[PATCH] D129566: AMDGPU: Turn off force init 16 input SGPRS for pal

[David Stuttard via Phabricator via llvm-commits]

dstuttard created this revision.
Herald added subscribers: kosarev, jsilvanus, foad, kerbowa, hiraditya, t-tye, tpr, yaxunl, nhaehnle, jvesely, kzhuravl, arsenm.
Herald added a project: All.
dstuttard requested review of this revision.
Herald added subscribers: llvm-commits, wdng.
Herald added a project: LLVM.

Pal uses a different mechanism for user sgprs.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D129566

Files:
  llvm/lib/Target/AMDGPU/SIISelLowering.cpp


Index: llvm/lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -2112,7 +2112,8 @@
                                            SIMachineFunctionInfo &Info,
                                            CallingConv::ID CallConv,
                                            bool IsShader) const {
-  if (Subtarget->hasUserSGPRInit16Bug()) {
+  if (Subtarget->hasUserSGPRInit16Bug() && !Subtarget->isAmdPalOS()) {
+    // Note: user SGPRs are handled by the front-end for pal based shaders
     // Pad up the used user SGPRs with dead inputs.
     unsigned CurrentUserSGPRs = Info.getNumUserSGPRs();
 
@@ -2175,7 +2176,8 @@
     CCInfo.AllocateReg(PrivateSegmentWaveByteOffsetReg);
   }
 
-  assert(!Subtarget->hasUserSGPRInit16Bug() || Info.getNumPreloadedSGPRs() >= 16);
+  assert(!Subtarget->hasUserSGPRInit16Bug() || Subtarget->isAmdPalOS() ||
+         Info.getNumPreloadedSGPRs() >= 16);
 }
 
 static void reservePrivateMemoryRegs(const TargetMachine &TM,


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D129566.443952.patch
Type: text/x-patch
Size: 1105 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220712/41ff8153/attachment.bin>