"Pierre R\. Mai" <pmai@pmsf.de> writes:

> Hi!
>=20
> With the support of Johannes Gr=F8dem, I have created a port of CMU CL
> to NetBSD/x86 1.5.2, which should be available in the experimental
> directory of the CMU CL distribution site (and in the course of the
> next days on most mirrors).
>=20
> Note though that the port is quite untested, and has the following
> specific restrictions/quirks:
>=20
> - NetBSD 1.5.2 on x86 doesn't provide access to the faulting memory
>   address in SIGSEGV handlers, hence the generational GC can't use a
>   hardware-based write-barrier in order to reduce the need to track
>   pointers from older generations to newer ones, resulting in degraded
>   GC performance.
>=20
> - AFAICT NetBSD only allows mmaping strictly less than 128 MB on each
>   call, so that I had to hack the memory management routines to mmap
>   the various areas in chunks of 64MB.  This is likely to cause
>   decreased VM performance, and might have introduced bugs in certain
>   corner-cases.
>=20
>   Furthermore I don't know if the 64MB figure will work on all
>   systems, or is dependent on the machine/configuration in question.
>=20
> - The FFI code seems to work, but hasn't received more than cursory
>   testing...
>=20
> Regs, Pierre.
>=20
> --=20
> Pierre R. Mai <pmai@pmsf.de>                    http://www.pmsf.de/pmai/
>  The most likely way for the world to be destroyed, most experts agree,
>  is by accident. That's where we come in; we're computer professionals.
>  We cause accidents.                           -- Nathaniel Borenstein



CMUCL is available at cvs2.cons.org:/pub/lisp/cmucl with the above
mentioned experimental directory as a daughter of this root.

The more folk who might want to take a look at this, the better for
both projects.=20=20




                        John Towler