[QUOTE=Puzzle-Peter;254985]I am confident Mark will be able to solve the issues I'm having with running the script in Linux. If / when he does, I might make k=63G my CRUS goal for the year :grin:[/QUOTE]

:surprised

The fixed linux version should be ready later today. Note that it is only an issue on linux as it was due to a CPU specific build of GMP. I rebuilt GMP on Steven's box as "generic x86". He was linking with the Ubuntu distribution and I have no idea how that one is built. It will be better going forward as I'll have more control over the version of GMP that gets linked with the linux version.

Great job on getting k to 1G. I'm gonna take some lower ranges up to n=100K, starting with k=200M-210M, and in the longer run to 300M.

Reserving 1G - 1.1G

I'll use this to test Mark's latest modification and - in case I encounter no more errors - evaluate the speed benefit of 64 bit Linux vs. 32 bit Windows. Plus the Linux boxes have higher clock frequencies (well, most of them have). After this, I will decide where to put my long term goal.

[QUOTE=Puzzle-Peter;255070]I'll use this to test Mark's latest modification and - in case I encounter no more errors - evaluate the speed benefit of 64 bit Linux vs. 32 bit Windows. Plus the Linux boxes have higher clock frequencies (well, most of them have). After this, I will decide where to put my long term goal.[/QUOTE]

Seems to be running smoothly, so I fed a few cores. Therefore I reserve 1.1G - 5.7G to n=25k. I created a little script to update min_k, so I can stop in the morning and restart in the evening without too much hassle. Let's see how this works out...

[QUOTE=Puzzle-Peter;255134]Therefore I reserve 1.1G - 5.7G to n=25k.[/QUOTE]

:surprised:surprised:flex::bow:

[QUOTE=gd_barnes;255146]:surprised:surprised:flex::bow:[/QUOTE]

:grin:

It will take several weeks. And we will have to think about a way of exchanging the results. Mailing 10M blocks seems to be rather bothersome...

[QUOTE=Puzzle-Peter;255148]:grin:

It will take several weeks. And we will have to think about a way of exchanging the results. Mailing 10M blocks seems to be rather bothersome...[/QUOTE]

Several weeks? How many cores will you have on it? 30? 40? 50? It took my I7 (equivalent of ~5-6 cores) 8-10 days to do a k=100M range to n=25K. So assuming 40 CPU days to do k=100M means that your range would take 40*57=2080 CPU days. So 40 cores would take ~55-60 days.

Is that in the ball park of what you are thinking?

There is one thing that I want to bring up here: The boredom factor. I and many others have run up against that on these kinds of efforts. If you can stomach finding billions of teeny primes for weeks and months at a time, then it is something that could be done.

[QUOTE=gd_barnes;255152]Several weeks? How many cores will you have on it? 30? 40? 50? It took my I7 (equivalent of ~5-6 cores) 8-10 days to do a k=100M range to n=25K. So assuming 40 CPU days to do k=100M means that your range would take 40*57=2080 CPU days. So 40 cores would take ~55-60 days.

Is that in the ball park of what you are thinking?[/QUOTE]

That is about what I was thinking, yes. I can give better estimates tomorrow when all those machines (50 cores) have crunched for one night, so I can extrapolate the results.

[QUOTE=gd_barnes;255152]
There is one thing that I want to bring up here: The boredom factor. I and many others have run up against that on these kinds of efforts. If you can stomach finding billions of teeny primes for weeks and months at a time, then it is something that could be done.[/QUOTE]

That's why I try to always have several things running at a time. On the other hand I am nearly done with all k's for SR5 covering n ranges adding up to 45000. Before that I spent several months doing nothing but 27121 tests for Prime Grid. So it's not too easy to bore me. Plus I realized you guys don't ask for updates every other day, so I know I can always put a few cores on another task. That helps too.

But I have to admit I am relieved that R301 is close to 25k...

[QUOTE=Puzzle-Peter;255148]:grin:

It will take several weeks. And we will have to think about a way of exchanging the results. Mailing 10M blocks seems to be rather bothersome...[/QUOTE]
For larger files, you can use [url=http://www.sendspace.com]Sendspace[/url], though that too can be a bit of a pain if you have to use it a lot (what with all the ads it has). If you have a very large amount of data you need to transfer to Gary, it would probably be best for me to set you up to make direct SCP/FTP/SFTP transfers into the noprimeleftbehind.net server (from whence Gary can copy them directly to his external USB hard drive, which IIRC is where he keeps all these CRUS results).

Too bad. I had tested this on one of the newer boxes. The others started automatically in the evening and most of them errored out with "FATAL - kernel too old" messages.

Therefore I UNreserve 1.1G - 5.7G and concentrate on other subprojects. Sorry for that! I will return to R3 now and then for small ranges...

R3 k=3677878 tested from n=950k to n=1M, no prime. k released.