Message boards : Number crunching : Versions for x86_64 platforms avaible ?
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
Daniel Koch

Send message
Joined: 25 Jul 05
Posts: 1
Credit: 151,195
RAC: 0
Message 8926 - Posted: 25 Jul 2005, 23:29:05 UTC

Are there any client versions for x86_64-unkown-linux-gnu platforms available ? If not, will there be any in the future ?? If there was only a source code available...

Daniel Koch
ID: 8926 · Report as offensive     Reply Quote
Travis DJ

Send message
Joined: 29 Sep 04
Posts: 196
Credit: 207,040
RAC: 0
Message 8928 - Posted: 26 Jul 2005, 2:40:11 UTC - in response to Message 8926.  

In short, not for LHC@Home. Sixtrack does not stand to benefit from "64-bit" technologies at this time. For reasons why, see this thread.
ID: 8928 · Report as offensive     Reply Quote
Desti

Send message
Joined: 16 Jul 05
Posts: 84
Credit: 1,875,851
RAC: 0
Message 8937 - Posted: 26 Jul 2005, 12:51:23 UTC - in response to Message 8928.  

> In short, not for LHC@Home. Sixtrack does not stand to benefit from "64-bit"
> technologies at this time. For reasons why, see <a> href="http://lhcathome.cern.ch/forum_thread.php?id=1460">this[/url] thread.
>

But it would be nice too have it just for better compatibility.
Linux Users Everywhere @ BOINC
[url=http://lhcathome.cern.ch/team_display.php?teamid=717]
ID: 8937 · Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 2 Sep 04
Posts: 545
Credit: 148,912
RAC: 0
Message 8939 - Posted: 26 Jul 2005, 17:09:45 UTC - in response to Message 8937.  

> But it would be nice too have it just for better compatibility.

As stated in the other thead, making a 64 bit version conveys no benefits, does not make it "more" compatible. All it does is add another version that has to be tested to see if the outputs are equivelent to the current crop.

There is nothing magic about a 64-bit capable processor and making things 64-bit only delivers benefits for certain classes of programs of which sixtrack is not one. The primary place where you get benefit is in large databases where having a large physical memory space allows you to have more stuff resident in the physical space.

You also get an even larger virtual space ... so your ability to use large data sets is also increased. However, you are not going to see classes of programs running on your home PC that require that capability.
ID: 8939 · Report as offensive     Reply Quote
Desti

Send message
Joined: 16 Jul 05
Posts: 84
Credit: 1,875,851
RAC: 0
Message 8940 - Posted: 26 Jul 2005, 18:01:41 UTC - in response to Message 8939.  

> > But it would be nice too have it just for better compatibility.
>
> As stated in the other thead, making a 64 bit version conveys no benefits,
> does not make it "more" compatible.

If i have a x86_64 system without working ia32 emulation, i'm not able to run 32 bit programs.
Linux Users Everywhere @ BOINC
[url=http://lhcathome.cern.ch/team_display.php?teamid=717]
ID: 8940 · Report as offensive     Reply Quote
Heffed

Send message
Joined: 2 Sep 04
Posts: 71
Credit: 8,657
RAC: 0
Message 8961 - Posted: 27 Jul 2005, 9:25:47 UTC - in response to Message 8940.  

> If i have a x86_64 system without working ia32 emulation, i'm not able to run
> 32 bit programs.

Do you have one?
ID: 8961 · Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 2 Sep 04
Posts: 545
Credit: 148,912
RAC: 0
Message 8963 - Posted: 27 Jul 2005, 11:20:38 UTC

I have a Xeon-64 bit capable. But, running windows I have 32-bit compatibility so the programs run in 32-bit mode.

Now what is interesting about that is that some of the Science Applications seem to run faster on the 64-bit computer, with others slower. But, this is only on a week of data ... so, it may be just a skew in the work unit processing times. YOu can still see the old average processing time page on the old site. I am still updating that with data from the BOINC View logs.
ID: 8963 · Report as offensive     Reply Quote
Travis DJ

Send message
Joined: 29 Sep 04
Posts: 196
Credit: 207,040
RAC: 0
Message 8974 - Posted: 27 Jul 2005, 17:02:21 UTC - in response to Message 8940.  
Last modified: 27 Jul 2005, 17:15:51 UTC

> If i have a x86_64 system without working ia32 emulation, i'm not able to run
> 32 bit programs.

"IA-32 emulation" is a feature of Intel's Itanium (Itanic) processors. IA-64 is Intel's 64-bit technology the Itanium is based upon. IA-64 is 100% incompatible with AMD64, which was developed by AMD and adopted by Intel for their x86 processor line and renamed EMT64.

