smyalygames/FiniteVolumeGPU
1
0
Fork 0
mirror of https://github.com/smyalygames/FiniteVolumeGPU.git synced 2026-01-14 15:48:43 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Fixed order again

Browse Source
This commit is contained in:
André R. Brodtkorb
2018-11-15 16:47:13 +01:00
parent dcb849b705
commit 7592ad5b9f
22 changed files with 758 additions and 619 deletions
Download Patch File Download Diff File Expand all files Collapse all files

76
GPUSimulators/WAF.py
View File

@@ -46,71 +46,65 @@ class WAF (Simulator.BaseSimulator):
dt: Size of each timestep (90 s)
g: Gravitational accelleration (9.81 m/s^2)
"""
def __init__(self, \
context, \
h0, hu0, hv0, \
nx, ny, \
dx, dy, dt, \
g, \
order=2, \
boundary_conditions=BoundaryCondition(), \
def __init__(self,
context,
h0, hu0, hv0,
nx, ny,
dx, dy, dt,
g,
boundary_conditions=BoundaryCondition(),
block_width=16, block_height=16):
# Call super constructor
super().__init__(context, \
nx, ny, \
dx, dy, dt, \
super().__init__(context,
nx, ny,
dx, dy, dt*2,
block_width, block_height);
self.g = np.float32(g)
self.order = np.int32(order)
self.boundary_conditions = boundary_conditions.asCodedInt()
#Get kernels
self.kernel = context.get_prepared_kernel("cuda/SWE2D_WAF.cu", "WAFKernel", \
"iiffffiiPiPiPiPiPiPi", \
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("iiffffiiPiPiPiPiPiPi")
#Create data by uploading to device
self.u0 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
self.u0 = Common.ArakawaA2D(self.stream,
nx, ny,
2, 2,
[h0, hu0, hv0])
self.u1 = Common.ArakawaA2D(self.stream, \
nx, ny, \
2, 2, \
self.u1 = Common.ArakawaA2D(self.stream,
nx, ny,
2, 2,
[None, None, None])
def step(self, dt):
if (self.order == 1):
self.substepDimsplit(dt, substep=(self.nt % 2))
elif (self.order == 2):
self.substepDimsplit(dt, substep=0)
self.substepDimsplit(dt, substep=1)
else:
raise(NotImplementedError("Order {:d} is not implemented".format(self.order)))
self.substepDimsplit(dt*0.5, substep=0)
self.substepDimsplit(dt*0.5, substep=1)
self.t += dt
self.nt += 1
self.nt += 2
def substepDimsplit(self, dt, substep):
self.kernel.prepared_async_call(self.grid_size, self.block_size, self.stream, \
self.nx, self.ny, \
self.dx, self.dy, dt, \
self.g, \
Simulator.stepOrderToCodedInt(step=substep, order=self.order), \
self.boundary_conditions, \
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.kernel.prepared_async_call(self.grid_size, self.block_size, self.stream,
self.nx, self.ny,
self.dx, self.dy, dt,
self.g,
substep,
self.boundary_conditions,
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