|
Boost : |
From: Joel Young (jdy_at_[hidden])
Date: 2004-05-13 12:17:29
I believe there is a numeric error in big_integer. I am working
to isolate it to a single operation.
I can find the error now because the library is fast enough to
run one of my sims. I get a different value when running with
rational<big_integer>
then with any of these choices (which give identical results)
rational<long long>
rational<big_integer (cln)>
rational<big_integer (gmp)>
cln::cl_RA
Not that it is much use but the wrong answers are:
Inonsistent Value:
38485535445/562949953421312
6375930541/562949953421312
Consistent Value:
42780502741/562949953421312
10670897837/562949953421312
Again, I'll try to narrow down the exact point where things go
wrong but it might take a while.
Joel
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk