[Remarks] Refactor remark diagnostic emission in a RemarkStreamer

[thegameg]

Summary

This allows us to store more info about where we're emitting the remarks without cluttering LLVMContext.

[JDevlieghere]

clang/lib/CodeGen/CodeGenAction.cpp

279

This doesn't seem to "set a remark streamer", but rather create one. Shouldn't the method name reflect that, maybe something like createRemarkStreamer? Or is this a pattern used in the LLVMContext?

llvm/include/llvm/IR/LLVMContext.h

258

You should probably mention that if a RemarkStreamer is set, it will be replaced.

llvm/include/llvm/IR/RemarkStreamer.h

1

Old license

llvm/lib/IR/RemarkStreamer.cpp

1

Old license

[thegameg]

Thanks Jonas!

[JDevlieghere]

LGTM

clang/lib/CodeGen/CodeGenAction.cpp

void HandleTranslationUnit(ASTContext &C) override {
         OptRecordFile = llvm::make_unique<llvm::ToolOutputFile>(
             CodeGenOpts.OptRecordFile, EC, sys::fs::F_None);
         if (EC) {
           Diags.Report(diag::err_cannot_open_file) <<
             CodeGenOpts.OptRecordFile << EC.message();
           return;
         }
 
-        Ctx.setDiagnosticsOutputFile(
-            llvm::make_unique<yaml::Output>(OptRecordFile->os()));
+        Ctx.setRemarkStreamer(CodeGenOpts.OptRecordFile, OptRecordFile->os());
 
         if (CodeGenOpts.getProfileUse() != CodeGenOptions::ProfileNone)
           Ctx.setDiagnosticsHotnessRequested(true);
       }
 
       // Link each LinkModule into our module.
       if (LinkInModules())
         return;

llvm/include/llvm/Analysis/OptimizationRemarkEmitter.h

public:
   /// Take a lambda that returns a remark which will be emitted.  Second
   /// argument is only used to restrict this to functions.
   template <typename T>
   void emit(T RemarkBuilder, decltype(RemarkBuilder()) * = nullptr) {
     // Avoid building the remark unless we know there are at least *some*
     // remarks enabled. We can't currently check whether remarks are requested
     // for the calling pass since that requires actually building the remark.
 
-    if (F->getContext().getDiagnosticsOutputFile() ||
+    if (F->getContext().getRemarkStreamer() ||
         F->getContext().getDiagHandlerPtr()->isAnyRemarkEnabled()) {
       auto R = RemarkBuilder();
       emit((DiagnosticInfoOptimizationBase &)R);
     }
   }
 
   /// Whether we allow for extra compile-time budget to perform more
   /// analysis to produce fewer false positives.
   ///
   /// This is useful when reporting missed optimizations.  In this case we can
   /// use the extra analysis (1) to filter trivial false positives or (2) to
   /// provide more context so that non-trivial false positives can be quickly
   /// detected by the user.
   bool allowExtraAnalysis(StringRef PassName) const {
-    return (F->getContext().getDiagnosticsOutputFile() ||
+    return (F->getContext().getRemarkStreamer() ||
             F->getContext().getDiagHandlerPtr()->isAnyRemarkEnabled(PassName));
   }
 
 private:
   const Function *F;
 
   BlockFrequencyInfo *BFI;
 

llvm/include/llvm/CodeGen/MachineOptimizationRemarkEmitter.h

public:
   /// Whether we allow for extra compile-time budget to perform more
   /// analysis to be more informative.
   ///
   /// This is useful to enable additional missed optimizations to be reported
   /// that are normally too noisy.  In this mode, we can use the extra analysis
   /// (1) to filter trivial false positives or (2) to provide more context so
   /// that non-trivial false positives can be quickly detected by the user.
   bool allowExtraAnalysis(StringRef PassName) const {
-    return (MF.getFunction().getContext().getDiagnosticsOutputFile() ||
-            MF.getFunction().getContext()
-            .getDiagHandlerPtr()->isAnyRemarkEnabled(PassName));
+    return (
+        MF.getFunction().getContext().getRemarkStreamer() ||
+        MF.getFunction().getContext().getDiagHandlerPtr()->isAnyRemarkEnabled(
+            PassName));
   }
 
   /// Take a lambda that returns a remark which will be emitted.  Second
   /// argument is only used to restrict this to functions.
   template <typename T>
   void emit(T RemarkBuilder, decltype(RemarkBuilder()) * = nullptr) {
     // Avoid building the remark unless we know there are at least *some*
     // remarks enabled. We can't currently check whether remarks are requested
     // for the calling pass since that requires actually building the remark.
 
