AMDGPU: [AMDGPU] Assembler: add .hsa_code_object_metadata directive for functime metadata V2.0

Summary:
Added pair of directives .hsa_code_object_metadata/.end_hsa_code_object_metadata.
Between them user can put YAML string that would be directly put to the generated note. E.g.:
'''
.hsa_code_object_metadata
    {
        amd.MDVersion: [ 2, 0 ]
    }
.end_hsa_code_object_metadata
'''
Based on D25046

Reviewers: vpykhtin, nhaustov, yaxunl, tstellarAMD

Subscribers: arsenm, kzhuravl, wdng, nhaehnle, mgorny, tony-tye

Differential Revision: https://reviews.llvm.org/D27619

llvm-svn: 290097

lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

@@ -119,7 +119,7 @@ void AMDGPUAsmPrinter::EmitStartOfAsmFile(Module &M) {
                                     "AMD", "AMDGPU");
 
   // Emit runtime metadata.
-  TS->emitRuntimeMetadata(M);
+  TS->EmitRuntimeMetadata(M);
 }
 
 bool AMDGPUAsmPrinter::isBlockOnlyReachableByFallthrough(

lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp

@@ -23,6 +23,7 @@
 #include "llvm/ADT/StringSwitch.h"
 #include "llvm/ADT/Twine.h"
 #include "llvm/CodeGen/MachineValueType.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCInst.h"
@@ -45,6 +46,8 @@
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Support/SMLoc.h"
 #include "llvm/Support/TargetRegistry.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Support/MathExtras.h"
 #include <algorithm>
 #include <cassert>
 #include <cstdint>
@@ -678,6 +681,7 @@ private:
   bool ParseDirectiveMajorMinor(uint32_t &Major, uint32_t &Minor);
   bool ParseDirectiveHSACodeObjectVersion();
   bool ParseDirectiveHSACodeObjectISA();
+  bool ParseDirectiveRuntimeMetadata();
   bool ParseAMDKernelCodeTValue(StringRef ID, amd_kernel_code_t &Header);
   bool ParseDirectiveAMDKernelCodeT();
   bool ParseSectionDirectiveHSAText();
@@ -1747,6 +1751,46 @@ bool AMDGPUAsmParser::ParseDirectiveHSACodeObjectISA() {
   return false;
 }
 
+bool AMDGPUAsmParser::ParseDirectiveRuntimeMetadata() {
+  std::string Metadata;
+  raw_string_ostream MS(Metadata);
+
+  getLexer().setSkipSpace(false);
+
+  bool FoundEnd = false;
+  while (!getLexer().is(AsmToken::Eof)) {
+    while (getLexer().is(AsmToken::Space)) {
+      MS << ' ';
+      Lex();
+    }
+
+    if (getLexer().is(AsmToken::Identifier)) {
+      StringRef ID = getLexer().getTok().getIdentifier();
+      if (ID == ".end_amdgpu_runtime_metadata") {
+        Lex();
+        FoundEnd = true;
+        break;
+      }
+    }
+
+    MS << Parser.parseStringToEndOfStatement()
+       << getContext().getAsmInfo()->getSeparatorString();
+
+    Parser.eatToEndOfStatement();
+  }
+
+  getLexer().setSkipSpace(true);
+
+  if (getLexer().is(AsmToken::Eof) && !FoundEnd)
+    return TokError("expected directive .end_amdgpu_runtime_metadata not found");
+
+  MS.flush();
+
+  getTargetStreamer().EmitRuntimeMetadata(Metadata);
+
+  return false;
+}
+
 bool AMDGPUAsmParser::ParseAMDKernelCodeTValue(StringRef ID,
                                                amd_kernel_code_t &Header) {
   SmallString<40> ErrStr;
@@ -1853,6 +1897,9 @@ bool AMDGPUAsmParser::ParseDirective(AsmToken DirectiveID) {
   if (IDVal == ".hsa_code_object_isa")
     return ParseDirectiveHSACodeObjectISA();
 