It's not possible to purchase an AMD Athlon64 or an EMT64-enabled Intel Pentium 4 that is incapable of running 32-bit applications. 64-bit mode is much like 32-bit "protected" mode; an enhancement of the original 8/16bit "real" mode available on all x86 CPUs.

That's why your statement is so curious. Which processor do you own? Vendor & Model? Operating system?

ID: 8974 · Report as offensive     Reply Quote
Desti

Send message
Joined: 16 Jul 05
Posts: 84
Credit: 1,875,851
RAC: 0
Message 9001 - Posted: 28 Jul 2005, 13:52:15 UTC - in response to Message 8974.  

> > If i have a x86_64 system without working ia32 emulation, i'm not able to
> run
> > 32 bit programs.
>
> "IA-32 emulation" is a feature of Intel's Itanium (Itanic) processors. IA-64
> is Intel's 64-bit technology the Itanium is based upon. IA-64 is 100%
> incompatible with AMD64, which was developed by AMD and adopted by Intel for
> their x86 processor line and renamed EMT64.
>
> It's not possible to purchase an AMD Athlon64 or an EMT64-enabled Intel
> Pentium 4 that is incapable of running 32-bit applications. 64-bit mode is
> much like 32-bit "protected" mode; an enhancement of the original 8/16bit
> "real" mode available on all x86 CPUs.
>
> That's why your statement is so curious. Which processor do you own? Vendor
> & Model? Operating system?
>
>


Take a look at this:

http://www.gentoo.org/proj/en/base/amd64/technotes/index.xml?part=1&chap=4


Linux Users Everywhere @ BOINC
[url=http://lhcathome.cern.ch/team_display.php?teamid=717]
ID: 9001 · Report as offensive     Reply Quote
ebahapo
Avatar

Send message
Joined: 18 Sep 04
Posts: 30
Credit: 104,162
RAC: 0
Message 14475 - Posted: 5 Aug 2006, 8:29:17 UTC

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?
ID: 14475 · Report as offensive     Reply Quote
Toby

Send message
Joined: 1 Sep 04
Posts: 137
Credit: 1,691,526
RAC: 383
Message 14500 - Posted: 7 Aug 2006, 18:36:12 UTC

Most Linux distros can run 32 bit binaries if set up correctly. I have an x86_64 system running gentoo. It can run the 32 bit version of sixtrack just fine as can be seen here. You just have to download the 32 bit binary and make an app_info.xml file. I made a post about this on our team message boards a while ago. You can read it here.
- A member of The Knights Who Say NI!
My BOINC stats site
ID: 14500 · Report as offensive     Reply Quote
ebahapo
Avatar

Send message
Joined: 18 Sep 04
Posts: 30
Credit: 104,162
RAC: 0
Message 14581 - Posted: 21 Aug 2006, 21:54:19 UTC - in response to Message 14475.  

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too...
ID: 14581 · Report as offensive     Reply Quote
ebahapo
Avatar

Send message
Joined: 18 Sep 04
Posts: 30
Credit: 104,162
RAC: 0
Message 14691 - Posted: 20 Sep 2006, 15:39:28 UTC - in response to Message 14581.  

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too...

HashClash now supports AMD64 with a 32-bit Linux application too...
ID: 14691 · Report as offensive     Reply Quote
ebahapo
Avatar

Send message
Joined: 18 Sep 04
Posts: 30
Credit: 104,162
RAC: 0
Message 14881 - Posted: 27 Sep 2006, 19:32:32 UTC - in response to Message 14691.  

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too...

HashClash now supports AMD64 with a 32-bit Linux application too...

HashClash now supports AMD64 on Windows with a 32-bit application...


ID: 14881 · Report as offensive     Reply Quote
ebahapo
Avatar

Send message
Joined: 18 Sep 04
Posts: 30
Credit: 104,162
RAC: 0
Message 14890 - Posted: 28 Sep 2006, 22:55:37 UTC

FWIW, running two instances of the client, one the 32-bit client, the other, the 64-bit client, on the same 4-core system, but limiting each client to 2 cores, I can compare the relative performance of 32-bit and 64-bit SIMAP's HMMER: the 64-bit version is about 7% faster.

By enabling vectorization (supported by default on AMD64), the SIMAP developers observed other 8% improvement.

Bottom line: porting the project application to AMD64 has the potential to improve performance by 15%!

ID: 14890 · Report as offensive     Reply Quote
ebahapo
Avatar

Send message
Joined: 18 Sep 04
Posts: 30
Credit: 104,162
RAC: 0
Message 15037 - Posted: 10 Oct 2006, 15:07:44 UTC - in response to Message 14881.  

