Added automatic AVX support - speedup +20% on CPU x86_64 Intel Skylake

AlexeyAB

2018-04-06 2a9f7e44ce1b73d3d56ef83f83e94f074ecac3f9

 src/cuda.c

@@ -96,6 +96,7 @@
    int i = cuda_get_device();
    if(!init[i]) {
        cublasCreate(&handle[i]);
      cublasStatus_t status = cublasSetStream(handle[i], get_cuda_stream());
        init[i] = 1;
    }
    return handle[i];