Optimized apps & updates

Message boards : Number crunching : Optimized apps & updates
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile Jonathan Jeckell

Send message
Joined: 25 Jun 99
Posts: 30
Credit: 16,549,509
RAC: 62
United States
Message 1755454 - Posted: 10 Jan 2016, 21:46:32 UTC

Sorry if this is answered elsewhere, but I didn't see it.

I installed the Lunaticks optimized v8 applications, but 8.01 on up to 8.03 have been released since then. Are the optimized apps still in play, or do I need to find out if Lunaticks has had commensurate updates as well?

Also, while I understand the immense workload the SETI team is under right now, I wish official optimized versions were available through the actual SETI application channel, like they had SSE4.1, SSSE3, and AVX version available for Mac OSX automatically. No disrespect to the Lunaticks team, but I would feel better from an information assurance standpoint, as well as convenience during updates.
ID: 1755454 · 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 1755456 - Posted: 10 Jan 2016, 21:54:06 UTC

The optimizers like Lunatics are waiting until the V8 applications are stable before starting to wave their magic wands.

(That said, I dare say they are doing some preparative work on things that they are fairly certain will not change as a result of using the stock v8 apps on main)
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1755456 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755461 - Posted: 10 Jan 2016, 22:27:57 UTC - in response to Message 1755454.  

There are already Apps available for your Intel Macs. All you have to do is download them, place them in the setiathome.berkeley.edu folder, reinstall BOINC, and they will be working on v8 tasks in a just a few minutes. The easiest method would be to download the package named "SETIv8 nVidia&CPU.zip". That one will give you the SSE41 CPU App and the OpenCL nVidia App. To use different Apps you would need to join app_info sections, however, the Intel_gpu and AVX Apps are also available.
http://www.arkayn.us/forum/index.php?topic=191.msg4368
ID: 1755461 · Report as offensive
Profile Jonathan Jeckell

Send message
Joined: 25 Jun 99
Posts: 30
Credit: 16,549,509
RAC: 62
United States
Message 1755590 - Posted: 11 Jan 2016, 16:18:16 UTC - in response to Message 1755461.  

There is a sticky thread in this forum that shows how to download and use v8.00 Lunaticks optimized applications. However, the official SETI app has moved on to 8.03 in some cases.

I just wanted to know if this requires commensurate updates to the Lunaticks version, which I have not used before this v8 change.

So the question wasn't how to get the Lunatick apps, but whether they need to be updated to keep up with the official platform changes.

Secondly, yes, while there are already apps for intel macs, the official SETI platform used to have three optimized flavors (SSE4.1, SSSE3, and AVX) that were in the official cache that automatically deployed if your computer had the right capabilities. I liked that from a management and information assurance standpoint. Those optimized apps seemed to perform substantially better than the plain vanilla v7 client for 64 bit Intel Macs, but I wonder if the v8 client already has one of those optimizations built in this time. I haven't captured enough performance data so far to tell for sure how these clients are performing in comparison to the v7.
ID: 1755590 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755592 - Posted: 11 Jan 2016, 16:45:26 UTC - in response to Message 1755590.  

Those Apps you keep speaking of were created by Me. They are Still listed here, http://www.arkayn.us/forum/index.php?topic=130.msg4193#msg4193
To get those accepted by SETI it took me around 2 months to refine them, two months to get them accepted into beta, and around two months hanging around beta. That means IF I started trying to get them onto beta Today, You Might see them on main in about 4 months. Considering what I went through Last time, I don't think they will Ever see Beta this time, so, the chances of you seeing My apps on main again are extremely slim. But they are available Now at the link I posted.

BTW Lunatics has Never posted an App for OSX or Linux.
ID: 1755592 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1755593 - Posted: 11 Jan 2016, 16:51:17 UTC

a) the sticky thread shows how to run the stock application (that's the one you get anyway when you attach to the project) under anonymous platform, e.g. with a app_info.xml.

b) Lunatics [please note spelling] are working hard on the next installer. It will have the GPU apps and optimised CPU apps (the latter could be better, but we are fully stretched, what with Josef Segur MIA) We will hand out 'something that works better than stock' for all the impatient users out there and then concentrate at a more sedate pace on further optimisations. The installer is Windows only, but Lunatics provide applications for Mac and Linux too. (manual install)

c)
I just wanted to know if this requires commensurate updates to the Lunaticks version, which I have not used before this v8 change.


