WIP: Cuda autotune

This PR introduces two one changes:

• rotate (32,1,1) depending on field strides to fastest dimension. So (1,1,32) for c-layout and (32,1,1) for fortran layout. So pystencils will be fast also for c-layout (this will always be performed)
• auto-tune the block dimensions to whatevers is fastest for a specific kernel on localhost. On first kernel call different layouts are tried and the kernel will be called henceforth with the fastest configuration (disk_cached). This could be intersting for OpenCL where we don't know which launch config is the fastest (on OpenCL the runtime can alternatively give a hint on that).

One drawback: the test calls are only correct if input and output fields do not overlap (so no in-place kernels).

pystencils/gpucuda/cudajit.py

@@ -78,7 +78,6 @@ def make_python_function(kernel_function_node, argument_dict=None, custom_backen
         except KeyError:
             full_arguments = argument_dict.copy()
             full_arguments.update(kwargs)
-            example_strides = [v.strides for v in kwargs.values() if isinstance(v, pycuda.gpuarray.GPUArray)][0]
             shape = _check_arguments(parameters, full_arguments)
 
             # TODO: use texture objects:
@@ -97,10 +96,9 @@ def make_python_function(kernel_function_node, argument_dict=None, custom_backen
                                                       tuple((k, v.strides, v.shape)
                                                             for k, v in kwargs.items()
                                                             if (isinstance(v, pycuda.gpuarray.GPUArray)))
-                                                      + (str(pystencils.show_code(kernel_function_node)),),
-                                                      example_strides))
+                                                      + (str(pystencils.show_code(kernel_function_node)),)))
             else:
-                block_and_thread_numbers = indexing.call_parameters(shape, example_strides)
+                block_and_thread_numbers = indexing.call_parameters(shape)
                 block_and_thread_numbers['block'] = tuple(int(i) for i in block_and_thread_numbers['block'])
                 block_and_thread_numbers['grid'] = tuple(int(i) for i in block_and_thread_numbers['grid'])