Linux CUDA 'Special' App finally available, featuring Low CPU use

Message boards : Number crunching : Linux CUDA 'Special' App finally available, featuring Low CPU use
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 5 · 6 · 7 · 8 · 9 · 10 · 11 . . . 83 · Next

AuthorMessage
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22158
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1840820 - Posted: 8 Jan 2017, 14:41:00 UTC

Arghhh - brain fade :-(
I just loaded the special app on a second PC - the one running a pair of GTX780s.... No prizes for guessing who forgot to make sure all the permission were set correctly before un-suspending SETI. Sorted with a quick attack of chmod and now it looks like things are behaving themselves (tasks aren't crashing out instantly...)
I've started with an un-roll of 8 (wild guess based on the 780 being a bit more pokey than a 750, but not as pokey as a 970). First glance suggest this is not "as good as it could be". I'll let things run for a few hours before I have a play.

My "big cruncher" has just lost a GTX970, but gained a GTX980 - part of a program of getting everything working again - it will be interesting to see what happens with the GTX980 card as it had a catastrophic fan failure (shredded bits of fan all over the place) when a foreign body was ingested (computer fans and M8 nuts don't mix very well). If all is well I'll put both GTX970 into my daily driver, then the big cruncher will have 2 off GTX980 and a GTX1080. Well that's the plan.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1840820 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1840833 - Posted: 8 Jan 2017, 16:26:22 UTC - in response to Message 1840820.  

Device 1: GeForce GTX 780, 3015 MiB, regsPerBlock 65536
computeCap 3.5, multiProcs 12


Try 24 for unroll, Peti's doubling the number seems to be working for me.

About half my tasks are down to 1:20 from ~3.5-4 min before tweaking.
ID: 1840833 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22158
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1840835 - Posted: 8 Jan 2017, 16:39:03 UTC

Thanks Brent - I'll let it run at 8 for a few hours just to get a feel for it, then ramp it up to 12 for a day or so, then lift it up to 16 (probably after Tuesday's outrage), and finally 24 later in the week.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1840835 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22158
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1840899 - Posted: 8 Jan 2017, 22:11:18 UTC

One interesting, un-quantified observation - the not only has the CPU usage dropped when compared to the Linux stock application, but the performance has improved, with run CPU run times dropping by something around a third to half.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1840899 · Report as offensive
Sidewinder Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 15 Nov 09
Posts: 100
Credit: 79,432,465
RAC: 0
United States
Message 1840908 - Posted: 8 Jan 2017, 23:10:12 UTC - in response to Message 1840786.  

Tbar, do you think setting it any higher will keep the utilization higher on the 1050 Ti's? They typically stay at low 90's to mid-80's.

You could try it and see. In my experience raising the unroll on the lower end cards will slow down the VLAR tasks. Of course we don't have any BLC tasks at present, but, the unroll is mainly for the BLC tasks and a setting for Arecibo tasks may not be best for VLARs.

It might be better to add the <no_priority_change>1</no_priority_change> line to your cc_config.xml file. That line will set All BOINC tasks to nice 0, but usually will increase GPU usage. I have it on my machines and the machine with three 750Ti is also running 2 CPU tasks on a old Quad core CPU yet the GPU usage is around the low to mid 90s. The pfb & pfp settings are the same from the Windows CUDA Apps and just as with the Windows CUDA Apps they produce little to No advantage. They Can cause increased Inconclusive results on some cards though. On my Mac it Slows down the tasks unless the settings are maxed out and then you get Many Inconclusive results. The last thing this App needs is More Inconclusive results, so, I don't recommend using those settings.


It looked like increasing the unroll to 8 caused more frequent dips in to the low 70's and even the 60's on the 1050 Ti's. Setting the unroll back to 6 and putting the following into the cc_config seems to have the best results so far:

<process_priority>2</process_priority>
<process_priority_special>3</process_priority_special>

ID: 1840908 · Report as offensive
qbit
Volunteer tester
Avatar

Send message
Joined: 19 Sep 04
Posts: 630
Credit: 6,868,528
RAC: 0
Austria
Message 1841835 - Posted: 13 Jan 2017, 8:58:52 UTC

My RAC is 12.6k currently. That's more then twice (!) what I had on Windows with OpenCL. Ofc it has to do with the current lack of Guppies but still that's pretty crazy!
ID: 1841835 · Report as offensive
Rockhount
Avatar

