worker: improve MessagePort performance #31605
Closed
Conversation
This file contains hidden or 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
Use a JS function as the single entry point for emitting `.onmessage()` calls, avoiding the overhead of manually constructing each message event object in C++. confidence improvement accuracy (*) (**) (***) worker/echo.js n=100000 sendsPerBroadcast=1 payload='object' workers=1 *** 16.34 % ±1.16% ±1.54% ±1.99% worker/echo.js n=100000 sendsPerBroadcast=1 payload='string' workers=1 *** 24.41 % ±1.50% ±1.99% ±2.58% worker/echo.js n=100000 sendsPerBroadcast=10 payload='object' workers=1 *** 26.66 % ±1.54% ±2.05% ±2.65% worker/echo.js n=100000 sendsPerBroadcast=10 payload='string' workers=1 *** 32.72 % ±1.60% ±2.11% ±2.73% worker/messageport.js n=1000000 payload='object' *** 40.28 % ±1.48% ±1.95% ±2.52% worker/messageport.js n=1000000 payload='string' *** 76.95 % ±2.19% ±2.90% ±3.75% Also fix handling exceptions returned from `MessagePort::New`.
addaleax added performance 
nodejs-github-bot added c++ 
Collaborator
jasnell approved these changes Feb 1, 2020
Collaborator
lundibundi approved these changes Feb 2, 2020
Trott approved these changes Feb 3, 2020
lundibundi approved these changes Feb 4, 2020
addaleax added the author ready Collaborator
Collaborator
Collaborator
Trott pushed a commit that referenced this pull request Feb 7, 2020
Use a JS function as the single entry point for emitting `.onmessage()` calls, avoiding the overhead of manually constructing each message event object in C++. confidence improvement accuracy (*) (**) (***) worker/echo.js n=100000 sendsPerBroadcast=1 payload='object' workers=1 *** 16.34 % ±1.16% ±1.54% ±1.99% worker/echo.js n=100000 sendsPerBroadcast=1 payload='string' workers=1 *** 24.41 % ±1.50% ±1.99% ±2.58% worker/echo.js n=100000 sendsPerBroadcast=10 payload='object' workers=1 *** 26.66 % ±1.54% ±2.05% ±2.65% worker/echo.js n=100000 sendsPerBroadcast=10 payload='string' workers=1 *** 32.72 % ±1.60% ±2.11% ±2.73% worker/messageport.js n=1000000 payload='object' *** 40.28 % ±1.48% ±1.95% ±2.52% worker/messageport.js n=1000000 payload='string' *** 76.95 % ±2.19% ±2.90% ±3.75% Also fix handling exceptions returned from `MessagePort::New`. PR-URL: #31605 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Denys Otrishko <shishugi@gmail.com> Reviewed-By: Rich Trott <rtrott@gmail.com>
Member
| Landed in a555be2 |
addaleax added a commit to addaleax/node that referenced this pull request Feb 8, 2020
a555be2 re-assigned `async_.data` to indicate success or failure of the constructor. As the `HandleWrap` implementation uses that field to access the `HandleWrap` instance from the libuv handle, this introduced two issues: - It implicitly assumed that casting `MessagePort*` → `void*` → `HandleWrap*` would be valid. - It made the `HandleWrap::OnClose()` function fail with a `nullptr` dereference if the constructor did fail. In particular, the second issue made test/parallel/test-worker-cleanexit-with-moduleload.js` crash at least once in CI. Since re-assigning `async_.data` isn’t actually necessary here (only a leftover from earlier versions of that commit), fix this by using a local variable instead, and add a `CHECK` that provides better error messages for this type of issue in the future. Refs: nodejs#31605
2 tasks
addaleax added a commit that referenced this pull request Feb 10, 2020
a555be2 re-assigned `async_.data` to indicate success or failure of the constructor. As the `HandleWrap` implementation uses that field to access the `HandleWrap` instance from the libuv handle, this introduced two issues: - It implicitly assumed that casting `MessagePort*` → `void*` → `HandleWrap*` would be valid. - It made the `HandleWrap::OnClose()` function fail with a `nullptr` dereference if the constructor did fail. In particular, the second issue made test/parallel/test-worker-cleanexit-with-moduleload.js` crash at least once in CI. Since re-assigning `async_.data` isn’t actually necessary here (only a leftover from earlier versions of that commit), fix this by using a local variable instead, and add a `CHECK` that provides better error messages for this type of issue in the future. Refs: #31605 PR-URL: #31696 Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: David Carlier <devnexen@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com>
codebytere pushed a commit that referenced this pull request Feb 17, 2020
Use a JS function as the single entry point for emitting `.onmessage()` calls, avoiding the overhead of manually constructing each message event object in C++. confidence improvement accuracy (*) (**) (***) worker/echo.js n=100000 sendsPerBroadcast=1 payload='object' workers=1 *** 16.34 % ±1.16% ±1.54% ±1.99% worker/echo.js n=100000 sendsPerBroadcast=1 payload='string' workers=1 *** 24.41 % ±1.50% ±1.99% ±2.58% worker/echo.js n=100000 sendsPerBroadcast=10 payload='object' workers=1 *** 26.66 % ±1.54% ±2.05% ±2.65% worker/echo.js n=100000 sendsPerBroadcast=10 payload='string' workers=1 *** 32.72 % ±1.60% ±2.11% ±2.73% worker/messageport.js n=1000000 payload='object' *** 40.28 % ±1.48% ±1.95% ±2.52% worker/messageport.js n=1000000 payload='string' *** 76.95 % ±2.19% ±2.90% ±3.75% Also fix handling exceptions returned from `MessagePort::New`. PR-URL: #31605 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Denys Otrishko <shishugi@gmail.com> Reviewed-By: Rich Trott <rtrott@gmail.com>
