Nim bindings for the CUDA libraries. The versions currently in use are

• CUDA 8.0
• CuDNN 5.1

Most libraries have working bindings. Out of these:

• most bindings are generated using c2nim and suitable directives (see the files inside /c2nim)
• cuda_runtime_api.h and vector_types.h were not parsable directly and have required some manual adjustment of the header files
• driver_types.h is parseable, but some adjustments have been made in the resulting nim files.

Ideally, once some improvements are available in c2nim, there should be no need to manually modify any files.

See a few examples under /examples. The examples can be run with the command nimble EXAMPLE_NAME, where EXAMPLE_NAME is one of the examples - for instance nimble fft.

API documentation lives under /htmldocs.

c2nim supports name mangling, which could be useful to simplify a few names (e.g. turn CUBLAS_STATUS_ARCH_MISMATCH into ARCH_MISMATCH, which can be qualified as cublasStatus_t.ARCH_MISMATCH in case of ambiguity).

Right now, no mangling is performed, because the API surface is large and not always consistent, so it felt simpler to leave it as is. This may change in a future release.

Under nimcuda/nimcuda there are a few templates that turn CUDA errors into Nim exceptions. They are all under the overloaded name check, so that one can do, for instance

check cudaMalloc(cast[ptr pointer](addr gpuRows), sizeof(rows))

(here cudaMalloc is one of the many functions that fail by returning an error code).