Update fft_cuda.h

include/matx/transforms/fft/fft_cuda.h

@@ -216,7 +216,7 @@ template <typename OutTensorType, typename InTensorType> class matxCUDAFFTPlan_t
       if (i.IsContiguous() && o.IsContiguous()) {
         // Previously we used cudaMemGetInfo to get free memory to determine batch size. This can be very slow, 
         // and for small FFTs this call can create extra latency. For now we'll just assume the user knows what 
-        // they're doing and not try to batch FFTs that are too small        
+        // they're doing and not try to batch FFTs that are too large        
         const auto shape = i.Shape();
         params.batch = std::accumulate(std::begin(shape), std::end(shape) - 1, 1, std::multiplies<index_t>());
         params.batch_dims = i.Rank() - 1;