Marc-Etienne <[EMAIL PROTECTED]> writes:
>We will soon hit cases like
>1419967,wayney,WP1
>1419967,SB,X
>1419967,MV1,J3
>(in file hrf3)
>where three different programs (Woltman, mersenne1 and MacLucas) give
>three different results. How will this be resolved? Has one of the
>programs a known bug in this area?
I have had this range reserved for double-checking (via George Woltman) since
before Prime95 v17 was released.
I queried George about this when I discovered the "problem", I thought it was
possible that the exponent had accidentally escaped being moved to the
double-checked list, but apparently the problem is that the residuals which
have already been reported to George do not agree with each other.
This exponent is not the only one of these I know about, though the other two
instances in the range I am working on have only two results reported.
When different programs report different results, Something Has Gone Wrong.
Could be a program bug, or a hardware glitch. We simply Don't Know. What I
intend to do when I get round to these exponents is to use another different
program (normally I'd use MacLucasUNIX because it's faster than the other
programs fft, mersenne1, mersenne2 which I have available, but there's no
reason why I shouldn't use any of the four). On completion I'll report the
residual to George & let him take it from there. (I don't know any of the
residual values returned already ...) I think it makes sense, if my residual
matches one of the others, that value is probably correct but shouldn't be
accepted without rerunning *again* using another independent program (Prime95
any version is different enough to what I will be using...) If the residual
I get happens to differ from all the other results, then probably the best
thing to do is to throw away *all* the results for that exponent & start from
scratch.
There are also a number of exponents which have two logged results but which
have been computed using essentially the same program (variants of Prime95 prior
to v17). These still require double-checking! I will work around to doing
the instances on my list, I will use MacLucasUNIX on either a Sun/Solaris or
an Alpha/Linux system, depending on which needs work at the time. If you run
Prime95 v17 on any of these exponents, you would just be doing work already
assigned to me, which doesn't bother me unduly.
Regards
Brian Beesley