-    if (MF.getFunction().getContext().getDiagnosticsOutputFile() ||
-        MF.getFunction().getContext().getDiagHandlerPtr()->isAnyRemarkEnabled()) {
+    if (MF.getFunction().getContext().getRemarkStreamer() ||
+        MF.getFunction()
+            .getContext()
+            .getDiagHandlerPtr()
+            ->isAnyRemarkEnabled()) {
       auto R = RemarkBuilder();
       emit((DiagnosticInfoOptimizationBase &)R);
     }
   }
 
 private:
   MachineFunction &MF;
 

llvm/include/llvm/IR/LLVMContext.h

 class Instruction;
 class LLVMContextImpl;
 class Module;
 class OptPassGate;
 template <typename T> class SmallVectorImpl;
 class SMDiagnostic;
 class StringRef;
 class Twine;
-
-namespace yaml {
-
-class Output;
-
-} // end namespace yaml
+class RemarkStreamer;
+class raw_ostream;
 
 namespace SyncScope {
 
 typedef uint8_t ID;
 
 /// Known synchronization scope IDs, which always have the same value.  All
 /// synchronization scope IDs that LLVM has special knowledge of are listed
 /// here.  Additionally, this scheme allows LLVM to efficiently check for
public:
   /// to be included in optimization diagnostics. If there is no minimum, this
   /// returns None.
   uint64_t getDiagnosticsHotnessThreshold() const;
 
   /// Set the minimum hotness value a diagnostic needs in order to be
   /// included in optimization diagnostics.
   void setDiagnosticsHotnessThreshold(uint64_t Threshold);
 
-  /// Return the YAML file used by the backend to save optimization
-  /// diagnostics.  If null, diagnostics are not saved in a file but only
-  /// emitted via the diagnostic handler.
-  yaml::Output *getDiagnosticsOutputFile();
-  /// Set the diagnostics output file used for optimization diagnostics.
-  ///
-  /// By default or if invoked with null, diagnostics are not saved in a file
-  /// but only emitted via the diagnostic handler.  Even if an output file is
-  /// set, the handler is invoked for each diagnostic message.
-  void setDiagnosticsOutputFile(std::unique_ptr<yaml::Output> F);
+  /// Return the streamer used by the backend to save remark diagnostics. If it
+  /// does not exist, diagnostics are not saved in a file but only emitted via
+  /// the diagnostic handler.
+  RemarkStreamer *getRemarkStreamer();
+  const RemarkStreamer *getRemarkStreamer() const;
+
+  /// Set the diagnostics output used for optimization diagnostics.
+  /// This filename may be embedded in a section for tools to find the
+  /// diagnostics whenever they're needed.
+  ///
+  /// By default, diagnostics are not saved in a file but only emitted via the
+  /// diagnostic handler.  Even if an output file is set, the handler is invoked
+  /// for each diagnostic message.
+  void setRemarkStreamer(StringRef Filename, raw_ostream &OS);
 
   /// Get the prefix that should be printed in front of a diagnostic of
   ///        the given \p Severity
   static const char *getDiagnosticMessagePrefix(DiagnosticSeverity Severity);
 
   /// Report a message to the currently installed diagnostic handler.
   ///
   /// This function returns, in particular in the case of error reporting

llvm/include/llvm/IR/RemarkStreamer.h

+//===- llvm/IR/RemarkStreamer.h - Remark Streamer ---------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file declares the main interface for outputting remarks.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_IR_REMARKSTREAMER_H
+#define LLVM_IR_REMARKSTREAMER_H
+
+#include "llvm/IR/DiagnosticInfo.h"
+#include "llvm/Support/YAMLTraits.h"
+#include "llvm/Support/raw_ostream.h"
+#include <string>
+#include <vector>
+
+namespace llvm {
+/// Streamer for remarks.
+class RemarkStreamer {
+  /// The filename that the remark diagnostics are emitted to.
+  const std::string Filename;
+  /// The open raw_ostream that the remark diagnostics are emitted to.
+  raw_ostream &OS;
+
+  /// The YAML streamer.
+  yaml::Output YAMLOutput;
+
+public:
+  RemarkStreamer(StringRef Filename, raw_ostream& OS);
+  /// Return the filename that the remark diagnostics are emitted to.
+  StringRef getFilename() const { return Filename; }
+  /// Return stream that the remark diagnostics are emitted to.
+  raw_ostream &getStream() { return OS; }
+  /// Emit a diagnostic through the streamer.
+  void emit(const DiagnosticInfoOptimizationBase &Diag);
+};
+} // end namespace llvm
+
+#endif // LLVM_IR_REMARKSTREAMER_H

llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp

 #include "llvm/IR/GlobalObject.h"
 #include "llvm/IR/GlobalValue.h"
 #include "llvm/IR/GlobalVariable.h"
 #include "llvm/IR/Instruction.h"
 #include "llvm/IR/Mangler.h"
 #include "llvm/IR/Metadata.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Operator.h"
+#include "llvm/IR/RemarkStreamer.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/Value.h"
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCCodePadder.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCDirectives.h"
 #include "llvm/MC/MCDwarf.h"
 #include "llvm/MC/MCExpr.h"

llvm/lib/IR/CMakeLists.txt

add_llvm_library(LLVMCore
   ModuleSummaryIndex.cpp
   Operator.cpp
   OptBisect.cpp
   Pass.cpp
   PassInstrumentation.cpp
   PassManager.cpp
   PassRegistry.cpp
   PassTimingInfo.cpp
+  RemarkStreamer.cpp
   SafepointIRVerifier.cpp
   ProfileSummary.cpp
   Statepoint.cpp
   Type.cpp
   TypeFinder.cpp
   Use.cpp
   User.cpp
   Value.cpp

llvm/lib/IR/DiagnosticInfo.cpp

 #include "llvm/IR/DerivedTypes.h"
 #include "llvm/IR/DiagnosticPrinter.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalValue.h"
 #include "llvm/IR/Instruction.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Metadata.h"
 #include "llvm/IR/Module.h"
+#include "llvm/IR/RemarkStreamer.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/Value.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/Path.h"
 #include "llvm/Support/Regex.h"
 #include "llvm/Support/ScopedPrinter.h"

llvm/lib/IR/LLVMContext.cpp

 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringMap.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/Twine.h"
 #include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/DiagnosticPrinter.h"
 #include "llvm/IR/Metadata.h"
 #include "llvm/IR/Module.h"
+#include "llvm/IR/RemarkStreamer.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/raw_ostream.h"
 #include <cassert>
 #include <cstdlib>
 #include <string>
 #include <utility>
 

 
 void LLVMContext::setDiagnosticsHotnessThreshold(uint64_t Threshold) {
   pImpl->DiagnosticsHotnessThreshold = Threshold;
 }
 uint64_t LLVMContext::getDiagnosticsHotnessThreshold() const {
   return pImpl->DiagnosticsHotnessThreshold;
 }
 
-yaml::Output *LLVMContext::getDiagnosticsOutputFile() {
-  return pImpl->DiagnosticsOutputFile.get();
+RemarkStreamer *LLVMContext::getRemarkStreamer() {
+  return pImpl->RemarkStreamer.get();
 }
-
-void LLVMContext::setDiagnosticsOutputFile(std::unique_ptr<yaml::Output> F) {
-  pImpl->DiagnosticsOutputFile = std::move(F);
+const RemarkStreamer *LLVMContext::getRemarkStreamer() const {
+  return const_cast<LLVMContext *>(this)->getRemarkStreamer();
+}
+void LLVMContext::setRemarkStreamer(StringRef Filename, raw_ostream &OS) {
+  pImpl->RemarkStreamer = llvm::make_unique<RemarkStreamer>(Filename, OS);
 }
 
 DiagnosticHandler::DiagnosticHandlerTy
 LLVMContext::getDiagnosticHandlerCallBack() const {
   return pImpl->DiagHandler->DiagHandlerCallback;
 }
 
 void *LLVMContext::getDiagnosticContext() const {
case DS_Remark:
     return "remark";
   case DS_Note:
     return "note";
   }
   llvm_unreachable("Unknown DiagnosticSeverity");
 }
 
 void LLVMContext::diagnose(const DiagnosticInfo &DI) {
-  if (auto *OptDiagBase = dyn_cast<DiagnosticInfoOptimizationBase>(&DI)) {
-    yaml::Output *Out = getDiagnosticsOutputFile();
-    if (Out) {
-      // For remarks the << operator takes a reference to a pointer.
-      auto *P = const_cast<DiagnosticInfoOptimizationBase *>(OptDiagBase);
-      *Out << P;
-    }
-  }
+  if (auto *OptDiagBase = dyn_cast<DiagnosticInfoOptimizationBase>(&DI))
+    if (RemarkStreamer *RS = getRemarkStreamer())
+      RS->emit(*OptDiagBase);
+
   // If there is a report handler, use it.
   if (pImpl->DiagHandler &&
       (!pImpl->RespectDiagnosticFilters || isDiagnosticEnabled(DI)) &&
       pImpl->DiagHandler->handleDiagnostics(DI))
     return;
 
   if (!isDiagnosticEnabled(DI))
     return;

llvm/lib/IR/LLVMContextImpl.h

 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/StringSet.h"
 #include "llvm/BinaryFormat/Dwarf.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DebugInfoMetadata.h"
 #include "llvm/IR/DerivedTypes.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Metadata.h"
+#include "llvm/IR/RemarkStreamer.h"
 #include "llvm/IR/TrackingMDRef.h"
 #include "llvm/Support/Allocator.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/YAMLTraits.h"
 #include <algorithm>
 #include <cassert>
 #include <cstddef>
 #include <cstdint>
public:
 
