Bug - Einstein@Home GPU apps continue during CPU benchmarks

Jacob Klein
Jacob Klein
Joined: 22 Jun 11
Posts: 45
Credit: 114028547
RAC: 0
Topic 196980

Einstein@Home Developers:

The GPU applications (v1.33, and maybe also v1.34) appear to have a bug.
When the user chooses to Run CPU benchmarks (Advanced View -> Advanced -> Run CPU benchmarks)
... the Einstein@Home GPU applications continue to run, interfering with the benchmarks!

Details from 2 emails are below.
Could you please fix it, so the GPU applications properly suspend work, like all other applications do?

Let me know what you find!
Thanks,
Jacob Klein

> From:
> To: charlief@...keley.edu; davea@...keley.edu
> Date: Sat, 1 Jun 2013 08:49:26 -0400
> CC: rwalton@...keley.edu; djangofett@...net; boinc_alpha@...keley.edu
> Subject: Re: [boinc_alpha] Run Benchmarks in Client/Manager 7.0.65
>
> I did some testing on this, since I'm attached to several GPU and CPU projects.
> I tested with BOINC v7.0.3 x64 Alpha.
>
> Basically, by inspection of GPU Load/Usage using eVGA Precision-X, as well as CPU Usage in Windows Task Manager,
> I have confirmed that the Einstein@Home GPU apps:
> - "Binary Radio Pulsar Search (Arecibo) 1.33"
> - "Binary Radio Pulsar Search (Perseus Arm Survey) 1.33"
> ... both continue to run during CPU benchmarking, utilizing the full GPU and whatever CPU they need.
>
> Other notes:
> - All of my *other* projects and apps (that I have tasks for) appear to be correctly suspending tasks during the benchmark.
> - Einstein@Home's CPU app "Gamma-ray pulsar search #2 1.04" does properly suspend.
> - All of my projects and apps suspend properly when I choose Activity -> Suspend (or Tray -> Snooze), including the Einstein GPU apps.
>
> So...
> It sounds like the Einstein GPU apps are bugged.
> I'll check their forums to see if it's been reported, and if not, I'll try to remember to report it.
>
> Regards,
> Jacob
>
>
> > >>> On 31-May-2013 8:40 AM, Christian Beer wrote:
> > >>>> Hi Rom,
> > >>>>
> > >>>> I just noticed a strange behavior of the Manager. I paused all CPU
> > >>>> Projects and only have work for Einstein on GPU I then click Advanced ->
> > >>>> Run CPU Benchmarks Two things happen: all (not ready to report) Tasks get
> > >>>> the Status [Suspended - CPU Benchmarks in progress (0.2 CPUs + 1 NVIDIA
> > >>>> GPU)] but Einstein is not suspended (although it uses 0.2 CPU) the
> > >>>> progress indicator is also showing progress (as is top). This may be not
> > >>>> a problem with Einstein but there are other projects that use 1 CPU for
> > >>>> each GPU and this will interfere with Benchmark results I guess.
> > >>>>
> > >>>> I'm runing 7.0.65 on Debian.
> > >>>>
> > >>>> Regards Christian

Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5893653
RAC: 1

Bug - Einstein@Home GPU apps continue during CPU benchmarks

Yup

Quote:
6/1/2013 5:18:44 PM | Einstein@Home | [cpu_sched] Preempting p2030.20121027.G201.54+00.94.N.b2s0g0.00000_2384_2 (left in memory)
6/1/2013 5:18:44 PM | Einstein@Home | [task] task_state=SUSPENDED for p2030.20121027.G201.54+00.94.N.b2s0g0.00000_2384_2 from suspend

6/1/2013 5:18:46 PM | | [benchmark] Starting floating-point benchmark
6/1/2013 5:18:56 PM | | [benchmark] Ended floating-point benchmark
6/1/2013 5:19:01 PM | | [benchmark] Starting integer benchmark
6/1/2013 5:19:11 PM | | [benchmark] Ended integer benchmark
6/1/2013 5:19:12 PM | Einstein@Home | [checkpoint] result p2030.20121027.G201.54+00.94.N.b2s0g0.00000_2384_2 checkpointed
6/1/2013 5:19:14 PM | | [benchmark] Ended benchmark
6/1/2013 5:19:15 PM | | [benchmark] CPU 0 has finished
6/1/2013 5:19:15 PM | | [benchmark] CPU 1 has finished
6/1/2013 5:19:15 PM | | [benchmark] CPU 2 has finished
6/1/2013 5:19:15 PM | | [benchmark] 3 out of 3 CPUs done
6/1/2013 5:19:15 PM | | [benchmark] CPU 0: fp 3337381943.929339 int 10756316114.544062 intloops 182496000.000000 inttime 9.656462
6/1/2013 5:19:15 PM | | [benchmark] CPU 1: fp 3381767136.865214 int 0.000000 intloops 0.000000 inttime 0.000000
6/1/2013 5:19:15 PM | | [benchmark] CPU 2: fp 3370000658.129181 int 0.000000 intloops 0.000000 inttime 0.000000
6/1/2013 5:19:15 PM | | Benchmark results:
6/1/2013 5:19:15 PM | | Number of CPUs: 3
6/1/2013 5:19:15 PM | | 3363 floating point MIPS (Whetstone) per CPU
6/1/2013 5:19:15 PM | | 10756 integer MIPS (Dhrystone) per CPU
6/1/2013 5:19:15 PM | | [dcf] scaling all duration correction factors by 0.988825
Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 779204697
RAC: 1184622

Hi! Thanks for reporting

Hi!

Thanks for reporting this. We'll look into this.

HBE

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 987
Credit: 25171438
RAC: 0

We identified an issue in the

We identified an issue in the BOINC API code and proposed a fix that's currently being tested upstream. We'll provide updated binaries as soon as the fix got finalised...

Cheers,
Oliver

Einstein@Home Project

Jacob Klein
Jacob Klein
Joined: 22 Jun 11
Posts: 45
Credit: 114028547
RAC: 0

That's awesome news! Thank

That's awesome news! Thank you for the prompt response!

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 987
Credit: 25171438
RAC: 0

All GPU apps >= v1.36 (BRP5)

All GPU apps >= v1.36 (BRP5) will contain the upstream fix for this issue. They're currently being tested on albert...

Stay tuned,
Oliver

Einstein@Home Project

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 987
Credit: 25171438
RAC: 0

FWIW, BRP5 v1.36 now also

FWIW, BRP5 v1.36 now also available here on einstein. Give it a try...

Cheers,
Oliver

Einstein@Home Project

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.