+  if (IDVal == ".amdgpu_runtime_metadata")
+    return ParseDirectiveRuntimeMetadata();
+
   if (IDVal == ".amd_kernel_code_t")
     return ParseDirectiveAMDKernelCodeT();
 

lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp

@@ -93,6 +93,18 @@ void AMDGPUTargetAsmStreamer::EmitAMDGPUHsaProgramScopeGlobal(
   OS << "\t.amdgpu_hsa_program_global " << GlobalName << '\n';
 }
 
+void AMDGPUTargetAsmStreamer::EmitRuntimeMetadata(Module &M) {
+  OS << "\t.amdgpu_runtime_metadata\n";
+  OS << getRuntimeMDYAMLString(M);
+  OS << "\n\t.end_amdgpu_runtime_metadata\n";
+}
+
+void AMDGPUTargetAsmStreamer::EmitRuntimeMetadata(StringRef Metadata) {
+  OS << "\t.amdgpu_runtime_metadata";
+  OS << Metadata;
+  OS << "\t.end_amdgpu_runtime_metadata\n";
+}
+
 //===----------------------------------------------------------------------===//
 // AMDGPUTargetELFStreamer
 //===----------------------------------------------------------------------===//
@@ -105,25 +117,39 @@ MCELFStreamer &AMDGPUTargetELFStreamer::getStreamer() {
 }
 
 void
-AMDGPUTargetELFStreamer::EmitDirectiveHSACodeObjectVersion(uint32_t Major,
-                                                           uint32_t Minor) {
-  MCStreamer &OS = getStreamer();
-  MCSectionELF *Note =
-      OS.getContext().getELFSection(PT_NOTE::SectionName, ELF::SHT_NOTE,
-                                    ELF::SHF_ALLOC);
+AMDGPUTargetELFStreamer::EmitAMDGPUNote(const MCExpr* DescSZ,
+                                        PT_NOTE::NoteType Type,
+                              std::function<void(MCELFStreamer &)> EmitDesc) {
+  auto &S = getStreamer();
+  auto &Context = S.getContext();
 
   auto NameSZ = sizeof(PT_NOTE::NoteName);
-  OS.PushSection();
-  OS.SwitchSection(Note);
-  OS.EmitIntValue(NameSZ, 4);                                     // namesz
-  OS.EmitIntValue(8, 4);                                          // descz
-  OS.EmitIntValue(PT_NOTE::NT_AMDGPU_HSA_CODE_OBJECT_VERSION, 4); // type
-  OS.EmitBytes(StringRef(PT_NOTE::NoteName, NameSZ));             // name
-  OS.EmitValueToAlignment(4);
-  OS.EmitIntValue(Major, 4);                                      // desc
+
+  S.PushSection();
+  S.SwitchSection(Context.getELFSection(
+    PT_NOTE::SectionName, ELF::SHT_NOTE, ELF::SHF_ALLOC));
+  S.EmitIntValue(NameSZ, 4);                                  // namesz
+  S.EmitValue(DescSZ, 4);                                     // descz
+  S.EmitIntValue(Type, 4); // type
+  S.EmitBytes(StringRef(PT_NOTE::NoteName, NameSZ));          // name
+  S.EmitValueToAlignment(4, 0, 1, 0);                         // padding 0
+  EmitDesc(S);                                                // desc
+  S.EmitValueToAlignment(4, 0, 1, 0);                         // padding 0
+  S.PopSection();
+}
+
+void
+AMDGPUTargetELFStreamer::EmitDirectiveHSACodeObjectVersion(uint32_t Major,
+                                                           uint32_t Minor) {
+
+  EmitAMDGPUNote(
+    MCConstantExpr::create(8, getContext()),
+    PT_NOTE::NT_AMDGPU_HSA_CODE_OBJECT_VERSION,
+    [&](MCELFStreamer &OS){
+      OS.EmitIntValue(Major, 4);
       OS.EmitIntValue(Minor, 4);
-  OS.EmitValueToAlignment(4);
-  OS.PopSection();
+    }
+  );
 }
 
