Clang change: Do not inline hot callsites for samplepgo in thinlto compile phase.

Summary:
Because SamplePGO passes will be invoked twice in ThinLTO build: once at compile phase, the other at backend. We want to make sure the IR at the 2nd phase matches the hot part in pro
file, thus we do not want to inline hot callsites in the first phase.

Reviewers: tejohnson, eraman

Reviewed By: tejohnson

Subscribers: mehdi_amini, cfe-commits, Prazek

Differential Revision: https://reviews.llvm.org/D31202

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@298429 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/BackendUtil.cpp b/lib/CodeGen/BackendUtil.cpp
index 73730e06d144202aa97cfeb4c50117fd881daa63..d1aae08cb57fc39098ee015d4c37f3074490bdd7 100644 (file)
--- a/lib/CodeGen/BackendUtil.cpp
+++ b/lib/CodeGen/BackendUtil.cpp
@@ -318,8 +318,13 @@ void EmitAssemblyHelper::CreatePasses(legacy::PassManager &MPM,
                                      !CodeGenOpts.DisableLifetimeMarkers);
     PMBuilder.Inliner = createAlwaysInlinerLegacyPass(InsertLifetimeIntrinsics);
   } else {
+    // We do not want to inline hot callsites for SamplePGO module-summary build
+    // because profile annotation will happen again in ThinLTO backend, and we
+    // want the IR of the hot path to match the profile.
     PMBuilder.Inliner = createFunctionInliningPass(
-        CodeGenOpts.OptimizationLevel, CodeGenOpts.OptimizeSize);
+        CodeGenOpts.OptimizationLevel, CodeGenOpts.OptimizeSize,
+        (!CodeGenOpts.SampleProfileFile.empty() &&
+         CodeGenOpts.EmitSummaryIndex));
   }
 
   PMBuilder.OptLevel = CodeGenOpts.OptimizationLevel;

diff --git a/test/CodeGen/Inputs/pgo-sample-thinlto-summary.prof b/test/CodeGen/Inputs/pgo-sample-thinlto-summary.prof
new file mode 100644 (file)
index 0000000..b4ef47d
--- /dev/null
+++ b/test/CodeGen/Inputs/pgo-sample-thinlto-summary.prof
@@ -0,0 +1,2 @@
+bar:100:100
+ 2: 2000 foo:2000

diff --git a/test/CodeGen/pgo-sample-thinlto-summary.c b/test/CodeGen/pgo-sample-thinlto-summary.c
new file mode 100644 (file)
index 0000000..cb45ea4
--- /dev/null
+++ b/test/CodeGen/pgo-sample-thinlto-summary.c
@@ -0,0 +1,19 @@
+// RUN: %clang_cc1 -O2 -fprofile-sample-use=%S/Inputs/pgo-sample-thinlto-summary.prof %s -emit-llvm -o - 2>&1 | FileCheck %s -check-prefix=INLINE
+// RUN: %clang_cc1 -O2 -fprofile-sample-use=%S/Inputs/pgo-sample-thinlto-summary.prof %s -emit-llvm -flto=thin -o - 2>&1 | FileCheck %s -check-prefix=NOINLINE
+// Checks if hot call is inlined by normal compile, but not inlined by
+// thinlto compile.
+
+int baz(int);
+int g;
+
+void foo(int n) {
+  for (int i = 0; i < n; i++)
+    g += baz(i);
+}
+
+// INLINE-NOT: call{{.*}}foo
+// NOINLINE: call{{.*}}foo
+void bar(int n) {
+  for (int i = 0; i < n; i++)
+    foo(i);
+}