-
Notifications
You must be signed in to change notification settings - Fork 248
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Tracking issue for glam refactoring #304
Comments
I have rebased the spriv changes from master into my refactor branch and changed attributes so that outer types like |
So you know, once you've built for |
I've made a gist with the diassasmbled spirv comparing glam master to glam refactor here: https://gist.github.com/bitshifter/dcf83dc148b05a58d4fb1753949aa4fe I think the It looks like the use of traits is bloating the spirv and preventing inlining in some cases. It would be good to know if this is a Rust GPU compiler issue or a more general problem. I haven't got to the point where I was looking at performance and codegen on CPU, but I guess now I am worried that my current approach might have too much of a performance penalty. Anyway, it would be good to better understand the implications on the Rust GPU side. Thanks. I'm not locked in to my current solution although if if performance and codegen seems OK on regular Rust I would most likely stick with it, unless it turns out that it's going to cause a lot of problems for Rust GPU. |
This is at the point where the bulk of the work has been done, f64, i32 and u32 types have been added. Performance for f32 seems as good or better than before (for the SSE2 implementation). It might be good if someone from Embark could check that this branch isn't going to cause any bit issues for rust-gpu. I still have a bunch of tidying to do on the branch, but I imagine it will make it into master sometime in the next couple of weeks. |
This has been merged to master. It's a pretty significant refactor so any branches are most likely going to have conflicts. LMK if you have any questions. |
Sorry we've been quiet on this! Just smoke-tested and tried building rust-gpu against glam master, everything seems fine. I don't seem to be reproducing the OpName bloating shenanigans, so idk, but in general we're probably not too worried about binary size and performance right now, we're more focused on just making things work. |
Refactor completed! Probably should have closed this a while ago. |
Our main vector math crate, glam-rs is currently being refactored to remove code duplication. This is an issue for tracking that progress and to make sure glam-rs continues to work with these changes.
Glam issue: bitshifter/glam-rs#99
Glam branch: https://github.com/bitshifter/glam-rs/tree/refactor
The text was updated successfully, but these errors were encountered: