[PATCH] D105787: [WIP][AbstractAttributor] Fold function calls to `__kmpc_is_spmd_exec_mode` if possible
Johannes Doerfert via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Sun Jul 11 19:35:24 PDT 2021
jdoerfert added a comment.
We need to properly tack when SPMDCompatibility is known.
================
Comment at: llvm/lib/Transforms/IPO/OpenMPOpt.cpp:3350
case OMPRTL___kmpc_is_spmd_exec_mode:
+ return;
case OMPRTL___kmpc_for_static_fini:
----------------
I don't think we need this anymore, do we?
================
Comment at: llvm/lib/Transforms/IPO/OpenMPOpt.cpp:3431
+ : public StateWrapper<KernelInfoState, AbstractAttribute> {
+ using Base = StateWrapper<KernelInfoState, AbstractAttribute>;
+
----------------
Do we need/want a KernelInfoState here? Wouldn't a BooleanState suffice to do the tacking of valid/invalid and fixpoint. Plus the ` Optional<Constant *> SimplifiedValue;`.
================
Comment at: llvm/lib/Transforms/IPO/OpenMPOpt.cpp:3480
+ A.registerSimplificationCallback(
+ IRPosition::callsite_function(CB),
+ [&](const IRPosition &IRP, const AbstractAttribute *AA,
----------------
================
Comment at: llvm/lib/Transforms/IPO/OpenMPOpt.cpp:3565
+
+ indicateOptimisticFixpoint();
+
----------------
tianshilei1992 wrote:
> This might not be always true. What if later there is a kernel entry in another mode?
I don't think that can happen
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D105787/new/
https://reviews.llvm.org/D105787
More information about the llvm-commits
mailing list