Send message
Joined: 29 May 00
Posts: 34
Credit: 31,935,954
RAC: 29
Germany
Message 1841851 - Posted: 13 Jan 2017, 11:59:21 UTC

The RAC from my machine increased nearly four times from 3500 up to 13k.
This is only a slow i3-2100T and a 750Ti with cuda 6 app.

https://setiathome.berkeley.edu/show_host_detail.php?hostid=1931980
Regards from nothern Germany
Roman

SETI@home classic workunits 207,059
SETI@home classic CPU time 1,251,095 hours

ID: 1841851 · Report as offensive
qbit
Volunteer tester
Avatar

Send message
Joined: 19 Sep 04
Posts: 630
Credit: 6,868,528
RAC: 0
Austria
Message 1841879 - Posted: 13 Jan 2017, 14:17:30 UTC

Wow, that's damn nice, Rockhount!

I'm just a bit confused because I saw ppl here talking about a speedup of about 30% or so. Looks like either the OpenCL apps run a lot better on newer&faster cards like the 1080/70 or this linux app here runs especially well on older/slower cards like the 750(ti).
ID: 1841879 · Report as offensive
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 1841996 - Posted: 13 Jan 2017, 21:53:36 UTC - in response to Message 1841851.  

The RAC from my machine increased nearly four times from 3500 up to 13k.
This is only a slow i3-2100T and a 750Ti with cuda 6 app.

https://setiathome.berkeley.edu/show_host_detail.php?hostid=1931980


. . Hi Rockhount

. . Looking very good. Especially the 11 min runtimes for the almost VLAR arecibo tasks with 7 mins for normal AR tasks.

Stephen

.
ID: 1841996 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1842002 - Posted: 13 Jan 2017, 22:19:15 UTC

Hi All,

I'm glad to see some serious positive feedback. Thanks. THAT is what keeps me going further on -- to the final front-ear!


I know there are still some unexpected and not yet to found error(s) causing cases that may have an impact to your credit/RAC/or whatsoever measure you are looking at (e.g. continuous valid ones). The one in the parenthesis is my next goal, I guess... Unless something that I have had in my mind for about a fotnight goes ahead of it.

I think my mind is bubbling. I feel kind of light... (And I'm not in under of an influence of any drug.)

I'll keeeeep on optimizinng, but my next goal is on organizing the workload even more to filling the SM/SMX and at the same time keeping a watch over the traffic from and to the main (GPU) memory and utilizing as many as possible kernel blocks a) at run b) at waiting for to be run . And at the same time to improve the impossible, and to serve All Wonderful people who have spent time testing my (sometimes invalid) versions.


Should any of you running the m version feel like running faster or more precise please comment.

...etc..

Petri
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1842002 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 1842015 - Posted: 13 Jan 2017, 22:55:10 UTC - in response to Message 1841879.  

Wow, that's damn nice, Rockhount!

I'm just a bit confused because I saw ppl here talking about a speedup of about 30% or so. Looks like either the OpenCL apps run a lot better on newer&faster cards like the 1080/70 or this linux app here runs especially well on older/slower cards like the 750(ti).

Also the lack of Guppie work over the last week would have resulted in a huge boost to throughput. Over the next few days now that Guppies have been restored we will see things drop down to more normal levels.
Grant
Darwin NT
ID: 1842015 · Report as offensive
qbit
Volunteer tester
Avatar

Send message
Joined: 19 Sep 04
Posts: 630
Credit: 6,868,528
RAC: 0
Austria
Message 1842036 - Posted: 14 Jan 2017, 0:20:20 UTC
Last modified: 14 Jan 2017, 0:23:44 UTC

I agree, I mentioned that in a post before. And it will have extra impact on my RAC since I can just use unroll 2 because of low VRAM.
But based on what I saw before Centurion went down I suppose my RAC will settle at around 10k and that's still an impressive boost compared to OpenCL on Windows.
ID: 1842036 · Report as offensive
Profile Michel Makhlouta
Volunteer tester
Avatar

Send message
Joined: 21 Dec 03
Posts: 169
Credit: 41,799,743
RAC: 0
Lebanon
Message 1843475 - Posted: 21 Jan 2017, 9:49:29 UTC

it's getting hard to ignore this thread and avoid linux... can someone list the requirements and steps on a high level once and for all?

