Subject: Problem with Maxima 5.17.1 and ECL 8.12.0
From: Raymond Toy
Date: Wed, 17 Dec 2008 22:49:56 -0500
Robert Dodier wrote:
> On Wed, Dec 17, 2008 at 6:45 PM, Raymond Toy <toy.raymond at gmail.com> wrote:
>
>
>>> By the way, what does ECL report for MOST-POSITIVE-FIXNUM ?
>>>
>
>
>> Good catch! most-positive-fixnum is 61 bits long, 2^61-1. I guess
>> 115...75 is 2^60-1.
>>
>
> There must be more to this story. For Clisp and ECL on a Linux box
> on which the build succeeds, MOST-POSITIVE-FIXNUM is 2^24 - 1
> and 2^29 - 1, respectively. So those must be executing the default
> case. Maybe EVAL-WHEN is being handled differently?
> Or sometimes ECL tries harder to resolve undefined functions??
>
The difference is that with that large fixnum, newprime eventually calls
primep and primep has to use $jacobi. For the other cases, primep
appears to be able to use primep-small, which doesn't call $jacobi.
Maybe moving $jacobi into rat3c is enough?
>
>> Perhaps it's not a good idea to initialize bigprimes with #.foo.
>> Perhaps a regular initialization is good enough.
>>
>
> OK by me, but I don't know anything about how to get EVAL-WHEN
> to do what we want.
>
>
Actually, I'm not very good with that either.
Ray