SGI: Hardware

SGI's first Linux-based Visualization System - Page 2

Quote:
Name: Arun Ramakrishnan ( mailto:[email protected] ) 10/11/04

Andi Kleen ( mailto:[email protected] ) on 10/11/04 wrote:
---------------------------
>On the SGI page for the Prism they announced the MIPS->IA64 emulator Arun was talking about earlier:
>
>http://www.sgi.com/products/visualization/prism/features.html
>>
>
>"QuickTransit dynamic translator Run existing IRIX applications*"
>
>* Check with your local SGI sales representative for IRIX application compatability and pricing
>
>I suspect it is more difficult to emulate Irix on Linux
>than MIPS on IA64.
>

Andi,in case of IRIX it was a more of a problem of resources.
Here i think Intel may have some role in funding quicktransit and the quick transit guys had help from our kernel developers to make sure the performance of this product is really competitive.

This is just being used to bring apps from irix to linux mainly in the graphics side .
Some of those apps dont even have any source code left !!!
Some of others have been ported from irix to windows by ISVs but they dont believe in linux yet.
So this proves to be a good starting point for encouraging them to run it on Prizm first and then coax them to port it over to Linux.

Of course,Quicktransit cant help apps which use the Irix real time features,xfs GRIO features or IR(Infinite Reality) specific hardware hacks :-)

Byt hey something is better than nothing :-)

And the first few weeks of running Maya on Prizm under quick transit,it was already faster than a IR based system.
Now then what would happen if this app was running natively on IA64 Linux is what excites us.

Cheers,
Arun

Cheers,
Arun

_________________
"Ever tried. Ever failed. No matter. Try Again. Fail again. Fail better"
87Porsche wrote:


Way cool to see that Maya works on so many different SGI/mips configurations.
[quote="Jason"]Way cool to see that Maya works on so many different SGI/mips configurations.[/quote]
Yeah, but the list of known bugs has been growing since version 3.0.
Time to update the SGI.com Image Library !

_________________
We are here on Earth to do good to others. What the others are here for, I don't know.
87Porsche wrote:


That's the x86 version of Maya.
As we all know, the x86 code execution speed of the Itanic is best described as "pathetic" . :twisted:

Futhermore, this little blurb doesn't sound good:
Code:
Maya 6.0 is currently Unsupported on the following 64-bit systems:
Itanium:    Red Hat Enterprise Linux WS (Intel Itanium)
jan-jaap wrote:
As we all know, the x86 code execution speed of the Itanic is best described as "pathetic" . :twisted:


From what I understand the Itanic is a lot like RISC in some ways - execution speed is very dependent on smart compilers. So what do they have on Linux ? Gcc ? hmmm.
hamei wrote:
From what I understand the Itanic is a lot like RISC in some ways - execution speed is very dependent on smart compilers. So what do they have on Linux ? Gcc ? hmmm.


I believe the intel compiler (icc) is used on these machines. Hard to guess its performance without real tests though

Matthias

_________________
Life is what happens while we are making other plans
I once attended an SGI/Intel presentation that talked about the Intel compilers on IA64, and how the GCC stuff would (because of some such optimization) run at best 75% the speed and how the AMD folks would not longer be able to cross compile. Sounded like Microsoft tactics to me at the time....
Brombear wrote:
hamei wrote:
From what I understand the Itanic is a lot like RISC in some ways - execution speed is very dependent on smart compilers. So what do they have on Linux ? Gcc ? hmmm.


I believe the intel compiler (icc) is used on these machines. Hard to guess its performance without real tests though


I just know what I've read about developers in the HP camp screaming bloody murder about "no tools ! no tools ! where the hell are all the optimized tools you promised us ?" Meanwhile Carly bibble-babbles on about how wonderful everything is ... if you can believe the Register, HP users are not happy campers about Itanic.

