Thread 'GPU not receiving tasks when CPU computing disabled'

Message boards : Questions and problems : GPU not receiving tasks when CPU computing disabled
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 5 · Next

AuthorMessage
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104169 - Posted: 29 Apr 2021, 21:58:49 UTC

Hi,

When "Allow research to run on my CPU?" in the project preferences is set to no, the project is not sending tasks for the intel GPU. It appears to me that the client is requesting it, because it shows:
29-Apr-2021 12:32:22 [World Community Grid] Requesting new tasks for Intel GPU
29-Apr-2021 12:32:22 [World Community Grid] [sched_op] CPU work request: 0.00 seconds; 0.00 devices
29-Apr-2021 12:32:22 [World Community Grid] [sched_op] Intel GPU work request: 108243.50 seconds; 0.00 devices
29-Apr-2021 12:32:23 [World Community Grid] Scheduler request completed: got 0 new tasks


It is not at the limit of tasks for opng on world community grid(wcg). If I enable "Allow research to run on my CPU?", intel GPU tasks start being sent again. This has been repeatable for me. I came across it trying to reproduce a problem that another wcg user was having. I was only able to reproduce it on a machine with just an intel GPU. I also tried it on a machine with an intel and nvidia GPU, but that machine continued getting intel GPU tasks.

Here is a snip showing what happened when I enabled . I had the debug flags turned off this time. I can go back through the log to a previous one that shows it with sched_op_debug and work_fetch_debug or enable them and repeat it.
30-Apr-2021 00:12:16 [World Community Grid] Sending scheduler request: Requested by user.
30-Apr-2021 00:12:16 [World Community Grid] Requesting new tasks for Intel GPU
30-Apr-2021 00:12:17 [World Community Grid] Scheduler request completed: got 0 new tasks
30-Apr-2021 00:12:17 [World Community Grid] Project requested delay of 121 seconds
30-Apr-2021 00:14:19 [World Community Grid] update requested by user
30-Apr-2021 00:14:22 [World Community Grid] Sending scheduler request: Requested by user.
30-Apr-2021 00:14:22 [World Community Grid] Requesting new tasks for Intel GPU
30-Apr-2021 00:14:23 [World Community Grid] Scheduler request completed: got 0 new tasks
30-Apr-2021 00:14:23 [World Community Grid] Project requested delay of 121 seconds
30-Apr-2021 00:14:28 [World Community Grid] General prefs: from World Community Grid (last modified 30-Apr-2021 00:12:36)
30-Apr-2021 00:14:28 [World Community Grid] Host location: none
30-Apr-2021 00:14:28 [World Community Grid] General prefs: using your defaults
30-Apr-2021 00:14:28 [---] Preferences:
30-Apr-2021 00:14:28 [---]    max memory usage when active: 22838.06 MB
30-Apr-2021 00:14:28 [---]    max memory usage when idle: 29363.22 MB
30-Apr-2021 00:14:28 [---]    max disk usage: 1.98 GB
30-Apr-2021 00:14:28 [---]    (to change preferences, visit a project web site or select Preferences in the Manager)
30-Apr-2021 00:16:25 [World Community Grid] update requested by user
30-Apr-2021 00:16:28 [World Community Grid] Sending scheduler request: Requested by user.
30-Apr-2021 00:16:28 [World Community Grid] Requesting new tasks for CPU and Intel GPU
30-Apr-2021 00:16:29 [World Community Grid] Scheduler request completed: got 2 new tasks
30-Apr-2021 00:16:29 [World Community Grid] Project requested delay of 121 seconds
30-Apr-2021 00:16:31 [World Community Grid] Started download of fb36d9a7d451bd460b6d3755f62304c8.pdbqt
30-Apr-2021 00:16:31 [World Community Grid] Started download of fba66e5bfc0897aabddeae317b24ec34.gpf
30-Apr-2021 00:16:33 [World Community Grid] Finished download of fba66e5bfc0897aabddeae317b24ec34.gpf
30-Apr-2021 00:16:33 [World Community Grid] Started download of 8d1b0f86d53e819ef868211a43cf4ffe.job
30-Apr-2021 00:16:34 [World Community Grid] Finished download of fb36d9a7d451bd460b6d3755f62304c8.pdbqt
30-Apr-2021 00:16:34 [World Community Grid] Finished download of 8d1b0f86d53e819ef868211a43cf4ffe.job
30-Apr-2021 00:16:34 [World Community Grid] Started download of 41e8a8afd8b26acc3415c426c9f9a427.zip
30-Apr-2021 00:16:34 [World Community Grid] Started download of 55964423eab329ab44bfe14385c5cbf5.tpr
30-Apr-2021 00:16:35 [World Community Grid] Finished download of 41e8a8afd8b26acc3415c426c9f9a427.zip
30-Apr-2021 00:16:36 [World Community Grid] Finished download of 55964423eab329ab44bfe14385c5cbf5.tpr


