Execute LBM scenario with OpenCL

It compiles so it's correct I guess... 

lbmpy/lbstep.py

@@ -18,7 +18,7 @@ from pystencils.timeloop import TimeLoop
 class LatticeBoltzmannStep:
 
     def __init__(self, domain_size=None, lbm_kernel=None, periodicity=False,
-                 kernel_params=MappingProxyType({}), data_handling=None, name="lbm", optimization=None,
+                 kernel_params=MappingProxyType({}), data_handling=None, name="lbm", optimization={},
                  velocity_data_name=None, density_data_name=None, density_data_index=None,
                  compute_velocity_in_every_step=False, compute_density_in_every_step=False,
                  velocity_input_array_name=None, time_step_order='stream_collide', flag_interface=None,
@@ -29,11 +29,15 @@ class LatticeBoltzmannStep:
             if domain_size is not None:
                 raise ValueError("When passing a data_handling, the domain_size parameter can not be specified")
 
+        target = optimization.get('target', 'cpu')
         if data_handling is None:
             if domain_size is None:
                 raise ValueError("Specify either domain_size or data_handling")
-            data_handling = create_data_handling(domain_size, default_ghost_layers=1,
-                                                 periodicity=periodicity, parallel=False)
+            data_handling = create_data_handling(domain_size,
+                                                 default_ghost_layers=1,
+                                                 periodicity=periodicity,
+                                                 default_target=target,
+                                                 parallel=False)
 
         if 'stencil' not in method_parameters:
             method_parameters['stencil'] = 'D2Q9' if data_handling.dim == 2 else 'D3Q27'
@@ -47,7 +51,6 @@ class LatticeBoltzmannStep:
             q = len(lbm_kernel.method.stencil)
         else:
             q = len(get_stencil(method_parameters['stencil']))
-        target = optimization['target']
 
         self.name = name
         self._data_handling = data_handling
@@ -58,7 +61,7 @@ class LatticeBoltzmannStep:
         self.density_data_index = density_data_index
         self._optimization = optimization
 
-        self._gpu = target == 'gpu'
+        self._gpu = target == 'gpu' or target == 'opencl'
         layout = optimization['field_layout']
 
         alignment = False

lbmpy/scenarios.py

@@ -85,7 +85,13 @@ def create_lid_driven_cavity(domain_size=None, lid_velocity=0.005, lbm_kernel=No
     """
     assert domain_size is not None or data_handling is not None
     if data_handling is None:
-        data_handling = create_data_handling(domain_size, periodicity=False, default_ghost_layers=1, parallel=parallel)
+        optimization = kwargs.get('optimization', None)
+        target = optimization.get('target', None) if optimization else None
+        data_handling = create_data_handling(domain_size,
+                                             periodicity=False,
+                                             default_ghost_layers=1,
+                                             parallel=parallel,
+                                             default_target=target)
     step = LatticeBoltzmannStep(data_handling=data_handling, lbm_kernel=lbm_kernel, name="ldc", **kwargs)
 
     my_ubb = UBB(velocity=[lid_velocity, 0, 0][:step.method.dim])
@@ -99,7 +105,7 @@ def create_lid_driven_cavity(domain_size=None, lid_velocity=0.005, lbm_kernel=No
 def create_channel(domain_size=None, force=None, pressure_difference=None, u_max=None, diameter_callback=None,
                    duct=False, wall_boundary=NoSlip(), parallel=False, data_handling=None, **kwargs):
     """Create a channel scenario (2D or 3D).
-    
+
     The channel can be driven either by force, velocity inflow or pressure difference. Choose one and pass
     exactly one of the parameters 'force', 'pressure_difference' or 'u_max'.
 

lbmpy_tests/test_lbstep.py

@@ -48,6 +48,35 @@ def test_data_handling_3d():
         np.testing.assert_almost_equal(results[0], arr)
 
 
+def test_data_handling_2d_opencl():
+    pytest.importorskip('pyopencl')
+    import pystencils.opencl.opencljit
+    pystencils.opencl.opencljit.init_globally()
+    print("--- LDC 2D test ---")
+    results = []
+    for parallel in [True, False] if parallel_available else [False]:
+        for gpu in [True, False] if gpu_available else [False]:
+            if parallel and gpu and not hasattr(wLB, 'cuda'):
+                continue
+
+            print("Testing parallel: %s\tgpu: %s" % (parallel, gpu))
+            opt_params = {'target': 'opencl' if gpu else 'cpu',
+                          'gpu_indexing_params': {'block_size': (8, 4, 2)}}
+            if parallel:
+                from pystencils.datahandling import ParallelDataHandling
+                blocks = wLB.createUniformBlockGrid(blocks=(2, 3, 1), cellsPerBlock=(5, 5, 1),
+                                                    oneBlockPerProcess=False)
+                dh = ParallelDataHandling(blocks, dim=2)
+                rho = ldc_setup(data_handling=dh, optimization=opt_params)
+                results.append(rho)
+            else:
+                rho = ldc_setup(domain_size=(10, 15), parallel=False, optimization=opt_params)
+                results.append(rho)
+    for i, arr in enumerate(results[1:]):
+        print("Testing equivalence version 0 with version %d" % (i + 1,))
+        np.testing.assert_almost_equal(results[0], arr)
+
+
 def test_data_handling_2d():
     print("--- LDC 2D test ---")
     results = []