Skip to content

Optimize query_cache_hit to reduce code size of the query hot path. #107529

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
Feb 8, 2023
Merged

Optimize query_cache_hit to reduce code size of the query hot path. #107529

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
e60ccfc
Don't inline query_cache_hit to reduce code size of the query hot path.
Zoxc Jan 31, 2023
9539737
Make an optimal cold path for query_cache_hit
Zoxc Feb 1, 2023
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
1 change: 1 addition & 0 deletions compiler/rustc_data_structures/src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,7 @@
#![feature(associated_type_bounds)]
#![feature(auto_traits)]
#![feature(cell_leak)]
#![feature(core_intrinsics)]
#![feature(extend_one)]
#![feature(hash_raw_entry)]
#![feature(hasher_prefixfree_extras)]
Expand Down
39 changes: 21 additions & 18 deletions compiler/rustc_data_structures/src/profiling.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,6 +88,7 @@ use std::borrow::Borrow;
use std::collections::hash_map::Entry;
use std::error::Error;
use std::fs;
use std::intrinsics::unlikely;
use std::path::Path;
use std::process;
use std::sync::Arc;
Expand Down Expand Up @@ -395,11 +396,18 @@ impl SelfProfilerRef {
/// Record a query in-memory cache hit.
#[inline(always)]
pub fn query_cache_hit(&self, query_invocation_id: QueryInvocationId) {
Copy link
Member

Choose a reason for hiding this comment

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

The same should probably be done for the other profiling events as well

Copy link
Contributor Author

Choose a reason for hiding this comment

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

The problem was that it generated code for TimingGuard which was unused. The other events do make use of it so it outlining doesn't help.

self.instant_query_event(
|profiler| profiler.query_cache_hit_event_kind,
query_invocation_id,
EventFilter::QUERY_CACHE_HITS,
);
#[inline(never)]
#[cold]
fn cold_call(profiler_ref: &SelfProfilerRef, query_invocation_id: QueryInvocationId) {
profiler_ref.instant_query_event(
|profiler| profiler.query_cache_hit_event_kind,
query_invocation_id,
);
}

if unlikely(self.event_filter_mask.contains(EventFilter::QUERY_CACHE_HITS)) {
cold_call(self, query_invocation_id);
}
}

/// Start profiling a query being blocked on a concurrent execution.
Expand Down Expand Up @@ -444,20 +452,15 @@ impl SelfProfilerRef {
&self,
event_kind: fn(&SelfProfiler) -> StringId,
query_invocation_id: QueryInvocationId,
event_filter: EventFilter,
) {
drop(self.exec(event_filter, |profiler| {
let event_id = StringId::new_virtual(query_invocation_id.0);
let thread_id = get_thread_id();

profiler.profiler.record_instant_event(
event_kind(profiler),
EventId::from_virtual(event_id),
thread_id,
);

TimingGuard::none()
}));
let event_id = StringId::new_virtual(query_invocation_id.0);
let thread_id = get_thread_id();
let profiler = self.profiler.as_ref().unwrap();
profiler.profiler.record_instant_event(
event_kind(profiler),
EventId::from_virtual(event_id),
thread_id,
);
}

pub fn with_profiler(&self, f: impl FnOnce(&SelfProfiler)) {
Expand Down
2 changes: 1 addition & 1 deletion compiler/rustc_query_system/src/dep_graph/graph.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ impl DepNodeIndex {
}

impl From<DepNodeIndex> for QueryInvocationId {
#[inline]
#[inline(always)]
fn from(dep_node_index: DepNodeIndex) -> Self {
QueryInvocationId(dep_node_index.as_u32())
}
Expand Down
12 changes: 3 additions & 9 deletions compiler/rustc_query_system/src/query/plumbing.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -346,9 +346,7 @@ where
{
match cache.lookup(&key) {
Some((value, index)) => {
if std::intrinsics::unlikely(tcx.profiler().enabled()) {
tcx.profiler().query_cache_hit(index.into());
}
tcx.profiler().query_cache_hit(index.into());
tcx.dep_graph().read_index(index);
Some(value)
}
Expand Down Expand Up @@ -408,9 +406,7 @@ where
panic!("value must be in cache after waiting")
};

if std::intrinsics::unlikely(qcx.dep_context().profiler().enabled()) {
qcx.dep_context().profiler().query_cache_hit(index.into());
}
qcx.dep_context().profiler().query_cache_hit(index.into());
query_blocked_prof_timer.finish_with_query_invocation_id(index.into());

(v, Some(index))
Expand Down Expand Up @@ -776,9 +772,7 @@ where
// Ensure that only one of them runs the query.
let cache = Q::query_cache(qcx);
if let Some((_, index)) = cache.lookup(&key) {
if std::intrinsics::unlikely(qcx.dep_context().profiler().enabled()) {
qcx.dep_context().profiler().query_cache_hit(index.into());
}
qcx.dep_context().profiler().query_cache_hit(index.into());
return;
}

Expand Down