ReductionCogBase
is an abstract class to perform a full reduction:
it contains two parts:
IgnoredPPs
summands whose PPs are to be ignored
Active
the part which will be reduced
Thanks to the limited operations allowed on a ReductionCog
, all PPs in
IgnoredPPs
are guaranteed bigger than those in the Active
part.
With a ReductionCog
F
you can compute:
ActiveLPP(F)
the LPP of the Active
part
IsActiveZero(F)
is the Active
part zero?
F.myMoveToNextLM()
move the LM
of the Active
part to the IgnoredPPs
F.myReduce(f)
reduce the Active
part with f
F.myAssignReset(f)
the Active
part gets f
; f
and IgnoredPPs
get 0
F.myAssignReset(f, fLen)
same as above but faster for geobucket implementation
F.myRelease(f)
F
gets the total value of f
; f
gets 0
F.myOutput(out)
The idea is that LM will be reduced first; if the result is not 0 it
will be "set aside and ignored" and the new LM of the Active
part will be
reduced, and so on.
The result of myReduce
is defined up to an invertible (in the
coefficient ring) constant factor.
Constructors are
ReductionCog NewRedCogPolyField(const SparsePolyRing& P); ReductionCog NewRedCogPolyGCD(const SparsePolyRing& P); ReductionCog NewRedCogGeobucketField(const SparsePolyRing& P); ReductionCog NewRedCogGeobucketGCD(const SparsePolyRing& P);
In the "GCD" version, myRelease
makes poly content free.
In the "Field" version: myRelease
does NOT make poly monic.
... I can't remember why I made this choice....
ReductionCog F = ChooseReductionCogGeobucket(myGRingInfoValue); F->myAssignReset(f, fLen); while ( !IsActiveZero(F) ) { (..) // find reducer g or break F->myReduce(g); } F->myRelease(f);
In general the geobucket implementation are to be preferred
RedCog::PolyFieldImpl
RingElem
.
RedCog::PolyGCDImpl
RedCog::GeobucketFieldImpl
IgnoredPPs
and a
geobucket for the Active
part
RedCog::GeobucketGCDImpl
IgnoredPPs
and a
geobucket for the Active
part
two coefficients [RingElem] and a counter