The class SmallFpImpl is a very low level implementation class for fast
arithmetic in a small, prime finite field. It is not intended for use
by casual CoCoALib users, who should instead see the documentation in
QuotientRing (in particular the function NewZZmod), or possibly the
documentation in RingFp, RingFpLog, and RingFpDouble.
The class SmallFpImpl offers the possibility of efficient arithmetic in
small, prime finite fields. This efficiency comes at a cost: the interface
is rather unnatural. The emphasis is on speed rather than convenience;
this speed depends on many functions being inlined.
The overall structure is modelled on that of ring and RingElem:
namely, operations on values are via member functions of SmallFpImpl.
The class SmallFpImpl records the modulus, while the actual values are
of type SmallFpImpl::value, and record only the residue class. Also
see below for the special type SmallFpImpl::NonRedValue.
The ctor for a SmallFpImpl object takes 1 or 2 args:
SmallFpImpl(p) - create a SmallFpImpl for prime p; error if p is not prime, or too large.
SmallFpImpl(p,conv) - specify export convention conv: either SymmResidues or NonNegResidues
The default export convention is SymmResidues (unless changed in the GlobalManager).
This convention may be either
GlobalSettings::SymmResidues or GlobalSettings::NonNegResidues;
the default convention is determined by the GlobalManager.
Note if the first argment is of type SmallPrime then the constructor
skips testing for primality.
Let ModP be a SmallFpImpl object.
SmallFpImpl::IsGoodCtorArg(p) -- returns true if p is a valid SmallFpImpl ctor arg; otherwise false
SmallFpImpl::ourMaxModulus() -- returns largest ctor arg allowed by the implementation
ModP.myModulus() -- returns the prime p (as a long)
ModP.myMaxIters() -- see section on unnormalized computation
All operations (except for zero, one, IsZero, IsOne, == and !=)
must be effected by calling member functions of the SmallFpImpl class.
The member function myReduce is effectively a ctor. Here is a brief summary.
long n;
BigInt N;
BigRat q;
SmallFpImpl::value a, b, c;
a = zero(SmallFp); // equiv to a = ModP.myReduce(0);
b = one(SmallFp); // equiv to b = ModP.myReduce(1);
IsZero(a); // equiv to (a == ModP.myReduce(0))
IsOne(b); // equiv to (b == ModP.myReduce(1))
a == b; // test for equality
a != b; // logical negation of (a == b)
ModP.myReduce(n); // reduce mod p
ModP.myReduce(N); // reduce mod p
ModP.myReduce(q); // reduce mod p
ModP.myExportNonNeg(a); // returns the least non negative preimage (of type long), between 0 and p-1.
ModP.myExportSymm(a); // returns a symmetric preimage (of type long), between -p/2 and p/2.
ModP.myExport(a); // returns a preimage (of type long) between -p/2 and p-1; see note below!
ModP.myNegate(a); // -a mod p, additive inverse
ModP.myRecip(a); // inv(a), multiplicative inverse
ModP.myAdd(a, b); // (a+b)%p;
ModP.mySub(a, b); // (a-b)%p;
ModP.myMul(a, b); // (a*b)%p;
ModP.myDiv(a, b); // (a*inv(b))%p; where inv(b) is inverse of b
ModP.myPower(a, n); // (a^n)%p; where ^ means "to the power of"
ModP.myIsZeroAddMul(a,b,c) // a = (a+b*c)%p; result is (a==0)
ModP.myAddMul(a,b,c) // (a+b*c)%p
We suggest using the function myExport principally for values to be printed;
in other contexts we recommend using myExportNonNeg if possible.
Code calling myExport should assume only that the value returned is
between -p/2 and p-1; the actual range of return values is determined
by the convention specified when the SmallFpImpl object was constructed.
The normal mod p arithmetic operations listed above always produce a normalized result, but this normalization incurs a run-time cost. In some loops (e.g. for an inner product) it may be possible to compute several iterations before having to normalize the result.
SmallFpImpl supports this by offering the type SmallFpImpl::NonRedValue
for unnormalized values; this type is effectively an unsigned integer,
and such values may be added and multiplied without normalization
(but also without overflow checks!) using the usual + and * operators
(and also += and *=).
SmallFpImpl offers the following three functions to help implement
a delayed normalization strategy.
SmallFpImpl::NonRedValue a;
ModP.myNormalize(a); -- FULL normalization of a, result is a SmallFpImpl::value
ModP.myHalfNormalize(a); -- *fast*, PARTIAL normalization of a, result is a NonRedValue
ModP.myMaxIters(); -- see comment below
The value of myMaxIters() is the largest number of unnormalized
products (of normalized values) which may safely be added to a "half
normalized" value without risking overflow. The half normalization
operation is quick (at most a comparison and a subtraction).
Naturally, the final result must be fully normalized. See example
program ex-SmallFp1.C for a working implementation.
Most functions are implemented inline, and no sanity checks are
performed (except when CoCoA_DEBUG is enabled). The constructor
does do some checking.
SmallFpImpl::value_t must be an unsigned integral type; it is a
typedef to a type specified in CoCoA/config.H -- this should allow
fairly easy platform-specific customization.
This code is valid only if the square of myModulus can be represented
in a SmallFpImpl::value_t; the constructor checks this condition.
Most functions do not require myModulus to be prime, though division
becomes only a partial map if it is composite; and the function
myIsDivisible is correct only if myModulus is prime. Currently the
constructor rejects non-prime moduli.
The code assumes that each value modulo p is represented as the least
non-negative residue (i.e. the values are represented as integers in
the range 0 to p-1 inclusive). This decision is linked to the fact
that SmallFpImpl::value_t is an unsigned type.
The constants myResidueUPBValue and myIterLimit are to allow efficient
exploitation of non-reduced multiplication (e.g. when trying to
compute an inner product modulo p). See example program ex-SmallFp1.C
The return type of NumBits is int even though the result is
always non-negative -- I do not like unsigned values.
Should there be a myIsMinusOne function?