How long did it take MIPSPro to become really effective ?
hamei wrote:
jan-jaap wrote:
As we all know, the x86 code execution speed of the Itanic is best described as "pathetic" . :twisted:

From what I understand the Itanic is a lot like RISC in some ways - execution speed is very dependent on smart compilers. So what do they have on Linux ? Gcc ? hmmm.

x86 code is slow on Itanium because it is run in some sort of emulation mode. SGI uses the Intel compilers on their Linux systems. I've heard that they're very good and getting better. GCC as always is behind compared to the official compilers.

Also, Pro/E on Linux is still around. But PTC droped support for Itanium (in favor of x86-64/EMT64) maybe six months ago.
From what i hear the HP itanium compilers are the best however they are only available on HP-UX and OpenVMS (well when the VMS port is finished) as for Itanium performance running x86 code. Well most of the slowness was due to the fact that the chip had dedicated area to x86 however very little attention was put into improving it. Last i heard it gave the same performance as a low Pentium II (on Itanium 2) it's been ditched from future versions and been replaced with a software approach, akin to FX32! that ran on Alpha. Intel actually got that and other DEC IP (eg. Alpha) after the Compaq and HP merger.

_________________
"Ever tried. Ever failed. No matter. Try Again. Fail again. Fail better"
hamei wrote:
Brombear wrote:
hamei wrote:
From what I understand the Itanic is a lot like RISC in some ways - execution speed is very dependent on smart compilers. So what do they have on Linux ? Gcc ? hmmm.


I believe the intel compiler (icc) is used on these machines. Hard to guess its performance without real tests though


I just know what I've read about developers in the HP camp screaming bloody murder about "no tools ! no tools ! where the hell are all the optimized tools you promised us ?"


It's not icc and ifc/ifort, but ecc and efc/efort on Itanium systems. And there is one performance tool you can run on an itanium, because of its specific counter registers included in the CPU core.
The tool is called HistX and can be downloaded from the SGI site:

http://www.sgi.com/products/evaluation/altix_histx/

But i admit, i miss ssrun/cvd/perfex on Itanium systems. Also, there's a lot of funky performance issues like some code running like mad on the PIV Xeon will crawl on Itanium. And you're completely dependent on ecc for your optimisations, so no pragma's like on MIPS will help you. I have seen and tested an Itanium2 1.5 GHz with 3MB cache to be slightly slower than a PIV Xeon 3.05GHz with respect to floating point fortran code. And considering a dual Itanium2 machine costs triple the amount of a dual PIV machine, the choice is easily made...
RageX wrote:
GCC as always is behind compared to the official compilers.


You may be right for non-x86 platforms. The fact that itanic leaves a lot of work to the compiler, and the companies involved then patented that work doesn't help either. But on x86, gcc-3.4 will perform roughly on par with icc.

In fact, numerical kernels like atlas perform very close to cpu peak performance when compiled with either of them. If you want to bash a compiler, try MSVC.

Don't take my word for it, but do some tests yourself. I did :D
i like the cabinet style...
skywriter wrote:
i like the cabinet style...

:D :lol:

me too (damn!)
squeen wrote:
skywriter wrote:
i like the cabinet style...

:D :lol:

me too (damn!)


the nice part about it is that when people start wanting to do case conversions for their peecees, it'll already be done !
:?
SGI still talks about a five year IRIX commitment.

Also the 1GHZ R16K is due out "soon".

After that.....?
hamei wrote:
squeen wrote:
skywriter wrote:
i like the cabinet style...

:D :lol:

me too (damn!)


the nice part about it is that when peopel start wanting to do case conversions for their peecees, it'll already be done !


bwahahahahahahaa *cough* hahaha.a.... *choke* hahahehehehe .... hehe.. phew... damn that was funny :) so much so, i will enable smiles in this post!
squeen wrote:
SGI still talks about a five year IRIX commitment.


is that like they just EOL'd the crimson, 4 years after y2k busted it? and support in the kernel was dropped in 6.5? now THATS commitment!