Could use better error handling. Would like to have defined error codes that are in the CUDA library, and pass those back to python where they would be handled. We sort of do this now, but the numbers are not systematic, or assigned meaningful strings. This might involve an enum type and string array in the library or something.
Would also want python tests to actually check we are catching and printing the same.
Could use better error handling. Would like to have defined error codes that are in the CUDA library, and pass those back to python where they would be handled. We sort of do this now, but the numbers are not systematic, or assigned meaningful strings. This might involve an enum type and string array in the library or something.
Would also want python tests to actually check we are catching and printing the same.