Intel® LLVM-based C/C++ and Fortran compilers,

icx

,

icpx

, and

ifx

, support OpenMP* offloading onto GPUs. When using OpenMP, the programmer inserts device directives in the code to direct the compiler to offload certain parts of the application onto the GPU. Offloading compute-intensive code can yield better performance.

This section covers the following topics related to OpenMP offloading, and how to improve the performance of offloaded code.