12:07:08 <RRSAgent> RRSAgent has joined #webgpu
12:07:08 <RRSAgent> logging to https://www.w3.org/2018/10/24-webgpu-irc
12:07:10 <Zakim> Zakim has joined #webgpu
12:07:12 <arnog> arnog has joined #webgpu
12:07:15 <tidoust> RRSAgent, make logs public
12:07:17 <sangchul> sangchul has joined #webgpu
12:07:20 <binji> binji has joined #webgpu
12:07:23 <sangwhan> sangwhan has joined #webgpu
12:07:24 <tidoust> Meeting: WebGPU TPAC breakout session
12:07:30 <tidoust> Chair: Corentin
12:07:39 <harry> harry has joined #webgpu
12:07:42 <halo> halo has joined #webgpu
12:07:42 <tidoust> scribenick: tidoust
12:07:50 <tidoust> Topic: WebGPU overview
12:08:00 <onishi> onishi has joined #webgpu
12:08:10 <krit> krit has joined #webgpu
12:08:13 <tidoust> Corentin: I chair the GPU for the Web Community Group, developing WebGPU.
12:08:17 <kazho> kazho has joined #webgpu
12:08:30 <tidoust> ... First question that usually gets raised is: why not WebGL instead?
12:08:34 <cpn> cpn has joined #webgpu
12:08:55 <tidoust> ... WebGL has a lot of historical bagage, which creates lot of CPU overhead. And also it does not map very well to modern graphic cards.
12:08:56 <JerryLin> JerryLin has joined #WebGPU
12:09:07 <ningxinhu> ningxinhu has joined #webgpu
12:09:16 <tidoust> ... The driver may be doing stuff for 100ms in some cases, resulting in dropped frames.
12:09:38 <tidoust> ... WebGL does not expose computation capabilities either.
12:09:53 <tidoust> ... And also, recently, Apple deprecated OpenGL.
12:10:10 <tidoust> ... WebGPU is like WebGL but more modern and without issue.
12:10:23 <wseltzer> wseltzer has joined #webgpu
12:10:32 <wseltzer> rrsagent, pointer?
12:10:32 <RRSAgent> See https://www.w3.org/2018/10/24-webgpu-irc#T12-10-32
12:10:33 <tidoust> ... It's inspired by modern GPU APIs (Metal, D3D12, Vulkan). It isn't any of them in particular, but close to them in particular.
12:10:49 <tidoust> ... Enables things such as faster games, more efficient machine learning on the Web, etc.
12:10:56 <Huangweichai> Huangweichai has joined #webgpu
12:11:26 <tidoust> ... Who's doing WebGPU? It's a W3C CG, and maybe soon as WG. Main browser vendors are engaged. Significant contributions from Intel and Yandex as well, for instance.
12:11:58 <tidoust> ... When will it be ready? We don't know, but we're making good progress on the API that allows scheduling of the computation on the GPU.
12:12:16 <tidoust> ... The discussions on the shading language are more difficult and progress has been slower on that front than expected.
12:12:50 <tidoust> ... Early next year, we'll have a good-enough-though-not-completely-stable API that we can share around and play with.
12:13:04 <tidoust> ?1: How is this API going to be consumed by the upper layer?
12:13:29 <xiaoqian> s/?1/Jungkee
12:13:34 <tidoust> Corentin: WebGPU is an abstraction layer over doing graphics. We don't expect a lot of people will use WebGPU directly for actual Web sites.
12:13:52 <takio> takio has joined #webgpu
12:13:53 <tidoust> ... The expectation is that most of the usage will be through frameworks.
12:14:01 <wseltzer> ... or game engines
12:14:17 <tidoust> ... Babylon.js, three.js are considering support for instance.
12:14:25 <wseltzer> ... unity and unreal
12:14:45 <tidoust> anssik: If you're moving to WG, how will the scope be different?
12:15:03 <tidoust> Corentin: The WG would be for stamping the spec, for IP reasons.
12:15:19 <tidoust> ?2: Do you consider an OpenCL replacement as well?
