[llvm-commits] [llvm] r122685 - in /llvm/trunk: lib/Target/README.txt lib/Transforms/Scalar/LoopIdiomRecognize.cpp test/Transforms/LoopIdiom/basic.ll
Chris Lattner
sabre at nondot.org
Sat Jan 1 23:58:36 PST 2011
Author: lattner
Date: Sun Jan 2 01:58:36 2011
New Revision: 122685
URL: http://llvm.org/viewvc/llvm-project?rev=122685&view=rev
Log:
Allow loop-idiom to run on multiple BB loops, but still only scan the loop
header for now for memset/memcpy opportunities. It turns out that loop-rotate
is successfully rotating loops, but *DOESN'T MERGE THE BLOCKS*, turning "for
loops" into 2 basic block loops that loop-idiom was ignoring.
With this fix, we form many *many* more memcpy and memsets than before, including
on the "history" loops in the viterbi benchmark, which look like this:
for (j=0; j<MAX_history; ++j) {
history_new[i][j+1] = history[2*i][j];
}
Transforming these loops into memcpy's speeds up the viterbi benchmark from
11.98s to 3.55s on my machine. Woo.
Modified:
llvm/trunk/lib/Target/README.txt
llvm/trunk/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
llvm/trunk/test/Transforms/LoopIdiom/basic.ll
Modified: llvm/trunk/lib/Target/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/README.txt?rev=122685&r1=122684&r2=122685&view=diff
==============================================================================
--- llvm/trunk/lib/Target/README.txt (original)
+++ llvm/trunk/lib/Target/README.txt Sun Jan 2 01:58:36 2011
@@ -414,14 +414,6 @@
//===---------------------------------------------------------------------===//
-[LOOP RECOGNITION]
-
-viterbi speeds up *significantly* if the various "history" related copy loops
-are turned into memcpy calls at the source level. We need a "loops to memcpy"
-pass.
-
-//===---------------------------------------------------------------------===//
-
[LOOP OPTIMIZATION]
SingleSource/Benchmarks/Misc/dt.c shows several interesting optimization
Modified: llvm/trunk/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopIdiomRecognize.cpp?rev=122685&r1=122684&r2=122685&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopIdiomRecognize.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopIdiomRecognize.cpp Sun Jan 2 01:58:36 2011
@@ -126,11 +126,6 @@
bool LoopIdiomRecognize::runOnLoop(Loop *L, LPPassManager &LPM) {
CurLoop = L;
- // We only look at trivial single basic block loops.
- // TODO: eventually support more complex loops, scanning the header.
- if (L->getBlocks().size() != 1)
- return false;
-
// The trip count of the loop must be analyzable.
SE = &getAnalysis<ScalarEvolution>();
if (!SE->hasLoopInvariantBackedgeTakenCount(L))
@@ -142,6 +137,11 @@
TD = getAnalysisIfAvailable<TargetData>();
if (TD == 0) return false;
+ // TODO: We currently only scan the header of the loop, because it is the only
+ // part that is known to execute and we don't want to make a conditional store
+ // into an unconditional one in the preheader. However, there can be diamonds
+ // and other things in the loop that would make other blocks "always executed"
+ // we should get the full set and scan each block.
BasicBlock *BB = L->getHeader();
DEBUG(dbgs() << "loop-idiom Scanning: F[" << BB->getParent()->getName()
<< "] Loop %" << BB->getName() << "\n");
Modified: llvm/trunk/test/Transforms/LoopIdiom/basic.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopIdiom/basic.ll?rev=122685&r1=122684&r2=122685&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/LoopIdiom/basic.ll (original)
+++ llvm/trunk/test/Transforms/LoopIdiom/basic.ll Sun Jan 2 01:58:36 2011
@@ -21,6 +21,30 @@
; CHECK-NOT: store
}
+; This is a loop that was rotated but where the blocks weren't merged. This
+; shouldn't perturb us.
+define void @test1a(i8* %Base, i64 %Size) nounwind ssp {
+bb.nph: ; preds = %entry
+ br label %for.body
+
+for.body: ; preds = %bb.nph, %for.body
+ %indvar = phi i64 [ 0, %bb.nph ], [ %indvar.next, %for.body.cont ]
+ %I.0.014 = getelementptr i8* %Base, i64 %indvar
+ store i8 0, i8* %I.0.014, align 1
+ %indvar.next = add i64 %indvar, 1
+ br label %for.body.cont
+for.body.cont:
+ %exitcond = icmp eq i64 %indvar.next, %Size
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+; CHECK: @test1a
+; CHECK: call void @llvm.memset.p0i8.i64(i8* %Base, i8 0, i64 %Size, i32 1, i1 false)
+; CHECK-NOT: store
+}
+
+
define void @test2(i32* %Base, i64 %Size) nounwind ssp {
entry:
%cmp10 = icmp eq i64 %Size, 0
More information about the llvm-commits
mailing list