[AMDGPU] Avoid processing functions in amdgpu-propagate-attributes pass for shaders

[piotr]

Summary

The pass amdgpu-propagate-attributes ("Early/Late propagate attributes
from kernels to functions") is currently run also for shaders, where
it does nothing. Modify the check so the pass only processes functions
for kernels.

[arsenm]

Needs test (although I am working towards removing this pass)

[piotr]

I was thinking about this, but not sure what the test would demonstrate, as the patch just limits the number of cases for which this pass works. With this patch, there are no test changes in the existing set of tests, as expected.

This patch causes an early exit for calling conventions where isShader() is true. We run early inlining in these cases.

[arsenm]

In D109961#3009438, @piotr wrote:

I was thinking about this, but not sure what the test would demonstrate, as the patch just limits the number of cases for which this pass works. With this patch, there are no test changes in the existing set of tests, as expected.

This patch causes an early exit for calling conventions where isShader() is true. We run early inlining in these cases.

It would show not touching the attributes on the function

llvm/lib/Target/AMDGPU/AMDGPUPropagateAttributes.cpp

for (Clone &C : Clones)
     if (C.OrigF == OrigF && PropsNeeded == C.Properties)
       return C.NewF;
 
   return nullptr;
 }
 
 bool AMDGPUPropagateAttributes::process(Module &M) {
   for (auto &F : M.functions())
-    if (AMDGPU::isEntryFunctionCC(F.getCallingConv()))
+    if (AMDGPU::isKernel(F.getCallingConv()))
       Roots.insert(&F);
 
-  return process();
+  return Roots.empty() ? false : process();
 }
 
 bool AMDGPUPropagateAttributes::process(Function &F) {
   Roots.insert(&F);
   return process();
 }
 
 bool AMDGPUPropagateAttributes::process() {
   bool Changed = false;
   SmallSet<Function *, 32> NewRoots;
   SmallSet<Function *, 32> Replaced;
 
-  if (Roots.empty())
-    return false;
+  assert(!Roots.empty());
   Module &M = *(*Roots.begin())->getParent();
 
   do {
     Roots.insert(NewRoots.begin(), NewRoots.end());
     NewRoots.clear();
 
     for (auto &F : M.functions()) {
       if (F.isDeclaration())
bool AMDGPUPropagateAttributesEarly::runOnFunction(Function &F) {
   if (!TM) {
     auto *TPC = getAnalysisIfAvailable<TargetPassConfig>();
     if (!TPC)
       return false;
 
     TM = &TPC->getTM<TargetMachine>();
   }
 
-  if (!AMDGPU::isEntryFunctionCC(F.getCallingConv()))
+  if (!AMDGPU::isKernel(F.getCallingConv()))
     return false;
 
   return AMDGPUPropagateAttributes(TM, false).process(F);
 }
 
 bool AMDGPUPropagateAttributesLate::runOnModule(Module &M) {
   if (!TM) {
     auto *TPC = getAnalysisIfAvailable<TargetPassConfig>();