Posts by Sebastian M. Bobrecki
|
1)
Message boards :
Number crunching :
The Server Issues / Outages Thread - Panic Mode On! (117)
(Message 2023542)
Posted 18 Dec 2019 by  Sebastian M. Bobrecki
Post: I got some, but still no working downloads for me. |
|
2)
Message boards :
Number crunching :
The Server Issues / Outages Thread - Panic Mode On! (117)
(Message 2016550)
Posted 25 Oct 2019 by Sebastian M. Bobrecki
Post: It's a server issue- there's no way there are almost a million shorties out there waiting to be validated. For as much as we rage against results that take forever to be returned, the vast majority are returned within a few days.I think the behavior match with what I think before. I looked at logs of my clients and yesterday was a lot more Arecibo tasks than usual. Actually more Arecibo than blc. According to haveland it was about 500-600k more "awaiting validation" than usual. With over 120k results returned per hour it's not that much. I also saw yesterday that my validation inconclusive increased from usual below 100 to over 150. And I'm not a mega rac cruncher. Today I see it starts dropping with some increase in invalids. Those aren't shorties, they are noise bombs, because they bomb out early (in a matter of seconds). Shorties are processed all the way through, but don't take nearly a long as an average mid range WU to process.This is more a matter of naming convention. For me naming those few seconds tasks noise bombs makes a negative feelings (as noise is something we don't want). And actually they could be what we are all looking for. But as I wrote this is just a matter of naming convention. |
|
3)
Message boards :
Number crunching :
The Server Issues / Outages Thread - Panic Mode On! (117)
(Message 2016492)
Posted 24 Oct 2019 by Sebastian M. Bobrecki
Post: If I remember correctly "Results returned and awaiting validation" are results awaiting for quorum. So it is not a server problem. I'm guessing that this could be due to lots of Arecibo shorties, which ends with "-9 result_overflow" and don't match between opt. and normal app., so they requires another wingman. |
|
4)
Message boards :
Number crunching :
Data Chat
(Message 2005713)
Posted 4 Aug 2019 by Sebastian M. Bobrecki
Post: ...If this is the case, than this means that this is not a Julian date. Following wikipedia: ... Ordinarily calculating the fractional portion of the JD is straightforward; the number of seconds that have elapsed in the day divided by the number of seconds in a day, 86,400. ... |
|
5)
Message boards :
Number crunching :
Data Chat
(Message 2005477)
Posted 3 Aug 2019 by Sebastian M. Bobrecki
Post: ...One correction. The Part 4 is not a offset. It's a fraction of day. It is calculated by dividing given number of seconds pas midnight by number of seconds in the day (witch is normally 86400, but can be 86399 or 86401 if leap seconds are required ). And back to get the number of seconds you have to multiply this fraction by number of seconds in day. For example in: blc11_2bit_guppi_58692_01635_HIP80413_0117 there is 01635 as fractional part. So to get number of seconds past midnight just do: 86400s x 0.01635 = 1412.64s = 23m 32.64s Thanks to that there is sub-second accuracy. |
|
6)
Message boards :
SETI Perspectives :
Seeing the Unseeable: The Black Hole Image
(Message 2004202)
Posted 25 Jul 2019 by Sebastian M. Bobrecki
Post: The stars existed for many years before the event which created the black hole. They were not approaching each other then...Actually they are approaching each other. If you look closely, on the poster that I linked earlier, for BH scenario at relative time 0 distance between stars was about 5500 solar radius. After about 5.5 million years distance shrunk to about 4000 solar radius and then the Common Envelope phase begins. It stands for another about 700 thousands years during witch distance is reduced to about 40 solar radius. Then there is SuperNova explosion during witch some of the matter is thrown into space. From the remaining matter, a BH is formed. As some of the matter was thrown away and as BH is very dense and have much smaller radius than the star from witch it was created, relative distance rises to about 47 solar radius. After next about 14 million years distance shrunk again to about 46 solar radius and BH begins to draw matter from a companion star, and create accretion disk. |
|
7)
Message boards :
Number crunching :
Don't know where it should go? Stick it here!
(Message 2002845)
Posted 16 Jul 2019 by Sebastian M. Bobrecki
Post: I thought we were doing great, but according to Eric's speech, it looks like we're only analyzing 1% of GBT data. He starts around 6:30, and talks about this around 15:45 - https://www.youtube.com/watch?v=dYmsQB_Gjrg |
|
8)
Message boards :
SETI Perspectives :
Seeing the Unseeable: The Black Hole Image
(Message 2001301)
Posted 6 Jul 2019 by Sebastian M. Bobrecki
Post: I'm a little curious about the, quite frequently heard, issue, that when a member of a binary system collapses to a black hole, it strips material off the companion star. I would expect the black hole to have pretty much the same mass as the original star, perhaps slightly less, which was not stripping material from its neighbour when it was a regular star. Sure, the mass is now concentrated into a much smaller object, and the gravitational field close to the hole is stronger than it was, but it is much smaller., and hence, further away from the companion. How is this issue justified?It doesn't happen immediately after collapse. It starts after some time as their distance decreases accordingly. For examples of evolutions look at this poster. |
|
9)
Message boards :
Number crunching :
Panic Mode On (114) Server Problems?
