[darknoon]

It would be a more fair comparison to https://deeplearnjs.org/, since CPU training is not standard practice for neural nets these days.

[nsthorat]

We've done some initial tests ourselves. WASM doesn't yet support SIMD so WebGL tends to be 5-10x faster. SIMD is actively being worked on by many smart people in Chromium / other browsers, so I would expect to see huge wins in the near term future. When that happens, deeplearn.js will have a WASM backend. WASM has a much better memory management story (destructors on the C++ side) so I'm super excited about its future.

[seanmcdirmid]

Is SIMD going to close that performance gap for a problem that otherwise fits well on a GPU? Wouldn't the ideal by WASM support for GPU access?

[fulafel]

Indeed WASM supports GPU access through the normal JS WebGL API. Many asm.js/WebAssembly apps use the GPU in this way.

Though compiling WebAssembly to GPU code would be a very interesting thing to explore, too.

[nsthorat]

This probably won't buy you anything. The API you still have is WebGL, and JS is not the bottleneck.

[make3]

GPUs are at least an order of magnitude faster at training neural nets than CPUs. This is why companies like Google very large amounts of multi-thousand dollar GPUs like the Tesla V100. If cpu WASM is ever faster than WebGL, it's just that the webGL implementation is suboptimal (or that WebGL has too much overhead for real GPU compute)