pornel

27 ms·

511.
by pornel 2y ago
Less controversially, when you write C, you write for a virtual machine described by the C spec, not your actual hardware. Your C optimizer is emulating that VM when performing symbolic execution, and the compiler backend is cross-compiling
512.
by pornel 2y ago
This style is coming back — Rivian R3X looks eerily similar, just with more even proportions.
513.
by pornel 2y ago
Channels are only problematic if they're the only tool you have in your toolbox, and you end up using them where they don't belong. BTW, you can create a deadlock equivalent with channels if you write "wait for A, reply with
514.
by pornel 2y ago
The point of the design wasn't to be pretty, but to be different, edgy, and look rugged. But they made a mistake by announcing the design long before they were able to produce it. By the time they had built it, the hype cycle was over,
515.
by pornel 2y ago
Lifetime tracking isn't just for safety or resource management. It also helps write correct code, especially parallel one where shared vs mutable matters. Unit testing of shaders is usually a pain. Traditionally they're a black bo
516.
by pornel 2y ago
https://www.scratchapixel.com https://www.youtube.com/c/acerola_t https://www.shadertoy.com https://google.github.io/tour-of-wgsl/
517.
by pornel 2y ago
But this also means that almost all of the wear is from trucks. This whole meme comes from junk science ( https://www.emissionsanalytics.com/news/pollution-tyre-wear-... ) > we found that the car emitted 5.8 grams per
518.
by pornel 2y ago
Heavy particles and gaseous emissions are not comparable in such a simplistic way. If you take a dump on the street it doesn't mean you caused 50 million times more emissions than the EPA limits for ICE car exhaust. For example, iron f
519.
by pornel 2y ago
LLMs on GPUs have a lot of computational inefficiencies and untapped parallelism. GPUs have been designed for more diverse workloads with much smaller working sets. LLM inference is ridiculously DRAM-bound. We currently have 10×-200× too
520.
by pornel 2y ago
AI will be cheap to run. The hardware for AI is getting cheaper and more efficient, and the models are getting less wasteful too. Just a few years ago GPT-3.5 used to be a secret sauce running on the most expensive GPU racks, and now models
521.
by pornel 2y ago
It could fail if the generated C code triggered Undefined Behavior. For example, signed overflow is UB in C, but defined in Rust. Generated code can't simply use the + operator. C has type-based alias analysis that makes some type cast
522.
by pornel 2y ago
The interop is already great via PyO3, except when people want to build the Rust part from source, but are grumpy about having to install the Rust compiler. This hack is a Rust compiler back-end. Backends get platform-specific instructions
523.
by pornel 2y ago
USSR once fucked up its agriculture and purged biological research, because Stalin thought survival of the fittest didn't fit his political narrative, and banned genetics for being an imperial ideology: https://en.wikipedia.
524.
by pornel 2y ago
Internally the compiler tracks what "niches" exist in types, like minimum and maximum valid values, and takes the unused values for enum tags. One thing it can't use is padding in structs, because references to individual fie
525.
by pornel 2y ago
> We were pretty "soft" (dislike the term) on the land borders and on immigration That's absolutely not the impression US gives to the people outside. The visa system is soft only on one specific demographic it deems worth
526.
by pornel 2y ago
Markdown is semantically less expressive than man's markup. man goes into details like explicitly marking up command's arguments. You can convert man to Markdown easily using full capabilities of Markdown, but Markdown lacks seman
527.
by pornel 2y ago
Why would they, if they're ideologically opposed to what the social security symbolizes? Breaking the system with false positives is in their interest. It allows them to demonstrate it's broken, and gives them an excuse to dismant
528.
by pornel 2y ago
From European perspective, Russia's latest invasion was a wake-up call that the peace isn't permanent, and diplomacy and economic power may not be sufficient to keep it. Effectiveness of drones and advanced missiles in the war was
529.
by pornel 2y ago
They link to actual issues in their bug tracker, so if it was a joke, it was an impressive long con.
530.
by pornel 2y ago
Rust can be optionally compiled in a panic=abort mode, but by default panics are recoverable. From implementation perspective Rust panics are almost identical to C++ exceptions. For servers that must not suddenly die, it's wise to use
531.
by pornel 2y ago
Even if you rewrite C code to the same "shape" as Rust, it won't become equally easy to statically analyze. The C language doesn't give the same guarantees, so it doesn't benefit from the same restrictions. For exam
532.
by pornel 2y ago
> Could such bugs be avoided in C using the right tools and strategies "right tools and strategies" is very open-ended, almost tautological — if you didn't catch the bug, then obviously you haven't used the right tool
533.
by pornel 2y ago
I don't think it's a common concern in Rust. It used to be a problem in Internet Explorer. It's a footgun in Swift, but Rust's exclusive ownership and immutability make cycles very difficult to create by accident. If yo
534.
by pornel 2y ago
Weights in neural networks don't always need to be precise. Not all weights are equally useful to the network. There seems to be a lot of redundancy that can be replaced with approximations. This technique seems a bit similar to lossy
535.
by pornel 2y ago
Keep in mind that the "experts" are selected per layer, so it's not even a single expert selection you can correlate with a token, but an interplay of abstract features across many experts at many layers.
536.
by pornel 2y ago
In image compression "lossless" is a term of the art. What you're doing is a practically useful quality degradation, but it's not lossless .
537.
by pornel 2y ago
No, it's more like sharding of parameters. There's no understandable distinction between the experts.
538.
by pornel 2y ago
This is where Rust makes a massive difference. I can change iter() to par_iter() and have it work flawlessly on the first try.
539.
by pornel 2y ago
The warp model in GPUs is great at hiding the DRAM latency. The GPU isn't idly waiting for DRAM. All threads that need a memory access are in a hardware queue, and data coming from the DRAM immediately dequeues a thread and runs the wo
540.
by pornel 2y ago
Facebook and Google+ tried to do this with their realname policies. It doesn't work as well as one would expect: • Toxic assholes are not deterred by their name being attached to what they're saying, because they think they'r

More ›