[clang] Add clang::nooutline Attribute #163666

lenary · 2025-10-16T00:33:40Z

This change:

Adds a [[clang::nooutline]] function attribute for C and C++. There
is no equivalent GNU syntax for this attribute, so no __attribute__
syntax.
Uses the presence of [[clang::nooutline]] to add the nooutline
attribute to IR function definitions.
Adds test for the above.

The nooutline attribute disables both the Machine Outliner (enabled at
Oz for some targets), and the IR Outliner (disabled by default).

lenary · 2025-10-16T00:33:54Z

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

This stack of pull requests is managed by Graphite. Learn more about stacking.

llvmbot · 2025-10-16T00:34:35Z

@llvm/pr-subscribers-clang

@llvm/pr-subscribers-clang-codegen

Author: Sam Elliott (lenary)

Changes

This change:

Adds a [[clang::nooutline]] function attribute for C and C++. There
is no equivalent GNU syntax for this attribute, so no __attribute__
syntax.
Uses the presence of [[clang::nooutline]] to add the nooutline
attribute to IR function definitions.
Adds test for the above.

The nooutline attribute disables both the Machine Outliner (enabled at
Oz for some targets), and the IR Outliner (disabled by default).

Full diff: https://github.com/llvm/llvm-project/pull/163666.diff

6 Files Affected:

(modified) clang/include/clang/Basic/Attr.td (+7)
(modified) clang/lib/CodeGen/CodeGenModule.cpp (+3)
(added) clang/test/CodeGen/attr-nooutline.c (+16)
(modified) clang/test/Misc/pragma-attribute-supported-attributes-list.test (+1)
(added) clang/test/Sema/attr-nooutline.c (+7)
(added) clang/test/Sema/attr-nooutline.cpp (+7)

diff --git a/clang/include/clang/Basic/Attr.td b/clang/include/clang/Basic/Attr.td
index 22e60aa9fe312..b8a61ba4cbac9 100644
--- a/clang/include/clang/Basic/Attr.td
+++ b/clang/include/clang/Basic/Attr.td
@@ -2355,6 +2355,13 @@ def NoInline : DeclOrStmtAttr {
   let SimpleHandler = 1;
 }
 
+def NoOutline : DeclOrStmtAttr {
+  let Spellings = [CXX11<"clang", "nooutline">, C23<"clang", "nooutline">];
+  let Subjects = SubjectList<[Function], ErrorDiag>;
+  let Documentation = [Undocumented];
+  let SimpleHandler = 1;
+}
+
 def NoMips16 : InheritableAttr, TargetSpecificAttr<TargetMips32> {
   let Spellings = [GCC<"nomips16">];
   let Subjects = SubjectList<[Function], ErrorDiag>;
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 8d019d4b2da25..ab267236ed579 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -2820,6 +2820,9 @@ void CodeGenModule::SetLLVMFunctionAttributesForDefinition(const Decl *D,
       B.addAttribute(llvm::Attribute::MinSize);
   }
 
+  if (D->hasAttr<NoOutlineAttr>())
+    B.addAttribute(llvm::Attribute::NoOutline);
+
   F->addFnAttrs(B);
 
   unsigned alignment = D->getMaxAlignment() / Context.getCharWidth();
diff --git a/clang/test/CodeGen/attr-nooutline.c b/clang/test/CodeGen/attr-nooutline.c
new file mode 100644
index 0000000000000..b9f175da24cb5
--- /dev/null
+++ b/clang/test/CodeGen/attr-nooutline.c
@@ -0,0 +1,16 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-attributes --version 6
+// RUN: %clang_cc1 -emit-llvm %s -triple x86_64-unknown-linux-gnu -disable-O0-optnone -o - | FileCheck %s
+
+
+// CHECK: Function Attrs: noinline nooutline nounwind
+// CHECK-LABEL: define dso_local i32 @t1(
+// CHECK-SAME: i32 noundef [[X:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[X_ADDR:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    store i32 [[X]], ptr [[X_ADDR]], align 4
+// CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[X_ADDR]], align 4
+// CHECK-NEXT:    ret i32 [[TMP0]]
+//
+[[clang::nooutline]] int t1(int x) {
+  return x;
+}
diff --git a/clang/test/Misc/pragma-attribute-supported-attributes-list.test b/clang/test/Misc/pragma-attribute-supported-attributes-list.test
index 73d4cb1769ed5..7ef758dbea9eb 100644
--- a/clang/test/Misc/pragma-attribute-supported-attributes-list.test
+++ b/clang/test/Misc/pragma-attribute-supported-attributes-list.test
@@ -123,6 +123,7 @@
 // CHECK-NEXT: NoMerge (SubjectMatchRule_function, SubjectMatchRule_variable)
 // CHECK-NEXT: NoMicroMips (SubjectMatchRule_function)
 // CHECK-NEXT: NoMips16 (SubjectMatchRule_function)
+// CHECK-NEXT: NoOutline (SubjectMatchRule_function)
 // CHECK-NEXT: NoProfileFunction (SubjectMatchRule_function)
 // CHECK-NEXT: NoRandomizeLayout (SubjectMatchRule_record)
 // CHECK-NEXT: NoSanitize (SubjectMatchRule_function, SubjectMatchRule_objc_method, SubjectMatchRule_variable_is_global)
diff --git a/clang/test/Sema/attr-nooutline.c b/clang/test/Sema/attr-nooutline.c
new file mode 100644
index 0000000000000..83848c22d1003
--- /dev/null
+++ b/clang/test/Sema/attr-nooutline.c
@@ -0,0 +1,7 @@
+// RUN: %clang_cc1 %s -verify -fsyntax-only
+
+[[clang::nooutline]] int a; // expected-error {{'clang::nooutline' attribute only applies to functions}}
+
+[[clang::nooutline]] void t1(void);
+
+[[clang::nooutline(2)]] void t2(void); // expected-error {{'clang::nooutline' attribute takes no arguments}}
diff --git a/clang/test/Sema/attr-nooutline.cpp b/clang/test/Sema/attr-nooutline.cpp
new file mode 100644
index 0000000000000..b6c9b3995081a
--- /dev/null
+++ b/clang/test/Sema/attr-nooutline.cpp
@@ -0,0 +1,7 @@
+// RUN: %clang_cc1 -verify -fsyntax-only %s -Wno-c++17-extensions
+
+[[clang::nooutline]] int a; // expected-error {{'clang::nooutline' attribute only applies to functions}}
+
+[[clang::nooutline]] void t1(void);
+
+[[clang::nooutline(2)]] void t2(void); // expected-error {{'clang::nooutline' attribute takes no arguments}}

philnik777

General comment: I really dislike the spelling. The first time I read this I interpreted it as "noout-line" not "no-outline".

philnik777 · 2025-10-16T07:49:23Z

clang/include/clang/Basic/Attr.td

+def NoOutline : DeclOrStmtAttr {
+  let Spellings = [CXX11<"clang", "nooutline">, C23<"clang", "nooutline">];
+  let Subjects = SubjectList<[Function], ErrorDiag>;
+  let Documentation = [Undocumented];


New attributes shouldn't be undocumented.

I have added documentation.

philnik777 · 2025-10-16T07:49:37Z

clang/include/clang/Basic/Attr.td

 }

+def NoOutline : DeclOrStmtAttr {
+  let Spellings = [CXX11<"clang", "nooutline">, C23<"clang", "nooutline">];


I don't think this is done anywhere else.

NoInline does this too but I presume that’s because it’s old; it should also be updated to use Clang<>

Sorry, yes I was following NoInline. fixed.

Sirraide · 2025-10-16T16:34:59Z

clang/include/clang/Basic/Attr.td

 }

+def NoOutline : DeclOrStmtAttr {
+  let Spellings = [CXX11<"clang", "nooutline">, C23<"clang", "nooutline">];


Suggested change

let Spellings = [CXX11<"clang", "nooutline">, C23<"clang", "nooutline">];

let Spellings = [Clang<"nooutline">];

Done, thanks for the pointer.

Sirraide · 2025-10-16T16:35:45Z

clang/include/clang/Basic/Attr.td

 }

+def NoOutline : DeclOrStmtAttr {
+  let Spellings = [CXX11<"clang", "nooutline">, C23<"clang", "nooutline">];


NoInline does this too but I presume that’s because it’s old; it should also be updated to use Clang<>

Sirraide · 2025-10-16T16:36:57Z

clang/test/CodeGen/attr-nooutline.c

@@ -0,0 +1,16 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-attributes --version 6
+// RUN: %clang_cc1 -emit-llvm %s -triple x86_64-unknown-linux-gnu -disable-O0-optnone -o - | FileCheck %s


We should also have a C++ codegen test, but that can just be an extra RUN line w/ -x c++.

Done.

I don't know if there's something I'm missing by using Function, which I should be testing.

For instance:

C++ Lambdas?

C++ Methods?

Blocks?

Objective-C Methods?

I presume C++ Methods will work with this, the others I'm less sure of.

C++ Lambdas?

