Seti@Home optimized science apps and information
Optimized Seti@Home apps => Discussion Forum => Topic started by: Pappa on 15 May 2009, 03:42:05 pm
-
Richard
Here is the new part... BTW
http://setiathome.berkeley.edu/ap_faq.php New
http://setiathome.berkeley.edu/astropulse_faq.php Old
I found the Notes Section You are talking about
The platform lines have been removed from the MarkJ excerpt.
That's great - but there's also a reference under 'Notes:' (second major section - second line) - a reference to changing the (now removed) platform tags. I think that whole line is now redundant and can be removed - no replacement needed.
I've now updated the BOINC Wiki myself.
As for the Astropulse FAQ Josh has added some things (graphics) but yes the text needs an overhaul.
The FAQ (http://setiathome.berkeley.edu/astropulse_faq.php) is a bit odd - it seems to have gone back to an old version "An app_info.xml file will be made available here", replacing the link to the Advanced Users page which was there previously. But the deadline is still stated to be 14 days, which has never been true for all the nine months that AP has been running.
But my link, and concern, was about the Advanced Users (http://setiathome.berkeley.edu/ap_advanced.php) page, which has live download links to extremely old applications and the sample (pre-optimisation) app_info files which Urs and I wrote.
-
Richard
Here is the new part... BTW
http://setiathome.berkeley.edu/ap_faq.php New
http://setiathome.berkeley.edu/astropulse_faq.php Old
Ah - yes, my mistake. Your 'new' version (ap_faq) is the one which is currently linked from the front page: and that's a problem, because it has the live link to the advanced stuff, which is by now frankly rubbish. It should be removed and de-linked.
And the FAQ still says the deadline is 14 days.
-
Okay, in the NC forum, a machine that had a corrupt/missing Optimized App attempted to download the missing file.
http://setiathome.berkeley.edu/forum_thread.php?id=53645&nowrap=true#895297
-
Okay this works
The first time Boinc complains it can not find the files and then starts the D/L
It then complains
5/16/2009 1:22:04 PM SETI@home Beta Test Finished download of setigraphics_6.03_windows_intelx86.exe
5/16/2009 1:22:04 PM SETI@home Beta Test Finished download of astropulse_5.05_windows_intelx86.exe
5/16/2009 1:22:04 PM SETI@home Beta Test Finished download of ap_graphics_5.05_windows_intelx86.exe
5/16/2009 1:22:04 PM SETI@home Beta Test Started download of ap405.jpg
5/16/2009 1:22:04 PM SETI@home Beta Test [error] Application file setigraphics_6.03_windows_intelx86.exe missing signature
5/16/2009 1:22:04 PM SETI@home Beta Test [error] BOINC cannot accept this file
5/16/2009 1:22:04 PM SETI@home Beta Test [error] Checksum or signature error for setigraphics_6.03_windows_intelx86.exe
5/16/2009 1:22:04 PM SETI@home Beta Test [error] Application file astropulse_5.05_windows_intelx86.exe missing signature
5/16/2009 1:22:04 PM SETI@home Beta Test [error] BOINC cannot accept this file
AS You restart Boinc everything is happy
<app_info>
<app>
<name>setiathome_enhanced</name>
<user_friendly_name>SETI@home Enhanced</user_friendly_name>
</app>
<app>
<name>astropulse</name>
<user_friendly_name>AstroPulse</user_friendly_name>
</app>
<file_info>
<name>setiathome_6.08_windows_intelx86__cuda.exe</name>
<nbytes>1445888.000000</nbytes>
<executable/>
<url>http://boinc2.ssl.berkeley.edu/beta/download/setiathome_6.08_windows_intelx86__cuda.exe</url>
<md5_cksum>F7736D5913A4FF0AAD2CDB02F8A04D94</md5_cksum>
</file_info>
<file_info>
<name>cudart.dll</name>
<nbytes>192512.000000</nbytes>
<url>http://boinc2.ssl.berkeley.edu/beta/download/cudart.dll</url>
<md5_cksum>32C707FA2E3C6DFBA0B11F18678EC215</md5_cksum>
</file_info>
<file_info>
<name>cufft.dll</name>
<nbytes>389120.000000</nbytes>
<url>http://boinc2.ssl.berkeley.edu/beta/download/cufft.dll</url>
<md5_cksum>D34AACA43AB178A52F08D823F8A6BA49</md5_cksum>
</file_info>
<file_info>
<name>libfftw3f-3-1-1a_upx.dll</name>
<nbytes>448600.000000</nbytes>
