Posts by Eddie Cottongim

1) Questions and Answers : Web site : Dead links from Obtaining Source Code (Message 1849362)
Posted 17 Feb 2017 by Profile Eddie Cottongim Project Donor
Post:
This page has mostly dead or obsolete links:
http://setiathome.berkeley.edu/sah_source_code.php

This page is somewhat better - has a link to SVN which I think it the right place, but there's still a nightly tarball link which hasn't been updated in four years.
https://setiathome.berkeley.edu/sah_porting.php
2) Questions and Answers : Unix/Linux : export.h error building/running on Debian SPARC (Message 1830432)
Posted 13 Nov 2016 by Profile Eddie Cottongim Project Donor
Post:
I'm about convinced that the export.h error is nothing to worry about. (Really only need to "make" from the client directory.)

As for the runtime error, I learned how to run the reference WU, and that crashes too, so will work from there.
3) Questions and Answers : Unix/Linux : export.h error building/running on Debian SPARC (Message 1830399)
Posted 13 Nov 2016 by Profile Eddie Cottongim Project Donor
Post:
Also, I've seen this from a failed WU run:
cat stderr1.txt
18:12:20 (20848): Can't open init data file - running in standalone mode
18:12:20 (20848): Can't open init data file - running in standalone mode
SETI@home error -5 Can't open file
(work_unit.sah) in read_wu_state() errno=2

File: worker.cpp
Line: 135

The mention of "read_wu" made me think the earlier build error with readwu.cpp might be important.
4) Questions and Answers : Unix/Linux : export.h error building/running on Debian SPARC (Message 1830366)
Posted 13 Nov 2016 by Profile Eddie Cottongim Project Donor
Post:
I've built BOINC with no problems, but I hit an error on seti@home:

"readwu.cpp:47:20: error: export.h: No such file or directory"

followed by a bunch of symbols not found. A client is created, but it fails with an error:
"2016-11-12 20:20:50 [SETI@home] Unrecoverable error for result 02fe16aa.4639.5801.15.42.80.vlar_2 (process exited with code 193 (0xc1))"
Figured I might as well get the build working cleanly, since I don't have much insight into what that error means.

I see that others have hit this error on old forum posts, but don't see any resolutions.

I installed the packages for kernel headers, libtao, and libidl-dev to try to fix this, but no luck.

Is export.h is part of, or generated by IDL? That's why I was trying those IDL packages, even though I've never seen them in a list of dependencies for BOINC or SETI@Home.

My source tree is a few months old (I've been trying this that long) - The client reports 8.0.0.

Thanks for any ideas!
5) Questions and Answers : Unix/Linux : Rasperry Pi WUs getting stuck (Message 1786909)
Posted 12 May 2016 by Profile Eddie Cottongim Project Donor
Post:
I have a Pi B+ v1.2. I'm running seti@home, and the workunits seem to get stuck after a few hours. The "setiathome_8.02" tasks no longer appears. If I do a suspend/resume via manager, then it starts again and runs for a few more hours. When it has died, I find the attached stack trace. It was doing this on a workunit, and I got tired of suspend/resume, so I aborted and got another, which is also doing the same thing.

The system ran the s@h Beta without this happening. I also have a Pi 2 running without problems.

