[llvm-branch-commits] [llvm] AMDGPU: Start using subarch in attributor instead of subtarget (PR #212207)

[Matt Arsenault via llvm-branch-commits]

================
@@ -153,7 +153,13 @@ class AMDGPUInformationCache : public InformationCache {
                          BumpPtrAllocator &Allocator,
                          SetVector<Function *> *CGSCC, TargetMachine &TM)
       : InformationCache(M, AG, Allocator, CGSCC), TM(TM),
-        CodeObjectVersion(AMDGPU::getAMDHSACodeObjectVersion(M)) {}
+        CodeObjectVersion(AMDGPU::getAMDHSACodeObjectVersion(M)),
+        HasApertureRegs(
+            AMDGPU::hasApertureRegs(M.getTargetTriple().getSubArch())),
+        SupportsGetDoorbellID(
+            AMDGPU::supportsGetDoorbellID(M.getTargetTriple().getSubArch())),
+        ArchAttr(AMDGPU::getArchAttrAMDGCN(
----------------
arsenm wrote:

Missed optimization 

https://github.com/llvm/llvm-project/pull/212207