def create_ops_par_loop(trees, ops_kernel, parameters, block, name_to_ops_dat,
accessible_origin, par_to_ops_stencil, dims):
it_range = []
for tree in trees:
if isinstance(tree, IterationTree):
for bounds in [it.bounds() for it in tree]:
it_range.extend(bounds)
range_array = Array(name='%s_range' % ops_kernel.name,
dimensions=(DefaultDimension(
name='range', default_value=len(it_range)), ),
dtype=np.int32,
scope='stack')
range_array_init = Expression(
ClusterizedEq(Eq(range_array, ListInitializer(it_range))))
ops_par_loop_call = Call(namespace['ops_par_loop'], [
Literal(ops_kernel.name),
Literal('"%s"' % ops_kernel.name), block, dims, range_array, *[
create_ops_arg(p, accessible_origin, name_to_ops_dat,
par_to_ops_stencil) for p in parameters
]
])
return [range_array_init], ops_par_loop_call
def test_create_ops_dat_function(self):
grid = Grid(shape=(4))
u = Function(name='u', grid=grid, space_order=2)
block = OpsBlock('block')
name_to_ops_dat = {}
result = create_ops_dat(u, name_to_ops_dat, block)
assert name_to_ops_dat['u'].name == namespace['ops_dat_name'](u.name)
assert name_to_ops_dat['u']._C_typename == namespace['ops_dat_type']
assert result[0].expr.lhs.name == namespace['ops_dat_dim'](u.name)
assert result[0].expr.rhs.params == (Integer(4), )
assert result[1].expr.lhs.name == namespace['ops_dat_base'](u.name)
assert result[1].expr.rhs.params == (Zero(), )
assert result[2].expr.lhs.name == namespace['ops_dat_d_p'](u.name)
assert result[2].expr.rhs.params == (Integer(2), )
assert result[3].expr.lhs.name == namespace['ops_dat_d_m'](u.name)
assert result[3].expr.rhs.params == (Integer(-2), )
assert result[4].expr.lhs == name_to_ops_dat['u']
assert type(result[4].expr.rhs) == namespace['ops_decl_dat']
assert result[4].expr.rhs.args == (
block, 1, Symbol(namespace['ops_dat_dim'](u.name)),
Symbol(namespace['ops_dat_base'](u.name)),
Symbol(namespace['ops_dat_d_m'](u.name)),
Symbol(namespace['ops_dat_d_p'](u.name)), Byref(u.indexify(
(0, ))), Literal('"%s"' % u._C_typedata), Literal('"u"'))
def create_ops_par_loop(trees, ops_kernel, parameters, block, name_to_ops_dat,
accessible_origin, par_to_ops_stencil, dims):
it_range = []
devito_to_ops_indexer = 1
for tree in trees:
if isinstance(tree, IterationTree):
for i in tree:
it_range.extend(
[i.symbolic_min, i.symbolic_max + devito_to_ops_indexer])
range_array = Array(name='%s_range' % ops_kernel.name,
dimensions=(DefaultDimension(
name='range', default_value=len(it_range)), ),
dtype=np.int32,
scope='stack')
range_array_init = Expression(
ClusterizedEq(Eq(range_array, ListInitializer(it_range))))
ops_args = []
for p in parameters:
ops_arg = create_ops_arg(p, accessible_origin, name_to_ops_dat,
par_to_ops_stencil)
ops_args.append(
ops_arg.ops_type(ops_arg.ops_name, ops_arg.elements_per_point,
ops_arg.dtype, ops_arg.rw_flag))
ops_par_loop_call = Call(namespace['ops_par_loop'], [
Literal(ops_kernel.name),
Literal('"%s"' % ops_kernel.name), block, dims, range_array, *ops_args
])
return [range_array_init], ops_par_loop_call
def create_ops_arg(p, name_to_ops_dat, par_to_ops_stencil):
if p.is_Constant:
return namespace['ops_arg_gbl'](
Byref(Constant(name=p.name[1:])), 1,
Literal('"%s"' % dtype_to_cstr(p.dtype)), namespace['ops_read'])
else:
return namespace['ops_arg_dat'](
name_to_ops_dat[p.name], 1, par_to_ops_stencil[p],
Literal('"%s"' % dtype_to_cstr(p.dtype)),
