chaboud

40 ms·

421.
by chaboud 9y ago
We saw substantial performance improvements from the increased register count of AMD64 when I lead the porting effort of a video editing/compositing suite back in the early days. Because our engine took advantage of memory-mapped unnam
422.
by chaboud 9y ago
Registers may be stored to RAM (though there are often local register files in the CPU). If you're on a modern x86 platform, you may have encrypted memory support. For Intel, look up SGX. For AMD, look up SEV. Each of these is way
423.
by chaboud 9y ago
You can interview for that? (Not a joke) Very few of the job functions we rely on in my space can be reasonably handled in 45 minutes. Resultantly, every interview question is indirect. We either reduce problems to what we hope are agg
424.
by chaboud 9y ago
I had a few simple reasons: 1) PoE wired backhaul. 2) Way more management and configuration control. 3) Separability of routing, switching, and wireless. 4) Simple L2TP VPN endpoint support. I'm very happy with my choice, and I just
425.
by chaboud 9y ago
Or just marketing as... goes to 11....
426.
by chaboud 9y ago
That is, effectively, what I was saying. The list is dumb. The best thing we can do is move on.
427.
by chaboud 9y ago
In the top 10 list, two cities in the top five are a ~1 hour drive apart. That makes them one commute zone for a lot of workers. How do I know? I commute from San Francisco to Sunnyvale most days, and I see a lot of people doing the same
428.
by chaboud 9y ago
Right, though some side effects (e.g. Status registers) might be affected. (One reason strict modes blow floating point optimization, generally.)
429.
by chaboud 9y ago
Yeah. I'm partially unconvinced. That is to say that, if the compiler Dev wants to take on the completeness responsibility, cool. If not, I'm okay with that, too. Adding the optimization and screwing up the edge cases is the onl
430.
by chaboud 9y ago
I'd be more curious what Clang did for values that would result in different behaviors for some ranges of a. I understand that there are limits, and that there is range checking in literals for escalation to, for instance, 64-bit valu
431.
by chaboud 9y ago
Since it's a *=, I'd expect the operation to occur in double, quantize accordingly, and handle overflow as double-to-int would, rather than int multiplication would. For 10.0 specifically, double is more than large enough to, as f
432.
by chaboud 9y ago
While there may be some compiler misses, I started doubting the whole thing when I hit: "Missed simplification of multiplication by integer-valued floating-point constant Variant of the above code with the constant changed slightly: in
433.
by chaboud 9y ago
Citations? I'm happy to read any research you have. (You can also come over, crack a beer, and record balloon pops and acoustic instruments... and an electric guitar... and electric kazoo.) The xiph-cited studies I've seen show
434.
by chaboud 9y ago
So, having come from an image-processing background to put in 15 years of experience at audio processing companies, I've seen the xiph arguments, or something like them, several times. Heck, I worked for Sony (where that precious 22KH
435.
by chaboud 9y ago
I was arguing against the author's assertions that: 1) Euclidean geometry is a plane (though, on re-reading, she said that a plane is pure Euclidean geometry). So, to be fair to the author, that may well be me over-reading her stateme
436.
by chaboud 9y ago
I would expect that to work to a point, but coherence and interlocking are eventually going to rear their ugly heads. I've created and lived with multiple inter-process high-performance multimedia/data systems (e.g. video special
437.
by chaboud 9y ago
Oh, this is painful to read... As a physics dropout (hey, '90s tech boom) who works with theatrical costumers on a regular basis as the "engineering" guy (who sews, rhinestones, and hot-glues with the rest of them) there&#x27
438.
by chaboud 9y ago
Unfortunately, it's wholly impractical to do real work with message busses alone if you want to maximize performance. See: Every performance inter-process system ever... Could we cover a number of cases with copy-on-write semantics and
439.
by chaboud 9y ago
Unfortunately, busses in many metropolitan areas run on staggered schedules. It's why middle school starts at 7:20am while grade school starts at 8:10 in some cities. A typical service day for a school bus can be pretty packed.