I've noticed that many, if not all work units seem to start out moving at a slow rate and speed up as percentage completed increases. I can't recall whether this also occurred with V6 seti.

IMHO, this is something I've noticed even when crunching SETI units before the BOINC era. That behaviour has always been around and hasn't changed ever since.

Thanks Eric for your explanations & information sofar. Looking forward to further tweaks and improvements,

Kind regards

---

To boldly crunch ...

You're replying to a post. Have you read the post? For if you would have, you could've known why present calculations with v7 are so much slower.

---

I am not reporting longer times on my Linux box by comparing SETI@home 6.03 and 7.01. I was using an AP 6.01 by Lunatics and when SETI@home 7.01 appeared I renamed the app_info.xml hoping to get an AP 7.01 too. But stock AP 6.01 is much slower than the Lunatics app, so I went back to the Lunatics version. Only CPU, no GPU.
Tullio

---

Eric: Dedisperse, dechirp, "N" factor...Websters gonna love you...and can u say "Infinity times Infinity"? - (Boom)

And - a great job you and the staff are doing. Thanks to all.

Unfortunately I am not good at computations or anything to do with computer programs. I am a layman. I joined because I believe the S.E.T.I. project is on the right track and will be the first to discover that we have other life in the universe. It is fun to watch how much gets done by me doing little to nothing. Can you explain what a signal would most likely look like?

Unfortunately I am not good at computations or anything to do with computer programs. I am a layman. I joined because I believe the S.E.T.I. project is on the right track and will be the first to discover that we have other life in the universe. It is fun to watch how much gets done by me doing little to nothing. Can you explain what a signal would most likely look like?

A small spike on a graph.

---

You will be assimilated...bunghole!

A more helpful and useful response would not just be ask the person to re-read the thread, but also mention what he should be looking for and notice.

So the increase in computational times _may_ be a result of size increases in the work-units. Or Lunatics may still be faster than v7 even if the size of work-units coincidentally increased about the same time? Can we get a _constructive_ answer?


Personally, I was using 6 this year before 7 and I want see and find some stats on that. I had used Lunatics for years previous and that was far better than 6 but I don't know which would be better now -- again, I just wanted to see some stats on 6 vs 7.

Actually, if there were published stats, and people could see them and check out the engines and stuff, the bug preventing cuda updates/optimizations previously may have been noticed in stats rather than by Eric looking through the code.


I'd like to see how much better my CPU (class) is now with v7, and how it did with v6 and perhaps how effective Lunatics are/were and how all the CUDA's and GPU's are doing (so we could have noticed the bug/discrepancy).

Stats like those could even let us see when WU increase in size and how different WU may have been processed.


Perhaps Seti could get some statisticians to crank seti processing data for their own scientic purposes and not waste Seti resources, although Seti must already be doing most of this if they know which optimal version of CUDA (7+ versions now) to send GPUs.

A more helpful and useful response would not just be ask the person to re-read the thread, but also mention what he should be looking for and notice.

More helpfull responses are more likely if the question were asked in the appropriate forum.

---

Grant
Darwin NT

A more helpful and useful response would not just be ask the person to re-read the thread, but also mention what he should be looking for and notice.

I wasn't saying one should (re)read the thread, only to read the first post that one is answering to, especially so when the first post can yield the answer to the so sought question. On the other hand, I am not going to hold one's hand, or go quote parts of the first post, if they couldn't even take the time to go read it in the first place.

---

So how do you search for the unexpected? I understand the basic idea was looking for signals in the H2 noise. is that still the case or have we moved on to other spectra. Please excuse my ignorance, I was away from Physics for over 30 years, though I have tried to read the popular literature. I was a Nuclear Engineer for over 20 years until the field dried up.

---

Sounds like a good question for the Seti@home
Science forum.
http://setiathome.berkeley.edu/forum_forum.php?id=9

As someone who knows just enough to make wrong statements, I'll ask...

Isn't the way the FT is calculated in practice is by taking the FT of the AC of the signal? So the AC looks like a sample of the signal(?) What do you get by looking at the AC that you don't get by looking at the signal?

The signals we are looking for are a few percent of the noise level in the 2.5MHz band. We're basically looking at the Fourier transform in order to increase the signal to noise ratio for the signal by a factor of 300 or so by confining the signal to a single channel while spreading the noise out over 131072 channels. That's the basic principle. The details are a bit more complicated.

---

@SETIEric@qoto.org (Mastodon)

Since the autocorrection is now in effect, does that mean the past work units need to be re-processed?

---

The first priority will continue to be analyzing new data, including the data from the Green Bank Telescope observations of Kepler planetary systems. (That's what I'm trying to work on this week.)

The second priority is filling in the gaps where some data wasn't processed because older versions of the data splitter tended to quit when they found bad data rather than working around it. We've fixed those problem.

The third priority will be to reanalyze data that was processed before we developed to software radar blanker. That data so contaminated with radar that it's difficult to find non-radar signals in it.

Then, if we aren't being flooded with new data, we might put some of the old tapes on to get autocorrelation and the extra sensitivity that the new polyphase splitter would get us.

---

@SETIEric@qoto.org (Mastodon)

Sounds like a good plan.

And a stupid question. If you are re-doing data, is it possible to just look at the areas that were skipped, or do you have to go through the entire data as if it was new?

---

For the areas that were skipped (it might be portions of a data file or one of the 14 receivers that got skipped) we're just doing the portions that were skipped.

For the data that wasn't radar blanked with the software blanker we'll have to do it all. (IIRC, there are about 18 months of calendar time in that set).

---

@SETIEric@qoto.org (Mastodon)

According to this article many US universities do tons of classified military jobs for Pentagon: http://demilitarize.org/fact-sheets/enfact-sheet-pentagon-universities/

Boinc is only top free software in the world that surely reaches to the 200 countries of the planet into 338k pc's so what credentials can prove that UC Berkely based Boinc/Seti software is clean?!

As of the article MIT is pretty much Pentagons R&D facility! How much does MIT involved into Boinc?!

Well since PRISM awakining everybody is waking up me hope.

---

Mandtugai!

AFAIK no military intelligence in the world would care about a program searching for alien intelligence. The Royal Air Force has just closed a UFO documentation office involving just 2 people. There are no UFO.
Tullio

---

Of course there are flying objects which have not been identified but they are not from outer space.

---