Cache the flopcost of a sumfact kernel object

python/dune/codegen/sumfact/permutation.py

@@ -3,6 +3,7 @@
 import itertools
 
 from dune.codegen.options import get_option
+from dune.codegen.sumfact.tabulation import quadrature_points_per_direction
 from dune.codegen.ufl.modified_terminals import Restriction
 
 

python/dune/codegen/sumfact/symbolic.py

@@ -14,7 +14,7 @@ from dune.codegen.sumfact.permutation import (flop_cost,
                                               sumfact_cost_permutation_strategy,
                                               sumfact_quadrature_permutation_strategy,
                                               )
-from dune.codegen.sumfact.tabulation import BasisTabulationMatrixBase, BasisTabulationMatrixArray
+from dune.codegen.sumfact.tabulation import BasisTabulationMatrixBase, BasisTabulationMatrixArray, quadrature_points_per_direction
 from dune.codegen.loopy.target import dtype_floatingpoint, type_floatingpoint
 from dune.codegen.loopy.vcl import ExplicitVCLCast, VCLLowerUpperLoad
 from dune.codegen.tools import get_leaf, maybe_wrap_subscript, remove_duplicates
@@ -564,6 +564,7 @@ class SumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable):
 
         # Precompute and cache a number of keys
         self._cached_cache_key = None
+        self._cached_flop_cost = {}
 
     #
     # The methods/fields needed to get a well-formed pymbolic node
@@ -824,7 +825,10 @@ class SumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable):
     def operations(self):
         """ The total number of floating point operations for the kernel
         to be carried out """
-        return flop_cost(self.matrix_sequence_cost_permuted)
+        qp = quadrature_points_per_direction()
+        if qp not in self._cached_flop_cost:
+            self._cached_flop_cost[qp] = flop_cost(self.matrix_sequence_cost_permuted)
+        return self._cached_flop_cost[qp]
 
 
 # Extract the argument list and store it on the class. This needs to be done
@@ -873,6 +877,7 @@ class VectorizedSumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable)
 
         # Precompute and cache a number of keys
         self._cached_cache_key = None
+        self._cached_flop_cost = {}
 
     def __getinitargs__(self):
         return (self.kernels, self.horizontal_width, self.vertical_width, self.buffer, self.insn_dep)
@@ -1135,4 +1140,7 @@ class VectorizedSumfactKernel(SumfactKernelBase, ImmutableRecord, prim.Variable)
     def operations(self):
         """ The total number of floating point operations for the kernel
         to be carried out """
-        return flop_cost(self.matrix_sequence_cost_permuted)
+        qp = quadrature_points_per_direction()
+        if qp not in self._cached_flop_cost:
+            self._cached_flop_cost[qp] = flop_cost(self.matrix_sequence_cost_permuted)
+        return self._cached_flop_cost[qp]