[2.38] Taking some IOS GC branches for WPE as well#1422
[2.38] Taking some IOS GC branches for WPE as well#1422tomasz-karczewski-red wants to merge 3 commits intoWebPlatformForEmbedded:wpe-2.38from
Conversation
magomez
left a comment
magomez
left a comment
There was a problem hiding this comment.
Please, we have here 4 changes that should go in 4 separate commits. Well, 3, as the change in the MemoryPressureHandler fractions should not be there. Please, remove the fractions change and split the PR in 3 different commits.
| WTF_EXPORT_PRIVATE bool MemoryPressureHandler::ReliefLogger::s_loggingEnabled = false; | ||
|
|
||
| #if PLATFORM(IOS_FAMILY) | ||
| #if PLATFORM(IOS_FAMILY) || PLATFORM(WPE) |
There was a problem hiding this comment.
We cannot change the default values of the fractions, cause this affects all of the devices. As I told you in previous PRs, there's API to change these values for your platform only, please use it.
The aim is to make the GC more reliable on low-memory platforms. In FullGCActivityCallback, the change is to do not bail out of GC early.
The aim is to make the GC more reliable on low-memory platforms. In MemoryRelease, the change is to request immediate GC in releaseCriticalMemory.
The aim is to make the GC more reliable on low-memory platforms. In WorkerGlobalScope::deleteJSCodeAndGC the change is to collect even if called with 'Synchronous::No'
tomasz-karczewski-red
force-pushed
the
RDK-588_GC_finetuning_use_IOS_branches
branch
from
7cd0bde to
8c1c5e5
Compare
|
sure - I've just updated the PR |
|
@modeveci , this is technically fine for merge, but it's adding changes to how the memory release is done and may have an impact in performance, so it should be verified on your side first. |
|
Thanks @magomez, we'll investigate if there are any performance issues. |
|
Hi |
4 participants
The aim is to make the GC more reliable on low-memory platforms. IOS branches applied: