Setting up Linux to crunch CUDA90 and above for Windows users

Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 127 · 128 · 129 · 130 · 131 · 132 · 133 . . . 162 · Next

AuthorMessage
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 2008594 - Posted: 21 Aug 2019, 6:56:50 UTC

@TBar

. . Hi, I don't know if this is significant to you but I have an update on the failure to terminate BoincManager and Boinc Client when exiting the manager window. I normally use the icon in the task bar to close the app. Right click, bring up the exit dialogue box, select terminate the client and then select close. This NEVER works in the 7.14.2 version. The Manager Window closes but the manager and the client apps still run in the background. However, I have discovered that if I use the Manager menu options instead it works, with a short delay. So now I select files from the menu bar, then exit Boinc Manager and then end the client in the exit dialog and voila, it actually works. So the issue seems to be related to the path via the task bar icon. For what that is worth.

. . I hope that helps someone, somehow ...

Stephen

? ?
ID: 2008594 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2008598 - Posted: 21 Aug 2019, 8:42:08 UTC - in response to Message 2008594.  

I thought Richard said the icon in the Tray was a kludge that hadn't worked in a long time. Closing via the Manager dialog has always been the preferred and correct way as far as I remember. The tray icon was built with the old deprecated WxWidgets I believe.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2008598 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2008604 - Posted: 21 Aug 2019, 10:03:00 UTC - in response to Message 2008598.  

I thought Richard said the icon in the Tray was a kludge that hadn't worked in a long time. Closing via the Manager dialog has always been the preferred and correct way as far as I remember. The tray icon was built with the old deprecated WxWidgets I believe.
I don't think I said that.

As a primarily Windows user, the tray icon - labelled 'Exit' in the Windows build - works as intended: invoking the exit dialog (if enabled), and stopping the client (if that option is selected).

The description of the Linux icon as displaying a 'Close' option is suspicious: that implies a change in visibility, rather than functionality.

Although I have a Linux machine now (Mint/Ubuntu), I can't test that function: I'm running BOINC v7.16.1 take 2, and as threatened the tray icon and the menu control options have been removed. I intend to make a fuss about that tomorrow evening.
ID: 2008604 · Report as offensive     Reply Quote
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 2008638 - Posted: 21 Aug 2019, 15:55:41 UTC - in response to Message 2008604.  

I thought Richard said the icon in the Tray was a kludge that hadn't worked in a long time. Closing via the Manager dialog has always been the preferred and correct way as far as I remember. The tray icon was built with the old deprecated WxWidgets I believe.
I don't think I said that.

As a primarily Windows user, the tray icon - labelled 'Exit' in the Windows build - works as intended: invoking the exit dialog (if enabled), and stopping the client (if that option is selected).

The description of the Linux icon as displaying a 'Close' option is suspicious: that implies a change in visibility, rather than functionality.

Although I have a Linux machine now (Mint/Ubuntu), I can't test that function: I'm running BOINC v7.16.1 take 2, and as threatened the tray icon and the menu control options have been removed. I intend to make a fuss about that tomorrow evening.

. . OK, my bad. I was tired (as usual) and that was a quick off the cuff description. If I use The 'Files' option from the menu bar I can invoke the 'Exit Boinc' option to bring up the exit dialog box, Ctrl-Q does the same. Right clicking on the task bar "kludge" :) does nothing if I just click on it, the options sub window closes as soon as I release the button. But if I hold the button while moving the mouse I can select the 'Exit' option (to me Exit and close are synonymous) to also bring up the exit dialog. But this is the strange part, when the exit dialog is called from the menu bar or via Ctrl-Q it works, but when called from the task bar icon it fails to end the apps. Which is why I was surprised that it worked from the other methods.

Stephen

? ?
ID: 2008638 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2008643 - Posted: 21 Aug 2019, 16:47:32 UTC - in response to Message 2008638.  

To explain the distinction I was trying to make - here is the File menu structure for the Windows build:



'Close window', Ctrl-W, acts the same as the 'Red-X' button in the top-right corner of the screen: it removes the Manager from the visible screen (by minimising it to the notification area / system tray - depending on your preferred terminology era), but leaves both Manager and Client processes running.

'Exit BOINC' will - after dialogs, if enabled - shut down the Manager process completely. If the exit dialog has 'Stop running tasks ...' checked, that's implemented by shutting down the Client process too.

That's what I meant by saying that 'Close' and 'Exit' have distinct and different meanings in this context.
ID: 2008643 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2008649 - Posted: 21 Aug 2019, 17:41:26 UTC

Sorry about the misquote Richard, didn't mean to put words in your mouth. I vaguely remembered something about the Tray.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2008649 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2008650 - Posted: 21 Aug 2019, 17:47:14 UTC - in response to Message 2008604.  

