Revert "[AMDGPU] Remove Code Object V3 (#67118)"

This reverts commit 544d91280c26fd5f7acd70eac4d667863562f4cc.

8 files changed, 110 insertions, 40 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
index 8d0ef67..aadc4a6 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -341,6 +341,9 @@ bool AMDGPUAsmPrinter::doInitialization(Module &M) {
 
   if (TM.getTargetTriple().getOS() == Triple::AMDHSA) {
     switch (CodeObjectVersion) {
+    case AMDGPU::AMDHSA_COV3:
+      HSAMetadataStream.reset(new HSAMD::MetadataStreamerMsgPackV3());
+      break;
     case AMDGPU::AMDHSA_COV4:
       HSAMetadataStream.reset(new HSAMD::MetadataStreamerMsgPackV4());
       break;
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
index b51a876..5060cd3 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
@@ -49,14 +49,14 @@ namespace AMDGPU {
 namespace HSAMD {
 
 //===----------------------------------------------------------------------===//
-// HSAMetadataStreamerV4
+// HSAMetadataStreamerV3
 //===----------------------------------------------------------------------===//
 
-void MetadataStreamerMsgPackV4::dump(StringRef HSAMetadataString) const {
+void MetadataStreamerMsgPackV3::dump(StringRef HSAMetadataString) const {
   errs() << "AMDGPU HSA Metadata:\n" << HSAMetadataString << '\n';
 }
 
-void MetadataStreamerMsgPackV4::verify(StringRef HSAMetadataString) const {
+void MetadataStreamerMsgPackV3::verify(StringRef HSAMetadataString) const {
   errs() << "AMDGPU HSA Metadata Parser Test: ";
 
   msgpack::Document FromHSAMetadataString;
@@ -78,7 +78,7 @@ void MetadataStreamerMsgPackV4::verify(StringRef HSAMetadataString) const {
 }
 
 std::optional<StringRef>
-MetadataStreamerMsgPackV4::getAccessQualifier(StringRef AccQual) const {
+MetadataStreamerMsgPackV3::getAccessQualifier(StringRef AccQual) const {
   return StringSwitch<std::optional<StringRef>>(AccQual)
       .Case("read_only", StringRef("read_only"))
       .Case("write_only", StringRef("write_only"))
@@ -86,7 +86,7 @@ MetadataStreamerMsgPackV4::getAccessQualifier(StringRef AccQual) const {
       .Default(std::nullopt);
 }
 
-std::optional<StringRef> MetadataStreamerMsgPackV4::getAddressSpaceQualifier(
+std::optional<StringRef> MetadataStreamerMsgPackV3::getAddressSpaceQualifier(
     unsigned AddressSpace) const {
   switch (AddressSpace) {
   case AMDGPUAS::PRIVATE_ADDRESS:
@@ -107,7 +107,7 @@ std::optional<StringRef> MetadataStreamerMsgPackV4::getAddressSpaceQualifier(
 }
 
 StringRef
-MetadataStreamerMsgPackV4::getValueKind(Type *Ty, StringRef TypeQual,
+MetadataStreamerMsgPackV3::getValueKind(Type *Ty, StringRef TypeQual,
                                         StringRef BaseTypeName) const {
   if (TypeQual.contains("pipe"))
     return "pipe";
@@ -134,7 +134,7 @@ MetadataStreamerMsgPackV4::getValueKind(Type *Ty, StringRef TypeQual,
                    : "by_value");
 }
 
-std::string MetadataStreamerMsgPackV4::getTypeName(Type *Ty,
+std::string MetadataStreamerMsgPackV3::getTypeName(Type *Ty,
                                                    bool Signed) const {
   switch (Ty->getTypeID()) {
   case Type::IntegerTyID: {
@@ -173,7 +173,7 @@ std::string MetadataStreamerMsgPackV4::getTypeName(Type *Ty,
 }
 
 msgpack::ArrayDocNode
-MetadataStreamerMsgPackV4::getWorkGroupDimensions(MDNode *Node) const {
+MetadataStreamerMsgPackV3::getWorkGroupDimensions(MDNode *Node) const {
   auto Dims = HSAMetadataDoc->getArrayNode();
   if (Node->getNumOperands() != 3)
     return Dims;
@@ -184,20 +184,14 @@ MetadataStreamerMsgPackV4::getWorkGroupDimensions(MDNode *Node) const {
   return Dims;
 }
 
