about multiprecision

Dear Mat,

COuld OpenACC support the codes of multiprecision ?

our customer’s codes is multiprecision fortran…and because of time issue, they think it is possible to optmize them with CUDA fortran. They hope to try OpenACC.

Hi Teslalady,

Can you please clarify what is meant by “multiprecision”?

I think you mean double precision (i.e. 64-bit) floating point numbers, in which case, then yes, OpenACC does support double precision.

  • Mat

multiprecision means:4-times precision,8-times precision…OpenMP and CUDA support multiprecison…so I just want to know whether PGI OpenACC support it. I’m just studying the OpenACC

Sorry, but I still need more information on exactly what you’re looking for.

In searching, I do see multiple Multiprecision libraries which allow the users to use arbitrary precision arithmetic (For example: http://en.wikipedia.org/wiki/GNU_Multiple_Precision_Arithmetic_Library) and I do see that at least one is implemented using OpenMP. However, these are not part of the OpenMP language itself, but rather an external library. Is this what you’re looking for?

  • Mat