Thomas de Zeeuw
2fcc6c2cb0
Box Futures in #[tokio::test]
...
This reduces the amount of copies of the Runtime::block_on and related
functions the compiler has to generate and LLVM process. We've seen it
reduce the compilation time of our tests (some 1900 of them) from 40s
down to 12s, with no impact on the runtime of tests.
Below is an output of llvm-lines for our tests.
Before:
Lines Copies Function name
----- ------ -------------
8954414 156577 (TOTAL)
984626 (11.0%, 11.0%) 9289 (5.9%, 5.9%) std:🧵 :local::LocalKey<T>::try_with
648093 (7.2%, 18.2%) 1857 (1.2%, 7.1%) tokio::runtime::scheduler::current_thread::CoreGuard::block_on::{{closure}}
557100 (6.2%, 24.5%) 3714 (2.4%, 9.5%) tokio::park:🧵 :CachedParkThread::block_on
551679 (6.2%, 30.6%) 7430 (4.7%, 14.2%) tokio::coop::with_budget::{{closure}}
514389 (5.7%, 36.4%) 3714 (2.4%, 16.6%) tokio::runtime::scheduler::current_thread::Context::enter
326832 (3.6%, 40.0%) 1857 (1.2%, 17.8%) tokio::runtime::scheduler::current_thread::CurrentThread::block_on
291549 (3.3%, 43.3%) 1857 (1.2%, 19.0%) tokio::runtime::scheduler::current_thread::CoreGuard::enter
261907 (2.9%, 46.2%) 7430 (4.7%, 23.7%) tokio::coop::budget
189468 (2.1%, 48.3%) 7430 (4.7%, 28.5%) tokio::coop::with_budget
137418 (1.5%, 49.8%) 3714 (2.4%, 30.8%) tokio::runtime::enter::Enter::block_on
126276 (1.4%, 51.3%) 1857 (1.2%, 32.0%) tokio::runtime::Runtime::block_on
124419 (1.4%, 52.6%) 1857 (1.2%, 33.2%) tokio::macros::scoped_tls::ScopedKey<T>::set
118897 (1.3%, 54.0%) 3715 (2.4%, 35.6%) core::option::Option<T>::or_else
111420 (1.2%, 55.2%) 1857 (1.2%, 36.8%) tokio::runtime::scheduler::current_thread::CurrentThread::block_on::{{closure}}
109408 (1.2%, 56.4%) 2105 (1.3%, 38.1%) <core::future::from_generator::GenFuture<T> as core::future::future::Future>::poll
105893 (1.2%, 57.6%) 9289 (5.9%, 44.0%) std:🧵 :local::LocalKey<T>::with
96564 (1.1%, 58.7%) 1857 (1.2%, 45.2%) tokio::runtime::scheduler::current_thread::Context::run_task
90993 (1.0%, 59.7%) 7428 (4.7%, 50.0%) tokio::runtime::scheduler::current_thread::CoreGuard::block_on::{{closure}}::{{closure}}
90515 (1.0%, 60.7%) 2105 (1.3%, 51.3%) core::pin::Pin<&mut T>::map_unchecked_mut
89136 (1.0%, 61.7%) 1857 (1.2%, 52.5%) tokio::runtime::scheduler::multi_thread::MultiThread::block_on
After:
Lines Copies Function name
----- ------ -------------
3188618 41634 (TOTAL)
109408 (3.4%, 3.4%) 2105 (5.1%, 5.1%) <core::future::from_generator::GenFuture<T> as core::future::future::Future>::poll
90515 (2.8%, 6.3%) 2105 (5.1%, 10.1%) core::pin::Pin<&mut T>::map_unchecked_mut
56220 (1.8%, 8.0%) 1874 (4.5%, 14.6%) alloc::boxed::Box<T>::pin
48333 (1.5%, 9.5%) 2179 (5.2%, 19.8%) core::ops::function::FnOnce::call_once
28587 (0.9%, 10.4%) 1 (0.0%, 19.8%) XXXXXXXXXXXXXXXXXXX
18730 (0.6%, 11.0%) 1873 (4.5%, 24.3%) alloc::boxed::Box<T,A>::into_pin
16190 (0.5%, 11.5%) 2 (0.0%, 24.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
15870 (0.5%, 12.0%) 2 (0.0%, 24.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
15250 (0.5%, 12.5%) 1 (0.0%, 24.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXX
12801 (0.4%, 12.9%) 2 (0.0%, 24.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
12801 (0.4%, 13.3%) 2 (0.0%, 24.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
12630 (0.4%, 13.7%) 2105 (5.1%, 29.4%) <core::future::from_generator::GenFuture<T> as core::future::future::Future>::poll::{{closure}}
12613 (0.4%, 14.1%) 2 (0.0%, 29.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
12613 (0.4%, 14.5%) 2 (0.0%, 29.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
12613 (0.4%, 14.9%) 2 (0.0%, 29.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
12613 (0.4%, 15.3%) 2 (0.0%, 29.4%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
11395 (0.4%, 15.7%) 96 (0.2%, 29.7%) alloc::alloc::box_free
11364 (0.4%, 16.0%) 1891 (4.5%, 34.2%) <T as core::convert::Into<U>>::into
11238 (0.4%, 16.4%) 1873 (4.5%, 38.7%) alloc::boxed::<impl core::convert::From<alloc::boxed::Box<T,A>> for core::pin::Pin<alloc::boxed::Box<T,A>>>::from
10735 (0.3%, 16.7%) 2 (0.0%, 38.7%) XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
Note that I have replaced our test functions with XXX. As you can
clearly see they're not in the top 20 in the before output, while
they're in the after oput.
Further note that the amount of copies have been reduced from 156577 to
41634.
2022-11-29 14:28:13 +00:00