rfoo

23 ms·

271.
by rfoo 2y ago
No. Like the siblings said, say, you have a program which spends 10% time in Python code between these numpy calls. The code is still not scalable, because you can run at most 10 such threads in a Python process before you hit the hard li
272.
by rfoo 2y ago
These "accelerated sub-languages" are still driven by, well, Python glue. That's why we need free-threading and faster Python. We want the glue to be faster because it's currently the most accessible glue to the communit
273.
by rfoo 2y ago
> Python is never really going to be 'fast' no matter what is done to it because its semantics make most important optimizations impossible Scientific computing community have a bunch of code calling numpy or whatever stuff. Th
274.
by rfoo 2y ago
There definitely is a Chinese ethnicity. What you are talking about sounds like claiming that the US is not a country built by immigrants because there are Indians. I can understand why you insist on such ideas if you are, well, the so-call
275.
by rfoo 2y ago
CUDA is not merely software control. Let's face it: it's tied to their architecture and evolves with their architecture. Ignore whatever patent and software copyright you are talking about, they are simply the best at implementing
276.
by rfoo 2y ago
In a worse world, worse is better.
277.
by rfoo 2y ago
I've always been running a headless Linux VM wherever I do any dev work, Windows or Mac. Works very well. Glad to see that WSL2 and OrbStack promotes this workflow. For now my only complaint is I prefer full-blown VMs instead of WSL-li
278.
by rfoo 2y ago
I'd add that a lot of the described advantages come from culture. For web applications manual memory management is 100% a friction instead of a relief. But the culture in Rust community in general, at least for the past ten years or so
279.
by rfoo 2y ago
That's an interesting idea. The downside (I could imagine) would be we have to convince people to implement THE RIGHT MODEL in every build tools in the world, but if that's possible then indeed it would work a lot better.
280.
by rfoo 2y ago
> Kill the language and replace it with configuration No. Hard no. Please don't tell me you are going to kill a fine (even though awkward) configuration language and replace it with (potentially text-templated) yaml.
281.
by rfoo 2y ago
That's the first thing came to my mind too. While a little bit too much of a guess, it's quite possible that whatever three letter agency finally had a high-confidence note on who was behind the XZ backdoor and decided to issue an
282.
by rfoo 2y ago
Otherwise how could we join the <x>.cpp fancy gang? We'd have to name the project "agent.js" which is super boring! /s
283.
by rfoo 2y ago
China has been doing this for a while already. Not only porting a country's software on Linux, they are also porting to their own platforms. [1] [2] [3] [4] Currently it's full of rough edges and barely (if at all) works. But I do
284.
by rfoo 2y ago
Time to... (check notes)... break up Google and somehow sanction Google so they themselves can't use Android anymore and you get Fuchsia on the Pixel line. That's how you push alternative OS thesedays.
285.
by rfoo 2y ago
Honestly, I think Fuchsia was fine, there's just no reason to deploy it yet.
286.
by rfoo 2y ago
It's quite sad that application interoperability requires parsing text passed via pipes instead of exchanging structured information. Like others said, worse is better.
287.
by rfoo 2y ago
Agree on this being stupid. > my bet is rather on awful security and infrastructure practices For sure. As far as I know ByteDance does not have an established culture of always building secure systems. You don't need to be a master
288.
by rfoo 2y ago
> but what on earth would be his motive Rumors said that his motivation would be to just actively sabotage colleague's work because managers decided to give priority on GPU resources to those who were working on DiT models, and he w
289.
by rfoo 2y ago
> such as by using a memory safe language or hardware capabilities that prevent memory safety vulnerabilities Heh, now I do believe that my partially-memory-corruption-based side job may be in danger. If it was just "by using a memo
290.
by rfoo 2y ago
It may be done on purpose? To me this looks like a satirical piece in response to some "nooooo how dare you introduce lifetime annotations to C++"-style comments on the same author (Sean Baxter)'s C++ lifetime annotation prop
291.
by rfoo 2y ago
Pretty sure Google is going to allow anyone to "sudo". ... in the Linux virtual machine.
292.
by rfoo 2y ago
> nobody would be using anything but ZFS if it was in tree I'd pick xfs anytime. I don't know if it was already fixed, but ~1y ago had a really miserable experience using ZFS on my NVMe drives, zfs is like 100x slower in random
293.
by rfoo 2y ago
> pricing is astronomical. you also have to "contact sales" to get access to anything outside of a free trial You don't have to contact sales if you are a Chinese-speaking customer. And pricing is fine. ByteDance has a dif
294.
by rfoo 2y ago
The funny part of your example is, decoding AArch64 instructions (argubly, "parsing") is also easier than decoding x86_64. Yet GP's point still holds.
295.
by rfoo 2y ago
> Apple historically used to have a deservedly good reputation for this. Are they? Apple only started their bug bounty program (with monetary rewards) merely 5 years ago, 12 years after first iOS release and well after everyone else. The
296.
by rfoo 2y ago
Impressive safety metrics! I wish OAI include "% Rejections on perfectly safe prompts" in this table, too.
297.
by rfoo 2y ago
I usually just point the target branch of MR 2 to MR 1. After merging MR 1 GitLab automatically change it to the default branch so it's more or less okay. However this makes updating these MRs very rebase heavy and as said in OP it is
298.
by rfoo 2y ago
If you are stuck with git, try git-branchless. At least it makes rebasing your patch stack easier (git sync && git submit).
299.
by rfoo 2y ago
You can try to be Rust and refuse such access rules. Yes, I know it sometimes makes sense. Rust rejects perfectly fine codes too. People are really into such things.
300.
by rfoo 2y ago
For those who are convinced and want a slightly more convenient tool for doing this: try yadm.

More ›