Now you should restart boinc-client and try it. At this moment you run without a wisdom-file, so don't expect that it run too fast at this point.
did the above two steps but put on a 23 hour delay. reset the project but no work
Quote:
The script is currently under development. The output is maybe a little bit confusing ;).
Quote:
finished in seconds with the following issues:
./boinc-app-eah-brp/debian/extra/create_wisdomf_eah_brp.sh I: Computing wisdom file for BRP4 and FGRP projects. This will take several hours if not days.
I: Generating wisdom for (2^22)*3 sample projects (BRP) W: Not recognising platform 'armv7l' as sufficiently capable for 2^26 bit projects fftw-wisdom:
This is normal behavior. The script generates a wisdom for the FGRP (2^26) too. But there is no need for that wisdom on arm-devices. So the script skips the generation of that.
That is okay, there is no system-wisdom at that point.
The problem with the short runtime of the wisdom-generation is caused by a bug in the current libfftw3 debian/ubuntu-package on ARM-devices. fftw can not access the performance-timer on ARM-devices. So it can't make the necessary measurements to generate a wisdom. Instead it only guesses the right settings and put that into the wisdomfile. Thats why it only takes seconds to compute.
We fixed that. But the fixed version(3.3.6p2-1) is only available under debian sid/unstable or ubuntu artful/devel at the moment.
I guess you are running ubuntu xenial on your odroid?
yes
Quote:
There are several ways to get the new Version of the fftw3-package. This is only one way and not the best way to do it.
will look into doing this tomorrow
Quote:
edit your sources.list and add the following lines to it:
I am too swamped to address the backporting to earlier Ubuntu releases myself. Any volunteers out there? Ubuntu has the fantastic well-established concepts of PPA (Personal Package Archives) that would do the job to auto-build the package upon submitting a so called source package. Everyone could do that, really, also computer-enthusiastic non-programmers, ... I am too swamped to promise any long term commitment myself on this, so I better reject any active involvement on Ubuntu backports upfront. The backports for Debian I'll help out with.
did the above two steps but put on a 23 hour delay. reset the project but no work
I need some more infos to help. Could provide the output of boinc's eventlog?
In Boinc-Manager go to Tools -> Event Log; (Use the copy all Button)
I did a reset on the project and what follows is the content of "Event Log". There are no WUs/tasks showing up in Boinc_Manager.
Tue 23 May 2017 02:37:13 PM EDT | Einstein@Home | Resetting project
Tue 23 May 2017 02:37:17 PM EDT | Einstein@Home | Master file download succeeded
Tue 23 May 2017 02:37:22 PM EDT | Einstein@Home | Sending scheduler request: To fetch work.
Tue 23 May 2017 02:37:22 PM EDT | Einstein@Home | Requesting new tasks for CPU
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | Scheduler request completed: got 0 new tasks
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | No work sent
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | (reached daily quota of 6 tasks)
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | Project has no jobs available
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of einstein_icon.png
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Android.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Arecibo_full.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Arecibo_platform.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Fermi_grsky.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Fermi_satellite.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of GW_BBH1.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of GW_BBH2.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_Hanford.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_Livingston.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_laser.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_optics.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_schematic.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_seisisol.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_vacuum.jpg
Ahh, okay. It doesn't seems to found a app_info.xml. It seems that the installation using dpkg -i havn't worked *strange*.
Could you prove that? Go to your boinc-project-folder: /var/lib/boinc-client/projects/einstein.phys.uwm.edu in my case.
There should be a link to a app_info.xml and a einsteinbinary_BRP4. Both should have boinc as owner.
If that's not the case, you have to manually copy the einsteinbinary_BRP4 to this folder change the owner to boinc and create a app_info.xml. If you need help with that write me a PM ;)
Ahh, okay. It doesn't seems to found a app_info.xml. It seems that the installation using dpkg -i havn't worked *strange*.
