owlbite

19 ms·

151.
by owlbite 4y ago
Yeah, I see from other comments he went the sovereign idiot route. An interesting observation is I'm 95% sure back in my (european) home country, they'd just take the clean up action to court and seek forced entry on grounds of ha
152.
by owlbite 4y ago
I mean sure, fighting things in court and losing is going to cost you more money than admitting defeat. However, at some point the council should just say "you had your chance to rectify the issue, we hired a contractor to fix that for
153.
by owlbite 4y ago
If the data is Sparse (which is not uncommon for large matrices in the real world), you can exploit the sparsity to do significantly better then O(n**3).
154.
by owlbite 4y ago
Same old story - as soon as you start using a metric to incentivize people, they optimize to the metric. If the metric is not well aligned with what you actually wanted (and I mean optimizing it is what you want, not currently correlated be
155.
by owlbite 4y ago
Does anyone know if these algorithms are practical on real problems on real machines, or is the constant term large enough you would normally just use a more traditional algorithm for most work?
156.
by owlbite 5y ago
The main problem is each subsection of a chapter in NR is really a subfield in and of itself, and to adequately describe a decent modern implementation is at least one if not multiple books. First make sure you're familiar with backgro
157.
by owlbite 5y ago
Don't roll your own numerics. There are normally many subtle gotchas that affect numerical stability of a method, as well as great attention to detail required to handle degenerate edge cases. And that's before we get to subtle bu
158.
by owlbite 5y ago
But its algorithms are often incorrect, inefficient and numerically unstable. Just use a library written by an expert in whatever your area of interest is.
159.
by owlbite 5y ago
The nakedly predatory user-hostile nature of many large actors on the web really gave them no choice.
160.
by owlbite 5y ago
Completely different to the order I use, which basically amounts to "how hard is it to fix this later". So most important questions are: (1) API must be as correct as possible, especially if its exposed outside of the immediate te
161.
by owlbite 5y ago
Right, but portable code != portable performance. See also: OpenCL. There's also the observation "keep simd vectors small but many" (e.g. Apple's arm chips) over "super long vectors" (intel avx512) is superior
162.
by owlbite 5y ago
I must admit, I've yet to find a performance relevant loop that I can't do in simd that doesn't also have dependencies on previous iterations such that no magic instruction set is going to help unless it's capable of tim
163.
by owlbite 5y ago
What have been the biggest advances in linker technology in the last decade or so?
164.
by owlbite 5y ago
cynically: computation is easier to mark than reasoning.
165.
by owlbite 5y ago
How does this compare to the native BLAS in the Accelerate library?
166.
by owlbite 5y ago
End-to-end encryption != no encryption. It's seems they're quite happy as long as authorized parties (adtech, intelligence agencies, etc) can man-in-the-middle the thing. Of course, tech companies never get breached and are except
167.
by owlbite 5y ago
My cousin seemed to have it worked out. Do as many relevant AP classes for credit as possible, finish the first two years worth of credits at a community college for $cheap and then transfer to a brand-name school for the final part. My gen
168.
by owlbite 5y ago
Non-linear optimization is very popular in many fields (non-linear = real world, optimization = find me the least cost/most profit/lowest energy). Nearly all scientific problems can be reformulated as an optimization problem. Trai
169.
by owlbite 5y ago
Self driving cars seem like a dangerous threat vector if an adversarial image can be deployed in such a way as to cause them to commit dangerous maneuvers on demand.
170.
by owlbite 5y ago
As mentioned by at least one other comment (buried in a thread), in reality, don't do this. If you care about the performance of the loop /that/ much, just hand code the vectorization/assembly yourself. Use some sort of
171.
by owlbite 5y ago
Unless, of course, you have some algorithm that depends on signed zeros. Which is basically the same with all the optimizations the article complains about. I'd suggest -ffp-contract=fast is a good idea for 99% of code. It's only
172.
by owlbite 5y ago
I'd guess the aim is to be secure on all components (most of these things have their own processor(s)). If you can compromise one component you can move from there to compromise another one, until you get to something worthwhile. I don
173.
by owlbite 5y ago
So it depends on how you want to detect overflow. Trapping on math exceptions or otherwise pulling an emergency break doesn't really fit with modern hardware design if you want performance (for one thing, what do you do with the rest o
174.
by owlbite 5y ago
Rename to: -fdisable-math-safeties
175.
by owlbite 5y ago
I'm not covinced companies are out to "make their old products feel worse over time". It's just the inevitable consequence of the steady march of technological progress. A 10-20% per year performance improvements and new
176.
by owlbite 5y ago
You probably have to be very clever to set this up so they can't be Company A (US) Ltd registering very little profit as they have to pay royalties to Company A (Tax Haven) limited without also inflicting massive collateral damage to s
177.
by owlbite 5y ago
Do we actually have a source for the polarization factoid delivered here? The only source I can find with a quick google is a 2020 study based on relatively old data from about 8 western countries. Further this focused on party-aligned pola
178.
by owlbite 5y ago
One of the problems I see is that we are now in the era of what one might call "population scale mind control", actually delivering on the promises of propaganda. Given enough data you can find a large enough subset of the populat
179.
by owlbite 5y ago
Sure, but if a cigarette company commission a whole load of research, sat on the "bad" results and publicized the "good" ones, and then a whistleblower came forwards and showed the world the bad ones they had sat on...
180.
by owlbite 5y ago
I mean, don't do it for the "green". Do it for longer battery life for your users.

More ›