12:15:44 <tidoust> Corentin: OpenCL had some security issues, which is why WebCL did not progress, but the functionality should be there.
12:15:52 <anssik> s/?2/Arno/
12:16:21 <tidoust> ?3: Do you have some typical use cases that compare use of WebGPU vs. use of WebGL?
12:16:29 <tidoust> Corentin: One reason is advanced games.
12:16:47 <xiaoqian> s/?3/Zhiqiang
12:16:50 <tidoust> ... Another reason is that you'll be able to push more scene complexity, because the CPU overhead is smaller.
12:17:07 <tidoust> Zhiqiang: Do we have some concrete benchmarks?
12:17:26 <tidoust> Corentin: We had some early prototype, but we're still working on integrating the new proposal inside browsers.
12:18:01 <tidoust> ?4: Some ongoing effort to port "Aquarium" for benchmark.
12:18:14 <anssik> s/?4/Yang/
12:18:33 <tidoust> Corentin: Intel is working on WebGL2 Compute, but it's unclear when it will be done, and also won't work on Mac.
12:19:04 <tidoust> shangwan: Is non browser GPU also a goal for this?
12:19:09 <tidoust> Corentin: Yes, definitely.
12:19:23 <wseltzer> s/shangwan/sangwhan/
12:20:12 <tidoust> Topic: WHLSL
12:20:29 <tidoust> Myles: Want to talk about a new shading language.
12:20:35 <Kangz> WHLSL
12:20:59 <tidoust> ... Corentin explained that WebGPU needs to work across hardware platforms. All of these platforms have different shading languages.
12:21:32 <tidoust> ... Requirements for WebGPU: language needs to be safe! That's a non starter if the language isn't safe.
12:21:43 <tidoust> ... It obviously also needs to be well-specified.
12:21:59 <tidoust> ... Then translatable across the 3 different platforms.
12:22:15 <tidoust> ... It needs to be performance.
12:22:30 <tidoust> ... Some features need to be evolved together between WebGPU and the shading language.
12:22:54 <tidoust> ... Lastly, since it's new, it would be good if it could be easy to read and write.
12:23:08 <tidoust> ... Really quick tour through the existing languages:
12:23:42 <tidoust> ... Metal Shading language is C++ based. But not safe since you can cast pointer into integers back and forth.
12:24:03 <tidoust> ... HLSL isn't fully defined.
12:24:37 <tidoust> ... GLSL is the language that WebGL shaders uses. GLSL tends out to be a very difficult language to use because of its historical baggage.
12:25:01 <tidoust> ... SPIR-V is an interesting choice but it isn't one that humans write.
12:25:27 <tidoust> ... JS is also intersesting, but not typed-enough, e.g. a+b. It would be very slow.
12:25:58 <tidoust> ... WebAssembly is yet another interesting idea. But not fully adpated for GPUs, including memory layout.
12:26:22 <tidoust> ... Taking all that into account, we came up with a new one, WHLSL.
12:26:49 <tidoust> ... The spec contains rules, which is more than a grammar. It defines things such as how types work, and the execution environment.
12:27:13 <tidoust> ... There is also a compiler, which produces SPIR-V, MSL, HLSL
12:27:30 <tidoust> ... This is not in any browser yet.
12:28:16 <ningxinhu> ningxinhu has joined #webgpu
12:29:07 <tidoust> [showing an example]
12:29:19 <t_homma> t_homma has joined #webgpu
12:29:51 <tidoust> Myles: The example shows the source, the result of the compilation, here for a Mac, and the rendering.
12:30:10 <tidoust> ... I could talk about the details of the language.
12:30:32 <tidoust> anssik: How did you come up with this proposal? Massaging the different proposals?
12:30:53 <tidoust> Myles: Most of the work was done by the Webkit team. We're in the process of inviting other perspectives.
12:30:54 <sangwhan> q+
12:31:44 <tidoust> Corentin: In the group, there are people who believe that the browser should support SPIR-V and others that browsers should support WHLSL directly. Complicated discussions.
