mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-24 21:42:54 +02:00
027f71d443
SCEV currently fails to compute loop counts for nonunit stride
loops. This comes up frequently. It prevents loop optimization and
forces vectorization to insert extra loop checks.
For example:
void foo(int n, int *x) {
for (int i = 0; i < n; i += 3) {
x[i] = i;
x[i+1] = i+1;
x[i+2] = i+2;
}
}
We need to properly handle the case in which limit > INT_MAX-stride. In
the above case: n > INT_MAX-3. In this case the loop counter will step
beyond the limit and overflow at the same time. However, knowing that
signed integer overlow in undefined, we can assume the loop test
behavior is arbitrary after overflow. This obeys both C undefined
behavior rules, and the more strict LLVM poison value rules.
I'm finally fixing this in response to Hal Finkel's persistence.
The most probable reason that we never optimized this before is that
we were being careful to handle case where the developer expected a
side-effect free infinite loop relying on overflow:
for (int i = 0; i < n; i += s) {
++j;
}
return j;
If INT_MAX+1 is a multiple of s and n > INT_MAX-s, then we might
expect an infinite loop. However there are plenty of ways to achieve
this effect without relying on undefined behavior of signed overflow.
llvm-svn: 193015
|
||
|---|---|---|
| .. | ||
| 2007-07-15-NegativeStride.ll | ||
| 2007-08-06-MisinterpretBranch.ll | ||
| 2007-08-06-Unsigned.ll | ||
| 2007-09-27-LargeStepping.ll | ||
| 2007-11-14-SignedAddRec.ll | ||
| 2007-11-18-OrInstruction.ll | ||
| 2008-02-11-ReversedCondition.ll | ||
| 2008-02-12-SMAXTripCount.ll | ||
| 2008-02-15-UMax.ll | ||
| 2008-05-25-NegativeStepToZero.ll | ||
| 2008-06-12-BinomialInt64.ll | ||
| 2008-07-12-UnneededSelect1.ll | ||
| 2008-07-12-UnneededSelect2.ll | ||
| 2008-07-19-InfiniteLoop.ll | ||
| 2008-07-19-WrappingIV.ll | ||
| 2008-07-29-SGTTripCount.ll | ||
| 2008-07-29-SMinExpr.ll | ||
| 2008-08-04-IVOverflow.ll | ||
| 2008-08-04-LongAddRec.ll | ||
| 2008-11-02-QuadraticCrash.ll | ||
| 2008-11-15-CubicOOM.ll | ||
| 2008-11-18-LessThanOrEqual.ll | ||
| 2008-11-18-Stride1.ll | ||
| 2008-11-18-Stride2.ll | ||
| 2008-12-08-FiniteSGE.ll | ||
| 2008-12-11-SMaxOverflow.ll | ||
| 2008-12-14-StrideAndSigned.ll | ||
| 2008-12-15-DontUseSDiv.ll | ||
| 2009-01-02-SignedNegativeStride.ll | ||
| 2009-04-22-TruncCast.ll | ||
| 2009-05-09-PointerEdgeCount.ll | ||
| 2009-07-04-GroupConstantsWidthMismatch.ll | ||
| 2010-09-03-RequiredTransitive.ll | ||
| 2011-03-09-ExactNoMaxBECount.ll | ||
| 2011-04-26-FoldAddRec.ll | ||
| 2011-10-04-ConstEvolve.ll | ||
| 2012-03-26-LoadConstant.ll | ||
| 2012-05-18-LoopPredRecurse.ll | ||
| 2012-05-29-MulAddRec.ll | ||
| and-xor.ll | ||
| avoid-infinite-recursion-0.ll | ||
| avoid-infinite-recursion-1.ll | ||
| avoid-smax-0.ll | ||
| avoid-smax-1.ll | ||
| div-overflow.ll | ||
| do-loop.ll | ||
| fold.ll | ||
| how-far-to-zero.ll | ||
| load.ll | ||
| max-trip-count-address-space.ll | ||
| max-trip-count.ll | ||
| nsw-offset.ll | ||
| nsw.ll | ||
| pointer-sign-bits.ll | ||
| pr3909.ll | ||
| scev-aa.ll | ||
| scev-invalid.ll | ||
| sext-inreg.ll | ||
| sext-iv-0.ll | ||
| sext-iv-1.ll | ||
| sext-iv-2.ll | ||
| sle.ll | ||
| smax.ll | ||
| SolveQuadraticEquation.ll | ||
| trip-count2.ll | ||
| trip-count3.ll | ||
| trip-count4.ll | ||
| trip-count5.ll | ||
| trip-count6.ll | ||
| trip-count7.ll | ||
| trip-count8.ll | ||
| trip-count9.ll | ||
| trip-count10.ll | ||
| trip-count11.ll | ||
| trip-count12.ll | ||
| trip-count.ll | ||
| undefined.ll | ||
| unreachable-code.ll | ||
| unsimplified-loop.ll | ||
| xor-and.ll | ||
| zext-wrap.ll | ||