[clang] 7a6c894 - [SampleFDO] Add use-sample-profile function attribute.

[Wei Mi via cfe-commits]

Author: Wei Mi
Date: 2020-06-02T17:23:17-07:00
New Revision: 7a6c89427c9babc8e4a69e8a2b61bbf4a4b80c56

URL: https://github.com/llvm/llvm-project/commit/7a6c89427c9babc8e4a69e8a2b61bbf4a4b80c56
DIFF: https://github.com/llvm/llvm-project/commit/7a6c89427c9babc8e4a69e8a2b61bbf4a4b80c56.diff

LOG: [SampleFDO] Add use-sample-profile function attribute.

When sampleFDO is enabled, people may expect they can use
-fno-profile-sample-use to opt-out using sample profile for a certain file.
That could be either for debugging purpose or for performance tuning purpose.
However, when thinlto is enabled, if a function in file A compiled with
-fno-profile-sample-use is imported to another file B compiled with
-fprofile-sample-use, the inlined copy of the function in file B may still
get its profile annotated.

The inconsistency may even introduce profile unused warning because if the
target is not compiled with explicit debug information flag, the function
in file A won't have its debug information enabled (debug information will
be enabled implicitly only when -fprofile-sample-use is used). After it is
imported into file B which is compiled with -fprofile-sample-use, profile
annotation for the outline copy of the function will fail because the
function has no debug information, and that will trigger  profile unused
warning.

We add a new attribute use-sample-profile to control whether a function
will use its sample profile no matter for its outline or inline copies.
That will make the behavior of -fno-profile-sample-use consistent.

Differential Revision: https://reviews.llvm.org/D79959

Added: 
    clang/test/CodeGen/use-sample-profile-attr.c
    llvm/test/Transforms/Inline/inline-incompat-attrs.ll
    llvm/test/Transforms/Inline/partial-inline-incompat-attrs.ll
    llvm/test/Transforms/SampleProfile/Inputs/use-sample-profile-attr.prof
    llvm/test/Transforms/SampleProfile/use-sample-profile-attr.ll

Modified: 
    clang/lib/CodeGen/CodeGenFunction.cpp
    llvm/include/llvm/IR/Attributes.td
    llvm/lib/Transforms/IPO/SampleProfile.cpp
    llvm/test/LTO/Resolution/X86/load-sample-prof-icp.ll
    llvm/test/LTO/Resolution/X86/load-sample-prof-lto.ll
    llvm/test/LTO/Resolution/X86/load-sample-prof.ll
    llvm/test/Transforms/SampleProfile/Inputs/profile-symbol-list.ll
    llvm/test/Transforms/SampleProfile/branch.ll
    llvm/test/Transforms/SampleProfile/calls.ll
    llvm/test/Transforms/SampleProfile/cold-indirect-call.ll
    llvm/test/Transforms/SampleProfile/cov-zero-samples.ll
    llvm/test/Transforms/SampleProfile/coverage-warning.ll
    llvm/test/Transforms/SampleProfile/discriminator.ll
    llvm/test/Transforms/SampleProfile/early-inline.ll
    llvm/test/Transforms/SampleProfile/entry_counts.ll
    llvm/test/Transforms/SampleProfile/entry_counts_cold.ll
    llvm/test/Transforms/SampleProfile/entry_counts_missing_dbginfo.ll
    llvm/test/Transforms/SampleProfile/fnptr.ll
    llvm/test/Transforms/SampleProfile/function_metadata.ll
    llvm/test/Transforms/SampleProfile/gcc-simple.ll
    llvm/test/Transforms/SampleProfile/indirect-call-gcc.ll
    llvm/test/Transforms/SampleProfile/indirect-call.ll
    llvm/test/Transforms/SampleProfile/inline-callee-update.ll
    llvm/test/Transforms/SampleProfile/inline-cold-callsite-samplepgo.ll
    llvm/test/Transforms/SampleProfile/inline-cold.ll
    llvm/test/Transforms/SampleProfile/inline-combine.ll
    llvm/test/Transforms/SampleProfile/inline-coverage.ll
    llvm/test/Transforms/SampleProfile/inline-mergeprof.ll
    llvm/test/Transforms/SampleProfile/inline-stats.ll
    llvm/test/Transforms/SampleProfile/inline-topdown.ll
    llvm/test/Transforms/SampleProfile/inline.ll
    llvm/test/Transforms/SampleProfile/nolocinfo.ll
    llvm/test/Transforms/SampleProfile/offset.ll
    llvm/test/Transforms/SampleProfile/profile-format-compress.ll
    llvm/test/Transforms/SampleProfile/profile-format.ll
    llvm/test/Transforms/SampleProfile/profile-sample-accurate.ll
    llvm/test/Transforms/SampleProfile/propagate.ll
    llvm/test/Transforms/SampleProfile/remap.ll
    llvm/test/Transforms/SampleProfile/remarks.ll
    llvm/test/Transforms/SampleProfile/section-accurate-samplepgo.ll
    llvm/test/Transforms/SampleProfile/syntax.ll
    llvm/test/Transforms/SampleProfile/warm-inline-instance.ll

Removed: 
    


################################################################################
diff  --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index d6622a435b53..0fa795d696e5 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -791,6 +791,9 @@ void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy,
   if (CGM.getCodeGenOpts().ProfileSampleAccurate)
     Fn->addFnAttr("profile-sample-accurate");
 
+  if (!CGM.getCodeGenOpts().SampleProfileFile.empty())
+    Fn->addFnAttr("use-sample-profile");
+
   if (D && D->hasAttr<CFICanonicalJumpTableAttr>())
     Fn->addFnAttr("cfi-canonical-jump-table");
 

diff  --git a/clang/test/CodeGen/use-sample-profile-attr.c b/clang/test/CodeGen/use-sample-profile-attr.c
new file mode 100644
index 000000000000..ba6db07c3226
--- /dev/null
+++ b/clang/test/CodeGen/use-sample-profile-attr.c
@@ -0,0 +1,20 @@
+// Test use-sample-profile attribute is present only when SampleFDO
+// is enabled.
+//
+// RUN: %clang_cc1 -O2 -fno-experimental-new-pass-manager \
+// RUN:     -fprofile-sample-use=%S/Inputs/pgo-sample.prof %s -emit-llvm -o - \
+// RUN:     2>&1 | FileCheck %s
+// RUN: %clang_cc1 -O2 -fexperimental-new-pass-manager \
+// RUN:     -fprofile-sample-use=%S/Inputs/pgo-sample.prof %s -emit-llvm -o - \
+// RUN:     2>&1 | FileCheck %s
+// RUN: %clang_cc1 -O2 -fno-experimental-new-pass-manager %s \
+// RUN:     -emit-llvm -o - 2>&1 | FileCheck %s --check-prefix=NOATTR
+// RUN: %clang_cc1 -O2 -fexperimental-new-pass-manager %s -emit-llvm -o - \
+// RUN:     2>&1 | FileCheck %s --check-prefix=NOATTR
+
+// CHECK: define{{.*}} @func{{.*}} #[[ATTRID:[0-9]+]]
+// CHECK: attributes #[[ATTRID]] = {{.*}} "use-sample-profile"
+// NOATTR: define{{.*}} @func{{.*}} #[[ATTRID:[0-9]+]]
+// NOATTR-NOT: attributes #[[ATTRID]] = {{.*}} "use-sample-profile"
+
+int func(int a) { return a; }

diff  --git a/llvm/include/llvm/IR/Attributes.td b/llvm/include/llvm/IR/Attributes.td
index 209b8cca8e4d..ee48f40ca284 100644
--- a/llvm/include/llvm/IR/Attributes.td
+++ b/llvm/include/llvm/IR/Attributes.td
@@ -237,6 +237,7 @@ def UnsafeFPMath : StrBoolAttr<"unsafe-fp-math">;
 def NoJumpTables : StrBoolAttr<"no-jump-tables">;
 def NoInlineLineTables : StrBoolAttr<"no-inline-line-tables">;
 def ProfileSampleAccurate : StrBoolAttr<"profile-sample-accurate">;
+def UseSampleProfile : StrBoolAttr<"use-sample-profile">;
 
 class CompatRule<string F> {
   // The name of the function called to check the attribute of the caller and
@@ -255,6 +256,7 @@ def : CompatRule<"isEqual<SanitizeHWAddressAttr>">;
 def : CompatRule<"isEqual<SanitizeMemTagAttr>">;
 def : CompatRule<"isEqual<SafeStackAttr>">;
 def : CompatRule<"isEqual<ShadowCallStackAttr>">;
+def : CompatRule<"isEqual<UseSampleProfileAttr>">;
 
 class MergeRule<string F> {
   // The name of the function called to merge the attributes of the caller and

