def __init__(self,
context,
rho,
rho_u,
rho_v,
E,
nx,
ny,
dx,
dy,
g,
gamma,
theta=1.3,
cfl_scale=0.9,
boundary_conditions=BoundaryCondition(),
block_width=16,
block_height=8):
# Call super constructor
super().__init__(context, nx, ny, dx, dy, boundary_conditions,
cfl_scale, 2, block_width, block_height)
self.g = np.float32(g)
self.gamma = np.float32(gamma)
self.theta = np.float32(theta)
#Get kernels
module = context.get_module("cuda/EE2D_KP07_dimsplit.cu",
defines={
'BLOCK_WIDTH': self.block_size[0],
'BLOCK_HEIGHT': self.block_size[1]
},
compile_args={
'no_extern_c': True,
'options': ["--use_fast_math"],
},
jit_compile_args={})
self.kernel = module.get_function("KP07DimsplitKernel")
self.kernel.prepare("iiffffffiiPiPiPiPiPiPiPiPiP")
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, nx, ny, 2, 2,
[rho, rho_u, rho_v, E])
self.u1 = Common.ArakawaA2D(self.stream, nx, ny, 2, 2,
[None, None, None, None])
self.cfl_data = gpuarray.GPUArray(self.grid_size, dtype=np.float32)
dt_x = np.min(self.dx / (np.abs(rho_u / rho) + np.sqrt(gamma * rho)))
dt_y = np.min(self.dy / (np.abs(rho_v / rho) + np.sqrt(gamma * rho)))
self.dt = min(dt_x, dt_y)
self.cfl_data.fill(self.dt, stream=self.stream)
def __init__(self,
context,
h0,
hu0,
hv0,
nx,
ny,
dx,
dy,
g,
cfl_scale=0.9,
boundary_conditions=BoundaryCondition(),
block_width=16,
block_height=16):
# Call super constructor
super().__init__(context, nx, ny, dx, dy, boundary_conditions,
cfl_scale, 2, block_width, block_height)
self.g = np.float32(g)
#Get kernels
module = context.get_module("cuda/SWE2D_WAF.cu",
defines={
'BLOCK_WIDTH': self.block_size[0],
'BLOCK_HEIGHT': self.block_size[1]
},
compile_args={
'no_extern_c': True,
'options': ["--use_fast_math"],
},
jit_compile_args={})
self.kernel = module.get_function("WAFKernel")
self.kernel.prepare("iiffffiiPiPiPiPiPiPiP")
#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])
self.cfl_data = gpuarray.GPUArray(self.grid_size, dtype=np.float32)
dt_x = np.min(self.dx / (np.abs(hu0 / h0) + np.sqrt(g * h0)))
dt_y = np.min(self.dy / (np.abs(hv0 / h0) + np.sqrt(g * h0)))
dt = min(dt_x, dt_y)
self.cfl_data.fill(dt, stream=self.stream)