[AMDGPU] Emit kernel debug properties as code object metadata

Differential Revision: https://reviews.llvm.org/D30969

llvm-svn: 298558

llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUCodeObjectMetadata.h

@@ -280,23 +280,23 @@ struct Metadata final {
 namespace CodeProps {
 
 namespace Key {
-/// \brief Key for Kernel::CodeProps::mKernargSegmentSize.
+/// \brief Key for Kernel::CodeProps::Metadata::mKernargSegmentSize.
 constexpr char KernargSegmentSize[] = "KernargSegmentSize";
-/// \brief Key for Kernel::CodeProps::mWorkgroupGroupSegmentSize.
+/// \brief Key for Kernel::CodeProps::Metadata::mWorkgroupGroupSegmentSize.
 constexpr char WorkgroupGroupSegmentSize[] = "WorkgroupGroupSegmentSize";
-/// \brief Key for Kernel::CodeProps::mWorkitemPrivateSegmentSize.
+/// \brief Key for Kernel::CodeProps::Metadata::mWorkitemPrivateSegmentSize.
 constexpr char WorkitemPrivateSegmentSize[] = "WorkitemPrivateSegmentSize";
-/// \brief Key for Kernel::CodeProps::mWavefrontNumSGPRs.
+/// \brief Key for Kernel::CodeProps::Metadata::mWavefrontNumSGPRs.
 constexpr char WavefrontNumSGPRs[] = "WavefrontNumSGPRs";
-/// \brief Key for Kernel::CodeProps::mWorkitemNumVGPRs.
+/// \brief Key for Kernel::CodeProps::Metadata::mWorkitemNumVGPRs.
 constexpr char WorkitemNumVGPRs[] = "WorkitemNumVGPRs";
-/// \brief Key for Kernel::CodeProps::mKernargSegmentAlign.
+/// \brief Key for Kernel::CodeProps::Metadata::mKernargSegmentAlign.
 constexpr char KernargSegmentAlign[] = "KernargSegmentAlign";
-/// \brief Key for Kernel::CodeProps::mGroupSegmentAlign.
+/// \brief Key for Kernel::CodeProps::Metadata::mGroupSegmentAlign.
 constexpr char GroupSegmentAlign[] = "GroupSegmentAlign";
-/// \brief Key for Kernel::CodeProps::mPrivateSegmentAlign.
+/// \brief Key for Kernel::CodeProps::Metadata::mPrivateSegmentAlign.
 constexpr char PrivateSegmentAlign[] = "PrivateSegmentAlign";
-/// \brief Key for Kernel::CodeProps::mWavefrontSize.
+/// \brief Key for Kernel::CodeProps::Metadata::mWavefrontSize.
 constexpr char WavefrontSize[] = "WavefrontSize";
 } // end namespace Key
 
@@ -349,6 +349,63 @@ struct Metadata final {
 
 } // end namespace CodeProps
 
+//===----------------------------------------------------------------------===//
+// Kernel Debug Properties Metadata.
+//===----------------------------------------------------------------------===//
+namespace DebugProps {
+
+namespace Key {
+/// \brief Key for Kernel::DebugProps::Metadata::mDebuggerABIVersion.
+constexpr char DebuggerABIVersion[] = "DebuggerABIVersion";
+/// \brief Key for Kernel::DebugProps::Metadata::mReservedNumVGPRs.
+constexpr char ReservedNumVGPRs[] = "ReservedNumVGPRs";
+/// \brief Key for Kernel::DebugProps::Metadata::mReservedFirstVGPR.
+constexpr char ReservedFirstVGPR[] = "ReservedFirstVGPR";
+/// \brief Key for Kernel::DebugProps::Metadata::mPrivateSegmentBufferSGPR.
+constexpr char PrivateSegmentBufferSGPR[] = "PrivateSegmentBufferSGPR";
+/// \brief Key for
+///     Kernel::DebugProps::Metadata::mWavefrontPrivateSegmentOffsetSGPR.
+constexpr char WavefrontPrivateSegmentOffsetSGPR[] =
+    "WavefrontPrivateSegmentOffsetSGPR";
+} // end namespace Key
+
+/// \brief In-memory representation of kernel debug properties metadata.
+struct Metadata final {
+  /// \brief Debugger ABI version. Optional.
+  std::vector<uint32_t> mDebuggerABIVersion = std::vector<uint32_t>();
+  /// \brief Consecutive number of VGPRs reserved for debugger use. Must be 0 if
+  /// mDebuggerABIVersion is not set. Optional.
+  uint16_t mReservedNumVGPRs = 0;
+  /// \brief First fixed VGPR reserved. Must be uint16_t(-1) if
+  /// mDebuggerABIVersion is not set or mReservedFirstVGPR is 0. Optional.
+  uint16_t mReservedFirstVGPR = uint16_t(-1);
+  /// \brief Fixed SGPR of the first of 4 SGPRs used to hold the scratch V# used
+  /// for the entire kernel execution. Must be uint16_t(-1) if
+  /// mDebuggerABIVersion is not set or SGPR not used or not known. Optional.
+  uint16_t mPrivateSegmentBufferSGPR = uint16_t(-1);
+  /// \brief Fixed SGPR used to hold the wave scratch offset for the entire
+  /// kernel execution. Must be uint16_t(-1) if mDebuggerABIVersion is not set
+  /// or SGPR is not used or not known. Optional.
+  uint16_t mWavefrontPrivateSegmentOffsetSGPR = uint16_t(-1);
+
+  /// \brief Default constructor.
+  Metadata() = default;
+
+  /// \returns True if kernel debug properties metadata is empty, false
+  /// otherwise.
+  bool empty() const {
+    return !notEmpty();
+  }
+
+  /// \returns True if kernel debug properties metadata is not empty, false
+  /// otherwise.
+  bool notEmpty() const {
+    return !mDebuggerABIVersion.empty();
+  }
+};
+
+} // end namespace DebugProps
+
 namespace Key {
 /// \brief Key for Kernel::Metadata::mName.
 constexpr char Name[] = "Name";
@@ -362,6 +419,8 @@ constexpr char Attrs[] = "Attrs";
 constexpr char Args[] = "Args";
 /// \brief Key for Kernel::Metadata::mCodeProps.
 constexpr char CodeProps[] = "CodeProps";
+/// \brief Key for Kernel::Metadata::mDebugProps.
+constexpr char DebugProps[] = "DebugProps";
 } // end namespace Key
 
 /// \brief In-memory representation of kernel metadata.
@@ -378,6 +437,8 @@ struct Metadata final {
   std::vector<Arg::Metadata> mArgs = std::vector<Arg::Metadata>();
   /// \brief Code properties metadata. Optional.
   CodeProps::Metadata mCodeProps = CodeProps::Metadata();
+  /// \brief Debug properties metadata. Optional.
+  DebugProps::Metadata mDebugProps = DebugProps::Metadata();
 
   /// \brief Default constructor.
   Metadata() = default;

llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUCodeObjectMetadataStreamer.cpp

@@ -177,6 +177,22 @@ struct MappingTraits<Kernel::CodeProps::Metadata> {
   }
 };
 
+template <>
+struct MappingTraits<Kernel::DebugProps::Metadata> {
+  static void mapping(IO &YIO, Kernel::DebugProps::Metadata &MD) {
+    YIO.mapOptional(Kernel::DebugProps::Key::DebuggerABIVersion,
+                    MD.mDebuggerABIVersion, std::vector<uint32_t>());
+    YIO.mapOptional(Kernel::DebugProps::Key::ReservedNumVGPRs,
+                    MD.mReservedNumVGPRs, uint16_t(0));
+    YIO.mapOptional(Kernel::DebugProps::Key::ReservedFirstVGPR,
+                    MD.mReservedFirstVGPR, uint16_t(-1));
+    YIO.mapOptional(Kernel::DebugProps::Key::PrivateSegmentBufferSGPR,
+                    MD.mPrivateSegmentBufferSGPR, uint16_t(-1));
+    YIO.mapOptional(Kernel::DebugProps::Key::WavefrontPrivateSegmentOffsetSGPR,
+                    MD.mWavefrontPrivateSegmentOffsetSGPR, uint16_t(-1));
+  }
+};
+
 template <>
 struct MappingTraits<Kernel::Metadata> {
   static void mapping(IO &YIO, Kernel::Metadata &MD) {
@@ -190,6 +206,8 @@ struct MappingTraits<Kernel::Metadata> {
       YIO.mapOptional(Kernel::Key::Args, MD.mArgs);
     if (!MD.mCodeProps.empty() || !YIO.outputting())
       YIO.mapOptional(Kernel::Key::CodeProps, MD.mCodeProps);
+    if (!MD.mDebugProps.empty() || !YIO.outputting())
+      YIO.mapOptional(Kernel::Key::DebugProps, MD.mDebugProps);
   }
 };
 
@@ -574,6 +592,25 @@ void MetadataStreamer::emitKernelCodeProps(
   CodeProps.mWavefrontSize = KernelCode.wavefront_size;
 }
 
