On Jun 10, 2008, at 00:40, Albert Graef wrote:
> Ryan Schmidt wrote:
>
>> I would love to know how the arithmetic test (or maybe q in general)
>> does its math though. I can't get any other app to tell me that 2 * 2
>> == 4.4501477170164897e-308, and finding out how it's computing that
>> might help me isolate what's wrong with my computer.
>
> You can find all the arithmetic routines in qmfuns.c, look for
> qmadd et
> al. Bigint arithmetic is done using the gmp routines, double
> arithmetic
> using the C floating point operators. Apart from that, it's just
> boxing/unboxing, which are just accesses to a C struct, nothing
> unusual
> there.
>
> The bigint arithmetic seems to work, since the first two results
> are ok.
> It's the mixed double/bigint arithmetic that fails there. The
> bigint ->
> double promotion is done using another GMP routine, mpz_get_d, I don't
> see how that could go wrong unless GMP is broken on that machine.
>
> You could try whether just plain double arithmetic works. If you
> can run
> the interpreter from the command line, try something like '2.0*2.0'
> instead of '2*2.0', or edit test.q accordingly and run 'make test'.
Thanks. Turns out my gmp is broken, and I know why [1] and I'm
working to fix it.
The fun output from q:
==> 1*1
1
==> 1.0*1.0
1.0
==> 1*1.0
2.22507385850824e-308
==> 1.0*1
2.22507385850824e-308
[1] http://gmplib.org/list-archives/gmp-bugs/2008-June/001031.html
|
