Upgraded LLR program
- Hi, All !
The new LLR has arrived !
It uses Irrational Base Discrete Weighted Transforms to square and multiply=
I succeeded to update C init. code and Normalization Assembler code in the =
sources of the
George Woltman's Gwnums system, in order to make this possible.
Unfortunately, there are severe limitations for the k values (in this first=
So, this program uses IBDWT only if k <= 511 in non SSE2 code, and (alas...=
) k <= 31 in
SSE2 code ; for larger k's, it works exactly like the previous LLR version,=
using Proth mode
and rational bases FFT. But for really small k's, it is a major improvement=
With P4 and SSE2, it tests 3*2^n-1 numbers more than four times faster than=
previous LLR !
So, Paul Underwood and the 3-2-1 project members would be happy !
- Excellent job on the great speed improvment made Jean!!!
Where can we get this latest version at?
http://www.mersenne.org/gimps sitll shows old version.
Also, is the speed improvement just for k*2^n-1? In other words,
3*2^n+1 doesn't get a speed improvent?
Thanks for the responses.
the program seems to have a bug if a "save file" is loaded. A save
file is created when you press "test-stop" -- SO DON'T PRESS IT. If
LLR has crashes on you:
1) stop the process: kill llr.exe
2) delete the save files -- these start with 'x' or 'y' and have 7
digits after that in their file name.
3) restart the new llr and set "options->minutes between disk writes"
to 99999 so that no save file is ever going to be created.
I have told Jean about this problem and he said that he would look
into after a holiday to India!
- Can someone please let me know the most up to date source for
k*2^n-1 ranges reserved. According to
k=23 has been tested to , is this still accurate? I'd
like to reserve it and search a bit higher.
- --- In firstname.lastname@example.org, "andrew_j_walker" <ajw01@u...>
>Thanks to everyone who replied to me. To summarise, the link
> Can someone please let me know the most up to date source for
> k*2^n-1 ranges reserved. According to
> k=23 has been tested to , is this still accurate? I'd
> like to reserve it and search a bit higher.
> Andrew Walker
above is as accurate as it could be, it appears many people have
searched higher without informing Wilfred!
http://www.15k.org/ has a search for multiples of 15*k*2^n-1
and for k*2^n-1 for k<300. I'm going to start k=5 which will largely
be verification for small k.