Forum > Linux
Initial public release - SSE2 Linux optimized Seti@Home client (5.15)
Simon:
Since I now have an SSE3 box to test on, I'm going to test some more on it, but have not found *any* performance difference between SSE2 and 3. So, for maximum compatibility, and since everything else is just not tested enough, I went with SSE2.
It'll take a bit more time, but SSE and SSE3 and MMX are definitely in the works.
HTH,
Simon.
Simon:
--- Quote from: Metod, S56RKO on 30 Jun 2006, 06:02:24 am ---
--- Quote from: Simon on 26 Jun 2006, 03:10:34 pm ---Now if I can only figure out why my SSE builds only work on some Intel chips and not anywhere else...time to install the dev tools on yet another host.
--- End quote ---
If I remember right from the days I did my optimized seti (the old one) binary using ICC, there can be some older chips that don't take well binary compiled using too many -ax branches. They crashed miserably on my P3 Tualatin. The solution was to compile binary specifficaly for that using -xK ...
BR,
Metod
--- End quote ---
Thanks Metod :)
All my build scripts use -xW -xK -xP etc. I've found that compiling with -axK etc. leads to around 1-3% slower times and does not really do that much for compatibility anyway. I'd rather build an extra specialized client than lose performance, but that's just me.
I could also use -xK -axK which actually does different things than just -axK, but I haven't tried that yet.
My newest SSE build runs fine on an AMD Duron 1.2GHz, so things are looking good in that regard. It's not as much faster than the default as my SSE2 builds, but I'll put that down to my Duron having only 64k cache (normal P3s have 256+64, AFAIK).
It is, however, faster than Crunch3r's SSE build on my Duron, so that's in line with my previous Linux builds ;D
Regards,
Simon.
Metod, S56RKO:
--- Quote from: Simon on 30 Jun 2006, 06:42:27 pm ---My newest SSE build runs fine on an AMD Duron 1.2GHz, so things are looking good in that regard. It's not as much faster than the default as my SSE2 builds, but I'll put that down to my Duron having only 64k cache (normal P3s have 256+64, AFAIK).
It is, however, faster than Crunch3r's SSE build on my Duron, so that's in line with my previous Linux builds ;D
--- End quote ---
If you need a beta-tester ... I'd love to run your SSE-only binaries on my dual P3 Coppermine and P3 Tualatin.
Simon:
Okay Metod - you've been bumped :)
You should see a new forum and downloads section now.
Regards,
Simon.
Dirk Sadowski:
Hello Simon,
now I use a Boinc- client 64 Bit and your 32 Bit SSE2 for LINUX- 64 Bit... :)
Why you didnīt told me, that your LINUX SSE2 app is faster than Crunch3rs Windows SSE2 (On AMD K8) ? ;)
Because I saw an other AMD K8 3200+, he use your app... and I looked to the performance on our PCs... :)
Thanks a lot again for your good work!
I have a small question:
...again... ;)
Is there a command like in the Windows Version:
"report all finished WUs, get the current statistics, and get maybe new work"
So all finished WUs will be reported and I can shut down my PC...
Greetings!
BTW. Whatīs the different between your uncompressed and UPX- compressed apps?
You know, Iīm a very new LINUX user... ;)
There are no different performance?
Only the installation is an other way, or?
Maybe like with uncompressed and WINZIP- compressed in Windows?
--------------------------------------------------
Update infos from your side:
Along with the automatic test and installer package, I've updated the downloads to Release 1.3 of the science applications. Windows.
This update is a cosmetic fix - it inserts information about the true CPU speed into the stderr section of each result.
--------------------------------------------------
Can you insert this feature for the LINUX apps too?
BTW #2: ;)
Because you told me, that I can start the Boinc- client in the console:
./boinc -return_results_immediately
But this doesnīt work... :(
The uploaded WUs will not be reported immediately...
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version