Start of BOINC log:
29-Apr-2021 12:25:51 [---] Starting BOINC client version 7.16.5 for windows_x86_64
29-Apr-2021 12:25:51 [---] log flags: file_xfer, sched_ops, task, sched_op_debug, work_fetch_debug
29-Apr-2021 12:25:51 [---] Libraries: libcurl/7.47.1 OpenSSL/1.0.2s zlib/1.2.8
29-Apr-2021 12:25:51 [---] Data directory: C:\ProgramData\BOINC
29-Apr-2021 12:25:51 [---] Running under account benjamin
29-Apr-2021 12:25:51 [---] OpenCL: Intel GPU 0: Intel(R) UHD Graphics 620 (driver version 26.20.100.7262, device version OpenCL 2.1 NEO, 13050MB, 13050MB available, 221 GFLOPS peak)
29-Apr-2021 12:25:51 [---] OpenCL CPU: Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz (OpenCL driver vendor: Intel(R) Corporation, driver version 7.6.0.0814, device version OpenCL 2.1 (Build 0))
29-Apr-2021 12:25:52 [---] Windows processor group 0: 8 processors
29-Apr-2021 12:25:52 [---] Host name: DESKTOP-[snip]
29-Apr-2021 12:25:52 [---] Processor: 8 GenuineIntel Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz [Family 6 Model 142 Stepping 10]
29-Apr-2021 12:25:52 [---] Processor 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 fma cx16 sse4_1 sse4_2 movebe popcnt aes f16c rdrandsyscall nx lm avx avx2 vmx smx tm2 pbe fsgsbase bmi1 hle smep bmi2
29-Apr-2021 12:25:52 [---] OS: Microsoft Windows 10: Professional x64 Edition, (10.00.18362.00)
29-Apr-2021 12:25:52 [---] Memory: 31.86 GB physical, 36.61 GB virtual
29-Apr-2021 12:25:52 [---] Disk: 2.00 GB total, 1.11 GB free
29-Apr-2021 12:25:52 [---] Local time is UTC +3 hours
29-Apr-2021 12:25:52 [---] WSL detected:
29-Apr-2021 12:25:52 [---]    [Ubuntu] (default): Linux Ubuntu (Ubuntu 18.04.3 LTS [4.4.0-18362-Microsoft])
29-Apr-2021 12:25:52 [---] VirtualBox version: 6.0.10
29-Apr-2021 12:25:52 [World Community Grid] Found app_config.xml
29-Apr-2021 12:25:52 [World Community Grid] arp1: Max -1 concurrent jobs
29-Apr-2021 12:25:52 [World Community Grid] hst1: Max -1 concurrent jobs
29-Apr-2021 12:25:52 [---] Config: GUI RPC allowed from any host
29-Apr-2021 12:25:52 [---] Config: GUI RPCs allowed from:
29-Apr-2021 12:25:52 [---]     192.168.1.32
29-Apr-2021 12:25:52 [---] Config: use all coprocessors
29-Apr-2021 12:25:57 [World Community Grid] General prefs: from World Community Grid (last modified 29-Apr-2021 09:43:04)
29-Apr-2021 12:25:57 [World Community Grid] Host location: none
29-Apr-2021 12:25:57 [World Community Grid] General prefs: using your defaults
29-Apr-2021 12:25:57 [---] Preferences:
29-Apr-2021 12:25:57 [---]    max memory usage when active: 22838.06 MB
29-Apr-2021 12:25:57 [---]    max memory usage when idle: 29363.22 MB
29-Apr-2021 12:25:57 [---]    max disk usage: 1.98 GB
29-Apr-2021 12:25:57 [---]    (to change preferences, visit a project web site or select Preferences in the Manager)
29-Apr-2021 12:25:57 [---] [work_fetch] Request work fetch: Prefs update
29-Apr-2021 12:25:57 [---] [work_fetch] Request work fetch: Startup
29-Apr-2021 12:25:57 [---] Setting up project and slot directories
29-Apr-2021 12:25:57 [---] Checking active tasks
29-Apr-2021 12:25:57 [---] Using account manager BOINCstatsBAM!
29-Apr-2021 12:25:57 [World Community Grid] URL http://www.worldcommunitygrid.org/; Computer ID [snip]; resource share 300
29-Apr-2021 12:25:57 [---] Setting up GUI RPC socket
29-Apr-2021 12:25:57 [---] Checking presence of 396 project files
29-Apr-2021 12:25:57 Initialization completed


Hopefully this is enough information (at least to start) for you guys to be able to help.

Thank you,
Benjamin
ID: 104169 · Report as offensive
ProfileKeith Myers
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 17 Nov 16
Posts: 890
United States
Message 104171 - Posted: 30 Apr 2021, 1:34:51 UTC

Actually it is stating it isn't asking for gpu work.

29-Apr-2021 12:32:22 [World Community Grid] [sched_op] Intel GPU work request: 108243.50 seconds; 0.00 devices

It should be stating at least 1.00 devices to get any gpu work.

There is issues with Intel iGPU and the project. Read some of the posts at WCG OpenPandemics sub-project to trick the project into sending gpu work.
Temporarily change your preferences to get the cpu OPN1 tasks and start a task, update the project again to lock that in and see if it then sends some gpu work, then change the preferences again to no cpu tasks.
Either finish the OPN1 cpu task or abort it.
ID: 104171 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 104176 - Posted: 30 Apr 2021, 6:15:46 UTC

Actually, the 'devices' field is only non-zero if you have absolutely no work left left and the device - Intel GPU - is idle.

108243.50 seconds (over 30 hours) is a lot of work to ask for in a single gulp, and for a project under stress (deliberately - they're running a stress test!), it will make life even harder for the project server. Under normal circumstances - unless you have an intermittent internet connection or something - your machine will be contacting the server at least once an hour to report results, and it's better to ask for a top-up then.
ID: 104176 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104177 - Posted: 30 Apr 2021, 7:30:23 UTC - in response to Message 104171.  

