I set the GPU App to require only half a core. Feedback is welcome.
Regarding the overshoot of tasks, this is an issue with BOINC's runtime estimation, which is difficult to handle. If you're not running any other App of this project on the machine, it should adjust itself after running some dozens of these tasks. I'll see what I can do to mitigate that on the server side, too. Regarding server side management these processors' behavior is pretty new, and the app is also still under development.
the app instantly terminated on my m1 with an error
Application:Binary Radio Pulsar Search (Arecibo,GBT,A) v2.08 (Apple_M-opencl)
arm64-apple-darwin
Quote:
Stderr output
<core_client_version>7.22.2</core_client_version>
<![CDATA[
<message>
process exited with code 6 (0x6, -250)</message>
<stderr_txt>
[17:58:40][13531][INFO ] Application startup - thank you for supporting Einstein@Home!
[17:58:40][13531][INFO ] Starting data processing...
[17:58:40][13531][INFO ] Using Metal device "Apple M1"
[17:58:40][13531][INFO ] Checkpoint file unavailable: p2030.1711377073.G60.20-03.65.C.b2s0g0.00000_3481.cpt (No such file or directory).
------> Starting from scratch...
[17:58:40][13531][INFO ] Header contents:
------> Original WAPP file: ./p2030.20160623.G60.20-03.65.C.b2s0g0.00000_DM734.10
------> Sample time in microseconds: 65.4762
------> Observation time in seconds: 274.62705
------> Time stamp (MJD): 57562.305947446097
------> Number of samples/record: 0
------> Center freq in MHz: 1214.289551
------> Channel band in MHz: 0.336182022
------> Number of channels/record: 960
------> Nifs: 1
------> RA (J2000): 195732.3456
------> DEC (J2000): 221420.4799
------> Galactic l: 0
------> Galactic b: 0
------> Name: G60.20-03.65.C
------> Lagformat: 0
------> Sum: 1
------> Level: 3
------> AZ at start: 0
------> ZA at start: 0
------> AST at start: 0
------> LST at start: 0
------> Project ID: --
------> Observers: --
------> File size (bytes): 0
------> Data size (bytes): 0
------> Number of samples: 4194304
------> Trial dispersion measure: 734.1 cm^-3 pc
------> Scale factor: 0.000969215
[17:58:40][13531][INFO ] Seed for random number generator is 1161917594.
-[MTLComputePipelineDescriptorInternal setComputeFunction:withType:]:722: failed assertion `computeFunction must not be nil.'
Guys, we think we found the culprit causing the signal 6 errors on Sonoma, finally!
The details can be found here but it boils down to you having to install BOINC 7.24.3 to fix the permission issue. If a simple update doesn't help you might want to uninstall and fully reinstall BOINC to make sure.
The good thing is that investigating the issue also allowed us to improve the numerical stability of our app such that it should validate much better :)
We're going to release a new version (2.08) without the debugging stuff soon.
Thanks for your patience! This one took a while to figure out because we couldn't reproduce the error in our labs.
Cheers,
Oliver
That looks like the problem Oliver mentioned here. If you didn't install BOINC Client 7.24.3 recently, please do so.
Thanks for the prompt response!
I have ran the update, will report if GPU task is successful once i get some since currently Einstein did not assign any
Thanks for the prompt response!
I have ran the update, will report if GPU task is successful once i get some since currently Einstein did not assign any
follow up: after the update apps now come in and seem to compute properly at least for Arecibo GBT,A
Thanks for the solution. I'm seeing some tasks running without error so far :) Will report back once they reach validation.
EDIT: One thing I am curious about (possibly a tangent), I didn't see a prompt in the BOINC client that there was a newer 7.x version to update to, like normal. I suspect others who have this hardware might also not see an update and still be on a 7.22.x version, like I was. Not sure if anything can be done on the project side to help folks update to latest...
