[ThinLTO] Add an option to force summary call edges cold for debugging

Summary:
Useful to selectively disable importing into specific modules for
debugging/triaging/workarounds.

Reviewers: eraman

Subscribers: inglorion, llvm-commits

Differential Revision: https://reviews.llvm.org/D45062

llvm-svn: 328909

llvm/include/llvm/IR/ModuleSummaryIndex.h

@@ -404,6 +404,13 @@ class FunctionSummary : public GlobalValueSummary {
   /// <CalleeValueInfo, CalleeInfo> call edge pair.
   using EdgeTy = std::pair<ValueInfo, CalleeInfo>;
 
+  /// Types for -force-summary-edges-cold debugging option.
+  enum ForceSummaryHotnessType : unsigned {
+    FSHT_None,
+    FSHT_AllNonCritical,
+    FSHT_All
+  };
+
   /// An "identifier" for a virtual function. This contains the type identifier
   /// represented as a GUID and the offset from the address point to the virtual
   /// function pointer, where "address point" is as defined in the Itanium ABI:

llvm/lib/Analysis/ModuleSummaryAnalysis.cpp

@@ -49,6 +49,7 @@
 #include "llvm/Object/SymbolicFile.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/Casting.h"
+#include "llvm/Support/CommandLine.h"
 #include <algorithm>
 #include <cassert>
 #include <cstdint>
@@ -58,6 +59,18 @@ using namespace llvm;
 
 #define DEBUG_TYPE "module-summary-analysis"
 
+// Option to force edges cold which will block importing when the
+// -import-cold-multiplier is set to 0. Useful for debugging.
+FunctionSummary::ForceSummaryHotnessType ForceSummaryEdgesCold =
+    FunctionSummary::FSHT_None;
+cl::opt<FunctionSummary::ForceSummaryHotnessType, true> FSEC(
+    "force-summary-edges-cold", cl::Hidden, cl::location(ForceSummaryEdgesCold),
+    cl::desc("Force all edges in the function summary to cold"),
+    cl::values(clEnumValN(FunctionSummary::FSHT_None, "none", "None."),
+               clEnumValN(FunctionSummary::FSHT_AllNonCritical,
+                          "all-non-critical", "All non-critical edges."),
+               clEnumValN(FunctionSummary::FSHT_All, "all", "All edges.")));
+
 // Walk through the operands of a given User via worklist iteration and populate
 // the set of GlobalValue references encountered. Invoked either on an
 // Instruction or a GlobalVariable (which walks its initializer).
@@ -268,6 +281,8 @@ computeFunctionSummary(ModuleSummaryIndex &Index, const Module &M,
         auto ScaledCount = PSI->getProfileCount(&I, BFI);
         auto Hotness = ScaledCount ? getHotness(ScaledCount.getValue(), PSI)
                                    : CalleeInfo::HotnessType::Unknown;
+        if (ForceSummaryEdgesCold != FunctionSummary::FSHT_None)
+          Hotness = CalleeInfo::HotnessType::Cold;
 
         // Use the original CalledValue, in case it was an alias. We want
         // to record the call edge to the alias in that case. Eventually
@@ -318,7 +333,9 @@ computeFunctionSummary(ModuleSummaryIndex &Index, const Module &M,
   // sample PGO, to enable the same inlines as the profiled optimized binary.
   for (auto &I : F.getImportGUIDs())
     CallGraphEdges[Index.getOrInsertValueInfo(I)].updateHotness(
-        CalleeInfo::HotnessType::Critical);
+        ForceSummaryEdgesCold == FunctionSummary::FSHT_All
+            ? CalleeInfo::HotnessType::Cold
+            : CalleeInfo::HotnessType::Critical);
 
   bool NonRenamableLocal = isNonRenamableLocal(F);
   bool NotEligibleForImport =

llvm/lib/Bitcode/Writer/BitcodeWriter.cpp

@@ -89,6 +89,9 @@ static cl::opt<unsigned>
 cl::opt<bool> WriteRelBFToSummary(
     "write-relbf-to-summary", cl::Hidden, cl::init(false),
     cl::desc("Write relative block frequency to function summary "));
+
+extern FunctionSummary::ForceSummaryHotnessType ForceSummaryEdgesCold;
+
 namespace {
 
 /// These are manifest constants used by the bitcode writer. They do not need to
@@ -3436,7 +3439,8 @@ void ModuleBitcodeWriterBase::writePerModuleFunctionSummaryRecord(
   for (auto &RI : FS->refs())
     NameVals.push_back(VE.getValueID(RI.getValue()));
 