Actually it is stating it isn't asking for gpu work.

29-Apr-2021 12:32:22 [World Community Grid] [sched_op] Intel GPU work request: 108243.50 seconds; 0.00 devices

It should be stating at least 1.00 devices to get any gpu work.

There is issues with Intel iGPU and the project. Read some of the posts at WCG OpenPandemics sub-project to trick the project into sending gpu work.
Temporarily change your preferences to get the cpu OPN1 tasks and start a task, update the project again to lock that in and see if it then sends some gpu work, then change the preferences again to no cpu tasks.
Either finish the OPN1 cpu task or abort it.

Hi Keith,
That is basically what I said I did to reproduce the issue. I set the preferences to not allow CPU and stopped getting intel GPU work. Then set it to allow and started receiving intel GPU work. As Richard said, 0.00 devices does not mean it is not requesting work, I can paste lines showing it saying 0.00 and getting work if you would like.
ID: 104177 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104178 - Posted: 30 Apr 2021, 7:34:22 UTC - in response to Message 104176.  

Actually, the 'devices' field is only non-zero if you have absolutely no work left left and the device - Intel GPU - is idle.

108243.50 seconds (over 30 hours) is a lot of work to ask for in a single gulp, and for a project under stress (deliberately - they're running a stress test!), it will make life even harder for the project server. Under normal circumstances - unless you have an intermittent internet connection or something - your machine will be contacting the server at least once an hour to report results, and it's better to ask for a top-up then.

Hi Richard,
It is true that that is pretty high. I normally do not have the cache set so high. I just temporarily increased while trying different things to try to reproduce the issue that someone else (binii) was having.

I guess what I am asking since I know that you a lot more about the inner workings of BOINC than I do is:
Is this is a BOINC issue(client or server) or a WCG server specific issue?
If it is a WCG server specific issue, what data should I record during the stress test when work unit availability is not a factor to show the issue and help them find and fix it?
ID: 104178 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 104179 - Posted: 30 Apr 2021, 8:06:14 UTC - in response to Message 104178.  

I guess what I am asking since I know that you a lot more about the inner workings of BOINC than I do is:
Is this is a BOINC issue(client or server) or a WCG server specific issue?
If it is a WCG server specific issue, what data should I record during the stress test when work unit availability is not a factor to show the issue and help them find and fix it?
Hi there.

I've been working for a while on the generic BOINC server code, but I'm very new to the specifics of WCG - I only joined at the start of the Covid-19 GPU beta test. And we have many, many, variables flying around at once, and being changed at short notice.

But about 30 minutes ago, I logged:

30/04/2021 08:33:28 | World Community Grid | Sending scheduler request: To fetch work.
30/04/2021 08:33:28 | World Community Grid | Requesting new tasks for NVIDIA GPU and Intel GPU
30/04/2021 08:33:28 | World Community Grid | [sched_op] NVIDIA GPU work request: 560.14 seconds; 0.00 devices
30/04/2021 08:33:28 | World Community Grid | [sched_op] Intel GPU work request: 150.10 seconds; 0.00 devices
30/04/2021 08:33:29 | World Community Grid | Scheduler request completed: got 5 new tasks
30/04/2021 08:33:29 | World Community Grid | [sched_op] estimated total NVIDIA GPU task duration: 704 seconds
30/04/2021 08:33:29 | World Community Grid | [sched_op] estimated total Intel GPU task duration: 2906 seconds
So, at that much-reduced level, it's working as designed. My cache settings, during this test, are 0.1 days + 0.01 days: I have 63 tasks on that machine, running two-up on each of two NVidia GPUs (GTX 1660 and GTX 1650). I also have a 'update every five minutes', script running, not that it's really needed during the stress test.

So, I suspect, the 'little and often' approach works best under current circumstances. That means your experience is perhaps best put down to strange things happening when a BOINC server is stressed beyond its normal limits. Similar things were seen at SETI@Home at busy times.
ID: 104179 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 104180 - Posted: 30 Apr 2021, 8:31:02 UTC

Two more logs - I updated the cache to 0.3 days.

30/04/2021 09:17:27 | World Community Grid | Sending scheduler request: To fetch work.
30/04/2021 09:17:27 | World Community Grid | Requesting new tasks for NVIDIA GPU and Intel GPU
30/04/2021 09:17:27 | World Community Grid | [sched_op] NVIDIA GPU work request: 20370.37 seconds; 0.00 devices
30/04/2021 09:17:27 | World Community Grid | [sched_op] Intel GPU work request: 8172.86 seconds; 0.00 devices
30/04/2021 09:17:28 | World Community Grid | Scheduler request completed: got 22 new tasks
30/04/2021 09:17:28 | World Community Grid | [sched_op] estimated total NVIDIA GPU task duration: 19356 seconds
30/04/2021 09:17:28 | World Community Grid | [sched_op] estimated total Intel GPU task duration: 0 seconds

