It's a programing language that aims to be more or less hardware platform independent from the programmers point of view, you write one program at it can be run on many different hardware platforms. Eg. Nvidia GPUs, AMD GPUs, Intel GPUs and normal CPUs should be able to run the same program. The main goal is to be able to use the computing power of many different hardware platforms to get the job done faster.
Currently FGRP is not at all affected by the 'Run CPU versions of applications for which GPU versions are available' setting, which is a bug we'll hope to fix in the next app version release.
BM
Should be fixed now.
BM
This might be a bit more trouble than its worth, but it would be nice if 'Run CPU versions of applications for which GPU versions are available' was 'Run CPU versions of applications for which GPU versions are available that your computer is able to run' or maybe different check boxes for each type (CPU/GPU/IntelGPU)
It looks like FGRP3s are available but I was not getting them due to the setting. With a 1GB card I can't run those on the GPU. Then with the S6casA down I was getting no tasks.
Are there still CPU BRP4s around? I wouldn't want them as I can run the GPU version but I would like FGRP3 because I can't do those otherwise.
Are there still CPU BRP4s around? I wouldn't want them as I can run the GPU version but I would like FGRP3 because I can't do those otherwise.
Nope no CPU BRP4 tasks around, they only go to Intel GPU, Android and Raspberry PI. So you should not receive them at all, the GPU version is called BRP4G and is a separate app.
So when I see OpenCL, I should probably read it...Thank you . .
Quote:
you write one program at it can be run on many different hardware platforms. Eg. Nvidia GPUs, AMD GPUs, Intel GPUs and normal CPUs should be able to run the same program.
I am seeing quite long calculation times on my main computer (http://einsteinathome.org/host/7964949/tasks&offset=0&show_names=1&state=0&appid=26) for FGRP3 tasks on CPU (I7-3770 @3.40GHz with HT on). It takes about 100000 seconds for each to complete, two of the CPU cores are reserved to feed the GPU's (Intel HD Graphics 4000 & Nvidia GTX 650 Ti, 1 MB) and six for CPU tasks.
I have also a laptop with I7-2820QM @2.30GHz without HT (running CPU tasks only) and it takes about 32000 seconds for each FGRP3 task to finish.
What could be the reason for such a difference in performance? You would expect the hyper threading to result max 100% time difference, not 200% seen now, especially with those clock speeds.
Both computers are running with Intel Turbo Boost on, so they run a bit faster than stock speed.
Am about ready to report one
)
Am about ready to report one that had 3 errors on it.
http://einsteinathome.org/workunit/184835781
CPU's seem more reliable.
And what does open CL mean?
Thank you.
RE: And what does open CL
)
OpenCL = Open Computing Language
It's a programing language that aims to be more or less hardware platform independent from the programmers point of view, you write one program at it can be run on many different hardware platforms. Eg. Nvidia GPUs, AMD GPUs, Intel GPUs and normal CPUs should be able to run the same program. The main goal is to be able to use the computing power of many different hardware platforms to get the job done faster.
RE: RE: Currently FGRP is
)
This might be a bit more trouble than its worth, but it would be nice if 'Run CPU versions of applications for which GPU versions are available' was 'Run CPU versions of applications for which GPU versions are available that your computer is able to run' or maybe different check boxes for each type (CPU/GPU/IntelGPU)
It looks like FGRP3s are available but I was not getting them due to the setting. With a 1GB card I can't run those on the GPU. Then with the S6casA down I was getting no tasks.
Are there still CPU BRP4s around? I wouldn't want them as I can run the GPU version but I would like FGRP3 because I can't do those otherwise.
RE: Are there still CPU
)
Nope no CPU BRP4 tasks around, they only go to Intel GPU, Android and Raspberry PI. So you should not receive them at all, the GPU version is called BRP4G and is a separate app.
So when I see OpenCL, I
)
So when I see OpenCL, I should probably read it...Thank you . .
RE: GW work will remain
)
Mirrors working again, turned GW work back on.
BM
BM
According to the online task
)
According to the online task list, the server has tried to give my machine (8749294) several new gravitational wave tasks. None have arrived:
[error] Can't parse workunit in scheduler reply: unexpected XML tag or syntax
[error] No close tag in scheduler reply
I am seeing quite long
)
I am seeing quite long calculation times on my main computer (http://einsteinathome.org/host/7964949/tasks&offset=0&show_names=1&state=0&appid=26) for FGRP3 tasks on CPU (I7-3770 @3.40GHz with HT on). It takes about 100000 seconds for each to complete, two of the CPU cores are reserved to feed the GPU's (Intel HD Graphics 4000 & Nvidia GTX 650 Ti, 1 MB) and six for CPU tasks.
I have also a laptop with I7-2820QM @2.30GHz without HT (running CPU tasks only) and it takes about 32000 seconds for each FGRP3 task to finish.
What could be the reason for such a difference in performance? You would expect the hyper threading to result max 100% time difference, not 200% seen now, especially with those clock speeds.
Both computers are running with Intel Turbo Boost on, so they run a bit faster than stock speed.
RE: According to the online
)
I am getting the same message, but on only one of my 3 machines. The units are all Windows Gamma-ray pulsar search #3 v1.11 (FGRPopencl-nvidia).
Shall I abort the completed workunits and see if I can download new ones?
Looks like we ran into a
)
Looks like we ran into a limit with the recent mirror configuration.
I'll try to fix it on the server side, don't abort any finished WUs.
BM
BM