🚀 go-pugleaf

I want to port to ESIX V4 to do some code generation experiments. I run into
a series of problems and feel lost. What setup is  correct? The version I
have builds great .s output calls the as fine and core dumps. The .o files
can be used to produce good output.

Also what is really going on with bcopy() etc. When do we build it and when
do we use memcpy() cccp.c seems to want to use these but not build them. I
suspect I've got something #defined wrong.

By the way one of the code generation experiments I want to try is a new
keyword nodefault: for switch statements. It means if not compiled for
optomize and a default: occurs abort(). If compiled for optomize assume that
no default may occur and jump the rails if the programmer is wrong. I think
this will make C more usefull for simulators etc by reducing the overhead of
switch statements. In short it means "trust me" it must be one of the cases
don't check. If there are cases for 0, 4, 8, and 12 the expression may be
added to the start of a table with no compares or shifting, and a branch
taken. If the programmer said "trust me" and was wrong tough. This may also
reduce some extrainious warnings about uninitialized variables.

I hope to have something usefull to add to gcc soon.


P.S. I have a 4 color incremental garbage collector based on Baker's March
SIGPLAN article. It deals with arbitrary data objects provided there are
descriptors. It never moves objects and may do its work when the system
is not busy. So far it looks great, any intrest.

-- 
cef@mwc.com        Mark Williams Co       Philosophy is the art of
708-291-6700 x119  60 Revere Drive        misunderstanding terms in a
708-291-6750 FAX   Northbrook, Il 60062   specialized language, designed
Charles (Froggy) Fiterman		  for that purpose.