This was primarily done to allow the developers to get work on the opencl_intel_100 plan class.

---

6/14/2013 2:29:32 AM | SETI@home Beta Test | Resetting project
6/14/2013 2:29:36 AM | SETI@home Beta Test | update requested by user
6/14/2013 2:29:37 AM | SETI@home Beta Test | [sched_op] Starting scheduler request
6/14/2013 2:29:37 AM | SETI@home Beta Test | [sched_op] Fetching master file
6/14/2013 2:29:40 AM | SETI@home Beta Test | [sched_op] Got master file; parsing
6/14/2013 2:29:40 AM | SETI@home Beta Test | [sched_op] Found 1 scheduler URLs in master file
6/14/2013 2:29:40 AM | SETI@home Beta Test | Master file download succeeded
6/14/2013 2:29:45 AM | SETI@home Beta Test | [sched_op] Starting scheduler request
6/14/2013 2:29:45 AM | SETI@home Beta Test | Sending scheduler request: Requested by user.
6/14/2013 2:29:45 AM | SETI@home Beta Test | Requesting new tasks for intel_gpu
6/14/2013 2:29:45 AM | SETI@home Beta Test | [sched_op] CPU work request: 0.00 seconds; 0.00 devices
6/14/2013 2:29:45 AM | SETI@home Beta Test | [sched_op] intel_gpu work request: 878623.34 seconds; 0.00 devices
6/14/2013 2:29:47 AM | SETI@home Beta Test | Scheduler request completed: got 0 new tasks
6/14/2013 2:29:47 AM | SETI@home Beta Test | [sched_op] Server version 701
6/14/2013 2:29:47 AM | SETI@home Beta Test | No tasks sent
6/14/2013 2:29:47 AM | SETI@home Beta Test | No tasks are available for AstroPulse v6
6/14/2013 2:29:47 AM | SETI@home Beta Test | No tasks are available for SETI@home v7 large workunits
6/14/2013 2:29:47 AM | SETI@home Beta Test | Project has no tasks available
6/14/2013 2:29:47 AM | SETI@home Beta Test | Project requested delay of 7 seconds
6/14/2013 2:29:47 AM | SETI@home Beta Test | [sched_op] Deferring communication for 7 sec
6/14/2013 2:29:47 AM | SETI@home Beta Test | [sched_op] Reason: requested by project

Here's your work request. The "no work available is weird because there are about 1000 AP results available. While it's recognizing that you are have a version for the opencl_intel_100 plan class, and it's recognizing that you are requesting work for an intel gpu, it's not recognizing that the anonymous platform version uses the intel GPU. Could you post the appropriate section of your app_info.xml?

There seems to be some confusion in the server code between "intel" and "intel_gpu" as an identifier. That could be the cause.

And can you move your app_info.xml out of the way, and test whether it gives you anything without anonymous apps around? Thanks.



2013-06-13 15:28:14.4861 [PID=27322] Request: [USER#747] [HOST#63590] [IP 77.37.226.252] client 7.0.52
2013-06-13 15:28:14.4867 [PID=27322] [version] (opencl_intel_100) using client-supplied flops 34.863508G
2013-06-13 15:28:14.4867 [PID=27322] [quota] effective ncpus 4 ngpus 1
2013-06-13 15:28:14.4868 [PID=27322] [quota] max jobs per RPC: 40
2013-06-13 15:28:14.4868 [PID=27322] [quota] Overall limits on jobs in progress:
2013-06-13 15:28:14.4868 [PID=27322] [send] Not using matchmaker scheduling; Not using EDF sim
2013-06-13 15:28:14.4868 [PID=27322] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2013-06-13 15:28:14.4868 [PID=27322] [send] Intel GPU: req 878581.89 sec, 0.00 instances; est delay 0.00
2013-06-13 15:28:14.4868 [PID=27322] [send] work_req_seconds: 878581.89 secs
2013-06-13 15:28:14.4868 [PID=27322] [send] available disk 9.00 GB, work_buf_min 432000
2013-06-13 15:28:14.4868 [PID=27322] [send] active_frac 0.930513 on_frac 0.926358
2013-06-13 15:28:14.4868 [PID=27322] [send] Anonymous platform app versions:
2013-06-13 15:28:14.4869 [PID=27322] app: astropulse_v6 version 608 cpus 0.04 flops 34.863508G
2013-06-13 15:28:14.4869 [PID=27322] [send] CPU features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 popcnt aes syscall nx lm vmx smx tm2 pbe
2013-06-13 15:28:14.4876 [PID=27322] [send] don't need more work
2013-06-13 15:28:14.4881 [PID=27322] Sending reply to [HOST#63590]: 0 results, delay req 7.00
2013-06-13 15:28:14.4882 [PID=27322] Scheduler ran 0.007 seconds

---

If one does not use app_info.xml, then how does one request work for the intel GPU in this case? In preferences, it is missing the option to select intel GPU.

Edit: Un-selecting CPU ends up with GPU tasks downloaded. So that's good. What if I wanted CPU tasks, but not GPU tasks?

Edit 2: My second machine refuses to ask for work, even though the intel GPU is sitting idle.

---

Dublin, California
Team: SETI.USA

Edit: Un-selecting CPU ends up with GPU tasks downloaded. So that's good. What if I wanted CPU tasks, but not GPU tasks?

That probably means your CPU is overcommitted with work for other projects or for this beta. Since the GPU app uses non-zero CPU time, that could prevent it from getting results. On the other hand it could mean that the server is estimating that the GPU app will use more than one CPU. I'll try to add some more debugging output to the server tomorrow.

Edit 2: My second machine refuses to ask for work, even though the intel GPU is sitting idle.

I looked at the requests. Your machine is asking for 0 seconds of Intel GPU work. That could mean that your GPU is overcommitted with work for other projects, or it could be a bug in the BOINC client. I'd say let it run a couple days and see if it eventually asks for non-zero work.

Thanks for the info!

---

Thanks. I'll let the two machines run exclusively for a few days, even if idle. FWIW, they are running 7.1.15.

But why no GPU selection in preferences for Intel, just like AMD and nVidia? These kind of machines need that.

Edit: Colllatz has a selection for Intel GPU. Just FYI.

Example: https://dl.dropboxusercontent.com/u/55884901/intel_gpu_prefs.jpg

to be clear, here are the two machines:

machine that got tasks so far: http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=62655

machine that can't/won't get tasks: http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=62709

---

Dublin, California
Team: SETI.USA

First task ran for 59 seconds and then failed. Task 5433708

Can't figure out from the stderr output what the reason for the failure is.

---

Can't figure out from the stderr output what the reason for the failure is.

This could have been an opencl driver issue. I have re-installed the Intel drivers. Now I just have to wait for the next task. Boinc keeps asking for nVidia tasks and if I unselect that option on my account page I get "Not requesting tasks: don't need"

But why no GPU selection in preferences for Intel, just like AMD and nVidia?

I'll have to figure out how to add one.

---

But why no GPU selection in preferences for Intel, just like AMD and nVidia?

I'll have to figure out how to add one.

IIRC, the 'Use CPU', 'Use ATI', 'Use NVIDIA', options are made visible and enabled automatically when applications defined for those resources are added to the project.

http://boinc.berkeley.edu/trac/browser/boinc-v2/html/inc/prefs.inc has:

42	// <project_preferences>
43	//    <resource_share>4</resource_share>
44	//    [ <allow_beta_work>0|1</allow_beta_work> ]
45	//    [ <no_cpu>1</no_cpu> ]
46	//    [ <no_cuda>1</no_cuda> ]
47	//    [ <no_ati>1</no_ati> ]
48	//    <project-specific>
49	//            ... (arbitrary project-specific XML)
50	//    </project-specific>
51	//    <home>
52	//       ...
53	//    </home>
54	// </project_preferences>

Whether David is treating "intel_gpu" as an arbitrary resource, or just hasn't got round to adding it yet, I couldn't say. But it should be consistent across all projects (which rather argues against 'arbitrary project-specific') when he does get there.

And I rather suspect that having a placeholder for external app testing, but no actual installed app, probably wouldn't be enough to trigger automatic visibility. That would be another one for the web-code maintainers, which probably means David again.

This is what is happening with my two hosts now. I de-select all three: CPU, ATI and nVidia, and then BOINC won't ask for any tasks, even with the GPU idle. So then I select only CPU, and BOINC requests CPU tasks, but not intel GPU tasks.

Not sure why it worked last night on one of the machines. But it's not working at all now.

I tried ATI only, and nVidia only, and both ATI and nVidia just for grins, and the same result. BOINC won't ask for tasks.

FWIW, it does work with an app_info.xml. But that defeats the purpose.

---

Dublin, California
Team: SETI.USA

It might be the options thing. Looking into it now.

Neither of you Intel GPU hosts are asking for Intel GPU work.

---

I think I may have found the problem. Hidden in the code is a check for plan classes that contain the string "intel_gpu". If no plan classes contain that string the code presumes that you don't support Intel GPUs.

So I renamed the plan class to "opencl_intel_gpu_100". We'll see if that helps.

---

I'll try it out. Which of the CPU/ATI/nVidia boxes should we be checking, if any?

Edit: Look at that! Intel GPU is an option. Thanks!

---

Dublin, California
Team: SETI.USA

Eric, my host works w/o app_info on beta. It was request for stock app.

Использовать ЦП
начиная с версии 6.10+ да
Использовать ГП ATI
начиная с версии 6.10+ да
Использовать ГП NVIDIA
начиная с версии 6.10+ да
Use Intel GPU
Enforced by version 7.0+ да

but no work still.

Yeah, I have only intel GPU selected, and no work. BOINC just won't ask for it, even though the GPU is idle.


9225 SETI@home Beta Test 6/14/2013 1:06:15 PM update requested by user
9226 SETI@home Beta Test 6/14/2013 1:06:17 PM [sched_op] Starting scheduler request
9227 SETI@home Beta Test 6/14/2013 1:06:17 PM Sending scheduler request: Requested by user.
9228 SETI@home Beta Test 6/14/2013 1:06:17 PM Not requesting tasks: don't need
9229 SETI@home Beta Test 6/14/2013 1:06:17 PM [sched_op] CPU work request: 0.00 seconds; 0.00 devices
9230 SETI@home Beta Test 6/14/2013 1:06:17 PM [sched_op] intel_gpu work request: 0.00 seconds; 0.00 devices
9231 SETI@home Beta Test 6/14/2013 1:06:19 PM Scheduler request completed
9232 SETI@home Beta Test 6/14/2013 1:06:19 PM [sched_op] Server version 701
9233 SETI@home Beta Test 6/14/2013 1:06:19 PM Project requested delay of 7 seconds
9234 SETI@home Beta Test 6/14/2013 1:06:19 PM [sched_op] Deferring communication for 00:00:07
9235 SETI@home Beta Test 6/14/2013 1:06:19 PM [sched_op] Reason: requested by project

I'll continue to not crunch any other intel GPU projects, to see if it changes its mind.

(Does anyone else find the new message of "Not requesting tasks: don't need" annoying? It's like BOINC is saying "you're not the boss of me!")

Edit: I even tried using <fetch_on_update>1</fetch_on_update>, and BOINC still will not ask for work.

<fetch_on_update>0|1</fetch_on_update>
When updating a project, request work even if not highest priority project. New in 7.0.54

---

Dublin, California
Team: SETI.USA

my host asks for work, quite a lot seconds... but still doesn't recive any work.

Does anyone else find the new message of "Not requesting tasks: don't need" annoying?

I do...
Strange thing is that the first time Boinc asked for work after it asked for intel_gpu work. Then I unchecked some options (v7, CPU, ATi GPU) and it stopped asking for intel_gpu work.

It's not Boinc, because when I update Einstein I get this:

14-6-2013 22:30:42 | Einstein@Home | update requested by user
14-6-2013 22:30:43 | Einstein@Home | Sending scheduler request: Requested by user.
14-6-2013 22:30:43 | Einstein@Home | Requesting new tasks for intel_gpu
14-6-2013 22:30:45 | Einstein@Home | Scheduler request completed: got 0 new tasks

Collatz works as well, just finished a 8000 credit task in 5 hours on the Intel GPU

my host asks for work, quite a lot seconds... but still doesn't recive any work.

I think you need to change the plan class in your app_info.xml to be opencl_intel_gpu_100, it looks like you're still requesting for opencl_intel_100.

---