+void MetadataStreamer::emitKernelDebugProps(
+    const amd_kernel_code_t &KernelCode) {
+  if (!(KernelCode.code_properties & AMD_CODE_PROPERTY_IS_DEBUG_SUPPORTED))
+    return;
+
+  auto &DebugProps = CodeObjectMetadata.mKernels.back().mDebugProps;
+
+  // FIXME: Need to pass down debugger ABI version through features. This is ok
+  // for now because we only have one version.
+  DebugProps.mDebuggerABIVersion.push_back(1);
+  DebugProps.mDebuggerABIVersion.push_back(0);
+  DebugProps.mReservedNumVGPRs = KernelCode.reserved_vgpr_count;
+  DebugProps.mReservedFirstVGPR = KernelCode.reserved_vgpr_first;
+  DebugProps.mPrivateSegmentBufferSGPR =
+      KernelCode.debug_private_segment_buffer_sgpr;
+  DebugProps.mWavefrontPrivateSegmentOffsetSGPR =
+      KernelCode.debug_wavefront_private_segment_offset_sgpr;
+}
+
 void MetadataStreamer::begin(const FeatureBitset &Features, const Module &Mod) {
   emitVersion();
   emitIsa(Features);
@@ -593,6 +630,7 @@ void MetadataStreamer::emitKernel(const Function &Func,
   emitKernelAttrs(Func);
   emitKernelArgs(Func);
   emitKernelCodeProps(KernelCode);
+  emitKernelDebugProps(KernelCode);
 }
 
 ErrorOr<std::string> MetadataStreamer::toYamlString() {

llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUCodeObjectMetadataStreamer.h

@@ -76,6 +76,8 @@ private:
 
   void emitKernelCodeProps(const amd_kernel_code_t &KernelCode);
 
+  void emitKernelDebugProps(const amd_kernel_code_t &KernelCode);
+
 public:
   MetadataStreamer() = default;
   ~MetadataStreamer() = default;

llvm/test/CodeGen/AMDGPU/code-object-metadata-kernel-debug-props.ll

+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx700 -filetype=obj -o - < %s | llvm-readobj -amdgpu-code-object-metadata -elf-output-style=GNU -notes | FileCheck --check-prefix=CHECK --check-prefix=GFX700 --check-prefix=NOTES %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx800 -filetype=obj -o - < %s | llvm-readobj -amdgpu-code-object-metadata -elf-output-style=GNU -notes | FileCheck --check-prefix=CHECK --check-prefix=GFX800 --check-prefix=NOTES %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -filetype=obj -o - < %s | llvm-readobj -amdgpu-code-object-metadata -elf-output-style=GNU -notes | FileCheck --check-prefix=CHECK --check-prefix=GFX900 --check-prefix=NOTES %s
+
+declare void @llvm.dbg.declare(metadata, metadata, metadata)
+
+; CHECK: ---
+; CHECK:  Version: [ 1, 0 ]
+
+; CHECK:  Kernels:
+; CHECK:    - Name: test
+; CHECK:      DebugProps:
+; CHECK:        DebuggerABIVersion:                [ 1, 0 ]
+; CHECK:        ReservedNumVGPRs:                  4
+; CHECK:        ReservedFirstVGPR:                 11
+; CHECK:        PrivateSegmentBufferSGPR:          0
+; CHECK:        WavefrontPrivateSegmentOffsetSGPR: 11
+define amdgpu_kernel void @test(i32 addrspace(1)* %A) #0 !dbg !7 !kernel_arg_addr_space !12 !kernel_arg_access_qual !13 !kernel_arg_type !14 !kernel_arg_base_type !14 !kernel_arg_type_qual !15 {
+entry:
+  %A.addr = alloca i32 addrspace(1)*, align 4
+  store i32 addrspace(1)* %A, i32 addrspace(1)** %A.addr, align 4
+  call void @llvm.dbg.declare(metadata i32 addrspace(1)** %A.addr, metadata !16, metadata !17), !dbg !18
+  %0 = load i32 addrspace(1)*, i32 addrspace(1)** %A.addr, align 4, !dbg !19
+  %arrayidx = getelementptr inbounds i32, i32 addrspace(1)* %0, i64 0, !dbg !19
+  store i32 777, i32 addrspace(1)* %arrayidx, align 4, !dbg !20
+  %1 = load i32 addrspace(1)*, i32 addrspace(1)** %A.addr, align 4, !dbg !21
+  %arrayidx1 = getelementptr inbounds i32, i32 addrspace(1)* %1, i64 1, !dbg !21
+  store i32 888, i32 addrspace(1)* %arrayidx1, align 4, !dbg !22
+  %2 = load i32 addrspace(1)*, i32 addrspace(1)** %A.addr, align 4, !dbg !23
+  %arrayidx2 = getelementptr inbounds i32, i32 addrspace(1)* %2, i64 2, !dbg !23
+  store i32 999, i32 addrspace(1)* %arrayidx2, align 4, !dbg !24
+  ret void, !dbg !25
+}
+
+attributes #0 = { noinline nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="gfx800" "target-features"="+16-bit-insts,+amdgpu-debugger-emit-prologue,+amdgpu-debugger-insert-nops,+amdgpu-debugger-reserve-regs,+dpp,+fp64-fp16-denormals,+s-memrealtime,-fp32-denormals" "unsafe-fp-math"="false" "use-soft-float"="false" }
+
+!llvm.dbg.cu = !{!0}
+!opencl.ocl.version = !{!3}
+!llvm.module.flags = !{!4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 5.0.0", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2)
+!1 = !DIFile(filename: "code-object-metadata-kernel-debug-props.cl", directory: "/some/random/directory")
+!2 = !{}
+!3 = !{i32 1, i32 0}
+!4 = !{i32 2, !"Dwarf Version", i32 2}
+!5 = !{i32 2, !"Debug Info Version", i32 3}
+!6 = !{!"clang version 5.0.0"}
+!7 = distinct !DISubprogram(name: "test", scope: !1, file: !1, line: 1, type: !8, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: false, unit: !0, variables: !2)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10}
+!10 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{i32 1}
+!13 = !{!"none"}
+!14 = !{!"int*"}
+!15 = !{!""}
+!16 = !DILocalVariable(name: "A", arg: 1, scope: !7, file: !1, line: 1, type: !10)
+!17 = !DIExpression(DW_OP_constu, 1, DW_OP_swap, DW_OP_xderef)
+!18 = !DILocation(line: 1, column: 30, scope: !7)
+!19 = !DILocation(line: 2, column: 3, scope: !7)
+!20 = !DILocation(line: 2, column: 8, scope: !7)
+!21 = !DILocation(line: 3, column: 3, scope: !7)
+!22 = !DILocation(line: 3, column: 8, scope: !7)
+!23 = !DILocation(line: 4, column: 3, scope: !7)
+!24 = !DILocation(line: 4, column: 8, scope: !7)
+!25 = !DILocation(line: 5, column: 1, scope: !7)

