Subject:
|
Re: Any LDView users with quad-core machine?
|
Newsgroups:
|
lugnet.cad.dev
|
Date:
|
Wed, 12 Mar 2008 19:31:59 GMT
|
Viewed:
|
10513 times
|
| |
| |
In lugnet.cad.dev, Travis Cobbs wrote:
> In lugnet.cad.dev, Kevin L. Clague wrote:
> > Hi Travis,
> >
> > Congratulations on getting SMP working in LDView.
>
> Thanks.
>
>
> > If you can run on Solaris, I have an 8 core with 8 hardware thread box in my
> > office I could try things on. Maybe you'll only use 32 given the above
> > algorithm.
>
> Yowza! In theory it will run on Solaris as long as QT 3.3 is installed, but I'm
> not sure if anyone has tried that. Note also that it would have to be a 32 bit
> compile, which may or may not be easy on Solaris. The current source also needs
> libjpg, libpng, and boost to be installed in order to build. If you want to try
> building it, or can give me a user account to do it myself, it would certainly
> be cool.
>
> On your box, it would in theory use 34 total threads (1 foreground thread, 32
> background threads for conditionals, and 1 background thread for sorting).
> Having said that, the sorting generally takes very little time, so it wouldn't
> be very apparent that that thread was using a processor. Also, unless you have
> a heinously fast graphics card (probably faster than any in existence), the
> conditional calculations aren't going to keep 32 cores busy for the time it
> takes to draw the rest of the model. My dual-core Athlon can basically get the
> conditionals calculated in the background fast enough that performance with and
> without conditionals is very similar. That means that I'm being limited by my
> video card (an nVidia 7800GT).
>
> Also, while I believe my code that detects the number of available processors
> will work on a Solaris box, it of course hasn't been tested there. (I use
> sysconf(_SC_NPROCESSORS_ONLN) if _SC_NPROCESSORS_ONLN is defined, and
> sysconf(_SC_NPROCESSORS_CONF) if _SC_NPROCESSORS_CONF is defined but
> _SC_NPROCESSORS_ONLN isn't defined or the other sysconf returns -1. (The first
> is the number of processors online; the second is the number of processors on
> the machine.) I don't know if these would return 8 or 64, though.
>
>
> > I have access to a two processor chip machine with 2x8x8 strands that I might
> > be able to boot Solaris on.
>
> What does that machine have on it? The 8-core 8-thread processors are Sparcs,
> right?
>
> --Travis
Hi Travis,
Qt 4 provides threads. It can tell you how many ideal threads there are
(presumably the number of hardware strands).
Yes, the instruction set is SPARC, with 64 bit integer registers (31 of
those), plus the usual 32, 64, 128 bit floating point (lots of those too).
The SPARC chip has a 4M L2 cache (8K I-cache, and 8K D-cache per core). It
also has PCI express access to the graphics card. I've been working (on and
off) on getting an SMP mandelbrot generator (13 instruction loop, run lots of
times) working. Also it has 2 1Gb ethernet ports and 2 10Gb ethernet ports.
I'd guess I have 4 to 8 G of FBDIMMs in four banks (for the single chip
version) with the one I have. The two (or four) chip version has two banks of
FBDIMMs per chip.
I can give it a try sometime, but for now I'm hammering on LPub 4.0.
Kevin
|
|
Message has 1 Reply: | | Re: Any LDView users with quad-core machine?
|
| (...) I actually saw the "ideal num threads" call in the QT 4 documentation when I was trying to figure out how to get the number of CPUs on a Linux box. However, even though the sysconf calls are in a QT-only section of LDView's code, I can't use (...) (17 years ago, 12-Mar-08, to lugnet.cad.dev)
|
Message is in Reply To:
| | Re: Any LDView users with quad-core machine?
|
| (...) Thanks. (...) Yowza! In theory it will run on Solaris as long as QT 3.3 is installed, but I'm not sure if anyone has tried that. Note also that it would have to be a 32 bit compile, which may or may not be easy on Solaris. The current source (...) (17 years ago, 12-Mar-08, to lugnet.cad.dev)
|
7 Messages in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|