-void MetadataStreamerMsgPackV4::emitVersion() {
+void MetadataStreamerMsgPackV3::emitVersion() {
   auto Version = HSAMetadataDoc->getArrayNode();
-  Version.push_back(Version.getDocument()->getNode(VersionMajorV4));
-  Version.push_back(Version.getDocument()->getNode(VersionMinorV4));
+  Version.push_back(Version.getDocument()->getNode(VersionMajorV3));
+  Version.push_back(Version.getDocument()->getNode(VersionMinorV3));
   getRootMetadata("amdhsa.version") = Version;
 }
 
-void MetadataStreamerMsgPackV4::emitTargetID(
-    const IsaInfo::AMDGPUTargetID &TargetID) {
-  getRootMetadata("amdhsa.target") =
-      HSAMetadataDoc->getNode(TargetID.toString(), /*Copy=*/true);
-}
-
-void MetadataStreamerMsgPackV4::emitPrintf(const Module &Mod) {
+void MetadataStreamerMsgPackV3::emitPrintf(const Module &Mod) {
   auto Node = Mod.getNamedMetadata("llvm.printf.fmts");
   if (!Node)
     return;
@@ -210,7 +204,7 @@ void MetadataStreamerMsgPackV4::emitPrintf(const Module &Mod) {
   getRootMetadata("amdhsa.printf") = Printf;
 }
 
-void MetadataStreamerMsgPackV4::emitKernelLanguage(const Function &Func,
+void MetadataStreamerMsgPackV3::emitKernelLanguage(const Function &Func,
                                                    msgpack::MapDocNode Kern) {
   // TODO: What about other languages?
   auto Node = Func.getParent()->getNamedMetadata("opencl.ocl.version");
@@ -229,7 +223,7 @@ void MetadataStreamerMsgPackV4::emitKernelLanguage(const Function &Func,
   Kern[".language_version"] = LanguageVersion;
 }
 
-void MetadataStreamerMsgPackV4::emitKernelAttrs(const Function &Func,
+void MetadataStreamerMsgPackV3::emitKernelAttrs(const Function &Func,
                                                 msgpack::MapDocNode Kern) {
 
   if (auto Node = Func.getMetadata("reqd_work_group_size"))
@@ -254,7 +248,7 @@ void MetadataStreamerMsgPackV4::emitKernelAttrs(const Function &Func,
     Kern[".kind"] = Kern.getDocument()->getNode("fini");
 }
 
-void MetadataStreamerMsgPackV4::emitKernelArgs(const MachineFunction &MF,
+void MetadataStreamerMsgPackV3::emitKernelArgs(const MachineFunction &MF,
                                                msgpack::MapDocNode Kern) {
   auto &Func = MF.getFunction();
   unsigned Offset = 0;
@@ -267,7 +261,7 @@ void MetadataStreamerMsgPackV4::emitKernelArgs(const MachineFunction &MF,
   Kern[".args"] = Args;
 }
 
-void MetadataStreamerMsgPackV4::emitKernelArg(const Argument &Arg,
+void MetadataStreamerMsgPackV3::emitKernelArg(const Argument &Arg,
                                               unsigned &Offset,
                                               msgpack::ArrayDocNode Args) {
   auto Func = Arg.getParent();
@@ -332,7 +326,7 @@ void MetadataStreamerMsgPackV4::emitKernelArg(const Argument &Arg,
                 AccQual, TypeQual);
 }
 
-void MetadataStreamerMsgPackV4::emitKernelArg(
+void MetadataStreamerMsgPackV3::emitKernelArg(
     const DataLayout &DL, Type *Ty, Align Alignment, StringRef ValueKind,
     unsigned &Offset, msgpack::ArrayDocNode Args, MaybeAlign PointeeAlign,
     StringRef Name, StringRef TypeName, StringRef BaseTypeName,
@@ -381,7 +375,7 @@ void MetadataStreamerMsgPackV4::emitKernelArg(
   Args.push_back(Arg);
 }
 
-void MetadataStreamerMsgPackV4::emitHiddenKernelArgs(
+void MetadataStreamerMsgPackV3::emitHiddenKernelArgs(
     const MachineFunction &MF, unsigned &Offset, msgpack::ArrayDocNode Args) {
   auto &Func = MF.getFunction();
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
@@ -454,10 +448,9 @@ void MetadataStreamerMsgPackV4::emitHiddenKernelArgs(
   }
 }
 
