[AMDGPU] Update pal metadata for v3.6 and fix v3.0 (llvm#135196)

Update entry_point for all pal versions below 3.6.
3.6 and above removes entry_point.

Author: dstutt

llvm/lib/Target/AMDGPU/Utils/AMDGPUPALMetadata.cpp

@@ -23,6 +23,7 @@
 #include "llvm/MC/MCExpr.h"
 #include "llvm/Support/AMDGPUMetadata.h"
 #include "llvm/Support/EndianStream.h"
+#include "llvm/Support/VersionTuple.h"
 
 using namespace llvm;
 using namespace llvm::AMDGPU;
@@ -259,13 +260,16 @@ void AMDGPUPALMetadata::setEntryPoint(unsigned CC, StringRef Name) {
   getHwStage(CC)[".entry_point_symbol"] =
       MsgPackDoc.getNode(Name, /*Copy=*/true);
 
-  // Set .entry_point which is defined
-  // to be _amdgpu_<stage> and _amdgpu_cs for non-shader functions
-  SmallString<16> EPName("_amdgpu_");
-  raw_svector_ostream EPNameOS(EPName);
-  EPNameOS << getStageName(CC) + 1;
-  getHwStage(CC)[".entry_point"] =
-      MsgPackDoc.getNode(EPNameOS.str(), /*Copy=*/true);
+  // For PAL version 3.6 and above, entry_point is no longer required.
+  if (getPALVersion() < VersionTuple(3, 6)) {
+    // Set .entry_point which is defined to be _amdgpu_<stage>_main and
+    // _amdgpu_cs_main for non-shader functions.
+    SmallString<16> EPName("_amdgpu_");
+    raw_svector_ostream EPNameOS(EPName);
+    EPNameOS << getStageName(CC) + 1 << "_main";
+    getHwStage(CC)[".entry_point"] =
+        MsgPackDoc.getNode(EPNameOS.str(), /*Copy=*/true);
+  }
 }
 
 // Set the number of used vgprs in the metadata. This is an optional
@@ -1052,6 +1056,10 @@ unsigned AMDGPUPALMetadata::getPALMajorVersion() { return getPALVersion(0); }
 
 unsigned AMDGPUPALMetadata::getPALMinorVersion() { return getPALVersion(1); }
 
+VersionTuple AMDGPUPALMetadata::getPALVersion() {
+  return VersionTuple(getPALVersion(0), getPALVersion(1));
+}
+
 // Set the field in a given .hardware_stages entry
 void AMDGPUPALMetadata::setHwStage(unsigned CC, StringRef field, unsigned Val) {
   getHwStage(CC)[field] = Val;

llvm/lib/Target/AMDGPU/Utils/AMDGPUPALMetadata.h

@@ -21,7 +21,6 @@
 namespace llvm {
 
 class Module;
-class StringRef;
 
 class AMDGPUPALMetadata {
 public:
@@ -155,6 +154,7 @@ class AMDGPUPALMetadata {
 
   unsigned getPALMajorVersion();
   unsigned getPALMinorVersion();
+  VersionTuple getPALVersion();
 
   void setHwStage(unsigned CC, StringRef field, unsigned Val);
   void setHwStage(unsigned CC, StringRef field, bool Val);

llvm/test/CodeGen/AMDGPU/amdpal-cs.ll

@@ -8,7 +8,7 @@
 ; GCN-NEXT: amdpal.pipelines:
 ; GCN-NEXT:   - .hardware_stages:
 ; GCN-NEXT:       .cs:
-; GCN-NEXT:         .entry_point:    _amdgpu_cs
+; GCN-NEXT:         .entry_point:    _amdgpu_cs_main
 ; GCN-NEXT:         .entry_point_symbol:    cs_amdpal
 ; GCN-NEXT:         .scratch_memory_size: 0
 ; GCN:     .registers:

llvm/test/CodeGen/AMDGPU/amdpal-es.ll

@@ -7,7 +7,7 @@
 ; GCN-NEXT: amdpal.pipelines:
 ; GCN-NEXT:   - .hardware_stages:
 ; GCN-NEXT:       .es:
-; GCN-NEXT:         .entry_point:    _amdgpu_es
+; GCN-NEXT:         .entry_point:    _amdgpu_es_main
 ; GCN-NEXT:         .entry_point_symbol:    es_amdpal
 ; GCN-NEXT:         .scratch_memory_size: 0
 ; GCN:     .registers:

llvm/test/CodeGen/AMDGPU/amdpal-gs.ll

@@ -8,7 +8,7 @@
 ; GCN-NEXT: amdpal.pipelines:
 ; GCN-NEXT:   - .hardware_stages:
 ; GCN-NEXT:       .gs:
-; GCN-NEXT:         .entry_point:    _amdgpu_gs
+; GCN-NEXT:         .entry_point:    _amdgpu_gs_main
 ; GCN-NEXT:         .entry_point_symbol:    gs_amdpal
 ; GCN-NEXT:         .scratch_memory_size: 0
 ; GCN:     .registers:

llvm/test/CodeGen/AMDGPU/amdpal-ls.ll

@@ -7,7 +7,7 @@
 ; GCN-NEXT: amdpal.pipelines:
 ; GCN-NEXT:   - .hardware_stages:
 ; GCN-NEXT:       .ls:
-; GCN-NEXT:         .entry_point:    _amdgpu_ls
+; GCN-NEXT:         .entry_point:    _amdgpu_ls_main
 ; GCN-NEXT:         .entry_point_symbol:    ls_amdpal
 ; GCN-NEXT:         .scratch_memory_size: 0
 ; GCN:     .registers:

llvm/test/CodeGen/AMDGPU/amdpal-psenable.ll

@@ -11,7 +11,7 @@
 ; GCN-NEXT: amdpal.pipelines:
 ; GCN-NEXT:   - .hardware_stages:
 ; GCN-NEXT:       .ps:
-; GCN-NEXT:         .entry_point:    _amdgpu_ps
+; GCN-NEXT:         .entry_point:    _amdgpu_ps_main
 ; GCN-NEXT:         .entry_point_symbol:    amdpal_psenable
 ; GCN-NEXT:         .scratch_memory_size: 0
 ; GCN:     .registers:

llvm/test/CodeGen/AMDGPU/amdpal-vs.ll

@@ -8,7 +8,7 @@
 ; GCN-NEXT: amdpal.pipelines:
 ; GCN-NEXT:   - .hardware_stages:
 ; GCN-NEXT:       .vs:
-; GCN-NEXT:         .entry_point:    _amdgpu_vs
+; GCN-NEXT:         .entry_point:    _amdgpu_vs_main
 ; GCN-NEXT:         .entry_point_symbol:    vs_amdpal
 ; GCN-NEXT:         .scratch_memory_size: 0
 ; GCN:     .registers:

llvm/test/CodeGen/AMDGPU/amdpal.ll

@@ -86,7 +86,7 @@ declare void @llvm.amdgcn.raw.ptr.buffer.store.f32(float, ptr addrspace(8), i32,
 ; PAL-NEXT: amdpal.pipelines:
 ; PAL-NEXT:   - .hardware_stages:
 ; PAL-NEXT:       .cs:
-; PAL-NEXT:         .entry_point:    _amdgpu_cs
+; PAL-NEXT:         .entry_point:    _amdgpu_cs_main
 ; PAL-NEXT:         .entry_point_symbol:    scratch2_cs
 ; PAL-NEXT:         .scratch_memory_size: 0x10
 ; PAL-NEXT:         .sgpr_count:     0x

llvm/test/CodeGen/AMDGPU/elf-notes.ll

@@ -66,7 +66,7 @@
 ; OSABI-PAL-ELF: amdpal.pipelines:
 ; OSABI-PAL-ELF:   - .hardware_stages:
 ; OSABI-PAL-ELF:       .cs:
-; OSABI-PAL-ELF:         .entry_point:    _amdgpu_cs
+; OSABI-PAL-ELF:         .entry_point:    _amdgpu_cs_main
 ; OSABI-PAL-ELF:         .entry_point_symbol:    elf_notes
 ; OSABI-PAL-ELF:         .scratch_memory_size: 0
 ; OSABI-PAL-ELF:         .sgpr_count:     96