i know i am supposed to search and read the posts, but this will make it easier on everyone to switch to linux. i will write a detailed tutorial for people who aren't familiar with linux, i just need the requirements (kernel and drivers version, boinc version and etc...). Then i will test it on my 2x1070 host and write the tutorial in parallel.
ID: 1843475 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1843484 - Posted: 21 Jan 2017, 10:49:49 UTC - in response to Message 1843475.  
Last modified: 21 Jan 2017, 11:01:04 UTC

Hi,

Hardware for crunching only could be something like a 4-12 core processor, 8-16 Gb RAM, 128Gb SSD and a modern GPU. A gtx780, 750i, 980 or a 10x0.

I used an USB drive to install Linux. https://help.ubuntu.com/community/Installation/FromUSBStick

0) There is an administrative user called root and you choose the password during the install. Keep it safe. You create a username for yourself and a password at the install time too.
1) I have run many flavours of Linux. Currently I'm running Ubuntu. https://www.ubuntu.com/
2) My desktop manager is called GNOME. Other variants exist and any of them will do the job. Google: how to install gnome ubuntu
3) I use drivers from NVIDIA. My version is 375.10. A newer version is available. http://www.nvidia.com/object/unix.html. Easiest is to use system settings and select a suitable NVIDIA driver there. The default public domain driver can not do Seti. You must change to Nvidia driver way or another.
4) Boinc can be installed a) from command line (terminal window): sudo apt-get install boinc-client b) somehow from the desktop. The sudo command will ask the administrative password.
5) You have to set up the app_init.xml and maybe an app_config.xml to run Anonymous platform apps. Test with any available cuda app. When you get all previous stuff working I'll supply you with the latest special app.

It is a long time since I did the installation so I do not remember all the quirks and hacks needed.

Petri
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1843484 · Report as offensive
Sidewinder Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 15 Nov 09
Posts: 100
Credit: 79,432,465
RAC: 0
United States
Message 1843505 - Posted: 21 Jan 2017, 13:28:09 UTC - in response to Message 1843475.  
Last modified: 21 Jan 2017, 13:29:31 UTC

If you want to use Arch or Antergos (the distro I use) here are the steps that I used more or less:

### OS ###
sudo pacman -S nvidia cuda nvidia-libgl
# if prompted, choose "xf86-input-evdev"

sudo pacman -S gtk3 xorg-server gtk2 xorg-xauth xorg-xhost
sudo pacman -S p7zip screenfetch nvidia-settings htop boinc
sudo pacman -S cantarell-fonts ttf-google-fonts

# edit sshd_config for boincmgr
Edit /etc/ssh/sshd_config:
	Port 22
	ListenAddress 0.0.0.0
	AllowTcpForwarding yes
	X11Forwarding yes
	X11UseLocalhost yes

# reboot machine

### BOINC ###
sudo ln -s /var/lib/boinc/gui_rpc_auth.cfg /home/<user>/gui_rpc_auth.cfg
sudo chmod 644 /var/lib/boinc/gui_rpc_auth.cfg

sudo nano /var/lib/boinc/cc_config.xml:

	<cc_config>
		<options>
			<allow_remote_gui_rpc>1</allow_remote_gui_rpc>
			<process_priority>2</process_priority>
			<process_priority_special>3</process_priority_special>
		</options>
	</cc_config>


sudo chown boinc:boinc /var/lib/boinc/cc_config.xml

boinccmd --project_attach http://setiathome.berkeley.edu <account_key>

# after boinc is running with seti@home, quit (boinccmd --quit) to setup custom app(s)

# You can use a custom app_info or the one included with the archive
nano /var/lib/boinc/projects/setiathome.berkeley.edu/app_info.xml
chown boinc:boinc /var/lib/boinc/projects/setiathome.berkeley.edu/app_info.xml

# might be unnecessary, but backup just in case
sudo mv /var/lib/boinc/projects/setiathome.berkeley.edu/libcudart.so.6.0 /var/lib/boinc/projects/setiathome.berkeley.edu/libcudart.so.6.0.old
sudo mv /var/lib/boinc/projects/setiathome.berkeley.edu/libcufft.so.6.0 /var/lib/boinc/projects/setiathome.berkeley.edu/libcufft.so.6.0.old

