Message boards :
News :
Distributing 4-bit workunits
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 . . . 6 · Next
Author | Message |
---|---|
![]() Send message Joined: 15 Mar 05 Posts: 1547 Credit: 26,981,856 RAC: 717 ![]() ![]() |
It's for both Arecibo and GBT. ![]() |
Send message Joined: 30 Dec 13 Posts: 258 Credit: 12,340,341 RAC: 4 ![]() |
I'm guessing that since these blc3_2bit_guppi have file sizes of 704KB that they are the 4-bit work units? |
![]() Send message Joined: 15 Mar 05 Posts: 1547 Credit: 26,981,856 RAC: 717 ![]() ![]() |
That's correct. ![]() |
Send message Joined: 30 Dec 13 Posts: 258 Credit: 12,340,341 RAC: 4 ![]() |
I'm seeing a lot of -9 overflows with work units starting with 24mr10 Anyone else seeing that? Or is it just my machine? |
Send message Joined: 15 Jun 16 Posts: 45 Credit: 1,836,741 RAC: 0 ![]() |
I'm seeing a lot of -9 overflows with work units starting with 24mr10 I only had one of them. Suspended current work to run it, and it lasted 10 seconds. Looking dodgy. Grant Darwin NT. |
![]() Send message Joined: 11 Nov 12 Posts: 853 Credit: 2,999,363 RAC: 267 ![]() |
Well I have set one of currently unused machines to crunch here. As they say every little helps!! |
Send message Joined: 15 Jun 16 Posts: 45 Credit: 1,836,741 RAC: 0 ![]() |
I'm seeing a lot of -9 overflows with work units starting with 24mr10 Just got another one, it's crunching normally. EDIT- That and another crunched OK, another 2 only lasted 10 and 11 seconds. Grant Darwin NT. |
Send message Joined: 15 Jun 16 Posts: 45 Credit: 1,836,741 RAC: 0 ![]() |
Well I have set one of currently unused machines to crunch here. Only another 3,300 WUs to finish off. :-) Grant Darwin NT. |
![]() ![]() Send message Joined: 18 Aug 05 Posts: 2423 Credit: 15,878,738 RAC: 0 ![]() |
Got 4-bits Arecibo tasks: <receiver_cfg> <s4_id>11</s4_id> <name>Arecibo 1.4GHz Array, Beam 4, Pol 0</name> <beam_width>0.0500000007</beam_width> <center_freq>1420</center_freq> <latitude>18.3538056</latitude> <longitude>-66.7552222</longitude> <elevation>497</elevation> <diameter>168</diameter> <az_orientation>180</az_orientation> <az_corr_coeff length=99 encoding="x-csv"> -37,-6.05,92.35,-731.21,-1013.97,-24.53,-11.19,9.18,106.04,3.02,-1.74, -3.46,1.29 </az_corr_coeff> <zen_corr_coeff length=99 encoding="x-csv"> -57.55,-95.56,-4.13,141.69,677.51,-10.41,-7.71,-10.39,0.08,0.43,-0.62, 0.03,-0.36 </zen_corr_coeff> <array_az_ellipse>329.06</array_az_ellipse> <array_za_ellipse>384.005</array_za_ellipse> <array_angle>-60</array_angle> <min_vgc>512</min_vgc> <polarization>linear</polarization> </receiver_cfg> <recorder_cfg> <name>seti_dr2_ALFA</name> <bits_per_sample>2</bits_per_sample> <sample_rate>2500000</sample_rate> <beams>14</beams> <version>1.99000001</version> </recorder_cfg> <splitter_cfg> <version>0.300000012</version> <data_type>encoded</data_type> <fft_len>256</fft_len> <ifft_len>1</ifft_len> <filter>polyphase</filter> <window>hanning</window> <samples_per_wu>1048576</samples_per_wu> <highpass>0</highpass> <blanker_filter>randomize</blanker_filter> <pfb_ntaps>16</pfb_ntaps> <pfb_width_factor>1.04999995</pfb_width_factor> <wu_bits_per_sample>4</wu_bits_per_sample> </splitter_cfg> News about SETI opt app releases: https://twitter.com/Raistmer |
![]() ![]() Send message Joined: 18 Aug 05 Posts: 2423 Credit: 15,878,738 RAC: 0 ![]() |
As I could recall there was some discussion with Josef Segur to enable compression inside workunit. There are even test compressed tasks available in Lunatics's collection (processed OK by known modern MB apps). Maybe worth to enable such compression now? News about SETI opt app releases: https://twitter.com/Raistmer |
Send message Joined: 15 Jun 16 Posts: 45 Credit: 1,836,741 RAC: 0 ![]() |
Would compression have much effect on file sizes? From memory truly random data doesn't compress much (if at all), so with WUs the only parts that would compress would be the added reference data. Grant Darwin NT. |
Send message Joined: 3 Jan 07 Posts: 1451 Credit: 3,266,661 RAC: 8 ![]() |
