Posts by Agentb

InfoMessage
1) Message boards : News : Project management changes proposed
Message 80757
Posted 5 Sep 2017 by ProfileAgentb
Perhaps I'm not understanding how the new document works, apologies i have only spent 20 minutes trying to find the changes.

Can someone show me the changes are between current and v2?

The wiki shows a number of edits of v2 as far as i can see, not the changes from the existing document.

Returning to accreditation of authors. The members of the Committee (assuming all contributed) should be listed and given credit. If BOINC is to be a meritocracy, then credit for the work needs to be given to people, it is their work provided freely i expect, and they justly deserve credit. If anyone contributes they should be credited.

It is good practice in Open Source projects (and probably all things) to give credit and list authors individually. Every science paper and RFC credits its authors.

If a new document is being created to address an existing problem, then perhaps a "History" explaining the current process and problems - the one sentence in Kevin's post is very important, and describes why a new document is needed.
2) Message boards : News : Project management changes proposed
Message 80709
Posted 3 Sep 2017 by ProfileAgentb
In July 2017 a committee was formed to propose improvements to the BOINC project. Two draft documents are now available for discussion and review.

First, thanks for proposing the drafts, i know a lot of effort has gone in, i'm sure there are multiple contributors and some consensus. Often the discussion better describes the change. Authors and contributors should always be credited.

If the documents are draft / approved etc, can this be immediately clear from the document?

Publishing them as web pages is probably not conducive to change control. Searching for Governance for example via a search on the boinc website reveals the draft first and "master" second.

The current "master" Governance version i believe is via a link from here http://boinc.berkeley.edu/trac/wiki/ProjectGovernance

I don't know what the "master" of the development work flow is maybe here http://boinc.berkeley.edu/trac/wiki/DevProcess ?

Both documents Governance and Development should be under source and project control.

What is the process (now / future) for changes to them?

Both draft documents do not show the changes clearly from original, and if changes are proposed, those changes should be made clear.

Can the changes from current masters please be shown clearly?

I know it's a minor thing but grammatical style. Some of the new sections in the Governance document switches to "second person" . Easily found by searching for "you ".
3) Message boards : GPUs : [Solved] GPU detection failed. error code 512
Message 80390
Posted 21 Aug 2017 by ProfileAgentb
I'm not sure why the normal install is not working for you - i guess you started from here https://wiki.archlinux.org/index.php/BOINC ?

There are some suggestions about GPU detection errors mentioned there as well.
4) Message boards : Projects : For the betterment of BOINC
Message 80333
Posted 19 Aug 2017 by ProfileAgentb
..still though a lack of transparency is disappointing.
Most all information will become available when we're done. It's not that it's so super secret what we're doing, but a lot of what we come up with and what's in the agenda and minutes, may not even make it into the end result. Just making sure we're not giving all kinds of false hope.

Transparency is very important i feel for an Open Source project.

Raising hopes is engagement! False hopes occur when expectations are raised which are not delivered, and lack of information is a "non-delivery".

I also hoped for more openness.
5) Message boards : Questions and problems : Moving BOINC to a new OS installation on Linux Mint.
Message 80315
Posted 18 Aug 2017 by ProfileAgentb
Probably the easiest way to create the boinc group and user - the usermod just tidies upa description field

sudo adduser --system --group boinc  --home /var/lib/boinc-client/
sudo usermod  --comment "BOINC Core user" boinc

This should create the home directory and the user / group records correctly, although the install normally does that.
edit: Juha beat me to it! You may need to add boinc to the video group if you plan to use the GPU for crunching
6) Message boards : News : BOINC Workshop 2017
Message 79804
Posted 26 Jul 2017 by ProfileAgentb
https://travel.state.gov/content/passports/en/go/schengen-fact-sheet.html

I believe in total reciprocity... don't think USA requires visa for stay duration up to 3 months for citizens of the EU Schengen area either.

The US requires a visa for EU citizens of Bulgaria, Croatia, Cyprus, Poland and Romania which are all part of the Schengen area.

This I believe is the main reason behind the the law change.
7) Message boards : Questions and problems : Running on an SSD
Message 79738
Posted 23 Jul 2017 by ProfileAgentb
To give you an idea about how SSD have improved, i have been running an SSD 7x24 crunching at Einstein for over two years and the SMART wear level indicator suggests i am at 96% meaning at the current rate i have ~20 years left before it wears out! I suspect it will die for other reasons long before that. It you are really concerned, just ensure you have plenty of RAM so that paging or swapping does not occur, and that is a more a general recommendation rather than a boinc specific one.

Very few projects put any stress on disk components, but you can reduce that further by reducing checkpoint frequency.

Almost all of the heat generated will be GPU or CPU (or the PSU itself) so keeping other components away from the exhaust of these is good design.

HTH.
8) Message boards : Questions and problems : GPU detection failed. error code 139
Message 79643
Posted 16 Jul 2017 by ProfileAgentb
I had a brainwave and removed VERY carefully all the extraneous NVIDIA packages and this was the result:

The problem me thinks is solved!


OK that's good news, but i'm curious why the nvidia packages were installed.
9) Message boards : Questions and problems : GPU detection failed. error code 139
Message 79637
Posted 16 Jul 2017 by ProfileAgentb
How exactly did you install boinc?

I used the synaptic package manager

OK that is what you used to do the installing, but exactly what did you install with synaptic? There are several boinc (and other stuff) things you could have installed with synaptic. I think you can view the logs through the synaptic gui.

Would it be an idea to dig around for extraneous nvidia packages?
yes - but this

# ldconfig -p | grep -i nvidia


should show nothing. If it shows a lot of libraries then you have installed an nVidia package and that needs removing, especially if it shows a libOpenCL.so

echo $LD_LIBRARY_PATH


should also show nothing.
10) Message boards : Questions and problems : GPU detection failed. error code 139
Message 79634
Posted 16 Jul 2017 by ProfileAgentb
How exactly did you install boinc?

sudo apt-get install ....

for some reason you have specific nVidia icds / libraries reporting

check the icds are correct

# cd /etc/OpenCL/vendors

you should see only two files
amdocl32.icd  amdocl64.icd


and the content of each file will be a library name eg libamdocl64.so in the second.

check existence of these libraries in the library cache - you should see

# ldconfig -p | grep amdocl
	libamdocl64.so (libc6,x86-64) => /opt/amdgpu-pro/lib/x86_64-linux-gnu/libamdocl64.so
	libamdocl32.so (libc6) => /opt/amdgpu-pro/lib/i386-linux-gnu/libamdocl32.so
	libamdocl12cl64.so (libc6,x86-64) => /opt/amdgpu-pro/lib/x86_64-linux-gnu/libamdocl12cl64.so
	libamdocl12cl32.so (libc6) => /opt/amdgpu-pro/lib/i386-linux-gnu/libamdocl12cl32.so
11) Message boards : Questions and problems : GPU detection failed. error code 139
Message 79630
Posted 16 Jul 2017 by ProfileAgentb
Have done the driver upgrade and the file is still missing.

Am now "googling" libcaticalart.so
Don't do that. The messages indicate it can not find libraries for CAL and nVidia, and i don't think you have (or need) them.

You should see something like

Sun 16 Jul 2017 18:34:11 BST |  | Starting BOINC client version 7.6.33 for x86_64-pc-linux-gnu
Sun 16 Jul 2017 18:34:11 BST |  | log flags: file_xfer, sched_ops, task, coproc_debug
Sun 16 Jul 2017 18:34:11 BST |  | Libraries: libcurl/7.47.0 OpenSSL/1.0.2g zlib/1.2.8 libidn/1.32 librtmp/2.3
Sun 16 Jul 2017 18:34:11 BST |  | Data directory: /var/lib/boinc-client
Sun 16 Jul 2017 18:34:11 BST |  | [coproc] launching child process at /usr/bin/boinc
Sun 16 Jul 2017 18:34:11 BST |  | [coproc] relative to directory /
Sun 16 Jul 2017 18:34:11 BST |  | [coproc] with data directory /var/lib/boinc-client
Sun 16 Jul 2017 18:34:11 BST |  | OpenCL: AMD/ATI GPU 0: AMD Radeon (TM) RX 480 Graphics (driver version 2348.3, device version OpenCL 1.2 AMD-APP (2348.3), 5819MB, 5819MB available, 5935 GFLOPS peak)
Sun 16 Jul 2017 18:34:11 BST |  | [coproc] NVIDIA: libcuda.so: cannot open shared object file: No such file or directory
Sun 16 Jul 2017 18:34:11 BST |  | [coproc] ATI: libaticalrt.so: cannot open shared object file: No such file or directory
Sun 16 Jul 2017 18:34:11 BST |  | [coproc] clGetDeviceInfo failed to get CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD for device 0


I'll ask again, could you run clinfo and post the output.
12) Message boards : Questions and problems : GPU detection failed. error code 139
Message 79626
Posted 16 Jul 2017 by ProfileAgentb
boinc need to be part of the video group, /etc/group should show a line like
video:x:44:boinc,agentb


code 139 is congruent to the segfault.

a) # clinfo may be revealing along with coproc_debug event log flag

and

b) there are newer drivers at AMD (as mentioned by Ritchie)

Good luck

Edit: I am having no problem with 4.4
13) Message boards : BOINC client : The old scheduling problem strikes again
Message 79616
Posted 15 Jul 2017 by ProfileAgentb
And why is it removing the task from memory in violation of the leave suspended tasks in memory setting, is it to clear the gpu's memory?
Yes, GPU apps are always removed from memory when suspended. GPUs don't have the facility to swap stale memory images out to a paging file on disk, so a suspended task would always continue to occupy real, physical, memory, which might be in short supply.

Am i right in saying, the GPU task actually falls back to last checkpoint and restarts from that point?
14) Message boards : The Lounge : These are no longer with us, may they have peace.
Message 79598
Posted 13 Jul 2017 by ProfileAgentb
Liu Xiaobo, 61. Writer, human rights activist and Nobel Peace Prize laureate.
15) Message boards : GPUs : BOINC can't detect RX 480
Message 79525
Posted 10 Jul 2017 by ProfileAgentb
ROCm is probably worth looking at. If i had some time i might try to get it working. There are some articles on phoronix about it.

I am not clear why you are not running boinc 7.6.33, GPU detection is more forgiving in recent versions.
16) Message boards : The Lounge : The Seti is Slumbering Cafe
Message 79473
Posted 5 Jul 2017 by ProfileAgentb
Now, what shall I go play now? Mafia 3, Skyrim the Special Edition, Elder Scrolls Morrowind, or Tier 1 of Wildlands? Sigh. :)


Hmm i'd suggest giving the digeridoo a rest until you throat gets better. You'd be the only didj player at the festival.
17) Message boards : Questions and problems : Moving BOINC data file folder isn't working
Message 79441
Posted 2 Jul 2017 by ProfileAgentb
Personally, i'd would return everything back to normal and check working / you haven't broken ownership and permissions. I would not put it in my home directory - for several reasons.

Juha's suggestion in this related thread is how i have done and would do a move.

If you have sudo privs then if the second drive is mounted as /hdd1 then .
# Stop boinc then
cd /var/lib
sudo cp -rp boinc-client /hdd1/boinc-client 
sudo mv boinc-client boinc-backup
sudo ln -s /hdd1/boinc-client

# restart boinc

of course you need to understand what these do and work out a recovery if it goes wrong.

good luck!
18) Message boards : Questions and problems : Optimising AMD cards for maximum efficiency
Message 79072
Posted 16 Jun 2017 by ProfileAgentb
, I have several (more thirsty) quad cores like the Q9650 that are currently doing nothing. Is it worth it or needed to have those extra cores?
http://ark.intel.com/products/35428/Intel-Core2-Quad-Processor-Q9650-12M-Cache-3_00-GHz-1333-MHz-FSB

To repeat - you need to look at the applications you want to run - some apps use virtually no CPU, and offload it all to the GPU. I can't help you with that, you have to make a choice.

You are worried about maybe 20W on a CPU with 2x180W GPUS smoking away in the case (or more) ? Really?

I think for the same load the quad core would use a lot less power and be more efficient per watt. The quad core has more L2 cache and higher bus speed.

Some apps benefit a lot from a fast CPU to drive the GPU - pushing data in and out of RAM, you can't make a steam engine go fast with a plastic teaspoon to shovel the coal.
19) Message boards : Questions and problems : Optimising AMD cards for maximum efficiency
Message 79052
Posted 15 Jun 2017 by ProfileAgentb
You will get many different opinions and all will be right!

My 0.02 gbp

Firstly different applications use GPUs in very very different ways and different configurations and operating systems will be optimal.

The rule of thumb leaving a CPU free is just that.

I have seen as many as 8 GPUs however even with two GPUs, a case (and then room) cooling becomes THE issue.

As the number of cards increase the PCI / CPU / RAM bottlenecks start to have an impact.

Quad channel 4x2GB memory will be a lot faster than 1x8GB for certain apps.

All i would say is develop your method this is roughly what i do

Stability is the main thing for me restarts and crashes are bad, for all reasons.

Look at the applications you want to run and which suit your hardware / memory/ VRAM etc read forums etc

THEN

- choose OS you are confident in
- be absolutely certain the cooling and power are up to the task - i can't stress this enough. Overspend on PSU and case cooling. Efficient PSUs pay for themselves and reduce case and room temperatures.
- start from a simple base line - run each application / GPU solo to get a baseline for each application
- record stable run (and validation success) times over steady state.
- Monitor temps, CPU, memory and power usage - build a spreadsheet and this will help over the life - you can go back to see what the temps were when you started and see if it is running hotter or slower.
- if you are getting invalids or errors stop and find out why - you can't tune a faulty engine.
- increase task loads and keep recording run times and temps etc.
- the fewer GPU apps you run the more stable. For example you may find running project X one month then project Y the next is more stable than running X and Y at the same time.
- overclock by small amounts over long periods (if you must).
- stability trumps speed. %(errors+invalids) should be <0.5% imho. If you monitor one thing monitor this stat.
- ignore CPU crunching until you have the GPU side optimised - and what optimal exactly means depends on what you want. Power / temp / times etc
- repeat for CPU tasks disabling GPU.
- return to the GPU and start adding the CPU tasks and see how they affect GPU times.

- call it done, and leave it you can spend too much time...

good luck
20) Message boards : Promotion : An Idea about promoting BOINC in Local School System
Message 79038
Posted 15 Jun 2017 by ProfileAgentb
Another idea is to set up a team for each class or each school, then you would be able to see each team members contribution.

For example

https://einsteinathome.org/community/stats/teams

You can drill into each team and see membership contribution

Statistic sites are described here and they can show different data.

http://boinc.berkeley.edu/links.php#stats

You need to be aware different projects assign values to credit differently, so comparing different projects is similar to comparing currencies.

Good luck.
Next 20

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.