Function: nfsplitting
Section: number_fields
C-Name: nfsplitting
Prototype: GDG
Help: nfsplitting(P,{d}): defining polynomial over Q for the splitting field of
 P, that is the smallest field over which P is totally split. P can also be
 given by a nf structure. If d is given, it must be a multiple of the splitting
 field degree.
Doc: defining polynomial over~$\Q$ for the splitting field of
 $\var{P}\in\Q[x]$, that is the smallest field over which $P$ is totally split.
 If irreducible, the polynomial $P$ can also be given by a~\kbd{nf} structure,
 which is more efficient. If $d$ is given, it must be a multiple of the
 splitting field degree. Note that if $P$ is reducible the splitting field
 degree can be smaller than the degree of $P$.
 \bprog
 ? K = nfinit(x^3-2);
 ? nfsplitting(K)
 %2 = x^6 + 108
 ? nfsplitting(x^8-2)
 %3 = x^16 + 272*x^8 + 64
 ? S = nfsplitting(x^6-8) // reducible
 %4 = x^4+2*x^2+4
 ? lift(nfroots(subst(S,x,a),x^6-8))
 %5 = [-a,a,-1/2*a^3-a,-1/2*a^3,1/2*a^3,1/2*a^3+a]
 @eprog
 \noindent
 Specifying the degree of the splitting field can make the computation faster.
 \bprog
 ? nfsplitting(x^17-123);
 time = 3,607 ms.
 ? poldegree(%)
 %2 = 272
 ? nfsplitting(x^17-123,272);
 time = 150 ms.
 ? nfsplitting(x^17-123,273);
  *** nfsplitting: Warning: ignoring incorrect degree bound 273
 time = 3,611 ms.
 @eprog
 \noindent
 The complexity of the algorithm is polynomial in the degree $d$ of the
 splitting field and the bitsize of $T$; if $d$ is large the result will
 likely be unusable, e.g. \kbd{nfinit} will not be an option:
 \bprog
 ? nfsplitting(x^6-x-1)
 [... degree 720 polynomial deleted ...]
 time = 11,020 ms.
 @eprog