diff  --git a/llvm/lib/Transforms/IPO/SampleProfile.cpp b/llvm/lib/Transforms/IPO/SampleProfile.cpp
index 475f6bc8e9b7..a6449f49aa68 100644
--- a/llvm/lib/Transforms/IPO/SampleProfile.cpp
+++ b/llvm/lib/Transforms/IPO/SampleProfile.cpp
@@ -1042,6 +1042,7 @@ bool SampleProfileLoader::inlineHotFunctions(
           if (R != SymbolMap.end() && R->getValue() &&
               !R->getValue()->isDeclaration() &&
               R->getValue()->getSubprogram() &&
+              R->getValue()->hasFnAttribute("use-sample-profile") &&
               isLegalToPromote(*I, R->getValue(), &Reason)) {
             uint64_t C = FS->getEntrySamples();
             auto &DI =
@@ -1785,7 +1786,7 @@ SampleProfileLoader::buildFunctionOrder(Module &M, CallGraph *CG) {
 
   if (!ProfileTopDownLoad || CG == nullptr) {
     for (Function &F : M)
-      if (!F.isDeclaration())
+      if (!F.isDeclaration() && F.hasFnAttribute("use-sample-profile"))
         FunctionOrderList.push_back(&F);
     return FunctionOrderList;
   }
@@ -1795,7 +1796,7 @@ SampleProfileLoader::buildFunctionOrder(Module &M, CallGraph *CG) {
   while (!CGI.isAtEnd()) {
     for (CallGraphNode *node : *CGI) {
       auto F = node->getFunction();
-      if (F && !F->isDeclaration())
+      if (F && !F->isDeclaration() && F->hasFnAttribute("use-sample-profile"))
         FunctionOrderList.push_back(F);
     }
     ++CGI;

diff  --git a/llvm/test/LTO/Resolution/X86/load-sample-prof-icp.ll b/llvm/test/LTO/Resolution/X86/load-sample-prof-icp.ll
index 2c8549c388ea..8ebb0a134ffd 100644
--- a/llvm/test/LTO/Resolution/X86/load-sample-prof-icp.ll
+++ b/llvm/test/LTO/Resolution/X86/load-sample-prof-icp.ll
@@ -17,7 +17,7 @@ target triple = "x86_64-unknown-linux-gnu"
 ; CHECK-LABEL: @test
 ; Checks that the call instruction is promoted to direct call and has
 ; profile count annotated on the direct call.
-define void @test(void ()*) !dbg !7 {
+define void @test(void ()*) #0 !dbg !7 {
   %2 = alloca void ()*
   store void ()* %0, void ()** %2
   %3 = load void ()*, void ()** %2
@@ -28,6 +28,8 @@ define void @test(void ()*) !dbg !7 {
 
 declare void @bar() local_unnamed_addr
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4, !5}
 !llvm.ident = !{!6}

diff  --git a/llvm/test/LTO/Resolution/X86/load-sample-prof-lto.ll b/llvm/test/LTO/Resolution/X86/load-sample-prof-lto.ll
index 47287d16425e..37c016591cb4 100644
--- a/llvm/test/LTO/Resolution/X86/load-sample-prof-lto.ll
+++ b/llvm/test/LTO/Resolution/X86/load-sample-prof-lto.ll
@@ -20,7 +20,7 @@
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
 
-define void @foo() local_unnamed_addr !dbg !7 {
+define void @foo() #0 !dbg !7 {
 entry:
   tail call void @bar(), !dbg !10
   ret void, !dbg !11
@@ -28,6 +28,8 @@ entry:
 
 declare void @bar() local_unnamed_addr
 
+attributes #0 = {"local_unnamed_addr" "use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4, !5}
 !llvm.ident = !{!6}

diff  --git a/llvm/test/LTO/Resolution/X86/load-sample-prof.ll b/llvm/test/LTO/Resolution/X86/load-sample-prof.ll
index 8d09cc7327fe..1bcee25b07af 100644
--- a/llvm/test/LTO/Resolution/X86/load-sample-prof.ll
+++ b/llvm/test/LTO/Resolution/X86/load-sample-prof.ll
@@ -11,7 +11,7 @@
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
 
-define void @foo() local_unnamed_addr !dbg !7 {
+define void @foo() #0 !dbg !7 {
 entry:
   tail call void @bar(), !dbg !10
   ret void, !dbg !11
@@ -19,6 +19,8 @@ entry:
 
 declare void @bar() local_unnamed_addr
 
+attributes #0 = {"local_unnamed_addr" "use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4, !5}
 !llvm.ident = !{!6}

diff  --git a/llvm/test/Transforms/Inline/inline-incompat-attrs.ll b/llvm/test/Transforms/Inline/inline-incompat-attrs.ll
new file mode 100644
index 000000000000..88cdaaeae1da
--- /dev/null
+++ b/llvm/test/Transforms/Inline/inline-incompat-attrs.ll
@@ -0,0 +1,50 @@
+; RUN: opt < %s -passes=inline -inline-threshold=100 -S | FileCheck %s
+
+;; caller1/caller2/callee1/callee2 test functions with incompatible attributes
+;; won't be inlined into each other.
+
+define i32 @callee1(i32 %x) {
+  %x1 = add i32 %x, 1
+  %x2 = add i32 %x1, 1
+  %x3 = add i32 %x2, 1
+  call void @extern()
+  ret i32 %x3
+}
+
+define i32 @callee2(i32 %x) #0 {
+  %x1 = add i32 %x, 1
+  %x2 = add i32 %x1, 1
+  %x3 = add i32 %x2, 1
+  call void @extern()
+  ret i32 %x3
+}
+
+define i32 @caller1(i32 %y1) {
+;; caller1 doesn't have use-sample-profile attribute but callee2 has,
+;; so callee2 won't be inlined into caller1.
+;; caller1 and callee1 don't have use-sample-profile attribute, so
+;; callee1 can be inlined into caller1.
+; CHECK-LABEL: @caller1(
+; CHECK: call i32 @callee2
+; CHECK-NOT: call i32 @callee1
+  %y2 = call i32 @callee2(i32 %y1)
+  %y3 = call i32 @callee1(i32 %y2)
+  ret i32 %y3
+}
+
+define i32 @caller2(i32 %y1) #0 {
+;; caller2 and callee2 both have use-sample-profile attribute, so
+;; callee2 can be inlined into caller2.
+;; caller2 has use-sample-profile attribute but callee1 doesn't have,
+;; so callee1 won't be inlined into caller2.
+; CHECK-LABEL: @caller2(
+; CHECK-NOT: call i32 @callee2
+; CHECK: call i32 @callee1
+  %y2 = call i32 @callee2(i32 %y1)
+  %y3 = call i32 @callee1(i32 %y2)
+  ret i32 %y3
+}
+
+declare void @extern()
+
+attributes #0 = { "use-sample-profile" }

diff  --git a/llvm/test/Transforms/Inline/partial-inline-incompat-attrs.ll b/llvm/test/Transforms/Inline/partial-inline-incompat-attrs.ll
new file mode 100644
index 000000000000..25e4c588d088
--- /dev/null
+++ b/llvm/test/Transforms/Inline/partial-inline-incompat-attrs.ll
@@ -0,0 +1,49 @@
+; RUN: opt < %s -passes=partial-inliner -S 2>&1| FileCheck %s
+
+define i32 @callee1(i32 %arg) {
+bb:
+  %tmp = icmp slt i32 %arg, 0
+  br i1 %tmp, label %bb1, label %bb2
+
+bb1:
+  br i1 undef, label %bb4, label %bb2
+
+bb2:
+  br i1 undef, label %bb4, label %bb5
+
+bb4:
+  %xx1 = phi i32 [ 1, %bb1 ], [ 9, %bb2 ]
+  %xx2 = phi i32 [ 1, %bb1 ], [ 9, %bb2 ]
+  %xx3 = phi i32 [ 1, %bb1 ], [ 9, %bb2 ]
+  tail call void (...) @extern() #2
+  br label %bb5
+
+bb5:
+  %tmp6 = phi i32 [ 1, %bb2 ], [ 9, %bb4 ]
+  ret i32 %tmp6
+}
+
+declare void @extern(...)
+
+define i32 @caller1(i32 %arg) {
+bb:
+;; partial inliner inlines callee to caller.
+; CHECK-LABEL: @caller1
+; CHECK: br i1
+; CHECK: br i1
+; CHECK-NOT: call i32 @callee1(
+  %tmp = tail call i32 @callee1(i32 %arg)
+  ret i32 %tmp
+}
+
+define i32 @caller2(i32 %arg) #0 {
+bb:
+;; partial inliner won't inline callee to caller because they have
+;; incompatible attributes.
+; CHECK-LABEL: @caller2
+; CHECK: call i32 @callee1(
+  %tmp = tail call i32 @callee1(i32 %arg)
+  ret i32 %tmp
+}
+
+attributes #0 = { "use-sample-profile" }

diff  --git a/llvm/test/Transforms/SampleProfile/Inputs/profile-symbol-list.ll b/llvm/test/Transforms/SampleProfile/Inputs/profile-symbol-list.ll
index c4623fbf4113..f30df4e6ba4f 100644
--- a/llvm/test/Transforms/SampleProfile/Inputs/profile-symbol-list.ll
+++ b/llvm/test/Transforms/SampleProfile/Inputs/profile-symbol-list.ll
@@ -90,10 +90,10 @@ while.end:                                        ; preds = %while.body
 ; Function Attrs: nofree nounwind
 declare dso_local i32 @printf(i8* nocapture readonly, ...) local_unnamed_addr #3
 