Although I have a Linux machine now (Mint/Ubuntu), I can't test that function: I'm running BOINC v7.16.1 take 2, and as threatened the tray icon and the menu control options have been removed. I intend to make a fuss about that tomorrow evening.


Yes, I have been running the old 7.14.2 Manager with the 7.16.1 client. Perfectly happy with how that pair works together. I will not be using the 7.16.1 Manager at all since I use the Manager dialog all day long to exit the client many times a day. If I ran some unattended server remotely somewhere I guess the loss of functionality would not be missed. But I interact with my hosts directly every day many times a day. Not being able to stop the client immediately would drive me nuts.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2008650 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2008654 - Posted: 21 Aug 2019, 18:46:01 UTC - in response to Message 2008650.  

I find that the Terminal window in Mint helpfully remembers my command history, even across sessions. Terminal plus two up-arrows currently brings back

sudo systemctl stop boinc-client
and once I've entered the sudo password once, the session remembers that I'm SuperUser for long enough for me to do what I need to do - like

sudo systemctl start boinc-client
ID: 2008654 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2008662 - Posted: 21 Aug 2019, 19:22:31 UTC - in response to Message 2008654.  

I find that the Terminal window in Mint helpfully remembers my command history, even across sessions. Terminal plus two up-arrows currently brings back

sudo systemctl stop boinc-client
and once I've entered the sudo password once, the session remembers that I'm SuperUser for long enough for me to do what I need to do - like

sudo systemctl start boinc-client

Yes, that works except for the fact I don't run the daemon on any of my normal PC hosts. The only host running the daemon is the RPi3B+

I guess I would have to resort to the kill command in Terminal.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2008662 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2008680 - Posted: 21 Aug 2019, 21:18:36 UTC - in response to Message 2008662.  

boinccmd --quit?
ID: 2008680 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2008681 - Posted: 21 Aug 2019, 21:26:47 UTC - in response to Message 2008680.  

Tried that. Have to specify all the other junk like hostname and passphrase
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2008681 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 2008709 - Posted: 22 Aug 2019, 0:50:58 UTC - in response to Message 2008604.  
Last modified: 22 Aug 2019, 1:04:51 UTC

...and as threatened the tray icon and the menu control options have been removed. I intend to make a fuss about that tomorrow evening.
I thought it was a joke. Why would anyone remove the App's ON/OFF switch? I'm truly mystified.
Well, I can see where this is going... I see another run such as 7.2.33, where I ran it for three years because it didn't have the Jumping Tasks page Bug, then finally built my Own newer version without the jumping Tasks Page.
I'll place the Stray Comma bug-fix in the current 7.16.1 and call it good. It seems the 7.16.1 Manager is back to requiring Dependencies, so we'll just use the 7.14.2 Manager with 7.16.1 boinc. Then hunker down and wait for some sanity to return. Heck some people still use 6.10.something....
ID: 2008709 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2008746 - Posted: 22 Aug 2019, 7:43:43 UTC - in response to Message 2008709.  

...and as threatened the tray icon and the menu control options have been removed. I intend to make a fuss about that tomorrow evening.
I thought it was a joke. Why would anyone remove the App's ON/OFF switch? I'm truly mystified.
Well, my argument tonight will go along the lines of.....

The very best programmers design for every eventuality, error trap everything, write full and clear error messages, and provide full documentation. Their programs are self-evident and never fail, so the error messages and documentation are completely redundant. These programmers are usually unemployed, because they are too slow and too expensive.

Good programmers consider their users - all their users - and design accordingly.

Poor programmers can't see the wider picture, and only design for "people like us".

Unfortunately, the design of the Linux builds has been taken over by the third group - they volunteered, when the money to employ Type 1 or Type 2 programmers was withdrawn by the NSF. "People like us" means, in this context, institutional academic programmers in big research labs. Their computers are provided and controlled by their employers: corporate security policy controls what software can be installed, what updates can be installed (and when), and so on.

People like them don't have access to sudo, so they can't turn a service back on if it's been turned off. That's why they think it's a good idea to remove the 'off' switch.

I've written that I've recently started dabbling in Linux/Mint. People like me own the hardware outright, only allow a single person - themselves - to use it, and are thus automatically their own sysadmins: there is nowhere else to run to when something needs updating. We have access to sudo and root. We can turn things back on, so the argument for removing the off switch collapses.

Them and us. The story of power down the ages. I will be arguing - not for the first time - that the source code should cater for everyone: if there needs to be a 'lockdown' version and an 'open' version, that should be done by configuration settings in the installation scripts.
ID: 2008746 · Report as offensive     Reply Quote
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 2008747 - Posted: 22 Aug 2019, 7:56:29 UTC - in response to Message 2008746.  