7z x Linux_CUDA6_Special-App.7z

sudo wget http://boinc2.ssl.berkeley.edu/beta/download/libcudart.so.6.0 -P /var/lib/boinc/projects/setiathome.berkeley.edu/
sudo wget http://boinc2.ssl.berkeley.edu/beta/download/libcufft.so.6.0 -P /var/lib/boinc/projects/setiathome.berkeley.edu/

sudo cp ./Linux_CUDA6_Special-App/setiathome_x41p_zi+_x86_64-pc-linux-gnu_cuda60 /var/lib/boinc/projects/setiathome.berkeley.edu/
# you can also copy MBv8_8.0r3305_ssse3_x86_64-pc-linux-gnu if you want to use that app

sudo chown boinc:boinc /var/lib/boinc/projects/setiathome.berkeley.edu/*

sudo systemctl enable boinc
sudo systemctl restart boinc

### Useful commands/utilities ###
$ htop
or
$ top -sc
$ watch -n1 nvidia-smi
$ boincmgr

ID: 1843505 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1843562 - Posted: 21 Jan 2017, 18:42:00 UTC
Last modified: 21 Jan 2017, 18:45:24 UTC

Mine was fairly simple,
- Burn "ubuntu-14.04.5-desktop-amd64.iso" (Google will find it)
- Boot from DVD and install
- Get rid of default NVIDIA drivers (Terminal Window).
sudo apt-get purge 'nvidia*'
sudo apt-get autoremove
Reboot
Selected most recent available NVIDIA drivers in Ubuntu Software Center, under additional drivers
Reboot

- Install BOINC from Ubuntu Software Center
- Will be installed in '/var/lib/boinc-client' directory
Manually Install 'Special App' as described in the notes contained with it (and here).

BOINC likes to take ownership of files and won't let you edit, So ... (terminal Window)
sudo chmod -R 0777 /var/lib/boinc-client

NOTE: If a person is fairly familiar with how to get around using DOS on a computer they shouldn't have too much problems getting Ubuntu "Special App" up and running. A plain stock install is really, really easy!
ID: 1843562 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22158
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1843587 - Posted: 21 Jan 2017, 19:32:18 UTC

In the last couple of years the installation and configuration of Linux has leapt forwards a long way. I've been using "Mint" for my most recent builds, so bear that in mind while reading the following.
All Brent's steps, above, with the exception of the one command line starting "sudo" can be done using the GUI.
It is worth suspending SETI before installing the special app, and then re-running the command line before resuming SETI.
Also, you can do it from a USB stick if you don't have a bootable DVD to hand.
And it is a very rapid installation - from "cold hardware" to a system running SETI is about 30 minutes, and another ten minutes to get the special apps installed.


I found Mint has a very good guide to making the bootable USB stick, and that the hardest thing can be getting the target PC to default to "boot from USB", and every BIOS is different. It is worth re-setting to "boot from hard disk" once you've done the installation, because some BIOS appear to behave better set like this.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1843587 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1843821 - Posted: 22 Jan 2017, 19:19:27 UTC - in response to Message 1843562.  

I should add to my list ...

When adding executables into the project directory you must set the executable flag ...
sudo chmod -R 0777 /var/lib/boinc-client

Will do that.
ID: 1843821 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22158
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1843823 - Posted: 22 Jan 2017, 19:29:00 UTC

...not doing that the first time I did an installation of the "special app" it took me about half an hour to work out why I was getting so many errors.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1843823 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1845266 - Posted: 29 Jan 2017, 22:07:04 UTC

Question,
Where did you get BOINC v7.6.xx for Linux?
I see several people are running it, but I really don't see where you got it from or installed anything different than I did.
My searches end with nothing different.
I would really like to use the priority flags so that I can run more CPU cores, but v7.2.42 doesn't allow it.

Also, is there another (better or needs more testing) 'special' version that is 'public'?
I have: setiathome v8 enhanced x41p_zi+, Cuda 6.00 special

Thanks.
ID: 1845266 · Report as offensive
Previous · 1 . . . 5 · 6 · 7 · 8 · 9 · 10 · 11 . . . 83 · Next

Message boards : Number crunching : Linux CUDA 'Special' App finally available, featuring Low CPU use


 
©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.