add transport solver

transport_cl.py

@@ -17,8 +17,8 @@ _m = 1
 _wbord = 0.
 
 # grid size
-_nx = 512
-_ny = 512
+_nx = 1024
+_ny = 1024
 Lx = 1.
 Ly = 1.
 
@@ -110,14 +110,18 @@ def solve_ocl(m = _m, nx = _nx, ny = _ny, Tmax = _Tmax, dx = _dx, dy = _dy,
 
     # OpenCL init
     ctx = cl.create_some_context()
-    queue = cl.CommandQueue(ctx, properties=cl.command_queue_properties.PROFILING_ENABLE)
     mf = cl.mem_flags
 
+    # compile OpenCL C program
+    prg = cl.Program(ctx, source).build()
+
     # create OpenCL buffers
     wn_gpu = cl.Buffer(ctx, mf.READ_WRITE, size=(nx * ny * np.dtype('float32').itemsize))
     wnp1_gpu = cl.Buffer(ctx, mf.READ_WRITE, size=(nx * ny * np.dtype('float32').itemsize))
 
-    prg = cl.Program(ctx, source).build()
+    # create a queue (for submitting opencl operations)
+    queue = cl.CommandQueue(ctx, properties=cl.command_queue_properties.PROFILING_ENABLE)
+
 
     # init data
     event = prg.init_sol(queue, (nx * ny, ), (32, ), wn_gpu)
@@ -125,11 +129,10 @@ def solve_ocl(m = _m, nx = _nx, ny = _ny, Tmax = _Tmax, dx = _dx, dy = _dy,
 
     # number of animation frames
     nbplots = 10
-
     itermax = int(np.floor(Tmax / dt))
     iterplot = int(itermax / nbplots)
 
-    # time loop on gpu
+    # time loop
     t = 0
     iter = 0
     elapsed = 0.;