Keep import function list for inlinee profile update

Summary:
When adjusting function entry counts after inlining, Funciton::setEntryCount is called without providing an import function list. The side effect of that is the previously set import function list will be dropped. The import function list is used by ThinLTO to help import hot cross module callee for LTO inlining, so dropping that during ThinLTO pre-link may adversely affect LTO inlining. The fix is to keep the list while updating entry counts for inlining.

Reviewers: wmi, davidxl, tejohnson

Subscribers: mehdi_amini, hiraditya, dexonsmith, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69736

lib/IR/Function.cpp

@@ -1533,6 +1533,11 @@ void Function::setEntryCount(ProfileCount Count,
   auto PrevCount = getEntryCount();
   assert(!PrevCount.hasValue() || PrevCount.getType() == Count.getType());
 #endif
+
+  auto ImportGUIDs = getImportGUIDs();
+  if (S == nullptr && ImportGUIDs.size())
+    S = &ImportGUIDs;
+
   MDBuilder MDB(getContext());
   setMetadata(
       LLVMContext::MD_prof,

lib/Transforms/Utils/InlineFunction.cpp

@@ -1527,8 +1527,6 @@ void llvm::updateProfileCallee(
   else
     newEntryCount = priorEntryCount + entryDelta;
 
-  Callee->setEntryCount(newEntryCount);
-
   // During inlining ?
   if (VMap) {
     uint64_t cloneEntryCount = priorEntryCount - newEntryCount;
@@ -1537,12 +1535,17 @@ void llvm::updateProfileCallee(
         if (auto *CI = dyn_cast_or_null<CallInst>(Entry.second))
           CI->updateProfWeight(cloneEntryCount, priorEntryCount);
   }
+
+  if (entryDelta) {
+    Callee->setEntryCount(newEntryCount);
+
     for (BasicBlock &BB : *Callee)
       // No need to update the callsite if it is pruned during inlining.
       if (!VMap || VMap->count(&BB))
         for (Instruction &I : BB)
           if (CallInst *CI = dyn_cast<CallInst>(&I))
             CI->updateProfWeight(newEntryCount, priorEntryCount);
+  }
 }
 
 /// This function inlines the called function into the basic block of the

test/Transforms/SampleProfile/Inputs/inline-callee-update.prof

@@ -0,0 +1,10 @@
+test_sample_loader_inline:63067:0
+ 1: sample_loader_inlinee:40000
+sample_loader_inlinee:3000:0
+ 1: direct_leaf_func:35000
+  11: 3000
+test_cgscc_inline:63067:0
+ 1: sample_loader_inlinee:1
+cgscc_inlinee:3000:0
+ 1: direct_leaf_func:35000
+  11: 3000

test/Transforms/SampleProfile/inline-callee-update.ll

@@ -0,0 +1,73 @@
+; Make sure Import GUID list for ThinLTO properly maintained while update function's entry count for inlining
+
+; RUN: opt < %s -passes='thinlto-pre-link<O2>' -pgo-kind=pgo-sample-use-pipeline -sample-profile-file=%S/Inputs/inline-callee-update.prof -S | FileCheck %s
+
+@y = global i32* ()* null, align 8
+@z = global i32* ()* null, align 8
+
+define i32* @sample_loader_inlinee() !dbg !3 {
+bb:
+  %tmp = call i32* @direct_leaf_func(i32* null), !dbg !4
+  %cmp = icmp ne i32* %tmp, null
+  br i1 %cmp, label %then, label %else
+
+then:                                             ; preds = %bb
+  %tmp1 = load i32* ()*, i32* ()** @z, align 8, !dbg !5
+  %tmp2 = call i32* %tmp1(), !dbg !5
+  ret i32* %tmp2
+
+else:                                             ; preds = %bb
+  ret i32* null
+}
+
+define i32* @cgscc_inlinee() !dbg !6 {
+bb:
+  %tmp = call i32* @direct_leaf_func(i32* null), !dbg !7
+  %cmp = icmp ne i32* %tmp, null
+  br i1 %cmp, label %then, label %else
+
+then:                                             ; preds = %bb
+  %tmp1 = load i32* ()*, i32* ()** @y, align 8, !dbg !8
+  %tmp2 = call i32* %tmp1(), !dbg !8
+  ret i32* %tmp2
+
+else:                                             ; preds = %bb
+  ret i32* null
+}
+
+define i32* @test_sample_loader_inline(void ()* %arg) !dbg !9 {
+bb:
+  %tmp = call i32* @sample_loader_inlinee(), !dbg !10
+  ret i32* %tmp
+}
+
+define i32* @test_cgscc_inline(void ()* %arg) !dbg !11 {
+bb:
+  %tmp = call i32* @cgscc_inlinee(), !dbg !12
+  ret i32* %tmp
+}
+
+declare i32* @direct_leaf_func(i32*)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!2}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1, isOptimized: false, runtimeVersion: 0, emissionKind: NoDebug)
+!1 = !DIFile(filename: "test.cc", directory: "/")
+!2 = !{i32 2, !"Debug Info Version", i32 3}
+!3 = distinct !DISubprogram(name: "sample_loader_inlinee", scope: !1, file: !1, line: 11, spFlags: DISPFlagDefinition, unit: !0)
+!4 = !DILocation(line: 12, scope: !3)
+!5 = !DILocation(line: 13, scope: !3)
+!6 = distinct !DISubprogram(name: "cgscc_inlinee", scope: !1, file: !1, line: 31, spFlags: DISPFlagDefinition, unit: !0)
+!7 = !DILocation(line: 32, scope: !6)
+!8 = !DILocation(line: 33, scope: !6)
+!9 = distinct !DISubprogram(name: "test_sample_loader_inline", scope: !1, file: !1, line: 3, spFlags: DISPFlagDefinition, unit: !0)
+!10 = !DILocation(line: 4, scope: !9)
+!11 = distinct !DISubprogram(name: "test_cgscc_inline", scope: !1, file: !1, line: 20, spFlags: DISPFlagDefinition, unit: !0)
+!12 = !DILocation(line: 21, scope: !11)
+
+; Make sure the ImportGUID stays with entry count metadata for ThinLTO-PreLink
+; CHECK: distinct !DISubprogram(name: "sample_loader_inlinee"
+; CHECK-NEXT: {!"function_entry_count", i64 1, i64 -9171813444624716006}
+; CHECK: distinct !DISubprogram(name: "cgscc_inlinee"
+; CHECK-NEXT: !{!"function_entry_count", i64 0, i64 -9171813444624716006}