Mikie, you know what's interesting? I'm running Ubuntu aswell... makes me wonder if it's less the processor that doesn't profit from SSE 2 than a prob with the OS... Bernd, Bikeman, do you think it's worth playing around with a live CD to check? I should have Debian, Gentoo and BackTrack lying around... plus FreeBSD but I'm not sure if that agrees with my laptop (ATI chipset).
Hmmm....I don't see how the Linux distro could have any effect on this. I would be very, very surprised if it had any measurable effect.
CU
H-B
Mikie, you know what's interesting? I'm running Ubuntu aswell... makes me wonder if it's less the processor that doesn't profit from SSE 2 than a prob with the OS... Bernd, Bikeman, do you think it's worth playing around with a live CD to check? I should have Debian, Gentoo and BackTrack lying around... plus FreeBSD but I'm not sure if that agrees with my laptop (ATI chipset).
I don't think that's it. Of my two machines that benefit most from the SSE2 app, one is Kubuntu Feisty Fawn, and the other is Ubuntu Hardy Heron.
I spent last 2 days installing Linux and understanding how to set it up in order
to run perfs test.
- My system is XP3800+ ; so SSE, SSE2 and SSE3 capable
- Choosen OS is Ubuntu 8.04 «Hardy Heron »
To be clear, i'm completely new to Linux world, except a few hours spent on Sun workstations more than 10 years ago...
[HS]
Damm, this distro don't have root account, so u need to sudo anything, not really user friendly as it was said to...
[HS/]
All went well, 4.49 seem to be a little bit faster than Windows 4.36. The 2 little red dots at the bottom right.
But that's not the point here.
Then I ran perfs-tests and that’s where it became weird.
Here are the results (no I didn’t mixed up the labels :D )
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_0 ### std. app
real 10m46.047s
user 10m39.444s
sys 0m3.272s
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_1 ### SSE app
real 10m56.846s
user 10m44.680s
sys 0m4.324s
./run.sh ./einstein_S5R3_4.491_i686-pc-linux-gnu_1 ### SSE2 app
real 10m38.326s
user 10m20.443s
sys 0m3.940s
So it was faster in non-SSE than in SSE ???
And SSE2 was only a little bit faster than non-SSE.
Then I restarted BOINC to test crunching 2 WU’s with SSE2 version to validate the results on a normal run.
And all WU’s errored-out lightening fast with 255 exit code. http://einsteinathome.org/task/97867646
The perf test was run in a separate folder so it didn’t mixed-up my files (I double checked with the archive I made prior to run perfs-tests.
I wanted to switch back to normal SSE version to verify, but the trashed WU’s made me reach daily quota.
The only info I found about “exit code 255� was in this thread: http://einsteinathome.org/node/193653I
But it shouldn’t be it for me, as I run non-64bit Ubuntu.
So I switched back crunching to XP, and I’ll investigated further tomorrow…. :D
PS : If anyone has a hint to spindown HDDs in Ubuntu, please PM me ;)
I tried built-in functions, hdparm, laptop-mode-tools package, looking in conf files... And not being able to make anygood...
And this is quite critical if i want to run Boinc under Ubuntu in the future, as this PC has MANY hdds ( 7 SCSI's, 4 IDE, 4 SATA... ). Designed as file storage server.
God created a few good looking guys.. and for the rest he put hairs on top..
Damm, this distro don't have root account, so u need to sudo anything, not
Actually, Ubuntu does have a root account. It's just that there's no password assigned to it as yet. If you need root privileges, enter:
sudo passwd root
and then enter a password when prompted. Then, whenever you're logged on with your normal user account, but temporarily need root privileges, just enter
The switcher decided your system supports SSE, then tried to start the SSE application. Read the execve(2) manual: 'man 2 execve' and 'info execve'
This function never returns if successful, and returns -1 on failure, setting the global variable 'errno' to a value indicating the cause of failure. Unfortunately, the E@H application does not give that information, but my best guess would be: 'file not found'. Check if the name of the SSE2 application is correect! Should be 'einstein_S5R3_4.49_i686-pc-linux-gnu_1' in the E@H project directory, probably '/var/lib/boinc/projects/einstein.phys.uwm.edu', if you installed boinc as a system service (aka daemon).
