[quote=jrk;172984]If I re-do my calculation assuming a sieve of 10T (10e12) I get 0.63 primes and 47%.[/quote]

I like your figures better! :smile:

I'll have to compare your formula for to the spreadsheet that I set up for base 2 that was based on some formulas that were given at some point by axn1 in another thread on this forum.


Gary

Amazing News regarding k=3677878
 

Hello everyone who is interested.

It appears that for Riesel base=3, k=3677878 that Prime95 version 25.11 is 3.5 times as fast as availeable LLR versions. This means that regarding my n<=1M reservation for the k=3677878, is actually going to have an ETA around new year and maybe as early as november this year in stead of an ETA around late 2010 or early 2011. Only problem I've seen with the new versions of Prime95 and PFGW is that PFGW archive is not proberly compressed, hence the .exe file used to execute the OpenPFGW commands is missing from the compressed files. However with the new speed increases and if they are not inflected by the fact that one uses the stop at prime for each k command, this means that we will actually be able to come a lot further, and I might actually reconsider to either do some more Riesel base 3 testings or take the Sierp base 63 a bit further, since now it will be more fun to do these projects since more can be gained with little additional work required. :smile:

Regards

KEP

I haven't used Prime95. Can someone independently check Kenneth's speed difference for some other base? This seems like an usually big difference. Perhaps Rogue could try Prime95 on base 58 or Max could test it on base 94 if it's not too much trouble. If so, we may be able to use Prime95 for much more than base 3.

I'm wondering if Prime95 is comparable to Phrot on speed. Perhaps it makes sense to test LLR, Prime95, and Phrot on any new base that is reserved.

Kenneth, BTW, you probably should have been using Phrot on base 3. For base 63, it also probably should have been used if Rogue has a built in stop on prime option for a specific k like PFGW does.

[quote=gd_barnes;182045]I haven't used Prime95. Can someone independently check Kenneth's speed difference for some other base? This seems like an usually big difference. Perhaps Rogue could try Prime95 on base 58 or Max could test it on base 94 if it's not too much trouble. If so, we may be able to use Prime95 for much more than base 3.

I'm wondering if Prime95 is comparable to Phrot on speed. Perhaps it makes sense to test LLR, Prime95, and Phrot on any new base that is reserved.

Kenneth, BTW, you probably should have been using Phrot on base 3. For base 63, it also probably should have been used if Rogue has a built in stop on prime option for a specific k like PFGW does.[/quote]
Gary, see the thread I started a few days ago about Prime95 v25.11 and PFGW 3.1.0. It explains everything in there. In fact, that's the primary reason why I reserved base 94 in the first place: to test Prime95. :smile:

[QUOTE=gd_barnes;182045]I haven't used Prime95. Can someone independently check Kenneth's speed difference for some other base? This seems like an usually big difference. Perhaps Rogue could try Prime95 on base 58 or Max could test it on base 94 if it's not too much trouble. If so, we may be able to use Prime95 for much more than base 3.

I'm wondering if Prime95 is comparable to Phrot on speed. Perhaps it makes sense to test LLR, Prime95, and Phrot on any new base that is reserved.

Kenneth, BTW, you probably should have been using Phrot on base 3. For base 63, it also probably should have been used if Rogue has a built in stop on prime option for a specific k like PFGW does.[/QUOTE]

Well I did test Phrot at a lower n value, however LLR beat out Phrot, and that was why I abandoned the try to use Phrot since it would delay the effort of going to n=1M. Also as I understand it, even if Phrot should be faster also on base 63, using the stop on prime function, makes Proth stop after first primed k, and then it stops testing all k's unfortunantly. However if PFGW (though there seems to be a problem with the zip package) is also like 3.5 times faster than LLR at testing sierp base 63, it will mean that a test at n=25K would only take ~25 seconds. And a test at n=5K would take only a little more than a second :smile:

Actual timings using Prime95:

Prime95 PRP test of Sierp base 63 k=37565844 n=5000 takes ~2 seconds
Prime95 PRP test of Sierp base 63 k=37565844 n=25000 takes ~62 seconds

This means that there is for Sierp base 63 a 1/3 speed increase (~33.3%), not as magnificent as for Riesel base 3 however still a speed increase :smile:

Hope I answered everything. Take care everyone :smile:

Regards

Kenneth

Ps Can anyone (Max or Rogue) tell me where and how to get the new version of PFGW, since I get an "unexpected ending" error message on the zip compressed file after download completes?
PPs Does anyone know of a program that is able to split a file according to line number in stead of splitting it according to bytes like HJsplit does?

[quote=KEP;182104]Ps Can anyone (Max or Rogue) tell me where and how to get the new version of PFGW, since I get an "unexpected ending" error message on the zip compressed file after download completes?[/quote]
I've been able to download it without a problem, so I could repackage it and send it to you. However, I'd recommend against using PFGW for now; it has a bug that makes it output the first character of its residues incorrectly (which could complicate future doublechecking). This bug should be fixed in the next version; meanwhile, Prime95 operates at the exact same speed (I tested it), but without the residue bug, so I'd recommend using that instead for now.
[quote]PPs Does anyone know of a program that is able to split a file according to line number in stead of splitting it according to bytes like HJsplit does?[/quote]
If you have access to a Linux system (or Windows with Cygwin), the command "split -l [I]number-of-lines filename[/I]" will do the trick. If not, you can download the Linux "split" utility at [url]http://unxutils.sourceforge.net/[/url].

[QUOTE=mdettweiler;182122]I've been able to download it without a problem, so I could repackage it and send it to you. However, I'd recommend against using PFGW for now; it has a bug that makes it output the first character of its residues incorrectly (which could complicate future doublechecking). This bug should be fixed in the next version; meanwhile, Prime95 operates at the exact same speed (I tested it), but without the residue bug, so I'd recommend using that instead for now.

If you have access to a Linux system (or Windows with Cygwin), the command "split -l [I]number-of-lines filename[/I]" will do the trick. If not, you can download the Linux "split" utility at [url]http://unxutils.sourceforge.net/[/url].[/QUOTE]

Thanks for the links. If you can send me a repacked version of PFGW I would really like that. I'm not sure if you can e-mail a package containing an exe file to my e-mail, however if that is not possible I'm sure you find another solution.

Thanks for the links, I'll take a look at the program split.exe!

Regards

Kenneth

[quote=KEP;182129]Thanks for the links. If you can send me a repacked version of PFGW I would really like that. I'm not sure if you can e-mail a package containing an exe file to my e-mail, however if that is not possible I'm sure you find another solution.[/quote]
Okay, email on the way.

Riesel base 3 k=3677878 is now at n=610K, with testing times around 50 minutes on my Dual core. Wow this is fast and the remaining tests is approximately 6800 :smile:

Regards

Kenneth

Riesel base 3 k=3677878 is now at n=736K (and partially tested to n=825K). In total less than 4500 tests remain untill n=1M is reached. No prime yet. ETA is about 30-40 days from now, with an expected testing time at n=1M of around 3 hours. Currently testing is hammered on 5 cores, and as soon as this riesel base 3 testings complete, all 5 cores will be deployed to tackle the sierp base 63 conjecture :)

Take care everyone.

Kenneth!

Am impressive amount of work Kenneth. You've really stuck with it this time. That's more than I could say about myself. Nice work!

It would be something to get a humongous prime out of the effort. Good luck! :smile: