Re: [discuss] AMD x86-64 GCC development

[Andreas Jaeger <aj at suse dot de>]

>>>>> Richard Kenner writes:

Richard>     Richard, I'm confused by your statement.  AFAIR you're on the x86-64
Richard>     lists since the beginning and should have read all the discussions
Richard>     that were going on.  

Richard> I'm on a half dozen lists.  I try to follow as much of what's going on
Richard> as I can, but I can't be sure I get everything correct.

Richard> This came up in a discussion today where I'd assumed that once the NDA
Richard> ended, the GCC x86-84 development had moved, but I didn't see the
Richard> corresponding messages, so I was confused.

Richard>     The gcc work for AMD x86-64 is done on a *public* CVS tree that has
Richard>     been started from the gcc.gnu.org CVS tree at the end of september and
Richard>     officially announced 

Richard> Why is this not being done in the main GCC tree?

The code we currently have is not yet stable.  Having it in a separate
tree - that's publically accessable via anoncvs by everybody - gives
the chance to change things.  I'm more a Glibc than a GCC developer
and discussed things with Ulrich Drepper.  He told me to get the port
in a good shape - and then submit it to glibc.  Jan Hubicka, the main
GCC x86-64 developer, discussed the GCC situation with Mark Mitchell
and AFAIK [1] they decided that x86-64 should not go into the GCC 3.0 tree
to not destabilize the ia32 backend [2]

Richard>     We've created separate CVS trees for (so far) GCC, Binutils, Glibc and
Richard>     the Linux kernel.  When the x86-64 port to GNU/Linux is in a stable
Richard>     state, the CVS tree will be merged with the mainstream tree of the
Richard>     different projects (according to the rules set up by the projects),
Richard>     e.g. with GCC's tree at gcc.gnu.org with submissions to
Richard>     patches@gcc-gnu.org with proper ChangeLogs etc.

Richard> But isn't this going to be a lot more work than if the development
Richard> were done from the beginning in the main GCC tree?

Nope, it's less work.  We don't have a tree that's not even able to
compile glibc [3].  We can easily change stuff without breaking
other's folks code - and since we do try to get the patches to the
generic code as soon as possible, only a small number of files will
need merging with the main tree.  As soon as the code is stable enough
the merging will be done - and then we might consider again where the
rest of the x86-64 GCC development will happen.

The only other solution that I would see is a separate branch on the
GCC tree but I personally don't see much advantage in this at the
moment.

Andreas

Footnotes: 
[1]  Mark and Jan might correct me here.

[2]  ia32 and x86-64 share a lot of code.

[3]  That's AFAIK the current situation in regard to gcc CVS mainline
     on ia32.
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj