1 files changed, 67 insertions, 0 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
index 57a4660bc1eb..13a65f1ad601 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -27,8 +27,10 @@
 #include "SIMachineFunctionInfo.h"
 #include "TargetInfo/AMDGPUTargetInfo.h"
 #include "Utils/AMDGPUBaseInfo.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/BinaryFormat/ELF.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
+#include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h"
 #include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCContext.h"
@@ -415,6 +417,10 @@ uint16_t AMDGPUAsmPrinter::getAmdhsaKernelCodeProperties(
         amdhsa::KERNEL_CODE_PROPERTY_ENABLE_WAVEFRONT_SIZE32;
   }
 
+  if (CurrentProgramInfo.DynamicCallStack) {
+    KernelCodeProperties |= amdhsa::KERNEL_CODE_PROPERTY_USES_DYNAMIC_STACK;
+  }
+
   return KernelCodeProperties;
 }
 
@@ -506,6 +512,9 @@ bool AMDGPUAsmPrinter::runOnMachineFunction(MachineFunction &MF) {
 
   emitFunctionBody();
 
+  emitResourceUsageRemarks(MF, CurrentProgramInfo, MFI->isModuleEntryFunction(),
+                           STM.hasMAIInsts());
+
   if (isVerbose()) {
     MCSectionELF *CommentSection =
         Context.getELFSection(".AMDGPU.csdata", ELF::SHT_PROGBITS, 0);
@@ -875,6 +884,9 @@ void AMDGPUAsmPrinter::getSIProgramInfo(SIProgramInfo &ProgInfo,
     LDSAlignShift = 9;
   }
 
+  ProgInfo.SGPRSpill = MFI->getNumSpilledSGPRs();
+  ProgInfo.VGPRSpill = MFI->getNumSpilledVGPRs();
+
   ProgInfo.LDSSize = MFI->getLDSSize();
   ProgInfo.LDSBlocks =
       alignTo(ProgInfo.LDSSize, 1ULL << LDSAlignShift) >> LDSAlignShift;
@@ -1180,3 +1192,58 @@ void AMDGPUAsmPrinter::getAnalysisUsage(AnalysisUsage &AU) const {
   AU.addPreserved<AMDGPUResourceUsageAnalysis>();
   AsmPrinter::getAnalysisUsage(AU);
 }
+
+void AMDGPUAsmPrinter::emitResourceUsageRemarks(
+    const MachineFunction &MF, const SIProgramInfo &CurrentProgramInfo,
+    bool isModuleEntryFunction, bool hasMAIInsts) {
+  if (!ORE)
+    return;
+
+  const char *Name = "kernel-resource-usage";
+  const char *Indent = "    ";
+
+  // If the remark is not specifically enabled, do not output to yaml
+  LLVMContext &Ctx = MF.getFunction().getContext();
+  if (!Ctx.getDiagHandlerPtr()->isAnalysisRemarkEnabled(Name))
+    return;
+
+  auto EmitResourceUsageRemark = [&](StringRef RemarkName,
+                                     StringRef RemarkLabel, auto Argument) {
+    // Add an indent for every line besides the line with the kernel name. This
+    // makes it easier to tell which resource usage go with which kernel since
+    // the kernel name will always be displayed first.
+    std::string LabelStr = RemarkLabel.str() + ": ";
+    if (!RemarkName.equals("FunctionName"))
+      LabelStr = Indent + LabelStr;
+
+    ORE->emit([&]() {
+      return MachineOptimizationRemarkAnalysis(Name, RemarkName,
+                                               MF.getFunction().getSubprogram(),
+                                               &MF.front())
+             << LabelStr << ore::NV(RemarkName, Argument);
+    });
+  };
+
+  // FIXME: Formatting here is pretty nasty because clang does not accept
+  // newlines from diagnostics. This forces us to emit multiple diagnostic
+  // remarks to simulate newlines. If and when clang does accept newlines, this
+  // formatting should be aggregated into one remark with newlines to avoid
+  // printing multiple diagnostic location and diag opts.
+  EmitResourceUsageRemark("FunctionName", "Function Name",
+                          MF.getFunction().getName());
+  EmitResourceUsageRemark("NumSGPR", "SGPRs", CurrentProgramInfo.NumSGPR);
+  EmitResourceUsageRemark("NumVGPR", "VGPRs", CurrentProgramInfo.NumArchVGPR);
+  if (hasMAIInsts)
+    EmitResourceUsageRemark("NumAGPR", "AGPRs", CurrentProgramInfo.NumAccVGPR);
+  EmitResourceUsageRemark("ScratchSize", "ScratchSize [bytes/lane]",
+                          CurrentProgramInfo.ScratchSize);
+  EmitResourceUsageRemark("Occupancy", "Occupancy [waves/SIMD]",
+                          CurrentProgramInfo.Occupancy);
+  EmitResourceUsageRemark("SGPRSpill", "SGPRs Spill",
+                          CurrentProgramInfo.SGPRSpill);
+  EmitResourceUsageRemark("VGPRSpill", "VGPRs Spill",
+                          CurrentProgramInfo.VGPRSpill);
+  if (isModuleEntryFunction)
+    EmitResourceUsageRemark("BytesLDS", "LDS Size [bytes/block]",
+                            CurrentProgramInfo.LDSSize);
+}