Message boards :
Number crunching :
Linux CUDA 'Special' App finally available, featuring Low CPU use
Message board moderation
Previous · 1 . . . 30 · 31 · 32 · 33 · 34 · 35 · 36 . . . 83 · Next
Author | Message |
---|---|
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
The zi3v scans the Wu and if it finds any suspects it runs the part pf the wu again with unroll 1. That idea came from jason_gee. I tried and coded it and That is what I'm running now. It may be more accurate and a bit slower. Just keep testing.It seems to be better on the Pulses, but now I'm seeing a few Bad Best gaussian results. Can you also have it check the Best gaussians? https://setiathome.berkeley.edu/workunit.php?wuid=2573994560 Best gaussian: peak=3.208247, mean=0.5376316, ChiSq=1.405401, time=37.75, d_freq=1419524825.06, score=-1.281584, null_hyp=2.177316, chirp=-74.201, fft_len=16k Best gaussian: peak=3.839457, mean=0.570847, ChiSq=1.367193, time=88.08, d_freq=1419521976.84, score=-1.280418, null_hyp=2.15331, chirp=91.898, fft_len=16k I'm waiting on these; https://setiathome.berkeley.edu/workunit.php?wuid=2572586193 https://setiathome.berkeley.edu/workunit.php?wuid=2548202405 https://setiathome.berkeley.edu/workunit.php?wuid=2537036980 The new build is being run at Beta here, https://setiweb.ssl.berkeley.edu/beta/results.php?hostid=76256 |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
After 300 tasks at Beta, I see an example of the Bad Best gaussian; https://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=9814735 setiathome v8 enhanced x41p_zi3v Best gaussian: peak=4.179945, mean=0.5265673, ChiSq=1.137641, time=89.76, d_freq=1418819127.86, score=1.12354, null_hyp=2.163572, chirp=48.412, fft_len=16k SSE2x Linux64 Build 3584 Best gaussian: peak=4.148942, mean=0.5508794, ChiSq=1.387269, time=26, d_freq=1418820668.26, score=1.099337, null_hyp=2.292666, chirp=75.539, fft_len=16k Otherwise the only 3 Inconclusives out of 300 tasks were against two Intel iGPUs and a Laptop running CUDA 32. Not bad. |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
That's great to hear. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13732 Credit: 208,696,464 RAC: 304 |
Otherwise the only 3 Inconclusives out of 300 tasks were against two Intel iGPUs and a Laptop running CUDA 32. Not bad. 1% i'd consider Excellent. Grant Darwin NT |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Yep, looking at a few tasks it looks as though the idea worked. +1 for 3am lightbulb moments. I guess how costly it is will come out over time. Will see if I can spot a Gaussian issue. Usually those issues come from minor precision issues in the normalisation sums (easily checked and remedied). @Petri33 please email latest and I'll update the alpha folder ASAP. Earlier than was planned, due to Petri's pushing the technology so far, it opens multiple doors for 'next level' pre-scans. Namely Wavelet &/or Convolutional Neural Network based feature recognition, followed up with sparser traditional Fourier processing so as to get the same numbers. [Sidenote:] At present, still wrestling with home & work issues myself, though We get semi decent internet here in a month or so. Sick of working for the benefit of other bosses, I'll probably end up streaming during development at some point, and work in the Open Source stuff somehow. Probably priority would be on survival and having a laugh, we'll see. "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. |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
Yep, looking at a few tasks it looks as though the idea worked. +1 for 3am lightbulb moments. I guess how costly it is will come out over time. Will see if I can spot a Gaussian issue. Usually those issues come from minor precision issues in the normalisation sums (easily checked and remedied). @Petri33 please email latest and I'll update the alpha folder ASAP. . . Does any of this portend a possible Windows version anytime in the forseeable future ?? Stephen ?? |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Yep, looking at a few tasks it looks as though the idea worked. +1 for 3am lightbulb moments. I guess how costly it is will come out over time. Will see if I can spot a Gaussian issue. Usually those issues come from minor precision issues in the normalisation sums (easily checked and remedied). @Petri33 please email latest and I'll update the alpha folder ASAP. Sure. I've built Windows ones on multiple occasions, which is how I stumbled on the unroll validation problem and possible fix (lots of help from TBar and Jeff Buck in spotting the patterns solved that). I only sat on them because the scope of damage saturating results with 1000 Windows hosts would potentially compromise the project too far. Probably there will still be niggles to address, and the high-end device requirements will need tailoring to solve, Though advanced user 'Use at you own risk' Windows builds become viable once again. "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. |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . Hmm, I would be up for that. Stephen :) |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
After 300 tasks at Beta, I see an example of the Bad Best gaussian; https://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=9814735Hmmm, maybe it's the 'other' Apps that's having gaussian problems. Besides the above task, I have two others that were decided in favor of CUDA. The Above with SETI@home v8 v8.05: Best gaussian: peak=4.179937, mean=0.5265682, ChiSq=1.13764, time=89.76, d_freq=1418819127.86, score=1.123425, null_hyp=2.163565, chirp=48.412, fft_len=16k Then these; https://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=9816865 & https://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=9815080 Anyway, Beta started sending out Arecibo VLARs, so, I stopped that. Back to Main for now. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
It's possible. We need to always reference to Win32 stock CPU, just because Eric Made it. It's not that 1-5% is problematic, just that every goal needs a reference. "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. |
petri33 Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 |
@Petri33 please email latest and I'll update the alpha folder ASAP. Emailed. To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
. . Hmm, I would be up for that. If I understood the conversation right, I would be willing to alpha/beta test any windows-based Seti program too. Tom A proud member of the OFA (Old Farts Association). |
Jim1348 Send message Joined: 13 Dec 01 Posts: 212 Credit: 520,150 RAC: 0 |
I am still at a loss to get BOINC installed. This works for me on Ubuntu 16 (and 17). I don't know why they make it such a hassle, or why no one has a standard set of installation instructions. I had to gather them up in bits and pieces. Note that it includes installing the files (“gui_rpc_auth.cfg†and “remote_hosts.cfgâ€) that are used for monitoring BOINC over a LAN using BOINCTasks on a Windows machine, so you may not need them. Note that "USER" is of course your user name.
