Attached is a patch to use array instead of hash table to convert
character codes between EUC-JP and Unicode.
With this patch on my configuration (0.9.6.3, FreeBSD 6.0-RC1, x86),
25264128 -> 24875008 (380KB smaller)
(time (dotimes (i 100) (load "eucjp.impure.lisp"))) in tests directory:
121.85585 -> 119.62863 seconds of user run time (1.86% faster)