Message boards :
SETI@home Enhanced :
Setiathome_v8 for Raspberry Pi?
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · Next
Author | Message |
---|---|
Send message Joined: 22 Jul 15 Posts: 21 Credit: 113,162 RAC: 0 |
mimo, Tom and Mark, please put the following app_config.xml in your Seti Beta project directories: The information provided by this is quite useful. Thanks Claggy. It shows that the two 8.01 apps currently on Beta are identical and both support neon if it is available. I compiled the armv6l and armv7l apps on a Pi1 and Pi2 respectively. This is the output from my Pi1. It explains to me why both apps run at the same speed on a Pi2. Clever coding by whoever wrote the software. It looks like we only need one app to support Linux armhf devices. <core_client_version>7.4.23</core_client_version> <![CDATA[ <stderr_txt> setiathome_v8 8.00 Revision: 3304 g++ (Raspbian 4.9.2-10) 4.9.2 libboinc: BOINC 7.7.0 Work Unit Info: ............... WU true angle range is : 2.721674 features: half thumb fastmult vfp edsp java tls Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.007918 0.00000 test vfp_GetPowerSpectrum 0.002828 0.00000 test neon_GetPowerSpectrum not supported on CPU vfp_GetPowerSpectrum 0.002828 0.00000 choice v_ChirpData 0.347058 0.00000 test fpu_ChirpData 0.227311 0.94721 test fpu_opt_ChirpData 0.347020 0.00000 test fpu_opt_ChirpData 0.347020 0.00000 choice v_Transpose 0.031991 0.00000 test v_Transpose2 0.031932 0.00000 test v_Transpose4 0.032902 0.00000 test v_Transpose8 0.068533 0.00000 test fftwf_transpose 0.035408 0.00000 test v_pfTranspose2 0.085403 0.00000 test v_pfTranspose4 0.056317 0.00000 test v_pfTranspose8 0.103865 0.00000 test v_vfpTranspose2 0.030517 0.00000 test v_vfpTranspose2 0.030517 0.00000 choice FPU opt folding 0.013012 0.00000 test opt VFP folding 0.010628 0.17083 test opt NEON folding not supported on CPU FPU opt folding 0.013012 0.00000 choice Test duration 36.78 seconds Flopcounter: 16155043434416.027344 Spike count: 1 Autocorr count: 1 Pulse count: 0 Triplet count: 2 Gaussian count: 0 00:43:18 (28421): called boinc_finish(0) </stderr_txt> ]]> This is the output from the armv6l app on one of my Pi2s: <core_client_version>7.4.23</core_client_version> <![CDATA[ <stderr_txt> setiathome_v8 8.00 Revision: 3304 g++ (Raspbian 4.9.2-10) 4.9.2 libboinc: BOINC 7.7.0 Work Unit Info: ............... WU true angle range is : 2.722039 features: half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.003331 0.00000 test vfp_GetPowerSpectrum 0.001433 0.00000 test neon_GetPowerSpectrum 0.002506 0.00000 test vfp_GetPowerSpectrum 0.001433 0.00000 choice v_ChirpData 0.185788 0.00000 test fpu_ChirpData 0.173306 0.94721 test fpu_opt_ChirpData 0.208163 0.00000 test v_ChirpData 0.185788 0.00000 choice v_Transpose 0.164408 0.00000 test v_Transpose2 0.085524 0.00000 test v_Transpose4 0.043768 0.00000 test v_Transpose8 0.082944 0.00000 test fftwf_transpose 0.028893 0.00000 test v_pfTranspose2 0.072353 0.00000 test v_pfTranspose4 0.039296 0.00000 test v_pfTranspose8 0.063141 0.00000 test v_vfpTranspose2 0.085758 0.00000 test fftwf_transpose 0.028893 0.00000 choice FPU opt folding 0.005477 0.00000 test opt VFP folding 0.004625 0.17082 test opt NEON folding 0.004049 0.00000 test opt NEON folding 0.004049 0.00000 choice Test duration 32.95 seconds Flopcounter: 16155100705679.298828 Spike count: 0 Autocorr count: 3 Pulse count: 0 Triplet count: 0 Gaussian count: 0 03:37:44 (15303): called boinc_finish(0) </stderr_txt> ]]> Now we just need to figure out the ChirpData issue Claggy mentions above. - Tom |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
