diff options
Add basic AA pattern support for OpenCL example
Diffstat (limited to 'lid_driven_cavity/opencl')
-rw-r--r-- | lid_driven_cavity/opencl/AA.py | 105 | ||||
-rw-r--r-- | lid_driven_cavity/opencl/AB.py (renamed from lid_driven_cavity/opencl/simulation.py) | 31 | ||||
-rw-r--r-- | lid_driven_cavity/opencl/common.py | 36 | ||||
-rw-r--r-- | lid_driven_cavity/opencl/ldc_2d.py | 74 |
4 files changed, 190 insertions, 56 deletions
diff --git a/lid_driven_cavity/opencl/AA.py b/lid_driven_cavity/opencl/AA.py new file mode 100644 index 0000000..f317cd0 --- /dev/null +++ b/lid_driven_cavity/opencl/AA.py @@ -0,0 +1,105 @@ +import pyopencl as cl +mf = cl.mem_flags + +import numpy + +class Memory: + def __init__(self, descriptor, geometry, context, float_type): + self.context = context + self.float_type = float_type + + self.size_x = geometry.size_x + self.size_y = geometry.size_y + self.size_z = geometry.size_z + self.volume = self.size_x * self.size_y * self.size_z + + self.pop_size = descriptor.q * self.volume * self.float_type(0).nbytes + self.moments_size = 3 * self.volume * self.float_type(0).nbytes + + self.cl_pop = cl.Buffer(self.context, mf.READ_WRITE, size=self.pop_size) + + self.cl_moments = cl.Buffer(self.context, mf.WRITE_ONLY, size=self.moments_size) + + def gid(self, x, y, z = 0): + return z * (self.size_x*self.size_y) + y * self.size_x + x; + +class Lattice: + def __init__(self, geometry, kernel_src, descriptor, platform = 0, precision = 'single'): + self.geometry = geometry + self.descriptor = descriptor + + self.float_type = { + 'single': (numpy.float32, 'float'), + 'double': (numpy.float64, 'double'), + }.get(precision, None) + + self.platform = cl.get_platforms()[platform] + self.layout = None + + self.context = cl.Context( + properties=[(cl.context_properties.PLATFORM, self.platform)]) + + self.queue = cl.CommandQueue(self.context) + + self.memory = Memory(descriptor, self.geometry, self.context, self.float_type[0]) + self.tick = False + + self.compiler_args = { + 'single': '-cl-single-precision-constant -cl-fast-relaxed-math', + 'double': '-cl-fast-relaxed-math' + }.get(precision, None) + + self.build_kernel(kernel_src) + + self.program.equilibrilize_all( + self.queue, self.geometry.size(), self.layout, self.memory.cl_pop).wait() + + self.tick_tasks = [] + self.tock_tasks = [] + + def build_kernel(self, src): + self.program = cl.Program(self.context, src).build(self.compiler_args) + + def schedule_tick(self, f, cells, *params): + self.tick_tasks += [ (eval("self.program.%s" % f), cells, params) ] + + def schedule_tock(self, f, cells, *params): + self.tock_tasks += [ (eval("self.program.%s" % f), cells, params) ] + + def evolve(self): + if self.tick: + self.tick = False + for f, cells, params in self.tick_tasks: + f(self.queue, cells.size(), self.layout, self.memory.cl_pop, cells.get(), *params) + else: + self.tick = True + for f, cells, params in self.tock_tasks: + f(self.queue, cells.size(), self.layout, self.memory.cl_pop, cells.get(), *params) + + def sync(self): + self.queue.finish() + + def get_moments(self): + moments = numpy.ndarray(shape=(self.memory.volume*(self.descriptor.d+1),1), dtype=self.float_type[0]) + + self.program.collect_moments_all( + self.queue, self.geometry.size(), self.layout, self.memory.cl_pop, self.memory.cl_moments) + + cl.enqueue_copy(self.queue, moments, self.memory.cl_moments).wait(); + + return moments + +HelperTemplate = """ +__kernel void equilibrilize_all(__global ${float_type}* f_next) +{ + const unsigned int gid = ${index.gid('get_global_id(0)', 'get_global_id(1)')}; + equilibrilize_tick(f_next, gid); +} + +__kernel void collect_moments_all(__global ${float_type}* f, + __global ${float_type}* moments) +{ + const unsigned int gid = ${index.gid('get_global_id(0)', 