C++ Methods?

Blocks?

Objective-C Methods?

It would be a good idea to add tests for all of those yes.

Sirraide · 2025-10-16T16:42:22Z

General comment: I really dislike the spelling. The first time I read this I interpreted it as "noout-line" not "no-outline".

Compounds with the same vowel to the left and right of a word boundary always look a bit weird yeah; that said, I personally don’t think it’s too bad; the naming is just a bit unfortunate.

Sirraide · 2025-10-16T16:45:43Z

Oh yeah, there should also be a release note that mentions that we’re adding a new attribute.

lenary · 2025-10-16T20:23:30Z

Thanks for the reviews. They're very helpful. Working on those changes now.

General comment: I really dislike the spelling. The first time I read this I interpreted it as "noout-line" not "no-outline".

I went through other Clang attributes with similar no * naming, to see what the status quo is.

Clang attributes starting "no" without punctuation to split words:

noinline (Function, Statement)
nomerge (Function, Statement, Var)
noduplicate (Function)
nouwtable (Function, Variable)
noescape (Parameter)
noderef (Type?)

Clang attributes starting "no" with punctuation to split words:

no_stack_protector (Function)
no_specializations (Function, Class, Variable)
no_sanitize_memory (Function)
no_thread_safety_analysis (Function)
no_speculative_load_hardening (Function)
no_builtin (Function)
no_destroy (Variable)

To me this split is pretty even between with an underscore and without, but I would prefer the symmetry with noinline and nomerge, if possible.

This change: - Adds a `[[clang::nooutline]]` function attribute for C and C++. There is no equivalent GNU syntax for this attribute, so no `__attribute__` syntax. - Uses the presence of `[[clang::nooutline]]` to add the `nooutline` attribute to IR function definitions. - Adds test for the above. The `nooutline` attribute disables both the Machine Outliner (enabled at Oz for some targets), and the IR Outliner (disabled by default).

Sirraide · 2025-10-16T20:53:53Z

clang/docs/ReleaseNotes.rst

+- Added a new attribute, ``[[clang::nooutline]]`` to suppress outlining from
+  annotated functions. This uses the LLVM `nooutline` attribute.


There is a dedicated section called ‘Attribute Changes in Clang’ below

Sirraide · 2025-10-16T20:57:17Z

clang/test/Sema/attr-nooutline.cpp

@@ -0,0 +1,7 @@
+// RUN: %clang_cc1 -verify -fsyntax-only %s -Wno-c++17-extensions


Suggested change

// RUN: %clang_cc1 -verify -fsyntax-only %s -Wno-c++17-extensions

// RUN: %clang_cc1 -verify -fsyntax-only %s

I don’t think we need to pass -Wno-c++17-extensions here since we’re not using any from what I can tell (and Clang defaults to C++17 anyway).

erichkeane · 2025-10-16T21:04:34Z

clang/include/clang/Basic/Attr.td

 }

+def NoOutline : DeclOrStmtAttr {
+  let Spellings = [Clang<"nooutline">];


Suggested change

let Spellings = [Clang<"nooutline">];

let Spellings = [Clang<"no_outline">];

perhaps? Makes it... somewhat consistent with the force_inline/etc type attributes, and makes it a better mental 'parse'.

I wrote a longer comment about this, see above.

I see, thanks.

Hmm... my mental parse is still not really picking up the actual meaning of what this is (and as I'm the second one to mention that, perhaps something we should care about?). The rest don't have that parse problem for me. While I appreciate the symmetry, at the same time readability should, IMO trump it.

erichkeane · 2025-10-16T21:05:57Z

clang/include/clang/Basic/AttrDocs.td

+  }
+
+  }];
+}


Can you explain why one would not want this to happen? Perhaps a quick example? And can you show a psuedo-code version of x2/x3 (though I'd probably suggest only a single-one of those examples, they aren't unique enough to have 2) as to what it might look like as an 'outlined' version?

erichkeane · 2025-10-16T21:07:06Z

clang/test/CodeGen/attr-nooutline.c

+// C: Function Attrs: noinline nooutline nounwind optnone
+// C-LABEL: define dso_local i32 @t1(
+// C-SAME: i32 noundef [[X:%.*]]) #[[ATTR0:[0-9]+]] {
+// C-NEXT:  [[ENTRY:.*:]]


The body in these tests are probably not worth it since all we care about are the FunctionAttrs line and the one with the declaration. The rest seems fragile for no value.

I agree with the fragility aspect, is there a better way to automatically test these or should I just write the checks by hand?

