Posts by Janus

21) Message boards : Problems and Help : Sorry, I leave this project !! (Message 15453)
Posted 13 Aug 2018 by Profile Janus
Post:
For the same host, same running time, credits vary from 1 to 3.

Oh it can be much much worse than that. Check out some of the graphs from the Payday announcement thread.
22) Message boards : Number crunching : Tasks are not MT (Message 15452)
Posted 13 Aug 2018 by Profile Janus
Post:
Looks like the new plan class may have been missing a flag specifically required to instruct BOINC to pass the number of threads. Really weird that it has been working somewhat for some people and not others.

I've tried cancelling the WU I had that was stuck at 1. The new one I got from the server correctly launched as a proper multithreaded unit. It may be fixed, maybe not, since it seemed a bit random. How does it look at your end?
I don't know if all downloaded WUs are stuck, so I just cancelled everything on my client to get a fresh WU.

You can check very quickly if it launched correctly or not by opening the file "stderr.txt" in the slot directory. If it has "--nthreads detected with value" (and a number) among the first 20 lines or so then it works properly, otherwise it is stuck at 1-thread.
23) Message boards : Number crunching : Estimated runtime, credits,......... (Message 15450)
Posted 13 Aug 2018 by Profile Janus
Post:
Still no any answer from Admin !!!
Nothing !!!


Well, mmonnin is correct. There is not much more to add.

CreditNew is very unreliable for WUs without an estimated GFlop workload. We sometimes run a corrective credit algorithm that brings the variance down considerably.

Two things in the pipeline to change the credit situation:
1) Run Payday more often. Currently it is mostly run when the DB is being purged for big upgrades. It would be nice to be able to run it almost immediately after a session completes, or even replace CreditNew with Payday entirely but it is difficult to do that.
2) Provide a wild guess for runtime estimation up front. This has the to potential to make things better but could also make them way worse.
24) Message boards : Number crunching : Tasks are not MT (Message 15449)
Posted 13 Aug 2018 by Profile Janus
Post:
Really having a hard time nailing this issue but was finally able to reproduce it locally today.

BOINC clearly states that the WU is multithreaded but does not pass the number of threads that the WU should be started with (this differs from client to client depending on settings) to Glue3. Glue3 defaults to 1 thread and passes 1 thread to Blender.
25) Message boards : Comments and discussion : 3456 (Message 15445)
Posted 11 Aug 2018 by Profile Janus
Post:
This renders very quickly (around 0.5 secs per frame), maybe consider rendering this kind of animation locally on your own computer to avoid the delay of submitting it to a renderfarm?
There is a lot of overhead involved both in the review process as well as when sending out the workunits to the computers of the participants who donate resources to the renderfarm. All in all it ends up taking a lot longer than 0.5 seconds.
26) Message boards : Comments and discussion : 3453 (Message 15444)
Posted 11 Aug 2018 by Profile Janus
Post:
This session contains a particle physics simulation but the simulation data was not included in the .blend-file. Unfortunately Blender stores the fluid data on the local machine.

It also renders really fast, maybe you should render this locally instead?

The session was rejected
27) Message boards : Comments and discussion : 3455 (Message 15443)
Posted 11 Aug 2018 by Profile Janus
Post:
This session contains a particle physics simulation but the simulation data was not included in the .blend-file. Unfortunately Blender stores the fluid data on the local machine.

Also it completes almost instantaneously - maybe not the best fit for rendering on a renderfarm?

The session was rejected.
28) Message boards : Comments and discussion : 3442 (Message 15442)
Posted 11 Aug 2018 by Profile Janus
Post:
Check out the results over at #3450, seems to look ok with the new clients that skip unpacking for Cycles sessions.

I just now re-assigned the duplicate session to your user, forgot to do that when it was created as a test.
29) Message boards : Comments and discussion : 3452 (Message 15440)
Posted 3 Aug 2018 by Profile Janus
Post:
This session contains a physics simulation but the simulation data was not included in the .blend-file. Unfortunately Blender stores the fluid data on the local machine.

We're looking to support fluid simulation on the renderfarm but so far it is an unsupported feature unless the fluid data is baked, exported and then included in the .blend-file.

The session was rejected.
30) Message boards : Comments and discussion : 3450 (Message 15438)
Posted 31 Jul 2018 by Profile Janus
Post:
The OpenCL test WU that was sent out for this session didn't seem to succeed at all. No reply was received from any of the machines that received it.
It will take a little while longer for the WUs to time out and hopefully, by then, some logs will show what the problem is.

Unfortunately I don't personally have modern enough OpenCL hardware available for testing yet.
31) Message boards : Number crunching : Tasks are not MT (Message 15435)
Posted 30 Jul 2018 by Profile Janus
Post:
I've been trying to narrow down this issue further, it seems BOINC sometimes simply doesn't send the multithreading parameter to Glue3 and in that case Glue3 will default to 1 thread. What is weird is that this can happen differently even for two instances of the same WU like in this case where one host rendered with multithreading and the other didn't. The very top of the debug output shows the parsed parameters and in one case --nthreads is simply missing.

Still looking into this.
32) Message boards : Problems and Help : Process exited with code 253 on Ubuntu 18.04. Level 17.10 works. (Message 15434)
Posted 30 Jul 2018 by Profile Janus
Post:
It looks like a fix can be sent out without having to generate a new client, this will be tested in the next session.
33) Message boards : Problems and Help : Process exited with code 253 on Ubuntu 18.04. Level 17.10 works. (Message 15432)
Posted 30 Jul 2018 by Profile Janus
Post:
Confirmed the issue on Ubuntu 18.04 live CD.
LC_ALL=C apparently avoids the issue.

Looking into it.
34) Message boards : Number crunching : Tasks are not MT (Message 15431)
Posted 30 Jul 2018 by Profile Janus
Post:
Please keep an eye on this over the next few sessions. Recently the plan classes (that contain data about how BOINC should schedule work on clients) and the code relating to them were updated to include a new OpenCL plan-class that we are testing out these days (unfortunately the OpenCL client seems to be broken...). During the change the MT plan class was also upgraded from a max of 64 threads to a max of 128 threads. It is possible that something weird is going on after this change.

Keep in mind that the first few minutes and the last part of WU will almost always be either single-threaded or very few cores as Blender is reading the scene file and performing final postprocessing respectively.

Also keep a keen eye on the type of WU, there has been a few GPU test units recently - they use just one CPU thread but 100% of a GPU, which is similar to what you describe but they do say that they are GPU tasks and not MT like yours.

[Edit] I can confirm that the WUs in question unintentionally contain an instruction that limits the number of active threads to 1. Looking into why this is happening.
35) Message boards : Comments and discussion : 3450 (Message 15424)
Posted 29 Jul 2018 by Profile Janus
Post:
Great to hear!

There's still some CPU used during the start-up and up to 1 full thread for managing the GPU.
36) Message boards : Server backend and mirrors : Scheduler, feeder (Message 15423)
Posted 29 Jul 2018 by Profile Janus
Post:
Most of the scheduling issues have been fixed and the service restarted.

I'll look into the remaining issue with a few stuck WUs tomorrow afternoon.
37) Message boards : Problems and Help : Process exited with code 253 on Ubuntu 18.04. Level 17.10 works. (Message 15422)
Posted 29 Jul 2018 by Profile Janus
Post:
We usually don't see issues with the unzip process. Since the issue happened in "loadlocale" what languages is this system configured with?
38) Message boards : Comments and discussion : 3450 (Message 15419)
Posted 29 Jul 2018 by Profile Janus
Post:
Using this session to slowly and carefully run a few more GPU tests:
1) The code for GPU indexing has been updated to match the most recent changes in how Cycles uses GPUs
2) Preliminary test of OpenCL support for AMD GPUs
3) Test of configuring client binaries across multiple platforms inside BOINC using just one set of files
39) Message boards : Server backend and mirrors : Scheduler, feeder (Message 15416)
Posted 29 Jul 2018 by Profile Janus
Post:
Update: Still looking into scheduler issues. Fixed the issue with CPU units not going out and now instead have an issue with reported results not getting properly registered as having come back.
40) Message boards : Server backend and mirrors : Scheduler, feeder (Message 15413)
Posted 27 Jul 2018 by Profile Janus
Post:
Taking the site down for a few hours to investigate the scheduling issues without workunits flowing through it at the same time.


Previous 20 · Next 20