CUDA Fortran provides native support for the NVIDIA CUDA architecture from within the Fortran programming language. CUDA Fortran is supported by both the IBM XL Fortran compiler and the NVIDIA Fortran compiler, available as part of the NVIDIA HPC SDK. NVIDIA CUDA Fortran includes a Fortran 2003 compiler and tool chain for programming NVIDIA GPUs using Fortran. NVIDIA CUDA Fortran is currently supported on Linux. Windows support is in process.
Fortran is a key programming language used by high performance computing developers. It is the language of choice in many application domains including computational fluid dynamics (including weather and ocean modeling), finite-element analysis, molecular dynamics, and quantum chemistry.
Many popular HPC packages were originally developed in Fortran and continue to be enhanced over time. A Fortran compiler allows developers to program in their native language (Fortran) to recompile these codes to take advantage of new performance and architectural features.
The NVIDIA CUDA Fortran Compiler provides Fortran language support for NVIDIA’s CUDA-enabled GPUs. Fortran developers with data parallel problems will be able to use this compiler to harness the massive parallel computing capability of NVIDIA GPUs to create high performance applications for scientific computing.
NVIDIA has a long history of embracing and supporting standards, as a wider choice of languages improves the number and scope of applications that can exploit parallel computing on the GPU. With support for the C, Java, and Python languages and DirectCompute APIs and now Fortran, the performance advantage of GPU Computing is accessible to mainstream developers. NVIDIA is the only processor company to offer this breadth of development environments for the GPU.
Apply to NVIDIA's CUDA Registered Developer Program for early access to new releases and more.
Fortran Code Samples
CUDA Fortran Documentation