Optimisation flags for CUDA?

I’ve looked on the man page for pgfortran and was wandering if any of the optimisation flags will work with CUDA or if they’re specifically for cpu code?

The only CUDA specific flags on the man page I could find were the options for -Mcuda such as those which set which compute capability you want to compile for, maxregcount and ones which allow you to keep the generated cuda binaries for example.

Hi Tom,

pgfortran will perform optimization on CUDA Fortran code where it make sense.

  • Mat

Cool thanks