[HLSL] Allow SV_GroupIndex for lib profile.

[python3kgae]

Summary

Lib profile could include all kind of entry functions.

[aaron.ballman]

This looks reasonable to me, but I leave the final sign-off to @beanz.

clang/test/CodeGenHLSL/GlobalDestructors.hlsl

1–2

Strips an introduced trailing whitespace.

[beanz]

clang/lib/Sema/SemaDeclAttr.cpp

6901

I think the underlying issue here is that some attributes should only be validated on entry points that are being compiled as entry points in the current compilation action. Please file a GitHub issue for this and put the issue URL in the comment here.

[python3kgae]

Update comments and format.

[beanz]

LGTM.

clang/lib/Sema/SemaDeclAttr.cpp

if (HLSLNumThreadsAttr *NT = D->getAttr<HLSLNumThreadsAttr>()) {
     return nullptr;
   }
   return ::new (Context) HLSLNumThreadsAttr(Context, AL, X, Y, Z);
 }
 
 static void handleHLSLSVGroupIndexAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   using llvm::Triple;
   Triple Target = S.Context.getTargetInfo().getTriple();
-  if (Target.getEnvironment() != Triple::Compute) {
+  if (Target.getEnvironment() != Triple::Compute &&
+      Target.getEnvironment() != Triple::Library) {
+    // FIXME: it is OK for a compute shader entry and pixel shader entry live in
+    // same HLSL file. Issue https://github.com/llvm/llvm-project/issues/57880.
     uint32_t Pipeline =
         (uint32_t)S.Context.getTargetInfo().getTriple().getEnvironment() -
         (uint32_t)llvm::Triple::Pixel;
     S.Diag(AL.getLoc(), diag::err_hlsl_attr_unsupported_in_stage)
         << AL << Pipeline << "Compute";
     return;
   }
 

clang/test/CodeGenHLSL/GlobalDestructors.hlsl

 // RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -S -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -std=hlsl202x -S -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=CHECK
 
 struct Tail {
   Tail() {
     add(1);
   }
 
   ~Tail() {
     add(-1);