Many moons ago I was preparing for an interview to join wasmer, back then they showed me that by running wasmer as a kernel extension in Linux they have made it run four orders of magnitude (10,000 times) faster. Given the WASI, I keep thinking more than the plugin usability of things like wasmer, what is the point of WASM not being accepted as a second binary in all operating systems? Windows could support .exe as well as .wasm for example and we would finally have a cross-plaform binary. Apple would never accept; Microsoft will do, and Linux can simply have,
Such speedup is not plausible at all. It had to be a flawed benchmark relying on some API being very slow in userspace (probably on purpose, not using it the right way). The raw execution speed of WASM can't change much just because it is in a kernel.
Judging by the slow speed of node here, I wonder if it was run without forcing optimizations. It doesn't do OSR (on-stack replacement), meaning that a benchmark script doesn't get a chance to tier up from the baseline compiler to the fully optimized tier.
Running with
node --no-liftoff
avoids this problem.
(This is not an issue on the Web, where code must return to the event loop anyhow, allowing tiering up to work, and node is optimized for that kind of workflow.)
It's genuinely great to see WASM runtimes are getting closer to native code while being sandboxed, but these benchmarks assume you're not bundling a whole Runtime + GC + stdlib inside the sandbox.
One fairly revealing microbenchmark for WASM runtimes is `int8` dot products & angular/cosine distances.
(My) NumKong [1] has implementations targeting both vanilla AVX2/Haswell and AVX2-VNNI/Alder Lake, which makes it easy to see where runtimes and code generators leave performance on the table.
I started a few Wasmtime/Cranelift PRs around this, but didn’t manage to finish them :facepalm: Might be a fun weekend project for someone interested in backend/codegen work.
It'd be nice to see memory use. Ideally base usage, then cost of running 1 of a thing, then cost of running 50 if a thing.
Personally low my 50% speed vs native does not sound like the worst to me. I've spent years using perl, java (in v4-v7 days), node... These have various ways they are less good than native, and we collectively seem to accept that. Ideally yes wasm gets faster but a 2x is pretty ok!
My only issue with WASM is that it forces me to contort my platform abstraction later to terrible web APIs. Other than that it's great. I can compile C++ with plain clang or the WASI SDK, and it runs everywhere at near native speeds with minimal boilerplate.
In my opinion, WebAssembly just needs to be better than what it replaced (asm.js) to be called a success story. And it definitely is better than asm.js.
Unlike HTML or CSS or JavaScript which are by definition large enabling technologies that enable completely new things, WASM is always that small step to make C/C++ and other native languages usable on the web. I never thought of WASM as a magic wand that makes web apps super fast and you probably shouldn’t.
>In my opinion, WebAssembly just needs to be better than what it replaced (asm.js) to be called a success story.
That would be only true if WASM was meant as a transitionary technology. But it has been with us for more than a decade, and it's only so-so. And if you think this, what do you think should replace it?
Is speeding up apps an aim of WebAssembly? I’m not being sarcastic… most of the slowdown you see in webapps is due to the DOM and/or network latency and WebAssembly offers nothing new there. The JavaScript language is very rarely the reason why an app is slow.
I’ve used WebAssembly to great success creating a unified codebase that runs on Node, an iOS app and on the web. It’s a niche but in that area it opens doors that were previously impossible to open.
Obviously it is, I've been writing a scientific library that runs in-browser and WebAssembly has been a blessing, it's easily 10x faster than the backend Python code it replaced, and 2-5x faster than I could achieved with JavaScript.
I agree. It is mostly a solution looking for a problem and has close to the same functional downsides as Flash or Applets other than "transparency".
Its 'sandbox' is also escapeable at both the engine level in V8, Gecko and JavaScriptCore (All in C/C++) and at the implementation level; especially if the software was converted from C to WASM.
So it is not the silver bullet people think it is.
reply