Skip to content

[CGData][Merger] Avoid merging the attached call target #121030

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Dec 27, 2024
Merged

[CGData][Merger] Avoid merging the attached call target #121030

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
16c2303
[CGData][Merger] Avoid merging the attached call target
kyulee-com Dec 24, 2024
d35dc74
Address comments from nocchijiang
kyulee-com Dec 25, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 9 additions & 3 deletions llvm/lib/CodeGen/GlobalMergeFunctions.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,11 +60,17 @@ static bool canParameterizeCallOperand(const CallBase *CI, unsigned OpIdx) {
if (Name.starts_with("__dtrace"))
return false;
}
if (isCalleeOperand(CI, OpIdx) &&
CI->getOperandBundle(LLVMContext::OB_ptrauth).has_value()) {
if (isCalleeOperand(CI, OpIdx)) {
// The operand is the callee and it has already been signed. Ignore this
// because we cannot add another ptrauth bundle to the call instruction.
return false;
if (CI->getOperandBundle(LLVMContext::OB_ptrauth).has_value())
return false;
} else {
// The target of the arc-attached call must be a constant and cannot be
Copy link
Contributor

@nocchijiang nocchijiang Dec 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there a general method to determine whether a bundle is unsafe to optimize? If so, we can avoid enumerating all of them.

Copy link
Contributor Author

@kyulee-com kyulee-com Dec 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not familiar with the bundle. Conservatively, we could bail out all bundle cases, but for now I bail out the known case only. I think if we run into invalid cases, the compilation will fail/crash so we can follow up.

// parameterized.
if (CI->isOperandBundleOfType(LLVMContext::OB_clang_arc_attachedcall,
OpIdx))
return false;
}
return true;
}
Expand Down
37 changes: 37 additions & 0 deletions llvm/test/CodeGen/AArch64/cgdata-no-merge-attached-call-garget.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,37 @@
; This test verifies that two similar functions, f1 and f2, are not merged
; when their attached call targets differ, since these targets cannot be parameterized.

; RUN: llc -mtriple=arm64-apple-darwin -enable-global-merge-func=true < %s | FileCheck %s

; CHECK-NOT: _f1.Tgm
; CHECK-NOT: _f2.Tgm

target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
target triple = "arm64-apple-darwin"

define i64 @f1(ptr %0) {
%2 = call ptr @g1(ptr %0, i32 0) minsize [ "clang.arc.attachedcall"(ptr @llvm.objc.unsafeClaimAutoreleasedReturnValue) ]
tail call void (...) @llvm.objc.clang.arc.noop.use(ptr %2)
%3 = call i64 @g2(ptr %2)
tail call void @objc_release(ptr %2)
%4 = tail call i64 @g3(i64 %3)
ret i64 %4
}

define i64 @f2(ptr %0) {
%2 = call ptr @g1(ptr %0, i32 0) minsize [ "clang.arc.attachedcall"(ptr @llvm.objc.retainAutoreleasedReturnValue) ]
tail call void (...) @llvm.objc.clang.arc.noop.use(ptr %2)
%3 = call i64 @g2(ptr %2)
tail call void @objc_release(ptr %2)
%4 = tail call i64 @g3(i64 %3)
ret i64 %4
}

declare ptr @g1(ptr, i32)
declare i64 @g2(ptr)
declare i64 @g3(i64)

declare void @llvm.objc.clang.arc.noop.use(...)
declare ptr @llvm.objc.unsafeClaimAutoreleasedReturnValue(ptr)
declare ptr @llvm.objc.retainAutoreleasedReturnValue(ptr)
declare void @objc_release(ptr)
Loading