[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

Tue Sep 28 12:01:10 PDT 2021

gandhi21299 added inline comments.

================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp:96-97
     if (Function* F = dyn_cast<Function>(A.getAliasee())) {
+      if (A.getLinkage() != GlobalValue::InternalLinkage)
+        continue;
       A.replaceAllUsesWith(F);
----------------
yaxunl wrote:
> If we do this for older GPU's (e.g. Tonga/redwood), IR's using aliases will fail on them. I don't think it is acceptable.
> 
> Is it possible to restrict this change to gfx9 and above? Or should we introduce some feature to indicate 'alias support' and use that to restrict this change to subtargets supporting this feature.
Restricting this change to gfx9 and above sounds simpler and more relevant with the problem as well.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109707/new/

https://reviews.llvm.org/D109707