(Message 1975314)
Posted 15 Jan 2019 by Sebastian M. Bobrecki
Post: I can take FRB being Fast Radio Burst, but anyone got an idea what the GJ is in blc34_2bit_guppi_58405_85972_GJ687_0028.4001.818.21.44.244.vlar_0 ? https://en.wikipedia.org/wiki/Star_catalogue#Gl.2C_GJ.2C_Wo https://en.wikipedia.org/wiki/Gliese_687 |
|
10)
Message boards :
Number crunching :
Panic Mode On (112) Server Problems?
(Message 1942766)
Posted 5 Jul 2018 by Sebastian M. Bobrecki
Post: ... In MySQL You can try something like this to obtain date: select timestampadd(SECOND,floor(86400*cast(replace(substring_index(substring_index("blc15_2bit_guppi_58227_21810_HIP67105_0060","_","5"),"_","-2"),"_",".") as DECIMAL(12,6))) ,"1858-11-17");
|
|
11)
Message boards :
Number crunching :
New dedicated system for SETI/BOINC
(Message 1032187)
Posted 9 Sep 2010 by Sebastian M. Bobrecki
Post: Hello lupo From my experience you will not feel any significant improvement in productivity from the use of SSD drives either in Seti or in most other projects. So will the DDR2 and DDR3. Always two i7 Xeons are better than a normal Core i7 because with HT they can count 16 threads and Core i7 only eighth. Even two older Core 2 based Xeon should be better than a normal Core i7 because it's always full 8 cores instead of threads. If you also want to count on the GPU better solution may be one CPU + motherboard that can handle more graphics cards + more graphics cards :) Depends on what your budget is. Memory size can be easily estimated assuming that the application on a CPU uses about 50MB and GPU around 150MB. Taking two i7 Xeons we have up to 16 applications on a CPU + four Geforce GTX480 each of which can count two or three tasks at once so it give 8 or 12 in total on GPU. 16 * 50MB + 12 * 150MB = 2600MB So the size of memory I would not worry:) I hope I helped. |
|
12)
Message boards :
Number crunching :
Pendings
(Message 992013)
Posted 26 Apr 2010 by Sebastian M. Bobrecki
Post: FYI: My pendings have dropped from a ratio of about 6-1 a couple of days ago (pending of 200K+) to about 140K tonight. Well, you're a lucky guy. My ratio of pending to the RAC is increased to 7.628591096. And also my RAC began to drop :( |
|
13)
Message boards :
Number crunching :
Pendings
(Message 990626)
Posted 20 Apr 2010 by Sebastian M. Bobrecki
Post: But generally we are all responsible for this situation. In short: 1). Server sometimes crashes/run out of WU. 2). To avoid dry run we are setting bigger queues. 3). Bigger queues = servers must generate more WU. 4). More generated WU = faster fetch from tapes/images. 5). Faster fetch from tapes = servers faster run out of WU. Look 1). 6). More WU = more files and more database records. 7). More files and db records = more servers load. 8). More servers load = server crashes more often. Look 1). I think the situation is hopeless as long as people do not change their behavior. |
|
14)
Message boards :
Number crunching :
Pendings
(Message 990621)
Posted 20 Apr 2010 by Sebastian M. Bobrecki
Post: Pending/RAC = 6,292374201 |
|
15)
Message boards :
Technical News :
Creep (Nov 17 2009)
(Message 948028)
Posted 18 Nov 2009 by Sebastian M. Bobrecki
Post: I have some thoughts to consideration. Maybe you should go back with primary database to jocelyn which is more stable and use mork for some cpu intensive tasks like software radar blanker. I know that mork have a lot of ram that will be wasted in such tasks. But it may fit to other server, maybe jocelyn? Like always, thanks for the update Matt. |
|
16)
Message boards :
Number crunching :
Post your BOINC Startup 'CUDA' Info
(Message 939354)
Posted 12 Oct 2009 by Sebastian M. Bobrecki
Post: NVIDIA GPU 0: GeForce GTX 295 (driver version 0, compute capability 1.3, 895MB, est. 106GFLOPS) NVIDIA GPU 0: GeForce GTX 275 (driver version 0, compute capability 1.3, 896MB, est. 120GFLOPS) |
|
17)
Message boards :
Number crunching :
Welcome to the 10 year club
(Message 937955)
Posted 6 Oct 2009 by Sebastian M. Bobrecki
Post:
How could it be? P3 was introduced on February 26, 1999. Yes I know it was so many years ago so everyone may don't remember. |
|
18)
Message boards :
Number crunching :
- POLL - : SETI@home on ATI GPU..