llvm/test/MC/AMDGPU/code-object-metadata-kernel-code-props.s

@@ -7,10 +7,10 @@
 // CHECK:    Kernels:
 // CHECK:      - Name: test_kernel
 // CHECK:        CodeProps:
-// CHECK:        KernargSegmentSize:         24
-// CHECK:        WorkitemPrivateSegmentSize: 16
-// CHECK:        WavefrontNumSGPRs:          6
-// CHECK:        WorkitemNumVGPRs:           12
+// CHECK:          KernargSegmentSize:         24
+// CHECK:          WorkitemPrivateSegmentSize: 16
+// CHECK:          WavefrontNumSGPRs:          6
+// CHECK:          WorkitemNumVGPRs:           12
 .amdgpu_code_object_metadata
   Version: [ 1, 0 ]
   Printf: [ '1:1:4:%d\n', '2:1:8:%g\n' ]

llvm/test/MC/AMDGPU/code-object-metadata-kernel-debug-props.s

+// RUN: llvm-mc -triple=amdgcn-amd-amdhsa -mcpu=gfx700 -show-encoding %s | FileCheck --check-prefix=CHECK --check-prefix=GFX700 %s
+// RUN: llvm-mc -triple=amdgcn-amd-amdhsa -mcpu=gfx800 -show-encoding %s | FileCheck --check-prefix=CHECK --check-prefix=GFX800 %s
+// RUN: llvm-mc -triple=amdgcn-amd-amdhsa -mcpu=gfx900 -show-encoding %s | FileCheck --check-prefix=CHECK --check-prefix=GFX900 %s
+
+// CHECK:  .amdgpu_code_object_metadata
+// CHECK:    Version: [ 1, 0 ]
+// CHECK:    Kernels:
+// CHECK:      - Name: test_kernel
+// CHECK:        DebugProps:
+// CHECK:          DebuggerABIVersion:                [ 1, 0 ]
+// CHECK:          ReservedNumVGPRs:                  4
+// CHECK:          ReservedFirstVGPR:                 11
+// CHECK:          PrivateSegmentBufferSGPR:          0
+// CHECK:          WavefrontPrivateSegmentOffsetSGPR: 11
+.amdgpu_code_object_metadata
+  Version: [ 1, 0 ]
+  Printf: [ '1:1:4:%d\n', '2:1:8:%g\n' ]
+  Kernels:
+    - Name:            test_kernel
+      DebugProps:
+        DebuggerABIVersion:                [ 1, 0 ]
+        ReservedNumVGPRs:                  4
+        ReservedFirstVGPR:                 11
+        PrivateSegmentBufferSGPR:          0
+        WavefrontPrivateSegmentOffsetSGPR: 11
+.end_amdgpu_code_object_metadata
\ No newline at end of file