30/04/2021 09:19:31 | World Community Grid | Sending scheduler request: To fetch work.
30/04/2021 09:19:31 | World Community Grid | Reporting 2 completed tasks
30/04/2021 09:19:31 | World Community Grid | Requesting new tasks for NVIDIA GPU and Intel GPU
30/04/2021 09:19:31 | World Community Grid | [sched_op] NVIDIA GPU work request: 10888.86 seconds; 0.00 devices
30/04/2021 09:19:31 | World Community Grid | [sched_op] Intel GPU work request: 8234.69 seconds; 0.00 devices
30/04/2021 09:19:32 | World Community Grid | Scheduler request completed: got 2 new tasks
30/04/2021 09:19:32 | World Community Grid | [sched_op] estimated total NVIDIA GPU task duration: 1760 seconds
30/04/2021 09:19:32 | World Community Grid | [sched_op] estimated total Intel GPU task duration: 0 seconds
I now have 107 tasks, and 7 of them are for the intel. We have a limit of 50 tasks per GPU, so my NV cards are maxxed out. Reporting 2 tasks gave me space for 2 replacements, but at that point, NV processing would have been stopped. None for Intel, although work was requested? I suspect server processing stops when one part hits a limit, and doesn't move on to the next part of the request. Best of luck in getting David Anderson to fix that one.
ID: 104180 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 104181 - Posted: 30 Apr 2021, 8:35:28 UTC
Last modified: 30 Apr 2021, 8:40:57 UTC

Confirmed - while I was typing,

30/04/2021 09:32:23 | World Community Grid | [sched_op] NVIDIA GPU work request: 8525.33 seconds; 0.00 devices
30/04/2021 09:32:23 | World Community Grid | [sched_op] Intel GPU work request: 5921.32 seconds; 0.00 devices
30/04/2021 09:32:24 | World Community Grid | Scheduler request completed: got 0 new tasks
30/04/2021 09:32:24 | World Community Grid | This computer has reached a limit on tasks in progress
Nothing for Intel while NVidia is at the limit. But if you report one, you can get a replacement:

30/04/2021 09:27:23 | World Community Grid | Reporting 1 completed tasks
30/04/2021 09:27:23 | World Community Grid | [sched_op] NVIDIA GPU work request: 10744.88 seconds; 0.00 devices
30/04/2021 09:27:23 | World Community Grid | [sched_op] Intel GPU work request: 8605.69 seconds; 0.00 devices
30/04/2021 09:27:26 | World Community Grid | Scheduler request completed: got 1 new tasks
30/04/2021 09:27:26 | World Community Grid | [sched_op] estimated total NVIDIA GPU task duration: 0 seconds
30/04/2021 09:27:26 | World Community Grid | [sched_op] estimated total Intel GPU task duration: 2906 seconds
ID: 104181 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104182 - Posted: 30 Apr 2021, 9:33:02 UTC - in response to Message 104181.  

Thank you for looking into it Richard. I apologize if I did not state this clearly before:
    I am unable to reproduce this on a machine that has an integrated intel + a discrete nvidia card. Only on one with just the integrated intel card.
    If cpu is disabled in the project prefs, it does not get any gpu tasks.
    If cpu is enabled in the project prefs, it fills up to the its limit on gpu tasks.



If that is not clear, please let me know and I will try to state it another way.

ID: 104182 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104183 - Posted: 30 Apr 2021, 9:39:01 UTC - in response to Message 104182.  

I forgot to mention that I have 29 CPU tasks and 12 GPU tasks on that machine at this moment. So I do not think it is hitting a limit of 50. :)
ID: 104183 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 104184 - Posted: 30 Apr 2021, 9:41:49 UTC - in response to Message 104182.  

That sounds like a related class of problem: if something blocks one part of the request, it doesn't fall through and process the Intel part. Either your 'CPU disabled'. or my 'NV at task limit' is enough to skip processing the intel part of the request.

I think the 'must enable CPU to get intel' variant was reported fairly early in the Beta test, and from memory Uplinger fixed it? Might be worth looking through those Beta messages. I'll try to look through the relevant bit of the server code, but this weekend is beginning to get busy.
ID: 104184 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104185 - Posted: 30 Apr 2021, 9:48:36 UTC

