Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
berkut
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
71 ms
·
721.
▲
by
berkut
13y ago
What? Not if they're CPU-bound they're not. Look at Dell/HP workstation/servers: Z600/800 | Z620/Z820 range - dual sockets, all of them. The cheaper configs might only come with one CPU, but they've got tw
722.
▲
by
berkut
13y ago
What? They're not uncommon for servers or high-performance workstations - Xeon dual/quad socket are very common. I've seen close to 250% speed increases with NUMA-aware code (and tying thread affinity down) - writing things s
723.
▲
by
berkut
13y ago
Excellent article - writing NUMA-aware code such that it's aware of where a thread's memory is compared to what socket it's on is very important for high-performance code. You can get serious speed gains by reducing latency.
724.
▲
by
berkut
13y ago
In 1977, NSA were planning to have 157 GB of storage: http://cryptome.org/2013/03/cryptologs/cryptolog_28.pdf
725.
▲
by
berkut
13y ago
> Not just different, inferior. Because they're not going to spend time working out by trial and error (it's possible based on timing and evaluating code) how many float ops / cycle each AMD chip can do. For their own chip
726.
▲
by
berkut
13y ago
You could look at the graphs - it's pretty obvious ICC wins in almost all benchmarks on all processors, not just one benchmark. Intel state they do different optimisations for different chips, and I'd guess it does it based on how
727.
▲
by
berkut
13y ago
Not since 2010 it doesn't: http://www.hardware.fr/articles/847-1/impact-compilateurs-ar... It can generate code which when run on AMD is faster than GCC and MSVC.
728.
▲
by
berkut
13y ago
It's a good conclusion I feel: this is always the issue with language benchmarks - who wrote the code, and how good were they with each of the languages. Similarly as the article points out, the compiler matters a lot: ICC can in certa
729.
▲
by
berkut
13y ago
In fact they do: Phil Schiller himself on 10th June at the WWDC keynote mentioned that we've ported Mari to OS X, and he was extremely impressed with it and the fact that we've ported it. It's heavily Qt-based, and doesn&#x
730.
▲
by
berkut
13y ago
Not quite: it's a highly-custom multiple selection mask button (with intelligent title based on the selection options) - doesn't really have any native equivalent on any platform as far as I'm aware: http://imgur.c
731.
▲
by
berkut
13y ago
It's got a custom overall stylesheet for the tab widgets as I don't like the OS X tab style or colour, so they're completely different. Qt "emulates" native control appearances by drawing them itself. This means it
732.
▲
by
berkut
13y ago
It's got a master stylesheet attached for the tabs, but admittedly, even without it, it's not identical. http://imgur.com/68b9eoO - is native Cocoa moc-up in XCode. I concede the borders are slightly different, bu
733.
▲
by
berkut
13y ago
Good points - thanks. Although having looked at hopper, I must say I'm not convinced it seems any more native than a Qt app on OS X could look if work was concentrated with this goal in mind. So as I said in my VirtualBox comment, I
734.
▲
by
berkut
13y ago
Which keyer? Primatte's had a few issues in 6.3 and previous. on 10.6 (and 10.7), memory allocation/paging is pretty atrocious if you haven't got much free mem left. Basically, OS X prioritises paging to disk over freeing up
735.
▲
by
berkut
13y ago
Can you give me an example of the "non-intuitive non-standard UI"? This may well be true, but it's quite possible it's just the developer was lazy and didn't do all the effort to make it look native. Edit: in fact,
736.
▲
by
berkut
13y ago
What version (OS X and Nuke) and how does it crash doing what as an example?
737.
▲
by
berkut
13y ago
C++ and Qt.
738.
▲
by
berkut
13y ago
Local widgets & theming yes, although generally we ship with our own themes by default as we make VFX apps, and the last thing artists want is bright UI to distract them. I don't believe there are general UX guidelines for Linuxes,
739.
▲
by
berkut
13y ago
Just use Qt. Works brilliantly, great API and documentation, and gets you 95% of the way to Windows versions and 85% of the way on OS X too. On OS X you'll have issues with native painting within OpenGL drawing and maybe event throttli
740.
▲
by
berkut
13y ago
We "officially support" RHEL 5.4 and RHEL 6.0 (and thus corresponding CentOSes). But we've got customers running everything from RHEL 4.5 to the latest and greatest Ubuntus, Mints and Arch. All with the same binary installers
741.
▲
by
berkut
13y ago
What rubbish. Any specifics, or are you just trolling? The company I work for develops high-end desktop software for Linux, Mac and Windows, and Linux generally causes us the least trouble and OS X (thanks mostly to atrocious graphics drive
742.
▲
by
berkut
13y ago
They also helped in the "Battle of the beams" beforehand, as well.
743.
▲
by
berkut
13y ago
Yeah. It used to be very easy to get copies of birth certificates and things (with almost no proof you're the person it's for), but at least in the UK, they tightened this up in around 2004.
744.
▲
by
berkut
13y ago
This is one of the reasons Mari's finally able to be ported to OS X, as was shown at WWDC.
745.
▲
by
berkut
13y ago
This is fine for VMs and maybe low-load servers, but the problem for high-performance servers is ARM sucks memory-wise as they don't have the patents on the cache hierarchies that Intel has (and AMD can use), and thus have very limited
746.
▲
by
berkut
13y ago
Actually the details aren't that surprising, given what processor and gpu tech's available at present. I'm surprised there haven't been leaks about Maverick and the Mac Pro - I work for The Foundry, and tomorrow Pixar ar
747.
▲
by
berkut
13y ago
Depends - as it's designed (using GRC's Paramarine - http://www.qinetiq.com/what/capabilities/maritime/Pages/nava... ) you can virtually unit test things like propulsion, stability and buoyancy.
748.
▲
by
berkut
13y ago
How does that work with timezones and summertime? They also then can't use caching - that's going to impact them a bit I think...
749.
▲
by
berkut
13y ago
I've tried one (dev one) and the resolution's crap - 1280 × 800 spit across both eyes. It's shocking how bad it was. The final one will have full HD, but even then I think it needs a quadrupling in res before it's even acceptable.
750.
▲
by
berkut
13y ago
Other than grading, compositing and sound, rendering's generally the final step - definitely, the modelling, lighting, animation and texturing have to be done before the rendering can be started.
More ›