Page 2 of 3

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:08 am
by Joe_H
Your log shows a CPU folding core running a CPU project WU on your CPU. Not a GPU job at all. Somehow the client managed to failsafe from your attempt to configure a slot for your Intel iGPU. There is no folding core that the systems would download to your PC process a GPU WU on your iGPU.

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:13 am
by JimboPalmer
-gpu-vendor amd

I think this says it all

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:18 am
by jetchko.jekov
And here is the log from my notebook.
Notice it even says: OpenCL: Not detected: clGetDeviceIDs() returned -1. And yet it runs ...

Code: Select all

*********************** Log Started 2020-04-11T17:16:57Z ***********************
17:16:57:************************* Folding@home Client *************************
17:16:57:    Website: https://foldingathome.org/
17:16:57:  Copyright: (c) 2009-2018 foldingathome.org
17:16:57:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
17:16:57:       Args: --child --lifeline 29653 /etc/fahclient/config.xml --run-as
17:16:57:             fahclient --pid-file=/var/run/fahclient.pid --daemon
17:16:57:     Config: /etc/fahclient/config.xml
17:16:57:******************************** Build ********************************
17:16:57:    Version: 7.5.1
17:16:57:       Date: May 12 2018
17:16:57:       Time: 22:51:07
17:16:57: Repository: Git
17:16:57:   Revision: 4705bf53c635f88b8fe85af7675557e15d491ff0
17:16:57:     Branch: master
17:16:57:   Compiler: GNU 4.4.7 20120313 (Red Hat 4.4.7-18)
17:16:57:    Options: -std=gnu++98 -O3 -funroll-loops
17:16:57:   Platform: linux2 4.14.0-3-amd64
17:16:57:       Bits: 64
17:16:57:       Mode: Release
17:16:57:******************************* System ********************************
17:16:57:        CPU: Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz
17:16:57:     CPU ID: GenuineIntel Family 6 Model 142 Stepping 9
17:16:57:       CPUs: 4
17:16:57:     Memory: 31.30GiB
17:16:57:Free Memory: 23.57GiB
17:16:57:    Threads: POSIX_THREADS
17:16:57: OS Version: 5.5
17:16:57:Has Battery: true
17:16:57: On Battery: false
17:16:57: UTC Offset: 2
17:16:57:        PID: 29655
17:16:57:        CWD: /var/lib/fahclient
17:16:57:         OS: Linux 5.5.15-200.fc31.x86_64 x86_64
17:16:57:    OS Arch: AMD64
17:16:57:       GPUs: 1
17:16:57:      GPU 0: Bus:0 Slot:2 Func:0 AMD:0 [Intel Corporation HD Graphics 620]
17:16:57:       CUDA: Not detected: Failed to open dynamic library 'libcuda.so':
17:16:57:             libcuda.so: cannot open shared object file: No such file or
17:16:57:             directory
17:16:57:     OpenCL: Not detected: clGetDeviceIDs() returned -1
17:16:57:***********************************************************************
[ -- cut -- ]
7:16:58:WU00:FS01:0xa7:*********************** Log Started 2020-04-11T17:16:57Z ***********************
17:16:58:WU00:FS01:0xa7:************************** Gromacs Folding@home Core ***************************
17:16:58:WU00:FS01:0xa7:       Type: 0xa7
17:16:58:WU00:FS01:0xa7:       Core: Gromacs
17:16:58:WU00:FS01:0xa7:       Args: -dir 00 -suffix 01 -version 705 -lifeline 29682 -checkpoint 15
17:16:58:WU00:FS01:0xa7:             -gpu-vendor amd -opencl-device 2 -gpu 2
17:16:58:WU00:FS01:0xa7:************************************ CBang *************************************
17:16:58:WU00:FS01:0xa7:       Date: Nov 5 2019
17:16:58:WU00:FS01:0xa7:       Time: 06:06:57
17:16:58:WU00:FS01:0xa7:   Revision: 46c96f1aa8419571d83f3e63f9c99a0d602f6da9
17:16:58:WU00:FS01:0xa7:     Branch: master
17:16:58:WU00:FS01:0xa7:   Compiler: GNU 8.3.0
17:16:58:WU00:FS01:0xa7:    Options: -std=c++11 -O3 -funroll-loops -fno-pie -fPIC
