Steinar H. Gunderson wrote:
On Mon, Oct 30, 2006 at 11:06:51AM +0100, Steinar H. Gunderson wrote:
I have no idea why you think these ranges are the only true ones and drop
automatically from nowhere. Prime95 ran with completely different ranges,
“completely� is, of course, an overstatement -- but they were different,
even
for non-SSE2 code.
The point was: They're not clear-cut and obvious, and if you step a bit
outside them, you'll still survive.
If you make the ffts too big you will be perfectly safe, if you make
them too small you won't.
I would expect the boundary IS rather sharp but harder than most people
think to predict and it varies slightly (a bit or two) from one example
to the next.
The old code was i believe being more conservative on the safe/large
side than it needed to be so it would be using ffts larger than needed
and thus be slower than it needed to be for the cases on the borderline.
The newer sizes take a slight risk of overflow which in the rare event
it is a problem is almost always detected early and corrected by
changing to a larger fft. The worst theoretical case would be it not
noticed till the double check (I wonder if this has EVER actually
happened?). In return it is faster for those numbers it takes the risk on.
*******************
Regarding those who think 100000 computers on this project is not enuf,
seems it is doing quite well to me. Setting records and getting results
that no one expected right and left. The fact that a few other projects
have larger numbers of contributers could be due to the abstract nature
of the project and nothing to do with the mode it runs in. For goodness
sake, this is MATHEMATICS! You know how many people LIKE MATH these
days? :-)
"screen saver mode" really isn't very efficient for this. It wont
scavenge all those cycles that computers waste in between keystrokes and
mouse activity. People surfing the net and doing email actually spend
MOST of the computer's time waiting while the human reads the screen.
About the only people who might really notice slight differences in
response are either gamers who are so hyped up on caffeine that the
screen refreshing at 70 hz "strobes" their abused eyeballs or people
whose system's are so starved for ram that they are paging and the 10mb
or so mprime uses makes the paging worse. Or both. Perhaps they SHOULD
be doing [EMAIL PROTECTED] or protein folding in a screen saver then and not
mprime, those other projects are worthy enuf that we should not need to
steal their cycles from them anyway.
If your machine does have enuf memory that you can spare 10 mb for
mprime (that generally would mean that its not already threshing the
paging file) then its pretty unlikely that you will really be able to
notice whether it is running or not. It is really good at getting out of
the way and staying in the background.
_______________________________________________
Prime mailing list
[email protected]
http://hogranch.com/mailman/listinfo/prime