|
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
Thanks Jim for the insight on how to change permissions for USER access to the 'simple' install of BOINC. Notes have been added to my tickle trunk :D Brent EDIT: Jim, did you get BOINC v7.6.33 from the Ubuntu Software in v16/17? Ubuntu 14.04 distributes BOINC v7.2.42 Just curious ... |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
Thanks Jim for the insight on how to change permissions for USER access to the 'simple' install of BOINC. . . If you are running the later versions of Ubuntu you can run the repository version of BOINC and you get 7.6.33, but then you have your BOINC files scattered over several folders/directories. If you want BOINC to be installed under your 'Home' directory for ease of access and backing up then you need the Seti version of BOINC and you get 7.2.42 but are restricted to the earlier version of Ubuntu 14.04. Horses for courses. Stephen .. |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
I am still at a loss to get BOINC installed. . . Those pointers could solve several issues I am having. Even though I am running 14.04 and BOINC 7.2.42, like Brent they are going to be saved here. Stephen :) |
Jim1348 Send message Joined: 13 Dec 01 Posts: 212 Credit: 520,150 RAC: 0 |
EDIT: Ubuntu 16.04 gives you BOINC 7.6.31, whereas Ubuntu 16.10 (and 17) gives you BOINC 7.6.33. There is really no difference between them for Linux. I have had no problems with them. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
It's possible. We need to always reference to Win32 stock CPU, just because Eric Made it. It's not that 1-5% is problematic, just that every goal needs a reference.The Plot thickens... Ran a few hundred more tasks and ran across this one, https://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=9821403 setiathome v8 enhanced x41p_zi3v Best gaussian: peak=4.149051, mean=0.5146711, ChiSq=1.19371, time=79.69, d_freq=1420106366.11, score=2.326251, null_hyp=2.25355, chirp=-71.774, fft_len=16k SSE3xj Win32 Build 3584 Best gaussian: peak=3.718165, mean=0.5199579, ChiSq=1.359161, time=81.37, d_freq=1420106245.69, score=2.270708, null_hyp=2.337314, chirp=-71.774, fft_len=16k I ran it with My CPU App, SSE4.1xjf OS X 64bit Build 3344 Best gaussian: peak=4.14904, mean=0.5146721, ChiSq=1.193707, time=79.69, d_freq=1420106366.11, score=2.326065, null_hyp=2.253538, chirp=-71.774, fft_len=16k You can run it yourself if you wish, you might want to hurry though; http://boinc2.ssl.berkeley.edu/beta/download/62/04oc08ab.31484.890.13.47.11 The third Host is also running SETI@home v8 v8.22 (opencl_nvidia_SoG) windows_intelx86, so, I suspect My results will be overruled. That adds the Windows version to the Linux version of SoG that appear to be giving the wrong Best Gaussian results on occasion. I believe I remember a discussion about this very problem at Beta some time ago, I think it was agreed the Best Gaussian didn't necessarily have to be listed with the signals to be correct. It's probably still there if someone wants to track it down. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Saved for analysis when I can. Will probably see what win32 CPU stock and ye-olde Cuda turn up. There are some differences between the way stock cpu and 3rd party cpu sum, with the former having been refined to greater portability & less cumulative error. Wouldn't have thought that alone could explain that great of a difference in numbers though, so other explanations could be in the compilation used etc. The lack of 80-bit FPU availability in the later apps can be problematic there if not handled carefully. That's just some possibilities assuming no lingering bugs in either case though, which could easily have been buried in the noise of prior other issues amongst different applications. Overall I'd suspect the SSE3xj Win32 gaussian values are 'more correct', as since the powerspectrum values are positive any increase in cumulative error tends to push peaks higher rather than lower. [Edit: hmmm, a different time value too] "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. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Overall I'd suspect the SSE3xj Win32 gaussian values are 'more correct', as since the powerspectrum values are positive any increase in cumulative error tends to push peaks higher rather than lower. [Edit: hmmm, a different time value too]It's basically a replay of yesterday's task where the Windows CPU App agreed with CUDA zi3v over Build 3584 SoG; After some thought I remember looking at this some weeks ago. The CUDA Special App matched the CPU on two different Platforms for Best Gaussian. Now that a Widows CPU also matches, that makes it Three platforms.After 300 tasks at Beta, I see an example of the Bad Best gaussian; https://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=9814735Hmmm, maybe it's the 'other' Apps that's having gaussian problems. Besides the above task, I have two others that were decided in favor of CUDA. It shouldn't be that difficult, just look at some Hosts at Beta running CPU only, here's one; https://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=9810628 I suspect there will be others. |
©2024 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.