Revert r300746 (SCEV analysis for or instructions).
There have been multiple reports of this causing problems: a
compile-time explosion on the LLVM testsuite, and a stack
overflow for an opencl kernel.
llvm-svn: 300928
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 700c383..593f78b 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -5328,12 +5328,28 @@
break;
case Instruction::Or:
- // Use ValueTracking to check whether this is actually an add.
- if (haveNoCommonBitsSet(BO->LHS, BO->RHS, getDataLayout(), &AC,
- nullptr, &DT)) {
- // There aren't any common bits set, so the add can't wrap.
- auto Flags = SCEV::NoWrapFlags(SCEV::FlagNUW | SCEV::FlagNSW);
- return getAddExpr(getSCEV(BO->LHS), getSCEV(BO->RHS), Flags);
+ // If the RHS of the Or is a constant, we may have something like:
+ // X*4+1 which got turned into X*4|1. Handle this as an Add so loop
+ // optimizations will transparently handle this case.
+ //
+ // In order for this transformation to be safe, the LHS must be of the
+ // form X*(2^n) and the Or constant must be less than 2^n.
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(BO->RHS)) {
+ const SCEV *LHS = getSCEV(BO->LHS);
+ const APInt &CIVal = CI->getValue();
+ if (GetMinTrailingZeros(LHS) >=
+ (CIVal.getBitWidth() - CIVal.countLeadingZeros())) {
+ // Build a plain add SCEV.
+ const SCEV *S = getAddExpr(LHS, getSCEV(CI));
+ // If the LHS of the add was an addrec and it has no-wrap flags,
+ // transfer the no-wrap flags, since an or won't introduce a wrap.
+ if (const SCEVAddRecExpr *NewAR = dyn_cast<SCEVAddRecExpr>(S)) {
+ const SCEVAddRecExpr *OldAR = cast<SCEVAddRecExpr>(LHS);
+ const_cast<SCEVAddRecExpr *>(NewAR)->setNoWrapFlags(
+ OldAR->getNoWrapFlags());
+ }
+ return S;
+ }
}
break;