I am benchmarking some code that uses cudafy. I see that in GASS I can use cuFuncSetCachConfig to change from the default L1/Shared(16KB/48KB) to 48KB/16KB. I would like to use this in my bench marking. I am using a fermi class card. However as expected the
following code does not work with cudafied code
CUDA cuda = new CUDA(0, true);
CUfunction cuf = cuda.GetModuleFunction("Mul");
CUResult cur = CUDADriver.cuFuncSetCacheConfig(cuf, CUFunctionCache.PreferL1);
The method Mul cannot be found. Looking at the source for CUfunction it is simply an intptr containing the address of the function Mul. Is there any way that htis can be obtained in cudafy? or some alternative way in cudafy to set the L1/Shared memory sizes?
Thanks in advance