To prevent problems with misspelled names, you can do the application selection manually, by linking (see 'man 1 ln' and 'info ln') the one you want to test to 'einstein_S5R3_4.49_i686-pc-linux-gnu' (this is the switcher, so make a copy of it first).
The switcher decided your system supports SSE, then tried to start the SSE application. Read the execve(2) manual: 'man 2 execve' and 'info execve'
This function never returns if successful, and returns -1 on failure, setting the global variable 'errno' to a value indicating the cause of failure. Unfortunately, the E@H application does not give that information, but my best guess would be: 'file not found'. Check if the name of the SSE2 application is correect! Should be 'einstein_S5R3_4.49_i686-pc-linux-gnu_1' in the E@H project directory, probably '/var/lib/boinc/projects/einstein.phys.uwm.edu', if you installed boinc as a system service (aka daemon).
To prevent problems with misspelled names, you can do the application selection manually, by linking (see 'man 1 ln' and 'info ln') the one you want to test to 'einstein_S5R3_4.49_i686-pc-linux-gnu' (this is the switcher, so make a copy of it first).
Regards,
Waldi
Many thanks Waldi for your detailed help !
I'm pretty confident i didn't misspell the name as i copy/paste the name of the old SEE one to be sure. But maybe a leading or trailing space...
I suppose i can't access any more to the global variable 'errno' to check as i restarted since. But i'll do if it reoccur.
Everything went back to normal this morning as i just replace back to previous the SSE file (_1).
So i'll wait for the current WU's to finish and try again with the SSE2 one, triple checking the spell.
And by that time i will look further deep in all the resources you gave me, thanks again.
PS : and by the time i've be able to put hdd's in standby mode manualy, using sdpram.
God created a few good looking guys.. and for the rest he put hairs on top..
RE: Mikie, you know what's
)
Hmmm....I don't see how the Linux distro could have any effect on this. I would be very, very surprised if it had any measurable effect.
CU
H-B
RE: Mikie, you know what's
)
I don't think that's it. Of my two machines that benefit most from the SSE2 app, one is Kubuntu Feisty Fawn, and the other is Ubuntu Hardy Heron.
Okay then. Apparently SSE2
)
Okay then. Apparently SSE2 doesn't help much on Cores then... shame... can anyone confirm this?
RE: Okay then. Apparently
)
Actually, you might want to go ahead and run some production workunits with the SSE2 app. Even if it doesn't help, it at least won't hurt anything.
My results with the test
)
My results with the test unit:
OpenSuse 10.3 64Bit, AMD X2 5000+@3.22GHz
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_0 ### std. app
real 11m12.006s
user 11m6.754s
sys 0m2.108s
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_1 ### SSE app
real 7m19.363s
user 7m14.635s
sys 0m2.084s
./run.sh ./einstein_S5R3_4.491_i686-pc-linux-gnu_1 ### SSE2 app
real 6m53.435s
user 6m48.954s
sys 0m1.928s
OpenSuse 10.3 32Bit, AMD XP 3000+
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_0 ### std. app
real 20m55.525s
user 20m49.602s
sys 0m3.144s
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_1 ### SSE app
real 13m54.208s
user 13m48.868s
sys 0m2.804s
I can add a X2 6000+, if necessary, but I think there is not much difference to the X2 5000.
cu,
Michael
[edit] deleted wrong comment
Best result so far with SSE2
)
Best result so far with SSE2 app on a WU: 22,658.47 s
AMD Opteron 1210 1.8 GHz, SuSE Linux 10.3 32 bit
Tullio
Hi Dudes, I spent last 2
)
Hi Dudes,
I spent last 2 days installing Linux and understanding how to set it up in order
to run perfs test.
- My system is XP3800+ ; so SSE, SSE2 and SSE3 capable
- Choosen OS is Ubuntu 8.04 «Hardy Heron »
To be clear, i'm completely new to Linux world, except a few hours spent on Sun workstations more than 10 years ago...
[HS]
Damm, this distro don't have root account, so u need to sudo anything, not really user friendly as it was said to...
