Currently only double precision is used; I would like to make this flexible so that single precision can be used (allowing it to run on Intel Xe-based iGPUs that don't support 64-bit FP), or even quads in the case that very small nonlinear effects are to be studied
Currently only double precision is used; I would like to make this flexible so that single precision can be used (allowing it to run on Intel Xe-based iGPUs that don't support 64-bit FP), or even quads in the case that very small nonlinear effects are to be studied