[LTO][AIX] Invoking AIX System Assembler in LTO CodeGen

[qiongsiwu1]

Summary

This patch teaches LTOCodeGenerator to call into the AIX system assembler to generate object files. This is in contrast to the approach taken on other platforms, where the LTOCodeGenerate calls the integrated assembler to generate object files. We need to rely on the system assembler because the integrated assembler is incomplete at the moment.

[qiongsiwu1]

Update the diff to include the context.

[w2yehia]

llvm/lib/LTO/LTOCodeGenerator.cpp

119

We should either introduce the savetemps option in a separate patch (my preference), or implement it fully by calling Config.addSaveTemps in the LTOCodeGenerator constructor.
Also, the option spelling should be -save-temps to match the existing option available in other tools.

[MaskRay]

llvm/lib/LTO/LTOCodeGenerator.cpp

120

Drop the default value cl::init(false)

246

The code is AIX specific. The function should be named so instead of pretending it to be generic and used by other targets.

255

The code is AIX specific. The function should be named so instead of pretending it to be generic and used by other targets.

260

Avoid getpid. See createTemporaryFile in Support/FileSystem.h

[qiongsiwu1]

Addressing code review:

1. The lto-save-temp option is removed. We will add that later with a different patch.
2. The newly added AIX specific functions are explicitly named.
3. Minor changes to variable names (runSystemAssembler(SmallString<128> &Name) -> runSystemAssembler(SmallString<128> &AssemblyFile)) and removing unnecessary comments.

Thanks for the feedback @w2yehia and @MaskRay !

[qiongsiwu1]

Minor formatting update.

[qiongsiwu1]

Updating function names so they are more explicit.

[qiongsiwu1]

Gentle ping for review. Thanks!

[w2yehia]

LGTM

llvm/lib/LTO/LTOCodeGenerator.cpp

300

braces can be removed

[MaskRay]

llvm/lib/LTO/LTOCodeGenerator.cpp

120

This option needs an AIX specific name.

251

Remove this function. Inline its use.

[qiongsiwu1]

Addressing code review comments.

[MaskRay]

llvm/lib/LTO/LTOCodeGenerator.cpp

279

Too many blank lines which don't improve readability. The function has some small number of logically inherent units and within a unit a blank line may be unneeded.

[qiongsiwu1]

Addressing code review comments and fixing a test failure on Linux. I will land patch once the pre-commit CI finishes.

Thanks for the review!

llvm/include/llvm/LTO/legacy/LTOCodeGenerator.h

private:
   void applyScopeRestrictions();
   void preserveDiscardableGVs(
       Module &TheModule,
       llvm::function_ref<bool(const GlobalValue &)> mustPreserveGV);
 
   bool determineTarget();
   std::unique_ptr<TargetMachine> createTargetMachine();
 
+  bool useAIXSystemAssembler();
+  bool runAIXSystemAssembler(SmallString<128> &AssemblyFile);
+
   void emitError(const std::string &ErrMsg);
   void emitWarning(const std::string &ErrMsg);
 
   void finishOptimizationRemarks();
 
   LLVMContext &Context;
   std::unique_ptr<Module> MergedModule;
   std::unique_ptr<Linker> TheLinker;

llvm/lib/LTO/LTOCodeGenerator.cpp

 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/SubtargetFeature.h"
 #include "llvm/MC/TargetRegistry.h"
 #include "llvm/Remarks/HotnessThresholdParser.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/Host.h"
 #include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/Process.h"
 #include "llvm/Support/Signals.h"
 #include "llvm/Support/TargetSelect.h"
 #include "llvm/Support/ToolOutputFile.h"
 #include "llvm/Support/YAMLTraits.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Target/TargetOptions.h"
 #include "llvm/Transforms/IPO.h"
 #include "llvm/Transforms/IPO/Internalize.h"
cl::opt<std::string> RemarksFormat(
     "lto-pass-remarks-format",
     cl::desc("The format used for serializing remarks (default: YAML)"),
     cl::value_desc("format"), cl::init("yaml"));
 
 cl::opt<std::string> LTOStatsFile(
     "lto-stats-file",
     cl::desc("Save statistics to the specified file"),
     cl::Hidden);
+
+cl::opt<std::string> AIXSystemAssemblerPath(
+    "lto-aix-system-assembler",
+    cl::desc("Absolute path to the system assembler, picked up on AIX only"),
+    cl::value_desc("path"));
 }
 