These are short/easy enough I'd suggest just writing them by hand (or just deleting everything in the body of the function manually).

erichkeane · 2025-10-16T21:07:38Z

clang/include/clang/Basic/Attr.td


+def NoOutline : DeclOrStmtAttr {
+  let Spellings = [Clang<"nooutline">];
+  let Subjects = SubjectList<[Function], ErrorDiag>;


Your examples all use prototyped functions, is there a reason this cant be used on a no-prototype function? If so, this should reflect that (alternatively, write a codegen test to show its lack of importance).

philnik777 · 2025-10-17T06:48:02Z

Thanks for the reviews. They're very helpful. Working on those changes now.

General comment: I really dislike the spelling. The first time I read this I interpreted it as "noout-line" not "no-outline".

I went through other Clang attributes with similar no * naming, to see what the status quo is.

Clang attributes starting "no" without punctuation to split words:
* `noinline` (Function, Statement)

* `nomerge` (Function, Statement, Var)

* `noduplicate` (Function)

* `nouwtable` (Function, Variable)

* `noescape` (Parameter)

* `noderef` (Type?)
Clang attributes starting "no" with punctuation to split words:
* `no_stack_protector` (Function)

* `no_specializations` (Function, Class, Variable)

* `no_sanitize_memory` (Function)

* `no_thread_safety_analysis` (Function)

* `no_speculative_load_hardening` (Function)

* `no_builtin` (Function)

* `no_destroy` (Variable)
To me this split is pretty even between with an underscore and without, but I would prefer the symmetry with noinline and nomerge, if possible.

I'm really not a fan of that. I don't think noinline or nomerge have the same problem, since they don't have noo as the start, making the parsing significantly easier.

This was referenced Oct 16, 2025

[clang][Driver] Support Outline Flags on RISC-V and X86 #163664

Open

[outliners] Turn nooutline into an Enum Attribute #163665

Open

lenary marked this pull request as ready for review October 16, 2025 00:34

llvmbot added clang Clang issues not falling into any other category clang:frontend Language frontend issues, e.g. anything involving "Sema" clang:codegen IR generation bugs: mangling, exceptions, etc. labels Oct 16, 2025

lenary requested review from efriedma-quic and ornata October 16, 2025 00:43

lenary mentioned this pull request Oct 16, 2025

[clang] Ensure -mno-outline adds attributes #163692

Open

philnik777 reviewed Oct 16, 2025

View reviewed changes

Sirraide reviewed Oct 16, 2025

View reviewed changes

lenary added 3 commits October 16, 2025 13:49

Address reviewer feedback: Tests, Docs, TableGen

6337dd0

Release Notes

a729b26

lenary force-pushed the users/lenary/llvm-nooutline-enum-attr branch from 75103d1 to a84c645 Compare October 16, 2025 20:51

lenary force-pushed the users/lenary/clang-add-nooutline-attr branch from 358b089 to a729b26 Compare October 16, 2025 20:51

Sirraide reviewed Oct 16, 2025

View reviewed changes

Sirraide requested a review from erichkeane October 16, 2025 20:57

erichkeane reviewed Oct 16, 2025

View reviewed changes

zwuis requested a review from cor3ntin October 19, 2025 13:24

	let Spellings = [CXX11<"clang", "nooutline">, C23<"clang", "nooutline">];
	let Spellings = [Clang<"nooutline">];

		@@ -0,0 +1,16 @@
		// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-attributes --version 6
		// RUN: %clang_cc1 -emit-llvm %s -triple x86_64-unknown-linux-gnu -disable-O0-optnone -o - \| FileCheck %s

		- Added a new attribute, ``[[clang::nooutline]]`` to suppress outlining from
		annotated functions. This uses the LLVM `nooutline` attribute.

		@@ -0,0 +1,7 @@
		// RUN: %clang_cc1 -verify -fsyntax-only %s -Wno-c++17-extensions

	let Spellings = [Clang<"nooutline">];
	let Spellings = [Clang<"no_outline">];

+                }
+                }];
+              }

[clang] Add clang::nooutline Attribute #163666

Are you sure you want to change the base?

[clang] Add clang::nooutline Attribute #163666

Uh oh!

Conversation

lenary commented Oct 16, 2025

Uh oh!

lenary commented Oct 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented Oct 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

philnik777 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Sirraide commented Oct 16, 2025

Uh oh!

Sirraide commented Oct 16, 2025

Uh oh!

lenary commented Oct 16, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

philnik777 commented Oct 17, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

lenary commented Oct 16, 2025 •

edited

Loading

llvmbot commented Oct 16, 2025 •

edited

Loading