<bool>
RunVectorization("vectorize", cl::desc("Run vectorization passes"));
+static cl::opt<bool>
+UseGVNAfterVectorization("use-gvn-after-vectorization",
+ cl::init(false), cl::Hidden,
+ cl::desc("Run GVN instead of Early CSE after vectorization passes"));
+
PassManagerBuilder::PassManagerBuilder() {
OptLevel = 2;
SizeLevel = 0;
@@ -182,8 +187,10 @@
if (Vectorize) {
MPM.add(createBBVectorizePass());
MPM.add(createInstructionCombiningPass());
- if (OptLevel > 1)
- MPM.add(createGVNPass()); // Remove redundancies
+ if (OptLevel > 1 && UseGVNAfterVectorization)
+ MPM.add(createGVNPass()); // Remove redundancies
+ else
+ MPM.add(createEarlyCSEPass()); // Catch trivial redundancies
}
MPM.add(createAggressiveDCEPass()); // Delete dead instructions