-  bool HasProfileData = F.hasProfileData();
+  bool HasProfileData =
+      F.hasProfileData() || ForceSummaryEdgesCold != FunctionSummary::FSHT_None;
   for (auto &ECI : FS->calls()) {
     NameVals.push_back(getValueId(ECI.first));
     if (HasProfileData)

llvm/test/Transforms/FunctionImport/Inputs/funcimport_forcecold.ll

@@ -0,0 +1,4 @@
+define void @foo() {
+entry:
+  ret void
+}

llvm/test/Transforms/FunctionImport/funcimport_forcecold.ll

@@ -0,0 +1,36 @@
+; Test to ensure that building summary with -force-summary-edges-cold
+; blocks importing as expected.
+
+; "-stats" and "-debug-only" require +Asserts.
+; REQUIRES: asserts
+
+; First do with default options, which should import
+; RUN: opt -module-summary %s -o %t.bc
+; RUN: opt -module-summary %p/Inputs/funcimport_forcecold.ll -o %t2.bc
+; RUN: llvm-lto -thinlto -o %t3 %t.bc %t2.bc
+; RUN: opt -function-import -stats -print-imports -summary-file %t3.thinlto.bc %t.bc -S 2>&1 | FileCheck %s --check-prefix=IMPORT
+
+; Next rebuild caller module summary with non-critical edges forced cold (which
+; should affect all edges in this test as we don't have any sample pgo).
+; Make sure we don't import.
+; RUN: opt -force-summary-edges-cold=all-non-critical -module-summary %s -o %t.bc
+; RUN: llvm-lto -thinlto -o %t3 %t.bc %t2.bc
+; RUN: opt -function-import -stats -print-imports -summary-file %t3.thinlto.bc %t.bc -S 2>&1 | FileCheck %s --check-prefix=NOIMPORT
+
+; Next rebuild caller module summary with all edges forced cold.
+; Make sure we don't import.
+; RUN: opt -force-summary-edges-cold=all -module-summary %s -o %t.bc
+; RUN: llvm-lto -thinlto -o %t3 %t.bc %t2.bc
+; RUN: opt -function-import -stats -print-imports -summary-file %t3.thinlto.bc %t.bc -S 2>&1 | FileCheck %s --check-prefix=NOIMPORT
+
+define i32 @main() {
+entry:
+  call void @foo()
+  ret i32 0
+}
+
+; IMPORT: Import foo
+; NOIMPORT-NOT: Import foo
+; IMPORT: define available_externally void @foo()
+; NOIMPORT: declare void @foo()
+declare void @foo()

llvm/test/Transforms/FunctionImport/funcimport_forcecold_samplepgo.ll

@@ -0,0 +1,37 @@
+; Test to ensure that building summary with -force-summary-edges-cold
+; blocks importing as expected.
+
+; "-stats" and "-debug-only" require +Asserts.
+; REQUIRES: asserts
+
+; First do with default options, which should import
+; RUN: opt -module-summary %s -o %t.bc
+; RUN: opt -module-summary %p/Inputs/funcimport_forcecold.ll -o %t2.bc
+; RUN: llvm-lto -thinlto -o %t3 %t.bc %t2.bc
+; RUN: opt -function-import -stats -print-imports -summary-file %t3.thinlto.bc %t.bc -S 2>&1 | FileCheck %s --check-prefix=IMPORT
+
+; Next rebuild caller module summary with only non-critical edges forced cold,
+; which should still import in this case.
+; RUN: opt -force-summary-edges-cold=all-non-critical -module-summary %s -o %t.bc
+; RUN: llvm-lto -thinlto -o %t3 %t.bc %t2.bc
+; RUN: opt -function-import -stats -print-imports -summary-file %t3.thinlto.bc %t.bc -S 2>&1 | FileCheck %s --check-prefix=IMPORT
+
+; Next rebuild caller module summary with all edges forced cold.
+; Make sure we don't import.
+; RUN: opt -force-summary-edges-cold=all -module-summary %s -o %t.bc
+; RUN: llvm-lto -thinlto -o %t3 %t.bc %t2.bc
+; RUN: opt -function-import -stats -print-imports -summary-file %t3.thinlto.bc %t.bc -S 2>&1 | FileCheck %s --check-prefix=NOIMPORT
+
+define i32 @main() !prof !1 {
+entry:
+  call void @foo()
+  ret i32 0
+}
+
+; IMPORT: Import foo
+; NOIMPORT-NOT: Import foo
+; IMPORT: define available_externally void @foo()
+; NOIMPORT: declare void @foo()
+declare void @foo()
+
+!1 = !{!"function_entry_count", i64 110, i64 6699318081062747564}