From facca6e3f3a3b2988cb872834937cb204eb2a686 Mon Sep 17 00:00:00 2001 From: Andrew Trick Date: Thu, 22 May 2014 00:37:03 +0000 Subject: [PATCH] Fix a bug in SCEV's backedge taken count computation from my prior fix in Jan. This has to do with the trip count computation for loops with multiple exits, which is quite subtle. Most passes just ask for a single trip count number, so we must be conservative assuming any exit could be taken. Normally, we rely on the "exact" trip count, which was correctly given as "unknown". However, SCEV also gives a "max" back-edge taken count. The loops max BE taken count is conservatively a maximum over the max of each exit's non-exiting iterations count. Note that some exit tests can be skipped so the max loop back-edge taken count can actually exceed the max non-exiting iterations for some exits. However, when we know the loop *latch* cannot be skipped, we can directly use its max taken count disregarding other exits. I previously took the minimum here without checking whether the other exit could be skipped. The correct, and simpler thing to do here is just to directly use the loop latch's max non-exiting iterations as the loops max back-edge count. In the problematic test case, the first loop exit had a max of zero non-exiting iterations, but could be skipped. The loop latch was known not to be skipped but had max of one non-exiting iteration. We incorrectly claimed the loop back-edge could be taken zero times, when it is actually taken one time. Fixes Loop %for.body.i: Unpredictable backedge-taken count. Loop %for.body.i: max backedge-taken count is 1. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@209358 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Analysis/ScalarEvolution.cpp | 14 +++++----- .../ScalarEvolution/max-trip-count.ll | 26 +++++++++++++++++++ 2 files changed, 32 insertions(+), 8 deletions(-) diff --git a/lib/Analysis/ScalarEvolution.cpp b/lib/Analysis/ScalarEvolution.cpp index cef9966f9e8..dad8e07dadb 100644 --- a/lib/Analysis/ScalarEvolution.cpp +++ b/lib/Analysis/ScalarEvolution.cpp @@ -4413,7 +4413,7 @@ ScalarEvolution::ComputeBackedgeTakenCount(const Loop *L) { const SCEV *MaxBECount = getCouldNotCompute(); bool CouldComputeBECount = true; BasicBlock *Latch = L->getLoopLatch(); // may be NULL. - const SCEV *LatchMaxCount = nullptr; + bool LatchMustExit = false; SmallVector, 4> ExitCounts; for (unsigned i = 0, e = ExitingBlocks.size(); i != e; ++i) { ExitLimit EL = ComputeExitLimit(L, ExitingBlocks[i]); @@ -4431,16 +4431,14 @@ ScalarEvolution::ComputeBackedgeTakenCount(const Loop *L) { // skip some loop tests. Taking the max over the exits is sufficiently // conservative. TODO: We could do better taking into consideration // non-latch exits that dominate the latch. - if (EL.MustExit && ExitingBlocks[i] == Latch) - LatchMaxCount = EL.Max; - else + if (EL.MustExit && ExitingBlocks[i] == Latch) { + MaxBECount = EL.Max; + LatchMustExit = true; + } + else if (!LatchMustExit) MaxBECount = getUMaxFromMismatchedTypes(MaxBECount, EL.Max); } } - // Be more precise in the easy case of a loop latch that must exit. - if (LatchMaxCount) { - MaxBECount = getUMinFromMismatchedTypes(MaxBECount, LatchMaxCount); - } return BackedgeTakenInfo(ExitCounts, CouldComputeBECount, MaxBECount); } diff --git a/test/Analysis/ScalarEvolution/max-trip-count.ll b/test/Analysis/ScalarEvolution/max-trip-count.ll index 0cdbdf57a64..43a54b4f303 100644 --- a/test/Analysis/ScalarEvolution/max-trip-count.ll +++ b/test/Analysis/ScalarEvolution/max-trip-count.ll @@ -98,3 +98,29 @@ for.end: ; preds = %for.cond.for.end_cr ; CHECK: Determining loop execution counts for: @test ; CHECK-NEXT: backedge-taken count is ; CHECK-NEXT: max backedge-taken count is -1 + +; PR19799: Indvars miscompile due to an incorrect max backedge taken count from SCEV. +; CHECK-LABEL: @pr19799 +; CHECK: Loop %for.body.i: Unpredictable backedge-taken count. +; CHECK: Loop %for.body.i: max backedge-taken count is 1 +@a = common global i32 0, align 4 + +define i32 @pr19799() { +entry: + store i32 -1, i32* @a, align 4 + br label %for.body.i + +for.body.i: ; preds = %for.cond.i, %entry + %storemerge1.i = phi i32 [ -1, %entry ], [ %add.i.i, %for.cond.i ] + %tobool.i = icmp eq i32 %storemerge1.i, 0 + %add.i.i = add nsw i32 %storemerge1.i, 2 + br i1 %tobool.i, label %bar.exit, label %for.cond.i + +for.cond.i: ; preds = %for.body.i + store i32 %add.i.i, i32* @a, align 4 + %cmp.i = icmp slt i32 %storemerge1.i, 0 + br i1 %cmp.i, label %for.body.i, label %bar.exit + +bar.exit: ; preds = %for.cond.i, %for.body.i + ret i32 0 +}