-attributes #0 = { noinline norecurse nounwind readnone uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
-attributes #1 = { norecurse nounwind readnone uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
-attributes #2 = { nofree norecurse nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
-attributes #3 = { nofree nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { noinline norecurse nounwind readnone uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
+attributes #1 = { norecurse nounwind readnone uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
+attributes #2 = { nofree norecurse nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
+attributes #3 = { nofree nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4, !5}

diff  --git a/llvm/test/Transforms/SampleProfile/Inputs/use-sample-profile-attr.prof b/llvm/test/Transforms/SampleProfile/Inputs/use-sample-profile-attr.prof
new file mode 100644
index 000000000000..b9fc5d9c1efd
--- /dev/null
+++ b/llvm/test/Transforms/SampleProfile/Inputs/use-sample-profile-attr.prof
@@ -0,0 +1,15 @@
+foo:5860:10
+ 0: 1820
+ 1: 1820
+ 2: 1820
+goo:1820:20
+ 0: 1820
+main:225715:1
+ 2.1: 5553
+ 3: 5391
+ 3.1: foo:5860
+  0: 5279
+  1: 5279
+  2: 5279
+ 4: goo:5860
+  1: 5860

diff  --git a/llvm/test/Transforms/SampleProfile/branch.ll b/llvm/test/Transforms/SampleProfile/branch.ll
index 0976e39fa8a9..6e7a9eb7208c 100644
--- a/llvm/test/Transforms/SampleProfile/branch.ll
+++ b/llvm/test/Transforms/SampleProfile/branch.ll
@@ -150,7 +150,7 @@ declare i32 @atoi(i8*) #2
 
 declare i32 @printf(i8*, ...) #3
 
-attributes #0 = { uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind readonly "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
 attributes #3 = { "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }

diff  --git a/llvm/test/Transforms/SampleProfile/calls.ll b/llvm/test/Transforms/SampleProfile/calls.ll
index 4e4322868aff..f92bf547182a 100644
--- a/llvm/test/Transforms/SampleProfile/calls.ll
+++ b/llvm/test/Transforms/SampleProfile/calls.ll
@@ -20,7 +20,7 @@
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
 ; Function Attrs: nounwind uwtable
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !4 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !4 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -33,7 +33,7 @@ entry:
 }
 
 ; Function Attrs: uwtable
-define i32 @main() !dbg !7 {
+define i32 @main() #0 !dbg !7 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -85,6 +85,8 @@ while.end:                                        ; preds = %while.cond
 
 declare i32 @printf(i8*, ...) #2
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}

diff  --git a/llvm/test/Transforms/SampleProfile/cold-indirect-call.ll b/llvm/test/Transforms/SampleProfile/cold-indirect-call.ll
index b8a61e07ae7d..641300b843e4 100644
--- a/llvm/test/Transforms/SampleProfile/cold-indirect-call.ll
+++ b/llvm/test/Transforms/SampleProfile/cold-indirect-call.ll
@@ -1,21 +1,22 @@
 ; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/cold-indirect-call.prof -S | FileCheck %s
 ; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/cold-indirect-call.prof -S | FileCheck %s
 
-define i32 @foo(i32 ()* %func) !dbg !3 {
+define i32 @foo(i32 ()* %func) #0 !dbg !3 {
 ; CHECK: icmp {{.*}} @bar
 ; CHECK-NOT: icmp {{.*}} @baz
   %call = call i32 %func(), !dbg !4
   ret i32 %call
 }
 
-define i32 @bar() !dbg !5 {
+define i32 @bar() #0 !dbg !5 {
   ret i32 41, !dbg !6
 }
 
-define i32 @baz() !dbg !7 {
+define i32 @baz() #0 !dbg !7 {
   ret i32 42, !dbg !8
 }
 
+attributes #0 = {"use-sample-profile"}
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!2}

diff  --git a/llvm/test/Transforms/SampleProfile/cov-zero-samples.ll b/llvm/test/Transforms/SampleProfile/cov-zero-samples.ll
index 9eb312aa4c0a..30f3b14fc233 100644
--- a/llvm/test/Transforms/SampleProfile/cov-zero-samples.ll
+++ b/llvm/test/Transforms/SampleProfile/cov-zero-samples.ll
@@ -18,7 +18,7 @@ source_filename = "test/Transforms/SampleProfile/cov-zero-samples.ll"
 @N = global i64 8000000000, align 8, !dbg !0
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
-define i32 @_Z12never_calledi(i32 %i) !dbg !11 {
+define i32 @_Z12never_calledi(i32 %i) #1 !dbg !11 {
 entry:
   ret i32 0, !dbg !15
 }
@@ -26,7 +26,7 @@ entry:
 ; Function Attrs: nounwind readnone
 declare void @llvm.dbg.declare(metadata, metadata, metadata) #0
 
-define i32 @main() !dbg !17 {
+define i32 @main() #1 !dbg !17 {
 entry:
   %retval = alloca i32, align 4
   %sum = alloca i32, align 4
@@ -85,6 +85,7 @@ for.end:                                          ; preds = %for.cond
 declare i32 @printf(i8*, ...)
 
 attributes #0 = { nounwind readnone }
+attributes #1 = {"use-sample-profile"}
 
 !llvm.dbg.cu = !{!2}
 !llvm.module.flags = !{!8, !9}

diff  --git a/llvm/test/Transforms/SampleProfile/coverage-warning.ll b/llvm/test/Transforms/SampleProfile/coverage-warning.ll
index 3f683cec5298..7577aff59022 100644
--- a/llvm/test/Transforms/SampleProfile/coverage-warning.ll
+++ b/llvm/test/Transforms/SampleProfile/coverage-warning.ll
@@ -1,6 +1,6 @@
 ; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/coverage-warning.prof -sample-profile-check-record-coverage=90 -sample-profile-check-sample-coverage=100 -o /dev/null 2>&1 | FileCheck %s
 ; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/coverage-warning.prof -sample-profile-check-record-coverage=90 -sample-profile-check-sample-coverage=100 -o /dev/null 2>&1 | FileCheck %s
-define i32 @foo(i32 %i) !dbg !4 {
+define i32 @foo(i32 %i) #0 !dbg !4 {
 ; The profile has samples for line locations that are no longer present.
 ; Coverage does not reach 90%, so we should get this warning:
 ;
@@ -27,6 +27,8 @@ return:                                           ; preds = %if.end, %if.then
   ret i32 %1, !dbg !13
 }
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!6, !7}
 !llvm.ident = !{!8}

diff  --git a/llvm/test/Transforms/SampleProfile/discriminator.ll b/llvm/test/Transforms/SampleProfile/discriminator.ll
index 7b3b270d9f1f..4908c313dc5c 100644
--- a/llvm/test/Transforms/SampleProfile/discriminator.ll
+++ b/llvm/test/Transforms/SampleProfile/discriminator.ll
@@ -62,6 +62,7 @@ while.end:                                        ; preds = %while.cond
   ret i32 %4, !dbg !21
 }
 
+attributes #0 = {"use-sample-profile"}
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!7, !8}

diff  --git a/llvm/test/Transforms/SampleProfile/early-inline.ll b/llvm/test/Transforms/SampleProfile/early-inline.ll
index 8b857a449f07..b14316ceda60 100644
--- a/llvm/test/Transforms/SampleProfile/early-inline.ll
+++ b/llvm/test/Transforms/SampleProfile/early-inline.ll
@@ -54,8 +54,8 @@ define void @no_inline() #1 !dbg !17 {
 
 declare i32 @__gxx_personality_v0(...)
 
