I’d like to see a de facto standard way to proxy Vulkan across processes. It’s silly that every secure app has to reinvent it.
-
-
Replying to @pcwalton
Proxying OpenGL calls the way Chrome does makes me unhappy…GL is tricky and perf, security suffers. GPU remoting needs a holistic design.
2 replies 2 retweets 5 likes -
Replying to @pcwalton
At the W3C GPU group, we (esp chrome) are trying to design a modern GPU API that takes multiprocess as a first class consideration.
2 replies 0 retweets 0 likes -
Replying to @kainino0x
Can we share an IPC implementation? Heck, I’ll volunteer to implement ;D
1 reply 0 retweets 0 likes -
Replying to @pcwalton
We're absolutely interested in sharing as much implementation as desired across all browsers. The backends will be quite a piece of work.
1 reply 0 retweets 0 likes -
Replying to @kainino0x @pcwalton
Would love to continue this discussion with more folks in the room, maybe via public-gpu@w3.org? cc:
@DaKangz1 reply 0 retweets 0 likes -
Replying to @kainino0x @pcwalton
Yep, the ML is the best place to reach everyone, even if the group is for a Web facing API. Our prototype already has IPC, but is in C++ ;-)
1 reply 0 retweets 0 likes -
Replying to @DaKangz @kainino0x
How are you planning to prevent the new sandbox escapes that will occur when you introduce that much new API surface?
1 reply 0 retweets 0 likes -
Replying to @pcwalton @kainino0x
We validate the API calls are well-formed so you can't probe the driver directly. The IPC and state-tracking would obviously be fuzzed.
3 replies 0 retweets 0 likes
(Opened an issue: https://github.com/google/nxt-standalone/issues/34 …)
-
-
Replying to @pcwalton @kainino0x
Thanks! That's a better medium to have technical discussions than Twitter :)
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.