We saw in the first chapter that running the Euclidean
Algorithm in
can lead to a significant expression swell.
Let us try to estimate how coefficients grow.
That is, we want to estimate the maximal space
needed to store one coefficient of an intermediate polynomial.
Let
be the number of bits of a machine word.
we define
the length of
we have
![]() |
(1) |
with
and
we define
, the length of
and
.
of degree
such that all coefficients have denominator
have gcd 1
then we define the length of ![]() |
(2) |
of degree
can be represented with
![]() |
(3) |
be univariate polynomials over
.
.
be univariate polynomials over
holds.
Let ![]() |
(4) |
![]() |
(5) |
![]() |
(6) |
![]() |
(7) |
![]() |
(8) |
![]() |
(9) |
we obtain
![]() |
(10) |
and
.
such that
.
See [GG99].
These results provide us with an exponential upper bound for the
coefficient of the gcd computed by the Euclidean Algorithm.
But in fact a polynomial bound can be established.
Moreover there is an efficient modular approach for computing gcds in
and
as we shall see in the next sections.
Marc Moreno Maza