smyalygames/FiniteVolumeGPU
1
0
Fork 0
mirror of https://github.com/smyalygames/FiniteVolumeGPU.git synced 2025-11-29 17:28:03 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Refactoring CudaArray and ArakawaA grid

Browse Source
This commit is contained in:
André R. Brodtkorb
2018-08-23 20:44:49 +02:00
parent 5668e28f99
commit 918d22b257
10 changed files with 452 additions and 159 deletions
Download Patch File Download Diff File Expand all files Collapse all files

58
GPUSimulators/Common.py
View File

@@ -84,7 +84,6 @@ class CudaContext(object):
self.cuda_device = cuda.Device(0)
self.logger.info("Using '%s' GPU", self.cuda_device.name())
self.logger.debug(" => compute capability: %s", str(self.cuda_device.compute_capability()))
self.logger.debug(" => memory: %d MB", self.cuda_device.total_memory() / (1024*1024))
# Create the CUDA context
if (self.blocking):
@@ -92,6 +91,9 @@ class CudaContext(object):
self.logger.warning("Using blocking context")
else:
self.cuda_context = self.cuda_device.make_context(flags=cuda.ctx_flags.SCHED_AUTO)
free, total = cuda.mem_get_info()
self.logger.debug(" => memory: %d / %d MB available", int(free/(1024*1024)), int(total/(1024*1024)))
self.logger.info("Created context handle <%s>", str(self.cuda_context.handle))
@@ -294,7 +296,7 @@ class CudaArray2D:
"""
Uploads initial data to the CL device
"""
def __init__(self, stream, nx, ny, x_halo, y_halo, cpu_data):
def __init__(self, stream, nx, ny, x_halo, y_halo, cpu_data=None, dtype=np.float32):
self.logger = logging.getLogger(__name__)
self.nx = nx
self.ny = ny
@@ -307,16 +309,18 @@ class CudaArray2D:
#self.logger.debug("Allocating [%dx%d] buffer", self.nx, self.ny)
#Make sure data is in proper format
assert np.issubdtype(cpu_data.dtype, np.float32), "Wrong datatype: %s" % str(cpu_data.dtype)
assert cpu_data.itemsize == 4, "Wrong size of data type"
assert not np.isfortran(cpu_data), "Wrong datatype (Fortran, expected C)"
if cpu_data is not None:
assert cpu_data.itemsize == 4, "Wrong size of data type"
assert not np.isfortran(cpu_data), "Wrong datatype (Fortran, expected C)"
#Upload data to the device
if (cpu_data.shape == (ny_halo, nx_halo)):
if cpu_data is None:
self.data = pycuda.gpuarray.empty((ny_halo, nx_halo), dtype)
elif (cpu_data.shape == (ny_halo, nx_halo)):
self.data = pycuda.gpuarray.to_gpu_async(cpu_data, stream=stream)
elif (cpu_data.shape == (self.ny, self.nx)):
#Should perhaps use pycuda.driver.mem_alloc_data.pitch() here
self.data = pycuda.gpuarray.empty((ny_halo, nx_halo), cpu_data.dtype)
self.data = pycuda.gpuarray.empty((ny_halo, nx_halo), dtype)
#self.data.fill(0.0)
#Create copy object from host to device
@@ -337,7 +341,6 @@ class CudaArray2D:
#Perform the copy
copy(stream)
stream.synchronize()
else:
assert False, "Wrong data shape: %s vs %s / %s" % (str(cpu_data.shape), str((self.ny, self.nx)), str((ny_halo, nx_halo)))
@@ -390,36 +393,31 @@ class CudaArray2D:
"""
A class representing an Arakawa A type (unstaggered, logically Cartesian) grid
"""
class SWEDataArakawaA:
class ArakawaA2D:
"""
Uploads initial data to the CL device
"""
def __init__(self, stream, nx, ny, halo_x, halo_y, h0, hu0, hv0):
def __init__(self, stream, nx, ny, halo_x, halo_y, cpu_variables):
self.logger = logging.getLogger(__name__)
self.h0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, h0)
self.hu0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hu0)
self.hv0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hv0)
self.h1 = CudaArray2D(stream, nx, ny, halo_x, halo_y, h0)
self.hu1 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hu0)
self.hv1 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hv0)
"""
Swaps the variables after a timestep has been completed
"""
def swap(self):
self.h1, self.h0 = self.h0, self.h1
self.hu1, self.hu0 = self.hu0, self.hu1
self.hv1, self.hv0 = self.hv0, self.hv1
self.gpu_variables = []
for cpu_variable in cpu_variables:
self.gpu_variables += [CudaArray2D(stream, nx, ny, halo_x, halo_y, cpu_variable)]
def __getitem__(self, key):
assert type(key) == int, "Indexing is int based"
if (key > len(self.gpu_variables) or key < 0):
raise IndexError("Out of bounds")
return self.gpu_variables[key]
"""
Enables downloading data from CL device to Python
"""
def download(self, stream):
h_cpu = self.h0.download(stream, async=True)
hu_cpu = self.hu0.download(stream, async=True)
hv_cpu = self.hv0.download(stream, async=False)
return h_cpu, hu_cpu, hv_cpu
cpu_variables = []
for gpu_variable in self.gpu_variables:
cpu_variables += [gpu_variable.download(stream, async=True)]
stream.synchronize()
return cpu_variables

31
GPUSimulators/FORCE.py
View File

@@ -21,7 +21,7 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
"""
#Import packages we need
from GPUSimulators import Simulator
from GPUSimulators import Simulator, Common
@@ -60,9 +60,7 @@ class FORCE (Simulator.BaseSimulator):
# Call super constructor
super().__init__(context, \
h0, hu0, hv0, \
nx, ny, \
1, 1, \
dx, dy, dt, \
g, \
block_width, block_height);
@@ -73,6 +71,16 @@ class FORCE (Simulator.BaseSimulator):
BLOCK_WIDTH=self.local_size[0], \
BLOCK_HEIGHT=self.local_size[1])
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
1, 1, \
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
1, 1, \
[None, None, None])
def __str__(self):
return "First order centered"
@@ -84,13 +92,14 @@ class FORCE (Simulator.BaseSimulator):
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def download(self):
return self.u0.download(self.stream)

31
GPUSimulators/HLL.py
View File

@@ -20,7 +20,7 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
"""
#Import packages we need
from GPUSimulators import Simulator
from GPUSimulators import Simulator, Common
@@ -55,9 +55,7 @@ class HLL (Simulator.BaseSimulator):
# Call super constructor
super().__init__(context, \
h0, hu0, hv0, \
nx, ny, \
1, 1, \
dx, dy, dt, \
g, \
block_width, block_height);
@@ -68,6 +66,16 @@ class HLL (Simulator.BaseSimulator):
BLOCK_WIDTH=self.local_size[0], \
BLOCK_HEIGHT=self.local_size[1])
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
1, 1, \
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
1, 1, \
[None, None, None])
def __str__(self):
return "Harten-Lax-van Leer"
@@ -79,13 +87,14 @@ class HLL (Simulator.BaseSimulator):
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def download(self):
return self.u0.download(self.stream)

46
GPUSimulators/HLL2.py
View File

@@ -21,7 +21,7 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
#Import packages we need
import numpy as np
from GPUSimulators import Simulator
from GPUSimulators import Simulator, Common
@@ -59,9 +59,7 @@ class HLL2 (Simulator.BaseSimulator):
# Call super constructor
super().__init__(context, \
h0, hu0, hv0, \
nx, ny, \
2, 2, \
dx, dy, dt, \
g, \
block_width, block_height);
@@ -74,6 +72,16 @@ class HLL2 (Simulator.BaseSimulator):
BLOCK_WIDTH=self.local_size[0], \
BLOCK_HEIGHT=self.local_size[1])
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[None, None, None])
def __str__(self):
return "Harten-Lax-van Leer (2nd order)"
@@ -90,13 +98,13 @@ class HLL2 (Simulator.BaseSimulator):
self.g, \
self.theta, \
np.int32(0), \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def stepDimsplitYX(self, dt):
@@ -106,12 +114,14 @@ class HLL2 (Simulator.BaseSimulator):
self.g, \
self.theta, \
np.int32(1), \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def download(self):
return self.u0.download(self.stream)

31
GPUSimulators/KP07.py
View File

@@ -26,7 +26,7 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
#Import packages we need
import numpy as np
from GPUSimulators import Simulator
from GPUSimulators import Simulator, Common
@@ -60,9 +60,7 @@ class KP07 (Simulator.BaseSimulator):
# Call super constructor
super().__init__(context, \
h0, hu0, hv0, \
nx, ny, \
2, 2, \
dx, dy, dt, \
g, \
block_width, block_height);
@@ -75,6 +73,16 @@ class KP07 (Simulator.BaseSimulator):
BLOCK_WIDTH=self.local_size[0], \
BLOCK_HEIGHT=self.local_size[1])
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[None, None, None])
def __str__(self):
return "Kurganov-Petrova 2007"
@@ -88,13 +96,13 @@ class KP07 (Simulator.BaseSimulator):
self.g, \
self.theta, \
np.int32(substep), \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
def stepEuler(self, dt):
self.substepRK(dt, 0)
@@ -108,5 +116,4 @@ class KP07 (Simulator.BaseSimulator):
self.t += dt
def download(self):
return self.data.download(self.stream)
return self.u0.download(self.stream)

45
GPUSimulators/KP07_dimsplit.py
View File

@@ -26,7 +26,7 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
#Import packages we need
import numpy as np
from GPUSimulators import Simulator
from GPUSimulators import Simulator, Common
@@ -60,9 +60,7 @@ class KP07_dimsplit (Simulator.BaseSimulator):
# Call super constructor
super().__init__(context, \
h0, hu0, hv0, \
nx, ny, \
2, 2, \
dx, dy, dt, \
g, \
block_width, block_height);
@@ -75,6 +73,16 @@ class KP07_dimsplit (Simulator.BaseSimulator):
BLOCK_WIDTH=self.local_size[0], \
BLOCK_HEIGHT=self.local_size[1])
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[None, None, None])
def __str__(self):
return "Kurganov-Petrova 2007 dimensionally split"
@@ -91,13 +99,13 @@ class KP07_dimsplit (Simulator.BaseSimulator):
self.g, \
self.theta, \
np.int32(0), \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def stepDimsplitYX(self, dt):
@@ -107,13 +115,14 @@ class KP07_dimsplit (Simulator.BaseSimulator):
self.g, \
self.theta, \
np.int32(1), \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def download(self):
return self.u0.download(self.stream)