<url>http://boinc2.ssl.berkeley.edu/beta/download/libfftw3f-3-1-1a_upx.dll</url>
<md5_cksum>FAEF6807B4FE889A33647A661C52D99E</md5_cksum>
</file_info>
<file_info>
<name>seti_607.jpg</name>
<nbytes>9068.000000</nbytes>
<url>http://boinc2.ssl.berkeley.edu/beta/download/seti_607.jpg</url>
<md5_cksum>75AF6760157FDF8B2178971CB8F27851</md5_cksum>
</file_info>
<file_info>
<name>setiathome_6.03_windows_intelx86.exe</name>
<nbytes>406016.000000</nbytes>
<executable/>
<url>http://boinc2.ssl.berkeley.edu/beta/download/setiathome_6.03_windows_intelx86.exe</url>
<md5_cksum>D53249AADB1D72CC19DB36359E63425A</md5_cksum>
</file_info>
<file_info>
<name>setigraphics_6.03_windows_intelx86.exe</name>
<nbytes>267776.000000</nbytes>
<executable/>
<url>http://boinc2.ssl.berkeley.edu/beta/download/setigraphics_6.03_windows_intelx86.exe</url>
<md5_cksum>357A9FD375147BDF7545784B0891AAA9</md5_cksum>
</file_info>
<file_info>
<name>seti_603.jpg</name>
<nbytes>9068.000000</nbytes>
<url>http://boinc2.ssl.berkeley.edu/beta/download/seti_603.jpg</url>
<md5_cksum>75AF6760157FDF8B2178971CB8F27851</md5_cksum>
</file_info>
<file_info>
<name>astropulse_5.05_windows_intelx86.exe</name>
<nbytes>479232.000000</nbytes>
<executable/>
<url>http://boinc2.ssl.berkeley.edu/beta/download/astropulse_5.05_windows_intelx86.exe</url>
<md5_cksum>DE5D72065E2F2D1B30F39A2CFDBF6A93</md5_cksum>
</file_info>
<file_info>
<name>ap_graphics_5.05_windows_intelx86.exe</name>
<nbytes>294912.000000</nbytes>
<executable/>
<url>http://boinc2.ssl.berkeley.edu/beta/download/ap_graphics_5.05_windows_intelx86.exe</url>
<md5_cksum>1EBB7BFBC4D767118A325CE9E11D2F31</md5_cksum>
</file_info>
<file_info>
<name>ap405.jpg</name>
<nbytes>7653.000000</nbytes>
<url>http://boinc2.ssl.berkeley.edu/beta/download/ap405.jpg</url>
<md5_cksum>05C865530822A969E8BFBA67581F8D2A</md5_cksum>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.008666</avg_ncpus>
<max_ncpus>0.008666</max_ncpus>
<flops>5245714285.714286</flops>
<plan_class>cuda</plan_class>
<api_version>6.3.22</api_version>
<file_ref>
<file_name>setiathome_6.08_windows_intelx86__cuda.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
<open_name>cudart.dll</open_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
<open_name>cufft.dll</open_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
<open_name>libfftw3f-3-1-1a_upx.dll</open_name>
</file_ref>
<file_ref>
<file_name>seti_607.jpg</file_name>
<open_name>seti_logo</open_name>
</file_ref>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>603</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>1.000000</avg_ncpus>
<max_ncpus>1.000000</max_ncpus>
<flops>3026692970.287821</flops>
<file_ref>
<file_name>setiathome_6.03_windows_intelx86.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>setigraphics_6.03_windows_intelx86.exe</file_name>
<open_name>graphics_app</open_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
<open_name>libfftw3f-3-1-1a_upx.dll</open_name>
</file_ref>
<file_ref>
<file_name>seti_603.jpg</file_name>
<open_name>seti_logo</open_name>
</file_ref>
</app_version>
<app_version>
<app_name>astropulse</app_name>
<version_num>505</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>1.000000</avg_ncpus>
<max_ncpus>1.000000</max_ncpus>
<flops>3026692970.287821</flops>
<api_version>6.7.4</api_version>
<file_ref>
<file_name>astropulse_5.05_windows_intelx86.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>ap_graphics_5.05_windows_intelx86.exe</file_name>
<open_name>graphics_app</open_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
<open_name>libfftw3f-3-1-1a_upx.dll</open_name>
</file_ref>
<file_ref>
<file_name>ap405.jpg</file_name>
<open_name>seti_logo</open_name>
</file_ref>
</app_version>
</app_info>
-
With the various stop and restarts that I went through, anything from Seti is going to look for a Signing Key, because the MD5 checksum "matched" it kept the files in place (after the restart). The client_state.xml was fine (contained no errors) and matched what was in the app_info.xml
Yes I am going to be told to trim the app_info.xml again.