… the source code should cater for everyone: if there needs to be a 'lockdown' version and an 'open' version, that should be done by configuration settings in the installation scripts.
Particularly so if the goal is for as many people as possible to make use of BOINC for running their projects, and for as many people as possible to process work for those projects.
Grant
Darwin NT
ID: 2008747 · Report as offensive     Reply Quote
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22160
Credit: 416,307,556
RAC: 380
United Kingdom
Message 2008749 - Posted: 22 Aug 2019, 8:41:08 UTC

Also, while groups one and two may use a substantial proportion of the computers (but I doubt as think that there are more than a few dozen of such users) they do not make up the majority of users. However they are often the most vocal in the shouting of "it doesn't work the way I want it to therefore it is wrong". As initiated BOINC was to be for the "majority of average users", not a clique group of people who ignore the masses and just want to please themselves.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 2008749 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2008754 - Posted: 22 Aug 2019, 9:16:41 UTC - in response to Message 2008749.  

And the difficulty is going to lie in convincing the current Type 3 programmers of their falsity because of their inbuilt bias that "Real scientific programmers use Linux, everybody else has Windows (except a few wierdos on both sides who use Macs)". I won that argument the first time round (#3105, #3106), but clearly I didn't succeed in laying it to rest completely.
ID: 2008754 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2008822 - Posted: 22 Aug 2019, 19:25:53 UTC

Ouch. That was hard work, and I've got a headache coming on. Quickly, while it's fresh in my mind, and before I head out to the pub to make the headache worse...

A thinly-attended holiday meeting. BOINC v7.16 is delayed anyway, because Berkeley need to renew their code-signing certificate, and that's a very bureaucratic security process. I think it was clear from the BOINC side of the table that BOINC understands that there are many, if not a majority, of Linux users in the "own machine, own sysadmin, full control" camp - for whom the removal of the menus is inappropriate and unnecessary. So the BOINC repository code has "shut down connected client" and "Exit BOINC Manager" in place and - I think - will continue to do so. The removal of the system tray icon was different, and I wrote a piece about that, but forgot to post it after preview. I'll try to recreate it tomorrow.

But we also had a Fedora repository maintainer at the table, and apparently it was a Fedora decision to remove the menu items from this distribution. We were given this url for a list of patches applied by Fedora to the raw BOINC code during the packaging process: https://src.fedoraproject.org/rpms/boinc-client/tree/master. I see

manager_exit_menu_entry_removal.patch
manager_shut_down_connected_client.patch
in that list, but I haven't dug any deeper, and won't tonight.

So we seem to be between a rock and a hard place. Linux (as represented by Fedora) is saying "we package for distribution to secure sites as a service: user mode distribution should be by download from the BOINC website". BOINC is saying "We don't have the resources to maintain user-mode download builds any more: that should be left to the Linux distribution community".

What I don't, at this stage, understand is the relationship between LocutusOfBorg and Fedora. It feels as if Gianfranco is packaging from the Fedora tree, rather than the BOINC tree. I think I've volunteered to reach out to Gianfranco and try to clarify that. At some point, I also want to get my Linux build system to make me a Manager from certified BOINC sources, so I can see what v7.16.1 really looks like. But that didn't work last time I tried it. Going to be a busy bank holiday weekend.

Starting now, with beer. Seeya.
ID: 2008822 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2008827 - Posted: 22 Aug 2019, 20:06:32 UTC

Thanks for the update Richard. I hope the decision finally arrives at the "the majority of Linux users are user-mode users/sysadmins" as the conclusion.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2008827 · Report as offensive     Reply Quote
elec999 Project Donor

Send message
Joined: 24 Nov 02
Posts: 375
Credit: 416,969,548
RAC: 141
Canada
Message 2008919 - Posted: 23 Aug 2019, 11:56:48 UTC

Just wondering why my 1070tis machine is doing so bad,
ID: 8780060
Details | Tasks
Cross-project stats:
BOINCstats.com Free-DC 1070ti2 home 78,115.50 2,472,618 7.14.2 GenuineIntel
Intel(R) Xeon(R) CPU E3-1230 v3 @ 3.30GHz [Family 6 Model 60 Stepping 3]
(8 processors) [2] NVIDIA GeForce GTX 1070 Ti (4095MB) driver: 418.56 OpenCL: 1.2 Linux Ubuntu
Ubuntu 19.04 [5.0.0-21-generic|libc 2.29 (Ubuntu GLIBC 2.29-0ubuntu2)] 23 Aug 2019, 11:33:17 UTC

Thank you
ID: 2008919 · Report as offensive     Reply Quote
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 2008920 - Posted: 23 Aug 2019, 12:13:35 UTC - in response to Message 2008919.  

You are having lots of errors on that system. Primarily cuda initialization errors. It looks like one card crashed and isn’t working.

Reboot the system, check your power cables. What PSU is running this system?
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 2008920 · Report as offensive     Reply Quote
Previous · 1 . . . 127 · 128 · 129 · 130 · 131 · 132 · 133 . . . 162 · Next

Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.