I'm crunching tasks from BRP7/MeerKAT, but my workstation stopped receiving those tasks and started receiving tasks from the All-Sky Gravitational Wave search on O3. However, in the project preferences, I selected to receive only BRP7. Any idea what might be happening?
Make sure in your preferences that BOTH the compute and project preferences are on the same venue and try the comparison views of both and check that all items in one (whether work, home, school, etc.) are the same as the default venue.
Afterwards, in BOINC Manger (you are using the advanced view, aren't you?) do an Options >> Read Config Files. Then I usually do a relaunch of BOINC Manager, waiting ~1 minute to relaunch.
HTH
.....[EDIT].....
By the way, next time you need help with something, it might be advisable to place your post in the "Cruncher's Corner" forum.
Bernd, could you briefly try to explain the reason why the new low frequency search is running so much slower? most people are reporting that the new tasks are somewhere around 3x longer in runtime.
is it just the lower frequency is hard to parse? or is it mostly the setup parameters? i do see differences in how these tasks are setup in terms of the setup variables vs how they were on the HF tasks, many of them are scaled almost exactly 4x between the HF and Bu tasks. and I don't know if the number of segments is the correct metric to guage how much work the task is actually processing, but the new tasks are only processing 1/4 the segments that HF was, are each of the Bu segments just 4x larger?
are the new tasks doing a finer (more resolution) search than what HF was doing? is the freq range now wider than the 2Hz it was doing on HF?
i was curious about that since in the past the runtime would roughly be in line with VRAM use (hence when you changed the app from 1x pass with high VRAM 4GB+ to lower VRAM ~2GB and 2x passes, the app runtime basically doubled). so it's strange to see the VRAM use remain at basically the same ~2GB level, but also running in 1x pass and taking so much longer to run.
Thank you for the help GWGeorge007! Unfortunately, it’s still the same for me. :/ I tested it on another machine, but I’m still not receiving any Meerkat tasks. I apologize for posting here; I considered doing it there, but since I had already posted here and there was no option to delete it, I ended up leaving it here. I'll continue working on the All-Sky O3 tasks until this gets resolved. The All-Sky O3 tasks are indeed slower; as far as I remember, they used to be faster before.
Thank you for the help GWGeorge007! Unfortunately, it’s still the same for me. :/ I tested it on another machine, but I’m still not receiving any Meerkat tasks. I apologize for posting here; I considered doing it there, but since I had already posted here and there was no option to delete it, I ended up leaving it here. I'll continue working on the All-Sky O3 tasks until this gets resolved. The All-Sky O3 tasks are indeed slower; as far as I remember, they used to be faster before.
You may need to clear out your cache. Simply go to your BOINC Manager and expand it to full screen and then click on "Show all tasks". Then highlight all of the Einstein tasks that you have and then 'abort' them.
You should start getting BRP7/MeerKAT tasks then. I might also suggest the Options >> Read Config Files again to be sure.
Thank you for the help GWGeorge007! Unfortunately, it’s still the same for me. :/ I tested it on another machine, but I’m still not receiving any Meerkat tasks. I apologize for posting here; I considered doing it there, but since I had already posted here and there was no option to delete it, I ended up leaving it here. I'll continue working on the All-Sky O3 tasks until this gets resolved. The All-Sky O3 tasks are indeed slower; as far as I remember, they used to be faster before.
You may need to clear out your cache. Simply go to your BOINC Manager and expand it to full screen and then click on "Show all tasks". Then highlight all of the Einstein tasks that you have and then 'abort' them.
You should start getting BRP7/MeerKAT tasks then. I might also suggest the Options >> Read Config Files again to be sure.
HTH
Thank you very much, you solved my problem. I really appreciate the help <3
You may need to clear out your cache. Simply go to your BOINC Manager and expand it to full screen and then click on "Show all tasks". Then highlight all of the Einstein tasks that you have and then 'abort' them.
You should start getting BRP7/MeerKAT tasks then. I might also suggest the Options >> Read Config Files again to be sure.
HTH
Thank you very much, you solved my problem. I really appreciate the help <3
You're very welcome! Have a nice day crunching! :*)
I have many HF tasks left in my que (500+), but they are not registering as completed and valid, and no pending tasks either. I do have several 'errors' of HF tasks, but I only have Bu tasks showing completion with validation. The pending tasks are all showing Bu as well. What's up with that?
.....[EDIT].....
I have since resolved my dilemma with O3AS HF tasks with a Teammate's help.
Bernd, could you briefly try to explain the reason why the new low frequency search is running so much slower?
[...]
are the new tasks doing a finer (more resolution) search than what HF was doing? is the freq range now wider than the 2Hz it was doing on HF?
Basically yes. We are doing a finer (or, as we call it "deeper") search than before in the HF search. The frequency range of a single workunit is actually smaller (1Hz).
Basically yes. We are doing a finer (or, as we call it "deeper") search than before in the HF search. The frequency range of a single workunit is actually smaller (1Hz).
thanks.
would you quantify the new app to be performing about 4x the amount of computations as it was in HF? that would align with a lot of the input parameters being scaled by about 4x.
The BOINC time estimates seemed to be messed up. I'm currently running these new GW tasks in parallel with BRP7 tasks and BOINC seems to think the BRP7 ones take more than twice as long but actually the GW tasks take 8 - 9 times longer than BRP7. I have the impression that the slower than expected performance of the GW tasks is massively pushing the estimated time for the BRP7 tasks up (from 36 minutes to over 10 hours) and the faster the expected performance of BRP7 is pushing down the estimated time for the GW tasks.
As a result I wind up with too many GW tasks and not enough BRP7 being fetched. I normally only have a maximum of 1 day's worth of tasks but I'm getting multiple days of GW, perhaps more than can be done within the 7 day time limit.
FG_Madara wrote:I'm
)
Make sure in your preferences that BOTH the compute and project preferences are on the same venue and try the comparison views of both and check that all items in one (whether work, home, school, etc.) are the same as the default venue.
Afterwards, in BOINC Manger (you are using the advanced view, aren't you?) do an Options >> Read Config Files. Then I usually do a relaunch of BOINC Manager, waiting ~1 minute to relaunch.
HTH
.....[EDIT].....
By the way, next time you need help with something, it might be advisable to place your post in the "Cruncher's Corner" forum.
Proud member of the Old Farts Association
Bernd, could you briefly try
)
Bernd, could you briefly try to explain the reason why the new low frequency search is running so much slower? most people are reporting that the new tasks are somewhere around 3x longer in runtime.
is it just the lower frequency is hard to parse? or is it mostly the setup parameters? i do see differences in how these tasks are setup in terms of the setup variables vs how they were on the HF tasks, many of them are scaled almost exactly 4x between the HF and Bu tasks. and I don't know if the number of segments is the correct metric to guage how much work the task is actually processing, but the new tasks are only processing 1/4 the segments that HF was, are each of the Bu segments just 4x larger?
are the new tasks doing a finer (more resolution) search than what HF was doing? is the freq range now wider than the 2Hz it was doing on HF?
i was curious about that since in the past the runtime would roughly be in line with VRAM use (hence when you changed the app from 1x pass with high VRAM 4GB+ to lower VRAM ~2GB and 2x passes, the app runtime basically doubled). so it's strange to see the VRAM use remain at basically the same ~2GB level, but also running in 1x pass and taking so much longer to run.
_________________________________________________________________________
Thank you for the help!
)
Thank you for the help GWGeorge007! Unfortunately, it’s still the same for me. :/ I tested it on another machine, but I’m still not receiving any Meerkat tasks. I apologize for posting here; I considered doing it there, but since I had already posted here and there was no option to delete it, I ended up leaving it here. I'll continue working on the All-Sky O3 tasks until this gets resolved. The All-Sky O3 tasks are indeed slower; as far as I remember, they used to be faster before.
FG_Madara wrote: Thank you
)
You may need to clear out your cache. Simply go to your BOINC Manager and expand it to full screen and then click on "Show all tasks". Then highlight all of the Einstein tasks that you have and then 'abort' them.
You should start getting BRP7/MeerKAT tasks then. I might also suggest the Options >> Read Config Files again to be sure.
HTH
Proud member of the Old Farts Association
GWGeorge007 wrote: FG_Madara
)
Thank you very much, you solved my problem. I really appreciate the help <3
FG_Madara wrote: GWGeorge007
)
You're very welcome! Have a nice day crunching! :*)
Proud member of the Old Farts Association
Bernd, I have a question
)
Bernd, I have a question regarding O3AS HF tasks.
I have many HF tasks left in my que (500+), but they are not registering as completed and valid, and no pending tasks either. I do have several 'errors' of HF tasks, but I only have Bu tasks showing completion with validation. The pending tasks are all showing Bu as well. What's up with that?
.....[EDIT].....
I have since resolved my dilemma with O3AS HF tasks with a Teammate's help.
Thanks anyway.
Proud member of the Old Farts Association
Ian&Steve C. wrote:Bernd,
)
Basically yes. We are doing a finer (or, as we call it "deeper") search than before in the HF search. The frequency range of a single workunit is actually smaller (1Hz).
BM
Bernd Machenschalk
)
thanks.
would you quantify the new app to be performing about 4x the amount of computations as it was in HF? that would align with a lot of the input parameters being scaled by about 4x.
_________________________________________________________________________
The BOINC time estimates
)
The BOINC time estimates seemed to be messed up. I'm currently running these new GW tasks in parallel with BRP7 tasks and BOINC seems to think the BRP7 ones take more than twice as long but actually the GW tasks take 8 - 9 times longer than BRP7. I have the impression that the slower than expected performance of the GW tasks is massively pushing the estimated time for the BRP7 tasks up (from 36 minutes to over 10 hours) and the faster the expected performance of BRP7 is pushing down the estimated time for the GW tasks.
As a result I wind up with too many GW tasks and not enough BRP7 being fetched. I normally only have a maximum of 1 day's worth of tasks but I'm getting multiple days of GW, perhaps more than can be done within the 7 day time limit.