Mark - I looked through the code you linked. There is no mention of fftw_import_system_wisdom or any fftw_import for that matter, unless its in one of the includes. From the fftw docs it says it looks in /etc/fftw/wisdom for the system wisdom. Wisdom This is for a Pi2 and would also probably work on a Parallella but NOT the Pi Zero, B or Pi B+. If my B+ was running Stretch I could generate one for it but its running Jessie which has an older version of fftw. The wisdom files have to be generated for the same version of fftw that is being used by the app. (fftw-3.3.4 fftwf_wisdom #xca4daf64 #xc8f59ea6 #x586875c9 #x14018994 (fftwf_codelet_t1_4 0 #x11048 #x11048 #x0 #x10f6adc9 #xd9dae2d1 #x6b1f472c #xd5849a55) (fftwf_codelet_n1_64 0 #x11048 #x11048 #x0 #x44ad5c21 #x68343464 #xe4bbe970 #xe20f37be) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #x4ada737a #x8f11eeb5 #xddb31944 #xd61e48ef) (fftwf_codelet_n1_64 0 #x11048 #x11048 #x0 #xa4a64e81 #xdbcacf7d #x26e5d444 #x5f1dc192) (fftwf_codelet_n1_32 0 #x11048 #x11048 #x0 #x17325693 #xc21d61d1 #xd753940f #xb29ba4c4) (fftwf_codelet_r2cf_64 0 #x11048 #x11048 #x0 #x9d75d2bf #x7f5f2a0c #x2dff2a21 #xef5d11be) (fftwf_codelet_q1_4 0 #x11048 #x11048 #x0 #xe28d329e #x222b98fb #xb137951b #xfa3b384f) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #x8641f833 #xbd81619c #xb3bea8e0 #xab131e93) (fftwf_codelet_n1_32 0 #x11048 #x11048 #x0 #x5e624a61 #xe63ce5bb #xa070200f #xfdf1ad24) (fftwf_codelet_t1_32 0 #x11048 #x11048 #x0 #x9ee69f4a #x6a8d6e6c #xe8efd414 #x87e48fba) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #x9131a445 #x23a1aeda #xbe072318 #x90a4b596) (fftwf_ct_genericbuf_register 5 #x11048 #x11048 #x0 #x9f5ed26c #xfde5d368 #xeabd39db #x48515ead) (fftwf_codelet_r2cf_64 0 #x11448 #x11448 #x0 #xdc9b9c98 #x75a724ba #xb1895b5c #x348ad7ea) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #xe1efb4a4 #xb22b258c #xd3b9fdc4 #x6f541f4b) (fftwf_codelet_r2cf_32 0 #x11048 #x11048 #x0 #x5b844cea #x2a920b74 #xf82a54aa #xee1bc380) (fftwf_codelet_t1_8 0 #x11048 #x11048 #x0 #x743f1051 #x067f64af #x63eb2004 #x1f4ba210) (fftwf_codelet_n1_32 0 #x11048 #x11048 #x0 #x959a8189 #x531a3285 #xd2f4d4b2 #x8d515a33) (fftwf_codelet_n1_32 0 #x11048 #x11048 #x0 #xde222597 #x77efec0d #x47c88d24 #x8d859a94) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #xd5f3c653 #xff9ae8b7 #x1117daf1 #xd32171d9) (fftwf_reodft010e_r2hc_register 0 #x11040 #x11048 #x0 #x3f656c48 #x724e1d4c #x4cae1a6c #x04328e7d) (fftwf_codelet_n1_32 0 #x11048 #x11048 #x0 #xe064793f #x569bde2b #x56f73d35 #xfa150854) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #x076ffd7d #x528848e9 #xc5ba9827 #x5f9b2db6) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #x5f3eb4e1 #x0cb00c65 #x7b17191b #xf08cb99e) (fftwf_codelet_n1_16 0 #x11048 #x11048 #x0 #xd2d1dc80 #x2b7adad4 #x66b3f374 #x1c9b4580) (fftwf_codelet_hf_64 0 #x11048 #x11048 #x0 #x4b7606fc #x09d8ba9e #x61657d88 #xa7719fef) (fftwf_codelet_t1_32 0 #x11048 #x11048 #x0 #x52c16d5c #x78672292 #xb152d1c2 #x86ccca81) (fftwf_codelet_t1_64 0 #x11048 #x11048 #x0 #x1a1d813e #xfbac2c22 #x64337de8 #x46594bfd) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #xd2c09bd1 #x1f61b89d #xa82b12b6 #x5fca72cb) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #x4018e315 #xc69945e6 #x036dd400 #xb02b8904) (fftwf_codelet_n1_64 0 #x11048 #x11048 #x0 #x88c983e2 #x0f8bb65d #x9c0867d4 #x2500d51f) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #xbf6b49c6 #xe8902713 #x1b11fd2a #x3af5fd5e) (fftwf_rdft_indirect_register 1 #x11048 #x11048 #x0 #x00b20725 #xfe2286c5 #x99047c24 #x07cb8d59) (fftwf_codelet_t1_8 0 #x11048 #x11048 #x0 #x7a273437 #xbd621258 #x859a74d3 #x7639c6b7) (fftwf_ct_genericbuf_register 0 #x11048 #x11048 #x0 #x63a7f5a2 #x90744be9 #x1ae07cc5 #x62b30107) (fftwf_codelet_t1_4 0 #x11048 #x11048 #x0 #xae8767a9 #x551ea32f #xf88f40c1 #x69374d6d) (fftwf_codelet_n1_16 0 #x11048 #x11048 #x0 #xaaa33cd1 #xb263f7d8 #x0e91208a #x8e90cb5b) (fftwf_codelet_q1_8 0 #x11048 #x11048 #x0 #xed9be1c9 #x5e484970 #x0a34c047 #x9a62f6f2) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #xe4ecbec8 #x5af68578 #x2126ccab #xeb986b5a) (fftwf_codelet_r2cfII_32 0 #x11048 #x11048 #x0 #x7ef04cb6 #x382f909b #x68cbfeb0 #xdf48f380) (fftwf_codelet_t1_8 0 #x11048 #x11048 #x0 #xd33a7356 #x0498cc6f #xb6e32544 #x959f85bc) (fftwf_codelet_t1_8 0 #x11048 #x11048 #x0 #x3378193a #xf13a68c8 #x9762cbdf #x6d62549e) (fftwf_rdft_rank0_register 8 #x11048 #x11048 #x0 #x1160fed4 #x174482dc #x76c6c032 #x93691d13) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #x2079edd0 #x68d66fd4 #x063f6ab3 #xbf3e1613) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #x593773d9 #x2ae3a36a #x928e68af #xc8a37d68) (fftwf_codelet_n1_8 0 #x11048 #x11048 #x0 #x53f03e12 #x1cdb8ef8 #x83e36ba7 #x8ebcf0a1) (fftwf_codelet_n1_64 0 #x11048 #x11048 #x0 #xe10a4f89 #x79f5831f #x6ad85b28 #x06b20bf5) (fftwf_codelet_q1_4 0 #x11048 #x11048 #x0 #x0edc5f24 #x227c0867 #x7b6287cd #x88180ae1) (fftwf_codelet_r2cfII_64 0 #x11048 #x11048 #x0 #x4cac8088 #x8fcf5fdd #x58ff6170 #xa5681aee) (fftwf_codelet_n1_64 0 #x11048 #x11048 #x0 #xd4962c68 #x6d107867 #xd8ca3432 #xe26dbbfe) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #x2c49251b #xba5d8222 #x5de75edb #xca987682) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #x1ab0d001 #x877837d4 #x7816af06 #x038f268e) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #x4ca75661 #x503b8204 #xf8f788d8 #xdfae867d) (fftwf_codelet_t1_16 0 #x11048 #x11048 #x0 #xd17e8d64 #x641ca94a #x54b5d93f #xd008b4d2) (fftwf_codelet_n1_32 0 #x11048 #x11048 #x0 #xb970fcaa #x51ab2de2 #x99822498 #xc20f2f34) (fftwf_codelet_n1_64 0 #x11048 #x11048 #x0 #x5277952d #x7015940c #xac797b56 #x823a4353) (fftwf_dft_vrank_geq1_register 1 #x11048 #x11048 #x0 #xd32ea078 #xd94b02eb #x25b6801e #x455b660f) (fftwf_codelet_n1_16 0 #x11048 #x11048 #x0 #x15fef913 #x9630d295 #xdf912c30 #xacf10560) (fftwf_codelet_n1_32 0 #x11048 #x11048 #x0 #xd96df037 #x47380ef2 #x9db3e7f6 #xffbad50b) (fftwf_codelet_hf_32 0 #x11048 #x11048 #x0 #xc6ec8db1 #xc662d138 #x8fb51dbb #xeb99e443) ) |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
