[llvm] [SCEV] Rework wrap-flag-inference for sext-addrec (PR #217362)
Nikita Popov via llvm-commits
llvm-commits at lists.llvm.org
Fri Aug 21 01:10:39 PDT 2026
================
@@ -2044,91 +2044,17 @@ const SCEV *ScalarEvolution::getSignExtendExprImpl(SCEVUse Op, Type *Ty,
// operands (often constants). This allows analysis of something like
// this: for (signed char X = 0; X < 100; ++X) { int Y = X; }
if (match(Op, m_scev_AffineAddRec(m_SCEV(Start), m_SCEV(Step), m_Loop(L)))) {
+ // Redo the AddRec check, computing nsw this time.
const auto *AR = cast<SCEVAddRecExpr>(Op);
- unsigned BitWidth = getTypeSizeInBits(AR->getType());
-
- // The no-signed-wrap case is handled before the uniquing lookup above.
-
- // Check whether the backedge-taken count is SCEVCouldNotCompute.
- // Note that this serves two purposes: It filters out loops that are
- // simply not analyzable, and it covers the case where this code is
- // being called from within backedge-taken count analysis, such that
- // attempting to ask for the backedge-taken count would likely result
- // in infinite recursion. In the later case, the analysis code will
- // cope with a conservative value, and it will take care to purge
- // that value once it has finished.
- const SCEV *MaxBECount = getConstantMaxBackedgeTakenCount(L);
- if (!isa<SCEVCouldNotCompute>(MaxBECount)) {
- // Manually compute the final value for AR, checking for
- // overflow.
-
- // Check whether the backedge-taken count can be losslessly casted to
- // the addrec's type. The count is always unsigned.
- const SCEV *CastedMaxBECount =
- getTruncateOrZeroExtend(MaxBECount, Start->getType(), Depth);
- const SCEV *RecastedMaxBECount = getTruncateOrZeroExtend(
- CastedMaxBECount, MaxBECount->getType(), Depth);
- if (MaxBECount == RecastedMaxBECount) {
- Type *WideTy = IntegerType::get(getContext(), BitWidth * 2);
- // Check whether Start+Step*MaxBECount has no signed overflow.
- const SCEV *SMul =
- getMulExpr(CastedMaxBECount, Step, SCEV::FlagAnyWrap, Depth + 1);
- const SCEV *SAdd = getSignExtendExpr(
- getAddExpr(Start, SMul, SCEV::FlagAnyWrap, Depth + 1), WideTy,
- Depth + 1);
- const SCEV *WideStart = getSignExtendExpr(Start, WideTy, Depth + 1);
- const SCEV *WideMaxBECount =
- getZeroExtendExpr(CastedMaxBECount, WideTy, Depth + 1);
- const SCEV *OperandExtendedAdd =
- getAddExpr(WideStart,
- getMulExpr(WideMaxBECount,
- getSignExtendExpr(Step, WideTy, Depth + 1),
- SCEV::FlagAnyWrap, Depth + 1),
- SCEV::FlagAnyWrap, Depth + 1);
- if (SAdd == OperandExtendedAdd) {
- // Cache knowledge of AR NSW, which is propagated to this AddRec.
- setNoWrapFlags(const_cast<SCEVAddRecExpr *>(AR), SCEV::FlagNSW);
- // Return the expression with the addrec on the outside.
- Start =
- getExtendAddRecStart<SCEVSignExtendExpr>(AR, Ty, this, Depth + 1);
- Step = getSignExtendExpr(Step, Ty, Depth + 1);
- return getAddRecExpr(Start, Step, L, AR->getNoWrapFlags());
- }
- // Similar to above, only this time treat the step value as unsigned.
- // This covers loops that count up with an unsigned step.
- OperandExtendedAdd =
- getAddExpr(WideStart,
- getMulExpr(WideMaxBECount,
- getZeroExtendExpr(Step, WideTy, Depth + 1),
- SCEV::FlagAnyWrap, Depth + 1),
- SCEV::FlagAnyWrap, Depth + 1);
- if (SAdd == OperandExtendedAdd) {
- // If AR wraps around then
- //
- // abs(Step) * MaxBECount > unsigned-max(AR->getType())
- // => SAdd != OperandExtendedAdd
- //
- // Thus (AR is not NW => SAdd != OperandExtendedAdd) <=>
- // (SAdd == OperandExtendedAdd => AR is NW)
-
- setNoWrapFlags(const_cast<SCEVAddRecExpr *>(AR), SCEV::FlagNW);
-
- // Return the expression with the addrec on the outside.
- Start =
- getExtendAddRecStart<SCEVSignExtendExpr>(AR, Ty, this, Depth + 1);
- Step = getZeroExtendExpr(Step, Ty, Depth + 1);
- return getAddRecExpr(Start, Step, L, AR->getNoWrapFlags());
- }
- }
- }
-
+ inferNoWrapViaConstantRanges(AR);
----------------
nikic wrote:
This bit is less clear to me -- why do we need to rerun inferNoWrapViaConstantRanges() which already ran on construction? Is this about addrecs that get created via getAddRecExpr() rather than via IR? I guess we don't run in the inference logic for those...
Also this drops the `nw` bit leading to the test regression. Presumably on the premise that it's incorrect, but as commented on the other PRs, I don't think it's actually incorrect.
https://github.com/llvm/llvm-project/pull/217362
More information about the llvm-commits
mailing list