-
Notifications
You must be signed in to change notification settings - Fork 6k
[Impeller] migrate blur to calculating coefficients on the cpu #49512
Conversation
This comment was marked as outdated.
This comment was marked as outdated.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
Uber-nit: kernel as a name is sort of ambiguous. If we're able to get rid of the old gaussian, can we reclaim that name later?
Here's the thinking behind it. The fragment shader, by virtue of removing all the blur math, was left with just applying a Kernel. It's exactly the name for what it does. There are a few effects that can be implemented with just a Kernel and given Impeller's ethos of having a limited static set of shaders, it made more sense, to me, to make it available for everyone instead of naming it something specific. |
…140993) flutter/engine@0bbb4d6...f60d9a9 2024-01-04 [email protected] [Impeller] migrate blur to calculating coefficients on the cpu (flutter/engine#49512) If this roll has caused a breakage, revert this CL and stop the roller using the controls here: https://autoroll.skia.org/r/flutter-engine-flutter-autoroll Please CC [email protected],[email protected] on the revert to ensure that a human is aware of the problem. To file a bug in Flutter: https://github.com/flutter/flutter/issues/new/choose To report a problem with the AutoRoller itself, please file a bug: https://issues.skia.org/issues/new?component=1389291&template=1850622 Documentation for the AutoRoller is here: https://skia.googlesource.com/buildbot/+doc/main/autoroll/README.md
issue: flutter/flutter#131580
This factors out the gaussian blur fragment shader to be a more generic kernel fragment shader. We can't use this for the old gaussian blur because it's kernel is unbounded in one dimension. I implemented this by passing in the coefficients into the UBO so it can't support that.
Profile after change:
This is a 13% (21.25 / 25 ) drop in
average_gpu_frame_time
from the last profile in #49299.Pre-launch Checklist
///
).If you need help, consider asking for advice on the #hackers-new channel on Discord.