I don't quite understand the question. The app_info.xml is for those who were previously running Lunatics (or had their own app_info.xml on other platforms) and want to process v8 tasks without going stock first. In this case they need to manually update their app_info.xml and download the respective apps. If you never used Lunatics, you don;t need to do anything - you will get the correct applications anyway.

d)
So the question wasn't how to get the Lunatick apps, but whether they need to be updated to keep up with the official platform changes.


If you go anonymous you need to keep up manually with the updates, yes. for the majority of the windows users this means waiting for the Lunatics team to update the installer.

e)
Secondly, yes, while there are already apps for intel macs, the official SETI platform used to have three optimized flavors (SSE4.1, SSSE3, and AVX) that were in the official cache that automatically deployed if your computer had the right capabilities. I liked that from a management and information assurance standpoint. Those optimized apps seemed to perform substantially better than the plain vanilla v7 client for 64 bit Intel Macs, but I wonder if the v8 client already has one of those optimizations built in this time. I haven't captured enough performance data so far to tell for sure how these clients are performing in comparison to the v7.


Mac applications are a bit tricky at the moment. Eric doesn't build them himself, he has to ask Charlie Fenton, whose current status is volunteer (previously employed at boinc)
Therefore Mac is running as a bare minimum. More applications may or may not become available, depending on how much time the different persons have to produce further builds.


So to summarise, there are no updated Lunatics CPU apps out yet (if anybody else has built something they are not working under the Lunatics logo) so I'm not quite sure what you installed.

Only your android host is running anon.

so, given that you are running stock on all hosts, I don't understand you question.

If, however you were running app_info.xml on a Mac host, then yes, you would have had to download the latest application and change app_info.xml accrdingly.
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1755593 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1755596 - Posted: 11 Jan 2016, 17:01:30 UTC - in response to Message 1755592.  

Those Apps you keep speaking of were created by Me. They are Still listed here, http://www.arkayn.us/forum/index.php?topic=130.msg4193#msg4193
To get those accepted by SETI it took me around 2 months to refine them, two months to get them accepted into beta, and around two months hanging around beta. That means IF I started trying to get them onto beta Today, You Might see them on main in about 4 months. Considering what I went through Last time, I don't think they will Ever see Beta this time, so, the chances of you seeing My apps on main again are extremely slim. But they are available Now at the link I posted.

BTW Lunatics has Never posted an App for OSX or Linux.


Wrong.

http://lunatics.kwsn.info/index.php?action=downloads;cat=1 - Linux v7 apps
http://lunatics.kwsn.info/index.php?action=downloads;cat=30 osX AP

we (or rather Urs) has always worked on the Linux and Mac flavours and is responsible for all the GPU linux and mac apps you see in the v8 section of https://setiweb.ssl.berkeley.edu/beta/apps.php (the openCl ones)

We haven't posted any v8 apps yet to main, because we are still working on making them 'good' as opposed to 'working'
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1755596 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755599 - Posted: 11 Jan 2016, 17:09:00 UTC - in response to Message 1755596.  
Last modified: 11 Jan 2016, 17:10:18 UTC

BTW Lunatics has Never posted an App Installer for OSX or Linux.

There, Fixed it for ya. Happy?
ID: 1755599 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1755601 - Posted: 11 Jan 2016, 17:10:35 UTC - in response to Message 1755599.  

BTW Lunatics has Never posted an App Installer for OSX or Linux.

There, Fixed it for ya. Happy?

Neither have you.
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1755601 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755603 - Posted: 11 Jan 2016, 17:15:05 UTC - in response to Message 1755601.  

BTW Lunatics has Never posted an App Installer for OSX or Linux.

There, Fixed it for ya. Happy?

Neither have you.

I don't think I ever said I had. In fact, I'm sure of it. But I do have some nice Mac Apps.
;-)
ID: 1755603 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1755605 - Posted: 11 Jan 2016, 17:21:11 UTC - in response to Message 1755603.  

BTW Lunatics has Never posted an App Installer for OSX or Linux.

There, Fixed it for ya. Happy?

Neither have you.

