IV widening is sometimes a strictly harmful transform (some examples
of this are shown in tests 11, 12 in widen-loop-comp.ll). One of the
reasons of this is that sometimes SCEV fails to prove some facts after
part of guards has been widened.
Though each single such case looks like a bug that can be addressed,
it seems that disabling of IV widening may be profitable in some cases.
We want to have an option to do so. By default, existing behavior is
preserved and IV widening is on.
LoopPredication("indvars-predicate-loops", cl::Hidden, cl::init(true),
cl::desc("Predicate conditions in read only loops"));
+static cl::opt<bool>
+AllowIVWidening("indvars-widen-indvars", cl::Hidden, cl::init(true),
+ cl::desc("Allow widening of indvars to eliminate s/zext"));
+
namespace {
struct RewritePhi;
/// It would be simpler to delete uses as they are processed, but we must avoid
/// invalidating SCEV expressions.
PHINode *WidenIV::createWideIV(SCEVExpander &Rewriter) {
+ // Bail if we disallowed widening.
+ if(!AllowIVWidening)
+ return nullptr;
+
// Is this phi an induction variable?
const SCEVAddRecExpr *AddRec = dyn_cast<SCEVAddRecExpr>(SE->getSCEV(OrigPhi));
if (!AddRec)