Add parallel front end goal in 25h2 #330
Conversation
src/2025h2/parallel-front-end.md
Outdated
| - Enable parallel front end in bootstrap. | ||
| - Continue to improve parallel compilation performance, via parallelize the macro expansion and reduce data competition. | ||
| - Enable parallel front end in Cargo. | ||
| - Add more test cases for parallel front end in rustc-perf |
There was a problem hiding this comment.
Can add a couple more items
- Enable the parallel frontend by default on nightly, but not on stable, for preliminary testing (similarly to LLD)
- Write stabilization report for the feature and submit it for compiler FCP
There was a problem hiding this comment.
Are you and your team planning to work on this project goal with sparrow? Just the 6 items in sparrow's list seem like a good chunk of work for a single person.
There was a problem hiding this comment.
Are you and your team planning to work on this project goal with sparrow? Just the 6 items in sparrow's list seem like a good chunk of work for a single person.
Hopefully.
I just think a visible push for stabilization, initial stabilization report, and enabling it on nightly should happen sooner than many other items on the list.
There was a problem hiding this comment.
Thanks! I added the second one.
And enable it by default (even in nightly) might require more statistics, as it may cause performance regressions compared to single-threaded in many cases.
Parallel front end progressed well in the first half of the year. We resolved the deadlock issue caused by the deadlock handler, added the rustc-rayon that the parallel front end depends on to the rustc working tree, and optimized the work-stealing scheduling algorithm to avoid deadlocks in the parallelization of rustc queries.
Rendered