brrrrrm

30 ms·

391.
by brrrrrm 4y ago
This is really cool. Is the SLAM engine open source?
392.
by brrrrrm 4y ago
It’s particularly cool that this works on mobile
393.
Ask HN: Can you share websites that are pushing the utility of browsers forward?
188 points by brrrrrm 4y ago|128 comments
394.
by brrrrrm 4y ago
This might seem like a weird suggestion, but you could probably use WebAssembly. The spec is incredibly simple and portable. It'd allow you to use V8, cranelift, wasmer, wasm-micro-runtime and a whole host of other JITs/runtimes.
395.
by brrrrrm 4y ago
That might explain 10-20% utilization, but 0.4% utilization (four tenths of a percent!) should be ringing all sorts of alarm bells. It’s not even that small of a model (100x bigger than mobilenet v3)
396.
by brrrrrm 4y ago
Something seems a bit off about the GPU usage. The RTX 3090 is supposed to have ~300TFlops. If the 26.8GFlop YOLOX-s is running at 40fps, that means you’re getting 0.4% utilization - about 250x slower than expected peak…
397.
by brrrrrm 4y ago
I wrote a video editor in vanilla JS https://github.com/bwasti/mebm
398.
by brrrrrm 4y ago
This is an incredible API I wish existed in the browser (maybe over WebRTC?) addEventListener("fetch", (event) => { event.respondWith(new Response("Hello world")); });
399.
by brrrrrm 4y ago
What's the generated assembly look like? I suspect clang isn't smart enough to store things into registers. The latency of VPCOMPRESSB seems quite high (according to the table here at least https://uops.info/tabl
400.
by brrrrrm 4y ago
The complication involved with UTF-8 encoded space removal is immense and likely quite far out of scope.
401.
by brrrrrm 4y ago
hey I wrote that second one! It's incredibly flattering to see mebm recommended despite its relatively immature state. I've been considering resuming work on it, and this comment certainly provides a ton of motivation. thanks!
402.
List of Heuristic-Heavy Spots in Computer Science(twitter.com)
1 points by brrrrrm 4y ago|0 comments
403.
by brrrrrm 4y ago
Do you have a link to that thread? I’ve never quite understood the value of GC over RC save for vague claims of developer efficiency.
404.
by brrrrrm 5y ago
> ran into the restrictions on collection visibility Is there any more information about this? It seems really subtle!
405.
by brrrrrm 5y ago
The “WebAssembly object” doesn’t exist as you describe. You probably have an object that wraps WebAssembly function calls and primitive WebAssembly data (individual numbers). WebAssembly functions can only return numbers and can only take n
406.
JavaScript Garbage Collection with WebAssembly Is Possible Today(jott.live)
2 points by brrrrrm 5y ago|0 comments
407.
Etymology of the Word Yakuza(en.wikipedia.org)
2 points by brrrrrm 5y ago|0 comments
408.
by brrrrrm 5y ago
Thanks for the updated script! I added a comment to my code pointing to yours. I’m curious what the key slowdown in my attempt was.
409.
by brrrrrm 5y ago
here's an impl: https://jott.live/code/train_matrix_exp_estim_jax.py it works but surprisingly slow on my devices
410.
by brrrrrm 5y ago
I guess JAX would be a more fair comparison? It has a JIT compiler
411.
Browser cookies can silently cause tax payment fees to increase(twitter.com)
2 points by brrrrrm 5y ago|0 comments
412.
Wasmblr – C++ WebAssembly Assembler in a single header file(github.com)
11 points by brrrrrm 5y ago|0 comments
413.
by brrrrrm 5y ago
Unfortunately, you cannot. https://stackoverflow.com/questions/42806037/modify-canvas-f...
414.
by brrrrrm 5y ago
Thanks for the heads up! In that case, I'd prefer not to share this link with peers. Do you have an alternative source with similar high-level content to share?
415.
by brrrrrm 5y ago
> even during inference, the size of the parameters is completely dwarfed by the intermediate tensor representations What makes you say this?
416.
by brrrrrm 5y ago
A good visual introduction to neural networks can be found here: https://playground.tensorflow.org A parameter is a "weight" in this case (the lines drawn from neuron to neuron). The neurons are effectively runtime va
417.
by brrrrrm 5y ago
this is awesome, thanks for the link!
418.
by brrrrrm 5y ago
dang, still no SIMD for WebAssembly. Chrome and Firefox are meanwhile hitting 45gflops (85% of peak) on my M1 Air. Safari only gets to 12.5gflops
419.
by brrrrrm 5y ago
I’m not up to date on the latest mining algorithms but is it true that they’re all embarrassingly parallel to the point the WebGPU makes no difference? Fast inter-thread communication is the big addition here.
420.
by brrrrrm 5y ago
If you're using Chrome Canary with the "Unsafe WebGPU" flag turned on, yes. And it'll be a lot faster/more power hungry than before. Otherwise, nothing beyond the normal CPU based miners have been enabled. Adaptati

More ›