Added my B+ for app testing. Its running a "medium" overclock (900Mhz). Its Host 78198 |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
Umm forget what I said above about the import wisdom. Stupid IE11 find command doesn't work on that page. Had a closer look at the code and it does an import from string. It looks like it's looking in the default fftw folder for wisdom.sah, but if I am reading the code correctly will ignore it when it exceeds 512 bytes. I will have to check how big the one I posted above is, but I suspect it's more than 512 bytes. [edit] Yep it 5.4k... /etc/fftw $ ls wisdom.sah -lh |
Send message Joined: 29 May 06 Posts: 1037 Credit: 8,440,339 RAC: 0 |
The Stock code from seti_boinc generates a fftw wisdom file in the wu's slot directory every time the app runs for the first time, The optimised code from branches, generates a wisdom file named after the CPU found in the project directory, on the first generation(s) it does a fast plan, on the 2nd start it does a more thorough planning, after that no more generations are needed, saving that start up cost, it may be possible to port those changes. Claggy |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
The Stock code from seti_boinc generates a fftw wisdom file in the wu's slot directory every time the app runs for the first time, Or we could relax the 512 byte rule from stock which might be easier. I had a look in one of the slot directories on the Pi2 without a wisdom file and there is indeed one in the slot directory. I don't know how long they take to generate but when I did the one manually it was a minute or two which isn't much of the run time (around 27 hours in my case). |
Send message Joined: 4 Aug 08 Posts: 11 Credit: 1,437,079 RAC: 0 |
if ((file_size(wisdom_path.c_str(),wisdom_size)==0) && (wisdom_size>512)) { if ((wisdom=boinc_fopen(wisdom_path.c_str(),"r"))) { means open when size is more then 512 bytes, in my case more then 6kb so it is always loaded |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
Is wisdom_path.c_str() set to /etc/fftw/wisdom.sah I wonder. Anyway lets assume its reading it when its greater than 512 bytes as the slot directories will be getting cleaned out at the end of each work unit. That leaves the other code suggestions from Claggy to try. My Pi2 and Parallella's seem to be taking around 27 hours. The Pi B+ is looking like 60 hours (only 57% done after 36 hours). |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
Added my B+ for app testing. Its running a "medium" overclock (900Mhz). Its Host 78198 First wu came in at 228,006 seconds. Only 63.3 hours. |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
A couple of v8.02 tasks have shown up on one of my Pi2's. Any optimisations in this one? Should I abort the remaining v8.01 tasks? |
Send message Joined: 29 May 06 Posts: 1037 Credit: 8,440,339 RAC: 0 |
A couple of v8.02 tasks have shown up on one of my Pi2's. Any optimisations in this one? Should I abort the remaining v8.01 tasks? I got my Pi 2 to ask for work, it got sent a v8.01 (armv6l) task inspite of it not being listed on the applications page any longer, Bug? Or does something need reloading? Edit: A subsequent request got v8.02 work. Claggy |
Send message Joined: 22 Jul 15 Posts: 21 Credit: 113,162 RAC: 0 |
