>> Double-check assignments are identical to Lucas-Lehmer assignments, however
a
>> (D) also appears to the right of the exponent. These comparatively rare
tests
>> recompute earlier residues of questionable reliability.
>
>I thought part of the GIMPS process was to double check _all_ exponents on a
>different platform than first tested on (or the same platform with v17.x of
>the client). Am I mistaken about this?
as recently discussed, George has tweaked the latest version of Prime95 to
'shift' the lucas lehmer tests around significantly, and now feels that is an
adequate double check.
The fundamental problem is pentium's running prime95 and derivatives vastly
outnumber other platforms, especially when measuring 'p90 hours' or equivalent
measure. If we needed to double check on macintosh and unix platforms, we
would never complete as first run pentium numbers would be coming out so much
faster.
-jrp
