From 0001fb038cdf0800cc82aa4b147ffa357cc6c229 Mon Sep 17 00:00:00 2001
From: Dominic Kempf <dominic.kempf@iwr.uni-heidelberg.de>
Date: Wed, 19 Apr 2017 17:00:32 +0200
Subject: [PATCH] Move assembly of input for sf kernels to a dedicated object
---
python/dune/perftool/sumfact/accumulation.py | 9 +-
python/dune/perftool/sumfact/basis.py | 86 +++++++++++++++++---
python/dune/perftool/sumfact/realization.py | 47 +----------
python/dune/perftool/sumfact/symbolic.py | 39 ++++-----
4 files changed, 101 insertions(+), 80 deletions(-)
diff --git a/python/dune/perftool/sumfact/accumulation.py b/python/dune/perftool/sumfact/accumulation.py
index d24e100c..f01d2bd3 100644
--- a/python/dune/perftool/sumfact/accumulation.py
+++ b/python/dune/perftool/sumfact/accumulation.py
@@ -29,12 +29,13 @@ from dune.perftool.sumfact.tabulation import (basis_functions_per_direction,
from dune.perftool.sumfact.switch import (get_facedir,
get_facemod,
)
-from dune.perftool.sumfact.symbolic import SumfactKernel
+from dune.perftool.sumfact.symbolic import SumfactKernel, SumfactKernelInputBase
from dune.perftool.ufl.modified_terminals import extract_modified_arguments
from dune.perftool.tools import get_pymbolic_basename
from dune.perftool.error import PerftoolError
from dune.perftool.sumfact.quadrature import quadrature_inames
+from pytools import ImmutableRecord
import loopy as lp
import numpy as np
@@ -59,6 +60,10 @@ def accum_iname(restriction, bound, i):
return sumfact_iname(bound, "accum")
+class AlreadyAssembledInput(SumfactKernelInputBase, ImmutableRecord):
+ pass
+
+
@backend(interface="accum_insn", name="sumfact")
def generate_accumulation_instruction(visitor, accterm, measure, subdomain_id):
# When doing sumfactorization we want to split the test function
@@ -126,7 +131,7 @@ def generate_accumulation_instruction(visitor, accterm, measure, subdomain_id):
preferred_position=indices[-1] if accterm.new_indices else None,
accumvar=accum,
within_inames=jacobian_inames,
- coeff_func_index=coeff_func_index,
+ input=AlreadyAssembledInput(index=coeff_func_index),
)
from dune.perftool.sumfact.vectorization import attach_vectorization_info
diff --git a/python/dune/perftool/sumfact/basis.py b/python/dune/perftool/sumfact/basis.py
index cb318c7e..d7687986 100644
--- a/python/dune/perftool/sumfact/basis.py
+++ b/python/dune/perftool/sumfact/basis.py
@@ -33,19 +33,77 @@ from dune.perftool.pdelab.geometry import (local_dimension,
world_dimension,
)
from dune.perftool.loopy.buffer import initialize_buffer
-from dune.perftool.sumfact.symbolic import SumfactKernel
+from dune.perftool.sumfact.symbolic import SumfactKernel, SumfactKernelInputBase
from dune.perftool.options import get_option
from dune.perftool.pdelab.driver import FEM_name_mangling
from dune.perftool.pdelab.restriction import restricted_name
+from dune.perftool.pdelab.spaces import name_lfs, name_lfs_bound, lfs_child, name_leaf_lfs
from dune.perftool.tools import maybe_wrap_subscript
+from dune.perftool.pdelab.basis import shape_as_pymbolic
+from dune.perftool.sumfact.accumulation import sumfact_iname
-from pytools import product
+from ufl.functionview import select_subelement
+from ufl import VectorElement, TensorElement
+
+from pytools import product, ImmutableRecord
from loopy.match import Writes
import pymbolic.primitives as prim
+class LFSSumfactKernelInput(SumfactKernelInputBase, ImmutableRecord):
+ def __init__(self,
+ coeff_func=None,
+ coeff_func_index=None,
+ element=None,
+ component=None,
+ restriction=0,
+ ):
+ ImmutableRecord.__init__(self,
+ coeff_func=coeff_func,
+ coeff_func_index=coeff_func_index,
+ element=element,
+ component=component,
+ restriction=restriction,
+ )
+
+ def realize(self, sf, index, insn_dep):
+ lfs = name_lfs(self.element, self.restriction, self.component)
+ sub_element = select_subelement(self.element, self.component)