I don't think I ever said I had. In fact, I'm sure of it. But I do have some nice Mac Apps.
;-)

Do they have a < 5% inconclusive rate?
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1755605 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755609 - Posted: 11 Jan 2016, 17:33:18 UTC - in response to Message 1755605.  
Last modified: 11 Jan 2016, 17:43:41 UTC

They have about the same as any other Mac App I've run. Check them out yourself.
Beware, if you see something similar to;
setiathome v8 enhanced x41zc, Cuda 6.50 special
Compiled with NVCC 7.5, using 6.5 libraries. Modifications done by petri33.

That is an Alpha CUDA App being tested in hopes of speeding up CUDAs for Everyone. As with other Alpha code, things are bumpy at first. But, do checkout the others.
You can find most of them over here, http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=63959
ID: 1755609 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1755612 - Posted: 11 Jan 2016, 17:48:35 UTC - in response to Message 1755609.  
Last modified: 11 Jan 2016, 17:50:29 UTC

They have about the same as any other Mac App I've run. Check them out yourself.
Beware, if you see something similar to;
setiathome v8 enhanced x41zc, Cuda 6.50 special
Compiled with NVCC 7.5, using 6.5 libraries. Modifications done by petri33.

That is an Alpha CUDA App being tested in hopes of speeding up CUDAs for Everyone. As with other Alpha code, things are bumpy at first. But, do checkout the others.
You can find most of them over here, http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=63959

Don't have to tell me that, I've been codewalking part of petri's work and he himself admitted it wasn't quite up to scratch yet. He's working on it though.

You can't tell me how one of the basic measurements of quality control checks out?

Edit: and apologies to Jonathan for going off topic in his thread.
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1755612 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755615 - Posted: 11 Jan 2016, 18:04:44 UTC - in response to Message 1755612.  

They have about the same as any other Mac App I've run. Check them out yourself.
Beware, if you see something similar to;
setiathome v8 enhanced x41zc, Cuda 6.50 special
Compiled with NVCC 7.5, using 6.5 libraries. Modifications done by petri33.

That is an Alpha CUDA App being tested in hopes of speeding up CUDAs for Everyone. As with other Alpha code, things are bumpy at first. But, do checkout the others.
You can find most of them over here, http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=63959

Don't have to tell me that, I've been codewalking part of petri's work and he himself admitted it wasn't quite up to scratch yet. He's working on it though.

You can't tell me how one of the basic measurements of quality control checks out?

Edit: and apologies to Jonathan for going off topic in his thread.

You're the one worried about percentages. I have other things to do other than work up numbers for a dozen different Apps. When I look at the results and see Very few inconclusives, that's good enough for me. If you want percentages, work them up yourself. You do realize I've compiled about a couple dozen Apps in the last week or so? You want percentages? If it validates without giving a suspicious inconclusive that's good enough for this Volunteer.
ID: 1755615 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1755624 - Posted: 11 Jan 2016, 18:34:04 UTC - in response to Message 1755615.  

Those Apps you keep speaking of were created by Me.

To get those accepted by SETI it took me around 2 months to refine them, two months to get them accepted into beta, and around two months hanging around beta. That means IF I started trying to get them onto beta Today, You Might see them on main in about 4 months. Considering what I went through Last time, I don't think they will Ever see Beta this time, so, the chances of you seeing My apps on main again are extremely slim.

You're the one worried about percentages. I have other things to do other than work up numbers for a dozen different Apps. When I look at the results and see Very few inconclusives, that's good enough for me. If you want percentages, work them up yourself. You do realize I've compiled about a couple dozen Apps in the last week or so? You want percentages? If it validates without giving a suspicious inconclusive that's good enough for this Volunteer.

Ever thought those two statements might be related?
ID: 1755624 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755626 - Posted: 11 Jan 2016, 18:40:05 UTC - in response to Message 1755624.  

Those Apps you keep speaking of were created by Me.

To get those accepted by SETI it took me around 2 months to refine them, two months to get them accepted into beta, and around two months hanging around beta. That means IF I started trying to get them onto beta Today, You Might see them on main in about 4 months. Considering what I went through Last time, I don't think they will Ever see Beta this time, so, the chances of you seeing My apps on main again are extremely slim.

