azakai

17 ms·

211.
by azakai 4y ago
We actually have good data on this. Very little efficiency is lost in many cases - only around 14%: https://kripken.github.io/blog/wasm/2020/07/27/wasmboxc.html See also here, where wasm2c is the fa
212.
by azakai 4y ago
Ah, thanks! I should finish reading the entire article before commenting, sorry...
213.
by azakai 4y ago
IIUC steps 5-7 in the exploit cause around 2^32 oopses. I don't know much about the Linux kernel - could it perhaps have a limit on the number of oopses before it halts the entire system? The article explains why it is important to not
214.
by azakai 4y ago
SerenityOS's LadyBird browser is an impressive project, but it doesn't prove that you can rewrite something like Chromium in a few years. It does prove you can get something working. But it will take many more years to actually ge
215.
by azakai 4y ago
Very well written, and I like that it ends on this note: > Looking at my hypotheses, I was wrong on all counts The author expected some things, and had good reasons to, then did a lot of serious work, and ended up finding out somewhat di
216.
by azakai 4y ago
The main thread is a little special on the Web since it can't block, which can cause issues (like pthread_create doesn't immediately create an available pthread). There are workarounds for most of those issues (like pre-allocating
217.
by azakai 4y ago
In Emscripten that uses a pthread implementation layer built on top of Web Workers + a shared wasm Memory. Basically memory is shared, and you have atomic instructions, and each thread of execution gets its own Web Worker. That has some lim
218.
by azakai 4y ago
The JavaScript heaps are distinct, that's true, but there is a single shared wasm heap which is used from multiple threads. That is enough to implement the pthreads API. Applications like Photoshop and Google Earth use ptheads on the W
219.
by azakai 4y ago
I tried a few nontrivial examples and it gets them completely wrong - either missing the point entirely, or having the exact opposite of the intention. One issue is this appears to only process the lyrics of the song. But the actual audio m
220.
by azakai 4y ago
1/10th as popular as one of the by-far most popular programming languages of all time is still a huge success story.
221.
by azakai 4y ago
In general I agree with you, yes, market conditions change, and CEOs need to make sometimes tough choices in response. However, the specific context here is this quote from the CEO: > we were not as rigorous as we should have been in man
222.
by azakai 4y ago
One possibile way to take personal responsibility could be to add "and therefore I will reduce my compensation by the same percentage we are reducing headcount". (Based on other comments, that would be 6%.)
223.
by azakai 4y ago
There used to be 2GB limitations both on JS typed arrays and on the size of wasm memory, so this might refer to either or both of those. But browsers have been fixing that, here is Chrome doing so back in 2020: https://v8.dev&#x2
224.
by azakai 4y ago
The design of WASI and the component model are layered, though: you could start with V8 and build a runtime with full WASI etc. support. In fact, Node.js and Deno have both done exactly that, a V8 core with layered APIs on top including WAS
225.
by azakai 4y ago
Some (most?) do, yes, here is a V8 article on that for example: https://v8.dev/blog/wasm-code-caching
226.
by azakai 4y ago
Optimizing VMs often use lots of memory for various reasons. Specific costs in modern optimizing VMs include the memory for the JITed code, and with multiple tiers you end up JITing each function more than once. You also need memory to save
227.
by azakai 4y ago
Another thing that would be interesting to measure here is memory usage. QuickJS might be slower (as a minimal interpreter) but at least it should be much smaller than a full optimizing JS engine (running normally or compiled to wasm), both
228.
by azakai 4y ago
The 2GB limit was a temporary browser limitation. Modern browers mostly support 4GB now. And wasm64 is testable today, which supports 64-bit pointers and more than 4GB.
229.
by azakai 4y ago
Why are you certain such studies have not been done? Plenty of things have been published in that space. For example, https://academic.oup.com/jid/article/226/2/199/6582941 https://www.fr
230.
by azakai 4y ago
It's not unspecified "doctors and researchers". It's the overwheming consensus among medical professionals. The CDC recommends masks, as I linked. So does the WHO, https://www.who.int/news-room/quest
231.
by azakai 4y ago
Surgical masks and face shields are indeed designed specifically to protect from fluids. That is very important too. Separately from that, when there is a risk of airborne transmission as there is with Covid, doctors and nurses will wear ap
232.
by azakai 4y ago
> We simply don't know if n95 masks provide any protection -- personal or otherwise -- in the real world. That is simply not true. First, doctors and researchers consistently say that N95s are known to provide protection against Cov
233.
by azakai 4y ago
Yes, in fact, almost every successful language of the last decade has had such support: Mozilla funded a Rust team for years, Google funded a Go team, Apple a Swift team, Microsoft for TypeScript, etc. etc. There are examples of rising lang
234.
by azakai 4y ago
And to add to that, it also makes other people happy too, to contribute or just to follow. Really cool stuff, and meaningful.
235.
by azakai 4y ago
It is exactly that - an art project. Maybe it will also be useful some day, maybe not, but to me at least that's not the point. It's really cool that we can fund some art projects like this in the software industry! Not a perfect
236.
by azakai 4y ago
There has been a large amount of progress on wasm GC actually! If you haven't been attending the group meetings though then you might have missed it. In particular: * The GC MVP spec is mostly done at this point. * There are working to
237.
by azakai 4y ago
Julia is probably the most likely competitor for Python in the data science space. It's designed for such use cases and it's very good at it. Still, Python does have so much momentum that it's almost certainly going to contin
238.
by azakai 4y ago
There is often a speedup, but maybe not by orders of magnitude. I've usually seen 2x-3x or so.
239.
by azakai 4y ago
Interesting. That might be very specific to Rust, though. In that case, LLVM might not be the optimal backend for rustc. But this isn't an issue for most LLVM languages, which either assume parallelism in the build system, or they get
240.
by azakai 4y ago
In such cases one can sandbox LLVM, either in another process or using wasm etc., and that has been common for many years. It's practical to do so because it's a pure computation with a clear interface (compile source code to mach

More ›