Initial public release - SSE2 Linux optimized Seti@Home client (5.15)

[Simon]

Hi everyone,

a lot of people have been waiting for the first one to step forward and offer new optimized clients for PCs. So I'm going to take that leap and do it in the hope that many others may follow.

This client has been compiled using Intel's excellent compiler and library package (Links in the Links section, also in the How-To). It's version 5.15 and based on Eric Korpela's source tarballs.

It has been tested on a lot of different hardware - various Pentium 4, Pentium-D, Pentium-M as well as Athlon64-based machines. The results it produces have validated 100% in all tests.

Still, if you decide to use this client, it's at your own risk. Despite all testing that has gone into making sure it performs valid science and does so quickly, something might go wrong. YMMV.

That said, here's the download link:

Linux SSE2-optimized Seti@Home client

Please post whether it worked for you or not!
You will have to register before you can post.

Installation is very simple - when you unpack the archive, it will put all its files into a directory called something like "setiathome-5.15-sse2". Copy the contents of that folder (all files inside it, not the directory!) into BOINC/projects/setiathome.berkeley.edu. Then simply restart (stop and start) your BOINC client, it should recognize the app_info.xml and resume all your WUs for you.

Regards,
Simon

[pgy]

Hi Simon,

Congratulations on getting this far. Sadly, I can't use it. Any chance of sorting out plain SSE for Linux?

Best wishes,
pete

[Simon]

All I can say is, some more patience required

Have to test things before I release, and my SSE builds don't seem to work everywhere. I'm sure I'll fix it and offer MMX/SSE as well.

Regards,
Simon.

[Simon]

There was a slight problem with the first client - it did not run on all machines.
A new version that should now really work everywhere has been posted (the URL is still the same).

If it worked for you before, no need to get v2. If it didn't, please download the new package

Regards,
Simon.

[Simon]

Due to continued error reports, I've put a non-UPX compressed version online in the Downloads section -

Uncompressed Linux SSE2-optimized Seti@Home client

This is a direct download link.

Regards,
Simon

[s52d]

Hello!

Forum: black on black and very thin white lines,  but I managed to register.

Huh, uncompressed works on P4 no sse3 box, as previously reported.
Thanks a lot!
Maybe I overlooked some option in kernel?

Then laptop: segfault again.

cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 9
model name      : Intel(R) Celeron(R) M processor         1500MHz
stepping        : 5
cpu MHz         : 1498.751
cache size      : 512 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr mce cx8 apic sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 tm pbe
bogomips        : 2988.44


BR
Iztok

[Simon]

Hm, that's interesting.
What browser are you using again? All flavours of FireFox work fine (and do respect the CSS - all fonts are light so you can read them..).

Hans Dorn tested this executable on a Pentium-M system and they worked fine for him - also, these work fine on my Athlon64 SSE2 system as well, so I don't really know.

As to you forgetting something in your kernel, not sure - shouldn't depend on anything external since it's a static binary anyway.

I don't have a test system with a P-M here so it's kind of hard to see what's what, but I'll try and see what I can do.

Regards,
Simon

[s52d]

Hi!

Xeon: works on new, fails on old. (/proc/cpuinfo on seti forums).
Maybe... Failed boxes run 2.4 kernels, OK run 2.6.

BR
iztok

one more pentium-M (Celeron before)

tar -xzvf ../setiathome-5.15-sse2-v2-uncompressed.tar.gz 
setiathome-5.15-sse2/
setiathome-5.15-sse2/app_info.xml
setiathome-5.15-sse2/setiathome-5.15-sse2
setiathome-5.15-sse2/setiathome-5.15-sse2-AUTHORS
setiathome-5.15-sse2/setiathome-5.15-sse2-COPYING
setiathome-5.15-sse2/setiathome-5.15-sse2-COPYRIGHT
setiathome-5.15-sse2/setiathome-5.15-sse2-EXTRA_NOTE
setiathome-5.15-sse2/setiathome-5.15-sse2-README
setiathome-5.15-sse2/setiathome-5.15-sse2-README_APP
iztok@mlincek:~/ulov$ cd setiathome-5.15-sse2/
iztok@mlincek:~/ulov/setiathome-5.15-sse2$ ./setiathome-5.15-sse2 
Segmentation fault

iztok@mlincek:~/ulov/setiathome-5.15-sse2$ file ./setiathome-5.15-sse2
./setiathome-5.15-sse2: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked, stripped

iztok@mlincek:~/ulov/setiathome-5.15-sse2$ ldd ./setiathome-5.15-sse2
        not a dynamic executable

iztok@mlincek:~/ulov/setiathome-5.15-sse2$ uname -a
Linux mlincek 2.4.29 #2 Tue Feb 1 18:23:03 CET 2005 i686 unknown unknown GNU/Linux

iztok@mlincek:~/ulov/setiathome-5.15-sse2$ cat /proc/cpuinfo 
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 13
model name      : Intel(R) Pentium(R) M processor 1.80GHz
stepping        : 6
cpu MHz         : 1796.547
cache size      : 64 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr mce cx8 sep mtrr pge mca cmov pat clflush dts acpi mmx
                  fxsr sse sse2 ss tm pbe tm2 est
bogomips        : 3578.26

forum: ah, firefox 1.0.1, big CRT, bad eyes on my side.  Very thin white lines, do not bother.

--edited for better readability, this post and next (Simon)

[Simon]

Ah, that could be.

All my compilation and test boxes use 2.6 as kernel version.
I will experiment with compiling on 2.4 to see if that will make it more compatible or whether performance changes.

Regards,
Simon.

[Simon]

I've recompiled the clients on Linux 2.4.

I installed a 2.4 kernel and booted it, then tried to execute my binaries - it gave me a memory access error. So I made new ones under 2.4 and tested them there and under 2.6. The speed is the same, the results are okay, but the 2.4 works on 2.4 AND 2.6. So I'm switching the downloads to the 2.4-made binaries, which should now finally work for everyone (I can hope, anyway ).

The download links have been updated, the URLs are the same.

Regards,
Simon.

[s52d]

Thanks!

Very good news first:

343217099   82218989   17 Jun 2006 11:59:38 UTC   26 Jun 2006 12:57:49 UTC   Over   Success   Done   10,045.96   58.92   58.92
343217064   82218967   17 Jun 2006 11:59:38 UTC   23 Jun 2006 11:13:17 UTC   Over   Success   Done   23,267.57   58.87   58.87

(different AR, but impresive anyhow!)

And, normal good news:

works on celeron M (laptop) with 2.4 kernel.
While /usr/bin/file complains on header, it runs fine (2.4-upx). Cool, no need for uncompressed.
works on old Xeon, works on real Pentium-M laptop.

In fact, works on all my linux boxes with sse2, intel/amd CPUs.

Thanks for fast response, time for beer!

73 Iztok

[Simon]

Woo hoo

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.

Oh, and I was watching whether anyone reported problems So apart from some small compatibility issues that seem ironed out, it seems it's all good now. Phew!

Regards,
Simon.

[JavaPersona]

Simon,

Ni! Ni! Ni!

As a fellow KWSN and avid cruncher I congratulate you on the Linux app!  I am using it on 3 Linux hosts, using 2.6 kernels on FC4.  They are 2 P4 2.8 (HT'd) and a dual Xeon 3.06 (not HT'd).  All tasks have successfully completed and validated.

I may look into compiling my own Windows version; I really appreciate the instructions for that.  Thank you very much for your efforts.  Good Luck in the future! 

JavaPersona

[Metod, S56RKO]

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.

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

[s52d]

How about SSE3?

I looked at some fastest boxes on seti, and most of them (linux) are using
SSE3 Crunchr3 version.

Maybe it is worth trying?

BR 73
Iztok