12:32:11 <tidoust> ?6: How to WHLSL compare to HLSL in a more secure version?
12:32:20 <tidoust> Myles: That's basically what this is.
12:32:28 <tidoust> s/?6/Bryan
12:32:37 <tidoust> ack sangwhan
12:32:43 <tidoust> [scribe missed question]
12:33:31 <tidoust> Myles: The compiler in the demo outputs intermediary language that gets ingested by my machine directly, not running in the browser for now.
12:34:02 <tidoust> Corentin: The advantage that we have compared to compute functionality is that you can do matrix multiplication more easily. I can go way deeper on this, it's complicated.
12:34:08 <sangwhan> s/[scribe missed question]/sangwhan: is this proposal available for review anywhere? how is the demo implemented?/
12:34:11 <tidoust> Arno: What kind of use cases are you thinking about?
12:34:30 <Huangweichai> Huangweichai has joined #webgpu
12:34:33 <tidoust> ... Are you thinking about machine learning for instance?
12:34:41 <tidoust> Corentin: Yes.
12:35:12 <tidoust> ... It would help with machine learning, fluid computations, effects in rendering applications.
12:35:34 <tidoust> Arno: Where does the compilation happen?
12:35:59 <tidoust> Corentin: The results of the discussion on the computation language does not really change anything.
12:36:07 <tidoust> ... The compilation happens when you load the application.
12:36:12 <tidoust> Myles: Same as in WebGL.
12:36:26 <tidoust> Corentin: Capabilities don't matter for this, because both languages have the same ones.
12:37:06 <tidoust> Arno: If the compilation happens in the browser or if the compilation happens before, it's different. The application could generate shaders on-the-fly in the second case for instance.
12:37:48 <tidoust> Myles: The Web author could go back to the server, or serve the source and a compiler at the same time.
12:38:11 <tidoust> ?7: For WebAssembly, sometimes, the compilation time sometimes makes a difference.
12:38:51 <tidoust> Corentin: The compilation time is not so much of a problem. Register allocation is the most complicated issue in our context for various reasons, takes an insane amount of time.
12:39:00 <tidoust> ... Compilation time is not a differenciation criterion.
12:39:50 <tidoust> ... Every vendor comes with its own set of requirements. Goal is to have it run on all Windows 10 devices, a large number of Windows 7 devices, Android devices.
12:40:15 <tidoust> Myles: Metal and D3D have a small number of configurations that you can change. Vulkan has a zillion of such configurations.
12:40:57 <tidoust> sangwhan: Is the current proposal different from the language that was "shipped" in Safari?
12:41:18 <tidoust> Myles: Yes, different. That was a proof of concept to prove performance gains. Now we have data.
12:41:32 <tidoust> ?7: Some kind of software fallback?
12:41:38 <anssik> s/?7/Ben
12:41:51 <tidoust> Corentin: Yes, plugging things together, you'll be able to do that.
12:41:59 <tidoust> s/?7/Ben/g
12:42:14 <tidoust> ?8: Is is going to replace WebGL2 entirely?
12:42:40 <tidoust> Myles: WebGL is part of the Web Platform. We do want to say that WebGPU can be used in all areas where WebGL can be used.
12:42:59 <tidoust> Guido: [scribe missed question]
12:43:38 <tidoust> Corentin: Overhead target should be minimal. GPU performance is super important for us.
12:44:14 <tidoust> Ben: You mentioned WebAssembly API as a maybe. Anything WASM can do to help with that?
12:44:57 <tidoust> Corentin: We had exchanges on that with the CG. Host bindings is one. That should give us most of what you need.
12:45:39 <tidoust> Ben: Integers and floats are easier to pass in WASM. WebGL passes strings and arrays in a number of places, I believe.
12:45:50 <tidoust> ... Is there a way to make the API easier for pre-host bindings?
12:46:21 <tidoust> Corentin: We're looking at dictionary-based solutions, because it's not the fast path.
12:46:53 <tidoust> ?9: A possible future where WHLSL can be used in WebGL as well?
