Binary Radio Pulsar Search (Arecibo) "BRP4" - unbundled
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250498964
RAC: 34611
17 Jun 2013 13:11:48 UTC
Topic 197003
(moderation:
As BRR4 these days only runs on CPUs (and possibly slow ones), we have decided to switch back from "bundled" to "single-DM" Wus (one WU per dispersion measure).
This means that the BRP4 tasks generated from now on will run 1/8 of before, takes 1/8 of disk space and give 1/8 of the previous credit.
So overall nothing should change for you out there.
BM
BM
Binary Radio Pulsar Search (Arecibo) "BRP4" - unbundled
)
Why does BRP4 only run on CPUs now? It seems like it would be better to have the pulsar search applications run on GPUs as well so that CPUs can be used for gravitational wave searches, which are supposed to be the main focus of Einstein@home.
RE: Why does BRP4 only run
)
There are a number of reasons.
- With the radio pulsar search on Perseus Arm Survey data we have plenty of work to do for our GPUs that can't be done on CPUs.
- We have GPU apps for the Gamma-Ray search being tested, too
- The current analysis run for Gravitational Waves is ending soon, and the next one will take a while until we have it running on Einstein (see linked thread).
- Due to limitations in the infrastructure the Gamma-Ray pulsar search could not feed more than half of the project (CPUs) (until yesterday when we introduced longer running workunits).
- The Arecibo data search is the only search suitable for "small devices" (like Raspberry Pi's or Android phones and tablets) that have rather slow CPUs.
BM
BM
Today I've suddenly started
)
Today I've suddenly started receiving BRP4 (Arecibo) tasks for GPU again.
They are estimated to be 19.07MB [/code]
http://einsteinathome.org/task/387281946
http://einsteinathome.org/task/387125178
Looks like they were created today.
RE: Today I've suddenly
)
Thanks for the note. Nope, this was not intentional. A BOINC PHP script went berserk and caused some chaos in the database.
Should be fixed for now.
BM
BM
I'm recieveing GPU tasks too.
)
I'm recieveing GPU tasks too. But every second task has been finished with error so far. I'm getting some shorter tasks - around 4 minutes of work, but i get normal (ussual) tasks too.
Hi! Same with me! Suddenly
)
Hi!
Same with me! Suddenly again received around 50 'Arecibo' GPU WUs.
RUN time: ONLY 2.5 - 6.5 min; till now I've completed 10 of them, with half of them (the shorter ones - ?!?) finishing with 'Error while computing' with the message:
EXAMPLE:
http://einsteinathome.org/task/387272224
There were EXCEPTIONS: 3 tasks I received were 'standard' Arecibo WUs, with RUN time around 50 min, as expected. The one I completed until now, finished without problems.
QUESTION:
What to do with the remaining 40+ 'Arecibo' GPU WUs or how to solve the problem?
Thank You.
Denis
P.S.
My settings are:
GPU utilization factor of 0.5 and boxes checked for receiving ONLY GPU tasks - 'Perseus...' AND 'Arecibo';
P.P.S.
By the way, totally different topic:
I asked this question few months ago with no answer, so sorry to bother again:
WHEN can we expect an update on POSSIBLE NEW findings of pulsars with the BRP4 GPU - Arecibo project (since it is in its 'finishing' phase)?
If the answer or results were already posted I apologize, I did not have a lot of time to read the forums e.g. news.
“A little knowledge is a dangerous thing. So is a lot.” - Albert EINSTEIN
RE: What to do with the
)
It was an error that (short) BRP4 tasks were sent with GPU apps. Luckily this was fixed soon enough that these very short running tasks don't pose a problem for the project as a whole.
If you want to, you may cancel/abort these tasks. If you do nothing, a couple will probably error out anyway, and for those that finish successfully you will get ~65 credit for 6.5 min of GPU work. Your choice.
Rest assured that you will get notified as soon as there is a confirmed discovery of a new pulsar. The region of the sky the currently analyzed data comes from is named "outer galaxy". IIRC the density of known pulsars in that region is about 1/20 of that in the "inner galaxy", so I would estimate a discovery rate ratio of the same order: As we did discover a new pulsar in Arecibo data about every two weeks on average when searching the "inner galaxy", in the "outer galaxy" I would expect a pulsar discovery roughly once every year (assuming the same processing rate).
FWIW new data from Arecibo has been announced to us, but is currently stuck at Cornell due to technical problems. So this search will continue even when the data that has so far been transferred to us is finished.
BM
Edit: Nevertheless there might be an update pending to the list of "re-discoveries", known pulsars that were detected also by the Einstein@Home pipeline. I'll check that tomorrow.
BM
Hi! Thank You for the
)
Hi!
Thank You for the exact explanation.
It would really 'boost' my motivation, if I'd be one of the lucky ones to actually 'discover' a NEW pulsar, because for some time (hope not too long) I won't be able to do as much work (if any?) as a did for quite a LONG time now on a single, pretty average PC.
Nevertheless, I'm really glad, that I'm a small part of this great joint effort. Finding a pulsar would be a really great 'present', something, I'd be very proud of (I don't care, if it sounds silly...) but that is all.
Keep up the good work and I hope, some really important questions will eventually be answered.
Regards.
Denis
“A little knowledge is a dangerous thing. So is a lot.” - Albert EINSTEIN
I'm getting WU's that
)
I'm getting WU's that generate this kind of messages:
"29.6.2013 13:06:06 | Einstein@Home | Task PA0071_003C1_306_3 exited with zero status but no 'finished' file
29.6.2013 13:06:06 | Einstein@Home | If this happens repeatedly you may need to reset the project.
29.6.2013 13:06:06 | Einstein@Home | Restarting task PA0071_003C1_306_3 using einsteinbinary_BRP5 version 136 (BRP4cuda32nv301) in slot 0
"
I have reset it already but to no avail, so I came here and seems it must be because of this GPU-CPU mixup.
It takes about 2 minutes to get to 0,001% and then it starts all over :D
Now if there are many more of these WU's to come and I have to cancel them all manually, it's going to be a long and frustrating weekend :D
okay, now I'm getting CPU
)
okay, now I'm getting CPU WU's and they seem to be running, although one of them jumped from 0,000% straight to 5,000% just like that, while the others are increasing steadily.