+ shape = sub_element.value_shape() + (self.element.cell().geometric_dimension(),)
+
+ if isinstance(sub_element, (VectorElement, TensorElement)):
+ # Could be 0 but shouldn't be None
+ assert self.coeff_func_index is not None
+
+ lfs_pym = lfs_child(lfs,
+ (self.coeff_func_index,),
+ shape=shape_as_pymbolic(shape[:-1]),
+ symmetry=self.element.symmetry())
+
+ leaf_element = sub_element
+ if isinstance(sub_element, (VectorElement, TensorElement)):
+ leaf_element = sub_element.sub_elements()[0]
+
+ lfs = name_leaf_lfs(leaf_element, self.restriction)
+ basisiname = sumfact_iname(name_lfs_bound(lfs), "basis")
+ container = self.coeff_func(self.restriction)
+ if isinstance(sub_element, (VectorElement, TensorElement)):
+ from dune.perftool.pdelab.argument import pymbolic_coefficient as pc
+ coeff = pc(container, lfs_pym, basisiname)
+ else:
+ from dune.perftool.pdelab.argument import pymbolic_coefficient as pc
+ coeff = pc(container, lfs, basisiname)
+
+ assignee = prim.Subscript(prim.Variable("input_{}".format(sf.buffer)),
+ (prim.Variable(basisiname),) + (index,))
+ instruction(assignee=assignee,
+ expression=coeff,
+ depends_on=sf.insn_dep.union(insn_dep),
+ tags=frozenset({"sumfact_stage{}".format(sf.stage)}),
+ )
+
def name_sumfact_base_buffer():
count = get_counter('sumfact_base_buffer')
name = "buffer_{}".format(str(count))
@@ -102,14 +160,17 @@ def pymbolic_coefficient_gradient(element, restriction, component, coeff_func, v
if len(indices) == 2:
coeff_func_index = indices[0]
+ inp = LFSSumfactKernelInput(coeff_func=coeff_func,
+ coeff_func_index=coeff_func_index,
+ element=element,
+ component=component,
+ restriction=restriction,
+ )
+
# The sum factorization kernel object gathering all relevant information
sf = SumfactKernel(matrix_sequence=matrix_sequence,
- restriction=restriction,
preferred_position=indices[-1],
- coeff_func=coeff_func,
- coeff_func_index=coeff_func_index,
- element=element,
- component=component,
+ input=inp,
)
from dune.perftool.sumfact.vectorization import attach_vectorization_info
@@ -156,11 +217,14 @@ def pymbolic_coefficient(element, restriction, component, coeff_func, visitor_in
facemod=get_facemod(restriction),
basis_size=basis_size)
+ inp = LFSSumfactKernelInput(coeff_func=coeff_func,
+ element=element,
+ component=component,
+ restriction=restriction,
+ )
+
sf = SumfactKernel(matrix_sequence=matrix_sequence,
- restriction=restriction,
- coeff_func=coeff_func,
- element=element,
- component=component,
+ input=inp,
)
from dune.perftool.sumfact.vectorization import attach_vectorization_info
diff --git a/python/dune/perftool/sumfact/realization.py b/python/dune/perftool/sumfact/realization.py
index 4ff4dfb4..c03c5321 100644
--- a/python/dune/perftool/sumfact/realization.py
+++ b/python/dune/perftool/sumfact/realization.py
@@ -2,9 +2,6 @@
The code that triggers the creation of the necessary code constructs
to realize a sum factorization kernel
"""
-from ufl.functionview import select_subelement
-from ufl import VectorElement, TensorElement
-
from dune.perftool.generation import (barrier,
dump_accumulate_timer,
generator_factory,
@@ -21,7 +18,6 @@ from dune.perftool.loopy.buffer import (get_buffer_temporary,
from dune.perftool.pdelab.argument import pymbolic_coefficient
from dune.perftool.pdelab.basis import shape_as_pymbolic
from dune.perftool.pdelab.geometry import world_dimension
-from dune.perftool.pdelab.spaces import name_lfs, name_lfs_bound, lfs_child, name_leaf_lfs
from dune.perftool.options import get_option
from dune.perftool.pdelab.signatures import assembler_routine_name
from dune.perftool.sumfact.permutation import (sumfact_permutation_strategy,
@@ -63,7 +59,7 @@ def _realize_sum_factorization_kernel(sf):
# Set up the input for stage 1
if sf.stage == 1 and not get_option("fastdg"):
