News for hackers. Stuff that matters.
Re:GPU profiling for WebGPU workloads on Windows with Chrome by _factor on Thursday July 04, 2024 @12:23PM (#40876089)
I think the bigger issue is security. GPU access is too interlinked to kernel drivers. I’ve seen WebGPU and even WebGL crash otherwise stable systems.
Re:GPU profiling for WebGPU workloads on Windows with Chrome by pjmlp on Thursday July 04, 2024 @12:53PM (#40876338)
Yeah, however having a debugging tool, that developers have to explicitly open in developer tools, for debugging shaders and drivers that are always used by the browser anyway, doesn't change that.
Re:GPU profiling for WebGPU workloads on Windows with Chrome by stanleykm on Thursday July 04, 2024 @02:53PM (#40877154)
Agreed. Chrome already has flags to reduce the security posture for developers. It is not outrageous to expect you’d need to use those for profiling gpu shaders. This is a huge gap in the browser gpu ecosystem, especially since wgsl is missing features (like unrolling hints) that can result in dramatically different performance vs hlsl or whatever.
Re:GPU profiling for WebGPU workloads on Windows with Chrome by tw061023 on Thursday July 04, 2024 @11:21PM (#40879601)
WebGPU is supposed to be properly sandboxed.

The bigger issue is that WebGPU is basically dead on arrival for the same reasons WebGL is - it is impossible to get enough data to the client for it to actually matter, Awwwards-style design tricks notwithstanding.

I suppose browser vendors understand this and don't really care for either.

Re:GPU profiling for WebGPU workloads on Windows with Chrome by nmfisher on Thursday July 04, 2024 @11:26PM (#40879630)
What do you think could fix this? Access to genuinely permanent storage?
Re:GPU profiling for WebGPU workloads on Windows with Chrome by tw061023 on Friday July 05, 2024 @12:42AM (#40879886)
That's one part of it, yes. A browser API providing a few GBs of persistent storage with proper isolation and user management, obviously with some kind of compression/decompression going on to save both download times and loading times.

As an example, consider Infinity Blade, the poster child of mobile gaming: released in 2010, 595 MB download, 948 MB installed. Even the first version of WebGL is capable of providing this kind of experience, we just cannot get it to the user via browser.

Re:GPU profiling for WebGPU workloads on Windows with Chrome by pjmlp on Friday July 05, 2024 @03:29AM (#40880510)
I found out another fellow soul, that is exactly one of my complaints with Web 3D, the failure to 15 years later to provide the same experience as Infinity Blade, used by Apple to show off iPhone's OpenGL ES 3.0 capabilities.

Or Unreal Engine Citadel demo, originally done in Flash / C++.

Re:GPU profiling for WebGPU workloads on Windows with Chrome by ossobuco on Friday July 05, 2024 @07:37AM (#40881877)
That may be true (for now) for web apps, but what if you serve your app/game as a desktop app, for example with tauri?
Re:GPU profiling for WebGPU workloads on Windows with Chrome by _factor on Saturday July 06, 2024 @08:13AM (#40889943)
“WebGPU is supposed to be properly sandboxed.“

GPU must therefore provide open API for proprietary processing space.

Magic “packets” are therefore possible to execute arbitrary functions on “sandboxed” DMA devices.

Still a problem until we can audit the hardware. NV, and to a lesser extent AMD and ARC play somewhat open with a few omnipotent cards in their pockets. The prime of the issue is that gamers don’t care, only security professionals do. Because they’re the ones who see the 0-days fly by every day.