namespace['ops_read'] if p.read_only else namespace['ops_write'])
def _specialize_iet(self, iet, **kwargs):
warning("The OPS backend is still work-in-progress")
ops_init = Call(namespace['ops_init'], [0, 0, 2])
ops_partition = Call(namespace['ops_partition'], Literal('""'))
ops_exit = Call(namespace['ops_exit'])
ops_block = OpsBlock('block')
# Extract all symbols that need to be converted to ops_dat
dims = []
to_dat = set()
for section, trees in find_affine_trees(iet).items():
dims.append(len(trees[0].dimensions))
symbols = set(FindSymbols('symbolics').visit(trees[0].root))
symbols -= set(FindSymbols('defines').visit(trees[0].root))
to_dat |= symbols
# To ensure deterministic code generation we order the datasets to
# be generated (since a set is an unordered collection)
to_dat = filter_sorted(to_dat)
name_to_ops_dat = {}
pre_time_loop = []
for f in to_dat:
if f.is_Constant:
continue
pre_time_loop.extend(create_ops_dat(f, name_to_ops_dat, ops_block))
for n, (section, trees) in enumerate(find_affine_trees(iet).items()):
pre_loop, ops_kernel = opsit(trees, n)
pre_time_loop.extend(pre_loop)
self._ops_kernels.append(ops_kernel)
assert (d == dims[0] for d in dims), \
"The OPS backend currently assumes that all kernels \
have the same number of dimensions"
ops_block_init = Expression(
ClusterizedEq(
Eq(ops_block,
namespace['ops_decl_block'](dims[0], Literal('"block"')))))
self._headers.append(namespace['ops_define_dimension'](dims[0]))
self._includes.append('stdio.h')
body = [
ops_init, ops_block_init, *pre_time_loop, ops_partition, iet,
ops_exit
]
return List(body=body)
def create_ops_arg(p, accessible_origin, name_to_ops_dat, par_to_ops_stencil):
elements_per_point = 1
dtype = Literal('"%s"' % dtype_to_cstr(p.dtype))
if p.is_Constant:
ops_type = namespace['ops_arg_gbl']
ops_name = Byref(Constant(name=p.name[1:]))
rw_flag = namespace['ops_read']
else:
ops_type = namespace['ops_arg_dat']
accessible_info = accessible_origin[p.name]
ops_name = name_to_ops_dat[p.name] \
if accessible_info.time is None \
else name_to_ops_dat[accessible_info.origin_name].\
indexify([Add(accessible_info.time, accessible_info.shift)])
rw_flag = namespace['ops_read'] if p.read_only else namespace[
'ops_write']
ops_arg = OpsArgDecl(ops_type=ops_type,
ops_name=ops_name,
elements_per_point=elements_per_point,
dtype=dtype,
rw_flag=rw_flag)
return ops_arg
def to_ops_stencil(param, accesses):
dims = len(accesses[0])
pts = len(accesses)
stencil_name = namespace['ops_stencil_name'](dims, param.name, pts)
stencil_array = Array(
name=stencil_name,
dimensions=(DefaultDimension(name='len', default_value=dims * pts), ),
dtype=np.int32,
)
ops_stencil = OpsStencil(stencil_name.upper())
return ops_stencil, [
Expression(
ClusterizedEq(
Eq(stencil_array,
ListInitializer(list(itertools.chain(*accesses)))))),
Expression(
ClusterizedEq(
Eq(
ops_stencil, namespace['ops_decl_stencil'](
dims, pts, Symbol(stencil_array.name),
Literal('"%s"' % stencil_name.upper())))))
]
def create_ops_arg(p, accessible_origin, name_to_ops_dat, par_to_ops_stencil):
if p.is_Constant:
return namespace['ops_arg_gbl'](
Byref(Constant(name=p.name[1:])), 1,
Literal('"%s"' % dtype_to_cstr(p.dtype)), namespace['ops_read'])
else:
accessible_info = accessible_origin[p.name]
dat_name = name_to_ops_dat[p.name] \
if accessible_info.time is None \
else name_to_ops_dat[accessible_info.origin_name].\
indexify([accessible_info.time])
return namespace['ops_arg_dat'](
dat_name, 1, par_to_ops_stencil[p],
Literal('"%s"' % dtype_to_cstr(p.dtype)),
namespace['ops_read'] if p.read_only else namespace['ops_write'])
def test_create_ops_arg_constant(self):
a = Constant(name='*a')
res = create_ops_arg(a, {}, {})
assert res.name == namespace['ops_arg_gbl'].name
assert res.args == [
Byref(Constant(name='a')), 1,
Literal('"%s"' % dtype_to_cstr(a.dtype)), namespace['ops_read']
]
def test_create_ops_arg_constant(self):
a = Constant(name='*a')
ops_arg = create_ops_arg(a, {}, {}, {})
assert ops_arg.ops_type == namespace['ops_arg_gbl']
assert str(ops_arg.ops_name) == str(Byref(Constant(name='a')))
assert ops_arg.elements_per_point == 1
assert ops_arg.dtype == Literal('"%s"' % dtype_to_cstr(a.dtype))
assert ops_arg.rw_flag == namespace['ops_read']
def test_create_ops_arg_constant(self):
a = Constant(name='*a')
res = create_ops_arg(a, {}, {}, {})
assert type(res) == namespace['ops_arg_gbl']
assert str(res.args[0]) == str(Byref(Constant(name='a')))
assert res.args[1] == 1
assert res.args[2] == Literal('"%s"' % dtype_to_cstr(a.dtype))
assert res.args[3] == namespace['ops_read']
def test_create_ops_arg_function(self, read):
u = OpsAccessible('u', np.float32, read)
dat = OpsDat('u_dat')
stencil = OpsStencil('stencil')
info = AccessibleInfo(u, None, None)
res = create_ops_arg(u, {'u': info}, {'u': dat}, {u: stencil})
assert type(res) == namespace['ops_arg_dat']
assert res.args == (dat, 1, stencil,
Literal('"%s"' % dtype_to_cstr(u.dtype)),
namespace['ops_read']
if read else namespace['ops_write'])
def test_create_ops_arg_function(self, read):
u = OpsAccessible('u', dtype=np.float32, read_only=read)
dat = OpsDat('u_dat')
stencil = OpsStencil('stencil')
info = AccessibleInfo(u, None, None)
ops_arg = create_ops_arg(u, {'u': info}, {'u': dat}, {u: stencil})
assert ops_arg.ops_type == namespace['ops_arg_dat']
assert ops_arg.ops_name == OpsDat('u_dat')
assert ops_arg.elements_per_point == 1
assert ops_arg.dtype == Literal('"%s"' % dtype_to_cstr(u.dtype))
assert ops_arg.rw_flag == \
namespace['ops_read'] if read else namespace['ops_write']
def test_to_ops_stencil(self, _accesses):
param = Symbol('foo')
accesses = eval(_accesses)
stencil_name = 's2d_foo_%spt' % len(accesses)
stencil, result = to_ops_stencil(param, accesses)
assert stencil.name == stencil_name.upper()
assert result[0].expr.lhs.name == stencil_name
assert result[0].expr.rhs.params == tuple(itertools.chain(*accesses))
assert result[1].expr.lhs == stencil
assert type(result[1].expr.rhs) == namespace['ops_decl_stencil']
assert result[1].expr.rhs.args == (2, len(accesses),
Symbol(stencil_name),
Literal('"%s"' %
stencil_name.upper()))
def test_create_ops_dat_function(self):
grid = Grid(shape=(4))
u = Function(name='u', grid=grid, space_order=2)
block = OpsBlock('block')
name_to_ops_dat = {}
result = create_ops_dat(u, name_to_ops_dat, block)
assert name_to_ops_dat['u'].name == namespace['ops_dat_name'](u.name)
assert name_to_ops_dat['u']._C_typename == namespace['ops_dat_type']
assert result[0].expr.lhs.name == namespace['ops_dat_dim'](u.name)
assert result[0].expr.rhs.params == (Integer(4), )
assert result[1].expr.lhs.name == namespace['ops_dat_base'](u.name)
assert result[1].expr.rhs.params == (Zero(), )
assert result[2].expr.lhs.name == namespace['ops_dat_d_p'](u.name)
assert result[2].expr.rhs.params == (Integer(2), )
assert result[3].expr.lhs.name == namespace['ops_dat_d_m'](u.name)
assert result[3].expr.rhs.params == (Integer(-2), )
assert result[4].expr.lhs == name_to_ops_dat['u']
assert result[4].expr.rhs.name == namespace['ops_decl_dat'].name
assert result[4].expr.rhs.args == (
block, 1, Symbol(namespace['ops_dat_dim'](u.name)),
Symbol(namespace['ops_dat_base'](u.name)),
Symbol(namespace['ops_dat_d_m'](u.name)),
Symbol(namespace['ops_dat_d_p'](u.name)), Byref(u.indexify(
(0, ))), Literal('"%s"' % u._C_typedata), Literal('"u"'))
def test_create_ops_arg_constant(self):
a = Constant(name='*a')
res = create_ops_arg(a, {}, {})
assert res.name == namespace['ops_arg_gbl'].name
assert res.args == [
Byref(Constant(name='a')), 1,
Literal('"%s"' % dtype_to_cstr(a.dtype)), namespace['ops_read']
]
@pytest.mark.parametrize('read', [True, False])
def test_create_ops_arg_function(self, read):
u = OpsAccessible('u', np.float32, read)
dat = OpsDat('u_dat')
stencil = OpsStencil('stencil')
res = create_ops_arg(u, {'u': dat}, {u: stencil})
assert res.name == namespace['ops_arg_dat'].name
assert res.args == [
dat, 1, stencil,
Literal('"%s"' % dtype_to_cstr(u.dtype)),
namespace['ops_read'] if read else namespace['ops_write']
]
def create_ops_dat(f, name_to_ops_dat, block):
ndim = f.ndim - (1 if f.is_TimeFunction else 0)
dim = Array(name=namespace['ops_dat_dim'](f.name),
dimensions=(DefaultDimension(name='dim',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
base = Array(name=namespace['ops_dat_base'](f.name),
dimensions=(DefaultDimension(name='base',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
d_p = Array(name=namespace['ops_dat_d_p'](f.name),
dimensions=(DefaultDimension(name='d_p',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
d_m = Array(name=namespace['ops_dat_d_m'](f.name),
dimensions=(DefaultDimension(name='d_m',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
res = []
base_val = [Zero() for i in range(ndim)]
# If f is a TimeFunction we need to create a ops_dat for each time stepping
# variable (eg: t1, t2)
if f.is_TimeFunction:
time_pos = f._time_position
time_index = f.indices[time_pos]
time_dims = f.shape[time_pos]
dim_shape = sympify(f.shape[:time_pos] + f.shape[time_pos + 1:])
padding = f.padding[:time_pos] + f.padding[time_pos + 1:]
halo = f.halo[:time_pos] + f.halo[time_pos + 1:]
d_p_val = tuple(sympify([p[0] + h[0] for p, h in zip(padding, halo)]))
d_m_val = tuple(
sympify([-(p[1] + h[1]) for p, h in zip(padding, halo)]))
ops_dat_array = Array(name=namespace['ops_dat_name'](f.name),
dimensions=(DefaultDimension(
name='dat', default_value=time_dims), ),
dtype='ops_dat',
scope='stack')
dat_decls = []
for i in range(time_dims):
name = '%s%s%s' % (f.name, time_index, i)
name_to_ops_dat[name] = ops_dat_array.indexify(
[Symbol('%s%s' % (time_index, i))])
dat_decls.append(namespace['ops_decl_dat'](block, 1,
Symbol(dim.name),
Symbol(base.name),
Symbol(d_m.name),
Symbol(d_p.name),
Byref(f.indexify([i])),
Literal('"%s"' %
f._C_typedata),
Literal('"%s"' % name)))
ops_decl_dat = Expression(
ClusterizedEq(Eq(ops_dat_array, ListInitializer(dat_decls))))
else:
ops_dat = OpsDat("%s_dat" % f.name)
name_to_ops_dat[f.name] = ops_dat
d_p_val = tuple(
sympify([p[0] + h[0] for p, h in zip(f.padding, f.halo)]))
d_m_val = tuple(