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too...

HashClash now supports AMD64 with a 32-bit Linux application too...

HashClash now supports AMD64 on Windows with a 32-bit application...

Guess what? Leiden Classical now supports AMD64 on Linux with a 32-bit application. ;-)

ID: 15037 · Report as offensive     Reply Quote
Profile FalconFly
Avatar

Send message
Joined: 2 Sep 04
Posts: 121
Credit: 592,214
RAC: 0
Message 15038 - Posted: 10 Oct 2006, 22:35:01 UTC - in response to Message 15037.  

Not sure if it helps, but :

I run all AMD64 Systems on full native 64bit Linux (Fedora Core 3, 4 and 5). So far, all Projects worked just perfect with their 32bit Clients, if that was in question.
Scientific Network : 45000 MHz - 77824 MB - 1970 GB
ID: 15038 · Report as offensive     Reply Quote
River~~

Send message
Joined: 13 Jul 05
Posts: 456
Credit: 75,142
RAC: 0
Message 15052 - Posted: 11 Oct 2006, 6:27:31 UTC - in response to Message 15038.  
Last modified: 11 Oct 2006, 6:30:58 UTC

Not sure if it helps, but :

I run all AMD64 Systems on full native 64bit Linux (Fedora Core 3, 4 and 5). So far, all Projects worked just perfect with their 32bit Clients, if that was in question.


Thanks for the reminder. This can be a useful work-around.

I think the downside with running the 32-bit client is that it will only ever ask for the 32-bit app. If I am right then the you would never get the speed advantage if any of your chosen projects produced a 64-bit app. In exchange you get to run projects that do not offer work to the 64-bit platform.

Whether this is a good trade off is a personal judgment which includes how much you want to crunch particular projects, and how much you value performance gain (estimated earlier in this thread as ~15% on one project) if the wider app exists at all.

River~~

PS - for those who are new enough not to have picked up the distinction, the client is the BOINC software that you downloaded manually and installed. The app is the project-specific software that is downloaded behind the scenes without any human involvement. The issue here is that the software tries to match the bit width of the app to the client.
ID: 15052 · Report as offensive     Reply Quote
Profile FalconFly
Avatar

Send message
Joined: 2 Sep 04
Posts: 121
Credit: 592,214
RAC: 0
Message 15069 - Posted: 11 Oct 2006, 23:14:38 UTC - in response to Message 15052.  

Thinking about it, that's correct of course...

Not sure when Berkeley will start creating official 64bit BOINC Binaries, but I would assume they'd start looking more into it when there are enough Windows Vista users asking for 64bit support. That might take another year, however.
Scientific Network : 45000 MHz - 77824 MB - 1970 GB
ID: 15069 · Report as offensive     Reply Quote
Travis DJ

Send message
Joined: 29 Sep 04
Posts: 196
Credit: 207,040
RAC: 0
Message 15070 - Posted: 12 Oct 2006, 2:06:05 UTC
Last modified: 12 Oct 2006, 2:24:12 UTC

This topic seems to get asked over and over again .. the "64-bitness" refers to integer functions .. which explains the larger physical, virtual, and register memory sizes - and potentially faster integer functions. The thing about Sixtrack is the important parts of it aren't ran on the integer (ALU) but rather in the Floating-Point Unit (FPU). The FPU has 80-bit registers going back into the days of the 386DX series CPUs, IIRC.

Reasons why a 64-Bit native Sixtrack doesn't make sense:
1) the bulk of the work is executed in the FPU (up to 80-bit registers)
2) all 64 & 32 bit OSes for the x86 platform run 32-bit applications
3) the calculations are somewhat FPU specific (i.e. no PowerPC version)

..porting and testing for a 64-bit native applcation where no visible or only marginal performance boost is seen is not worth it. The real beauty is in MMX/SSEx computing, especially on a 64-bit x86 CPU. This topic has also been covered to some extent. However, Sixtrack produces unreliable results when MMX or SSE1-4 is used in 32-bit mode; imagine the results when additional and larger registers are introduced due to 64-bit enhancements that have to be accounted for. That is how sensitive this program is. Steam processing is not for every application.

Trust me I'm all about progress, doing things faster and better. In this case the tradeoff for those speedups is precision and that's what matters most. It would suck to blowout a section of the LHC when it goes live because the math was wrong.. it happened to NASA with the Mars Climate Orbiter when it crashed into Mars in 1999 where part of its mathematical programming was in English Units and the rest in Metric. :)
ID: 15070 · Report as offensive     Reply Quote
1 · 2 · 3 · Next

Message boards : Number crunching : Versions for x86_64 platforms avaible ?


©2024 CERN