HN user
AI accelerators used in DC are not really "graphic cards" any more, you ain't running gaming on it
I think the lighter 40 series cards like L40 still have OK graphics features. But otherwise yeah, after the Ampere generation graphics features went down the drain. The A100 and A40 cards can do graphics well but it already makes no sense in terms of power-to-performance ratio.
This is a feature I've been waiting for for a long time - finally, something that makes File Explorer _slower_.
Except literally everyone who needs a functioning queue system.
I'm not going to say Slurm is great (its API:s are awfully inconsistent and there's a lot of code churn between versions leading to subtle behaviour changes in prod), but it's an invaluable and reliable tool. As someone who manages Slurm clusters in academic HPC as a major part of my job, I'm not at all happy to see this and quite worried that the development and maintenance of Slurm will be broken by the inevitable market volatility.
There's a ton but it's pretty scattered. Yurii Nesterov's a big name, for example.
Yeah, I did a lot of traditional optimization problems during my Ph. D., this type of expression pops up all the time with higher-order gradient-based methods. You rescale or otherwise adjust the gradient based on some system-characteristic eigenvalues to promote convergence without overshooting too much.
No one is locked into Offive because of file formats
A lot of people are locked in because those import/export features are typically imperfect (or perhaps the documents themselves are) and will badly and often "invisibly" (to the non-Office user) break something.
Looking at Latex, I don't think hand tuning some parameters until you get right look in every single case is much better user experience...
Having written many papers, reports and my entire Ph. D. thesis in Latex, and also moved between LaTeX classes/templates when changing journals... I'm inclined to agree to an extent. I think every layout system has a final hand-tweaking component (like inline HTML in markdown for example), but LaTeX has a very steep learning curve once you go beyond the basic series of plots and paragraphs. There are so many tricks and hacks for padding and shifting and adjusting your layout, and some of them are "right" and others are "wrong" for really quite esoteric reasons (like which abstraction layer they work at, or some priority logic).
Of course in the end it's extremely powerful and still my favourite markup language when I need something more powerful than markdown (although reStructuredText is not so bad either). But it's really for professionals with the time to learn a layout system.
Then again there are other advantages to writing out the layout, when it comes to archiving and accessibility, due to the structured information contained in the markup beyond what is rendered. arXiv makes a point about this and forces you to submit LaTeX without rendering the PDF, so that they can really preserve it.
Plus with any kind of effort put into a standard browser setting you could easily have some granularity, like: accept anonymous ephemeral data collected to improve website, but not stuff shared with third parties, or anything collected for the purpose of tailoring content or recommendations for you.
Because they track usage stats for site development purposes, and there was no convergence on an agreed upon standard interface for browsers since nobody would respect it. Their banners are at least simple yes/no ones without dark patterns.
But yes, perhaps they should have worked with e.g. Mozilla to develop some kind of standard browser interface for this.
Yeah, but the only reason for this time wasteage is because website operators refuse to accept what would become the fallback default of "minimal", for which they would not need to seek explicit consent. It's a kind of arbitrage, like those scammy website that send you into redirect loops with enticing headlines.
The law is written to encourage such defaults if anything, it just wasn't profitable enough I guess.
I don't think the cookie law is that impractical? It's easy to comply with by just not storing non-essential user information. It would have been completely nondisruptive if platforms agreed to respect users' defaults via browser settings, and then converged on a common config interface.
It was made impractical by ad platforms and others who decided to use dark patterns, FUD and malicious compliance to deceive users into agreeing to be tracked.
The kind of CUDA you or I would write is not very hardware specific (a few constants here and there) but the kind of CUDA behind cuBLAS with a million magic flags, inline PTX ("GPU assembly") and exploitation of driver/firmware hacks is. It's like the difference between numerics code in C and and numerics code in C with tons of in-line assembly code for each one of a number of specific processors.
You can see similar things if you buy datacenter-grade CPUs from AMD or Intel and compare their per-model optimized BLAS builds and compilers to using OpenBLAS or swapping them around. The difference is not world ending but you can see maybe 50% in some cases.
funny how efficient markets work.
Can one really speak of efficient markets when there are multiple near molopolies at various steps in the production chain with massive integration, and infinity amounts of state spending in the process?
The problem with those are the iffy headless support. For Gnome it kind of works as long as you don't run into keyring issues (you will, the credentials management for gnome-remote-desktop is terrible compared to eg xrdp), don't need more than 1 desktop per user (you might need to support this for eg HPC, though there are workarounds), and you don't try to do hardware acceleration, headless, Wayland and Xwayland at the same time (which is a Mutter capabilities bug I have raised with them).
For KDE headless Wayland just doesn't work. Unfortunately they are only considering a narrow range of usecases, this has kind of always been the case since the whole "1 graphical session per user" thing, which doesn't map to a lot of use cases (though you can work around it it's very unintuitive).
I hope the XFCE devs can come up with something and that we can get good RDP support for that at least.
This is literally the best thing about Unix tools. You can write a compact little Awk script and pipe that with a few coreutils to parse mountains of text into sensible data. I use this kind of thing every day.
I spent a substantial part of my Ph.D. dealing with this stuff, existing frameworks and algorithms were not suitable for our exact use-case and implementation. I had more bugs/artefacts than I'd like to admit due to some off-by-less-than-0.5 in some particular direction in some particular frame...