lcc.cpp: fix abs -> fabs

Also some corrected information...

Timing UPDATED
--------------

Sorry the previous timing data was wrong.  Here are corrected values..

Here's what I get with g++ -O3 on two Linux machines with recent
versions of g++.  As always, you should take these with a grain of
salt.  Times per call in ns = nanoseconds.

                Fedora 31    Ubuntu 18
                g++-9.3.1    g++-7.5.0
                fwd   inv    fwd   inv
    old merc    207   461    217   522
    new merc    228   457    168   410
      etmerc    212   196    174   147

The new forward method is the 10% slower (resp 20% faster) on Fedora
31 (resp Ubuntu 18).  The new inverse method is the same speed (resp
20% faster) on Fedora 31 (resp Ubuntu 18).

Roughly speaking the speed comparison is a wash.  Maybe we should pay
attention more to the Fedora 31 results since these are with a newer
version of the compiler.  I would still make the argument that a 20%
time penalty (which in a full PROJ pipeline would probably be no more
than a 5% penalty) would be a worthwhile price to pay for a more
robust implementation of the projection.

0 files changed, 0 insertions, 0 deletions