You're the one worried about percentages. I have other things to do other than work up numbers for a dozen different Apps. When I look at the results and see Very few inconclusives, that's good enough for me. If you want percentages, work them up yourself. You do realize I've compiled about a couple dozen Apps in the last week or so? You want percentages? If it validates without giving a suspicious inconclusive that's good enough for this Volunteer.

Ever thought those two statements might be related?

If you want percentages, place them on Beta, wait a day, then hit the key for the readout. I Do hope you go through the results and weed out all the obvious Bad wingpeople before tallying the final results, That is the time consuming part.
ID: 1755626 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1755631 - Posted: 11 Jan 2016, 18:56:01 UTC - in response to Message 1755626.  
Last modified: 11 Jan 2016, 19:37:49 UTC

If you want percentages, place them on Beta, wait a day, then hit the key for the readout. I Do hope you go through the results and weed out all the obvious Bad wingpeople before tallying the final results, That is the time consuming part.

Yes. And I found one today where the problem wasn't obvious, so I'm re-running that one offline as we speak.

Edit - WU 2029956471. The offline test confirmed that the stock CPU app had found a triplet that the cuda app missed:

Triplet: peak=8.708315, time=91.87, period=1.166540, d_freq=1420162901.1828, chirp=-65.449670, fft_len=256

That would seem to be far enough above <triplet_thresh>8.12590027</triplet_thresh> as to indicate a problem (rather than simply one of the expected 5%), so it's going back to developer - and may even delay the submission of the application for stock use. That's what we do.
ID: 1755631 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755652 - Posted: 11 Jan 2016, 22:12:10 UTC - in response to Message 1755631.  

If you want percentages, place them on Beta, wait a day, then hit the key for the readout. I Do hope you go through the results and weed out all the obvious Bad wingpeople before tallying the final results, That is the time consuming part.

Yes. And I found one today where the problem wasn't obvious, so I'm re-running that one offline as we speak.

Edit - WU 2029956471. The offline test confirmed that the stock CPU app had found a triplet that the cuda app missed:

Triplet: peak=8.708315, time=91.87, period=1.166540, d_freq=1420162901.1828, chirp=-65.449670, fft_len=256

That would seem to be far enough above <triplet_thresh>8.12590027</triplet_thresh> as to indicate a problem (rather than simply one of the expected 5%), so it's going back to developer - and may even delay the submission of the application for stock use. That's what we do.

Yes, and I just went through my email and found new CUDA code from Petri dated Jan 8th. So, if I can Volunteer some time a little later I'll once again Pull the two NV cards from my machine and Compile a new Mac Alpha CUDA App. It's what I do in my spare time.
I noticed that was a Windows CUDA App. I'm curious how you are judging this mythical 5%. How are you verifying the machines you are running the percentages against? I'm seeing a good number of inconclusives from a few obviously misbehaving CPU machines. I've noticed some Macs running 10.7.4 that are having the same trouble with the current stock App they did with the v7 SSE41 & AVX Apps. For some reason, there seems to be a problem with OSX 10.7.4 and some SETI CPU Apps. The same machines that failed with the old v7 SSE41 & AVX Apps ran the SSSE3 version just fine.
ID: 1755652 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1755655 - Posted: 11 Jan 2016, 22:20:24 UTC - in response to Message 1755652.  

I'd delay on that for a little time, if I were you. Jason and I have tried two trial apps, one incorporating

memset(inv, -1, swi.analysis_cfg.triplet_pot_length*sizeof(int)); 

from https://setisvn.ssl.berkeley.edu/trac/changeset/3270, which both he and (he thinks) Petri missed.

But we still haven't found the missing triplet, so there's something else lurking too.
ID: 1755655 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1755665 - Posted: 11 Jan 2016, 23:10:28 UTC - in response to Message 1755655.  
Last modified: 11 Jan 2016, 23:11:05 UTC

Well, I searched the Jan 8th code and did find;
memset(inv, -1, sizeof(inv));
instead of
memset(inv, -1, swi.analysis_cfg.triplet_pot_length*sizeof(int));
I corrected the analyzeReport.cpp file.
However, searching for;
#elif (SIZEOF_LONG == 4);
#elif (SIZEOF_LONG == 4)
Finds neither one.
ID: 1755665 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : Optimized apps & updates


 
©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.