?iga Lenar?i? wrote:
>> Last time i have used 19f on the Mac i got severe errors in
>> computations.
>> Reverting to 19e solved the problem. I advise you to check
>> thorougly on
>> your Mac if 19f really works. I have a post in this newsgroup
>> presenting
>> a computation where 19f produced completely wrong results.
>> Apparently this
>> is only on the Mac since other people didn't see the problem on their
>> own machine (apparently not under OS X).
>>
>> --
>> Michel Talon
>
> I ran the testsuite - it ran without error. If there will be any
> reports about errors, I will make another .dmg, either with older
> CMUCL or SBCL. For now I'm assuming 19f will work :)
The computation that gave me erroneous results with 19f on a Mac is here:
http://www.math.utexas.edu/pipermail/maxima/2009/015847.html
One diagonalizes a symmetric matrix by the Jacobi method, and one compiles
eigens-by-jacobi.lisp to get eigens-by-jacobi.ssl2 before running the
computation. For me the extreme eigenvalues were very false. The correct
results are given in the following message of R. Toy:
- 2.045740417937529 and 2.876115587743105
If you get these values with your version, then you are not seeing the bug
i mentioned. I got something like -4 and +6.
Thanks a lot for your attention.
--
Michel Talon