[PATCH] D152414: AMDGPU: Do not not force alwaysinline with calls and module LDS enabled

Mon Jun 26 05:23:50 PDT 2023

arsenm added inline comments.

================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp:991-992

-  // Function calls are not supported, so make sure we inline everything.
-  addPass(createAMDGPUAlwaysInlinePass());
-  addPass(createAlwaysInlinerLegacyPass());
+  if (!AMDGPUTargetMachine::EnableFunctionCalls ||
+      !AMDGPUTargetMachine::EnableLowerModuleLDS) {
+    // Function calls are not supported, so make sure we inline everything.
----------------
jmmartinez wrote:
> Currently the inliner gives a boost to the inline-threshold for functions that take alloca instructions as arguments (ArgAllocaCost in AMDGPUTargetTransformInfo.cpp).
> 
> Should that inline bonus be removed too?
Those would be unrelated. That's just regular inline heuristics, not related to hacking out a call to avoid codegen limitations

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152414/new/

https://reviews.llvm.org/D152414