This is the mail archive of the
cgen@sourceware.org
mailing list for the CGEN project.
Re: opc-file-path
Dave,
Thanks for the quick response! I will create a minimal opc file for now,
but your response has me puzzled. In the 2 CVS trees where I have
looked(cgen and sid) there are several cpu's without .opc files.
Specifically, arm, ia32, sh and I do not see the cris.cpu.
I am a little new to this so I must be doing something fundamentally
wrong... I would appreciate any light you can shed on my situation...
Furthermore, which source repository is the 'official' copy. Are the
following two links sync'd at all?
http://sourceware.org/cgi-bin/cvsweb.cgi/src/cgen/cpu/?cvsroot=cgen
http://sources.redhat.com/cgi-bin/cvsweb.cgi/src/cgen/cpu/?cvsroot=sid
Steve
On Mon, 2008-09-15 at 11:52 -0400, Dave Brolley wrote:
> Steve Woloschek wrote:
> > I am working on a port of a new processor core and have run into an
> > issue when running desc-cpu.c[c|h].
> >
> > ./desc-cpu.scm:1038:32: In expression (opc-file-path):
> > ./desc-cpu.scm:1038:32: Unbound variable: opc-file-path
> >
> > It appears the scripts assume there is an '[cpu].opc' file. I noticed
> > that several processors in the /cpu directory do not have a .opc file.
> >
> The only one I ses is the cris.cpu file. The others are all part of
> archirectures that do have a .opc file. Not sure why cris doesn't have
> one. I've never tried to build it.
> > Any thoughts on how to get around this feature?
> >
> I would suggest creating a minimal .opc file. $src/cgen/cpu/i960.opc
> seems to be a minimal example you can clone. You can probably get rid of
> the CGEN_PRINT_INSN macro and keep the rest. At some point you will
> probably want to optimize the CGEN_DIS_HASH* macros for you particular port.
>
> I hope this helps,
> Dave