llr 3.8.2 released as dev-version
 

This for those of you who only read the PSP forum here at mersenneforum: George Woltman has released a new version of gwnum (26.2) with a lot of speedups. Jean Penné has compiled a development version of llr (3.8.2) with the new gwnum library.

I'm currently testing this. If everything seems OK, and I get a little help from runesk, we'll have updated clients quite soon here:

[URL]http://tools.1up.no/psp/prpclient/[/URL]

For more detailed info, read these threads:

[URL]http://www.mersenneforum.org/showthread.php?t=13957[/URL]
[URL]http://www.mersenneforum.org/showthread.php?t=13887[/URL]

GWnum 26.3 with a lot of bug fixes
 

is going to be released in a week or two. Why not wait until then?

Good point. :)

Although, we will have to wait for Jean Penné to compile a new llr too.

JP was quick to get this latest one out. It is an "easy" update to add into everything. Some people report 15% speed increases with some machines closer to 25%. Once the next round of updates they expect better speeds still.

20% speed up on llr testing is another reason to kill the sieve.


S.

[QUOTE=Sloth;231620]

20% speed up on llr testing is another reason to kill the sieve.


S.[/QUOTE]

My belive is this is not right moment to stop sieving. We are near the 10M area in LLR testing and when we go over to 10M test kandidates things will be 15-25% slower again (longer bit range). Right ?

[QUOTE=opyrt;231608]Good point. :)

Although, we will have to wait for Jean Penné to compile a new llr too.[/QUOTE]

I'm very disappointed that a new LLR is not yet available. PFGW has updated to GWNUM V26.4 which incorporates a lot of bug fixes.

LLR 3.8.4 dev. version is available for testing!
 

[QUOTE=Joe O;234264]I'm very disappointed that a new LLR is not yet available. PFGW has updated to GWNUM V26.4 which incorporates a lot of bug fixes.[/QUOTE]

Hi,

You may now download the Windows and Linux binaries of the development version 3.8.4 of LLR on my development directory :

[url]http://jpenne.free.fr/Development/[/url]

This version is identical to 3.8.2, apart from beeing linked with gwnum V26.4

I will upload the corresponding sources when "official" gwnum V26.4 will be released by George.

Regards,
Jean

Below is a comparison between 3.8.0 and 3.8.4.dev:
[code]
[B]3.8.0[/B]
7019*10^573413-1 is not prime. RES64: 2CE5D19B403686F0. OLD64: 86B174D1C0A394CE Time : 5457.451 sec.
7666*10^573413+1 is not prime. RES64: 6CA3801A0D62D64C. OLD64: 45EA804E282882E1 Time : 6629.058 sec.
7666*10^573437+1 is not prime. RES64: 1467BF6C6AF94E56. OLD64: 3D373E4540EBEAFF Time : 6618.179 sec.
4421*10^573478-1 is not prime. RES64: 0F836A8791496877. OLD64: 2E8A3F96B3DC3962 Time : 5435.556 sec.
4421*10^573508-1 is not prime. RES64: 41BF4A481270147A. OLD64: C53DDED837503D6B Time : 5438.300 sec.
7019*10^573569-1 is not prime. RES64: A2D19A672B82C159. OLD64: E874CF358288440A Time : 5447.202 sec.
7019*10^573617-1 is not prime. RES64: 1DB086228971AC98. OLD64: 591192679C5505C5 Time : 5434.558 sec.
7019*10^573749-1 is not prime. RES64: 246914663E6F31E9. OLD64: 6D3B3D32BB4D95B9 Time : 5456.040 sec.

[B]3.8.4[/B]
7019*10^573413-1 is not prime. RES64: 2CE5D19B403686F0. OLD64: 86B174D1C0A394CE Time : 5978.472 sec.
7666*10^573413+1 is not prime. RES64: 6CA3801A0D62D64C. OLD64: 45EA804E282882E1 Time : 5829.368 sec.
7666*10^573437+1 is not prime. RES64: 1467BF6C6AF94E56. OLD64: 3D373E4540EBEAFF Time : 5830.718 sec.
4421*10^573478-1 is not prime. RES64: 0F836A8791496877. OLD64: 2E8A3F96B3DC3962 Time : 4871.306 sec.
4421*10^573508-1 is not prime. RES64: 41BF4A481270147A. OLD64: C53DDED837503D6B Time : 4849.279 sec.
7019*10^573569-1 is not prime. RES64: A2D19A672B82C159. OLD64: E874CF358288440A Time : 6012.323 sec.
7019*10^573617-1 is not prime. RES64: 1DB086228971AC98. OLD64: 591192679C5505C5 Time : 6259.382 sec.
7019*10^573749-1 is not prime. RES64: 246914663E6F31E9. OLD64: 6D3B3D32BB4D95B9 Time : 5957.289 sec.[/code]
Overall I see an improvement over 3.8.0 / 3.8.1 version with one exception: 7019*10^n-1... Any ideas?

[QUOTE=Cruelty;235174]Overall I see an improvement over 3.8.0 / 3.8.1 version with one exception: 7019*10^n-1... Any ideas?[/QUOTE]

I suggest you post in this forum instead:
[URL]http://www.mersenneforum.org/forumdisplay.php?f=10[/URL]

You're more likely to get attention from the right people there. :)

GWnum version 26.4 is ready for testing.
 

Bug fixes are described [URL="http://www.mersenneforum.org/showpost.php?p=233281&postcount=2"]here.[/URL]

bug? - no results file
 

In both Windows and Linux, LLR isn't writing a Results file, neither upon startup nor when a prime is found. I wonder if anyone else is having this problem? the lresults file is getting written all right, & I can find the primes by grepping for a !, but it would be nice to have the results file.

Thanks,