Estimate the new primes

[xilman]

Originally Posted by Uncwilly

IIRC one could do something like bank swapping, address main memory then turn off access to main and access the cartridge. One can copy between cartridge and main memory. Maybe the Vic20 would be a better bet. They do have external memory expansion available.

Maybe a TI99/4A would be better.

I have one of those. Perhaps I should dig it out of the attic and see...

Paul

[Fusion_power]

I have both a working ti99/4a and a working commodore 64. No, you CAN'T have them, I do things with them from time to time.

[Nelson]

Originally Posted by ET_

nelson, I was joking...

Luigi

I realize that Luigi!

I have 2 Coleco Adams (Z80A) here that might make a good Cluster controller able to link about 240 6805's on it's serial interface plus hard drives and eliminate the need for floppies.

nelson

[ixfd64]

How about an M9000 server from Sun?

[Xyzzy]

http://www.mac-guild.org/raid.html

[Warped]

I can offer my ZX Spectrum.

[rgiltrap]

Originally Posted by ixfd64

How about an M9000 server from Sun?

Though I would love to use 256 CPU cores, the Mlucas code currently only runs up to 16 threads at present. Hopefully we'll reach 32 threads by the time M47 comes around. I don't think Glucas has been pushed past 16 threads either.

#threads    iteration    speedup vs    CPU
            time (sec)   1-thread      Utilization
--------    ---------    ----------    --------
 1          0.237         1.00x        100%
 2          0.138         1.72x         86%
 4          0.065         3.65x         91%
 8          0.038         6.24x         78%
16          0.022        10.8 x         68%

At 32 threads a speculative guess would be about 18x single thread performance on an SMP system. NUMA can be more difficult to scale due to the "Non-Uniform" latency between processors.

[T.Rex]

Originally Posted by rgiltrap

I don't think Glucas has been pushed past 16 threads either.

I do not remember limitations about the number of threads in Glucas. I think I used 32 threads at the beginning of the verification of Aug23th. I moved backed to 16 when I realized that HyperThreading was ON... I think I can use any number of threads. But I must check, don't remember the code...

I'm working (with the help of a colleague) to build data about a performance and scalability comparison of Glucas, Mlucas and Prime95. Will be interesting !
Already, it seems that Mlucas is more scalable than Prime95 :
With 8 cores on Harpertown 3Ghz, I've found that Mlucas runs 5.4 times faster than with only 1 core. Though Prime95 has a 3.3 factor only.
But this is only preliminary results... Ernst must confirm I'm using correct times.
Tony

[davieddy]

May I suggest:
http://mersenneforum.org/showthread.php?t=10645http://mersenneforum.org/showthread.php?t=10645

[uigrad]

Originally Posted by retina

I think that both are >10M digits (>33M exponent). This begin based upon my assumption that it would seem odd to hold back the announcement of the first discovery for a rather uninteresting (relativity speaking) 28M exponent.

Actually, the decision to hold off the announcements makes me think that M(Sept) is most likely small.

George wants a big press release for each new prime discovered. It would be hard to convince news agencies to run a story on a small discovery just a week after the big prize is handed out, so combining the stories into one release is absolutely necessary.


I have a T-99/4A that I could donate for the verification effort. With 16k memory each, it may take a large number of these to be able to handle just one FFT.

[davieddy]

Originally Posted by davieddy

Based on all the evidence, I guess exponents 42M for Aug, 38M for Sept.

Flatlander says 42M and 28M.

We have about 3 days to go before all is revealed.

Pats myself on back