A SETI@Home Astropulse optimized application UPDATE has been released. New application version is 5.03, r112 for Intel & AMD SSE & SSE3 compatible CPUs.

NOTE: This is a REQUIRED UPGRADE if you are currently running optimized application 5.00, r69 or r103

The packages are "transitional" meaning they contain optimized apps & app_info.xml file for w/ both version v5.00 & the new v5.03. The deprecated application is included since there are still v5.00 WU re-issues being sent or in current host queues.

The release post is at the stickied New Optimized Apps Links -- READ ONLY thread Here

Alternately, you can also download @ Lunatics site as well.

Several people have helped make this build possible. First and foremost, Raistmer & Jason G for leading this effort!!!
Special thanks go to Josef W. Segur & Richard Haselgrove for their expertise, input & meticulous testing of the countless development builds. Lastly, since appreciation to all the pre-release testers that clawed and chased after those elusive new AP 5.03 WUs & validations so this could get into your hands as quickly as possible : > )

To all of you we value your input and would like to hear from you at Lunatics'

Regards,
Lunatics Team

Thanks much for all your hard work on this app, it's greatly appreciated. I have a quick question, I would like to run it on my computers, currently getting ready to set it up on this machine, the startup messages list:

2/19/2009 8:43:16 PM Starting BOINC client version 6.6.9 for windows_intelx86
2/19/2009 8:43:16 PM log flags: task, file_xfer, sched_ops
2/19/2009 8:43:16 PM Libraries: libcurl/7.19.2 OpenSSL/0.9.8i zlib/1.2.3
2/19/2009 8:43:16 PM Data directory: C:\Documents and Settings\All Users\Application Data\BOINC
2/19/2009 8:43:16 PM Running under account Shannon
2/19/2009 8:43:16 PM Processor: 2 GenuineIntel Intel(R) Core(TM)2 Duo CPU T9300 @ 2.50GHz [x86 Family 6 Model 23 Stepping 6]
2/19/2009 8:43:16 PM Processor features: fpu tsc pae nx sse sse2 mmx
2/19/2009 8:43:16 PM OS: Microsoft Windows XP: Professional x86 Editon, Service Pack 2, (05.01.2600.00)
2/19/2009 8:43:16 PM Memory: 2.00 GB physical, 3.85 GB virtual
2/19/2009 8:43:16 PM Disk: 74.44 GB total, 59.74 GB free
2/19/2009 8:43:16 PM Local time is UTC -6 hours
2/19/2009 8:43:16 PM CUDA device: Quadro NVS 135M (driver version 17924, CUDA version 1.1, 128MB, est. 2GFLOPS)
2/19/2009 8:43:16 PM Not using a proxy
2/19/2009 8:43:16 PM Version change (6.6.4 -> 6.6.9)
2/19/2009 8:43:16 PM SETI@home URL: http://setiathome.berkeley.edu/; Computer ID: 4696039; location: home; project prefs: default
2/19/2009 8:43:16 PM General prefs: from SETI@home (last modified 26-May-2007 00:12:51)
2/19/2009 8:43:16 PM Computer location: home
2/19/2009 8:43:16 PM General prefs: no separate prefs for home; using your defaults
2/19/2009 8:43:16 PM Preferences limit memory usage when active to 1022.95MB
2/19/2009 8:43:16 PM Preferences limit memory usage when idle to 1943.60MB
2/19/2009 8:43:16 PM Preferences limit disk usage to 37.22GB
2/19/2009 8:43:16 PM Running CPU benchmarks
2/19/2009 8:43:16 PM Suspending computation - running CPU benchmarks
2/19/2009 8:43:26 PM SETI@home Sending scheduler request: Requested by user.
2/19/2009 8:43:26 PM SETI@home Reporting 1 completed tasks, requesting new tasks
2/19/2009 8:43:32 PM SETI@home Scheduler request completed: got 0 new tasks
2/19/2009 8:43:32 PM SETI@home Message from server: No work sent
2/19/2009 8:43:32 PM SETI@home Message from server: Can't use CUDA app for SETI@home Enhanced: Your CUDA device has insufficient memory
2/19/2009 8:43:32 PM SETI@home Message from server: CPU app exists for SETI@home Enhanced but no CPU work requested
2/19/2009 8:43:48 PM Benchmark results:
2/19/2009 8:43:48 PM Number of CPUs: 2
2/19/2009 8:43:48 PM 2479 floating point MIPS (Whetstone) per CPU
2/19/2009 8:43:48 PM 5519 integer MIPS (Dhrystone) per CPU
2/19/2009 8:43:49 PM SETI@home Restarting task 09ja09aa.4067.18477.14.8.184_0 using setiathome_enhanced version 603
2/19/2009 8:43:49 PM SETI@home Restarting task 09ja09aa.4067.18477.14.8.218_0 using setiathome_enhanced version 603


it looks like I did put in some enhanced app, would this one run in addition to the one that currently appears to be running, would it replace it, or am I just mistaken that I even have one? In the C:\program files\boinc dir, it lists these files, if it helps.

02/19/2009 08:43 PM <DIR> .
02/19/2009 08:43 PM <DIR> ..
04/28/2008 07:56 PM 1,077,248 AK_v8_win_SSE41.exe
04/28/2008 07:56 PM 146 AK_v8_win_SSE41.manifest
04/28/2008 07:56 PM 4,140,032 AK_v8_win_SSE41.pdb
04/30/2008 07:54 PM 951 app_info.xml
02/18/2009 12:07 PM 13,312 boinc.dll
02/18/2009 12:08 PM 745,472 boinc.exe
02/18/2009 12:11 PM 413,696 boinccmd.exe
01/16/2009 05:07 PM 284,646 BOINCGUIApp.ico
02/18/2009 12:10 PM 4,177,920 boincmgr.exe
02/18/2009 12:11 PM 8,192 boincsvcctrl.exe
02/18/2009 12:09 PM 53,248 boinctray.exe
01/16/2009 05:09 PM 242,471 ca-bundle.crt
01/16/2009 05:09 PM 35,821 COPYING
01/16/2009 05:09 PM 1,653 COPYRIGHT
01/16/2009 05:09 PM 192,512 cudart.dll
01/16/2009 05:08 PM 1,045,128 dbghelp.dll
01/16/2009 05:08 PM 163,088 dbghelp95.dll
01/16/2009 05:09 PM 212,992 libcurl.dll
01/16/2009 05:08 PM 1,019,904 libeay32.dll
02/19/2009 08:43 PM <DIR> locale
12/01/2006 09:03 PM 1,869 Microsoft.VC80.CRT.manifest
12/02/2006 05:22 AM 479,232 msvcm80.dll
12/01/2006 09:03 PM 548,864 msvcp80.dll
12/01/2006 09:03 PM 626,688 msvcr80.dll
01/16/2009 05:08 PM 94,944 srcsrv.dll
01/16/2009 05:08 PM 200,704 ssleay32.dll
01/16/2009 05:08 PM 133,048 symsrv.dll
01/16/2009 05:08 PM 1 symsrv.yes
01/16/2009 05:07 PM 61,952 zlib1.dll
28 File(s) 15,975,734 bytes

Just trying to get the most work done in the shortest amount of time, to maximize efficency. Also, as I don't think I have any of my machines hidden, would this one work on the other ones, and do I have to have v6.x to run it, as I know at least a couple are still running ver 5.x, they are most likely my older ones. I've always wanted to ask, is there a compelling reason to upgrade to 6 from 5, does it work better all around, or just on newer hardware.

Thanks

---

Thank you for getting this up, just a few hours after getting a reissue old AP wu.

I am glad you made it have both, cause I was going to make it work for both if you didn't. Made life easier. Just had to add the last part of the xml file, and copy the one executable over, stop and restart. Now I can run MB, AP4 and AP5.

Congrats at getting it out. My question is, what is the approximate average speed increase?

---

My movie https://vimeo.com/manage/videos/502242

Al:

Data directory: C:\Documents and Settings\All Users\Application Data\BOINC

This is where your files are, not program files\boinc.

So, NO you are not running the optimized apps.

Al:

You also need to look for a line in your Messages tab that reads:

2/21/2009 8:48:22 PM|SETI@home|Found app_info.xml; using anonymous platform

---

K, I guess I was just a little confused as I have an app_info.xml file in there for some reason, (contents below) that threw me off a little (is this the one I need to edit, or will there also be one in the C:\Documents and Settings\All Users\Application Data\BOINC dir as well?). One other question I have, I noticed that it says Processor features: fpu tsc pae nx sse sse2 mmx but when I bring up Intels Proc ID util, it says it supports Speedstep, Streaming SIMD Extensions 1 thru 4, is this the same as SSE4?


<app_info>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>AK_v8_win_SSE41.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>517</version_num>
<file_ref>
<file_name>AK_v8_win_SSE41.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>527</version_num>
<file_ref>
<file_name>AK_v8_win_SSE41.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>528</version_num>
<file_ref>
<file_name>AK_v8_win_SSE41.exe</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>

---

K, I guess I was just a little confused as I have an app_info.xml file in there for some reason, (contents below) that threw me off a little (is this the one I need to edit, or will there also be one in the C:\Documents and Settings\All Users\Application Data\BOINC dir as well?).

The only location for your app_info.xml file should be located in your BOINC data dir's setiathome.berkeley.edu subfolder, e.g. C:\Documents and Settings\All Users\Application Data\BOINC\projects\setiathome.berkeley.edu if using Windows XP.

One other question I have, I noticed that it says Processor features: fpu tsc pae nx sse sse2 mmx but when I bring up Intels Proc ID util, it says it supports Speedstep, Streaming SIMD Extensions 1 thru 4, is this the same as SSE4?

Intel's Proc ID Util & CPU-Z will be more accurate than BOINC. Yes, Streaming SIMD Extensions is the long name of SSE, so you should be able to support SSE, SSE2, SSE3, SSSE3 & SSE4.1.

---

Ok Al, maybe I can answer a couple of your questions. First, you did at one time try to get an Opt-App but you put it in the wrong place. You should have put it in the Seti folder in the boinc data file. boincdata/projects/setiathome

To find out for sure what instruction set you run, download CPUZ http://www.cpuid.com/cpuz.php and run it. That will help you find the right App.

And last, you won't be able to use that GPU. Seti Cuda requires a minimum of 256MB RAM. Your card only has 128MB.

---

PROUD MEMBER OF Team Starfire World BOINC

My question is, what is the approximate average speed increase?

Well...the newer AP is a bit of a different beast than the old. If you'll pardon the NASCAR analogy, optimized r112 is essentially the same race car as r103, but it's been modified to run on the new track that Berkeley has built. The WUs and analysis are a bit different and in some cases, take a bit longer than the previous AP 5.00 (Gen1). The new opt app is just as efficient as the prior r103, but new AP is slightly different than the old, so the ultimate speed-up is a little less than before, though still wicked-good :P Early indications are credit per hour for new AP + opt apps are actually higher than MB running latest AKv8. Soo....while still taking considerably longer to crunch individual WUs than MB, the "payoff" is actually quite good : > )

I'll defer to Joe for a qualified technical explanation of the processing & analysis differences between v5.00 & v5.03 ; > )

Gecko:

Your explanation is perfect. I talk tech talk about the same broken way you do, and understand it very well.

I've been behind a computer for too many years, I think (first computer touched in 1975 was a Dec PDP-11/45, I was 13 - now everyone can guess how old I am now).

first computer touched in 1975 was a Dec PDP-11/45,).

Where i work we still have a machine that's driven by a PDP-11/55. Yes, we are at the end of a very long corporate upgrade line.

---

Grant
Darwin NT

Gecko:

Your explanation is perfect. I talk tech talk about the same broken way you do, and understand it very well.

I've been behind a computer for too many years, I think (first computer touched in 1975 was a Dec PDP-11/45, I was 13 - now everyone can guess how old I am now).

Yeah..I'm w/ ya! First one I ever used (not owned), was TRS-80 @ '79, followed shortly thereafter by my first intro to BASIC class on Commodore Vic-20. Learned Pascal on Apple II+. Those were fun years.

Thanks for the info, I will correct the install to put it in the right place. I am looking for an Nvidia Quadro NVS 140M on fleabay to upgrade this Dell D830 laptop to either 256 or 512 meg of vid ram (I've heard supposedly both ways were availabe, either 128 dedicated + 128 shared, or 256+256, but as of yet have found neither, still looking)

I presume this package will enhance my performance, even though currently I don't have CUDA working, or should I get a different version if it's on non-cuda machines? Thanks for all the prompt advice/suggestions, I appreciate it.

Oh, one last question, is there an enhanced ap for lunix seti? I am hoping this weekend to get time to get the farm up and running, using Dotsch's
Dotsch/UX - A USB/Diskless/Harddisk BOINC Linux Distribution thread and I'm not sure about installing SETI on it, or more accurately getting it configured properly. I know that his distro has BOINC on it, but not seti. Suggestions on what version to use in a diskless environment, with (eventually) 8 quad core machines booting off the server? Is there even an enhanced app for Lunix? Thanks again guys.

---

Magic!! :))

Given all the caveats surrounding mention of the new Opt App for AP5.3, I expected a "modest" improvement in crunch time compared to Stock (perhaps halving it?). This was in the light of experience of r103 bringing crunch times of AP500 down from about 20 hours to about 6.5 hours on my rig.
But, assuming the progress bar in BM to be as linear for the new App as for the r103, I am at 25% after 2.5 hours using r112 compared to 40 hours to crunch an AP5.3 with the Stock App. So 4x the throughput is on the cards.

Looks like the wizards have done it again :)

F.

---

Alright... I finally tricked the server into giving me some ap503's. I only wanted one and I manually downloaded the stock app...as painful as it will be, so that I can get a baseline and see how much improvement I get.

So far: 17:28:45 @ 8.218%, projecting 212:41:37

That's on an Opteron 2210 @ 1.8GHz.

My P4 530 @ 3.0GHz at work was projecting about 220h last I was able to check. So even with drastically different core speeds, it appears the consensus between the two say it will take about 215h. Man I'm glad I'm only doing one of those with stock. I knew it would be painful, but I needed a baseline.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Man I'm glad I'm only doing one of those with stock. I knew it would be painful, but I needed a baseline.

with only one, the baseline reference value will be limited. Even not considering the 30/30 terminations, there is substantial execution time variation for AP units.

---

Man I'm glad I'm only doing one of those with stock. I knew it would be painful, but I needed a baseline.

with only one, the baseline reference value will be limited. Even not considering the 30/30 terminations, there is substantial execution time variation for AP units.

True, but with the previous ap500, my times were mostly consistent within about 5%. Every now and then with r69 and r103, the completion time would be +/- 20% from the "normal", with or without the early exit from 30/30.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

I looked at this thread and at Lunatics and I found that an Optimized 64bit Astropulse app for XP x64 or Vista x64 doesn't exist. When will this be fixed?

I'm sure when the project code has stabilised a little, and we have some solid background number of opt-stock validations to prove our current codebase validity, we will attempt a recompile. Do note that we have other priorities right now, mainly as AP validators seem to be turned off, all of which are delaying factors in moving further forward in any manner with respect to AstroPulse_v5.

For some insight:
That for most of our particular group of developers, Windows 32 bit is our primary platform is, in most cases, a matter of resources and simple economics. This is likely to gradually change, but I think an apt expression is 'pay peanuts get monkeys' (at least applicable in my case anyway :P)

If in some hurry, and have some skills & appropriate 64 bit tools, there is always the opportunity to download the source yourself and have a go. Just be aware that that the AP code has been in a recent state of relative flux, so some of the recent changes remain relatively unproven. (as compared to AKv8 builds for example)

Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

I looked at this thread and at Lunatics and I found that an Optimized 64bit Astropulse app for XP x64 or Vista x64 doesn't exist. When will this be fixed?

I'm sure when the project code has stabilised a little, and we have some solid background number of opt-stock validations to prove our current codebase validity, we will attempt a recompile. Do note that we have other priorities right now, mainly as AP validators seem to be turned off, all of which are delaying factors in moving further forward in any manner with respect to AstroPulse_v5.

For some insight:
That for most of our particular group of developers, Windows 32 bit is our primary platform is, in most cases, a matter of resources and simple economics. This is likely to gradually change, but I think an apt expression is 'pay peanuts get monkeys' (at least applicable in my case anyway :P)

If in some hurry, and have some skills & appropriate 64 bit tools, there is always the opportunity to download the source yourself and have a go. Just be aware that that the AP code has been in a recent state of relative flux, so some of the recent changes remain relatively unproven. (as compared to AKv8 builds for example)

Jason

I've always been under the impression that while 64-bit shows a small increase, it typically doesn't perform well enough to be worthwhile for compiling it. I think the small gains come from the ability to go larger than 2^32-1 without having to use extra CPU cycles to run an algorithm to be able to count higher than that, and that's where the slight increase in performance comes from. Could be wrong, but it makes sense to me.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

After more crunching time, I have observed this with stock: Opteron 2210 is faster than the P4 530. P4 is at 24h @ ~10%, Opteron is at 22h @ 11%. I'm almost thinking of just going ahead and finishing the crunching with opti. I know it's going to be between 215-240 hours with stock. Yeah, I'm just going to go ahead and do that.

[edit: much better. It's making 0.005-0.010% jumps per interval, and the ~66h ETA is dropping about 30s/interval. That will smooth out after a while, but I can see that's a huge improvement over stock.]

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)