-attributes #0 = {"target-features"="+sse4.1"}
-attributes #1 = {"target-features"="+sse4.2"}
+attributes #0 = {"target-features"="+sse4.1" "use-sample-profile"}
+attributes #1 = {"target-features"="+sse4.2" "use-sample-profile"}
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4}

diff  --git a/llvm/test/Transforms/SampleProfile/entry_counts.ll b/llvm/test/Transforms/SampleProfile/entry_counts.ll
index c7fc50313c01..f7e7ecc81e80 100644
--- a/llvm/test/Transforms/SampleProfile/entry_counts.ll
+++ b/llvm/test/Transforms/SampleProfile/entry_counts.ll
@@ -4,18 +4,20 @@
 ; According to the profile, function empty() was called 13,293 times.
 ; CHECK: {{.*}} = !{!"function_entry_count", i64 13294}
 
-define void @empty() !dbg !4 {
+define void @empty() #0 !dbg !4 {
 entry:
   ret void, !dbg !9
 }
 
 ; This function does not have profile, check if function_entry_count is -1
 ; CHECK: {{.*}} = !{!"function_entry_count", i64 -1}
-define void @no_profile() {
+define void @no_profile() #0 {
 entry:
   ret void
 }
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!6, !7}
 !llvm.ident = !{!8}

diff  --git a/llvm/test/Transforms/SampleProfile/entry_counts_cold.ll b/llvm/test/Transforms/SampleProfile/entry_counts_cold.ll
index dbc52b1ef187..95dda320b1fd 100644
--- a/llvm/test/Transforms/SampleProfile/entry_counts_cold.ll
+++ b/llvm/test/Transforms/SampleProfile/entry_counts_cold.ll
@@ -95,7 +95,7 @@ declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture) #2
 
 declare void @baz(...) #3
 
-attributes #0 = { nounwind ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { nounwind ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #1 = { nounwind readnone speculatable }
 attributes #2 = { argmemonly nounwind }
 attributes #3 = { "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }

diff  --git a/llvm/test/Transforms/SampleProfile/entry_counts_missing_dbginfo.ll b/llvm/test/Transforms/SampleProfile/entry_counts_missing_dbginfo.ll
index 45b42e04155d..a6def2f18ab4 100644
--- a/llvm/test/Transforms/SampleProfile/entry_counts_missing_dbginfo.ll
+++ b/llvm/test/Transforms/SampleProfile/entry_counts_missing_dbginfo.ll
@@ -105,7 +105,7 @@ declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture) #2
 
 declare void @baz(...) #3
 
-attributes #0 = { nounwind ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { nounwind ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #1 = { nounwind readnone speculatable }
 attributes #2 = { argmemonly nounwind }
 attributes #3 = { "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }

diff  --git a/llvm/test/Transforms/SampleProfile/fnptr.ll b/llvm/test/Transforms/SampleProfile/fnptr.ll
index 28319d8dd438..c22fc8403d87 100644
--- a/llvm/test/Transforms/SampleProfile/fnptr.ll
+++ b/llvm/test/Transforms/SampleProfile/fnptr.ll
@@ -127,6 +127,9 @@ declare i32 @rand() #1
 ; Function Attrs: nounwind
 declare i32 @printf(i8* nocapture readonly, ...) #1
 
+attributes #0 = {"use-sample-profile"}
+attributes #2 = {"use-sample-profile"}
+
 !llvm.module.flags = !{!0}
 !llvm.ident = !{!1}
 !llvm.dbg.cu = !{!26}

diff  --git a/llvm/test/Transforms/SampleProfile/function_metadata.ll b/llvm/test/Transforms/SampleProfile/function_metadata.ll
index 36d2d080fc0f..89c531b45529 100644
--- a/llvm/test/Transforms/SampleProfile/function_metadata.ll
+++ b/llvm/test/Transforms/SampleProfile/function_metadata.ll
@@ -10,12 +10,12 @@ declare void @bar()
 
 declare !dbg !13 void @bar_dbg()
 
-define void @bar_available() !dbg !14 {
+define void @bar_available() #0 !dbg !14 {
   ret void
 }
 
 ; CHECK: define void @test({{.*}} !prof ![[ENTRY_TEST:[0-9]+]]
-define void @test(void ()*) !dbg !7 {
+define void @test(void ()*) #0 !dbg !7 {
   %2 = alloca void ()*
   store void ()* %0, void ()** %2
   %3 = load void ()*, void ()** %2
@@ -26,7 +26,7 @@ define void @test(void ()*) !dbg !7 {
 }
 
 ; CHECK: define void @test_liveness({{.*}} !prof ![[ENTRY_TEST_LIVENESS:[0-9]+]]
-define void @test_liveness() !dbg !12 {
+define void @test_liveness() #0 !dbg !12 {
   call void @foo(), !dbg !20
   ret void
 }
@@ -42,6 +42,8 @@ define void @test_liveness() !dbg !12 {
 ; to bar. bar_available should not be included as it's within the same module.
 ; CHECK: ![[ENTRY_TEST_LIVENESS]] = !{!"function_entry_count", i64 1, i64 6699318081062747564, i64 -2012135647395072713, i64 -1522495160813492905}
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}

diff  --git a/llvm/test/Transforms/SampleProfile/gcc-simple.ll b/llvm/test/Transforms/SampleProfile/gcc-simple.ll
index 1ade694031cf..f778124c2200 100644
--- a/llvm/test/Transforms/SampleProfile/gcc-simple.ll
+++ b/llvm/test/Transforms/SampleProfile/gcc-simple.ll
@@ -136,7 +136,7 @@ for.end.6:                                        ; preds = %for.cond
 ; CHECK ![[PROF3]] = !{!"branch_weights", i32 1, i32 1}
 ; CHECK ![[PROF4]] = !{!"branch_weights", i32 1, i32 20238}
 
-attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
 attributes #3 = { nounwind }

diff  --git a/llvm/test/Transforms/SampleProfile/indirect-call-gcc.ll b/llvm/test/Transforms/SampleProfile/indirect-call-gcc.ll
index b1edb6bfe361..5298fab44ab1 100644
--- a/llvm/test/Transforms/SampleProfile/indirect-call-gcc.ll
+++ b/llvm/test/Transforms/SampleProfile/indirect-call-gcc.ll
@@ -6,7 +6,7 @@
 ; not work.
 ; XFAIL: host-byteorder-big-endian
 
-define void @test(void ()*) !dbg !3 {
+define void @test(void ()*) #0 !dbg !3 {
   %2 = alloca void ()*
   store void ()* %0, void ()** %2
   %3 = load void ()*, void ()** %2
@@ -15,6 +15,8 @@ define void @test(void ()*) !dbg !3 {
   ret void
 }
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!2}
 

diff  --git a/llvm/test/Transforms/SampleProfile/indirect-call.ll b/llvm/test/Transforms/SampleProfile/indirect-call.ll
index 3249cf5f185b..cc7fc7a75fbe 100644
--- a/llvm/test/Transforms/SampleProfile/indirect-call.ll
+++ b/llvm/test/Transforms/SampleProfile/indirect-call.ll
@@ -2,7 +2,7 @@
 ; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/indirect-call.compact.afdo -S | FileCheck %s
 
 ; CHECK-LABEL: @test
-define void @test(void ()*) !dbg !3 {
+define void @test(void ()*) #0 !dbg !3 {
   %2 = alloca void ()*
   store void ()* %0, void ()** %2
   %3 = load void ()*, void ()** %2
@@ -13,7 +13,7 @@ define void @test(void ()*) !dbg !3 {
 
 ; CHECK-LABEL: @test_inline
 ; If the indirect call is promoted and inlined in profile, we should promote and inline it.
-define void @test_inline(i64* (i32*)*, i32* %x) !dbg !6 {
+define void @test_inline(i64* (i32*)*, i32* %x) #0 !dbg !6 {
   %2 = alloca i64* (i32*)*
   store i64* (i32*)* %0, i64* (i32*)** %2
   %3 = load i64* (i32*)*, i64* (i32*)** %2
@@ -35,7 +35,7 @@ define void @test_inline(i64* (i32*)*, i32* %x) !dbg !6 {
 ; CHECK-LABEL: @test_inline_strip
 ; If the indirect call is promoted and inlined in profile, and the callee name
 ; is stripped we should promote and inline it.
-define void @test_inline_strip(i64* (i32*)*, i32* %x) !dbg !8 {
+define void @test_inline_strip(i64* (i32*)*, i32* %x) #0 !dbg !8 {
   %2 = alloca i64* (i32*)*
   store i64* (i32*)* %0, i64* (i32*)** %2
   %3 = load i64* (i32*)*, i64* (i32*)** %2
@@ -51,7 +51,7 @@ define void @test_inline_strip(i64* (i32*)*, i32* %x) !dbg !8 {
 ; CHECK-LABEL: @test_inline_strip_conflict
 ; If the indirect call is promoted and inlined in profile, and the callee name
 ; is stripped, but have more than 1 potential match, we should not promote.
-define void @test_inline_strip_conflict(i64* (i32*)*, i32* %x) !dbg !10 {
+define void @test_inline_strip_conflict(i64* (i32*)*, i32* %x) #0 !dbg !10 {
   %2 = alloca i64* (i32*)*
   store i64* (i32*)* %0, i64* (i32*)** %2
   %3 = load i64* (i32*)*, i64* (i32*)** %2
@@ -62,7 +62,7 @@ define void @test_inline_strip_conflict(i64* (i32*)*, i32* %x) !dbg !10 {
 