My computers are transitioning over to the 8.02 apps as they finish their 8.01 work. I built 8.02 from the current source code (3378) with the fix above. It is not specific to armv6 or armv7 processors and I tested it on both. It will run on either and use neon if it is available. The app_config.xml files will need to be edited to provide verbose output from the 8.02 app. Claggy can you help us with an updated app_config.xml file for verbose output? - Tom |
Send message Joined: 29 May 06 Posts: 1037 Credit: 8,440,339 RAC: 0 |
Claggy can you help us with an updated app_config.xml file for verbose output? <app_config> <app_version> <app_name>setiathome_v8</app_name> <cmdline>-verbose</cmdline> </app_version> </app_config> Just take out the planclass line, and it'll apply to any setiathome_v8 app Claggy |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
v8.02 is looking to be about the same run time as v8.01 for the few I've managed to complete so far - maybe slightly faster. v8.01 average 96-97k seconds v8.02 average 95k seconds (so far) |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
The Raspberry Pi foundation have just announced the availability of the Pi3 model B. It can run Pi/Pi2 apps but sports an 1.2Ghz ARM 64 bit CPU.(cortex A53) with built-in wifi and Bluetooth. No mention in the blog post about memory size. The are still running the current Raspbian image which is 32 bit but mention investigating 64 bit mode later. Oh and they are the same price as the Pi2 $35 US. |
Send message Joined: 29 May 06 Posts: 1037 Credit: 8,440,339 RAC: 0 |
Ordered one five minutes ago, spent 45 minutes trying to get logged on to farnells site to do it. Claggy |
Send message Joined: 4 Aug 08 Posts: 11 Credit: 1,437,079 RAC: 0 |
buying it too |
Send message Joined: 3 Jan 07 Posts: 1451 Credit: 3,272,268 RAC: 0 |
No mention in the blog post about memory size. RS Components in the UK had ~20,000 of these available for sale yesterday. Today, it's below 8,000 - selling like hot, er, pies? Raspberry Pi 3 Model B data sheet Memory 1GB LPDDR2 |
Send message Joined: 18 Oct 09 Posts: 48 Credit: 73,283 RAC: 0 |
There was some comment on the blog about the video core IV only being able to address 1Gb so that's why they still only have 1Gb of memory. I just bought my fourth Pi2 a fortnight ago. I also ordered the offical power supply and offical cases. Cases got back-ordered until April. One power supply dead on arrival, got a credit for it. Time to replace the Pi2's I guess. Might need to buy some more heatsinks though - I wonder if the SoC is the same size? |
Send message Joined: 29 May 06 Posts: 1037 Credit: 8,440,339 RAC: 0 |
My Pi 3 is up and running here and at Einstein: http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=78373 https://einstein.phys.uwm.edu/show_host_detail.php?hostid=12213079 vcgencmd measure_temp initially gave 74 C when running four Einstein, with a Heatsink freshly fitted, giving 78.8 C now when running three Seti and one Einstein a little later. vcgencmd measure_temp gave the frequency as frequency(45)=1170122000 initially, now gives consistently frequency(45)=1200000000 Need a case and a fan i think. Claggy |
©2025 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.