12:47:03 <tidoust> Myles: That hasn't been discussed.
12:47:22 <tidoust> [Some discussions on shading languages, compilers]
12:47:53 <tidoust> ningxinhu: Any plan to support tooling to port Metal Shading Language and others to WHLSL?
12:48:32 <tidoust> Corentin: First, HLSL is close to WHLSL. Also any shading language compiles to SPIR-V, so that's a common intermediary. But depends on the outcomes of related discussions.
12:48:59 <tidoust> ningxinhu: WHLSL is for humans, not for rendering?
12:49:14 <tidoust> Myles: It's actually for both. That's an explicit goal that I should have listed.
12:49:32 <tidoust> s/rendering/compilation output/
12:49:50 <tidoust> ?10: Possible to decompile SPIR-V into WHLSL?
12:49:55 <tidoust> Corentin: That should be roughly doable.
12:50:15 <tidoust> Myles: Any arbitrary SPIR-V program cannot, though.
12:50:44 <tidoust> Corentin: Some operators in SPIR-V are blocked.
12:51:02 <tidoust> ?8: is the goal to support multiple GPUs?
12:51:21 <tidoust> Corentin: That's explicitly something we thought about, but then it's too complicated and probably too niche on the Web.
12:51:35 <tidoust> ?8: Scientific research would benefit from it.
12:51:43 <tidoust> Myles: That's a great thing to explore in the future.
12:51:55 <tidoust> Corentin: Right. To be honest, I would be surprised.
12:52:08 <tidoust> sangwhan: GPU context switching?
12:52:46 <tidoust> Corentin: The idea is that there is no GPU switching. But you can choose low- or high-performance.
12:52:53 <tidoust> sangwhan: And that gives you the GPU model?
12:53:04 <tidoust> Myles: Unclear. We don't want to expose information unless needed.
12:53:30 <tidoust> Bryan: Will WebGPU WHLSL have extensions?
12:53:35 <tidoust> Myles: Unclear.
12:54:09 <tidoust> Corentin: We are thinking of an extension mechanism. For instance, you could imagine Raytracing extensions. But the plan is to remain conservative.
12:54:32 <tidoust> ?10: Is multiview in case?
12:54:52 <tidoust> Corentin: We haven't thought about multiview, but that's a bandage, to be honest.
12:55:06 <tidoust> ... Our current approach is to try to do a basic set of features that are really fast.
12:55:40 <tidoust> Myles: The purpose of multiviews is to increase performance. If that give much more performance, then that's a valuable use case.
12:56:24 <tidoust> [WHLSL is targeting feature levels on all existing platforms, so any card should be able to run it]
12:57:03 <tidoust> Myles: Google has started implementing this. Apple has started implementing this as well according to activity in public repo.
12:57:22 <tidoust> ... Mozilla is prototyping things.
12:58:24 <tidoust> wseltzer: Wendy from W3C. Just wanted to say hello. Good progress there. When you mentioned transition to a WG to get IP stamps, that's valuable, especially in the GPU space.
12:58:24 <sangchul> sangchul has joined #webgpu
12:58:55 <tidoust> ... We're happy to help with that transition. Of course, that requires approval by the W3C Advisory Committee. So when you think you're ready, please get in touch with us.
12:59:10 <tidoust> Corentin: Awesome. I believe Dean Jackson is looking into it as well.
12:59:17 <tidoust> RRSAgent, draft minutes v2
12:59:17 <RRSAgent> I have made the request to generate https://www.w3.org/2018/10/24-webgpu-minutes.html tidoust
13:09:50 <tidoust> tidoust has joined #webgpu
13:12:24 <xiaoqian> xiaoqian has joined #webgpu
13:21:13 <onishi> onishi has left #webgpu
13:27:30 <t_homma> t_homma has joined #webgpu
13:41:55 <t_homma> t_homma has left #webgpu
14:33:31 <Zakim> Zakim has left #webgpu
14:42:29 <xiaoqian> xiaoqian has joined #webgpu
15:54:39 <xiaoqian> xiaoqian has joined #webgpu