Here is a work request with sched_op_debug and work_fetch_debug turned on. This is with cpu being disabled in the profile preferences. The number at the beginning is added by boinc tasks. If that is distracting, I can copy it directly from the machine in the future:
24375	World Community Grid	4/30/2021 12:38:35 PM	update requested by user	
24376			4/30/2021 12:38:35 PM	[work_fetch] Request work fetch: project updated by user	
24377			4/30/2021 12:38:36 PM	[work_fetch] Request work fetch: Backoff ended for World Community Grid	
24378	World Community Grid	4/30/2021 12:38:39 PM	[sched_op] sched RPC pending: Requested by user	
24379	World Community Grid	4/30/2021 12:38:39 PM	piggyback_work_request()	
24380			4/30/2021 12:38:39 PM	[work_fetch] ------- start work fetch state -------	
24381			4/30/2021 12:38:39 PM	[work_fetch] target work buffer: 17280.00 + 4320.00 sec	
24382			4/30/2021 12:38:39 PM	[work_fetch] --- project states ---	
24383	World Community Grid	4/30/2021 12:38:39 PM	[work_fetch] REC 6753.904 prio -2.365 can request work	
24384			4/30/2021 12:38:39 PM	[work_fetch] --- state for CPU ---	
24385			4/30/2021 12:38:39 PM	[work_fetch] shortfall 147804.30 nidle 6.73 saturated 0.00 busy 0.00	
24386	World Community Grid	4/30/2021 12:38:39 PM	[work_fetch] share 0.000 blocked by project preferences 	
24387			4/30/2021 12:38:39 PM	[work_fetch] --- state for Intel GPU ---	
24388			4/30/2021 12:38:39 PM	[work_fetch] shortfall 9113.28 nidle 0.00 saturated 12486.72 busy 0.00	
24389	World Community Grid	4/30/2021 12:38:39 PM	[work_fetch] share 1.000  	
24390			4/30/2021 12:38:39 PM	[work_fetch] ------- end work fetch state -------	
24391	World Community Grid	4/30/2021 12:38:39 PM	piggyback: resource CPU	
24392	World Community Grid	4/30/2021 12:38:39 PM	piggyback: can't fetch CPU: blocked by project preferences	
24393	World Community Grid	4/30/2021 12:38:39 PM	piggyback: resource Intel GPU	
24394	World Community Grid	4/30/2021 12:38:39 PM	[work_fetch] using MC shortfall 9113.281509 instead of shortfall 9113.281509	
24395	World Community Grid	4/30/2021 12:38:39 PM	[work_fetch] set_request() for Intel GPU: ninst 1 nused_total 12.00 nidle_now 0.00 fetch share 1.00 req_inst 0.00 req_secs 9113.28	
24396	World Community Grid	4/30/2021 12:38:39 PM	[sched_op] Starting scheduler request	
24397	World Community Grid	4/30/2021 12:38:39 PM	[work_fetch] request: CPU (0.00 sec, 0.00 inst) Intel GPU (9113.28 sec, 0.00 inst)	
24398	World Community Grid	4/30/2021 12:38:39 PM	Sending scheduler request: Requested by user.	
24399	World Community Grid	4/30/2021 12:38:39 PM	Requesting new tasks for Intel GPU	
24400	World Community Grid	4/30/2021 12:38:39 PM	[sched_op] CPU work request: 0.00 seconds; 0.00 devices	
24401	World Community Grid	4/30/2021 12:38:39 PM	[sched_op] Intel GPU work request: 9113.28 seconds; 0.00 devices	
24402	World Community Grid	4/30/2021 12:38:40 PM	Scheduler request completed: got 0 new tasks	
24403	World Community Grid	4/30/2021 12:38:40 PM	[sched_op] Server version 701	
24404	World Community Grid	4/30/2021 12:38:40 PM	Project requested delay of 121 seconds	
24405	World Community Grid	4/30/2021 12:38:40 PM	[sched_op] Deferring communication for 00:02:01	
24406	World Community Grid	4/30/2021 12:38:40 PM	[sched_op] Reason: requested by project	
24407			4/30/2021 12:38:40 PM	[work_fetch] Request work fetch: RPC complete	


