Testcase simulating js execution of Backbone-SP3 is 13x-2x slower in different scenarios.
Categories
(Core :: JavaScript Engine, task, P3)
Tracking
()
People
(Reporter: mayankleoboy1, Unassigned)
References
(Blocks 1 open bug)
Details
Attachments
(1 file)
|
3.92 KB,
text/html
|
Details |
STR:
- Open testcase
- From the dropdown, select each mode in order.
- Run the test
- Refresh and go back to step2
In order of dropdown
Firefox: https://share.firefox.dev/4eEp0Z1 (460ms, 1.3s, 900ms, 220ms, 650ms)
Chrome: https://share.firefox.dev/4w2me6K (110ms, 100ms, 600ms, 100ms, 100ms)
|
Reporter | |
Comment 1•1 month ago
•
|
||
I ran backbone-sp3 in samply and fed the profile to Chatgpt. I also shared the link to SP3 github repo. I then asked it to find hotspots in the profile, and based on the pattern/shape of Backbone, create a minimal testcase.
The testcase is generated by chatgpt, so its resemblance with actual Backbone-SP3 should be verified.
This is the profile i gave to chatgpt: https://share.firefox.dev/4b2QyWQ
|
||
Updated•1 month ago
|
|
||
Comment 2•3 days ago
|
||
The biggest gap here is the scripted version, which is a deep chain of calls to a leaf function that does nothing. All these functions are defined nested in V8 inlines it all away; we inline most of it, but we still spend a lot of time guarding that the callee function is still what we expect. We have this design doc describing a potential solution to that problem, but it may require some finicky work in the parser to get the information we need.
Description
•