PGI community edition 16.10 and 1070 GTX

Can I use CUDA fortran programs using the PGI community edition 16.10 on a 1070 GTX?

I am aware it only supports single precision.

Hi laurence_jj,

Officially PGI only supports the Tesla product line of NVIDIA GPU. However in general, the GTX product line can be used provided that the target architecture is also a supported Tesla architecture. We have many users that use GTX for development before moving to Tesla for production.

GTX 1070 uses the “Pascal” architecture. Hence to compile for this target, add the CUDA Fortran flag “-Mcuda=cc60” or for OpenACC use “-ta=tesla:cc60”.

Hope this helps,