I've merged the modular arithmetic branch, which allows the compiler
to substitute (effectively) machine arithmetic for generic arithmetic
under the right circumstances. This has been in the compiler on x86
for a while now, but people will want to verify on non-x86 that I've
done it right. Simple tests include running Paul's test suite,
checking for failures in arithmetic or logical operations; and
disassembling sb-md5:update-md5-block, to see if it's all been
compiled inline or not (it should be, modulo rotate-byte not being
optimized for a particular architecture).
Running Paul's torture tester on non-x86 would probably also be a good
thing, to pick up some backend bugs that might still be lurking; in
the course of the modular branch I fixed several fundamental backend
problems on the alpha, and I'm sure more remain.
http://www-jcsu.jesus.cam.ac.uk/~csr21/ +44 1223 510 299/+44 7729 383 757
(set-pprint-dispatch 'number (lambda (s o) (declare (special b)) (format s b)))
(defvar b "~&Just another Lisp hacker~%") (pprint #36rJesusCollegeCambridge)