-msgpack::MapDocNode
-MetadataStreamerMsgPackV4::getHSAKernelProps(const MachineFunction &MF,
-                                             const SIProgramInfo &ProgramInfo,
-                                             unsigned CodeObjectVersion) const {
+msgpack::MapDocNode MetadataStreamerMsgPackV3::getHSAKernelProps(
+    const MachineFunction &MF, const SIProgramInfo &ProgramInfo,
+    unsigned CodeObjectVersion) const {
   const GCNSubtarget &STM = MF.getSubtarget<GCNSubtarget>();
   const SIMachineFunctionInfo &MFI = *MF.getInfo<SIMachineFunctionInfo>();
   const Function &F = MF.getFunction();
@@ -502,19 +495,18 @@ MetadataStreamerMsgPackV4::getHSAKernelProps(const MachineFunction &MF,
   return Kern;
 }
 
-bool MetadataStreamerMsgPackV4::emitTo(AMDGPUTargetStreamer &TargetStreamer) {
+bool MetadataStreamerMsgPackV3::emitTo(AMDGPUTargetStreamer &TargetStreamer) {
   return TargetStreamer.EmitHSAMetadata(*HSAMetadataDoc, true);
 }
 
-void MetadataStreamerMsgPackV4::begin(const Module &Mod,
+void MetadataStreamerMsgPackV3::begin(const Module &Mod,
                                       const IsaInfo::AMDGPUTargetID &TargetID) {
   emitVersion();
-  emitTargetID(TargetID);
   emitPrintf(Mod);
   getRootMetadata("amdhsa.kernels") = HSAMetadataDoc->getArrayNode();
 }
 
-void MetadataStreamerMsgPackV4::end() {
+void MetadataStreamerMsgPackV3::end() {
   std::string HSAMetadataString;
   raw_string_ostream StrOS(HSAMetadataString);
   HSAMetadataDoc->toYAML(StrOS);
@@ -525,7 +517,7 @@ void MetadataStreamerMsgPackV4::end() {
     verify(StrOS.str());
 }
 
-void MetadataStreamerMsgPackV4::emitKernel(const MachineFunction &MF,
+void MetadataStreamerMsgPackV3::emitKernel(const MachineFunction &MF,
                                            const SIProgramInfo &ProgramInfo) {
   auto &Func = MF.getFunction();
   if (Func.getCallingConv() != CallingConv::AMDGPU_KERNEL &&
@@ -551,6 +543,31 @@ void MetadataStreamerMsgPackV4::emitKernel(const MachineFunction &MF,
 }
 
 //===----------------------------------------------------------------------===//
+// HSAMetadataStreamerV4
+//===----------------------------------------------------------------------===//
+
+void MetadataStreamerMsgPackV4::emitVersion() {
+  auto Version = HSAMetadataDoc->getArrayNode();
+  Version.push_back(Version.getDocument()->getNode(VersionMajorV4));
+  Version.push_back(Version.getDocument()->getNode(VersionMinorV4));
+  getRootMetadata("amdhsa.version") = Version;
+}
+
+void MetadataStreamerMsgPackV4::emitTargetID(
+    const IsaInfo::AMDGPUTargetID &TargetID) {
+  getRootMetadata("amdhsa.target") =
+      HSAMetadataDoc->getNode(TargetID.toString(), /*Copy=*/true);
+}
+
+void MetadataStreamerMsgPackV4::begin(const Module &Mod,
+                                      const IsaInfo::AMDGPUTargetID &TargetID) {
+  emitVersion();
+  emitTargetID(TargetID);
+  emitPrintf(Mod);
+  getRootMetadata("amdhsa.kernels") = HSAMetadataDoc->getArrayNode();
+}
+
+//===----------------------------------------------------------------------===//
 // HSAMetadataStreamerV5
 //===----------------------------------------------------------------------===//
 
@@ -663,7 +680,7 @@ void MetadataStreamerMsgPackV5::emitHiddenKernelArgs(
 
 void MetadataStreamerMsgPackV5::emitKernelAttrs(const Function &Func,
                                                 msgpack::MapDocNode Kern) {
-  MetadataStreamerMsgPackV4::emitKernelAttrs(Func, Kern);
+  MetadataStreamerMsgPackV3::emitKernelAttrs(Func, Kern);
 
   if (Func.getFnAttribute("uniform-work-group-size").getValueAsBool())
     Kern[".uniform_work_group_size"] = Kern.getDocument()->getNode(1);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.h b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.h
index 18a7b5d..d2b3b89 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.h
@@ -62,7 +62,7 @@ protected:
                                msgpack::MapDocNode Kern) = 0;
 };
 