codebytere pushed a commit that referenced this pull request Feb 17, 2020
a555be2 re-assigned `async_.data` to indicate success or failure of the constructor. As the `HandleWrap` implementation uses that field to access the `HandleWrap` instance from the libuv handle, this introduced two issues: - It implicitly assumed that casting `MessagePort*` → `void*` → `HandleWrap*` would be valid. - It made the `HandleWrap::OnClose()` function fail with a `nullptr` dereference if the constructor did fail. In particular, the second issue made test/parallel/test-worker-cleanexit-with-moduleload.js` crash at least once in CI. Since re-assigning `async_.data` isn’t actually necessary here (only a leftover from earlier versions of that commit), fix this by using a local variable instead, and add a `CHECK` that provides better error messages for this type of issue in the future. Refs: #31605 PR-URL: #31696 Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: David Carlier <devnexen@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com>
Merged
targos pushed a commit that referenced this pull request Apr 18, 2020
Use a JS function as the single entry point for emitting `.onmessage()` calls, avoiding the overhead of manually constructing each message event object in C++. confidence improvement accuracy (*) (**) (***) worker/echo.js n=100000 sendsPerBroadcast=1 payload='object' workers=1 *** 16.34 % ±1.16% ±1.54% ±1.99% worker/echo.js n=100000 sendsPerBroadcast=1 payload='string' workers=1 *** 24.41 % ±1.50% ±1.99% ±2.58% worker/echo.js n=100000 sendsPerBroadcast=10 payload='object' workers=1 *** 26.66 % ±1.54% ±2.05% ±2.65% worker/echo.js n=100000 sendsPerBroadcast=10 payload='string' workers=1 *** 32.72 % ±1.60% ±2.11% ±2.73% worker/messageport.js n=1000000 payload='object' *** 40.28 % ±1.48% ±1.95% ±2.52% worker/messageport.js n=1000000 payload='string' *** 76.95 % ±2.19% ±2.90% ±3.75% Also fix handling exceptions returned from `MessagePort::New`. PR-URL: #31605 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Denys Otrishko <shishugi@gmail.com> Reviewed-By: Rich Trott <rtrott@gmail.com>
targos pushed a commit to targos/node that referenced this pull request Apr 25, 2020
a555be2 re-assigned `async_.data` to indicate success or failure of the constructor. As the `HandleWrap` implementation uses that field to access the `HandleWrap` instance from the libuv handle, this introduced two issues: - It implicitly assumed that casting `MessagePort*` → `void*` → `HandleWrap*` would be valid. - It made the `HandleWrap::OnClose()` function fail with a `nullptr` dereference if the constructor did fail. In particular, the second issue made test/parallel/test-worker-cleanexit-with-moduleload.js` crash at least once in CI. Since re-assigning `async_.data` isn’t actually necessary here (only a leftover from earlier versions of that commit), fix this by using a local variable instead, and add a `CHECK` that provides better error messages for this type of issue in the future. Refs: nodejs#31605 PR-URL: nodejs#31696 Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: David Carlier <devnexen@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com>
targos pushed a commit that referenced this pull request Apr 28, 2020
a555be2 re-assigned `async_.data` to indicate success or failure of the constructor. As the `HandleWrap` implementation uses that field to access the `HandleWrap` instance from the libuv handle, this introduced two issues: - It implicitly assumed that casting `MessagePort*` → `void*` → `HandleWrap*` would be valid. - It made the `HandleWrap::OnClose()` function fail with a `nullptr` dereference if the constructor did fail. In particular, the second issue made test/parallel/test-worker-cleanexit-with-moduleload.js` crash at least once in CI. Since re-assigning `async_.data` isn’t actually necessary here (only a leftover from earlier versions of that commit), fix this by using a local variable instead, and add a `CHECK` that provides better error messages for this type of issue in the future. Refs: #31605 PR-URL: #31696 Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: David Carlier <devnexen@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
5 participants
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.
Use a JS function as the single entry point for emitting
.onmessage()calls, avoiding the overhead of manually constructing each message
event object in C++.
Also fix handling exceptions returned from
MessagePort::New.Checklist
make -j4 test(UNIX), orvcbuild test(Windows) passes