2009/6/8 Tom Sharpless <[email protected]>:
>
> Hi allard
>
> LAPACK is a FORTRAN package, and Microsoft does not support FORTRAN.
> On Windows, you have to use CLAPACK, a C version generated largely by
> automatic translation.  CLAPACK still acts like FORTRAN, though, so
> you need several special headers and libraries to mediate between the
> C and FORTRAN environments (you also have to get used to 1-origin
> array indexing).  And you need a couple of "blas" libraries that
> provide the low-level arithmetic routines on which LAPACK depends
> (typically the blas codes are processor-specific, for speed, but there
> are generic versions too).
>
> And as with all Windows builds, you have to deal with the fact that
> you need separate debug and release versions of every library, that
> must have been built with the same compiler version that builds your
> app.
>
> So you would have to do a whole lot of preparative work before you
> could build Hugin with LAPACK on Windows.  The resulting gain of
> optimizer speed would be of little value, since stitching throughput
> is hardly limited by the optimizer.
>
> So in short, you will be much happier if you build without LAPACK.
>
> Indeed, LAPACK should be removed from Hugin, as it offers only
> trouble.  There are alternative hgih performance Levenberg-Marquardt
> codes in C++ (for example at AlgLib) if one really needed to tune up
> the optimizer.
>
> -- Tom

Hi, Tom,
as I've said I've no idea why MSVC requires LAPACK when it's clearly
optional. Moreover in current svn LAPACK have to be enabled explicitly
so it will not use LAPACK by default even when there is a LAPACK
library installed on system.

More to the speed. LAPACK is in fact a bit slower on my machine and on
Bruno's it's much slower. This lead to the disabling the LAPACK by
default even when there are necessary libraries.

I've explained why I've added LAPACK support under the rc2
announcement. In short because it gives me a bit better (not so bad as
without LAPACK) results with one of my test panoramas and doesn't
change results with the other ones. Moreover it's recommended to use
LAPACK in LEVMAR because it's more stable than it LU-decomposition or
whatever it uses.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"hugin and other free panoramic software" group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~----------~----~----~----~------~----~------~--~---

Reply via email to