Skip to content
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

Jump to bottom

Work around invalid DWARF bugs for fat LTO #89041

Merged
merged 1 commit into from
Sep 22, 2021
+20 −10
Merged

Work around invalid DWARF bugs for fat LTO #89041

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
18 changes: 16 additions & 2 deletions compiler/rustc_codegen_llvm/src/back/lto.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -325,6 +325,20 @@ fn fat_lto(
drop(linker);
save_temp_bitcode(&cgcx, &module, "lto.input");

// Fat LTO also suffers from the invalid DWARF issue similar to Thin LTO.
// Here we rewrite all `DICompileUnit` pointers if there is only one `DICompileUnit`.
// This only works around the problem when codegen-units = 1.
// Refer to the comments in the `optimize_thin_module` function for more details.
let mut cu1 = ptr::null_mut();
let mut cu2 = ptr::null_mut();
unsafe { llvm::LLVMRustLTOGetDICompileUnit(llmod, &mut cu1, &mut cu2) };
if !cu2.is_null() {
Copy link
Contributor

Choose a reason for hiding this comment

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

The Thin LTO use of this seems to have the opposite check? It only proceeds if cu2 is nul.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Oops... I didn't mean to do the opposite, actually.

However, after the patch, rustc seems to generate valid dwarf for binaries (such as ripgrep) while using full LTO. So I truly misunderstood something.

I am not sure how much missing debug info it will increase. But at least Linux perf does not show unknown userspace addresses in my practice...

Copy link
Contributor

Choose a reason for hiding this comment

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

I tried it out. It seems to be doing what I expect (puts everything in one huge CU), and this check makes sense to me (it's only worth trying to combine CUs if there is more than one CU), so it's probably just the comment that is wrong.

let _timer =
cgcx.prof.generic_activity_with_arg("LLVM_fat_lto_patch_debuginfo", &*module.name);
unsafe { llvm::LLVMRustLTOPatchDICompileUnit(llmod, cu1) };
save_temp_bitcode(cgcx, &module, "fat-lto-after-patch");
}

// Internalize everything below threshold to help strip out more modules and such.
unsafe {
let ptr = symbols_below_threshold.as_ptr();
Expand Down Expand Up @@ -748,7 +762,7 @@ pub unsafe fn optimize_thin_module(
// an error.
let mut cu1 = ptr::null_mut();
let mut cu2 = ptr::null_mut();
llvm::LLVMRustThinLTOGetDICompileUnit(llmod, &mut cu1, &mut cu2);
llvm::LLVMRustLTOGetDICompileUnit(llmod, &mut cu1, &mut cu2);
if !cu2.is_null() {
let msg = "multiple source DICompileUnits found";
return Err(write::llvm_err(&diag_handler, msg));
Expand Down Expand Up @@ -847,7 +861,7 @@ pub unsafe fn optimize_thin_module(
let _timer = cgcx
.prof
.generic_activity_with_arg("LLVM_thin_lto_patch_debuginfo", thin_module.name());
llvm::LLVMRustThinLTOPatchDICompileUnit(llmod, cu1);
llvm::LLVMRustLTOPatchDICompileUnit(llmod, cu1);
save_temp_bitcode(cgcx, &module, "thin-lto-after-patch");
}

Expand Down
8 changes: 2 additions & 6 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 @@ -2377,12 +2377,8 @@ extern "C" {
len: usize,
out_len: &mut usize,
) -> *const u8;
pub fn LLVMRustThinLTOGetDICompileUnit(
M: &Module,
CU1: &mut *mut c_void,
CU2: &mut *mut c_void,
);
pub fn LLVMRustThinLTOPatchDICompileUnit(M: &Module, CU: *mut c_void);
pub fn LLVMRustLTOGetDICompileUnit(M: &Module, CU1: &mut *mut c_void, CU2: &mut *mut c_void);
pub fn LLVMRustLTOPatchDICompileUnit(M: &Module, CU: *mut c_void);

pub fn LLVMRustLinkerNew(M: &'a Module) -> &'a mut Linker<'a>;
pub fn LLVMRustLinkerAdd(
Expand Down
4 changes: 2 additions & 2 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 @@ -1740,7 +1740,7 @@ LLVMRustGetBitcodeSliceFromObjectData(const char *data,
// Rewrite all `DICompileUnit` pointers to the `DICompileUnit` specified. See
// the comment in `back/lto.rs` for why this exists.
extern "C" void
LLVMRustThinLTOGetDICompileUnit(LLVMModuleRef Mod,
LLVMRustLTOGetDICompileUnit(LLVMModuleRef Mod,
DICompileUnit **A,
DICompileUnit **B) {
Module *M = unwrap(Mod);
Expand All @@ -1758,7 +1758,7 @@ LLVMRustThinLTOGetDICompileUnit(LLVMModuleRef Mod,
// Rewrite all `DICompileUnit` pointers to the `DICompileUnit` specified. See
// the comment in `back/lto.rs` for why this exists.
extern "C" void
LLVMRustThinLTOPatchDICompileUnit(LLVMModuleRef Mod, DICompileUnit *Unit) {
LLVMRustLTOPatchDICompileUnit(LLVMModuleRef Mod, DICompileUnit *Unit) {
Module *M = unwrap(Mod);

// If the original source module didn't have a `DICompileUnit` then try to
Expand Down