 void
@@ -132,26 +158,18 @@ AMDGPUTargetELFStreamer::EmitDirectiveHSACodeObjectISA(uint32_t Major,
                                                        uint32_t Stepping,
                                                        StringRef VendorName,
                                                        StringRef ArchName) {
-  MCStreamer &OS = getStreamer();
-  MCSectionELF *Note =
-      OS.getContext().getELFSection(PT_NOTE::SectionName, ELF::SHT_NOTE,
-                                    ELF::SHF_ALLOC);
-
   uint16_t VendorNameSize = VendorName.size() + 1;
   uint16_t ArchNameSize = ArchName.size() + 1;
+  
   unsigned DescSZ = sizeof(VendorNameSize) + sizeof(ArchNameSize) +
     sizeof(Major) + sizeof(Minor) + sizeof(Stepping) +
     VendorNameSize + ArchNameSize;
 
-  OS.PushSection();
-  OS.SwitchSection(Note);
-  auto NameSZ = sizeof(PT_NOTE::NoteName);
-  OS.EmitIntValue(NameSZ, 4);                              // namesz
-  OS.EmitIntValue(DescSZ, 4);                              // descsz
-  OS.EmitIntValue(PT_NOTE::NT_AMDGPU_HSA_ISA, 4);          // type
-  OS.EmitBytes(StringRef(PT_NOTE::NoteName, NameSZ));      // name
-  OS.EmitValueToAlignment(4);
-  OS.EmitIntValue(VendorNameSize, 2);                      // desc
+  EmitAMDGPUNote(
+    MCConstantExpr::create(DescSZ, getContext()),
+    PT_NOTE::NT_AMDGPU_HSA_ISA,
+    [&](MCELFStreamer &OS) {
+      OS.EmitIntValue(VendorNameSize, 2);
       OS.EmitIntValue(ArchNameSize, 2);
       OS.EmitIntValue(Major, 4);
       OS.EmitIntValue(Minor, 4);
@@ -160,8 +178,8 @@ AMDGPUTargetELFStreamer::EmitDirectiveHSACodeObjectISA(uint32_t Major,
       OS.EmitIntValue(0, 1); // NULL terminate VendorName
       OS.EmitBytes(ArchName);
       OS.EmitIntValue(0, 1); // NULL terminte ArchName
-  OS.EmitValueToAlignment(4);
-  OS.PopSection();
+    }
+  );
 }
 
 void
@@ -198,35 +216,27 @@ void AMDGPUTargetELFStreamer::EmitAMDGPUHsaProgramScopeGlobal(
   Symbol->setBinding(ELF::STB_GLOBAL);
 }
 
-void AMDGPUTargetELFStreamer::emitRuntimeMetadata(Module &M) {
-  auto &S = getStreamer();
-  auto &Context = S.getContext();
-
-  auto NameSZ = sizeof(PT_NOTE::NoteName); // Size of note name including trailing null.
-
-  S.PushSection();
-  S.SwitchSection(Context.getELFSection(
-      PT_NOTE::SectionName, ELF::SHT_NOTE, ELF::SHF_ALLOC));
-
+void AMDGPUTargetELFStreamer::EmitRuntimeMetadata(StringRef Metadata) {
   // Create two labels to mark the beginning and end of the desc field
   // and a MCExpr to calculate the size of the desc field.
+  auto &Context = getContext();
   auto *DescBegin = Context.createTempSymbol();
   auto *DescEnd = Context.createTempSymbol();
   auto *DescSZ = MCBinaryExpr::createSub(
     MCSymbolRefExpr::create(DescEnd, Context),
     MCSymbolRefExpr::create(DescBegin, Context), Context);
 