sympify([-(p[1] + h[1]) for p, h in zip(f.padding, f.halo)]))
dim_shape = sympify(f.shape)
ops_decl_dat = Expression(
ClusterizedEq(
Eq(
ops_dat,
namespace['ops_decl_dat'](block, 1, Symbol(dim.name),
Symbol(base.name),
Symbol(d_m.name),
Symbol(d_p.name),
Byref(f.indexify([0])),
Literal('"%s"' % f._C_typedata),
Literal('"%s"' % f.name)))))
res.append(Expression(ClusterizedEq(Eq(dim, ListInitializer(dim_shape)))))
res.append(Expression(ClusterizedEq(Eq(base, ListInitializer(base_val)))))
res.append(Expression(ClusterizedEq(Eq(d_p, ListInitializer(d_p_val)))))
res.append(Expression(ClusterizedEq(Eq(d_m, ListInitializer(d_m_val)))))
res.append(ops_decl_dat)
return res
def create_ops_dat(f, name_to_ops_dat, block):
ndim = f.ndim - (1 if f.is_TimeFunction else 0)
dim = Array(name=namespace['ops_dat_dim'](f.name),
dimensions=(DefaultDimension(name='dim',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
base = Array(name=namespace['ops_dat_base'](f.name),
dimensions=(DefaultDimension(name='base',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
d_p = Array(name=namespace['ops_dat_d_p'](f.name),
dimensions=(DefaultDimension(name='d_p',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
d_m = Array(name=namespace['ops_dat_d_m'](f.name),
dimensions=(DefaultDimension(name='d_m',
default_value=ndim), ),
dtype=np.int32,
scope='stack')
base_val = [Zero() for i in range(ndim)]
# If f is a TimeFunction we need to create a ops_dat for each time stepping
# variable (eg: t1, t2)
if f.is_TimeFunction:
time_pos = f._time_position
time_index = f.indices[time_pos]
time_dims = f.shape[time_pos]
dim_val = f.shape[:time_pos] + f.shape[time_pos + 1:]
d_p_val = f._size_nodomain.left[time_pos + 1:]
d_m_val = [-i for i in f._size_nodomain.right[time_pos + 1:]]
ops_dat_array = Array(name=namespace['ops_dat_name'](f.name),
dimensions=(DefaultDimension(
name='dat', default_value=time_dims), ),
dtype=namespace['ops_dat_type'],
scope='stack')
dat_decls = []
for i in range(time_dims):
name = '%s%s%s' % (f.name, time_index, i)
dat_decls.append(namespace['ops_decl_dat'](block, 1,
Symbol(dim.name),
Symbol(base.name),
Symbol(d_m.name),
Symbol(d_p.name),
Byref(f.indexify([i])),
Literal('"%s"' %
f._C_typedata),
Literal('"%s"' % name)))
ops_decl_dat = Expression(
ClusterizedEq(Eq(ops_dat_array, ListInitializer(dat_decls))))
# Inserting the ops_dat array in case of TimeFunction.
name_to_ops_dat[f.name] = ops_dat_array
else:
ops_dat = OpsDat("%s_dat" % f.name)
name_to_ops_dat[f.name] = ops_dat
dim_val = f.shape
d_p_val = f._size_nodomain.left
d_m_val = [-i for i in f._size_nodomain.right]
ops_decl_dat = Expression(
ClusterizedEq(
Eq(
ops_dat,
namespace['ops_decl_dat'](block, 1, Symbol(dim.name),
Symbol(base.name),
Symbol(d_m.name),
Symbol(d_p.name),
Byref(f.indexify([0])),
Literal('"%s"' % f._C_typedata),
Literal('"%s"' % f.name)))))
dim_val = Expression(ClusterizedEq(Eq(dim, ListInitializer(dim_val))))
base_val = Expression(ClusterizedEq(Eq(base, ListInitializer(base_val))))
d_p_val = Expression(ClusterizedEq(Eq(d_p, ListInitializer(d_p_val))))
d_m_val = Expression(ClusterizedEq(Eq(d_m, ListInitializer(d_m_val))))
return OpsDatDecl(dim_val=dim_val,
base_val=base_val,
d_p_val=d_p_val,
d_m_val=d_m_val,
ops_decl_dat=ops_decl_dat)
def make_ops_kernels(iet):
warning("The OPS backend is still work-in-progress")
affine_trees = find_affine_trees(iet).items()
