Fix ScalarEvolution's backedge-taken count computations to check for

overflow when computing a integer division to round up.

Thanks to Nick Lewycky for noticing this!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73862 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Dan Gohman 2009-06-21 23:46:38 +00:00
parent 14ee48a5ba
commit 51f53b7f5a
5 changed files with 51 additions and 11 deletions

View File

@ -332,6 +332,13 @@ namespace llvm {
const SCEVHandle &SymName,
const SCEVHandle &NewVal);
/// getBECount - Subtract the end and start values and divide by the step,
/// rounding up, to get the number of times the backedge is executed. Return
/// CouldNotCompute if an intermediate computation overflows.
SCEVHandle getBECount(const SCEVHandle &Start,
const SCEVHandle &End,
const SCEVHandle &Step);
/// getBackedgeTakenInfo - Return the BackedgeTakenInfo for the given
/// loop, lazily computing new values if the loop hasn't been analyzed
/// yet.

View File

@ -3788,6 +3788,33 @@ bool ScalarEvolution::isLoopGuardedByCond(const Loop *L,
return false;
}
/// getBECount - Subtract the end and start values and divide by the step,
/// rounding up, to get the number of times the backedge is executed. Return
/// CouldNotCompute if an intermediate computation overflows.
SCEVHandle ScalarEvolution::getBECount(const SCEVHandle &Start,
const SCEVHandle &End,
const SCEVHandle &Step) {
const Type *Ty = Start->getType();
SCEVHandle NegOne = getIntegerSCEV(-1, Ty);
SCEVHandle Diff = getMinusSCEV(End, Start);
SCEVHandle RoundUp = getAddExpr(Step, NegOne);
// Add an adjustment to the difference between End and Start so that
// the division will effectively round up.
SCEVHandle Add = getAddExpr(Diff, RoundUp);
// Check Add for unsigned overflow.
// TODO: More sophisticated things could be done here.
const Type *WideTy = IntegerType::get(getTypeSizeInBits(Ty) + 1);
SCEVHandle OperandExtendedAdd =
getAddExpr(getZeroExtendExpr(Diff, WideTy),
getZeroExtendExpr(RoundUp, WideTy));
if (getZeroExtendExpr(Add, WideTy) != OperandExtendedAdd)
return CouldNotCompute;
return getUDivExpr(Add, Step);
}
/// HowManyLessThans - Return the number of times a backedge containing the
/// specified less-than comparison will execute. If not computable, return
/// CouldNotCompute.
@ -3869,16 +3896,11 @@ HowManyLessThans(const SCEV *LHS, const SCEV *RHS,
// Finally, we subtract these two values and divide, rounding up, to get
// the number of times the backedge is executed.
SCEVHandle BECount = getUDivExpr(getAddExpr(getMinusSCEV(End, Start),
getAddExpr(Step, NegOne)),
Step);
SCEVHandle BECount = getBECount(Start, End, Step);
// The maximum backedge count is similar, except using the minimum start
// value and the maximum end value.
SCEVHandle MaxBECount = getUDivExpr(getAddExpr(getMinusSCEV(MaxEnd,
MinStart),
getAddExpr(Step, NegOne)),
Step);
SCEVHandle MaxBECount = getBECount(MinStart, MaxEnd, Step);;
return BackedgeTakenInfo(BECount, MaxBECount);
}

View File

@ -1,4 +1,9 @@
; RUN: llvm-as < %s | opt -analyze -scalar-evolution -disable-output |& grep {/u 3}
; RUN: llvm-as < %s | opt -analyze -scalar-evolution -disable-output \
; RUN: | grep {Loop bb: Unpredictable backedge-taken count\\.}
; ScalarEvolution can't compute a trip count because it doesn't know if
; dividing by the stride will have a remainder. This could theoretically
; be teaching it how to use a more elaborate trip count computation.
define i32 @f(i32 %x) nounwind readnone {
entry:

View File

@ -1,5 +1,9 @@
; RUN: llvm-as < %s | opt -scalar-evolution -analyze -disable-output \
; RUN: | grep {backedge-taken count is ((64 + (-64 smax (-1 + (-1 \\* %0))) + %0) /u 64)}
; RUN: | grep {Loop bb3\\.i: Unpredictable backedge-taken count\\.}
; ScalarEvolution can't compute a trip count because it doesn't know if
; dividing by the stride will have a remainder. This could theoretically
; be teaching it how to use a more elaborate trip count computation.
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
target triple = "x86_64-unknown-linux-gnu"

View File

@ -1,7 +1,9 @@
; RUN: llvm-as < %s | opt -indvars -loop-deletion | llvm-dis | grep phi | count 1
; XFAIL: *
; Indvars should be able to evaluate this loop, allowing loop deletion
; to delete it.
; Indvars can't evaluate this loop, because ScalarEvolution can't compute
; an exact trip count, because it doesn't know if dividing by the stride will
; have a remainder. It could be done with more aggressive VRP though.
define i32 @test(i32 %x_offs) nounwind readnone {
entry: