Suggestion Description
cudaMemcpy2D accepts fortran pointers to arrays with ranks greater than 2., while hip equivalent does not.
Is this a conscious choice? I have code which uses memcpy2d on 3- and 4-d arrays. I can obviously rewrite, but was hoping to keep changes with respect to cuda version minimal.
Operating System
No response
GPU
No response
ROCm Component
No response
Suggestion Description
cudaMemcpy2D accepts fortran pointers to arrays with ranks greater than 2., while hip equivalent does not.
Is this a conscious choice? I have code which uses memcpy2d on 3- and 4-d arrays. I can obviously rewrite, but was hoping to keep changes with respect to cuda version minimal.
Operating System
No response
GPU
No response
ROCm Component
No response