31
GPUSimulators/LxF.py
View File

@@ -21,7 +21,7 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
"""
#Import packages we need
from GPUSimulators import Simulator
from GPUSimulators import Simulator, Common
@@ -56,9 +56,7 @@ class LxF (Simulator.BaseSimulator):
# Call super constructor
super().__init__(context, \
h0, hu0, hv0, \
nx, ny, \
1, 1, \
dx, dy, dt, \
g, \
block_width, block_height);
@@ -68,6 +66,16 @@ class LxF (Simulator.BaseSimulator):
"iiffffPiPiPiPiPiPi", \
BLOCK_WIDTH=self.local_size[0], \
BLOCK_HEIGHT=self.local_size[1])
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
1, 1, \
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
1, 1, \
[None, None, None])
def __str__(self):
return "Lax Friedrichs"
@@ -80,13 +88,14 @@ class LxF (Simulator.BaseSimulator):
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def download(self):
return self.u0.download(self.stream)

14
GPUSimulators/Simulator.py
View File

@@ -48,9 +48,7 @@ class BaseSimulator:
"""
def __init__(self, \
context, \
h0, hu0, hv0, \
nx, ny, \
ghost_cells_x, ghost_cells_y, \
dx, dy, dt, \
g, \
block_width, block_height):
@@ -67,14 +65,6 @@ class BaseSimulator:
#Create a CUDA stream
self.stream = cuda.Stream()
#Create data by uploading to device
free, total = cuda.mem_get_info()
self.logger.debug("GPU memory: %d / %d MB available", int(free/(1024*1024)), int(total/(1024*1024)))
self.data = Common.SWEDataArakawaA(self.stream, \
nx, ny, \
ghost_cells_x, ghost_cells_y, \
h0, hu0, hv0)
#Save input parameters
#Notice that we need to specify them in the correct dataformat for the
@@ -94,7 +84,7 @@ class BaseSimulator:
self.global_size = ( \
int(np.ceil(self.nx / float(self.local_size[0]))), \
int(np.ceil(self.ny / float(self.local_size[1]))) \
)
)
"""
Function which simulates forward in time using the default simulation type
@@ -192,7 +182,7 @@ class BaseSimulator:
return self.t
def download(self):
return self.data.download(self.stream)
raise(NotImplementedError("Needs to be implemented in subclass"))
def synchronize(self):
self.stream.synchronize()

61
GPUSimulators/WAF.py
View File

@@ -22,7 +22,7 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
#Import packages we need
import numpy as np
from GPUSimulators import Simulator
from GPUSimulators import Simulator, Common
@@ -55,9 +55,7 @@ class WAF (Simulator.BaseSimulator):
# Call super constructor
super().__init__(context, \
h0, hu0, hv0, \
nx, ny, \
2, 2, \
dx, dy, dt, \
g, \
block_width, block_height);
@@ -68,6 +66,16 @@ class WAF (Simulator.BaseSimulator):
BLOCK_WIDTH=self.local_size[0], \
BLOCK_HEIGHT=self.local_size[1])
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
[None, None, None])
def __str__(self):
return "Weighted average flux"
@@ -79,30 +87,33 @@ class WAF (Simulator.BaseSimulator):
def stepDimsplitXY(self, dt):
self.kernel.prepared_async_call(self.global_size, self.local_size, self.stream, \
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
np.int32(0), \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
np.int32(0), \
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def stepDimsplitYX(self, dt):
self.kernel.prepared_async_call(self.global_size, self.local_size, self.stream, \
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
np.int32(1), \
self.data.h0.data.gpudata, self.data.h0.data.strides[0], \
self.data.hu0.data.gpudata, self.data.hu0.data.strides[0], \
self.data.hv0.data.gpudata, self.data.hv0.data.strides[0], \
self.data.h1.data.gpudata, self.data.h1.data.strides[0], \
self.data.hu1.data.gpudata, self.data.hu1.data.strides[0], \
self.data.hv1.data.gpudata, self.data.hv1.data.strides[0])
self.data.swap()
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
np.int32(1), \
self.u0[0].data.gpudata, self.u0[0].data.strides[0], \
self.u0[1].data.gpudata, self.u0[1].data.strides[0], \
self.u0[2].data.gpudata, self.u0[2].data.strides[0], \
self.u1[0].data.gpudata, self.u1[0].data.strides[0], \
self.u1[1].data.gpudata, self.u1[1].data.strides[0], \
self.u1[2].data.gpudata, self.u1[2].data.strides[0])
self.u0, self.u1 = self.u1, self.u0
self.t += dt
def download(self):
return self.u0.download(self.stream)