17:16:58:WU00:FS01:0xa7:   Platform: linux2 4.19.0-5-amd64
17:16:58:WU00:FS01:0xa7:       Bits: 64
17:16:58:WU00:FS01:0xa7:       Mode: Release
17:16:58:WU00:FS01:0xa7:************************************ System ************************************
17:16:58:WU00:FS01:0xa7:        CPU: Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz
17:16:58:WU00:FS01:0xa7:     CPU ID: GenuineIntel Family 6 Model 142 Stepping 9
17:16:58:WU00:FS01:0xa7:       CPUs: 4
17:16:58:WU00:FS01:0xa7:     Memory: 31.30GiB
17:16:58:WU00:FS01:0xa7:Free Memory: 23.55GiB
17:16:58:WU00:FS01:0xa7:    Threads: POSIX_THREADS
17:16:58:WU00:FS01:0xa7: OS Version: 5.5
17:16:58:WU00:FS01:0xa7:Has Battery: true
17:16:58:WU00:FS01:0xa7: On Battery: false
17:16:58:WU00:FS01:0xa7: UTC Offset: 2
17:16:58:WU00:FS01:0xa7:        PID: 29686
17:16:58:WU00:FS01:0xa7:        CWD: /var/lib/fahclient/work
17:16:58:WU00:FS01:0xa7:******************************** Build - libFAH ********************************
17:16:58:WU00:FS01:0xa7:    Version: 0.0.18
17:16:58:WU00:FS01:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
17:16:58:WU00:FS01:0xa7:  Copyright: 2019 foldingathome.org
17:16:58:WU00:FS01:0xa7:   Homepage: https://foldingathome.org/
17:16:58:WU00:FS01:0xa7:       Date: Nov 5 2019
17:16:58:WU00:FS01:0xa7:       Time: 06:13:26
17:16:58:WU00:FS01:0xa7:   Revision: 490c9aa2957b725af319379424d5c5cb36efb656
17:16:58:WU00:FS01:0xa7:     Branch: master
17:16:58:WU00:FS01:0xa7:   Compiler: GNU 8.3.0
17:16:58:WU00:FS01:0xa7:    Options: -std=c++11 -O3 -funroll-loops -fno-pie
17:16:58:WU00:FS01:0xa7:   Platform: linux2 4.19.0-5-amd64
17:16:58:WU00:FS01:0xa7:       Bits: 64
17:16:58:WU00:FS01:0xa7:       Mode: Release
17:16:58:WU00:FS01:0xa7:************************************ Build *************************************
17:16:58:WU00:FS01:0xa7:       SIMD: avx_256
17:16:58:WU00:FS01:0xa7:********************************************************************************
17:16:58:WU00:FS01:0xa7:Project: 13861 (Run 0, Clone 18728, Gen 47)
17:16:58:WU00:FS01:0xa7:Unit: 0x000000330d5a98395e730b0e4bbfc1fe
17:16:58:WU00:FS01:0xa7:Digital signatures verified
17:16:58:WU00:FS01:0xa7:Calling: mdrun -s frame47.tpr -o frame47.trr -x frame47.xtc -e frame47.edr -cpi state.cpt -cpt 15 -nt 1
17:16:58:WU00:FS01:0xa7:Steps: first=5875000 total=125000
17:16:59:WU00:FS01:0xa7:Completed 172 out of 125000 steps (0%)
17:20:11:WU00:FS01:0xa7:Completed 1250 out of 125000 steps (1%)
[ -- cut -- ]
22:58:41:WU00:FS01:0xa7:Completed 121250 out of 125000 steps (97%)
23:01:06:WU00:FS01:0xa7:Completed 122500 out of 125000 steps (98%)
23:03:29:WU00:FS01:0xa7:Completed 123750 out of 125000 steps (99%)
23:05:55:WU00:FS01:0xa7:Completed 125000 out of 125000 steps (100%)
23:05:56:WU00:FS01:0xa7:Saving result file ../logfile_01.txt
23:05:56:WU00:FS01:0xa7:Saving result file frame47.edr
23:05:56:WU00:FS01:0xa7:Saving result file frame47.trr
23:05:56:WU00:FS01:0xa7:Saving result file frame47.xtc
23:05:56:WU00:FS01:0xa7:Saving result file md.log
23:05:56:WU00:FS01:0xa7:Saving result file science.log
23:05:56:WU00:FS01:0xa7:Folding@home Core Shutdown: FINISHED_UNIT
23:05:56:WU00:FS01:FahCore returned: FINISHED_UNIT (100 = 0x64)
23:05:56:WU00:FS01:Sending unit results: id:00 state:SEND error:NO_ERROR project:13861 run:0 clone:18728 gen:47 core:0xa7 unit:0x000000330d5a98395e730b0e4bbfc1fe
23:05:56:WU00:FS01:Uploading 5.28MiB to 13.90.152.57
23:05:56:WU00:FS01:Connecting to 13.90.152.57:8080
23:06:02:WU00:FS01:Upload 62.73%
23:06:06:WU00:FS01:Upload complete
23:06:06:WU00:FS01:Server responded WORK_ACK (400)
23:06:06:WU00:FS01:Final credit estimate, 1129.00 points
23:06:06:WU00:FS01:Cleaning up