-class MetadataStreamerMsgPackV4 : public MetadataStreamer {
+class MetadataStreamerMsgPackV3 : public MetadataStreamer {
 protected:
   std::unique_ptr<msgpack::Document> HSAMetadataDoc =
       std::make_unique<msgpack::Document>();
@@ -89,8 +89,6 @@ protected:
 
   void emitVersion() override;
 
-  void emitTargetID(const IsaInfo::AMDGPUTargetID &TargetID);
-
   void emitPrintf(const Module &Mod);
 
   void emitKernelLanguage(const Function &Func, msgpack::MapDocNode Kern);
@@ -122,8 +120,8 @@ protected:
   }
 
 public:
-  MetadataStreamerMsgPackV4() = default;
-  ~MetadataStreamerMsgPackV4() = default;
+  MetadataStreamerMsgPackV3() = default;
+  ~MetadataStreamerMsgPackV3() = default;
 
   bool emitTo(AMDGPUTargetStreamer &TargetStreamer) override;
 
@@ -136,6 +134,19 @@ public:
                   const SIProgramInfo &ProgramInfo) override;
 };
 
+class MetadataStreamerMsgPackV4 : public MetadataStreamerMsgPackV3 {
+protected:
+  void emitVersion() override;
+  void emitTargetID(const IsaInfo::AMDGPUTargetID &TargetID);
+
+public:
+  MetadataStreamerMsgPackV4() = default;
+  ~MetadataStreamerMsgPackV4() = default;
+
+  void begin(const Module &Mod,
+             const IsaInfo::AMDGPUTargetID &TargetID) override;
+};
+
 class MetadataStreamerMsgPackV5 final : public MetadataStreamerMsgPackV4 {
 protected:
   void emitVersion() override;
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 21abfb4..3d70ed1 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -6478,6 +6478,11 @@ bool AMDGPULegalizerInfo::legalizeTrapIntrinsic(MachineInstr &MI,
       ST.getTrapHandlerAbi() != GCNSubtarget::TrapHandlerAbi::AMDHSA)
     return legalizeTrapEndpgm(MI, MRI, B);
 
+  const Module *M = B.getMF().getFunction().getParent();
+  unsigned CodeObjectVersion = AMDGPU::getCodeObjectVersion(*M);
+  if (CodeObjectVersion <= AMDGPU::AMDHSA_COV3)
+    return legalizeTrapHsaQueuePtr(MI, MRI, B);
+
   return ST.supportsGetDoorbellID() ?
          legalizeTrapHsa(MI, MRI, B) : legalizeTrapHsaQueuePtr(MI, MRI, B);
 }
diff --git a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
index 42af09e2..6b8c03c 100644
--- a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
+++ b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp
@@ -424,6 +424,7 @@ void AMDGPUTargetAsmStreamer::EmitAmdhsaKernelDescriptor(
   switch (CodeObjectVersion) {
   default:
     break;
+  case AMDGPU::AMDHSA_COV3:
   case AMDGPU::AMDHSA_COV4:
   case AMDGPU::AMDHSA_COV5:
     if (getTargetID()->isXnackSupported())
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 9c5b166..f8f1e6d 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -5992,6 +5992,11 @@ SDValue SITargetLowering::lowerTRAP(SDValue Op, SelectionDAG &DAG) const {
       Subtarget->getTrapHandlerAbi() != GCNSubtarget::TrapHandlerAbi::AMDHSA)
     return lowerTrapEndpgm(Op, DAG);
 
+  const Module *M = DAG.getMachineFunction().getFunction().getParent();
+  unsigned CodeObjectVersion = AMDGPU::getCodeObjectVersion(*M);
+  if (CodeObjectVersion <= AMDGPU::AMDHSA_COV3)
+    return lowerTrapHsaQueuePtr(Op, DAG);
+
   return Subtarget->supportsGetDoorbellID() ? lowerTrapHsa(Op, DAG) :
          lowerTrapHsaQueuePtr(Op, DAG);
 }
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
index 5fff19e..d123b38 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
@@ -128,6 +128,8 @@ std::optional<uint8_t> getHsaAbiVersion(const MCSubtargetInfo *STI) {
     return std::nullopt;
 
   switch (AmdhsaCodeObjectVersion) {
+  case 3:
+    return ELF::ELFABIVERSION_AMDGPU_HSA_V3;
   case 4:
     return ELF::ELFABIVERSION_AMDGPU_HSA_V4;
   case 5:
@@ -138,6 +140,12 @@ std::optional<uint8_t> getHsaAbiVersion(const MCSubtargetInfo *STI) {
   }
 }
 