You can easily test - take a sample datafile from your cache and run it through gz or zip compression tools (I doubt the project would go as far as LZMA, so not 7-zip for this test). 2-bit MB tasks shrink fron ~357 KB to ~267 KB, or 25%, because of the 64 character ASCII encoding. AP data is sent in a denser binary encoding, which barely compresses at all (except when the B3 P0 bit is stuck...) Now that our download channel seems to have enough bandwidth, would the server processing overhead of compression be worth a 25% saving? |
![]() Send message Joined: 15 Mar 05 Posts: 1547 Credit: 26,981,856 RAC: 717 ![]() ![]() |
We could just send binary data rather than encoded data to get most of the same benefit without loading the splitters or servers. Maybe I'll try that next. ![]() |
Send message Joined: 30 Dec 13 Posts: 258 Credit: 12,340,341 RAC: 4 ![]() |
I checked your "shorts" as well and they have the same results SETI@Home Informational message -9 result_overflow So seems there is a batch shorties in there Edit... SoG nonmodified 13 minutes, cuda 42 - 33 minutes, sah in the bull pen and waiting to start |
Send message Joined: 30 Dec 13 Posts: 258 Credit: 12,340,341 RAC: 4 ![]() |
Picked up 5 errors on the SaH app ERROR: Possible wrong computation state on GPU, host needs reboot or maintenance https://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24065993 https://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24065968 https://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24066051 https://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24065976 https://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24065955 GPUs run fine for cuda SoG and Cuda 42. Not sure why these were erroring out |
![]() ![]() Send message Joined: 18 Aug 05 Posts: 2423 Credit: 15,878,738 RAC: 0 ![]() |
Picked up 5 errors on the SaH app Too big autocorr value happened too often. That triggered sanity check. I discussed this with Eric - decision was leave it as is for now. News about SETI opt app releases: https://twitter.com/Raistmer |
![]() Send message Joined: 10 Mar 12 Posts: 1659 Credit: 12,789,696 RAC: 7,691 ![]() ![]() |
Hmm, interesting. With these new 4-bit WU's, it seems as if opencl_nvidia_sah, is coming back into play again. Up to now, opencl_nvidia_SoG has been chosen as the clearly faster app. I haven't been getting any opencl_nvidia_sah tasks for several days, but now after the shift to 4-bit WU's, I'm getting them again. Now though, it seems as if opencl_nvidia_sah will be competing with SoG with both Arecibo and Guppi tasks. At least that's what I see now initially. With the Guppi tasks, opencl_nvidia_sah seems to be almost 2 minutes faster, but nothing is conclusive yet. Could have been some freak stuff happening. |
Send message Joined: 30 Dec 13 Posts: 258 Credit: 12,340,341 RAC: 4 ![]() |
For me SaH caused some Error while computing. Never saw those with the SoG. I need to test Raistmer newest r3472 so I'll be back tomorrow and see if I can get some more SoG here. Like you, the server seems to think SaH is better but I know looking at the times, SoG was better. I think all those errors while downloading threw a confused the server into thinking SaH was better since only 5 with it vs 50 (download errors) |
![]() Send message Joined: 10 Mar 12 Posts: 1659 Credit: 12,789,696 RAC: 7,691 ![]() ![]() |
For me SaH caused some Error while computing. Never saw those with the SoG. Well, I don't know really. It seems to go up and down here, which one is faster. Same AR can at times be faster with opencl_nvidia_SoG, other times, same AR is faster with opencl_nvidia_sah. That goes for both Arecibo and Guppi tasks. I think I'm getting a tiny bit confused here. Ah well, at 60, I guess that's just to be expected, and it's just going to get worse over time :-) |
![]() Send message Joined: 10 Mar 12 Posts: 1659 Credit: 12,789,696 RAC: 7,691 ![]() ![]() |
Stock CPU app, also seems to work just fine with these 4-bit workunits: http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24070149 Although stock CPU app, is painfully slow, compared to MB8_win_x64_AVX_VS2010_r3330.exe. Stock CPU app, takes more than twice the time, almost 2.5 times slower. Edit: But a lot of that slowness, is due to me now running the iGPU with one task too, something I don't do on main with the opt CPU app. That slows down CPU apps so much that it's almost crazy. |
©2019 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.