[clang] [lld] [llvm] Integrated Distributed ThinLTO (DTLTO): Initial support (PR #126654)

[Nuri Amari via cfe-commits]

================
@@ -2142,3 +2190,327 @@ std::vector<int> lto::generateModulesOrdering(ArrayRef<BitcodeModule *> R) {
   });
   return ModulesOrdering;
 }
+
+namespace {
+// For this out-of-process backend no codegen is done when invoked for each
+// task. Instead we generate the required information (e.g. the summary index
+// shard,import list, etc..) to allow for the codegen to be performed
+// externally . This backend's `wait` function then invokes an external
+// distributor process to do backend compilations.
+class OutOfProcessThinBackend : public CGThinBackend {
+  using SString = SmallString<128>;
+
+  AddBufferFn AddBuffer;
+
+  BumpPtrAllocator Alloc;
+  StringSaver Saver{Alloc};
+
+  SString LinkerOutputFile;
+  StringRef LinkerVersion;
+  SString RemoteOptTool;
+  SString DistributorPath;
+  bool SaveTemps;
+
+  SmallVector<StringRef, 0> CodegenOptions;
+  DenseSet<StringRef> AdditionalInputs;
+
+  // Information specific to individual backend compilation job.
+  struct Job {
+    unsigned Task;
+    StringRef ModuleID;
+    StringRef Triple;
+    StringRef NativeObjectPath;
+    StringRef SummaryIndexPath;
+    ImportsFilesContainer ImportFiles;
+  };
+  // The set of backend compilations jobs.
+  SmallVector<Job> Jobs;
+
+  // A unique string to identify the current link.
+  SmallString<8> UID;
+
+public:
+  OutOfProcessThinBackend(
+      const Config &Conf, ModuleSummaryIndex &CombinedIndex,
+      ThreadPoolStrategy ThinLTOParallelism,
+      const DenseMap<StringRef, GVSummaryMapTy> &ModuleToDefinedGVSummaries,
+      AddStreamFn AddStream, AddBufferFn AddBuffer,
+      lto::IndexWriteCallback OnWrite, bool ShouldEmitIndexFiles,
+      bool ShouldEmitImportsFiles, StringRef LinkerOutputFile,
+      StringRef LinkerVersion, StringRef RemoteOptTool, StringRef Distributor,
+      bool SaveTemps)
+      : CGThinBackend(Conf, CombinedIndex, ModuleToDefinedGVSummaries, OnWrite,
+                      ShouldEmitIndexFiles, ShouldEmitImportsFiles,
+                      ThinLTOParallelism),
+        AddBuffer(std::move(AddBuffer)), LinkerOutputFile(LinkerOutputFile),
+        LinkerVersion(LinkerVersion), RemoteOptTool(RemoteOptTool),
+        DistributorPath(Distributor), SaveTemps(SaveTemps) {}
+
+  virtual void setup(unsigned MaxTasks) override {
+    UID = itostr(sys::Process::getProcessId());
+    Jobs.resize((size_t)MaxTasks);
+  }
+
+  Error start(
+      unsigned Task, BitcodeModule BM,
+      const FunctionImporter::ImportMapTy &ImportList,
+      const FunctionImporter::ExportSetTy &ExportList,
+      const std::map<GlobalValue::GUID, GlobalValue::LinkageTypes> &ResolvedODR,
+      MapVector<StringRef, BitcodeModule> &ModuleMap,
+      DenseMap<StringRef, std::string> &ModuleTriples) override {
+
+    StringRef ModulePath = BM.getModuleIdentifier();
+
+    SString ObjFilePath = sys::path::parent_path(LinkerOutputFile);
+    sys::path::append(ObjFilePath, sys::path::stem(ModulePath) + "." +
+                                       itostr(Task) + "." + UID + ".native.o");
+
+    Job &J = Jobs[Task - 1]; /*Task 0 is reserved*/
+    J = {Task,
+         ModulePath,
+         ModuleTriples[ModulePath],
+         Saver.save(ObjFilePath.str()),
+         Saver.save(ObjFilePath.str() + ".thinlto.bc"),
+         {}};
+
+    assert(ModuleToDefinedGVSummaries.count(ModulePath));
+    BackendThreadPool.async(
+        [=](Job &J, const FunctionImporter::ImportMapTy &ImportList) {
+          if (LLVM_ENABLE_THREADS && Conf.TimeTraceEnabled)
+            timeTraceProfilerInitialize(Conf.TimeTraceGranularity,
+                                        "thin backend");
+          if (auto E = emitFiles(ImportList, J.ModuleID, J.SummaryIndexPath,
+                                 J.ModuleID.str(), J.ImportFiles)) {
+            std::unique_lock<std::mutex> L(ErrMu);
+            if (Err)
+              Err = joinErrors(std::move(*Err), std::move(E));
+            else
+              Err = std::move(E);
+          }
+          if (LLVM_ENABLE_THREADS && Conf.TimeTraceEnabled)
+            timeTraceProfilerFinishThread();
+        },
+        std::ref(J), std::ref(ImportList));
+
+    return Error::success();
+  }
+
+  // Derive a set of Clang options that will be shared/common for all DTLTO
----------------
NuriAmari wrote:

With or without this patch to add integrated dthin-lto, I think something like this (automatic translation of a thin-link invocation into a matching Clang backend invocation) would be useful. Optimization level and flags forwarded through Clang directly to the linker (ex: `-Wl,-mllvm,-my_codegen_option`) are those I've run into most often.

For Buck we either have this [imho unfortunate script wrapping the backend Clang invocations](https://github.com/facebook/buck2/blob/main/prelude/cxx/dist_lto/tools/dist_lto_opt_gnu.py) that translates args from the thin-link invocation to the Clang backend invocation, or we specify both sets of flags manually. @teresajohnson @MaskRay How does Bazel handle this problem?

https://github.com/llvm/llvm-project/pull/126654