While running the program to add two simple matrix I get following error
CUDA runtime error: CUDA driver version is insufficient for CUDA runtime version (35) in magma_getdevice at interface_cuda/interface.cpp:547 Error: magmablasGetQueue requires magma_init() to be called first for MAGMA v1 compatability.
I have checked the other programs like sparse matrix calculation or eigen values functions are working but getting problem in Adding two matrices. I have also used magma_init() mentioned in error. Using following function.
magma_smalloc_cpu(&patterns, patternn2);
magma_smalloc_cpu(&onesmat, patterncols);
magma_smalloc_cpu(&col_mean, patternrows);
magma_smalloc_cpu(&meanmultiplier, patternn2);
// Values are assigned to each matrix and checked using loop its working
magmablas_sgeadd2(patternrows,patterncols,alpha,patterns,patternrows,beta,meanmultiplier,patternrows);
I am usng Cuda 7.5 version with magma 2.0.
If any one have idea about the error please answer, it would be really helpful for me to solve it.
Please read the error provided. Error messages were created to enable the compiler to complain about the error and potentially provide a possible solution.
CUDA driver version is insufficient for CUDA runtime version
That's it. According to this message, our CUDA driver is outdated and (here's a hidden potential solution) you should upgrade it.
Why? Because CUDA runtime uses some features of the newer driver, that the one you have doesn't support.