Other info: BOINC 7.0.27

May 11 06:02:48 raspberrypi kernel: [1056977.544250] Modules linked in: xt_multiport iptable_filter ip_tables x_tables snd_bcm2835 snd_pcm snd_seq snd_seq_device snd_timer snd bcm2835_gpiomem bcm2835_rng uio_pdrv_genirq uio
May 11 06:02:48 raspberrypi kernel: [1056977.560776] CPU: 0 PID: 7762 Comm: setiathome_8.02 Tainted: G      D         4.1.19+ #858
May 11 06:02:48 raspberrypi kernel: [1056977.569117] Hardware name: BCM2708
May 11 06:02:48 raspberrypi kernel: [1056977.572696] task: cd838da0 ti: daf18000 task.ti: daf18000
May 11 06:02:48 raspberrypi kernel: [1056977.578279] PC is at vfp_save_state+0x0/0x28
May 11 06:02:48 raspberrypi kernel: [1056977.582724] LR is at vfp_sync_hwstate+0x3c/0x48
May 11 06:02:48 raspberrypi kernel: [1056977.587428] pc : [<c000a948>]    lr : [<c000a58c>]    psr: 60000113
May 11 06:02:48 raspberrypi kernel: [1056977.587428] sp : daf19e78  ip : daf19e90  fp : daf19e8c
May 11 06:02:48 raspberrypi kernel: [1056977.599240] r10: 002efe08  r9 : daf18000  r8 : be863368
May 11 06:02:48 raspberrypi kernel: [1056977.604635] r7 : be863458  r6 : daf180f8  r5 : be863560  r4 : c0000788
May 11 06:02:48 raspberrypi kernel: [1056977.611329] r3 : daf180f8  r2 : c0842364  r1 : c0000788  r0 : daf180f8
May 11 06:02:48 raspberrypi kernel: [1056977.618025] Flags: nZCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
May 11 06:02:48 raspberrypi kernel: [1056977.625327] Control: 00c5387d  Table: 1adc8008  DAC: 00000015
May 11 06:02:48 raspberrypi kernel: [1056977.751325] [<c000a948>] (vfp_save_state) from [<c000a6f0>] (vfp_preserve_user_clear_hwstate+0x34/0xc0)
May 11 06:02:48 raspberrypi kernel: [1056977.760912] [<c000a6f0>] (vfp_preserve_user_clear_hwstate) from [<c0012d20>] (setup_sigframe+0x180/0x190)
May 11 06:02:48 raspberrypi kernel: [1056977.770662] [<c0012d20>] (setup_sigframe) from [<c0013150>] (do_signal+0x2d0/0x3e0)
May 11 06:02:48 raspberrypi kernel: [1056977.778501] [<c0013150>] (do_signal) from [<c0013458>] (do_work_pending+0xc8/0xd8)
May 11 06:02:48 raspberrypi kernel: [1056977.786253] [<c0013458>] (do_work_pending) from [<c000f7e4>] (work_pending+0xc/0x20)
May 11 06:02:48 raspberrypi kernel: [1056977.800442] ---[ end trace 6f0fb1a122aca842 ]---
6) Message boards : Number crunching : Raspberry Pi results (Message 1784669)
Posted 4 May 2016 by Profile Eddie Cottongim Project Donor
Post:
My Pi B is also getting stuck like this. If I suspend, then resume the task, it runs for a few hours, then sticks again.

Nothing interesting in the BOINC logs, but if I leave a shell going, I can see this:

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.064162] Internal error: Oops - undefined instruction: 0 [#6] ARM

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.156469] Process setiathome_8.02 (pid: 4853, stack limit = 0xdc12e188)

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.163338] Stack: (0xdc12fe78 to 0xdc130000)

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.167783] fe60:                                                       dc12fe90 dc12e000

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.176048] fe80: dc12feac dc12fe90 c000a6f0 c000a55c 00000000 be820368 00000000 daf7e3c8

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.184316] fea0: dc12fec4 dc12feb0 c0012d20 c000a6c8 dc12ffb0 dc12fec8 dc12ff8c dc12fec8

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.192581] fec0: c0013150 c0012bac 000d5378 14000000 0021b940 00000000 00000000 0000000e

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.200845] fee0: 00000000 00000080 00000000 00000000 00000000 00000000 00000003 dc12ff08

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.209110] ff00: c005b474 c005d5e0 00000078 0000000e 00000011 c07d6e1c 00000001 c07d6e1c

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.217374] ff20: c081b9b8 00000000 00000000 c005dc4c dc12ff4c dc12ff40 c005dc4c c005a9b8

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.225638] ff40: dc12ff5c dc12ff50 c001ff64 c002741c dc12ff74 dc12ff60 c002741c 0000000e

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.233903] ff60: c005df80 00000001 dc12e000 00000000 dc12ffb0 00000000 dc12e000 002efe08

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.242168] ff80: dc12ffac dc12ff90 c0013458 c0012e8c 0002b410 80000010 f200b200 00c5387d

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.250433] ffa0: 00000000 dc12ffb0 c000f7e4 c001339c 00f32e20 be8206cc ffffffdc 00f32f00

