| View previous topic :: View next topic |
| Author |
Message |
Martin
Guest
|
 Posted: Fri Aug 15, 2003 4:53 am Post subject: Re: Work unit age |
 |
|
Robi wrote:
[quote]Martin wrote:
Stratcat® wrote:
[...]
do, but I>m assuming for WU>s to be corroborative, they>d have to be
consistent within a reasonably narrow window of deviation.
The returned results are compared for an EXACT match. The results must
be IDENTICAL, otherwise there must have been a computational error.
Sorry, but no.
With the fft algorythms there are "rounding" errors which differ
from processor to processor (and OS? not sure here).
[...][/quote]
Yes...? Please explain further...
(Big hint: Have a look at where the rounding errors are. Note that
numbers for the results will be deliberately rounded for the analysis! (:-))
[quote]The only room for variation is for exactly where in the analysis a WU is
abandoned due to RFI. The whole result is discarded then in any case,
other than for marking that WU as unusable.
I honestly doubt that. Agreed, the analysis is stopped, but only due to
an overflow of data. That doesn>t mean that the WU is useless (or unusable
as you put it).
[/quote]
On what basis?...
[...]
Martin
--
----------
- Martin -
- 53N 1W -
---------- |
|
| |
|
 Back to top |
Robi
Guest
|
| Posted: Fri Aug 15, 2003 11:34 am Post subject: Re: Work unit age |
|
|
Martin wrote:
[quote]Robi wrote:
Martin wrote:
Stratcat® wrote:
[...]
do, but I>m assuming for WU>s to be corroborative, they>d have to be
consistent within a reasonably narrow window of deviation.
The returned results are compared for an EXACT match. The results must
be IDENTICAL, otherwise there must have been a computational error.
Sorry, but no.
With the fft algorythms there are "rounding" errors which differ
from processor to processor (and OS? not sure here).
[...]
Yes...? Please explain further...
(Big hint: Have a look at where the rounding errors are. Note that
numbers for the results will be deliberately rounded for the analysis! (:-))
[/quote]
ok, not fft rounding errors:
http://setiathome.ssl.berkeley.edu/newsletters/newsletter8.html
"There are several reasons why a result returned by a SETI@home volunteer
might be incorrect. The most common reason we get incorrect results is
processor malfunction. If a processor overheats, perhaps because there is
dust buildup inside the machine, or maybe it>s just a really hot day, the
first part of the chip to fail will be the most complex part, the floating-
point unit. A failure of the floating-point unit, which is responsible for
most of the calculations performed by SETI@home, will usually not cause a
computer to crash. It will cause the computer to generate incorrect results.
These innocent failures are responsible for most of the incorrect results
we see. The most common symptom of this problem is that every result from
malfunctioning computer contains hundreds of potential signals. Of course,
some valid results also contain hundreds of signals."
[quote]The only room for variation is for exactly where in the analysis a WU is
abandoned due to RFI. The whole result is discarded then in any case,
other than for marking that WU as unusable.
I honestly doubt that. Agreed, the analysis is stopped, but only due to
an overflow of data. That doesn>t mean that the WU is useless (or unusable
as you put it).
On what basis?...
[/quote]
"There are also many work units that were processed by more than one
version of the SETI@home client. More recent versions include analysis
that was not present in the early versions, so certain signals will
only be found with new versions"
IOW, a more advanced client could identify RFI and filter it out, catching
a signal in the background. Of course, such a client would first have to
be designed, but it shows that even if a WU has RFI it doesn>t mean it is
useless.
--
Robi
(2.7#@ 2.62 yrs) |
|
| |
|
| Back to top |
David Woolley
Guest
|
| Posted: Fri Aug 15, 2003 11:54 am Post subject: Re: Work unit age |
|
|
In article <yTJ_a.12237$yl6.8051@newsfep4-winn.server.ntli.net>,
Martin <ml_news@ddnospamddml1dd.co.uk.dd> (bogus) wrote:
[quote]The returned results are compared for an EXACT match. The results must
be IDENTICAL, otherwise there must have been a computational error.
[/quote]
The match is within a tolerance for rounding errors. Different compilers
may load and save registers at different places. Registers normally carry
more significant digits than the official number required for the type of
number.
Also, the chirping factors are calculated incrementally, with occasional
recalibrations, but if you stop and start the client, the restart values are
calculated directly, not by computing the last recalibration point value
and iterating forwards.
Both these factors introduce some indeterminacy in the numerical values, but
these indeterminacies are much smaller than those due to system noise (although
that is common to all clients) and therefore do not compromise the science.
[quote]The only room for variation is for exactly where in the analysis a WU is
abandoned due to RFI. The whole result is discarded then in any case,
[/quote]
WU>s are not abandoned for RFI, they are cut short when they exceed a
certain result file length, for whatever reason. Rounding error, may
cause borderline cases to be included in one set of results, but not
another, so the cut off may vary because of that, but the same build,
run without interruption, should stop at the same point.
Such a work unit could be the victim of a strong ETI signal, as well as
local RFI. It>s speculation that they are ignored entirely. Correlating
an overload at the same place but different times, might itself be
significant, although signals strong enough to overload the client may
well already have been detected by SERENDIP (whose data is used for S@H).
other than for marking that WU as unusable.
[quote]Reconfirming a result beyond 3 exact matches has ever diminishing value
for any further results returned.
[/quote]
The most that you can do after sufficient matches have been obtained is
to invalidate the whole work unit, as working data has started to be
deleted and is expensive to re-extract. |
|
| |
|
| Back to top |
pahis
Guest
|
| Posted: Sat Aug 16, 2003 2:57 am Post subject: Re: Work unit age |
|
|
[quote]
The returned results are compared for an EXACT match. The results must
be IDENTICAL, otherwise there must have been a computational error.
[/quote]
somehow i see this like bullshit.
if, and or, and to be, and why, someone will see this kind of computation
to be an measurement of power, or lenght, or just size,
why should somebody>s unique result of peak, or power, or direct signal,
somehow to be seen ,consired, or to be claimed as true, to be an exellent,
or worst scenario, or, jus>t be token as true, lie, or white lie, or false,
or?
when it is enough for what?
as you all see, my motherlanguage is not english....
pasi |
|
| |
|
| Back to top |
FalconFly
Guest
|
| Posted: Sat Aug 16, 2003 1:37 pm Post subject: Re: An embarrassment of riches - where>s the bottleneck..? |
|
|
[quote]That>s my attitude too. It>s pretty. I>m stuck with an AMD 233 Mhz,
doing my first WU. I>m at 25%, took mt 64 hrs CPU time to get this far,
so will take me probably 10 days to get done.
[/quote]
Then I>d recommend using the CLI Client instead of the GUI you seem currently using.
The AMD K6 233Mhz can complete a WorkUnit in roughly ~50 hours.
~10 days is the performance region of a 80486/100MHz Class system.
Greetings
FalconFly
Webmaster
http://www.falconfly.de
3dfx Archive
====================
Forum : http://www.falconfly-central.de |
|
| |
|
| Back to top |
~misfit~
Guest
|
| Posted: Tue Aug 19, 2003 11:30 am Post subject: Re: An embarrassment of riches - where>s the bottleneck..? |
|
|
"Stratcat®" <me@no.org> wrote in message
news:EeD_a.7078$Ih1.2298487@newssrv26.news.prodigy.com...
[quote]
"Nick M V Salmon" <spam_dump@btinternet.com> wrote in message
news:bhc54l$h7q$2@sparta.btinternet.com...
snip
I have 4 more sub 3.5Hrs/WU Athlons (Total 7.4GHz ) that I could put
online
with gear I already own but I haven>t because it would only be a
pointless
waste of electricity 'stats chasing' - roll on BOINC ASAP..!
Ciao...
[UK]_Nick...
No problem...send 'em to me. I>ll be more than happy to pay shipping, and
chase some stats!!!
[/quote]
I>ll split 'em with ya Strat. Whaddya say Nick? ;-)
--
~misfit~
---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.510 / Virus Database: 307 - Release Date: 14/08/2003 |
|
| |
|
| Back to top |
|