What this tells me is if there is a repository for Lunatics Apps and the url and MD5 checksum are written to the app_info.xml the a corrupt file might easily be replaced. I am going to guess that if the files are located at some location other than seti or a project that requires signing the signature checking will not be forced. The MD5 checksum verifies the file came down correctly.
-
I've been digging through the BOINC source, and am more than 90% certain of the following:
1. Download file verification uses either a signature or an MD5 checksum. If a file is considered an application file it wants the signature. If it's just a project file, the MD5 is used.
2. If a file is listed in a <file_ref> within an <app_version> it is automatically considered an application file, otherwise it's not.
3. If an application file <file_ref> came from an app_info.xml, BOINC doesn't delete the file even if it cannot be used.
4. The "cannot accept" means at that point BOINC won't link that app_version to tasks. That could cause freshly downloaded tasks to be deleted.
My conclusion is the <md5_cksum> entries aren't needed in the app_info.xml listed here, though if you had the AUTHORS, etc. they'd need those entries. They aren't protected by a <file_ref>.
=======================
Opinion: I think for now we should drop the idea of automatic downloads of individual files from the project. If a host doesn't have the right stuff to use a full set of Lunatics apps, and doesn't already have the required stock files, the installer should not complete. Maybe it would be possible to advise the user what's lacking, maybe just say the installer can't do what has been requested and advise where to ask for help. The basic advice would be to get the stock applications running, then try the installer again.
Joe
-
Joe et al
Go look at this. It appears with "progress" Boinc has gotten smarter in attempting to recover from corrupt files. It shows a Lunatics App that Boinc was attemting D/L from Seti.
Okay, in the NC forum, a machine that had a corrupt/missing Optimized App attempted to download the missing file.
http://setiathome.berkeley.edu/forum_thread.php?id=53645&nowrap=true#895297
-
Joe et al
Go look at this. It appears with "progress" Boinc has gotten smarter in attempting to recover from corrupt files. It shows a Lunatics App that Boinc was attemting D/L from Seti.
...
That's not new, and it isn't due to a corrupt file but a missing file. Once a file with <file_ref> from an app_info.xml is in the project folder, the core client doesn't check it other than to see if it is there. If a missing file is linked to some work, the core client makes a desperate attempt to download it from the project since that's the only known base URL. It's more a bug than a feature.
Joe
-
Is it possible to use current app_info.xml structure to assign different versions to different apps with the same plan?
I.e. is it possible to have both SSE4.1 and SSSE3x apps installed under 6.03 and 6.04 versions (for example) ?
-
Is it possible to use current app_info.xml structure to assign different versions to different apps with the same plan?
I.e. is it possible to have both SSE4.1 and SSSE3x apps installed under 6.03 and 6.04 versions (for example) ?
Yes, I think it is - we did that with the first (awkward) AP transition from v4.35 to v5.00 last November.
But the problem is, all new work fetched is assigned to the highest available version number - so in your example, once any v6.03 work currently in the cache is completed, the SSE4.1 app would never be used again unless you went through the 'rebranding' exercise.
-
Well, it's just what I wanna do ;D
As server will not respect SSE4.1 over SSSE3x performance gain for some of ARs anyway, I will use rebrandng script (the same that I use for CUDA) to slightly optimize CPU prcoessing at the same time when I do CUDA processing optimization so no additional BOINC shutdown will be involved.
More tests under heavy loaded system condtions are needed though.
-
Well, it's just what I wanna do ;D
Thought you might - that's why I mentioned it :P