bjoernager/rust
bjoernager/rust
1
Fork 0
Code Activity

Run LLVM coverage instrumentation passes before optimization passes

Browse source 
This matches the behavior of Clang and allows us to remove several
hacks which were needed to ensure functions weren't optimized away
before reaching the instrumentation pass.
This commit is contained in:
Amanieu d'Antras 2021-03-27 05:04:32 +00:00
parent 3debe9acb8
commit 26d260bfa4
9 changed files with 91 additions and 138 deletions
Download patch file Download diff file Expand all files Collapse all files

8
compiler/rustc_mir/src/monomorphize/partitioning/mod.rs
View file

@ -196,13 +196,7 @@ pub fn partition<'tcx>(
// Next we try to make as many symbols "internal" as possible, so LLVM has
// more freedom to optimize.
if !tcx.sess.link_dead_code() && !tcx.sess.instrument_coverage() {
// Disabled for `-Z instrument-coverage` because some LLVM optimizations can sometimes
// break coverage results. Tests that failed at certain optimization levels are now
// validated at those optimization levels (via `compile-flags` directive); for example:
// * `src/test/run-make-fulldeps/coverage/async.rs` broke with `-C opt-level=1`
// * `src/test/run-make-fulldeps/coverage/closure.rs` broke with `-C opt-level=2`, and
// also required disabling `generate_gcu_internal_copies` in `rustc_middle/mir/mono.rs`
if !tcx.sess.link_dead_code() {
let _prof_timer = tcx.prof.generic_activity("cgu_partitioning_internalize_symbols");
partitioner.internalize_symbols(cx, &mut post_inlining);
}