 LTOCodeGenerator::LTOCodeGenerator(LLVMContext &Context)
     : Context(Context), MergedModule(new Module("ld-temp.o", Context)),
       TheLinker(new Linker(*MergedModule)) {
   Context.setDiscardValueNames(LTODiscardValueNames);
   Context.enableDebugTypeODRUniquing();
 
if (Out.os().has_error()) {
     Out.os().clear_error();
     return false;
   }
 
   Out.keep();
   return true;
 }
 
+bool LTOCodeGenerator::useAIXSystemAssembler() {
+  const auto &Triple = TargetMach->getTargetTriple();
+  return Triple.isOSAIX();
+}
+
+bool LTOCodeGenerator::runAIXSystemAssembler(SmallString<128> &AssemblyFile) {
+  assert(useAIXSystemAssembler() &&
+         "Runing AIX system assembler when integrated assembler is available!");
+
+  // Set the system assembler path.
+  std::string AssemblerPath(llvm::AIXSystemAssemblerPath.empty()
+                                ? "/usr/bin/as"
+                                : llvm::AIXSystemAssemblerPath.c_str());
+
+  // Prepare inputs for the assember.
+  const auto &Triple = TargetMach->getTargetTriple();
+  const char *Arch = Triple.isArch64Bit() ? "-a64" : "-a32";
+  std::string ObjectFileName(AssemblyFile);
+  ObjectFileName[ObjectFileName.size() - 1] = 'o';
+  SmallVector<StringRef, 8> Args = {
+      "/bin/env",     "LDR_CNTRL=MAXDATA32=0x80000000@${LDR_CNTRL}",
+      AssemblerPath,  Arch,
+      "-many",        "-o",
+      ObjectFileName, AssemblyFile};
+
+  // Invoke the assembler.
+  int RC = sys::ExecuteAndWait(Args[0], Args);
+
+  // Handle errors.
+  if (RC < -1) {
+    emitError("LTO assembler exited abnormally");
+    return false;
+  }
+
+  if (RC < 0) {
+    emitError("Unable to invoke LTO assembler");
+    return false;
+  }
+
+  if (RC > 0) {
+    emitError("LTO assembler invocation returned non-zero");
+    return false;
+  }
+
+  // Cleanup.
+  remove(AssemblyFile.c_str());
+
+  // Fix the output file name.
+  AssemblyFile = ObjectFileName;
+
+  return true;
+}
+
 bool LTOCodeGenerator::compileOptimizedToFile(const char **Name) {
+  if (useAIXSystemAssembler())
+    setFileType(CGFT_AssemblyFile);
+
   // make unique temp output file to put generated code
   SmallString<128> Filename;
 
   auto AddStream = [&](size_t Task) -> std::unique_ptr<CachedFileStream> {
     StringRef Extension(Config.CGFileType == CGFT_AssemblyFile ? "s" : "o");
 
     int FD;
     std::error_code EC =
bool LTOCodeGenerator::compileOptimizedToFile(const char **Name) {
 
   // If statistics were requested, save them to the specified file or
   // print them out after codegen.
   if (StatsFile)
     PrintStatisticsJSON(StatsFile->os());
   else if (AreStatisticsEnabled())
     PrintStatistics();
 
+  if (useAIXSystemAssembler())
+    if (!runAIXSystemAssembler(Filename))
+      return false;
+
   NativeObjectPath = Filename.c_str();
   *Name = NativeObjectPath.c_str();
   return true;
 }
 
 std::unique_ptr<MemoryBuffer>
 LTOCodeGenerator::compileOptimized() {
   const char *name;