[LTO] Load sample profile in LTO link step.

[trentxintong]

Summary

Load sample profile in LTO link step.
ThinLTO calls populateModulePassManager to load the profile

[tejohnson]

Looks like the same issue exists in the new pass manager. The sample loader is added via buildModuleSimplificationPipeline, which is not called either directly or indirectly from buildLTODefaultPipeline. Can you add and test the fix there too? Thanks!

[trentxintong]

Add the same thing to the new pass manager as suggested by @tejohnson

[davidxl]

lib/Passes/PassBuilder.cpp

1006 ↗

(On Diff #174221)

I thought sample profile is already loaded twice -- one in FE and one in backend compilation. Is it not the case?

[trentxintong]

lib/Passes/PassBuilder.cpp

1006 ↗

(On Diff #174221)

I see thats the case for ThinLTO. but not for LTO.

[tejohnson]

lib/Passes/PassBuilder.cpp

1008 ↗

(On Diff #174221)

Nit: I think the typical format is to put the parameter name comment before the constant.

test/LTO/Resolution/X86/load-sample-prof-lto.ll

19 ↗

(On Diff #174221)

No need to use different check prefixes for the two pass managers since they are checking the same thing.

[trentxintong]

lib/Passes/PassBuilder.cpp

1008 ↗

(On Diff #174221)

A few lines up, line 976 in this file says different -:).

[tejohnson]

lib/Passes/PassBuilder.cpp

1008 ↗

(On Diff #174221)

Ok nevermind then, better to be consistent within this file.

[trentxintong]

Address @tejohnson comment. Thanks!

[tejohnson]

LGTM

llvm/trunk/lib/Passes/PassBuilder.cpp

 }
 
 ModulePassManager
 PassBuilder::buildLTODefaultPipeline(OptimizationLevel Level, bool DebugLogging,
                                      ModuleSummaryIndex *ExportSummary) {
   assert(Level != O0 && "Must request optimizations for the default pipeline!");
   ModulePassManager MPM(DebugLogging);
 
+  if (PGOOpt && !PGOOpt->SampleProfileFile.empty()) {
+    // Load sample profile before running the LTO optimization pipeline.
+    MPM.addPass(SampleProfileLoaderPass(PGOOpt->SampleProfileFile,
+                                        PGOOpt->ProfileRemappingFile,
+                                        false /* ThinLTOPhase::PreLink */));
+  }
+
   // Remove unused virtual tables to improve the quality of code generated by
   // whole-program devirtualization and bitset lowering.
   MPM.addPass(GlobalDCEPass());
 
   // Force any function attributes we want the rest of the pipeline to observe.
   MPM.addPass(ForceFunctionAttrsPass());
 
   // Do basic inference of function attributes from known properties of system

llvm/trunk/lib/Transforms/IPO/PassManagerBuilder.cpp

void PassManagerBuilder::populateModulePassManager(
   addExtensionsToPM(EP_OptimizerLast, MPM);
 
   // Rename anon globals to be able to handle them in the summary
   if (PrepareForLTO)
     MPM.add(createNameAnonGlobalPass());
 }
 
 void PassManagerBuilder::addLTOOptimizationPasses(legacy::PassManagerBase &PM) {
+  // Load sample profile before running the LTO optimization pipeline.
+  if (!PGOSampleUse.empty()) {
+    PM.add(createPruneEHPass());
+    PM.add(createSampleProfileLoaderPass(PGOSampleUse));
+  }
+
   // Remove unused virtual tables to improve the quality of code generated by
   // whole-program devirtualization and bitset lowering.
   PM.add(createGlobalDCEPass());
 
   // Provide AliasAnalysis services for optimizations.
   addInitialAliasAnalysisPasses(PM);
 
   // Allow forcing function attributes as a debugging and tuning aid.

llvm/trunk/test/LTO/Resolution/X86/load-sample-prof-lto.ll

+; Test that LTO pipeline loads profile.
+;
+; RUN: opt < %s -o %t.bc
+
+; Run the old pm LTO pipeline.
+; RUN: llvm-lto2 run -o %t.out %t.bc -save-temps \
+; RUN:   -r %t.bc,foo,px -r %t.bc,bar,x \
+; RUN:   -lto-sample-profile-file=%S/Inputs/load-sample-prof.prof
+; RUN: llvm-dis %t.out.0.4.opt.bc -o - | FileCheck %s
+
+; Run the new pm LTO pipeline.
+; RUN: llvm-lto2 run -o %t.out %t.bc -save-temps -use-new-pm \
+; RUN:   -r %t.bc,foo,px -r %t.bc,bar,x \
+; RUN:   -lto-sample-profile-file=%S/Inputs/load-sample-prof.prof
+; RUN: llvm-dis %t.out.0.4.opt.bc -o - | FileCheck %s
+
+; Make sure profile information is attached.
+; CHECK: !prof
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @foo() local_unnamed_addr !dbg !7 {
+entry:
+  tail call void @bar(), !dbg !10
+  ret void, !dbg !11
+}
+
+declare void @bar() local_unnamed_addr
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 6.0.0 ", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2)
+!1 = !DIFile(filename: "test.c", directory: "/tmp")
+!2 = !{}
+!3 = !{i32 2, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 6.0.0 "}
+!7 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 3, type: !8, isLocal: false, isDefinition: true, scopeLine: 3, isOptimized: true, unit: !0, retainedNodes: !2)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null}
+!10 = !DILocation(line: 4, column: 5, scope: !7)
+!11 = !DILocation(line: 5, column: 1, scope: !7)