I'm installing PETSc for complex numbers with the C99 standard. I'd like to have it installed using a higher precision than the default (double, I presume) since that will most likely make the LU preconditioner run better. I've tried several different commands that I could google, but it doesn't work. I've searched the manual but haven't found anything on the subject.

I tried invoking ./configure with the








in all four different combinations. The one time I used download-f2cblaslapack and precision=__float128 it did configure successfully but at the next stage when I compiled the libraries I got the following error for every library being compiled:

error: unknown type name ‘PetscScalar’

What do I have to write to have it configured at a higher precision?

  • 1
    $\begingroup$ this sounds like a bug in PETSc itself. Can you send the complete configure and make/test logs to petsc-maint@mcs.anl.gov for the case where you specified: --with-precision=__float128 --download-f2cblaslapack (as suggested by the PETSc documentation here). $\endgroup$ – Aron Ahmadia Aug 2 '12 at 12:57

Browse other questions tagged or ask your own question.