maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[LoopFlatten] Fix bug where SCEVCouldNotCompute object is used 

The SCEV method getBackedgeTakenCount() returns a SCEVCouldNotCompute
object if the backedge-taken count is unpredictable. This fix ensures
there is no longer an attempt to use such an object to find the trip
count.

Patch by: Rosie Sumpter.

Differential Revision: https://reviews.llvm.org/D106970
This commit is contained in:
Sjoerd Meijer 2021-07-28 18:35:08 +01:00
parent 03b8c69d06
commit bc43078fe8
2 changed files with 52 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
llvm/lib/Transforms/Scalar/LoopFlatten.cpp
View File

@ -170,8 +170,12 @@ static bool findLoopComponents(
// icmp ult %inc, tripcount -> icmp ult %j, tripcount-1, then we don't flatten
// the loop (yet).
TripCount = Compare->getOperand(1);
const SCEV *SCEVTripCount =
SE->getTripCountFromExitCount(SE->getBackedgeTakenCount(L));
const SCEV *BackedgeTakenCount = SE->getBackedgeTakenCount(L);
if (isa<SCEVCouldNotCompute>(BackedgeTakenCount)) {
LLVM_DEBUG(dbgs() << "Backedge-taken count is not predictable\n");
return false;
}
const SCEV *SCEVTripCount = SE->getTripCountFromExitCount(BackedgeTakenCount);
if (SE->getSCEV(TripCount) != SCEVTripCount) {
if (!IsWidened) {
LLVM_DEBUG(dbgs() << "Could not find valid trip count\n");

46
llvm/test/Transforms/LoopFlatten/loop-flatten-negative.ll
View File

@ -637,5 +637,51 @@ for.cond.cleanup:
ret void
}
; Backedge-taken count is not predictable.
%struct.Limits = type { i16, i16 }
define void @backedge_count(%struct.Limits* %lim) {
entry:
%N = getelementptr inbounds %struct.Limits, %struct.Limits* %lim, i32 0, i32 0
%M = getelementptr inbounds %struct.Limits, %struct.Limits* %lim, i32 0, i32 1
%0 = load i16, i16* %N, align 2
%cmp20 = icmp sgt i16 %0, 0
br i1 %cmp20, label %for.cond2.preheader.preheader, label %for.cond.cleanup
for.cond2.preheader.preheader:
%.pre = load i16, i16* %M, align 2
br label %for.cond2.preheader
for.cond2.preheader:
%1 = phi i16 [ %3, %for.cond.cleanup6 ], [ %0, %for.cond2.preheader.preheader ]
%2 = phi i16 [ %4, %for.cond.cleanup6 ], [ %.pre, %for.cond2.preheader.preheader ]
%i.021 = phi i32 [ %inc9, %for.cond.cleanup6 ], [ 0, %for.cond2.preheader.preheader ]
%cmp417 = icmp sgt i16 %2, 0
br i1 %cmp417, label %for.body7, label %for.cond.cleanup6
for.cond.cleanup:
ret void
for.cond.cleanup6.loopexit:
%.pre22 = load i16, i16* %N, align 2
br label %for.cond.cleanup6
for.cond.cleanup6:
%3 = phi i16 [ %.pre22, %for.cond.cleanup6.loopexit ], [ %1, %for.cond2.preheader ]
%4 = phi i16 [ %5, %for.cond.cleanup6.loopexit ], [ %2, %for.cond2.preheader ]
%inc9 = add nuw nsw i32 %i.021, 1
%conv = sext i16 %3 to i32
%cmp = icmp slt i32 %inc9, %conv
br i1 %cmp, label %for.cond2.preheader, label %for.cond.cleanup
for.body7:
%j.018 = phi i32 [ %inc, %for.body7 ], [ 0, %for.cond2.preheader ]
tail call void bitcast (void (...)* @g to void ()*)()
%inc = add nuw nsw i32 %j.018, 1
%5 = load i16, i16* %M, align 2
%conv3 = sext i16 %5 to i32
%cmp4 = icmp slt i32 %inc, %conv3
br i1 %cmp4, label %for.body7, label %for.cond.cleanup6.loopexit
}
declare dso_local void @f(i32*)
declare dso_local void @g(...)