'get_global_id(1)')}; + collect_moments_tock(f, gid, moments); +} +""" diff --git a/lid_driven_cavity/opencl/simulation.py b/lid_driven_cavity/opencl/AB.py index 37aab5b..7cbf1c6 100644 --- a/lid_driven_cavity/opencl/simulation.py +++ b/lid_driven_cavity/opencl/AB.py @@ -24,20 +24,6 @@ class Memory: def gid(self, x, y, z = 0): return z * (self.size_x*self.size_y) + y * self.size_x + x; -class CellList: - def __init__(self, context, queue, float_type, cells): - self.cl_cells = cl.Buffer(context, mf.READ_ONLY, size=len(cells) * numpy.uint32(0).nbytes) - self.np_cells = numpy.ndarray(shape=(len(cells), 1), dtype=numpy.uint32) - self.np_cells[:,0] = cells[:] - - cl.enqueue_copy(queue, self.cl_cells, self.np_cells).wait(); - - def get(self): - return self.cl_cells - - def size(self): - return (len(self.np_cells), 1, 1) - class Lattice: def __init__(self, geometry, kernel_src, descriptor, platform = 0, precision = 'single'): self.geometry = geometry @@ -103,3 +89,20 @@ class Lattice: cl.enqueue_copy(self.queue, moments, self.memory.cl_moments).wait(); return moments + +HelperTemplate = """ +__kernel void equilibrilize_all(__global ${float_type}* f_next, + __global ${float_type}* f_prev) +{ + const unsigned int gid = ${index.gid('get_global_id(0)', 'get_global_id(1)')}; + equilibrilize(f_next, f_prev, gid); + equilibrilize(f_prev, f_next, gid); +} + +__kernel void collect_moments_all(__global ${float_type}* f, + __global ${float_type}* moments) +{ + const unsigned int gid = ${index.gid('get_global_id(0)', 'get_global_id(1)')}; + collect_moments(f, gid, moments); +} +""" diff --git a/lid_driven_cavity/opencl/common.py b/lid_driven_cavity/opencl/common.py new file mode 100644 index 0000000..717c68f --- /dev/null +++ b/lid_driven_cavity/opencl/common.py @@ -0,0 +1,36 @@ +import pyopencl as cl +mf = cl.mem_flags + +import numpy + +import matplotlib +matplotlib.use('AGG') +import matplotlib.pyplot as plt + + +class CellList: + def __init__(self, context, queue, float_type, cells): + self.cl_cells = cl.Buffer(context, mf.READ_ONLY, size=len(cells) * numpy.uint32(0).nbytes) + self.np_cells = numpy.ndarray(shape=(len(cells), 1), dtype=numpy.uint32) + self.np_cells[:,0] = cells[:] + + cl.enqueue_copy(queue, self.cl_cells, self.np_cells).wait(); + + def get(self): + return self.cl_cells + + def size(self): + return (len(self.np_cells), 1, 1) + +def generate_moment_plots(lattice, moments): + for i, m in enumerate(moments): + print("Generating plot %d of %d." % (i+1, len(moments))) + + gid = lattice.memory.gid + velocity = numpy.reshape( + [ numpy.sqrt(m[gid(x,y)*3+1]**2 + m[gid(x,y)*3+2]**2) for x, y in lattice.geometry.inner_cells() ], + lattice.geometry.inner_size()) + + plt.figure(figsize=(10, 10)) + plt.imshow(velocity, origin='lower', cmap=plt.get_cmap('seismic')) + plt.savefig("result/ldc_2d_%02d.png" % i, bbox_inches='tight', pad_inches=0) diff --git a/lid_driven_cavity/opencl/ldc_2d.py b/lid_driven_cavity/opencl/ldc_2d.py index c080284..8d9e324 100644 --- a/lid_driven_cavity/opencl/ldc_2d.py +++ b/lid_driven_cavity/opencl/ldc_2d.py @@ -1,36 +1,16 @@ import numpy import time -import matplotlib -matplotlib.use('AGG') -import matplotlib.pyplot as plt - from boltzgen import Generator, Geometry from boltzgen.lbm.lattice import D2Q9 from boltzgen.lbm.model import BGK -from simulation import Lattice, CellList - -def MLUPS(cells, steps, time): - return cells * steps / time * 1e-6 - -def generate_moment_plots(lattice, moments): - for i, m in enumerate(moments): - print("Generating plot %d of %d." % (i+1, len(moments))) - - gid = lattice.memory.gid - velocity = numpy.reshape( - [ numpy.sqrt(m[gid(x,y)*3+1]**2 + m[gid(x,y)*3+2]**2) for x, y in lattice.geometry.inner_cells() ], - lattice.geometry.inner_size()) - - plt.figure(figsize=(10, 10)) - plt.imshow(velocity, origin='lower', cmap=plt.get_cmap('seismic')) - plt.savefig("result/ldc_2d_%02d.png" % i, bbox_inches='tight', pad_inches=0) +from common import CellList, generate_moment_plots nUpdates = 100000 nStat = 10000 -geometry = Geometry(256, 256) +geometry = Geometry(512, 512) print("Generating kernel using boltzgen...