[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: petsc-dev recompile needed




Mathieu,

    I try to avoid I but cannot say for sure.

    With hg there is an easy way to generate a patch with your
changes and then email to petsc-maint@xxxxxxxxxxx then we simply
apply it to the master.

   Barry


On Wed, 28 Jun 2006, Mathieu Taillefumier wrote:

Hy,

do your modifications include variable name like I (upper i) in your
structure. My question seems to be strange but it motivated by the fact
that the code will not be compliant if you use a C compiler for complex
types (Sorry for that, I am not defining the norms). In fact, I almost
made all corrections in the code in order to compile complex version of
PETSC with gcc (icc works also), but I didn't send any modifications
just because of time. The number of modified files is little and mostly
it corresponds to give a different name to the variables. I will do this
job again on the last development version of the library and send the
modified files to the mailing list if I don't find any others ways.
A second point should be noticed. I use the library (complex type
compiled with C compiler) and it seems to work correctly. At least I
didn't have any difference in the results. However, i think that this
option should be used with caution.

Mathieu

  I have changed some of the data structures for the various AIJ matrix
formats. On you next pull you will need to recompile all the matrix
code.

  cd src/mat
  make ACTION=libfast tree

 Let me know if the behavior of the matrices changes due to me update
to the code.

   Barry

I'm starting to get the various AIJ matrices to shared more common data
structures and code.