iswasmfast VS simd

At a glance, the bindings for wasm copy the data,

https://github.com/zandaqo/iswasmfast/blob/54bbb7b539c127185...

If the running code is short enough then that copy might easily make the wasm version much slower. That is indeed a known downside of wasm (calls to JS are somewhat slow, and copying of data even more so - wasm shines when you can avoid those things).

If it's not that, then a 10x difference suggests you are running into some kind of a VM bug or limitation.

WASM has its moments, as you can see in this[1] benchmark it outperforms JS and native addons on certain tasks.

Since the bottleneck with native addons is usually data copying/marshalling, and we have direct access to WebAssembly memory from the JavaScript side, using WebAssembly on this "shared" memory might become the best approach for computationally heavy tasks. I wrote about it a bit here[2].

[1] https://github.com/zandaqo/iswasmfast

A few years ago I did similar comparison but in context of Node.js and sans manual optimizations: https://github.com/zandaqo/iswasmfast

In my work, I have come to conclusion that it seldom pays off to go "native" when working with Node.js. More often than not, rewriting some computationally heavy code in C and sticking it as a native module yielded marginally better results when compared with properly optimized js code. Though, that doesn't negate other advantages of using said technologies: predictable performance from the start and re-using existing code base.

I'd love to see OpenCL or WasmSIMD support for RVV 071.

Wasm SIMD supports only 128-bit registers, no? https://github.com/WebAssembly/simd/blob/a78b98a6899c9e91a13095e560767af6e99d98fd/proposals/simd/SIMD.md

According to the WebAssembly Roadmap, the 128-bit packed SIMD Extension proposal has been accepted and is already implemented in every major runtime except Safari, and the Relaxed SIMD proposal is planned, with Firefox already having an experimental implementation in nightly-channel builds.

Thanks! You're probably right about spatial filtering with a low sphere count. SIMD is supported, that looks to work in every recent browser except Safari. Looking forward to trying that out. A really quick search of wasm-bindgen docs don't reveal an interface to use that SIMD support though. I wonder if you can unsafely inline some raw wasm like you can with asm!