 ; CHECK-LABEL: @test_noinline
 ; If the indirect call target is not available, we should not promote it.
-define void @test_noinline(void ()*) !dbg !12 {
+define void @test_noinline(void ()*) #0 !dbg !12 {
   %2 = alloca void ()*
   store void ()* %0, void ()** %2
   %3 = load void ()*, void ()** %2
@@ -75,7 +75,7 @@ define void @test_noinline(void ()*) !dbg !12 {
 ; CHECK-LABEL: @test_noinline_bitcast
 ; If the indirect call has been promoted to a direct call with bitcast,
 ; do not inline it.
-define float @test_noinline_bitcast(float ()*) !dbg !26 {
+define float @test_noinline_bitcast(float ()*) #0 !dbg !26 {
   %2 = alloca float ()*
   store float ()* %0, float ()** %2
 ; CHECK: icmp
@@ -87,7 +87,7 @@ define float @test_noinline_bitcast(float ()*) !dbg !26 {
 
 ; CHECK-LABEL: @test_norecursive_inline
 ; If the indirect call target is the caller, we should not promote it.
-define void @test_norecursive_inline() !dbg !24 {
+define void @test_norecursive_inline() #0 !dbg !24 {
 ; CHECK-NOT: icmp
 ; CHECK: call
   %1 = load void ()*, void ()** @y, align 8
@@ -95,7 +95,7 @@ define void @test_norecursive_inline() !dbg !24 {
   ret void
 }
 
-define i32* @return_arg(i32* readnone returned) !dbg !29{
+define i32* @return_arg(i32* readnone returned) #0 !dbg !29{
   ret i32* %0
 }
 
@@ -103,7 +103,7 @@ define i32* @return_arg(i32* readnone returned) !dbg !29{
 ; When the promoted indirect call returns a parameter that was defined by the
 ; return value of a previous direct call. Checks both direct call and promoted
 ; indirect call are inlined.
-define i32* @return_arg_caller(i32* (i32*)* nocapture) !dbg !30{
+define i32* @return_arg_caller(i32* (i32*)* nocapture) #0 !dbg !30{
 ; CHECK-NOT: call i32* @foo_inline1
 ; CHECK: if.true.direct_targ:
 ; CHECK-NOT: call
@@ -124,7 +124,7 @@ else:
 ; CHECK-LABEL: @branch_prof_valid
 ; Check the conditional branch generated by indirect call promotion won't
 ; have invalid profile like !{!"branch_weights", i32 0, i32 0}.
-define void @branch_prof_valid(void ()* %t0) !dbg !33 {
+define void @branch_prof_valid(void ()* %t0) #0 !dbg !33 {
   %t1 = alloca void ()*
   store void ()* %t0, void ()** %t1
   %t2 = load void ()*, void ()** %t1
@@ -137,49 +137,49 @@ define void @branch_prof_valid(void ()* %t0) !dbg !33 {
 @x = global i32 0, align 4
 @y = global void ()* null, align 8
 
-define i32* @foo_inline1(i32* %x) !dbg !14 {
+define i32* @foo_inline1(i32* %x) #0 !dbg !14 {
   ret i32* %x
 }
 
-define i32* @foo_inline_strip.suffix(i32* %x) !dbg !15 {
+define i32* @foo_inline_strip.suffix(i32* %x) #0 !dbg !15 {
   ret i32* %x
 }
 
-define i32* @foo_inline_strip_conflict.suffix1(i32* %x) !dbg !16 {
+define i32* @foo_inline_strip_conflict.suffix1(i32* %x) #0 !dbg !16 {
   ret i32* %x
 }
 
-define i32* @foo_inline_strip_conflict.suffix2(i32* %x) !dbg !17 {
+define i32* @foo_inline_strip_conflict.suffix2(i32* %x) #0 !dbg !17 {
   ret i32* %x
 }
 
-define i32* @foo_inline_strip_conflict.suffix3(i32* %x) !dbg !18 {
+define i32* @foo_inline_strip_conflict.suffix3(i32* %x) #0 !dbg !18 {
   ret i32* %x
 }
 
-define i32* @foo_inline2(i32* %x) !dbg !19 {
+define i32* @foo_inline2(i32* %x) #0 !dbg !19 {
   ret i32* %x
 }
 
-define void @foo_inline3() !dbg !35 {
+define void @foo_inline3() #0 !dbg !35 {
   ret void
 }
 
-define i32 @foo_noinline(i32 %x) !dbg !20 {
+define i32 @foo_noinline(i32 %x) #0 !dbg !20 {
   ret i32 %x
 }
 
-define void @foo_direct() !dbg !21 {
+define void @foo_direct() #0 !dbg !21 {
   ret void
 }
 
-define i32 @foo_direct_i32() !dbg !28 {
+define i32 @foo_direct_i32() #0 !dbg !28 {
   ret i32 0;
 }
 
 ; CHECK-LABEL: @test_direct
 ; We should not promote a direct call.
-define void @test_direct() !dbg !22 {
+define void @test_direct() #0 !dbg !22 {
 ; CHECK-NOT: icmp
 ; CHECK: call
   call void @foo_alias(), !dbg !23
@@ -188,6 +188,8 @@ define void @test_direct() !dbg !22 {
 
 @foo_alias = alias void (), void ()* @foo_direct
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!2}
 

diff  --git a/llvm/test/Transforms/SampleProfile/inline-callee-update.ll b/llvm/test/Transforms/SampleProfile/inline-callee-update.ll
index 36f13eae5539..65b8d358cbdb 100644
--- a/llvm/test/Transforms/SampleProfile/inline-callee-update.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-callee-update.ll
@@ -6,7 +6,7 @@
 @z = global i32* ()* null, align 8
 
 ; CHECK: define i32* @sample_loader_inlinee() {{.*}} !prof ![[ENTRY:[0-9]+]]
-define i32* @sample_loader_inlinee() !dbg !3 {
+define i32* @sample_loader_inlinee() #0 !dbg !3 {
 bb:
   %tmp = call i32* @direct_leaf_func(i32* null), !dbg !4
   %cmp = icmp ne i32* %tmp, null
@@ -22,7 +22,7 @@ else:                                             ; preds = %bb
 }
 
 ; CHECK: define i32* @cgscc_inlinee() {{.*}} !prof ![[ENTRY:[0-9]+]]
-define i32* @cgscc_inlinee() !dbg !6 {
+define i32* @cgscc_inlinee() #0 !dbg !6 {
 bb:
   %tmp = call i32* @direct_leaf_func(i32* null), !dbg !7
   %cmp = icmp ne i32* %tmp, null
@@ -37,13 +37,13 @@ else:                                             ; preds = %bb
   ret i32* null
 }
 
-define i32* @test_sample_loader_inline(void ()* %arg) !dbg !9 {
+define i32* @test_sample_loader_inline(void ()* %arg) #0 !dbg !9 {
 bb:
   %tmp = call i32* @sample_loader_inlinee(), !dbg !10
   ret i32* %tmp
 }
 
-define i32* @test_cgscc_inline(void ()* %arg) !dbg !11 {
+define i32* @test_cgscc_inline(void ()* %arg) #0 !dbg !11 {
 bb:
   %tmp = call i32* @cgscc_inlinee(), !dbg !12
   ret i32* %tmp
@@ -51,6 +51,8 @@ bb:
 
 declare i32* @direct_leaf_func(i32*)
 
+attributes #0 = {"use-sample-profile"}
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!2}
 

diff  --git a/llvm/test/Transforms/SampleProfile/inline-cold-callsite-samplepgo.ll b/llvm/test/Transforms/SampleProfile/inline-cold-callsite-samplepgo.ll
index de2fa7cdcbfb..d6a2fee6cf62 100644
--- a/llvm/test/Transforms/SampleProfile/inline-cold-callsite-samplepgo.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-cold-callsite-samplepgo.ll
@@ -5,12 +5,12 @@
 ; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/inline.prof -profile-sample-accurate -inline -S -inline-cold-callsite-threshold=0 | FileCheck %s --check-prefix ACCURATE
 
 declare void @extern()
-define void @callee() {
+define void @callee() #1 {
   call void @extern()
   ret void
 }
 