- assert sf.coeff_func
+ assert sf.input
# Get the input temporary!
input_setup = get_buffer_temporary(sf.buffer,
@@ -71,53 +67,18 @@ def _realize_sum_factorization_kernel(sf):
name="input_{}".format(sf.buffer)
)
- def _write_input(inputsf, index=0):
- # Write initial coefficients into buffer
- lfs = name_lfs(inputsf.element, inputsf.restriction, inputsf.component)
-
- sub_element = select_subelement(inputsf.element, inputsf.component)
- shape = sub_element.value_shape() + (inputsf.element.cell().geometric_dimension(),)
-
- if isinstance(sub_element, (VectorElement, TensorElement)):
- # Could be 0 but shouldn't be None
- assert inputsf.coeff_func_index is not None
-
- lfs_pym = lfs_child(lfs,
- (inputsf.coeff_func_index,),
- shape=shape_as_pymbolic(shape[:-1]),
- symmetry=inputsf.element.symmetry())
-
- leaf_element = sub_element
- if isinstance(sub_element, (VectorElement, TensorElement)):
- leaf_element = sub_element.sub_elements()[0]
-
- lfs = name_leaf_lfs(leaf_element, inputsf.restriction)
-
- basisiname = sumfact_iname(name_lfs_bound(lfs), "basis")
- container = inputsf.coeff_func(inputsf.restriction)
- if isinstance(sub_element, (VectorElement, TensorElement)):
- coeff = pymbolic_coefficient(container, lfs_pym, basisiname)
- else:
- coeff = pymbolic_coefficient(container, lfs, basisiname)
- assignee = prim.Subscript(prim.Variable(input_setup), (prim.Variable(basisiname),) + (index,))
- instruction(assignee=assignee,
- expression=coeff,
- depends_on=inputsf.insn_dep.union(insn_dep),
- tags=frozenset({"sumfact_stage{}".format(sf.stage)}),
- )
-
if sf.vectorized:
for i, inputsf in enumerate(sf.kernels):
- _write_input(inputsf, i)
+ inputsf.input.realize(inputsf, i, insn_dep)
else:
- _write_input(sf)
+ sf.input.realize(sf, 0, insn_dep)
insn_dep = insn_dep.union(frozenset({lp.match.Writes("input_{}".format(sf.buffer))}))
# Construct the direct_input for the FastDG case
direct_input = None
if get_option('fastdg') and sf.stage == 1:
- direct_input = sf.coeff_func(sf.restriction)
+ direct_input = sf.input.coeff_func(sf.input.restriction)
direct_output = None
if get_option('fastdg') and sf.stage == 3:
diff --git a/python/dune/perftool/sumfact/symbolic.py b/python/dune/perftool/sumfact/symbolic.py
index b254d5c6..88f884c8 100644
--- a/python/dune/perftool/sumfact/symbolic.py
+++ b/python/dune/perftool/sumfact/symbolic.py
@@ -14,6 +14,10 @@ import frozendict
import inspect
+class SumfactKernelInputBase(object):
+ pass
+
+
class SumfactKernelBase(object):
pass
@@ -24,13 +28,10 @@ class SumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable):
buffer=None,
stage=1,
preferred_position=None,
- restriction=0,
+ restriction=None,
within_inames=(),
insn_dep=frozenset(),
- coeff_func=None,
- coeff_func_index=None,
- element=None,
- component=None,
+ input=None,
accumvar=None,
):
"""Create a sum factorization kernel
@@ -101,6 +102,7 @@ class SumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable):
element: The UFL element
component: The treepath to the correct component of above element
accumvar: The accumulation variable to accumulate into
+ input: An SumfactKernelInputBase instance describing the input of the kernel
"""
# Assert the inputs!
assert isinstance(matrix_sequence, tuple)
@@ -112,6 +114,10 @@ class SumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable):
if preferred_position is not None:
assert isinstance(preferred_position, int)
+ if stage == 1:
+ assert isinstance(input, SumfactKernelInputBase)
+ restriction = input.restriction
+
if stage == 3:
assert isinstance(restriction, tuple)
@@ -161,7 +167,7 @@ class SumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable):
work on the same input coefficient (and are suitable for simultaneous
treatment because of that)
"""
- return (self.restriction, self.stage, self.coeff_func, self.coeff_func_index, self.element, self.component, self.accumvar)
+ return (self.input, self.restriction, self.accumvar)
#
# Some convenience methods to extract information about the sum factorization kernel
@@ -380,24 +386,9 @@ class VectorizedSumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable)
return self.kernels[0].within_inames
@property
- def coeff_func(self):
- assert len(set(k.coeff_func for k in self.kernels)) == 1
- return self.kernels[0].coeff_func
-
- @property
- def coeff_func_index(self):
- assert len(set(k.coeff_func_index for k in self.kernels)) == 1
- return self.kernels[0].coeff_func_index
-
- @property
- def element(self):
- assert len(set(k.element for k in self.kernels)) == 1
- return self.kernels[0].element
-
- @property
- def component(self):
- assert len(set(k.component for k in self.kernels)) == 1
- return self.kernels[0].component
+ def input(self):
+ assert len(set(k.input for k in self.kernels)) == 1
+ return self.kernels[0].input
@property
def accumvar(self):
--
GitLab