Could you prove that? Go to your boinc-project-folder: /var/lib/boinc-client/projects/einstein.phys.uwm.edu in my case.
There should be a link to a app_info.xml and a einsteinbinary_BRP4. Both should have boinc as owner.
If that's not the case, you have to manually copy the einsteinbinary_BRP4 to this folder change the owner to boinc and create a app_info.xml. If you need help with that write me a PM ;)
The directory: /var/lib/boinc-client/projects/einstein.phys.uwm.edu contains the following:
-rw-r--r-- 1 boinc boinc 489 Jul 182016 app_info.xml <--- not a link and note date probably wrong/incorrect file
First let me say that I have a single WU processing on the Odroid XU4. How this happened though I am not sure. For this reason I am replying to N30DG's original procedure and where I believe an problem may have occurred.
N30dG wrote:
You could build the BRP-App from our github-repositroy at your own. It's not that hard.
It was noted in an earlier post that the above command might have failed to generate the app_info.xml file. I noted that there was an app_info.xml file in /var/lib/boinc-client/projects/einstein.phys.uwm.edu and that it had a 2016 date. I removed it and reran the above dpkg command and this time a new app_info.xml link was generated in the project directory. There was also a new link for einsteinbinary_BRP4 -> /usr/lib/boinc-app-eah-brp/einsteinbinary_BRP4
I restarted boinc-client and reset it. Still no work until I made a change to my preference set (generic in my case for the XU4) - USE CPU -> NO, and Run CPU versions of apps for which GPU versions are available -> Yes
I then did a project update and received one WU which is estimating at 3 hours 47 minutes.
Not sure if any of this makes sense. The short version might be that I just "fat fingered" N30DG's procedure
It was noted in an earlier post that the above command might have failed to generate the app_info.xml file. I noted that there was an app_info.xml file in /var/lib/boinc-client/projects/einstein.phys.uwm.edu and that it had a 2016 date. I removed it and reran the above dpkg command and this time a new app_info.xml link was generated in the project directory. There was also a new link for einsteinbinary_BRP4 -> /usr/lib/boinc-app-eah-brp/einsteinbinary_BRP4
You tryed to use my optimized AArch64-App before, that can not work. This app only works with cortex-A53 in 64bit-mode! Thats why the installation of your self-build package fails. It couldn't link the app-info, because there was already a file with that name.
Quote:
I restarted boinc-client and reset it. Still no work until I made a change to my preference set (generic in my case for the XU4) - USE CPU -> NO, and Run CPU versions of apps for which GPU versions are available -> Yes
I then did a project update and received one WU which is estimating at 3 hours 47 minutes.
Pretty fast, considering that you don't have a proper wisdom at the moment.
Quote:
Not sure if any of this makes sense. The short version might be that I just "fat fingered" N30DG's procedure
8. Tell me if that worked for you, and I tell you how to get it even faster .
Try to upgrade your fftw-version to 3.3.6p2-1 before. Then run the create_wisdomf_eah_brp.sh again and copy the new wisdom to /etc/fftw/wisdomf
It should take some hours to compete.
I'm excited to hear about your run-times after that. I guess it will much blow 3h. Maybe below 2h.
N30dG wrote:Quote:Quote:4.
)
did the above two steps but put on a 23 hour delay. reset the project but no work
yes
will look into doing this tomorrow
Hello, Ubuntu already has
)
Hello, Ubuntu already has boinc-app-eah-brp in the upcoming "artful" distribution
(https://packages.ubuntu.com/search?keywords=boinc-app-eah-brp)
and so is the fftw3 3.3.6p2 version (https://packages.ubuntu.com/source/artful/fftw3).
I am too swamped to address the backporting to earlier Ubuntu releases myself. Any volunteers out there? Ubuntu has the fantastic well-established concepts of PPA (Personal Package Archives) that would do the job to auto-build the package upon submitting a so called source package. Everyone could do that, really, also computer-enthusiastic non-programmers, ... I am too swamped to promise any long term commitment myself on this, so I better reject any active involvement on Ubuntu backports upfront. The backports for Debian I'll help out with.