-  // Emit the note element for runtime metadata.
-  // Name and desc should be padded to 4 byte boundary but size of name and
-  // desc should not include padding 0's.
-  S.EmitIntValue(NameSZ, 4);                                  // namesz
-  S.EmitValue(DescSZ, 4);                                     // descz
-  S.EmitIntValue(PT_NOTE::NT_AMDGPU_HSA_RUNTIME_METADATA, 4); // type
-  S.EmitBytes(StringRef(PT_NOTE::NoteName, NameSZ));          // name
-  S.EmitValueToAlignment(4, 0, 1, 0);                         // padding 0
-  S.EmitLabel(DescBegin);
-  S.EmitBytes(getRuntimeMDYAMLString(M));                               // desc
-  S.EmitLabel(DescEnd);
-  S.EmitValueToAlignment(4, 0, 1, 0);                         // padding 0
-  S.PopSection();
+  EmitAMDGPUNote(
+    DescSZ,
+    PT_NOTE::NT_AMDGPU_HSA_RUNTIME_METADATA,
+    [&](MCELFStreamer &OS) {
+      OS.EmitLabel(DescBegin);
+      OS.EmitBytes(Metadata);
+      OS.EmitLabel(DescEnd);
+    }
+  );
+}
+
+void AMDGPUTargetELFStreamer::EmitRuntimeMetadata(Module &M) {
+  EmitRuntimeMetadata(getRuntimeMDYAMLString(M));
 }

lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.h

@@ -14,6 +14,7 @@
 #include "llvm/MC/MCStreamer.h"
 
 namespace llvm {
+#include "AMDGPUPTNote.h"
 
 class DataLayout;
 class Function;
@@ -24,6 +25,9 @@ class Module;
 class Type;
 
 class AMDGPUTargetStreamer : public MCTargetStreamer {
+protected:
+  MCContext &getContext() const { return Streamer.getContext(); }
+
 public:
   AMDGPUTargetStreamer(MCStreamer &S);
   virtual void EmitDirectiveHSACodeObjectVersion(uint32_t Major,
@@ -42,7 +46,9 @@ public:
 
   virtual void EmitAMDGPUHsaProgramScopeGlobal(StringRef GlobalName) = 0;
 
-  virtual void emitRuntimeMetadata(Module &M) = 0;
+  virtual void EmitRuntimeMetadata(Module &M) = 0;
+
+  virtual void EmitRuntimeMetadata(StringRef Metadata) = 0;
 };
 
 class AMDGPUTargetAsmStreamer : public AMDGPUTargetStreamer {
@@ -64,12 +70,18 @@ public:
 
   void EmitAMDGPUHsaProgramScopeGlobal(StringRef GlobalName) override;
 
-  void emitRuntimeMetadata(Module &M) override {}
+  void EmitRuntimeMetadata(Module &M) override;
+
+  void EmitRuntimeMetadata(StringRef Metadata) override;
 };
 
 class AMDGPUTargetELFStreamer : public AMDGPUTargetStreamer {
   MCStreamer &Streamer;
 
+  void EmitAMDGPUNote(const MCExpr* DescSize,
+                      AMDGPU::PT_NOTE::NoteType Type,
+                      std::function<void(MCELFStreamer &)> EmitDesc);
+
 public:
   AMDGPUTargetELFStreamer(MCStreamer &S);
 
@@ -90,7 +102,9 @@ public:
 
   void EmitAMDGPUHsaProgramScopeGlobal(StringRef GlobalName) override;
 
-  void emitRuntimeMetadata(Module &M) override;
+  void EmitRuntimeMetadata(Module &M) override;
+
+  void EmitRuntimeMetadata(StringRef Metadata) override;
 };
 
 }

test/MC/AMDGPU/hsa.s

@@ -14,6 +14,8 @@
 // ELF: 0020: 03000000 414D4400 04000700 07000000
 // ELF: 0030: 00000000 00000000 414D4400 414D4447
 // ELF: 0040: 50550000
+// We can't check binary representation of metadata note: it is different on
+// Windows and Linux because of carriage return on Windows
 
 // ELF: Symbol {
 // ELF: Name: amd_kernel_code_t_minimal
@@ -35,10 +37,29 @@
 .hsa_code_object_isa 7,0,0,"AMD","AMDGPU"
 // ASM: .hsa_code_object_isa 7,0,0,"AMD","AMDGPU"
 
