Linear operator equations Lf = g are considered in the context of boun
dary element methods, where the operator L is equivariant, i.e., commu
tes with the actions of a given finite symmetry group. By introducing
a generalization of Reynolds projectors, a decomposition of the identi
ty operator is constructed, which in turn allows the decomposition of
the problem Lf = g into a finite number of symmetric subproblems. The
data function g does not need to possess any symmetry properties. It i
s shown that analogous reductions are possible for discretizations. An
explicit construction of the corresponding reduced system matrices is
given. This effects a considerable reduction in the computational com
plexity. For example, in the case of the isometry group of the 3-cube,
the computational complexity of a direct linear equation solver for f
ull matrices is reduced by 99.65 percent. Specific decompositions of t
he identity are given for most of the significant finite isometry grou
ps acting on R2 and R3.