Here is what the log looks like when I enabled cpu in the profile preferences(including it receiving the preferences). It receives a gpu task on the very next update. The quick retry from the download failure was me telling it to retry the transfer manually.
24525	World Community Grid	4/30/2021 12:42:51 PM	Requesting new tasks for Intel GPU	
24526	World Community Grid	4/30/2021 12:42:51 PM	[sched_op] CPU work request: 0.00 seconds; 0.00 devices	
24527	World Community Grid	4/30/2021 12:42:51 PM	[sched_op] Intel GPU work request: 9570.44 seconds; 0.00 devices	
24528	World Community Grid	4/30/2021 12:42:52 PM	Scheduler request completed: got 0 new tasks	
24529	World Community Grid	4/30/2021 12:42:52 PM	[sched_op] Server version 701	
24530	World Community Grid	4/30/2021 12:42:52 PM	Project requested delay of 121 seconds	
24531	World Community Grid	4/30/2021 12:42:57 PM	General prefs: from World Community Grid (last modified 30-Apr-2021 12:42:21)	
24532	World Community Grid	4/30/2021 12:42:57 PM	Host location: none	
24533	World Community Grid	4/30/2021 12:42:57 PM	General prefs: using your defaults	
24534			4/30/2021 12:42:57 PM	Preferences:	
24535			4/30/2021 12:42:57 PM	   max memory usage when active: 22838.06 MB	
24536			4/30/2021 12:42:57 PM	   max memory usage when idle: 29363.22 MB	
24537			4/30/2021 12:42:57 PM	   max disk usage: 1.98 GB	
24538			4/30/2021 12:42:57 PM	   (to change preferences, visit a project web site or select Preferences in the Manager)	
24539			4/30/2021 12:42:57 PM	[work_fetch] Request work fetch: Prefs update	
24540	World Community Grid	4/30/2021 12:42:57 PM	[sched_op] Deferring communication for 00:02:01	
24541	World Community Grid	4/30/2021 12:42:57 PM	[sched_op] Reason: requested by project	
24542			4/30/2021 12:42:57 PM	[work_fetch] Request work fetch: RPC complete	
24543			4/30/2021 12:42:57 PM	choose_project(): 1619775777.156166	
24544			4/30/2021 12:42:57 PM	[work_fetch] ------- start work fetch state -------	
24545			4/30/2021 12:42:57 PM	[work_fetch] target work buffer: 17280.00 + 4320.00 sec	
24546			4/30/2021 12:42:57 PM	[work_fetch] --- project states ---	
24547	World Community Grid	4/30/2021 12:42:57 PM	[work_fetch] REC 6754.620 prio -1.354 can't request work: scheduler RPC backoff (116.11 sec)	
24548			4/30/2021 12:42:57 PM	[work_fetch] --- state for CPU ---	
24549			4/30/2021 12:42:57 PM	[work_fetch] shortfall 147927.40 nidle 6.73 saturated 0.00 busy 0.00	
24550	World Community Grid	4/30/2021 12:42:57 PM	[work_fetch] share 0.000  	
24551			4/30/2021 12:42:57 PM	[work_fetch] --- state for Intel GPU ---	
24552			4/30/2021 12:42:57 PM	[work_fetch] shortfall 9565.94 nidle 0.00 saturated 12034.06 busy 0.00	
24553	World Community Grid	4/30/2021 12:42:57 PM	[work_fetch] share 0.000  	
24554			4/30/2021 12:42:57 PM	[work_fetch] ------- end work fetch state -------	
24555	World Community Grid	4/30/2021 12:42:57 PM	choose_project: scanning	
24556	World Community Grid	4/30/2021 12:42:57 PM	skip: scheduler RPC backoff	
24557			4/30/2021 12:42:57 PM	[work_fetch] No project chosen for work fetch	
24558			4/30/2021 12:43:57 PM	choose_project(): 1619775837.160690	
24559			4/30/2021 12:43:57 PM	[work_fetch] ------- start work fetch state -------	
24560			4/30/2021 12:43:57 PM	[work_fetch] target work buffer: 17280.00 + 4320.00 sec	
24561			4/30/2021 12:43:57 PM	[work_fetch] --- project states ---	
24562	World Community Grid	4/30/2021 12:43:57 PM	[work_fetch] REC 6754.786 prio -1.353 can't request work: scheduler RPC backoff (56.10 sec)	
24563			4/30/2021 12:43:57 PM	[work_fetch] --- state for CPU ---	
24564			4/30/2021 12:43:57 PM	[work_fetch] shortfall 147948.23 nidle 6.73 saturated 0.00 busy 0.00	
24565	World Community Grid	4/30/2021 12:43:57 PM	[work_fetch] share 0.000  	
24566			4/30/2021 12:43:57 PM	[work_fetch] --- state for Intel GPU ---	
24567			4/30/2021 12:43:57 PM	[work_fetch] shortfall 9642.53 nidle 0.00 saturated 11957.47 busy 0.00	
24568	World Community Grid	4/30/2021 12:43:57 PM	[work_fetch] share 0.000  	
24569			4/30/2021 12:43:57 PM	[work_fetch] ------- end work fetch state -------	
24570	World Community Grid	4/30/2021 12:43:57 PM	choose_project: scanning	
24571	World Community Grid	4/30/2021 12:43:57 PM	skip: scheduler RPC backoff	
24572			4/30/2021 12:43:57 PM	[work_fetch] No project chosen for work fetch	
24573			4/30/2021 12:44:54 PM	[work_fetch] Request work fetch: Backoff ended for World Community Grid	
24574	World Community Grid	4/30/2021 12:44:54 PM	update requested by user	
24575			4/30/2021 12:44:54 PM	[work_fetch] Request work fetch: project updated by user	
24576	World Community Grid	4/30/2021 12:44:57 PM	[sched_op] sched RPC pending: Requested by user	
24577	World Community Grid	4/30/2021 12:44:57 PM	piggyback_work_request()	
24578			4/30/2021 12:44:57 PM	[work_fetch] ------- start work fetch state -------	
24579			4/30/2021 12:44:57 PM	[work_fetch] target work buffer: 17280.00 + 4320.00 sec	
24580			4/30/2021 12:44:57 PM	[work_fetch] --- project states ---	
24581	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] REC 6754.952 prio -2.348 can request work	
24582			4/30/2021 12:44:57 PM	[work_fetch] --- state for CPU ---	
24583			4/30/2021 12:44:57 PM	[work_fetch] shortfall 147997.97 nidle 6.73 saturated 0.00 busy 0.00	
24584	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] share 1.000  	
24585			4/30/2021 12:44:57 PM	[work_fetch] --- state for Intel GPU ---	
24586			4/30/2021 12:44:57 PM	[work_fetch] shortfall 9825.45 nidle 0.00 saturated 11774.55 busy 0.00	
24587	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] share 1.000  	
24588			4/30/2021 12:44:57 PM	[work_fetch] ------- end work fetch state -------	
24589	World Community Grid	4/30/2021 12:44:57 PM	piggyback: resource CPU	
