[llvm] [VPlan] Introduce VPIRFlags::getNWFlagsOrNone (NFC) (PR #207176)
Ramkumar Ramachandra via llvm-commits
llvm-commits at lists.llvm.org
Thu Jul 9 01:46:36 PDT 2026
================
@@ -4118,15 +4115,15 @@ class VPCurrentIterationPHIRecipe : public VPHeaderPHIRecipe {
class VPWidenCanonicalIVRecipe : public VPRecipeWithIRFlags {
public:
VPWidenCanonicalIVRecipe(VPRegionValue *CanonicalIV,
- const VPIRFlags::WrapFlagsTy &Flags = {false, false})
+ const VPIRFlags::WrapFlagsTy &Flags = {})
: VPRecipeWithIRFlags(VPRecipeBase::VPWidenCanonicalIVSC, CanonicalIV,
CanonicalIV->getType(), Flags) {}
~VPWidenCanonicalIVRecipe() override = default;
VPWidenCanonicalIVRecipe *clone() override {
auto *WideCanIV =
- new VPWidenCanonicalIVRecipe(getCanonicalIV(), getNoWrapFlags());
+ new VPWidenCanonicalIVRecipe(getCanonicalIV(), getNoWrapFlagsOrNone());
----------------
artagnon wrote:
Hm, not sure. I think the long-term plan is to allow multiple kinds of flags in VPIRFlags, for example WrapFlags with GEPNoWrapFlags -- it was discovered in https://github.com/llvm/llvm-project/pull/205908; the idea is to never have getting flags to assert. I think we currently verify if flags are valid for an opcode, so I'm not sure I see the value of asserting at the moment?
https://github.com/llvm/llvm-project/pull/207176
More information about the llvm-commits
mailing list