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.258701] ffc0: 00f32e00 0000002c 00f75590 00000040 002f1f60 002eff31 002efe08 002e8958

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.266966] ffe0: 00f75624 be820658 0002b3b8 0002b410 80000010 ffffffff 61657274 3072656d

Message from syslogd@raspberrypi at May  3 23:48:34 ...
 kernel:[429716.317656] Code: e12fff1e e1a0200d e1a0e009 eafffebe (eca00b20)


Didn't have this problem on Beta.
7) Message boards : News : SETI@home 8 released for Raspberry Pi. (Message 1780939)
Posted 21 Apr 2016 by Profile Eddie Cottongim Project Donor
Post:
Thanks very much for the Pi support - makes it a breeze to get these mini-boxes crunching.

For the person above who was having WU errors, I found s@h to be sensitive to overclocking. Mild overclocks that caused no other problems caused computation errors on my Pi2. With no overclock, I don't have the errors.
8) Message boards : Number crunching : Building seti app for the Raspberry Pi (Message 1763009)
Posted 7 Feb 2016 by Profile Eddie Cottongim Project Donor
Post:
Joined up with a Pi2/Raspian. Hope that helps! Would be great if support was semi-official so it's easier to keep up to date. I know you can only chase so many platforms, though. Live long and crunch!
9) Message boards : Number crunching : Memory Access Efficient Pulse Folding Algorithms (Message 782067)
Posted 13 Jul 2008 by Profile Eddie Cottongim Project Donor
Post:
I was wondering if anyone has come across pulse folding algorithms that minimize the required number of passes through memory. I haven't had much luck coming up with one myself. There are few papers that could be related but no definite hits yet.

Such an algorithm for the FFT exists, known as the "Four Step" FFT. It basically treats a big 1D FFT as a 2D FFT and deals with rows and columns at a time, achieving as low as two passes through memory at the cost of some more operations.

This has applications for FPGA implementation. FPGAs often don't have enough on board memory for an entire folding set. (Or if they do, it isn't just one big chunk, which is good and bad, but not simple in any case) They can have dozens of adders, but just one adder is enough to saturate a memory bus in a straightforward implementation like SETI@Home's. However, once the data set is small enough, they can work very rapidly.

Thanks!
10) Message boards : Number crunching : Folding Algorithm Characteristics (Message 772794)
Posted 24 Jun 2008 by Profile Eddie Cottongim Project Donor
Post:
Thanks for your reply. That helps a lot.

One interesting tidbit about the folding is the credit calculation:

analysis_state.FLOP_counter+=(PulsePotLen*0.1818181818182+400.0)*PulsePotLen;

I don't understand quite how the period seaches operate but I assume this line is a fair approximation of how many operations are done.

At 40,960 that comes out to around 320 megaflops per chunk, quite intensive!
11) Message boards : Number crunching : Folding Algorithm Characteristics (Message 770166)
Posted 19 Jun 2008 by Profile Eddie Cottongim Project Donor
Post:
Hi all,

I've been studying seti@home's pulse folding algorithms. I have a few questions (with a specific application in mind, but I think others may learn from this discussion as well).

First, I'll state what I believe to have happened up to this point so you can see if I'm way off:

Baseline Smoothing (once per WU), produces an 8 meg array consisting of ~1 million complex sampled points.

Dechirping (varies but thousands of times per WU), reads the previous array, outputs another 8 meg array of ~1 million complex sampled points

FFT: For each chirp rate, analyze at a variety of FFT sizes/bandwidths. FFT sizes from n=8 to 128k. FFT sizes actually used vary (8 to max at chirp rate 0, with less of the large sizes as chirprate goes to extremes). For each FFT size n, read the dechirped array, perform the FFT across the array (no overlap). Take the complex magnitude of the time values, giving us real data, and do a spike search.

Each of these passes at FFT size m will be stored as a "Power over time" (PoT) array. 4 meg, 1 million real values. This should be interpreted as a 2d array, with height being the number of frequency divisions (FFT size) and the width being the number of time divisions (number of samples(1048576)/FFT size).