Here is a PPA for the regular BOINC packages, which also includes SETI@Home for the time your distribution is not directly supported for BRP: https://launchpad.net/~costamagnagianfranco/+archive/ubuntu/boinc
did the above two steps but
)
I need some more infos to help. Could provide the output of boinc's eventlog?
In Boinc-Manager go to Tools -> Event Log; (Use the copy all Button)
N30dG wrote: did the above
)
I did a reset on the project and what follows is the content of "Event Log". There are no WUs/tasks showing up in Boinc_Manager.
Tue 23 May 2017 02:37:13 PM EDT | Einstein@Home | Resetting project
Tue 23 May 2017 02:37:17 PM EDT | Einstein@Home | Master file download succeeded
Tue 23 May 2017 02:37:22 PM EDT | Einstein@Home | Sending scheduler request: To fetch work.
Tue 23 May 2017 02:37:22 PM EDT | Einstein@Home | Requesting new tasks for CPU
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | Scheduler request completed: got 0 new tasks
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | No work sent
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | (reached daily quota of 6 tasks)
Tue 23 May 2017 02:37:26 PM EDT | Einstein@Home | Project has no jobs available
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of einstein_icon.png
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Android.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Arecibo_full.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Arecibo_platform.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Fermi_grsky.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of Fermi_satellite.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of GW_BBH1.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of GW_BBH2.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_Hanford.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_Livingston.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_laser.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_optics.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_schematic.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_seisisol.jpg
Tue 23 May 2017 02:37:28 PM EDT | Einstein@Home | Started download of LIGO_vacuum.jpg
Ahh, okay. It doesn't seems
)
Ahh, okay. It doesn't seems to found a app_info.xml. It seems that the installation using dpkg -i havn't worked *strange*.
Could you prove that? Go to your boinc-project-folder: /var/lib/boinc-client/projects/einstein.phys.uwm.edu in my case.
There should be a link to a app_info.xml and a einsteinbinary_BRP4. Both should have boinc as owner.
If that's not the case, you have to manually copy the einsteinbinary_BRP4 to this folder change the owner to boinc and create a app_info.xml. If you need help with that write me a PM ;)
N30dG wrote:Ahh, okay. It
)
The directory: /var/lib/boinc-client/projects/einstein.phys.uwm.edu contains the following:
-rw-r--r-- 1 boinc boinc 489 Jul 18 2016 app_info.xml <--- not a link and note date probably wrong/incorrect file
and
lrwxrwxrwx 1 boinc boinc 46 May 22 20:42 einsteinbinary_BRP4 -> /usr/lib/boinc-app-eah-brp/einsteinbinary_BRP4
ls -l /usr/lib/boinc-app-eah-brp/einsteinbinary_BRP4
-rwxr-xr-x 1 root root 68396 May 21 19:08 /usr/lib/boinc-app-eah-brp/einsteinbinary_BRP4 - note ownership
contents of app_info.xml :
<app_info>
<app>
<name>einsteinbinary_BRP4</name>
</app>
<file_info>
<name>einsteinbinary_BRP4_0.13_AARCH64-unknown-linux-gnu</name>
<executable/>
</file_info>
<app_version>
<app_name>einsteinbinary_BRP4</app_name>
<version_num>013</version_num>
<file_ref>
<file_name>einsteinbinary_BRP4_0.13_AARCH64-unknown-linux-gnu</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>
First let me say that I have
)
First let me say that I have a single WU processing on the Odroid XU4. How this happened though I am not sure. For this reason I am replying to N30DG's original procedure and where I believe an problem may have occurred.
