This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Converting the gcc backend to a library?
- To: Edward Jason Riedy <ejr at CS dot Berkeley dot EDU>
- Subject: Re: Converting the gcc backend to a library?
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Tue, 11 Jan 2000 04:25:15 -0700
- cc: "Martin v. Loewis" <martin at loewis dot home dot cs dot tu-berlin dot de>, gcc at gcc dot gnu dot org
- Reply-To: law at cygnus dot com
In message <200001110910.BAA10070@lotus.CS.Berkeley.EDU>you write:
> And "Martin v. Loewis" writes:
> -
> - Can you please give me (or the list) a pointer to the software you are
> - talking about? (unless you'd violate your license agreement by doing
> - so, of course).
>
> Tera's compilers: www.tera.com
Really?
> To be fair, I'm probably being paranoid about the backend. I used
> some bugs I recalled to break the automatic parallelization, but
> those bugs could well have been re-invented (or different but triggered
> in similar ways). The hw architecture is extremely different from
> anything gcc supports.
Yes, the hardware architecture is radically different than what GCC normally
targets. And I do suspect you're being a little over-paranoid.
> And the Edison-Group-based front-end is a separate executable, just as
> cc1, et al. are. There are some extra c++ symbols in the tcc executable
> that aren't used in gcc.c, but they looked relatively unimportant. I
> didn't press for the source beyond the initial ``no, look, it's only
> gcc.c''; I didn't have the time and I don't think I communicated my
> request well. (And to think I wanted to help find their bugs. Silly me.)
If all they're using is gcc, then it's not that big of a deal (personal
opinion). They still have to give you that source under the terms of the
GPL, but gcc.c isn't really the compiler, it's just a driver that knows how
to start up cpp, cc1, as, ld, etc.
If indeed they were using GNU cc1 which was reading output from the EDG
front-end, then they must give you the source for cc1 under the terms of the
GPL.
> It's much more likely that they did exactly what RMS doesn't want
> initially...
I doubt it. More likely they never wrote code which reads/writes trees or
RTL to feed into GNU cc1 but just replaced cc1 with their own and continued
to use the gcc.c driver.
> IDA/SRC seems to like starting with gcc)
Of course. Why re-invent the wheel? When I was with SRC we certainly hacked
GCC as our needs dictated. But there wasn't anyone with a strong enough GCC
background to do the kinds of things that people have suggested here.
That doesn't mean that IDA/SRC couldn't have hired or grown these kinds of
hackers over the years (I left long ago), but I would be rather surprised.
jeff