-define void @caller(i32 %y1) {
+define void @caller(i32 %y1) #1 {
 ; CHECK-LABEL: @caller
 ; CHECK-NOT: call void @callee
 ; ACCURATE-LABEL: @caller
@@ -28,4 +28,5 @@ define void @caller_accurate(i32 %y1) #0 {
   ret void
 }
 
-attributes #0 = { "profile-sample-accurate" }
+attributes #0 = { "profile-sample-accurate" "use-sample-profile" }
+attributes #1 = { "use-sample-profile" }

diff  --git a/llvm/test/Transforms/SampleProfile/inline-cold.ll b/llvm/test/Transforms/SampleProfile/inline-cold.ll
index abec880d22ab..6f21a3b1a756 100644
--- a/llvm/test/Transforms/SampleProfile/inline-cold.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-cold.ll
@@ -15,7 +15,7 @@
 
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !6 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !6 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -27,7 +27,7 @@ entry:
   ret i32 %add, !dbg !8
 }
 
-define i32 @main() !dbg !9 {
+define i32 @main() #0 !dbg !9 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -70,6 +70,8 @@ while.end:                                        ; preds = %while.cond
   ret i32 0, !dbg !23
 }
 
+attributes #0 = { "use-sample-profile" }
+
 declare i32 @printf(i8*, ...)
 
 !llvm.dbg.cu = !{!0}

diff  --git a/llvm/test/Transforms/SampleProfile/inline-combine.ll b/llvm/test/Transforms/SampleProfile/inline-combine.ll
index cdbaf3d646d3..efe491df5724 100644
--- a/llvm/test/Transforms/SampleProfile/inline-combine.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-combine.ll
@@ -16,7 +16,7 @@ $foo = comdat any
 
 $bar = comdat any
 
-define void @foo(%"class.llvm::FoldingSetNodeID"* %this) comdat align 2 !dbg !3 {
+define void @foo(%"class.llvm::FoldingSetNodeID"* %this) #0 align 2 !dbg !3 {
   %1 = alloca %"class.llvm::FoldingSetNodeID"*, align 8
   store %"class.llvm::FoldingSetNodeID"* %this, %"class.llvm::FoldingSetNodeID"** %1, align 8
   %2 = load %"class.llvm::FoldingSetNodeID"*, %"class.llvm::FoldingSetNodeID"** %1, align 8
@@ -27,10 +27,12 @@ define void @foo(%"class.llvm::FoldingSetNodeID"* %this) comdat align 2 !dbg !3
   ret void
 }
 
-define void @bar(%"class.llvm::SmallVectorImpl"* %this) comdat align 2 !dbg !8 {
+define void @bar(%"class.llvm::SmallVectorImpl"* %this) #0 align 2 !dbg !8 {
   ret void
 }
 
+attributes #0 = { "comdat" "use-sample-profile" }
+
 !llvm.module.flags = !{!0, !1}
 !llvm.ident = !{!2}
 !llvm.dbg.cu = !{!9}

diff  --git a/llvm/test/Transforms/SampleProfile/inline-coverage.ll b/llvm/test/Transforms/SampleProfile/inline-coverage.ll
index 134bbe8b80d0..5049ed38e0b4 100644
--- a/llvm/test/Transforms/SampleProfile/inline-coverage.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-coverage.ll
@@ -32,7 +32,7 @@
 ; 110% coverage check.
 ; CHECK: warning: coverage.cc:7: 78834 of 78834 available profile samples (100%) were applied
 
-define i64 @_Z3fool(i64 %i) !dbg !4 {
+define i64 @_Z3fool(i64 %i) #0 !dbg !4 {
 entry:
   %i.addr = alloca i64, align 8
   store i64 %i, i64* %i.addr, align 8
@@ -48,7 +48,7 @@ declare void @llvm.dbg.declare(metadata, metadata, metadata)
 
 declare i32 @rand()
 
-define i32 @main() !dbg !9 {
+define i32 @main() #0 !dbg !9 {
 entry:
   %retval = alloca i32, align 4
   %sum = alloca i64, align 8
@@ -87,6 +87,8 @@ for.end:                                          ; preds = %for.cond
   ret i32 %cond, !dbg !42
 }
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!13, !14}
 !llvm.ident = !{!15}

diff  --git a/llvm/test/Transforms/SampleProfile/inline-mergeprof.ll b/llvm/test/Transforms/SampleProfile/inline-mergeprof.ll
index 8b5989f7a623..08e9fa008fa8 100644
--- a/llvm/test/Transforms/SampleProfile/inline-mergeprof.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-mergeprof.ll
@@ -8,7 +8,7 @@
 
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
-define i32 @main() !dbg !6 {
+define i32 @main() #0 !dbg !6 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -22,7 +22,7 @@ entry:
   ret i32 0, !dbg !11
 }
 
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !12 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !12 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -44,7 +44,7 @@ if.else:                                          ; preds = %entry
   ret i32 %add, !dbg !15
 }
 
-define i32 @_Z3subii(i32 %x, i32 %y) !dbg !16 {
+define i32 @_Z3subii(i32 %x, i32 %y) #0 !dbg !16 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -56,6 +56,8 @@ entry:
   ret i32 %add, !dbg !18
 }
 
+attributes #0 = { "use-sample-profile" }
+
 declare i32 @printf(i8*, ...)
 
 !llvm.dbg.cu = !{!0}
@@ -94,4 +96,4 @@ declare i32 @printf(i8*, ...)
 ; MERGE: !{!"branch_weights", i32 11, i32 23}
 ; MERGE: !{!"branch_weights", i32 10}
 ; MERGE: name: "sub"
-; MERGE-NEXT: {!"function_entry_count", i64 3}
\ No newline at end of file
+; MERGE-NEXT: {!"function_entry_count", i64 3}

diff  --git a/llvm/test/Transforms/SampleProfile/inline-stats.ll b/llvm/test/Transforms/SampleProfile/inline-stats.ll
index f928aed6a97d..243960569dee 100644
--- a/llvm/test/Transforms/SampleProfile/inline-stats.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-stats.ll
@@ -19,7 +19,7 @@
 ; }
 ;
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !6 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !6 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -30,7 +30,7 @@ entry:
   %add = add nsw i32 %tmp, %tmp1, !dbg !8
   ret i32 %add, !dbg !8
 }
-define i32 @main() !dbg !9 {
+define i32 @main() #0 !dbg !9 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -72,6 +72,8 @@ while.end:                                        ; preds = %while.cond
 }
 declare i32 @printf(i8*, ...)
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4}
 !llvm.ident = !{!5}
@@ -101,4 +103,4 @@ declare i32 @printf(i8*, ...)
 !22 = !DILocation(line: 11, scope: !9)
 !23 = !DILocation(line: 12, scope: !9)
 
-; CHECK: 1 sample-profile - Number of functions inlined with context sensitive profile
\ No newline at end of file
+; CHECK: 1 sample-profile - Number of functions inlined with context sensitive profile

diff  --git a/llvm/test/Transforms/SampleProfile/inline-topdown.ll b/llvm/test/Transforms/SampleProfile/inline-topdown.ll
index 711440f0070c..3781f6ad2497 100644
--- a/llvm/test/Transforms/SampleProfile/inline-topdown.ll
+++ b/llvm/test/Transforms/SampleProfile/inline-topdown.ll
@@ -9,7 +9,7 @@
 
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !6 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !6 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -24,7 +24,7 @@ entry:
   ret i32 %add, !dbg !8
 }
 
-define i32 @_Z3subii(i32 %x, i32 %y) !dbg !9 {
+define i32 @_Z3subii(i32 %x, i32 %y) #0 !dbg !9 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -36,7 +36,7 @@ entry:
   ret i32 %add, !dbg !11
 }
 
-define i32 @main() !dbg !12 {
+define i32 @main() #0 !dbg !12 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -79,6 +79,8 @@ while.end:                                        ; preds = %while.cond
 
 declare i32 @printf(i8*, ...)
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4}
 !llvm.ident = !{!5}
@@ -120,4 +122,4 @@ declare i32 @printf(i8*, ...)
 ; TOPDOWN: @_Z3sumii
 ; TOPDOWN-NOT: call i32 @_Z3subii
 ; TOPDOWN: @main()
-; TOPDOWN: call i32 @_Z3subii
\ No newline at end of file
+; TOPDOWN: call i32 @_Z3subii

diff  --git a/llvm/test/Transforms/SampleProfile/inline.ll b/llvm/test/Transforms/SampleProfile/inline.ll
index bd7b024c5537..aae7dab52b7c 100644
--- a/llvm/test/Transforms/SampleProfile/inline.ll
+++ b/llvm/test/Transforms/SampleProfile/inline.ll
@@ -20,7 +20,7 @@
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
 ; Function Attrs: nounwind uwtable
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !4 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !4 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -33,7 +33,7 @@ entry:
 }
 