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:21 am
by jetchko.jekov
Joe_H wrote:Your log shows a CPU folding core running a CPU project WU on your CPU. Not a GPU job at all. Somehow the client managed to failsafe from your attempt to configure a slot for your Intel iGPU. There is no folding core that the systems would download to your PC process a GPU WU on your iGPU.
Hm? care to point out where is that exactly?

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:25 am
by Joe_H
You are using Core_A7, that is a CPU only folding core. It does not run on a GPU.

Project 13861 is a CPU project.

Again, THERE IS NO GPU FOLDING CORE FOR INTEL. I am writing that way to emphasize you can't use what is not programmed for being used.

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:29 am
by jetchko.jekov
Calm down, please and explain the meaning of this:

Code: Select all

05:48:32:WU00:FS01:0xa7:************************** Gromacs Folding@home Core ***************************
05:48:32:WU00:FS01:0xa7:       Type: 0xa7
05:48:32:WU00:FS01:0xa7:       Core: Gromacs
05:48:32:WU00:FS01:0xa7:       Args: -dir 00 -suffix 01 -version 705 -lifeline 43552 -checkpoint 15
05:48:32:WU00:FS01:0xa7:             -gpu-vendor amd -opencl-device 2 -gpu 2
05:48:32:WU00:FS01:0xa7:************************************ CBang *************************************
Also, there is no need to have a folding core especially for Intel GPUs, folding core for OpenCL is enough. Which Vendor GPU implements it should not matter.

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:47 am
by Joe_H
The CPU core is ignoring the switches that do not apply to it, here is what he A7 core is actually using:

Code: Select all

17:16:58:WU00:FS01:0xa7:Calling: mdrun -s frame47.tpr -o frame47.trr -x frame47.xtc -e frame47.edr -cpi state.cpt -cpt 15 -nt 1
Look, you are not the first person who thought they could misconfigure the the client and force it to do something it was not meant to do. Usually they end up with something that results in the request for a WU being given a "Unable to assign" message. IN your case it was only off enough so it defaulted to sending your system a CPU WU, and that was run as a CPU WU on your system.

What ended up asa command to run it was, and you clipped that out of the second log:

Code: Select all

11:12:24:WU01:FS01:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/cores.foldingathome.org/v7/lin/64bit/avx/Core_a7.fah/FahCore_a7 -dir 01 -suffix 01 -version 705 -lifeline 1244 -checkpoint 15 -gpu-vendor amd -opencl-platform 0 -opencl-device 2 -gpu 2
Everything was normal up to the '-checkpoint' parameter, the core wrapper process ignored everything after that, and assigned default values. For instance, it ran on a single CPU thread '-nt 1'.

