bug in 1d transport

simple_kernels.py

@@ -40,9 +40,9 @@ verif_cpu = np.fromfunction(lambda i: i * i + i, (taille, ), dtype = np.float32)
 
 prg = cl.Program(ctx, source).build()
 
-prg.K1(queue, (taille, ), None, x_gpu).wait()
+event = prg.K1(queue, (taille, ), None, x_gpu)
 prg.K2(queue, (taille, ), None, y_gpu).wait()
-prg.K3(queue, (taille, ), None, x_gpu, y_gpu, z_gpu).wait()
+prg.K3(queue, (taille, ), None , x_gpu, y_gpu, z_gpu, wait_for = [event]).wait()
 
 cl.enqueue_copy(queue, z_cpu, z_gpu)
 

testtranspose.py

@@ -5,23 +5,31 @@ from __future__ import absolute_import, print_function
 import pyopencl as cl
 import numpy as np
 
-
 with open('kernels.cl', 'r') as f:
     source = f.read()
-    
-ctx = cl.create_some_context()
 
+m = 2
+raf = 1
+deg = 1
+
+source = source.replace("THE_DEG", str(deg))
+source = source.replace("THE_RAF", str(raf))
+source = source.replace("THE_M", str(m))
+
+ctx = cl.create_some_context()
 
 prg = cl.Program(ctx, source).build()
 
-point = [ ( 0.0, 0.0, 0.0),
-          ( 1.0, 0.0, 0.0),
-          ( 1.0, 1.0, 0.0),
-          ( 0.0, 1.0, 0.0),
-          ( 0.0, 0.0, 1.0),
-          ( 1.0, 0.0, 1.0),
-          ( 1.0, 1.0, 1.0),
-          ( 0.0, 1.0, 1.0)]
+point_cpu = np.array([0.0, 0.0, 0.0,
+                  1.0, 0.0, 0.0,
+                  1.0, 1.0, 0.0,
+                  0.0, 1.0, 0.0,
+                  0.0, 0.0, 1.0,
+                  1.0, 0.0, 1.0,
+                  1.0, 1.0, 1.0,
+                  0.0, 1.0, 1.0], dtype = np.float32)
+
+
 
 element = [0,1,2,3,4,5,6,7]
 
@@ -32,4 +40,56 @@ face2node = [[0,1,5,4],
 	      [5,6,7,4],
 	      [0,3,2,1]]
 
+queue = cl.CommandQueue(ctx)
+mf = cl.mem_flags
+
+wsize = m * raf**3 * (deg+1)**3
+
+row_size = raf * (deg + 1)
+
+point_gpu = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf = point_cpu)
+wn_cpu = np.zeros(wsize, dtype = np.float32)
+wn_gpu = cl.Buffer(ctx, mf.WRITE_ONLY, size=(wsize * np.dtype('float32').itemsize))
+
+dtwn_cpu = np.zeros(wsize, dtype = np.float32)
+dtwn_gpu = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf = dtwn_cpu);
+
+ie = np.int32(0)
+
+prg.Init(queue, (row_size, row_size, row_size), None, ie, point_gpu, wn_gpu).wait()
+
+cl.enqueue_copy(queue, wn_cpu, wn_gpu)
+
+#print(wn_cpu)
+
+wn_out = wn_cpu.reshape((2*row_size**2,row_size))
+
+#print(wn_out)
+
+import matplotlib.pyplot as plt
+
+x_cpu = np.zeros(3 * row_size**3, dtype = np.float32)
+x_gpu = cl.Buffer(ctx, mf.WRITE_ONLY, x_cpu.nbytes)
+
+prg.get_nodes(queue, (row_size, row_size, row_size),
+              None, ie, point_gpu, x_gpu).wait()
+cl.enqueue_copy(queue, x_cpu, x_gpu).wait()
+
+plt.plot(x_cpu[0:row_size],wn_out[0])
+plt.plot(x_cpu[0:row_size],wn_out[1])
+
+#plt.show()
+#print(x_cpu[0:row_size])
+#print(x_cpu[0:row_size])
+
+tnow = np.float32(0)
+
+npts = deg+1
+
+prg.DGFlux(queue, (raf+1, row_size, row_size), (1, 1, 1), ie, point_gpu,tnow, wn_gpu,dtwn_gpu).wait()
+
+prg.DGVolume(queue, (row_size, row_size, row_size), (npts,npts,npts), ie, point_gpu, wn_gpu,dtwn_gpu).wait()
+
+cl.enqueue_copy(queue, dtwn_cpu, dtwn_gpu)
 
+print(dtwn_cpu)