+.amdgpu_runtime_metadata
+    {
+        amd.MDVersion: [ 2, 0 ]
+        amd.Kernels: [
+            { amd.KernelName: amd_kernel_code_t_test_all },
+            { amd.KernelName: amd_kernel_code_t_minimal }
+        ]
+    }
+.end_amdgpu_runtime_metadata
+
+// ASM: .amdgpu_runtime_metadata
+// ASM:     {
+// ASM:         amd.MDVersion: [ 2, 0 ]
+// ASM:         amd.Kernels: [
+// ASM:             { amd.KernelName: amd_kernel_code_t_test_all },
+// ASM:             { amd.KernelName: amd_kernel_code_t_minimal }
+// ASM:         ]
+// ASM:     }
+// ASM: .end_amdgpu_runtime_metadata
+
 .amdgpu_hsa_kernel amd_kernel_code_t_test_all
 .amdgpu_hsa_kernel amd_kernel_code_t_minimal
 
-
 amd_kernel_code_t_test_all:
 ; Test all amd_kernel_code_t members with non-default values.
 .amd_kernel_code_t

test/MC/AMDGPU/metadata.s

@@ -0,0 +1,35 @@
+// RUN: llvm-mc -triple amdgcn--amdhsa -mcpu=kaveri -show-encoding %s | FileCheck %s --check-prefix=ASM
+
+.amdgpu_runtime_metadata
+    { amd.MDVersion: [ 2, 0 ], amd.PrintfInfo: [ '1:1:4:%d\n', '2:1:8:%g\n' ], amd.Kernels:
+
+    - { amd.KernelName: test_char, amd.Language: OpenCL C, amd.LanguageVersion: [ 2, 0 ], amd.Args:
+        - { amd.ArgSize: 1, amd.ArgAlign: 1, amd.ArgKind: 0, amd.ArgValueType: 1, amd.ArgTypeName: char, amd.ArgAccQual: 0 }
+        - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 7, amd.ArgValueType: 9 }
+        - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 8, amd.ArgValueType: 9 }
+        - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 9, amd.ArgValueType: 9 }
+        - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 11, amd.ArgValueType: 1, amd.ArgAddrQual: 1 } }
+
+   - { amd.KernelName: test_ushort2, amd.Language: OpenCL C, amd.LanguageVersion: [ 2, 0 ], amd.Args:
+       - { amd.ArgSize: 4, amd.ArgAlign: 4, amd.ArgKind: 0, amd.ArgValueType: 4, amd.ArgTypeName: ushort2, amd.ArgAccQual: 0 }
+       - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 7, amd.ArgValueType: 9 }
+       - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 8, amd.ArgValueType: 9 }
+       - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 9, amd.ArgValueType: 9 }
+       - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 11, amd.ArgValueType: 1, amd.ArgAddrQual: 1 } }
+   }
+.end_amdgpu_runtime_metadata
+
+// ASM: { amd.MDVersion: [ 2, 0 ], amd.PrintfInfo: [ '1:1:4:%d\n', '2:1:8:%g\n' ], amd.Kernels:
+// ASM: - { amd.KernelName: test_char, amd.Language: OpenCL C, amd.LanguageVersion: [ 2, 0 ], amd.Args:
+// ASM:     - { amd.ArgSize: 1, amd.ArgAlign: 1, amd.ArgKind: 0, amd.ArgValueType: 1, amd.ArgTypeName: char, amd.ArgAccQual: 0 }
+// ASM:     - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 7, amd.ArgValueType: 9 }
+// ASM:     - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 8, amd.ArgValueType: 9 }
+// ASM:     - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 9, amd.ArgValueType: 9 }
+// ASM:     - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 11, amd.ArgValueType: 1, amd.ArgAddrQual: 1 } }
+// ASM: - { amd.KernelName: test_ushort2, amd.Language: OpenCL C, amd.LanguageVersion: [ 2, 0 ], amd.Args:
+// ASM:    - { amd.ArgSize: 4, amd.ArgAlign: 4, amd.ArgKind: 0, amd.ArgValueType: 4, amd.ArgTypeName: ushort2, amd.ArgAccQual: 0 }
+// ASM:    - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 7, amd.ArgValueType: 9 }
+// ASM:    - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 8, amd.ArgValueType: 9 }
+// ASM:    - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 9, amd.ArgValueType: 9 }
+// ASM:    - { amd.ArgSize: 8, amd.ArgAlign: 8, amd.ArgKind: 11, amd.ArgValueType: 1, amd.ArgAddrQual: 1 } }
+// ASM: }