 ; Function Attrs: uwtable
-define i32 @main() !dbg !7 {
+define i32 @main() #0 !dbg !7 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -78,6 +78,8 @@ while.end:                                        ; preds = %while.cond
 
 declare i32 @printf(i8*, ...) #2
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}

diff  --git a/llvm/test/Transforms/SampleProfile/nolocinfo.ll b/llvm/test/Transforms/SampleProfile/nolocinfo.ll
index ba4a78e3d849..0c201b6eb247 100644
--- a/llvm/test/Transforms/SampleProfile/nolocinfo.ll
+++ b/llvm/test/Transforms/SampleProfile/nolocinfo.ll
@@ -1,7 +1,7 @@
 ; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/nolocinfo.prof -S -pass-remarks=sample-profile 2>&1 | FileCheck %s
 ; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/nolocinfo.prof -S -pass-remarks=sample-profile 2>&1 | FileCheck %s
 
-define i32 @foo(i32 %i)  !dbg !4 {
+define i32 @foo(i32 %i) #0 !dbg !4 {
 entry:
   %i.addr = alloca i32, align 4
   %0 = load i32, i32* %i.addr, align 4
@@ -20,6 +20,8 @@ if.end:
   ret i32 1
 }
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}

diff  --git a/llvm/test/Transforms/SampleProfile/offset.ll b/llvm/test/Transforms/SampleProfile/offset.ll
index ba998b6e7f8a..8451efe62b8e 100644
--- a/llvm/test/Transforms/SampleProfile/offset.ll
+++ b/llvm/test/Transforms/SampleProfile/offset.ll
@@ -48,7 +48,7 @@ return:                                           ; preds = %if.else, %if.then
 ; Function Attrs: nounwind readnone
 declare void @llvm.dbg.declare(metadata, metadata, metadata) #1
 
-attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #1 = { nounwind readnone }
 
 !llvm.dbg.cu = !{!0}

diff  --git a/llvm/test/Transforms/SampleProfile/profile-format-compress.ll b/llvm/test/Transforms/SampleProfile/profile-format-compress.ll
index be4eae3cddcb..8303db8f8888 100644
--- a/llvm/test/Transforms/SampleProfile/profile-format-compress.ll
+++ b/llvm/test/Transforms/SampleProfile/profile-format-compress.ll
@@ -35,7 +35,7 @@
 ; CHECK: ![[IDX3]] = !{!"branch_weights", i32 1}
 
 ; Function Attrs: nounwind uwtable
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !4 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !4 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -48,7 +48,7 @@ entry:
 }
 
 ; Function Attrs: uwtable
-define i32 @main() !dbg !7 {
+define i32 @main() #0 !dbg !7 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -92,6 +92,8 @@ while.end:                                        ; preds = %while.cond
 
 declare i32 @printf(i8*, ...) #2
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}

diff  --git a/llvm/test/Transforms/SampleProfile/profile-format.ll b/llvm/test/Transforms/SampleProfile/profile-format.ll
index c42fb5a5e0e5..f00bc73511d8 100644
--- a/llvm/test/Transforms/SampleProfile/profile-format.ll
+++ b/llvm/test/Transforms/SampleProfile/profile-format.ll
@@ -37,7 +37,7 @@
 ; CHECK: ![[IDX3]] = !{!"branch_weights", i32 1}
 
 ; Function Attrs: nounwind uwtable
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !4 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !4 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -50,7 +50,7 @@ entry:
 }
 
 ; Function Attrs: uwtable
-define i32 @main() !dbg !7 {
+define i32 @main() #0 !dbg !7 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -94,6 +94,8 @@ while.end:                                        ; preds = %while.cond
 
 declare i32 @printf(i8*, ...) #2
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}

diff  --git a/llvm/test/Transforms/SampleProfile/profile-sample-accurate.ll b/llvm/test/Transforms/SampleProfile/profile-sample-accurate.ll
index bfade9d90106..0ec187c3bb86 100644
--- a/llvm/test/Transforms/SampleProfile/profile-sample-accurate.ll
+++ b/llvm/test/Transforms/SampleProfile/profile-sample-accurate.ll
@@ -48,7 +48,7 @@
 ; PROFSYMLIST: define i32 @_Z3sumii{{.*}}!prof ![[UNKNOWN_ID:[0-9]+]]
 ;
 ; Function Attrs: nounwind uwtable
-define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !4 {
+define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !4 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -61,7 +61,7 @@ entry:
 }
 
 ; Function Attrs: uwtable
-define i32 @main() !dbg !7 {
+define i32 @main() #0 !dbg !7 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -123,6 +123,8 @@ while.end:                                        ; preds = %while.cond
 
 declare i32 @printf(i8*, ...) #2
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}

diff  --git a/llvm/test/Transforms/SampleProfile/propagate.ll b/llvm/test/Transforms/SampleProfile/propagate.ll
index bd027937719f..b8db9f68c82a 100644
--- a/llvm/test/Transforms/SampleProfile/propagate.ll
+++ b/llvm/test/Transforms/SampleProfile/propagate.ll
@@ -201,9 +201,9 @@ entry:
 
 declare i32 @printf(i8*, ...) #3
 
-attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #1 = { nounwind readnone }
-attributes #2 = { norecurse uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #2 = { norecurse uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #3 = { "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
 
 !llvm.dbg.cu = !{!0}

diff  --git a/llvm/test/Transforms/SampleProfile/remap.ll b/llvm/test/Transforms/SampleProfile/remap.ll
index b104c07290f0..b7b3fb16bb94 100644
--- a/llvm/test/Transforms/SampleProfile/remap.ll
+++ b/llvm/test/Transforms/SampleProfile/remap.ll
@@ -8,7 +8,7 @@
 
 declare i1 @foo()
 
-define void @_ZN3foo3barERKN1M1XINS_6detail3quxEEE() !dbg !2 {
+define void @_ZN3foo3barERKN1M1XINS_6detail3quxEEE() #0 !dbg !2 {
 ; CHECK: Printing analysis 'Branch Probability Analysis' for function '_ZN3foo3barERKN1M1XINS_6detail3quxEEE':
 
 entry:
@@ -49,6 +49,8 @@ return:
   ret void
 }
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!4, !5}
 

diff  --git a/llvm/test/Transforms/SampleProfile/remarks.ll b/llvm/test/Transforms/SampleProfile/remarks.ll
index a5681fb9d542..16682d47522a 100644
--- a/llvm/test/Transforms/SampleProfile/remarks.ll
+++ b/llvm/test/Transforms/SampleProfile/remarks.ll
@@ -155,7 +155,7 @@ entry:
   ret i32 %conv, !dbg !58
 }
 
-attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
 attributes #1 = { nounwind argmemonly }
 attributes #2 = { nounwind readnone }
 attributes #3 = { nounwind "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }

diff  --git a/llvm/test/Transforms/SampleProfile/section-accurate-samplepgo.ll b/llvm/test/Transforms/SampleProfile/section-accurate-samplepgo.ll
index 6ac39a167689..9fc340237cb2 100644
--- a/llvm/test/Transforms/SampleProfile/section-accurate-samplepgo.ll
+++ b/llvm/test/Transforms/SampleProfile/section-accurate-samplepgo.ll
@@ -17,7 +17,7 @@ declare void @hot_func()
 ; ACCURATE: foo_not_in_profile{{.*}}!prof ![[ZERO_ID:[0-9]+]] !section_prefix ![[COLD_ID:[0-9]+]]
 ; The function not appearing in profile is cold when -profile-sample-accurate
 ; is on.
-define void @foo_not_in_profile() {
+define void @foo_not_in_profile() #1 {
   call void @hot_func()
   ret void
 }
@@ -31,7 +31,8 @@ define void @bar_not_in_profile() #0 {
   ret void
 }
 
-attributes #0 = { "profile-sample-accurate" }
+attributes #0 = { "profile-sample-accurate" "use-sample-profile" }
+attributes #1 = { "use-sample-profile" }
 
 ; CHECK: ![[NOPROFILE_ID]] = !{!"function_entry_count", i64 -1}
 ; CHECK: ![[ZERO_ID]] = !{!"function_entry_count", i64 0}

diff  --git a/llvm/test/Transforms/SampleProfile/syntax.ll b/llvm/test/Transforms/SampleProfile/syntax.ll
index 7114dfa61574..b6e84ec3f43a 100644
--- a/llvm/test/Transforms/SampleProfile/syntax.ll
+++ b/llvm/test/Transforms/SampleProfile/syntax.ll
@@ -16,10 +16,13 @@
 ; RUN: not opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/bad_samples.prof 2>&1 | FileCheck -check-prefix=BAD-SAMPLES %s
 ; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/bad_mangle.prof 2>&1 >/dev/null
 
