Switch back to using one threadpool #7898
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This reverts commit 0d9d214.
The previous method ended up reading lots of small byte buffers, and then joining them quadratically. This just fills a buffer.
ity
approved these changes
Jun 19, 2019
illicitonion
added a commit
to twitter/pants
that referenced
this pull request
Jul 3, 2019
As per https://docs.rs/tokio-threadpool/0.1.15/tokio_threadpool/fn.blocking.html when you use a blocking block in a task, it blocks the _entire_ task, not just the bit which you're marking blocking. This means that when we're in one giant join_all calling read_dir on a bunch of directories, we end up serializing them to only read one directory at a time. This change introduces a new logging::Executor (a slightly weird place for it, but it depends on the logging stuf, so...) which mirrors the tokio::Runtime and the futures::sync::oneshot::spawn APIs to conveniently allow blocking Futures to be spawned in their own tasks, so that they don't hinder parallelism within a task.
illicitonion
added a commit
to twitter/pants
that referenced
this pull request
Jul 3, 2019
As per https://docs.rs/tokio-threadpool/0.1.15/tokio_threadpool/fn.blocking.html when you use a blocking block in a task, it blocks the _entire_ task, not just the bit which you're marking blocking. This means that when we're in one giant join_all calling read_dir on a bunch of directories, we end up serializing them to only read one directory at a time. This change introduces a new logging::Executor (a slightly weird place for it, but it depends on the logging stuf, so...) which mirrors the tokio::Runtime and the futures::sync::oneshot::spawn APIs to conveniently allow blocking Futures to be spawned in their own tasks, so that they don't hinder parallelism within a task.
illicitonion
added a commit
to twitter/pants
that referenced
this pull request
Jul 3, 2019
As per https://docs.rs/tokio-threadpool/0.1.15/tokio_threadpool/fn.blocking.html when you use a blocking block in a task, it blocks the _entire_ task, not just the bit which you're marking blocking. This means that when we're in one giant join_all calling read_dir on a bunch of directories, we end up serializing them to only read one directory at a time. This change introduces a new logging::Executor (a slightly weird place for it, but it depends on the logging stuf, so...) which mirrors the tokio::Runtime and the futures::sync::oneshot::spawn APIs to conveniently allow blocking Futures to be spawned in their own tasks, so that they don't hinder parallelism within a task.
illicitonion
added a commit
to twitter/pants
that referenced
this pull request
Jul 5, 2019
tokio-fs has some weird semantics whereby using blocking in a larger task (e.g. mapping stats from the result of a readdir) forces the work to not be done in parallel. I'm pretty sure we _can_ make tokio-fs do what we want efficiently, but it's taking longer than I'd hoped, so I'm re-introducing the separate io threadpool so we don't have a performance regression while I investigate. This reverts a small portion of pantsbuild#7898 manually.
illicitonion
added a commit
to twitter/pants
that referenced
this pull request
Jul 5, 2019
tokio-fs has some weird semantics whereby using blocking in a larger task (e.g. mapping stats from the result of a readdir) forces the work to not be done in parallel. I'm pretty sure we _can_ make tokio-fs do what we want efficiently, but it's taking longer than I'd hoped, so I'm re-introducing the separate io threadpool so we don't have a performance regression while I investigate. This reverts a small portion of pantsbuild#7898 manually.
illicitonion
added a commit
that referenced
this pull request
Jul 10, 2019
* Introduce task_executor This is a single object we can pass around to allow all sorts of future running to happen, with logging happening properly, rather than needing to pass around different threadpools and manually sort out logging. * Core has an Executor not a Runtime * Scandir runs on io pool * PosixFS uses IO pool * ShardedLMDB uses io pool * Calculate fingeprint on io pool * Add TODO to move Executor some time * Add docstring to spawn_on_io_pool * Move Executor to its own crate * fmt
illicitonion
added a commit
that referenced
this pull request
Jul 10, 2019
* Introduce task_executor This is a single object we can pass around to allow all sorts of future running to happen, with logging happening properly, rather than needing to pass around different threadpools and manually sort out logging. * Core has an Executor not a Runtime * Scandir runs on io pool * PosixFS uses IO pool * ShardedLMDB uses io pool * Calculate fingeprint on io pool * Add TODO to move Executor some time * Add docstring to spawn_on_io_pool * Move Executor to its own crate * fmt
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This reverts #7848 and fixes the performance issue found in it.
The first commit is just a revert (with a few manual merge cleanups), the second commit fixes the performance issue.
Fixes #7896