def runTest(self):
nx, ny, nz, str_f, pt0, pt1, is_array = self.args
slices = common.slices_two_points(pt0, pt1)
str_fs = common.convert_to_tuple(str_f)
# instance
gpu_devices = common_gpu.gpu_device_list(print_info=False)
context = cl.Context(gpu_devices)
mainf_list = [gpu.Fields(context, device, nx, ny, nz) \
for device in gpu_devices]
mainf_list.append(cpu.Fields(nx, ny, nz))
nodef = Fields(mainf_list)
dtype = nodef.dtype
anx = nodef.accum_nx_list
getf = GetFields(nodef, str_f, (0, 0, 0),
(nodef.nx - 1, ny - 1, nz - 1))
setf = SetFields(nodef, str_f, pt0, pt1, is_array)
# generate random source
if is_array:
shape = common.shape_two_points(pt0, pt1, len(str_fs))
value = np.random.rand(*shape).astype(nodef.dtype)
split_value = np.split(value, len(str_fs))
split_value_dict = dict(zip(str_fs, split_value))
else:
value = np.random.ranf()
# host allocations
global_ehs = [np.zeros(nodef.ns, dtype) for i in range(6)]
eh_dict = dict(zip(['ex', 'ey', 'ez', 'hx', 'hy', 'hz'], global_ehs))
# verify
for str_f in str_fs:
if is_array:
eh_dict[str_f][slices] = split_value_dict[str_f]
else:
eh_dict[str_f][slices] = value
setf.set_fields(value)
gpu_getf = gpu.GetFields(mainf_list[0], str_fs, (0, 0, 0),
(nx - 1, ny - 1, nz - 1))
gpu_getf.get_event().wait()
getf.wait()
for str_f in str_fs:
original = eh_dict[str_f]
copy = getf.get_fields(str_f)
norm = np.linalg.norm(original - copy)
#if norm != 0:
#print '\ngpu getf\n', gpu_getf.get_fields(str_f)
#print original[slices]
#print copy[slices]
self.assertEqual(norm, 0, '%s, %g, %s' % (self.args, norm, str_f))
def runTest(self):
nx, ny, nz, str_f, pt0, pt1, is_array = self.args
slices = common.slices_two_points(pt0, pt1)
# generate random source
if is_array:
shape = common.shape_two_points(pt0, pt1)
value = np.random.rand(*shape).astype(np.float32)
else:
value = np.random.ranf()
# instance
gpu_devices = common_gpu.gpu_device_list(print_info=False)
context = cl.Context(gpu_devices)
mainf_list = [ cpu.Fields(nx, ny, nz) ]
mainf_list += [gpu.Fields(context, device, nx, ny, nz) \
for device in gpu_devices]
nodef = Fields(mainf_list)
dtype = nodef.dtype
anx = nodef.accum_nx_list
tfunc = lambda tstep: np.sin(0.03*tstep)
incident = IncidentDirect(nodef, str_f, pt0, pt1, tfunc, value)
# allocations for verify
eh = np.zeros(nodef.ns, dtype)
getf = GetFields(nodef, str_f, pt0, pt1)
# verify
eh[slices] = dtype(value) * dtype(tfunc(1))
e_or_h = str_f[0]
nodef.update_e()
nodef.update_h()
getf.wait()
original = eh[slices]
copy = getf.get_fields(str_f)
norm = np.linalg.norm(original - copy)
self.assertEqual(norm, 0, '%s, %g' % (self.args, norm))
def runTest(self):
nx, ny, nz, str_f, pt0, pt1 = self.args
slices = common.slices_two_points(pt0, pt1)
str_fs = common.convert_to_tuple(str_f)
# instance
gpu_devices = common_gpu.gpu_device_list(print_info=False)
context = cl.Context(gpu_devices)
