May be not exactly what you're looking for since there's 3D rendering involved but:
Do some fairly heavy CPU stuff with timers around them, but they're all using WebGL/OpenGL. The web versions have asm.js and wasm builds, and the same code also builds and runs natively on Windows, OSX or Linux, iOS or Android with GL, D3D11 or Metal as rendering backend. The most CPU-heavy demos are these: http://floooh.github.io/oryol-samples/wasm/BulletPhysicsBasic.html http://floooh.github.io/oryol-samples/wasm/BulletPhysicsCloth.html http://floooh.github.io/oryol-samples/wasm/Dragons.html The biggest / most surprising difference is the Dragons demo with 1k dragons: On my mid-2014 MBP the animation system takes: - natively: 4ms - Chrome/wasm: 12..14ms - Chrome/asm.js 13..15ms - FF Nightly/wasm: 7..10ms (quite heavy fluctuations) - FF Nightly/asm.js: 10.12ms This is doing scalar floating point ops on 32-bit floats, about 100 multiply-adds per bone (so at 64k bones about 6.5 million fp32 multiply-adds). Usually I'm not seeing such a big difference (3x..4x vs native), FF with wasm is doing pretty well though. All the heavy stuff happens in fairly small 'inner loops' with (hopefully) cache-friendly data layout, but I haven't analysed the native code too deeply yet (for cache misses etc...). Most real-world code I've seen so far seems to be around 1.2 to 2x slower than the native counterpart, so this gap is a bit surprising. Cheers, -Floh. Am Donnerstag, 6. Juli 2017 06:26:27 UTC+2 schrieb [email protected]: > > Are there any open source projects(that don't use webgl/opengl) that I can > build myself using clang as well as emscripten? I am trying to find out > native/wasm performance numbers. > -- You received this message because you are subscribed to the Google Groups "emscripten-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