# If there is no affine trees, then there is no loop to be optimized using OPS.
if not affine_trees:
return iet, {}
ops_init = Call(namespace['ops_init'], [0, 0, 2])
ops_partition = Call(namespace['ops_partition'], Literal('""'))
ops_exit = Call(namespace['ops_exit'])
# Extract all symbols that need to be converted to ops_dat
dims = []
to_dat = set()
for _, tree in affine_trees:
dims.append(len(tree[0].dimensions))
symbols = set(FindSymbols('symbolics').visit(tree[0].root))
symbols -= set(FindSymbols('defines').visit(tree[0].root))
to_dat |= symbols
# Create the OPS block for this problem
ops_block = OpsBlock('block')
ops_block_init = Expression(
ClusterizedEq(
Eq(ops_block, namespace['ops_decl_block'](dims[0],
Literal('"block"')))))
# To ensure deterministic code generation we order the datasets to
# be generated (since a set is an unordered collection)
to_dat = filter_sorted(to_dat)
name_to_ops_dat = {}
pre_time_loop = []
after_time_loop = []
for f in to_dat:
if f.is_Constant:
continue
pre_time_loop.extend(
list(create_ops_dat(f, name_to_ops_dat, ops_block)))
# Copy data from device to host
after_time_loop.extend(
create_ops_fetch(f, name_to_ops_dat, f.grid.time_dim.extreme_max))
# Generate ops kernels for each offloadable iteration tree
mapper = {}
ffuncs = []
for n, (_, tree) in enumerate(affine_trees):
pre_loop, ops_kernel, ops_par_loop_call = opsit(
tree, n, name_to_ops_dat, ops_block, dims[0])
pre_time_loop.extend(pre_loop)
ffuncs.append(ops_kernel)
mapper[tree[0].root] = ops_par_loop_call
mapper.update({i.root: mapper.get(i.root) for i in tree}) # Drop trees
iet = Transformer(mapper).visit(iet)
assert (d == dims[0] for d in dims), \
"The OPS backend currently assumes that all kernels \
have the same number of dimensions"
iet = iet._rebuild(body=flatten([
ops_init, ops_block_init, pre_time_loop, ops_partition, iet.body,
after_time_loop, ops_exit
]))
return iet, {
'includes': ['stdio.h', 'ops_seq.h'],
'ffuncs': ffuncs,
'headers': [namespace['ops_define_dimension'](dims[0])]
}
def _specialize_iet(self, iet, **kwargs):
warning("The OPS backend is still work-in-progress")
affine_trees = find_affine_trees(iet).items()
# If there is no affine trees, then there is no loop to be optimized using OPS.
if not affine_trees:
return iet
ops_init = Call(namespace['ops_init'], [0, 0, 2])
ops_partition = Call(namespace['ops_partition'], Literal('""'))
ops_exit = Call(namespace['ops_exit'])
# Extract all symbols that need to be converted to ops_dat
dims = []
to_dat = set()
for _, tree in affine_trees:
dims.append(len(tree[0].dimensions))
symbols = set(FindSymbols('symbolics').visit(tree[0].root))
symbols -= set(FindSymbols('defines').visit(tree[0].root))
to_dat |= symbols
# Create the OPS block for this problem
ops_block = OpsBlock('block')
ops_block_init = Expression(
ClusterizedEq(
Eq(ops_block,
namespace['ops_decl_block'](dims[0], Literal('"block"')))))
# To ensure deterministic code generation we order the datasets to
# be generated (since a set is an unordered collection)
to_dat = filter_sorted(to_dat)
name_to_ops_dat = {}
pre_time_loop = []
after_time_loop = []
for f in to_dat:
if f.is_Constant:
continue
pre_time_loop.extend(
list(create_ops_dat(f, name_to_ops_dat, ops_block)))
# To return the result to Devito, it is necessary to copy the data
# from the dat object back to the CPU memory.
after_time_loop.extend(
create_ops_fetch(f, name_to_ops_dat,
self.time_dimension.extreme_max))
# Generate ops kernels for each offloadable iteration tree
mapper = {}
for n, (_, tree) in enumerate(affine_trees):
pre_loop, ops_kernel, ops_par_loop_call = opsit(
tree, n, name_to_ops_dat, ops_block, dims[0])
pre_time_loop.extend(pre_loop)
self._func_table[namespace['ops_kernel_file'](ops_kernel.name)] = \
MetaCall(ops_kernel, False)
mapper[tree[0].root] = ops_par_loop_call
mapper.update({i.root: mapper.get(i.root)
for i in tree}) # Drop trees
iet = Transformer(mapper).visit(iet)
assert (d == dims[0] for d in dims), \
"The OPS backend currently assumes that all kernels \
have the same number of dimensions"
self._headers.append(namespace['ops_define_dimension'](dims[0]))
self._includes.extend(['stdio.h', 'ops_seq.h'])
body = [
ops_init, ops_block_init, *pre_time_loop, ops_partition, iet,
*after_time_loop, ops_exit
]
return List(body=body)