Gaussian fitting is done across the table at a constant frequency. (ie pick a row and walk across the columns in this table)

Again, loop through frequencies.
Loop through time in chunks to perform the triplet search and pulse folding. The size of the chunks is based on the slew rate of the telescope and/or some fields out of the WU header. It could be the entire PoT time series at zero slew rate(as long as 131072 entries for FFT length=8 which is probably the most common case), or it could be much smaller.

Ok, here's where I could use some help:
I'd like to understand what is really typical for chunk size (called AdvanceBy in the code) across the population of WU's and within WUs.

I am interested in getting the folding to work fast in a situation where there is very limited fast local memory - lets say 1024 entries (4 kbytes). Thats probably enough for the WU I'm looking at with pulse_pot_length=256, but I have no idea what the population out there looks like. If it often goes above 1024, I'd have a problem as folding would quickly become dependent on a slower memory bus. I don't think there is much way to modify the folding to hit memory less. It would pretty much require a complete pass for every pass of folding until the folds were small enough to fit in cache.

A minor additional question: What purpose does GetFixedPoT serve? Why does the PoT time series need to be a fixed length? To simplify the Gaussian fit?


I hope this makes sense. Thanks for your help.
12) Questions and Answers : Web site : Certificates busted (Message 120999)
Posted 9 Jun 2005 by Profile Eddie Cottongim Project Donor
Post:
Both the photo and modern certifates produce errors like this: (diploma works)

Warning: imagecreatefrompng(cert_images/whiteBackground.png): failed to open stream: No such file or directory in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 40

Warning: imageantialias(): supplied argument is not a valid Image resource in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 41

Warning: getimagesize(cert_images/whiteBackground.png): failed to open stream: No such file or directory in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 42

Warning: imagecolorallocate(): supplied argument is not a valid Image resource in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 45

Warning: imagecolorallocate(): supplied argument is not a valid Image resource in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 46

Warning: imagecolorallocate(): supplied argument is not a valid Image resource in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 47

Warning: imagecolorallocate(): supplied argument is not a valid Image resource in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 48

Warning: imagecolorallocate(): supplied argument is not a valid Image resource in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 49

Warning: imagecolorallocate(): supplied argument is not a valid Image resource in /disks/koloth/a/inet_services/boinc_www/share/projects/sah/html/project/cert3.php on line 50

(much, much more of this)
13) Questions and Answers : Web site : Date on Certificate still off (Message 91648)
Posted 27 Mar 2005 by Profile Eddie Cottongim Project Donor
Post:
Mine has also been bouncing around. For ages it was Sept 12, 2001, then recently it went to some time in 1999, then in the last few days its gone to Sept 15, 2001. Weird.
14) Questions and Answers : Wish list : Merge hosts (Message 86902)
Posted 17 Mar 2005 by Profile Eddie Cottongim Project Donor
Post:
I suppose one reason might be that you could merge a bunch of hosts to create a "superhost." People that care about host rankings (not me) might get annoyed.
15) Questions and Answers : Preferences : Merge host (Message 62999)
Posted 11 Jan 2005 by Profile Eddie Cottongim Project Donor
Post:
Weirdly, my machine seems to have changed CPU speed by 3 mhz, and thats enough to keep it from merging.
16) Questions and Answers : Windows : Bonic will not load - Windows 95b (Message 47976)
Posted 20 Nov 2004 by Profile Eddie Cottongim Project Donor
Post:
You probably need Winsock2, if you haven't already got it.
17) Message boards : Number crunching : My 200mhz PC received credit! (Message 10587)
Posted 21 Jul 2004 by Profile Eddie Cottongim Project Donor
Post:
Impressive. I've been trying to figure out how to get my 386sx-16 to do SETI@HOME - hard when the max ram is 4mb. (Not counting the memory on my gigantic Intel Aboveboards, 2mb each using on 72 120ns chips). Most schemes so far involve a soldering iron :/

> > > I have an old 386 SX-2 16mhz machine with Seti classic installed on
> it.
> > > Without the math co-processor (sold separately) it takes about 3
> years
> > per
> > > workunit. Adding a co-processor drops the time to about 1 year.
> :)
> >
> > Did you actually do/complete units on that computer? :o
> >
> >
>
> Do, yes. Complete, no. :)
> ----------------------------------------------------------------
> We are Borgholio. You will be assimilated...bunghole!
>





 
©2018 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.