Same with me. Using the “check for update” returned no update. However I checked the boinc website anyway and saw there was an update. This was a couple of weeks ago. Unfortunately I didn’t realize the implications, otherwise I could have helped people here. Disappointing.
So now that everything works, can we get a readout on the performance of the Mac GPUs? I am running the original M1 Mac mini, and my tasks take about ~1400 seconds. How would that compare, theoretically to other GPUs? And what about other apple GPU like Mac/pro/ultra? M1 vs m2 vs m3?
Same with me. Using the “check for update” returned no update. However I checked the boinc website anyway and saw there was an update. This was a couple of weeks ago. Unfortunately I didn’t realize the implications, otherwise I could have helped people here. Disappointing.
So now that everything works, can we get a readout on the performance of the Mac GPUs? I am running the original M1 Mac mini, and my tasks take about ~1400 seconds. How would that compare, theoretically to other GPUs? And what about other apple GPU like Mac/pro/ultra? M1 vs m2 vs m3?
As noted above the current GPU version isn't yet fully done. It's our first proof-of-concept native Metal app which still has the speed potential of a couple of factors more. Feel free, though, to start a thread in the Cruncher's Corner to gather timings from your fellow volunteers. Please keep in mind that it's not straight forward to compare those figures meaningfully as they'll depend not only on the hardware used, but also on the app version as well the runtime environment, e.g. the GPU and CPU load as well as the (unified!) memory consumption of everything else running on the system.
If you look back closely the error didn't change. We just augmented the output which gave us the needed clue.
Please note that the app isn't yet at its full potential since we're not done yet with porting all of the data analysis code. So stay tuned for more :)
Bernd Machenschalk wrote: I
)
the app instantly terminated on my m1 with an error
Oliver Behnke wrote: Guys,
)
That looks like the problem Oliver mentioned here. If you didn't install BOINC Client 7.24.3 recently, please do so.
BM
Thanks for the prompt
)
Thanks for the prompt response!
I have ran the update, will report if GPU task is successful once i get some since currently Einstein did not assign any
Marcin wrote: Thanks for the
)
follow up: after the update apps now come in and seem to compute properly at least for Arecibo GBT,A
Thanks for the solution. I'm
)
Thanks for the solution. I'm seeing some tasks running without error so far :) Will report back once they reach validation.
EDIT: One thing I am curious about (possibly a tangent), I didn't see a prompt in the BOINC client that there was a newer 7.x version to update to, like normal. I suspect others who have this hardware might also not see an update and still be on a 7.22.x version, like I was. Not sure if anything can be done on the project side to help folks update to latest...
Same with me. Using the
)
Same with me. Using the “check for update” returned no update. However I checked the boinc website anyway and saw there was an update. This was a couple of weeks ago. Unfortunately I didn’t realize the implications, otherwise I could have helped people here. Disappointing.
So now that everything works,
)
So now that everything works, can we get a readout on the performance of the Mac GPUs? I am running the original M1 Mac mini, and my tasks take about ~1400 seconds. How would that compare, theoretically to other GPUs? And what about other apple GPU like Mac/pro/ultra? M1 vs m2 vs m3?
Reno, NV Team: SETI.USA
jd wrote: Same with me.
)
There's a bug report upstream for this issue.
Einstein@Home Project
zombie67 wrote: So now that
)
As noted above the current GPU version isn't yet fully done. It's our first proof-of-concept native Metal app which still has the speed potential of a couple of factors more. Feel free, though, to start a thread in the Cruncher's Corner to gather timings from your fellow volunteers. Please keep in mind that it's not straight forward to compare those figures meaningfully as they'll depend not only on the hardware used, but also on the app version as well the runtime environment, e.g. the GPU and CPU load as well as the (unified!) memory consumption of everything else running on the system.
Oliver
Einstein@Home Project
Oliver Behnke wrote: If you
)
Any new news on this optimization effort?
Reno, NV Team: SETI.USA