mainf_list = [gpu.Fields(context, device, nx, ny, nz) \
for device in gpu_devices]
mainf_list.append(cpu.Fields(nx, ny, nz))
nodef = Fields(mainf_list)
dtype = nodef.dtype
anx = nodef.accum_nx_list
getf = GetFields(nodef, str_f, pt0, pt1)
# generate random source
global_ehs = [np.zeros(nodef.ns, dtype) for i in range(6)]
eh_dict = dict(zip(['ex', 'ey', 'ez', 'hx', 'hy', 'hz'], global_ehs))
for i, f in enumerate(mainf_list[:-1]):
nx, ny, nz = f.ns
ehs = common_random.generate_ehs(nx, ny, nz, dtype)
f.set_eh_bufs(*ehs)
for eh, geh in zip(ehs, global_ehs):
geh[anx[i]:anx[i + 1], :, :] = eh[:-1, :, :]
f = mainf_list[-1]
nx, ny, nz = f.ns
ehs = common_random.generate_ehs(nx, ny, nz, dtype)
f.set_ehs(*ehs)
for eh, geh in zip(ehs, global_ehs):
geh[anx[-2]:anx[-1] + 1, :, :] = eh[:]
# verify
getf.wait()
for str_f in str_fs:
original = eh_dict[str_f][slices]
copy = getf.get_fields(str_f)
norm = np.linalg.norm(original - copy)
self.assertEqual(norm, 0, '%s, %g' % (self.args, norm))
def runTest(self):
nx, ny, nz, str_f, pt0, pt1, is_array = self.args
slices = common.slices_two_points(pt0, pt1)
# generate random source
if is_array:
shape = common.shape_two_points(pt0, pt1)
value = np.random.rand(*shape).astype(np.float32)
else:
value = np.random.ranf()
# instance
gpu_devices = common_gpu.gpu_device_list(print_info=False)
context = cl.Context(gpu_devices)
mainf_list = [cpu.Fields(nx, ny, nz)]
mainf_list += [gpu.Fields(context, device, nx, ny, nz) \
for device in gpu_devices]
nodef = Fields(mainf_list)
dtype = nodef.dtype
anx = nodef.accum_nx_list
tfunc = lambda tstep: np.sin(0.03 * tstep)
incident = IncidentDirect(nodef, str_f, pt0, pt1, tfunc, value)
# allocations for verify
eh = np.zeros(nodef.ns, dtype)
getf = GetFields(nodef, str_f, pt0, pt1)
# verify
eh[slices] = dtype(value) * dtype(tfunc(1))
e_or_h = str_f[0]
nodef.update_e()
nodef.update_h()
getf.wait()
original = eh[slices]
copy = getf.get_fields(str_f)
norm = np.linalg.norm(original - copy)
self.assertEqual(norm, 0, '%s, %g' % (self.args, norm))
coeff_use = 'e'
precision_float = 'single'
# instances
gpu_devices = common_gpu.gpu_device_list(print_info=False)
context = cl.Context(gpu_devices)
device = gpu_devices[0]
gpuf = gpu.Fields(context, device, nx, ny, nz, coeff_use, precision_float)
buffer_dict = {}
if rank == 0:
buffer_dict['x+'] = BufferFields(gpuf, 'x+', tmax, ny, nz, coeff_use, precision_float)
elif rank == 1:
buffer_dict['x-'] = BufferFields(gpuf, 'x-', tmax, ny, nz, coeff_use, precision_float)
fields = Fields([gpuf], buffer_dict)
Core(fields)
Pbc(fields, 'yz')
if rank == 0:
print 'ns', fields.ns
print 'nbytes (MB)', nx*ny*nz * 9 * 4. / (1024**2)
tfunc = lambda tstep: 40 * np.sin(0.05 * tstep)
gpu.IncidentDirect(gpuf, 'ez', (220, 20, 0), (220, 20, -1), tfunc)
getf = gpu.GetFields(gpuf, 'ez', (0, 0, 0.5), (-1, -1, 0.5))
# plot
import matplotlib.pyplot as plt
plt.ion()
