[llvm] [AMDGPU] Infer amdgpu-no-flat-scratch-init attribute in AMDGPUAttributor (PR #94647)
Jun Wang via llvm-commits
llvm-commits at lists.llvm.org
Sat Aug 17 17:40:20 PDT 2024
================
@@ -677,6 +693,33 @@ struct AAAMDAttributesFunction : public AAAMDAttributes {
return !A.checkForAllCallLikeInstructions(DoesNotRetrieve, *this,
UsedAssumedInformation);
}
+
+ // Returns true if FlatScratchInit is needed, i.e., no-flat-scratch-init is
+ // not to be set.
+ bool needFlatScratchInit(Attributor &A) {
+ // This is called on each callee; false means callee shouldn't have
+ // no-flat-scratch-init.
+ auto CheckForNoFlatScratchInit = [&](Instruction &I) {
+ const auto &CB = cast<CallBase>(I);
+ const Value *CalleeOp = CB.getCalledOperand();
+ const Function *Callee = dyn_cast<Function>(CalleeOp);
+ if (!Callee) // indirect call
+ return CB.isInlineAsm();
----------------
jwanggit86 wrote:
Code has been updated after fixing a bug regarding indirect calls of known callees. I'm not sure using `checkForAllCallees()` is necessary for 2 reasons: (1) code is now very simple, boiling down to just checking for intrinsics (2) the first part of `AAAMDAttributesFunction::updateImpl()` already does similar work as `checkForAllCallees()`:
```cpp
ChangeStatus updateImpl(Attributor &A) override {
...
// Check for Intrinsics and propagate attributes.
const AACallEdges *AAEdges = A.getAAFor<AACallEdges>(
*this, this->getIRPosition(), DepClassTy::REQUIRED);
if (!AAEdges || AAEdges->hasNonAsmUnknownCallee())
return indicatePessimisticFixpoint();
...
for (Function *Callee : AAEdges->getOptimisticEdges()) { // Jun - what checkForAllCallees() does is essentially calling getOptimisticEdges() and then running the pred on the edges
Intrinsic::ID IID = Callee->getIntrinsicID();
if (IID == Intrinsic::not_intrinsic) {
const AAAMDAttributes *AAAMD = A.getAAFor<AAAMDAttributes>(
*this, IRPosition::function(*Callee), DepClassTy::REQUIRED);
if (!AAAMD)
return indicatePessimisticFixpoint();
*this &= *AAAMD;
continue;
}
...
```
https://github.com/llvm/llvm-project/pull/94647
More information about the llvm-commits
mailing list