Skip to content

Add support for storing code model to LLVM module IR #83044

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
Mar 14, 2021
Merged

Add support for storing code model to LLVM module IR #83044

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8357e57
Add support for storing code model to LLVM module IR
kubo39 Mar 11, 2021
7141838
Use only x86_64 flag
kubo39 Mar 12, 2021
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
2 changes: 1 addition & 1 deletion compiler/rustc_codegen_llvm/src/back/write.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,7 +140,7 @@ fn to_llvm_relocation_model(relocation_model: RelocModel) -> llvm::RelocModel {
}
}

fn to_llvm_code_model(code_model: Option<CodeModel>) -> llvm::CodeModel {
pub(crate) fn to_llvm_code_model(code_model: Option<CodeModel>) -> llvm::CodeModel {
match code_model {
Some(CodeModel::Tiny) => llvm::CodeModel::Tiny,
Some(CodeModel::Small) => llvm::CodeModel::Small,
Expand Down
8 changes: 8 additions & 0 deletions compiler/rustc_codegen_llvm/src/context.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
use crate::attributes;
use crate::back::write::to_llvm_code_model;
use crate::callee::get_fn;
use crate::coverageinfo;
use crate::debuginfo;
Expand Down Expand Up @@ -181,6 +182,13 @@ pub unsafe fn create_module(
}
}

// Linking object files with different code models is undefined behavior
// because the compiler would have to generate additional code (to span
// longer jumps) if a larger code model is used with a smaller one.
//
// See https://reviews.llvm.org/D52322 and https://reviews.llvm.org/D52323.
llvm::LLVMRustSetModuleCodeModel(llmod, to_llvm_code_model(sess.code_model()));

// If skipping the PLT is enabled, we need to add some module metadata
// to ensure intrinsic calls don't use it.
if !sess.needs_plt() {
Expand Down
1 change: 1 addition & 0 deletions compiler/rustc_codegen_llvm/src/llvm/ffi.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2326,6 +2326,7 @@ extern "C" {
pub fn LLVMRustUnsetComdat(V: &Value);
pub fn LLVMRustSetModulePICLevel(M: &Module);
pub fn LLVMRustSetModulePIELevel(M: &Module);
pub fn LLVMRustSetModuleCodeModel(M: &Module, Model: CodeModel);
pub fn LLVMRustModuleBufferCreate(M: &Module) -> &'static mut ModuleBuffer;
pub fn LLVMRustModuleBufferPtr(p: &ModuleBuffer) -> *const u8;
pub fn LLVMRustModuleBufferLen(p: &ModuleBuffer) -> usize;
Expand Down
8 changes: 8 additions & 0 deletions compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1260,6 +1260,14 @@ extern "C" void LLVMRustSetModulePIELevel(LLVMModuleRef M) {
unwrap(M)->setPIELevel(PIELevel::Level::Large);
}

extern "C" void LLVMRustSetModuleCodeModel(LLVMModuleRef M,
LLVMRustCodeModel Model) {
auto CM = fromRust(Model);
if (!CM.hasValue())
return;
unwrap(M)->setCodeModel(*CM);
}

// Here you'll find an implementation of ThinLTO as used by the Rust compiler
// right now. This ThinLTO support is only enabled on "recent ish" versions of
// LLVM, and otherwise it's just blanket rejected from other compilers.
Expand Down
18 changes: 18 additions & 0 deletions src/test/codegen/codemodels.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
// revisions: NOMODEL MODEL-SMALL MODEL-KERNEL MODEL-MEDIUM MODEL-LARGE
//[NOMODEL] compile-flags:
//[MODEL-SMALL] compile-flags: -C code-model=small
//[MODEL-KERNEL] compile-flags: --target x86_64-unknown-linux-gnu -C code-model=kernel
Copy link
Contributor

Choose a reason for hiding this comment

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

Pretty sure this is going to fail because not all builder have an x86_64-unknown-linux-gnu target (in particular, those where this is not the host target).

You need to either make this test only-x86_64, or drop the kernel code model part of the test, if that's not portable to other targets.

Copy link
Member

Choose a reason for hiding this comment

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

There's an alternative option of using:

// needs-llvm-components:

together with a #[no_core] example.

//[MODEL-MEDIUM] compile-flags: -C code-model=medium
//[MODEL-LARGE] compile-flags: -C code-model=large

#![crate_type = "lib"]

// MODEL-SMALL: !llvm.module.flags = !{{{.*}}}
// MODEL-SMALL: !{{[0-9]+}} = !{i32 1, !"Code Model", i32 1}
// MODEL-KERNEL: !llvm.module.flags = !{{{.*}}}
// MODEL-KERNEL: !{{[0-9]+}} = !{i32 1, !"Code Model", i32 2}
// MODEL-MEDIUM: !llvm.module.flags = !{{{.*}}}
// MODEL-MEDIUM: !{{[0-9]+}} = !{i32 1, !"Code Model", i32 3}
// MODEL-LARGE: !llvm.module.flags = !{{{.*}}}
// MODEL-LARGE: !{{[0-9]+}} = !{i32 1, !"Code Model", i32 4}
// NOMODEL-NOT: Code Model