Metal optimizations #2062
Merged
Metal optimizations #2062
Conversation
This file contains 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
- Significant changes to optimize memory usage. - Adjusted formatting of the metal shader code to improve readability. - Normalized naming conventions in shader code. - Abstracted repetitive code for attribute descriptors to a helper function.
|
@mitchellh I'm happy to test the builds on my Intel macOS 13 potato, but that machine is not set up for Zig builds. Could you enable the build workflows for this PR so I can grab the artefacts for testing? (I'm also open for suggestions what to test specifically to exercise all the codepaths.) |
|
Apparently I can't build the release from 3rd party forks :( Let me see what I can do. |
mitchellh
requested changes
Aug 8, 2024
| @@ -845,6 +678,41 @@ fn initImagePipeline(device: objc.Object, library: objc.Object) !objc.Object { | |||
| return pipeline_state; | |||
| } | |||
|
|
|||
| fn autoAttribute(T: type, attrs: objc.Object) void { | |||
There was a problem hiding this comment.
Well well well aren't we fancy. ❤️
|
asciinema: bug.cast.zip |
|
That's one way, I guess :) Happy to report this build looks fine and even fixes the regression with the earlier build I reported on Discord. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This PR significantly reworks the Metal shaders for memory efficiency. Additionally I've adjusted the formatting of the shader code for readability, and normalized the naming conventions.
There are two main changes that affect memory efficiency.
cols * rowstotaluchar4s representing the cell colors, which are drawn with a single fullscreen shader.[2]i32to[2]i16, it seems like a sane expectation to me that the bearings will never be more than 32K pixels in a given direction.Beyond this, I have added a couple other small optimizations, such as a fullscreen vertex shader that uses a single triangle, and using pixel coordinates to sample the glyph texture so that conversion to normalized coordinates isn't necessary.
Results
Experimentally I saw a ~20% reduction in total GPU memory use, and a ~80% reduction in max allocated buffer size.
TODO
Caution
There is a very real chance that this PR could have issues under macOS 13 for Intel processors, due to the known bugginess of the macOS 13 Intel Metal drivers. Comprehensive testing on an Intel mac running macOS 13 must be performed before this can be merged.