fig = plt.figure(figsize=(12,8))
from kemp.fdtd3d.node import Fields, Core, IncidentDirect, GetFields, Pbc from kemp.fdtd3d import gpu, cpu ny, nz = 140, 2 gpu_nx = 141 cpu_nx = 20 tmax, tgap = 150, 10 # instances gpu_devices = common_gpu.gpu_device_list(print_info=False) context = cl.Context(gpu_devices) mainf_list = [ cpu.Fields(cpu_nx, ny, nz) ] mainf_list += [gpu.Fields(context, device, gpu_nx, ny, nz) for device in gpu_devices] fields = Fields(mainf_list) Core(fields) Pbc(fields, 'xyz') nx = fields.nx tfunc = lambda tstep: np.sin(0.05 * tstep) #IncidentDirect(fields, 'ez', (20, 0, 0), (20, ny-1, nz-1), tfunc) IncidentDirect(fields, 'ez', (0, 20, 0), (nx-1, 20, nz-1), tfunc) getf = GetFields(fields, 'ez', (0, 0, nz/2), (nx-1, ny-1, nz/2)) #IncidentDirect(fields, 'ey', (20, 0, 0), (20, ny-1, nz-1), tfunc) #getf = GetFields(fields, 'ey', (0, 0, nz/2), (nx-1, ny-1, nz/2)) #IncidentDirect(fields, 'ex', (0, 20, 0), (nx-1, 20, nz-1), tfunc) #getf = GetFields(fields, 'ex', (0, 0, nz/2), (nx-1, ny-1, nz/2))
ny, nz = 140, 32
gpu_nx = 141
cpu_nx = 20
tmax, tgap = 150, 10
# instances
gpu_devices = common_gpu.gpu_device_list(print_info=False)
context = cl.Context(gpu_devices)
mainf_list = [
gpu.Fields(context, device, gpu_nx, ny, nz) for device in gpu_devices
]
mainf_list.append(cpu.Fields(cpu_nx, ny, nz))
#mainf_list = [ cpu.Fields(160, ny, nz) ]
fields = Fields(mainf_list, pbc='yz', mpi_shape=(2, 1, 1), tmax=tmax)
Core(fields)
nx = fields.nx
tfunc = lambda tstep: np.sin(0.05 * tstep)
#IncidentDirect(fields, 'ez', (20, 0, 0), (20, ny-1, nz-1), tfunc)
IncidentDirect(fields, 'ez', (0, 20, 0), (nx - 1, 20, nz - 1), tfunc)
getf = GetFields(fields, 'ez', (0, 0, nz / 2), (nx - 1, ny - 1, nz / 2))
if rank == 0:
buf = fields.buffer_dict['x+']
print 'buf instance_list', buf.instance_list
getf_buf = cpu.GetFields(buf, 'ez', (0, 0, nz / 2), (2, ny - 1, nz / 2))
#IncidentDirect(fields, 'ey', (20, 0, 0), (20, ny-1, nz-1), tfunc)
#getf = GetFields(fields, 'ey', (0, 0, nz/2), (nx-1, ny-1, nz/2))
precision_float = 'single'
# instances
gpu_devices = common_gpu.gpu_device_list(print_info=False)
context = cl.Context(gpu_devices)
device = gpu_devices[0]
gpuf = gpu.Fields(context, device, nx, ny, nz, coeff_use, precision_float)
buffer_dict = {}
if rank == 0:
buffer_dict['x+'] = BufferFields(gpuf, 'x+', tmax, ny, nz, coeff_use,
precision_float)
elif rank == 1:
buffer_dict['x-'] = BufferFields(gpuf, 'x-', tmax, ny, nz, coeff_use,
precision_float)
fields = Fields([gpuf], buffer_dict)
Core(fields)
Pbc(fields, 'yz')
if rank == 0:
print 'ns', fields.ns
print 'nbytes (MB)', nx * ny * nz * 9 * 4. / (1024**2)
tfunc = lambda tstep: 40 * np.sin(0.05 * tstep)
gpu.IncidentDirect(gpuf, 'ez', (220, 20, 0), (220, 20, -1), tfunc)
getf = gpu.GetFields(gpuf, 'ez', (0, 0, 0.5), (-1, -1, 0.5))
# plot
import matplotlib.pyplot as plt
plt.ion()
fig = plt.figure(figsize=(12, 8))