+bool isHsaAbiVersion3(const MCSubtargetInfo *STI) {
+  if (std::optional<uint8_t> HsaAbiVer = getHsaAbiVersion(STI))
+    return *HsaAbiVer == ELF::ELFABIVERSION_AMDGPU_HSA_V3;
+  return false;
+}
+
 bool isHsaAbiVersion4(const MCSubtargetInfo *STI) {
   if (std::optional<uint8_t> HsaAbiVer = getHsaAbiVersion(STI))
     return *HsaAbiVer == ELF::ELFABIVERSION_AMDGPU_HSA_V4;
@@ -166,6 +174,7 @@ unsigned getCodeObjectVersion(const Module &M) {
 
 unsigned getMultigridSyncArgImplicitArgPosition(unsigned CodeObjectVersion) {
   switch (CodeObjectVersion) {
+  case AMDHSA_COV3:
   case AMDHSA_COV4:
     return 48;
   case AMDHSA_COV5:
@@ -179,6 +188,7 @@ unsigned getMultigridSyncArgImplicitArgPosition(unsigned CodeObjectVersion) {
 // central TD file.
 unsigned getHostcallImplicitArgPosition(unsigned CodeObjectVersion) {
   switch (CodeObjectVersion) {
+  case AMDHSA_COV3:
   case AMDHSA_COV4:
     return 24;
   case AMDHSA_COV5:
@@ -189,6 +199,7 @@ unsigned getHostcallImplicitArgPosition(unsigned CodeObjectVersion) {
 
 unsigned getDefaultQueueImplicitArgPosition(unsigned CodeObjectVersion) {
   switch (CodeObjectVersion) {
+  case AMDHSA_COV3:
   case AMDHSA_COV4:
     return 32;
   case AMDHSA_COV5:
@@ -199,6 +210,7 @@ unsigned getDefaultQueueImplicitArgPosition(unsigned CodeObjectVersion) {
 
 unsigned getCompletionActionImplicitArgPosition(unsigned CodeObjectVersion) {
   switch (CodeObjectVersion) {
+  case AMDHSA_COV3:
   case AMDHSA_COV4:
     return 40;
   case AMDHSA_COV5:
@@ -762,6 +774,15 @@ std::string AMDGPUTargetID::toString() const {
   std::string Features;
   if (STI.getTargetTriple().getOS() == Triple::AMDHSA) {
     switch (CodeObjectVersion) {
+    case AMDGPU::AMDHSA_COV3:
+      // xnack.
+      if (isXnackOnOrAny())
+        Features += "+xnack";
+      // In code object v2 and v3, "sramecc" feature was spelled with a
+      // hyphen ("sram-ecc").
+      if (isSramEccOnOrAny())
+        Features += "+sram-ecc";
+      break;
     case AMDGPU::AMDHSA_COV4:
     case AMDGPU::AMDHSA_COV5:
       // sramecc.
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
index 1e0994d..bb2964f 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -42,12 +42,19 @@ namespace AMDGPU {
 
 struct IsaVersion;
 
-enum { AMDHSA_COV4 = 4, AMDHSA_COV5 = 5 };
+enum {
+  AMDHSA_COV3 = 3,
+  AMDHSA_COV4 = 4,
+  AMDHSA_COV5 = 5
+};
 
 /// \returns True if \p STI is AMDHSA.
 bool isHsaAbi(const MCSubtargetInfo &STI);
 /// \returns HSA OS ABI Version identification.
 std::optional<uint8_t> getHsaAbiVersion(const MCSubtargetInfo *STI);
+/// \returns True if HSA OS ABI Version identification is 3,
+/// false otherwise.
+bool isHsaAbiVersion3(const MCSubtargetInfo *STI);
 /// \returns True if HSA OS ABI Version identification is 4,
 /// false otherwise.
 bool isHsaAbiVersion4(const MCSubtargetInfo *STI);