It was noted in an earlier post that the above command might have failed to generate the app_info.xml file. I noted that there was an app_info.xml file in /var/lib/boinc-client/projects/einstein.phys.uwm.edu and that it had a 2016 date. I removed it and reran the above dpkg command and this time a new app_info.xml link was generated in the project directory. There was also a new link for einsteinbinary_BRP4 -> /usr/lib/boinc-app-eah-brp/einsteinbinary_BRP4
I restarted boinc-client and reset it. Still no work until I made a change to my preference set (generic in my case for the XU4) - USE CPU -> NO, and Run CPU versions of apps for which GPU versions are available -> Yes
I then did a project update and received one WU which is estimating at 3 hours 47 minutes.
Not sure if any of this makes sense. The short version might be that I just "fat fingered" N30DG's procedure
Zitat:It was noted in an
)
You tryed to use my optimized AArch64-App before, that can not work. This app only works with cortex-A53 in 64bit-mode! Thats why the installation of your self-build package fails. It couldn't link the app-info, because there was already a file with that name.
Pretty fast, considering that you don't have a proper wisdom at the moment.
Zitat:
Try to upgrade your fftw-version to 3.3.6p2-1 before. Then run the create_wisdomf_eah_brp.sh again and copy the new wisdom to /etc/fftw/wisdomf
It should take some hours to compete.
I'm excited to hear about your run-times after that. I guess it will much blow 3h. Maybe below 2h.
Quote:Try to upgrade
)
did the following:
1. downloaded 3.3.6 version from site
2. gunziped the file
3. tar -xvf fftw-3.3.6-pl2.tar
4. cd fftw-3.3.6-pl2
5. ./configure
6. make
7. make install
the above seemed to run fine.
ran
./boinc-app-eah-brp/debian/extra/create_wisdomf_eah_brp.sh this took maybe 2 minutes
sudo mkdir /etc/fftw
sudo mv /tmp/wisdomf /etc/fftw/
restarted boinc-cllient
also had to change account/preferences/project/usecpu to yes from no.
an update now is running 4 consecutive tasks with duration of ~4 hours.
Again not sure if the wisdom file was correctly generated. What follows is the content of the generate file:
(fftw-3.3.4 fftwf_wisdom #x4a633eef #xb5a95564 #x91014bdd #x9c85ce5f (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #x3ccfdb1a #xf8b7fb16 #xb777192b #xa86989c1) (fftwf_codelet_t1fv_32_neon 0 #x11048 #x11048 #x0 #x13a350ff #xccbd0d68 #x1df44d80 #x16ee0d41) (fftwf_codelet_hc2cfdftv_16_neon 0 #x11048 #x11048 #x0 #xead3fd9e #xffec487c #x5b618d5e #x8f8f3b12) (fftwf_codelet_r2cfII_16 2 #x11048 #x11048 #x0 #xf4d971ab #x381e69c1 #xc4398fe0 #x3f2135b1) (fftwf_codelet_t2fv_16_neon 0 #x11048 #x11048 #x0 #x3d0b62a7 #x15d0e0a0 #xd8a2423f #xa9a6da1c) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #xbddeb44e #xfd7343e7 #x3c8fc850 #x6888d042) (fftwf_codelet_t1fv_12_neon 0 #x11048 #x11048 #x0 #x821ed100 #xa1017c4b #x40993259 #x7860b2a1) (fftwf_codelet_r2cf_16 2 #x11048 #x11048 #x0 #x8f3ef9f7 #xe67e11ab #xe25a4700 #x8eed687a) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #xa65ca367 #xee5c44cb #x0578eeed #x986cea5e) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #xd97dcac9 #xd6110c1d #x25bf8814 #xe9a1ed91) (fftwf_codelet_n1fv_128_neon 0 #x11048 #x11048 #x0 #xdcd9ab89 #x9279272f #x45725e3d #xb22380a2)
[EDIT] the temp on the xu4 is averaging 92C
Zitat: did the following: 1.
)
You configured fftw wrong. You need to apply --enable-float --with-slow-timer --enable-neon to the ./configure.
And I'm not shure what make install does. I would suggest you to install the libfftw3-dev package from artful or devel instead.