Edi Weitz <[EMAIL PROTECTED]> writes:
> The bad thing: Something that's relatively new in CMUCL[1] seems to be
> responsible for a significant slowdown in GC times.
[...]
> [1] compared to 18e-pre, built 2002-12-19 which I've been using -
> downloaded as a pre-built binary from some CMUCL mirror
There seems to be another (?) problem in 18e-pre on FreeBSD.
Running (TEST 100000) in 18e-pre2 on FreeBSD didn't end after 5
minutes. Every time I interrupted it it seemed to be in
(COMMON-LISP::CYCLE-COUNT/FLOAT). Using the dummy CYCLE-COUNT/FLOAT
function instead, which just returns 0.0, it works similar to HEAD
18d 18e-pre 18e-pre/no cycle count HEAD
GC time 4.8 - 11.8 12.9
Could the general slowdown also have something to do with cycle
counts? It's certainly not weak hash tables or control stack
checking, because that's not in 18e-pre, sorry for the confusion.