\n") @@ -38,37 +18,35 @@ functions = ['collide_and_stream', 'equilibrilize', 'collect_moments', 'momenta_ extras = ['cell_list_dispatch'] precision = 'single' +streaming = 'AA' + +import AA +import AB + +Lattice = eval('%s.Lattice' % streaming) +HelperTemplate = eval('%s.HelperTemplate' % streaming) + +def MLUPS(cells, steps, time): + return cells * steps / time * 1e-6 generator = Generator( - model = BGK(D2Q9, tau = 0.6), + model = BGK(D2Q9, tau = 0.54), target = 'cl', precision = precision, + streaming = streaming, index = 'ZYX', layout = 'SOA') kernel_src = generator.kernel(geometry, functions, extras) -kernel_src += generator.custom(geometry, """ -__kernel void equilibrilize_all(__global ${float_type}* f_next, - __global ${float_type}* f_prev) -{ - const unsigned int gid = ${index.gid('get_global_id(0)', 'get_global_id(1)')}; - equilibrilize(f_next, f_prev, gid); - equilibrilize(f_prev, f_next, gid); -} - -__kernel void collect_moments_all(__global ${float_type}* f, - __global ${float_type}* moments) -{ - const unsigned int gid = ${index.gid('get_global_id(0)', 'get_global_id(1)')}; - collect_moments(f, gid, moments); -} -""") +kernel_src += generator.custom(geometry, HelperTemplate) print("Initializing simulation...\n") lattice = Lattice(geometry, kernel_src, D2Q9, precision = precision) gid = lattice.memory.gid +ghost_cells = CellList(lattice.context, lattice.queue, lattice.float_type, + [ gid(x,y) for x, y in geometry.cells() if x == 0 or y == 0 or x == geometry.size_x-1 or y == geometry.size_y-1 ]) bulk_cells = CellList(lattice.context, lattice.queue, lattice.float_type, [ gid(x,y) for x, y in geometry.inner_cells() if x > 1 and x < geometry.size_x-2 and y > 1 and y < geometry.size_y-2 ]) wall_cells = CellList(lattice.context, lattice.queue, lattice.float_type, @@ -76,9 +54,21 @@ wall_cells = CellList(lattice.context, lattice.queue, lattice.float_type, lid_cells = CellList(lattice.context, lattice.queue, lattice.float_type, [ gid(x,y) for x, y in geometry.inner_cells() if y == geometry.size_y-2 ]) -lattice.schedule('collide_and_stream_cells', bulk_cells) -lattice.schedule('velocity_momenta_boundary_cells', wall_cells, numpy.array([0.0, 0.0], dtype=lattice.float_type[0])) -lattice.schedule('velocity_momenta_boundary_cells', lid_cells, numpy.array([0.1, 0.0], dtype=lattice.float_type[0])) +if streaming == 'AB': + lattice.schedule('collide_and_stream_cells', bulk_cells) + lattice.schedule('velocity_momenta_boundary_cells', wall_cells, numpy.array([0.0, 0.0], dtype=lattice.float_type[0])) + lattice.schedule('velocity_momenta_boundary_cells', lid_cells, numpy.array([0.1, 0.0], dtype=lattice.float_type[0])) + +elif streaming == 'AA': + lattice.schedule_tick('collide_and_stream_tick_cells', bulk_cells) + lattice.schedule_tick('velocity_momenta_boundary_tick_cells', wall_cells, numpy.array([0.0, 0.0], dtype=lattice.float_type[0])) + lattice.schedule_tick('velocity_momenta_boundary_tick_cells', lid_cells, numpy.array([0.1, 0.0], dtype=lattice.float_type[0])) + + lattice.schedule_tock('equilibrilize_tick_cells', ghost_cells) + lattice.schedule_tock('collide_and_stream_tock_cells', bulk_cells) + lattice.schedule_tock('velocity_momenta_boundary_tock_cells', wall_cells, numpy.array([0.0, 0.0], dtype=lattice.float_type[0])) + lattice.schedule_tock('velocity_momenta_boundary_tock_cells', lid_cells, numpy.array([0.1, 0.0], dtype=lattice.float_type[0])) + print("Starting simulation using %d cells...\n" % lattice.geometry.volume) |