-define void @empty() {
+define void @empty() #0 {
 entry:
   ret void
 }
+
+attributes #0 = { "use-sample-profile" }
+
 ; NO-DEBUG: warning: No debug information found in function empty: Function profile not used
 ; MISSING-FILE: missing.prof: Could not open profile:
 ; BAD-FN-HEADER: error: {{.*}}bad_fn_header.prof: Could not open profile: Unrecognized sample profile encoding format

diff  --git a/llvm/test/Transforms/SampleProfile/use-sample-profile-attr.ll b/llvm/test/Transforms/SampleProfile/use-sample-profile-attr.ll
new file mode 100644
index 000000000000..78b575874d88
--- /dev/null
+++ b/llvm/test/Transforms/SampleProfile/use-sample-profile-attr.ll
@@ -0,0 +1,117 @@
+; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/use-sample-profile-attr.prof -S | FileCheck %s
+; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/use-sample-profile-attr.prof -S | FileCheck %s
+
+ at .str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
+
+;; @goo doesn't have use-sample-profile attribute so it won't have
+;; profile annotated.
+; CHECK-NOT: @goo{{.*}} !prof
+define void @goo() !dbg !26 {
+  ret void
+}
+
+;; @foo has use-sample-profile attribute so it will have profile annotated.
+; CHECK: @foo{{.*}} !prof ![[HDRCNT1:[0-9]+]]
+define i32 @foo(i32 %x, i32 %y) #0 !dbg !4 {
+entry:
+  %x.addr = alloca i32, align 4
+  %y.addr = alloca i32, align 4
+  store i32 %x, i32* %x.addr, align 4
+  store i32 %y, i32* %y.addr, align 4
+  %0 = load i32, i32* %x.addr, align 4, !dbg !11
+  %1 = load i32, i32* %y.addr, align 4, !dbg !11
+  %add = add nsw i32 %0, %1, !dbg !11
+  ret i32 %add, !dbg !11
+}
+
+;; @main has use-sample-profile attribute so it will have profile annotated.
+; CHECK: @main{{.*}} !prof ![[HDRCNT2:[0-9]+]]
+define i32 @main() #1 !dbg !7 {
+entry:
+  %retval = alloca i32, align 4
+  %s = alloca i32, align 4
+  %i = alloca i32, align 4
+  store i32 0, i32* %retval
+  store i32 0, i32* %i, align 4, !dbg !12
+  br label %while.cond, !dbg !13
+
+while.cond:                                       ; preds = %if.end, %entry
+  %0 = load i32, i32* %i, align 4, !dbg !14
+  %inc = add nsw i32 %0, 1, !dbg !14
+  store i32 %inc, i32* %i, align 4, !dbg !14
+  %cmp = icmp slt i32 %0, 400000000, !dbg !14
+  br i1 %cmp, label %while.body, label %while.end, !dbg !14
+
+while.body:                                       ; preds = %while.cond
+  %1 = load i32, i32* %i, align 4, !dbg !16
+  %cmp1 = icmp ne i32 %1, 100, !dbg !16
+  br i1 %cmp1, label %if.then, label %if.else, !dbg !16
+
+
+if.then:                                          ; preds = %while.body
+  %2 = load i32, i32* %i, align 4, !dbg !18
+  %3 = load i32, i32* %s, align 4, !dbg !18
+;; @foo is inlined because the callsite is hot and @foo has use-sample-profile
+;; attribute.
+; CHECK: if.then:
+; CHECK-NOT: call i32 @foo
+  %call = call i32 @foo(i32 %2, i32 %3), !dbg !18
+  store i32 %call, i32* %s, align 4, !dbg !18
+  br label %if.end, !dbg !18
+
+if.else:                                          ; preds = %while.body
+;; @goo is not inlined because @goo doesn't have use-sample-profile attribute.
+; CHECK: if.else:
+; CHECK: call void @goo
+  call void @goo(), !dbg !27
+  store i32 30, i32* %s, align 4, !dbg !20
+  br label %if.end
+
+if.end:                                           ; preds = %if.else, %if.then
+  br label %while.cond, !dbg !22
+
+while.end:                                        ; preds = %while.cond
+  %4 = load i32, i32* %s, align 4, !dbg !24
+  %call2 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i32 0, i32 0), i32 %4), !dbg !24
+  ret i32 0, !dbg !25
+}
+
+; CHECK: ![[HDRCNT1]] = !{!"function_entry_count", i64 11}
+; CHECK: ![[HDRCNT2]] = !{!"function_entry_count", i64 2}
+
+attributes #0 = {"use-sample-profile"}
+attributes #1 = {"use-sample-profile"}
+
+declare i32 @printf(i8*, ...)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!8, !9}
+!llvm.ident = !{!10}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, producer: "clang version 3.5 ", isOptimized: false, emissionKind: NoDebug, file: !1, enums: !2, retainedTypes: !2, globals: !2, imports: !2)
+!1 = !DIFile(filename: "calls.cc", directory: ".")
+!2 = !{}
+!4 = distinct !DISubprogram(name: "foo", line: 3, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: false, unit: !0, scopeLine: 3, file: !1, scope: !5, type: !6, retainedNodes: !2)
+!5 = !DIFile(filename: "calls.cc", directory: ".")
+!6 = !DISubroutineType(types: !2)
+!7 = distinct !DISubprogram(name: "main", line: 7, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: false, unit: !0, scopeLine: 7, file: !1, scope: !5, type: !6, retainedNodes: !2)
+!8 = !{i32 2, !"Dwarf Version", i32 4}
+!9 = !{i32 1, !"Debug Info Version", i32 3}
+!10 = !{!"clang version 3.5 "}
+!11 = !DILocation(line: 4, scope: !4)
+!12 = !DILocation(line: 8, scope: !7)
+!13 = !DILocation(line: 9, scope: !7)
+!14 = !DILocation(line: 9, scope: !15)
+!15 = !DILexicalBlockFile(discriminator: 2, file: !1, scope: !7)
+!16 = !DILocation(line: 10, scope: !17)
+!17 = distinct !DILexicalBlock(line: 10, column: 0, file: !1, scope: !7)
+!18 = !DILocation(line: 10, scope: !19)
+!19 = !DILexicalBlockFile(discriminator: 2, file: !1, scope: !17)
+!20 = !DILocation(line: 10, scope: !21)
+!21 = !DILexicalBlockFile(discriminator: 4, file: !1, scope: !17)
+!22 = !DILocation(line: 10, scope: !23)
+!23 = !DILexicalBlockFile(discriminator: 6, file: !1, scope: !17)
+!24 = !DILocation(line: 11, scope: !7)
+!25 = !DILocation(line: 12, scope: !7)
+!26 = distinct !DISubprogram(name: "goo", line: 11, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: false, unit: !0, scopeLine: 3, file: !1, scope: !5, type: !6, retainedNodes: !2)
+!27 = !DILocation(line: 11, column: 20, scope: !7)

diff  --git a/llvm/test/Transforms/SampleProfile/warm-inline-instance.ll b/llvm/test/Transforms/SampleProfile/warm-inline-instance.ll
index 622db49b7071..d7f1dcb31938 100644
--- a/llvm/test/Transforms/SampleProfile/warm-inline-instance.ll
+++ b/llvm/test/Transforms/SampleProfile/warm-inline-instance.ll
@@ -4,7 +4,7 @@
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
 ; Function Attrs: nounwind uwtable
-define i32 @foo(i32 %x, i32 %y) !dbg !4 {
+define i32 @foo(i32 %x, i32 %y) #0 !dbg !4 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -16,7 +16,7 @@ entry:
   ret i32 %add, !dbg !11
 }
 
-define i32 @goo(i32 %x, i32 %y) {
+define i32 @goo(i32 %x, i32 %y) #0 {
 entry:
   %x.addr = alloca i32, align 4
   %y.addr = alloca i32, align 4
@@ -29,7 +29,7 @@ entry:
 }
 
 ; Function Attrs: uwtable
-define i32 @main() !dbg !7 {
+define i32 @main() #0 !dbg !7 {
 entry:
   %retval = alloca i32, align 4
   %s = alloca i32, align 4
@@ -83,6 +83,8 @@ while.end:                                        ; preds = %while.cond
 
 declare i32 @printf(i8*, ...) #2
 
+attributes #0 = { "use-sample-profile" }
+
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!8, !9}
 !llvm.ident = !{!10}