[HS/]
I've been able to crunch 2 Wu's with the 4.49 opt app in SSE (not SSE2) using the automated switcher (witch worked well as it used the _1 version for my SSE CPU)
http://einsteinathome.org/task/97806227
http://einsteinathome.org/task/97805502
All went well, 4.49 seem to be a little bit faster than Windows 4.36. The 2 little red dots at the bottom right.
But that's not the point here.
Then I ran perfs-tests and that’s where it became weird.
Here are the results (no I didn’t mixed up the labels :D )
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_0 ### std. app
real 10m46.047s
user 10m39.444s
sys 0m3.272s
./run.sh ./einstein_S5R3_4.49_i686-pc-linux-gnu_1 ### SSE app
real 10m56.846s
user 10m44.680s
sys 0m4.324s
./run.sh ./einstein_S5R3_4.491_i686-pc-linux-gnu_1 ### SSE2 app
real 10m38.326s
user 10m20.443s
sys 0m3.940s
So it was faster in non-SSE than in SSE ???
And SSE2 was only a little bit faster than non-SSE.
Then I restarted BOINC to test crunching 2 WU’s with SSE2 version to validate the results on a normal run.
And all WU’s errored-out lightening fast with 255 exit code.
http://einsteinathome.org/task/97867646
The perf test was run in a separate folder so it didn’t mixed-up my files (I double checked with the archive I made prior to run perfs-tests.
I wanted to switch back to normal SSE version to verify, but the trashed WU’s made me reach daily quota.
The only info I found about “exit code 255� was in this thread:
http://einsteinathome.org/node/193653I
But it shouldn’t be it for me, as I run non-64bit Ubuntu.
So I switched back crunching to XP, and I’ll investigated further tomorrow…. :D
PS : If anyone has a hint to spindown HDDs in Ubuntu, please PM me ;)
I tried built-in functions, hdparm, laptop-mode-tools package, looking in conf files... And not being able to make anygood...
And this is quite critical if i want to run Boinc under Ubuntu in the future, as this PC has MANY hdds ( 7 SCSI's, 4 IDE, 4 SATA... ). Designed as file storage server.
God created a few good looking guys.. and for the rest he put hairs on top..
RE: Hi Dudes, Damm, this
)
Actually, Ubuntu does have a root account. It's just that there's no password assigned to it as yet. If you need root privileges, enter:
sudo passwd root
and then enter a password when prompted. Then, whenever you're logged on with your normal user account, but temporarily need root privileges, just enter
su - root
and enter the password.
RE: And all WU’s
)
- Detected CPU type 1
- execv returned: -1
The switcher decided your system supports SSE, then tried to start the SSE application. Read the execve(2) manual: 'man 2 execve' and 'info execve'
This function never returns if successful, and returns -1 on failure, setting the global variable 'errno' to a value indicating the cause of failure. Unfortunately, the E@H application does not give that information, but my best guess would be: 'file not found'. Check if the name of the SSE2 application is correect! Should be 'einstein_S5R3_4.49_i686-pc-linux-gnu_1' in the E@H project directory, probably '/var/lib/boinc/projects/einstein.phys.uwm.edu', if you installed boinc as a system service (aka daemon).
To prevent problems with misspelled names, you can do the application selection manually, by linking (see 'man 1 ln' and 'info ln') the one you want to test to 'einstein_S5R3_4.49_i686-pc-linux-gnu' (this is the switcher, so make a copy of it first).
Regards,
Waldi
Thank you Donald for the tip
)
Thank you Donald for the tip !
Many thanks Waldi for your detailed help !
I'm pretty confident i didn't misspell the name as i copy/paste the name of the old SEE one to be sure. But maybe a leading or trailing space...
I suppose i can't access any more to the global variable 'errno' to check as i restarted since. But i'll do if it reoccur.
Everything went back to normal this morning as i just replace back to previous the SSE file (_1).
So i'll wait for the current WU's to finish and try again with the SSE2 one, triple checking the spell.
And by that time i will look further deep in all the resources you gave me, thanks again.
PS : and by the time i've be able to put hdd's in standby mode manualy, using sdpram.
God created a few good looking guys.. and for the rest he put hairs on top..