Improved performance for group isomorphism/automorphisms #896
Commits on Nov 23, 2016
-
FIX: Ensure generating sets remain consistent.
Conflicts: lib/autsr.gi
-
FIX: Corrections to new automorphism code
Ensure generating sets remain consistent. Fixed variable name in copied code bit Conflicts: lib/autsr.gi
-
-
ENHANCE: Increase hardcoded domain size limit for reducing down from GL.
Utilize a given `autactbase' to help with pc group automorphisms.
-
ENHANCE: SmallGeneratingSet behaves better in cases of huge generator…
… numbers. Instead of abelian invariants use factors of index of G'.
-
ENHANCE: Use better series, projective action on vectors.
Also ensure that automorphism group is automorphismgroup and fix comment. Two small fixes in ordering if there is trivial orbit on lines
-
Modify assertions that were not any longer valid. Required blank lines in test file
-
-
ENHANCE: Use characteristic subgroup classes to split large group on …
…module. Both on top level and for lower modules.
-
ENHANCE: AutomorphismWreathembedding only if S_n-normalizer is transi…
…tive (to avoid bad blowup)
-
-
-
FIX: Catch base case trivial/trivial in subgroup lattice
which does not translate by niceo when called with an automorphism group.
-
-
ENHANCE: Instead of stabilizing subspaces repeatedly in GL,
write down the subgroup of GL that stabilizes all subspaces (and subspace orbits of length 2). Also Added assertion.
-
FIX: Special cases in subspace stabilizer
Fix diagonal case and suplicate spaces in subspace stabilizer. Also avoid characteristic options to be accidentally dragged through.
-
ENHANCE: Minor omissions: SubgroupByPcgs, StabilizerOp
SubgroupByPcgs sets subgroup order. Method for Dtabilizer that allows feeding in fake domain for hash function
-
ENHANCE: Improvements to `CompatiblePairs'
Ensure that solvable groups do stabilizer with pcgs. If possible, represent module by list of numbers, not list of matrices, using permutation representation of matrix group.
-
ENHANCE: Avoid SpaceAndOrbitStabilizer being overwhelmed
by a huge number of spaces -- in this case, if they are different, they will intersect small. FIX: Correct subspace inclusion -- larger space is not necc. triag.
-
ENHANCE: Pc automorphisms prepare better for CompatiblePairs
Ensure that many tails of the pcgs of a factor group correspond to characteristic subgroups. This way `CompatiblePairs` can reduce the orbit calculation into smaller steps.
-
-
-
Changed heuristic for factor group representation
to be less gung-ho on intersections, if there are lots.
-
-
ENHANCE: Improvements to factor permutation code
Allow degree up to 100000 (computers are bigger now) before hard Use maximal subgroups in trying perm rep. (Do not use maxsub in factor if called from maxsub) FIX: another recursion avoidance.
-
ENHANCE: In PcSeries do not induce if the pcgs is being built
There are two different situations in which PcSeries is called. The first is a good pcgs and we happen to want the series -- this is the existing code. The second case is that the series is needed to do anything with the pcgs. In this case inducing caused bad recursions. Avoid this.
-
FIX: ClosurePermGroup long run
In some situations, it seems that ClosurePermGroup can iterate very long time with evrification failing repeatedly. In this situation we now simply force a new stabilizer chain computation from scratch, than trying to bend the (apparently confused) chain in the right way. This is not a beautiful fix but a workaround, albeit one that should have litlle cost implications.
-
ENHANCE: If radical is elementary abelian, use subspace stabilizer code.
ENHANCE: Declare `SpaceAndOrbitStabilizer'
-
ENHANCE: Inherit radical/factor and frattini information for factor g…
…roup. Relax threshold for new permrep force. FIX: Change reading order to avoid forward declaration Also ensure a maximal subgroups computation for smaller permutation degree stays out of expensive operations.
-
