Remove Baseline's WarmUpCount IC
Categories
(Core :: JavaScript Engine: JIT, task, P3)
Tracking
()
Tracking | Status | |
---|---|---|
firefox71 | --- | fixed |
People
(Reporter: jandem, Assigned: jandem)
References
Details
Attachments
(3 files)
We never attach any stubs to this IC so it could just be a plain VM call and/or trampoline code. Furthermore, the Baseline Interpreter doesn't use this IC at all and it's not very nice to have this JIT-specific IC in JitScript.
Assignee | ||
Comment 1•5 years ago
|
||
Bug 1583172 touches this code too so let's wait for that to land first.
Updated•5 years ago
|
Assignee | ||
Comment 2•5 years ago
|
||
This lets us simplify the calling code because the VMFunction wrapper code now does
the right thing.
Assignee | ||
Comment 3•5 years ago
|
||
We never attached stubs to this and the Baseline Interpreter does not use this IC at all
so it's nicer and simpler to use a plain VM call.
Depends on D47880
Assignee | ||
Comment 4•5 years ago
|
||
-
Store the pointer in JitRuntime instead of JSContext.
-
Use UniquePtr for this pointer. I think this may also fix a memory leak
(when realloc failed we didn't free the old allocation). -
Call it ionOsrTempData instead of osrTempData.
-
Use placement new and remove an unnecessary memset.
Depends on D47881
Pushed by jdemooij@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/ed2fda501785 part 1 - Make VMFunction wrapper treat IonOsrTempData** as outparam. r=tcampbell https://hg.mozilla.org/integration/autoland/rev/47830868674f part 2 - Remove Baseline's WarmUpCount IC, use a plain VM call instead. r=tcampbell https://hg.mozilla.org/integration/autoland/rev/989b75468b7e part 3 - Clean up IonOsrTempData code a bit. r=tcampbell
Assignee | ||
Updated•5 years ago
|
Comment 6•5 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/ed2fda501785
https://hg.mozilla.org/mozilla-central/rev/47830868674f
https://hg.mozilla.org/mozilla-central/rev/989b75468b7e
Description
•