   LLVMContext::InlineAsmDiagHandlerTy InlineAsmDiagHandler = nullptr;
   void *InlineAsmDiagContext = nullptr;
 
   std::unique_ptr<DiagnosticHandler> DiagHandler;
   bool RespectDiagnosticFilters = false;
   bool DiagnosticsHotnessRequested = false;
   uint64_t DiagnosticsHotnessThreshold = 0;
-  std::unique_ptr<yaml::Output> DiagnosticsOutputFile;
+  std::unique_ptr<RemarkStreamer> RemarkStreamer;
 
   LLVMContext::YieldCallbackTy YieldCallback = nullptr;
   void *YieldOpaqueHandle = nullptr;
 
   using IntMapTy =
       DenseMap<APInt, std::unique_ptr<ConstantInt>, DenseMapAPIntKeyInfo>;
   IntMapTy IntConstants;
 

llvm/lib/IR/RemarkStreamer.cpp

+//===- llvm/IR/RemarkStreamer.cpp - Remark Streamer -*- C++ -------------*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains the implementation of the remark outputting as part of
+// LLVMContext.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/IR/RemarkStreamer.h"
+
+using namespace llvm;
+
+RemarkStreamer::RemarkStreamer(StringRef Filename, raw_ostream &OS)
+    : Filename(Filename), OS(OS),
+      YAMLOutput(OS, reinterpret_cast<void *>(this)) {
+  assert(!Filename.empty() && "This needs to be a real filename.");
+}
+
+void RemarkStreamer::emit(const DiagnosticInfoOptimizationBase &Diag) {
+  DiagnosticInfoOptimizationBase *DiagPtr =
+      const_cast<DiagnosticInfoOptimizationBase *>(&Diag);
+  YAMLOutput << DiagPtr;
+}

llvm/lib/LTO/LTO.cpp

lto::setupOptimizationRemarks(LLVMContext &Context,
   if (Count != -1)
     Filename += ".thin." + llvm::utostr(Count) + ".yaml";
 
   std::error_code EC;
   auto DiagnosticFile =
       llvm::make_unique<ToolOutputFile>(Filename, EC, sys::fs::F_None);
   if (EC)
     return errorCodeToError(EC);
-  Context.setDiagnosticsOutputFile(
-      llvm::make_unique<yaml::Output>(DiagnosticFile->os()));
+  Context.setRemarkStreamer(Filename, DiagnosticFile->os());
   DiagnosticFile->keep();
   return std::move(DiagnosticFile);
 }

llvm/tools/llc/llc.cpp

int main(int argc, char **argv) {
   if (RemarksFilename != "") {
     std::error_code EC;
     YamlFile =
         llvm::make_unique<ToolOutputFile>(RemarksFilename, EC, sys::fs::F_None);
     if (EC) {
       WithColor::error(errs(), argv[0]) << EC.message() << '\n';
       return 1;
     }
-    Context.setDiagnosticsOutputFile(
-        llvm::make_unique<yaml::Output>(YamlFile->os()));
+    Context.setRemarkStreamer(RemarksFilename, YamlFile->os());
   }
 
   if (InputLanguage != "" && InputLanguage != "ir" &&
       InputLanguage != "mir") {
     WithColor::error(errs(), argv[0])
         << "input language must be '', 'IR' or 'MIR'\n";
     return 1;
   }

llvm/tools/opt/opt.cpp

#endif
   if (RemarksFilename != "") {
     std::error_code EC;
     OptRemarkFile =
         llvm::make_unique<ToolOutputFile>(RemarksFilename, EC, sys::fs::F_None);
     if (EC) {
       errs() << EC.message() << '\n';
       return 1;
     }
-    Context.setDiagnosticsOutputFile(
-        llvm::make_unique<yaml::Output>(OptRemarkFile->os()));
+    Context.setRemarkStreamer(RemarksFilename, OptRemarkFile->os());
   }
 
   // Load the input module...
   std::unique_ptr<Module> M =
       parseIRFile(InputFilename, Err, Context, !NoVerify, ClDataLayout);
 
   if (!M) {
     Err.print(argv[0], errs());