Message boards :
Number crunching :
64 bit optimisation
Message board moderation
Author | Message |
---|---|
Paul Shellien Send message Joined: 6 Sep 03 Posts: 8 Credit: 1,171,683 RAC: 0 |
It's probably been asked a few times already, but is SETI optimised for 64 bit and if not, are they planning on doing so? Seems a logical step if most peeps have 64 bit Athlons (yes I have one) and also if Intel (boo hiss) are going 64 bit in the near future too <img border="0" src="http://boinc.mundayweb.com/one/stats.php?userID=515&prj=1&trans=off" /> |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
> It's probably been asked a few times already, but is SETI optimised for 64 bit > and if not, are they planning on doing so? Seems a logical step if most peeps > have 64 bit Athlons (yes I have one) and also if Intel (boo hiss) are going 64 > bit in the near future too There is no advantage to compile any of the BOINC components with 64-bit extensions at this time. The only part that might benefit from the 64-bit CPUs and compiling to 64-bit executables would be the MySQL database. Sorry. :) |
Chilean Send message Joined: 6 Apr 03 Posts: 498 Credit: 3,200,504 RAC: 0 |
|
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
True that simply compiling the SETI app for 64 bit won't gain much, but optimizing it for AMD processors does show an improvement. In this thread: http://setiweb.ssl.berkeley.edu/forum_thread.php?id=13068#100676 , Ned Slider reports that AMD optimized 32bit SETI runs around 13% faster, and 64bit perhaps a further 5%. Granted that these are under Linux, but I don't see why a similar benefit shouldn't be seen under Windows - 64bit Windows that is. Looking forward to trying it, although I've not yet seen an available version. (Is there a 64bit C compiler for Windows yet?) Never the less, the biggest SETI speed gains available are from large L2 cache, high FSB speed and dual chanel memory. |
Chilean Send message Joined: 6 Apr 03 Posts: 498 Credit: 3,200,504 RAC: 0 |
> True that simply compiling the SETI app for 64 bit won't gain much, but > optimizing it for AMD processors does show an improvement. > > In this thread: > http://setiweb.ssl.berkeley.edu/forum_thread.php?id=13068#100676 , Ned Slider > reports that AMD optimized 32bit SETI runs around 13% faster, and 64bit > perhaps a further 5%. > > Granted that these are under Linux, but I don't see why a similar benefit > shouldn't be seen under Windows - 64bit Windows that is. Looking forward to > trying it, although I've not yet seen an available version. (Is there a 64bit > C compiler for Windows yet?) > > Never the less, the biggest SETI speed gains available are from large L2 > cache, high FSB speed and dual chanel memory. > > > Yup, I got the optimized version for my Linux. |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 |
> > In this thread: > http://setiweb.ssl.berkeley.edu/forum_thread.php?id=13068#100676 , Ned Slider > reports that AMD optimized 32bit SETI runs around 13% faster, and 64bit > perhaps a further 5%. > While the above is true (actually, we're now getting 30-35% performance increases), hardly any of this appears to be due to optimizing for AMD arcitectures. About half is from using -ffast-math and the other half from using improved fast fourier transform routines. Chris Bosshard has more data on x86_64 than I do, and I'll certainly be working closely with him do make sure we have a fully optimized 64-bit client for seti (linux) shortly. *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 |
OK, I've just spoken with Chris (and Triton_b5) and we have confirmed the following: A true 64-bit seti client running on 64-bit linux is faster than a 32-bit seti client running on the same OS. Relative to the standard berkeley linux client, we have: Berkeley Reference 0% Athlon XP 32-bit +31% AMD64 (64-bit) +39% We don't have any data for a x86-64 seti client for AMD64 processor running on 32-bit linux. It might be the case that these extra benefits are only seen when running a 64 bit OS. Perhaps someone will be able to confirm this. Ned *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
TPR_Mojo Send message Joined: 18 Apr 00 Posts: 323 Credit: 7,001,052 RAC: 0 |
> OK, I've just spoken with Chris (and Triton_b5) and we have confirmed the > following: > > A true 64-bit seti client running on 64-bit linux is faster than a 32-bit seti > client running on the same OS. > > Relative to the standard berkeley linux client, we have: > > Berkeley Reference 0% > Athlon XP 32-bit +31% > AMD64 (64-bit) +39% > > We don't have any data for a x86-64 seti client for AMD64 processor running on > 32-bit linux. It might be the case that these extra benefits are only seen > when running a 64 bit OS. Perhaps someone will be able to confirm this. > > Ned > > Nice work Ned :) Are you sure that it is possible to run a 64-bit client process under a 32-bit operating system? Seems to me that something there is gonna not work - or the OS itself would strangle the client. > |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 |
> > OK, I've just spoken with Chris (and Triton_b5) and we have confirmed > the > > following: > > > > A true 64-bit seti client running on 64-bit linux is faster than a 32-bit > seti > > client running on the same OS. > > > > Relative to the standard berkeley linux client, we have: > > > > Berkeley Reference 0% > > Athlon XP 32-bit +31% > > AMD64 (64-bit) +39% > > > > We don't have any data for a x86-64 seti client for AMD64 processor > running on > > 32-bit linux. It might be the case that these extra benefits are only > seen > > when running a 64 bit OS. Perhaps someone will be able to confirm this. > > > > Ned > > > > > > Nice work Ned :) > Are you sure that it is possible to run a 64-bit client process under a 32-bit > operating system? Seems to me that something there is gonna not work - or the > OS itself would strangle the client. > > > Hi, I guess what I meant is a seti client compiled and run on a 32-bit version of linux but using -march=k8, opteron, athlon64, or athlon-fx (ie for the x86-64 instruction set). Not everyone with a 64-bit processor is running a 64-bit OS. Ned *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
spacemeat Send message Joined: 4 Oct 99 Posts: 239 Credit: 8,425,288 RAC: 0 |
you got it faster with -march=athlon64? mine ended up 10x slower than with -mcpu=i686. this is with 64 bit gentoo |
Neil Walker Send message Joined: 23 May 99 Posts: 288 Credit: 18,101,056 RAC: 0 |
> you got it faster with -march=athlon64? mine ended up 10x slower than with > -mcpu=i686. this is with 64 bit gentoo -march=k8 is the recommended setting for Gentoo, afaik. Anyway, in 2 days I'll be setting up 64-bit Gentoo on an AMD64 4000 so I'll be able to experiment. ;) Be lucky Neil |
Neil Walker Send message Joined: 23 May 99 Posts: 288 Credit: 18,101,056 RAC: 0 |
> you got it faster with -march=athlon64? mine ended up 10x slower than with > -mcpu=i686. this is with 64 bit gentoo I just did a bit of Googling and came across this page. You might want to give it a try. ;) Be lucky Neil |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.