Use vecLibFort instead of dotwrp

[tenomoto (Takeshi Enomoto)]

I added vecLibFort port to use it in ncarg. vecLibFort should fix all the problems with vecLib framework used from Fortran. The author of vecLibFort kindly sent me a pull request to dotwrp, which I haven't noticed for months, and I updated dotwrp. vecLibFort is more capable because it allows both static and dynamic linking and it includes a preloaded library. vecLibFort allows the use of vecLib in octave and qrupdate at the same time. So I switched from my dotwrp to vecLibFort in ncarg. I'll make dotwrp obsolete when all the ports using dotwrp (octave only?) switch to vecLibFort. The accelerate variant of petsc, itsol and hpl can use vecLibFort if they use vecLib from Fortran. Any other ports? Some ports in python?

[seanfarley (Sean Farley)]

Nice! Though, PETSc is written in C, so it's calling vecLib / Accelerate without Fortran. The Fortran wrappers that PETSc provide are just that, wrappers.

[tenomoto (Takeshi Enomoto)]

Thanks, Sean. You don't have any problems with PETSc. I'd like to remove you from cc and petsc from port, but I can't. Sorry for extra messages.

[dstrubbe (David Strubbe)]

I'm not sure what the "problems with vecLib framework used from Fortran" are supposed to be, and I have not heard of dotwrp before. For the hpl port, things seem to work currently. What change are you suggesting?

[tenomoto (Takeshi Enomoto)]

If hpl calls BLAS/LAPACK from C, vecLibFort or dotwrp is not necessary. vecLibFort is more complete addressing all the missing interface to BLAS/LAPACK. See README of vecLibFort. If hpl calls BLAS/LAPACK (cdotc, cdotu, zdotc, zdotu, sdot, snrm2, sasum, scnrm2, scasum) from Fortran, you need vecLbiFort.

[dstrubbe (David Strubbe)]

Ok I understand. hpl calls BLAS and LAPACK from Fortran, but only uses real double-precision mathematics, so vecLibFort does not appear necessary in this case.

[tenomoto (Takeshi Enomoto)]

Done in r125857.