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 6:58 am
by jetchko.jekov
Running processes in the system. I don't see wrapper ignoring anything.

Code: Select all

  43552 ?        SNl    0:00 /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/cores.foldingathome.org/v7/lin/64bit/avx/Core_a7.fah/FahCore_a7 -dir 00 -suffix 01 -version 705 -lifeline 34914 -checkpoint 15 -gpu-vendor amd -opencl-device 2 -gpu 2
  43556 ?        RNl   65:13 /var/lib/fahclient/cores/cores.foldingathome.org/v7/lin/64bit/avx/Core_a7.fah/FahCore_a7 -dir 00 -suffix 01 -version 705 -lifeline 43552 -checkpoint 15 -gpu-vendor amd -opencl-device 2 -gpu 2
  43560 ?        SNl    0:00 /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/cores.foldingathome.org/v7/lin/64bit/avx/Core_a7.fah/FahCore_a7 -dir 01 -suffix 01 -version 705 -lifeline 34914 -checkpoint 15 -np 2
  43564 ?        RNl  130:27 /var/lib/fahclient/cores/cores.foldingathome.org/v7/lin/64bit/avx/Core_a7.fah/FahCore_a7 -dir 01 -suffix 01 -version 705 -lifeline 43560 -checkpoint 15 -np 2
See, I am just curious, I want to understand.

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 7:03 am
by Joe_H
It may have those flags, but it is not passing them to the actual folding core, you will see that process as FAHCore_A7.

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 7:17 am
by jetchko.jekov
exactly
look at 2nd line

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 7:30 am
by Joe_H
The log shows what the core used, I have already posted that above. You can pass all kinds of garbage as input to programs, what they accepts another matter. In the case the core running showed exactly what it used in the log.

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 7:37 am
by jetchko.jekov
OK, thanks for the patience

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 9:32 am
by jetchko.jekov
Just 2 more questions:
How is it decided what core to be run on a particular slot?
Why is CPU only core run on GPU slot?

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 11:12 am
by JimboPalmer
Currently all CPUs run the Core_a7, although there are two branches it might take, SSE2 and avx_256.
https://en.wikipedia.org/wiki/SSE2
https://en.wikipedia.org/wiki/Advanced_ ... Extensions

Modern CPUs will use avx_256.

GPUs run one of two Cores

Core_21 has many outstanding projects started with it, you would not change in mid stream, so until those project complete, you will sometimes see Core_21 WUs.

The newest AMD cards (code names like Navi and RDNA, models like 5x00 XT) were incompatible with Core_21 and a newer Core_22 was written, since it is newer, it does better science, and all new Projects are run with it. (all the COVID-19 WUs are Core_22, as they are all new, when we get back to treating cancer, etc. expect Core_21 again and slightly lower PPD)

https://en.wikipedia.org/wiki/List_of_F ... home_cores

Re: Using INTEL GPUs

Posted: Mon Apr 13, 2020 11:27 am
by Neil-B
jetchko.jekov wrote:Just 2 more questions:
How is it decided what core to be run on a particular slot?
Why is CPU only core run on GPU slot?
In a normal setup the core slots are defined by the client - If a setup has been modified the client still tries to get the right core for the right compute (or so it seems in your case) or it simply doesn't work and throws an error.

The scientists decide whether their project will use CPU, GPU and they set their projects up for this ... The AS looks at the incoming request for work and allocates to a WS with CPU WUs if it is CPU resource and to a WS with GPU WUs if it is a GPU resource ... This is complicated by many factors which the client manages - number of CPU cores available, Type of Graphics Card, etc. ... In standard configuration the client and the whole process works relatively well but when attempts to modify the way the client works beyond the intended limits that is when this can go awry ... In your case the AS managed to actually send you an extra CPU WU to run on a second CPU slot and ignore the extraneous information re GPU (which to be honest I am surprised it managed this).

So to answer your second question - it didn't, it can't, it won't - what you are seeing (as far as I understand it) is the system running two CPU slots ... but I am sure the developer (when he has sorted all the other infrastructure and workflow issues) may want to have a look at this and ensure it doesn't happen again.