bjoernager/rust
bjoernager/rust
1
Fork 0
Code Activity

llvm: migrate to new parameter-bearing uwtable attr

Browse source 
In https://reviews.llvm.org/D114543 the uwtable attribute gained a flag
so that we can ask for sync uwtables instead of async, as the former are
much cheaper. The default is async, so that's what I've done here, but I
left a TODO that we might be able to do better.

While in here I went ahead and dropped support for removing uwtable
attributes in rustc: we never did it, so I didn't write the extra C++
bridge code to make it work. Maybe I should have done the same thing
with the `sync|async` parameter but we'll see.
This commit is contained in:
Augie Fackler 2022-02-14 14:01:19 -05:00
parent 52dd59ed21
commit 0958c8f4ca
5 changed files with 24 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

9
compiler/rustc_codegen_llvm/src/attributes.rs
View file

@ -59,8 +59,11 @@ pub fn sanitize<'ll>(cx: &CodegenCx<'ll, '_>, no_sanitize: SanitizerSet, llfn: &
/// Tell LLVM to emit or not emit the information necessary to unwind the stack for the function.
#[inline]
pub fn emit_uwtable(val: &Value, emit: bool) {
Attribute::UWTable.toggle_llfn(Function, val, emit);
pub fn emit_uwtable(val: &Value) {
// NOTE: We should determine if we even need async unwind tables, as they
// take have more overhead and if we can use sync unwind tables we
// probably should.
llvm::EmitUWTableAttr(val, true);
}
/// Tell LLVM if this function should be 'naked', i.e., skip the epilogue and prologue.
@ -275,7 +278,7 @@ pub fn from_fn_attrs<'ll, 'tcx>(
// You can also find more info on why Windows always requires uwtables here:
// https://bugzilla.mozilla.org/show_bug.cgi?id=1302078
if cx.sess().must_emit_unwind_tables() {
attributes::emit_uwtable(llfn, true);
attributes::emit_uwtable(llfn);
}
if cx.sess().opts.debugging_opts.profile_sample_use.is_some() {