!FZyQrssSlHEZqrYcOb:matrix.org

In WebGPU we Rust

914 Members
0.14 Release! 57 Servers

Load older messages


SenderMessageTime
6 Oct 2022
@i509vcb:matrix.orgi509vcbI assume anything you reference within a recorded render bundle must live as long as you continue to replay the bundle?18:43:46
@cwfitzgerald:matrix.orgcwfitzgeraldIt will live that long, but you don't need to keep it alive, the render bundle owns it18:44:44
@i509vcb:matrix.orgi509vcbOkay so under the hood the render bundle will keep anything it needs alive 18:45:14
@cwfitzgerald:matrix.orgcwfitzgeraldYes, once you call finish18:45:41
@i509vcb:matrix.orgi509vcbI assume the underlying vertex/instance buffers can't be mutated then or else a draw command could have not enough data18:46:53
@i509vcb:matrix.orgi509vcbSounds really like something you would use for something very static 18:47:12
@cwfitzgerald:matrix.orgcwfitzgeraldYou can mutate them fine18:47:12
@i509vcb:matrix.orgi509vcbWell the draw count couldn't be changed though?18:47:27
@cwfitzgerald:matrix.orgcwfitzgeraldCorrect18:47:37
@cwfitzgerald:matrix.orgcwfitzgerald(unless it was an indirect call) 18:47:45
@gents:matrix.orgGentsMmm I'm using indirect 😅 so render bundles will not actually make it anything better right? Or implementing secondary command buffers for them would make them more interesting?18:50:13
@i509vcb:matrix.orgi509vcbI am not working on replaying commands yet, but it could be useful in the case that I could be a rendering a surface into 3 dmabuf all of different formats18:52:26
@i509vcb:matrix.orgi509vcb(display 1, display 2, screen capture for example)18:53:01
@cwfitzgerald:matrix.orgcwfitzgerald
In reply to @i509vcb:matrix.org
I am not working on replaying commands yet, but it could be useful in the case that I could be a rendering a surface into 3 dmabuf all of different formats
Pipelines have the format baked in though
19:29:35
@i509vcb:matrix.orgi509vcbYeah I know of the pipelines as well19:29:57
@crowlkats:matrix.orgcrowlkats cwfitzgerald so investigated the trouble of running newer commits of cts: basically cts used to have an out/ directory which the deno script would use as it contained the transpiled typescript that is being used for the node runner, however at some point that out dir was taken out of git and is just gitignored now, and building it will need to have node installed in CI to generate it, though even then, i tried and no clue how to generate it... ideally cts would commit that folder in again, but unsure who to talk to about that. 20:12:53
@cwfitzgerald:matrix.orgcwfitzgeraldinstalling node in CI is also an option if that would be easier20:14:41
@cwfitzgerald:matrix.orgcwfitzgerald #WebGPU:matrix.org would be where to talk about it (besides CTS github issues) 20:14:57
@crowlkats:matrix.orgcrowlkats
In reply to @cwfitzgerald:matrix.org
#WebGPU:matrix.org would be where to talk about it (besides CTS github issues)
aight, will ask there
20:16:38
@cwfitzgerald:matrix.orgcwfitzgeraldthanks for looking into this!20:16:54
@crowlkats:matrix.orgcrowlkats
In reply to @cwfitzgerald:matrix.org
thanks for looking into this!
alright, cts is now working. some failing tests and no idea what the cause is for those; doubt i will be able to help too much with that though
22:00:14
@cwfitzgerald:matrix.orgcwfitzgeraldcool I can take it from here22:00:52
@crowlkats:matrix.orgcrowlkats also, i have no clue about this test.lst file, or how to properly have it update 22:01:07
@cwfitzgerald:matrix.orgcwfitzgeraldit's an intentional allow list so we only run stuff we can pass22:01:29
@cwfitzgerald:matrix.orgcwfitzgeraldI think what we should be doing probably is running the whole CTS and then diffing the results22:02:08
@crowlkats:matrix.orgcrowlkatsah ok22:02:36
@cwfitzgerald:matrix.orgcwfitzgerald crowlkats: what exactly does the cts_runner component do? 22:10:15
@crowlkats:matrix.orgcrowlkatsbasically just create a very slim deno binary with just webgpu & read and write file functionality22:12:00
@crowlkats:matrix.orgcrowlkats * basically just create a very slim deno binary with just deno_webgpu & read and write file functionality22:12:31
@unrelentingtech:mozilla.org@unrelentingtech:mozilla.org left the room.22:15:29

There are no newer messages yet.


Back to Room List