24590	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] using MC shortfall 5598.819289 instead of shortfall 147997.969053	
24591	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] set_request() for CPU: ninst 8 nused_total 32.26 nidle_now 6.73 fetch share 1.00 req_inst 0.00 req_secs 5598.82	
24592	World Community Grid	4/30/2021 12:44:57 PM	piggyback: resource Intel GPU	
24593	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] using MC shortfall 9825.446515 instead of shortfall 9825.446515	
24594	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] set_request() for Intel GPU: ninst 1 nused_total 12.00 nidle_now 0.00 fetch share 1.00 req_inst 0.00 req_secs 9825.45	
24595	World Community Grid	4/30/2021 12:44:57 PM	[sched_op] Starting scheduler request	
24596	World Community Grid	4/30/2021 12:44:57 PM	[work_fetch] request: CPU (5598.82 sec, 0.00 inst) Intel GPU (9825.45 sec, 0.00 inst)	
24597	World Community Grid	4/30/2021 12:44:57 PM	Sending scheduler request: Requested by user.	
24598	World Community Grid	4/30/2021 12:44:57 PM	Requesting new tasks for CPU and Intel GPU	
24599	World Community Grid	4/30/2021 12:44:57 PM	[sched_op] CPU work request: 5598.82 seconds; 0.00 devices	
24600	World Community Grid	4/30/2021 12:44:57 PM	[sched_op] Intel GPU work request: 9825.45 seconds; 0.00 devices	
24601	World Community Grid	4/30/2021 12:44:58 PM	Scheduler request completed: got 2 new tasks	
24602	World Community Grid	4/30/2021 12:44:58 PM	[sched_op] Server version 701	
24603	World Community Grid	4/30/2021 12:44:58 PM	Project requested delay of 121 seconds	
24604	World Community Grid	4/30/2021 12:44:58 PM	[sched_op] estimated total CPU task duration: 0 seconds	
24605	World Community Grid	4/30/2021 12:44:58 PM	[sched_op] estimated total Intel GPU task duration: 1957 seconds	
24606	World Community Grid	4/30/2021 12:44:58 PM	[sched_op] Deferring communication for 00:02:01	
24607	World Community Grid	4/30/2021 12:44:58 PM	[sched_op] Reason: requested by project	
24608			4/30/2021 12:44:58 PM	[work_fetch] Request work fetch: RPC complete	
24609	World Community Grid	4/30/2021 12:45:00 PM	Started download of fbdc6e0fda8d7f1d82ff231e39604eff.pdbqt	
24610	World Community Grid	4/30/2021 12:45:00 PM	Started download of 5e9fd141bb6c2972fac4bf18c362e0ae.gpf	
24611			4/30/2021 12:45:03 PM	choose_project(): 1619775903.196308	
24612			4/30/2021 12:45:03 PM	[work_fetch] ------- start work fetch state -------	
24613			4/30/2021 12:45:03 PM	[work_fetch] target work buffer: 17280.00 + 4320.00 sec	
24614			4/30/2021 12:45:03 PM	[work_fetch] --- project states ---	
24615	World Community Grid	4/30/2021 12:45:03 PM	[work_fetch] REC 6754.952 prio -1.391 can't request work: scheduler RPC backoff (116.18 sec)	
24616			4/30/2021 12:45:03 PM	[work_fetch] --- state for CPU ---	
24617			4/30/2021 12:45:03 PM	[work_fetch] shortfall 147464.92 nidle 6.73 saturated 0.00 busy 0.00	
24618	World Community Grid	4/30/2021 12:45:03 PM	[work_fetch] share 0.000  	
24619			4/30/2021 12:45:03 PM	[work_fetch] --- state for Intel GPU ---	
24620			4/30/2021 12:45:03 PM	[work_fetch] shortfall 7865.30 nidle 0.00 saturated 13734.70 busy 0.00	
24621	World Community Grid	4/30/2021 12:45:03 PM	[work_fetch] share 0.000  	
24622			4/30/2021 12:45:03 PM	[work_fetch] ------- end work fetch state -------	
24623	World Community Grid	4/30/2021 12:45:03 PM	choose_project: scanning	
24624	World Community Grid	4/30/2021 12:45:03 PM	skip: scheduler RPC backoff	
24625			4/30/2021 12:45:03 PM	[work_fetch] No project chosen for work fetch	
24626	World Community Grid	4/30/2021 12:45:18 PM	Finished download of fbdc6e0fda8d7f1d82ff231e39604eff.pdbqt	
24627	World Community Grid	4/30/2021 12:45:18 PM	Started download of 7e9c9c1984863d609dcce586ace8e2bd.job	
24628	World Community Grid	4/30/2021 12:45:19 PM	Finished download of 7e9c9c1984863d609dcce586ace8e2bd.job	
24629	World Community Grid	4/30/2021 12:45:19 PM	Started download of a6c534197320c09bd9a0a0c668a2a701.zip	
24630	World Community Grid	4/30/2021 12:45:20 PM	Finished download of a6c534197320c09bd9a0a0c668a2a701.zip	
24631	World Community Grid	4/30/2021 12:45:20 PM	Started download of ea562e1c93436bf135895322e81315d0.pdbqt	
24632	World Community Grid	4/30/2021 12:45:21 PM	Finished download of ea562e1c93436bf135895322e81315d0.pdbqt	
24633	World Community Grid	4/30/2021 12:45:21 PM	Started download of a0ca1ec3e780cf0d8286a242a889eef6.gpf	
24634	World Community Grid	4/30/2021 12:45:22 PM	Temporarily failed download of 5e9fd141bb6c2972fac4bf18c362e0ae.gpf: connect() failed	
24635	World Community Grid	4/30/2021 12:45:22 PM	Backing off 00:02:23 on download of 5e9fd141bb6c2972fac4bf18c362e0ae.gpf	
24636	World Community Grid	4/30/2021 12:45:22 PM	Finished download of a0ca1ec3e780cf0d8286a242a889eef6.gpf	
24637	World Community Grid	4/30/2021 12:45:22 PM	Started download of d5c255c381c30cdc8d55f941219b8d0e.job	
24638	World Community Grid	4/30/2021 12:45:22 PM	Started download of 15baf0bb509aa5c365c8bff3bdedbea4.zip	
24639			4/30/2021 12:45:23 PM	Project communication failed: attempting access to reference site	
24640	World Community Grid	4/30/2021 12:45:23 PM	Finished download of d5c255c381c30cdc8d55f941219b8d0e.job	
24641	World Community Grid	4/30/2021 12:45:24 PM	Finished download of 15baf0bb509aa5c365c8bff3bdedbea4.zip	
24642			4/30/2021 12:45:25 PM	Internet access OK - project servers may be temporarily down.	
24643	World Community Grid	4/30/2021 12:45:26 PM	Started download of 5e9fd141bb6c2972fac4bf18c362e0ae.gpf	
24644	World Community Grid	4/30/2021 12:45:27 PM	Finished download of 5e9fd141bb6c2972fac4bf18c362e0ae.gpf	
ID: 104185 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104186 - Posted: 30 Apr 2021, 9:50:19 UTC - in response to Message 104184.  

While it is possible, I do not recall that happening and I was an active participants in the betas. I will look though.
ID: 104186 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104187 - Posted: 30 Apr 2021, 9:52:38 UTC - in response to Message 104184.  

That sounds like a related class of problem: if something blocks one part of the request, it doesn't fall through and process the Intel part. Either your 'CPU disabled'. or my 'NV at task limit' is enough to skip processing the intel part of the request.

I think the 'must enable CPU to get intel' variant was reported fairly early in the Beta test, and from memory Uplinger fixed it? Might be worth looking through those Beta messages. I'll try to look through the relevant bit of the server code, but this weekend is beginning to get busy.

Do you mean on the server side or the client side? The log made me think that the client was requesting the intel gpu tasks.
ID: 104187 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 104188 - Posted: 30 Apr 2021, 10:02:49 UTC - in response to Message 104187.  

Do you mean on the server side or the client side? The log made me think that the client was requesting the intel gpu tasks.
I mean on the server side. The client makes the requests as expected, but the server seems to skip that part of the processing.
ID: 104188 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104189 - Posted: 30 Apr 2021, 10:09:48 UTC - in response to Message 104186.  

While it is possible, I do not recall that happening and I was an active participants in the betas. I will look though.

I looked through and did not find anything saying that uplinger had fixed the issue. Just him commenting that he was wondering if it was an issue.
ID: 104189 · Report as offensive
Grumpy Swede
Avatar

Send message
Joined: 30 Mar 20
Posts: 422
Sweden
Message 104190 - Posted: 30 Apr 2021, 10:10:47 UTC
Last modified: 30 Apr 2021, 10:12:07 UTC

I have absolutely no problems getting iGPU tasks. I have always had CPU tasks set to NO.

I have one iGPU, and one discrete GTX980 in the same computer, and I have tested to set even the GTX980 (NVIDIA) to NO, and keep the iGPU to YES. No problems getting tasks for the iGPU then either.
I tried the other way around too, and no problems getting tasks for NVIDIA.

So, in my case, CPU NO, and NVIDIA and iGPU to YES, delivers what both GPU's want. (even if I have no AMD GPU, I have AMD set to YES also)
ID: 104190 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104191 - Posted: 30 Apr 2021, 10:12:58 UTC - in response to Message 104188.  

Do you mean on the server side or the client side? The log made me think that the client was requesting the intel gpu tasks.
I mean on the server side. The client makes the requests as expected, but the server seems to skip that part of the processing.

Ok, thank you. So then the follow up question is: Is this a BOINC server issue or a WCG specific server issue?
If it is WCG specific, I wonder if I should post in the existing GPU support thread or open a new one in so it does not get lost among other issues in the pages.
ID: 104191 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104192 - Posted: 30 Apr 2021, 10:14:58 UTC - in response to Message 104190.  

I have absolutely no problems getting iGPU tasks. I have always had CPU tasks set to NO.

I have one iGPU, and one discrete GTX980 in the same computer, and I have tested to set even the GTX980 (NVIDIA) to NO, and keep the iGPU to YES. No problems getting tasks for the iGPU then either.
I tried the other way around too, and no problems getting tasks for NVIDIA.

So, in my case, CPU NO, and NVIDIA and iGPU to YES, delivers what both GPU's want. (even if I have no AMD GPU, I have AMD set to YES also)

Hi Grumpy_Swede, good to see you over here too! :)
I am only able to reproduce this issue when the iGPU is the only GPU that the machine has.
ID: 104192 · Report as offensive
goben_2003

Send message
Joined: 29 Apr 21
Posts: 50
Message 104193 - Posted: 30 Apr 2021, 10:17:18 UTC - in response to Message 104192.  

I have absolutely no problems getting iGPU tasks. I have always had CPU tasks set to NO.

I have one iGPU, and one discrete GTX980 in the same computer, and I have tested to set even the GTX980 (NVIDIA) to NO, and keep the iGPU to YES. No problems getting tasks for the iGPU then either.
I tried the other way around too, and no problems getting tasks for NVIDIA.

So, in my case, CPU NO, and NVIDIA and iGPU to YES, delivers what both GPU's want. (even if I have no AMD GPU, I have AMD set to YES also)

Hi Grumpy_Swede, good to see you over here too! :)
I am only able to reproduce this issue when the iGPU is the only GPU that the machine has.

No issues getting iGPU units on other machines. In fact the nvidia + intel gpu machine (which has cpu disabled) somehow got 59 intel gpu tasks.
ID: 104193 · Report as offensive
1 · 2 · 3 · 4 . . . 5 · Next

Message boards : Questions and problems : GPU not receiving tasks when CPU computing disabled

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.