(Message 937937)
Posted 6 Oct 2009 by Sebastian M. Bobrecki
Post: I don't have ATI card but I think that it should be application for ATI's GPU. But what You all should know, is that the "SETI Team" don't have resources (time, people, money,...) to make one. From information which I have, with CUDA application was same case, but Nvidia do big help with it. Majority of application code was written by Nvidia people. There is also huge difference in software side of CUDA compared to Stream, like availability of good working FFT library, SDK, examples,... Writing own FFT library is a huge thing, it's very complicated, needs optimizations, and sens of using given algorithm strongly depends on hardware. With CUDA Toolkit we see that FFT library is constantly evolving, so with Stream one it's need to have some developers which keep it live and to remove bugs. It of course maybe an open source project but "SETI Team" don't have time for that. And also who will better know ATI GPU than ATI's people. So maybe you should consider writing some petition to "AMD/ATI" or some open source programmers, not to "SETI Team". |
|
19)
Message boards :
Number crunching :
32b vs 64b
(Message 937936)
Posted 6 Oct 2009 by Sebastian M. Bobrecki
Post: To do some clearance in the matter of processor internals in x86 and x86_64 family: - ALU are capable to do operations on integers with or without sign of size 8bit, 16bit and 32bit for so called 32bit CPU and also for 64bit for 64bit CPU. It also can do operations on single or multiple bits, and on 4bit BCD code and 8bit condensed BCD code. - FPU (as from 8087) are capable to do operation on integers with sign of size 16bit,32bit and 64bit. On condensed (80bit ) BCD with the master byte representing sign. And also on floating point of size 32bit (single precision), 64bit (double precision), and 80bit (extended precision), but 80bit long are only for temporals stored in internal registers. There are no instruction to load or store 80bit fp values i memory. When you store content of 80bit register in memory it is rounded to single or double precision depends of instruction. In the byte code you have |ESCAPE MF 1|: - ESCAPE - mean the instruction is not for CPU (not for ALU or BIU), it also don't mean that it is for FPU it may be for 8089 I/O processor, but it's not important now. - MF - 00 - 32bit, 10 - 64bit - 1 - last bit of byte is set So even if internally are used extended precision, in memory it will be stored as single or double precision. And I think that it's clear that physical bus width don't have any impact on logical data format. Wit 8bit bus it only take more clock cycles to do transfer than for example on 32bit. But you know time was going forward and some day appeared new instructions sets like 3Dnow,SSE,SSE2,... Which are capable to do calculations on floating point data. And they are different from x87 instruction in that manner that if you have single precision input data then you also have single precision temporals, and analogically for double precision. So using different instructions may give slightly different result with same input data. There is a good, well known, example for this differences in simple Fortran code below: program hi real a,b,c,d real x,y,z a=.013 b=.027 c=.0937 d=.79 y=-a/b + (a/b+c)*EXP(d) print *,y z=(-a)/b + (a/b+c)*EXP(d) print *,z x=y-z print *,x end Compiled with x87 fp instructions: # g95 -o hi -mfpmath=387 -fzero -ftrace=full -fsloppy-char hi.for # ./hi 0.78587145 0.7858714 5.9604645E-8 And compiled with SSE fp instructions: # g95 -o hi -mfpmath=sse -msse2 -fzero -ftrace=full -fsloppy-char hi.for # ./hi 0.78587145 0.78587145 0. So as you can see if program have condition of this form: if ( number == 0 ) { ... } it may be a very big problem. That's why it not so simple do porting form x87 code to much faster SSE code. But using of SSE is some kind of standard with 64bit software. GCC for example automatically set -mfpmath=sse when -march is indicating 64bit CPU. There is also many more difference between processor running in 32bit mode and in 64bit mode, like increased number of registers in 64bit mode which are the fastest kind of memory for data storing, capability to use macro operations (for example, core2 can only use macro operations for 32bit data, core i7 for both 32 and 64 bit) and many more... |
|
20)
Message boards :
Technical News :
Phew (Oct 05 2009)
(Message 937802)
Posted 5 Oct 2009 by Sebastian M. Bobrecki
Post: Thanks for answer. I suspect that. I use xfs for years and never have similar problem with it. But with nfs shares there was a lot strange problems from corrupted files to locks on nonexistent ones. |
Next 20
©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.