Cray T3E port

Cray T3E port

[Roman Lechtchinsky]

Hi,

I hope to finish a first version of the Cray T3E port soon. What are the
requirements for integrating it into the GCC source tree? In particular,

  - Could someone please send me the copyright papers. What kind of
    disclaimer is needed from my employer? In particular, does it have to
    be in English?
  - Which CVS branch should I use?
  - GCC's config.guess says alpha-cray-unicosmk* while the master copy at
    ftp://ftp.gnu.org/pub/gnu/config/ has alphaev5-cray-unicosmk* as it
    should be. When will those two be merged?
  - I hope I haven't broken any other Alpha-based targets but I can't test
    that since I don't have access to other Alpha machines. Maybe someone
    could test that before I submit the patches?
  - How much functionality do I have to implement before the patch can be
    accepted? I've done only the C part but left out stuff like nested
    functions. I didn't touch any other languages (not even C++). The
    compiler bootstraps and the testsuite results are quite encouraging.
    Is that enough? I hope to add the missing parts in the future.
  - I can set up our T3E to run the tests once a week. Is that helpful and
    how do I go about it?
  - make check fails miserably because it throws away only the first 16
    bytes of the object files whereas the T3E uses something like 88 byte
    headers where it stores the compilation date among other things. Is
    there any workaround for this?

Bye

Roman

Re: Cray T3E port

[Joseph S. Myers]

On Sun, 6 May 2001, Roman Lechtchinsky wrote:

>   - make check fails miserably because it throws away only the first 16
>     bytes of the object files whereas the T3E uses something like 88 byte
>     headers where it stores the compilation date among other things. Is
>     there any workaround for this?

Write autoconf tests for how much of a header needs to be thrown away to
compare object files (including detecting any systems where removing some
fixed length of header isn't enough).  While you're about it, arrange for
the standard bootstrap/compare process to use the features only currently
used by "make gnucompare" in the gcc directory - that is, to use cmp
--ignore-initial=whatever if the installed cmp supports --ignore-initial.

(Other cmp implementations than GNU cmp may also support the same feature
with different syntax.  For example. man cmp on Solaris 8 says

     cmp [ -l ]  [ -s ]  file1  file2  [ skip1 ]  [ skip2 ]
...
                                                        skip1 and
     skip2 are initial byte offsets into file1 and file2  respec-
     tively,  and  may  be  either  octal or decimal; a leading 0
     denotes octal.

This could be autoconf-detected, as well.)

-- 
Joseph S. Myers
jsm28@cam.ac.uk

Re: Cray T3E port

[Andreas Jaeger]

Roman Lechtchinsky <rl@cs.tu-berlin.de> writes:

> [...]
> 
> I hope to finish a first version of the Cray T3E port soon. What are the
> requirements for integrating it into the GCC source tree? In particular,
> 
>   - Could someone please send me the copyright papers. What kind of
>     disclaimer is needed from my employer? In particular, does it have to
>     be in English?
>   - Which CVS branch should I use?

The head - that's what will go into gcc 3.1.

>   - GCC's config.guess says alpha-cray-unicosmk* while the master copy at
>     ftp://ftp.gnu.org/pub/gnu/config/ has alphaev5-cray-unicosmk* as it
>     should be. When will those two be merged?

Just ask for it ;-).

Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj

Re: Cray T3E port

[Gerald Pfeifer]

On Sun, 6 May 2001, Roman Lechtchinsky wrote:
>   - Could someone please send me the copyright papers. What kind of
>     disclaimer is needed from my employer?

I handled this by private mail.

>   - Which CVS branch should I use?

The main branch.

>   - I hope I haven't broken any other Alpha-based targets but I can't test
>     that since I don't have access to other Alpha machines. Maybe someone
>     could test that before I submit the patches?

That would make sense, though you could submit the patches and then ask
people to test.

>   - How much functionality do I have to implement before the patch can be
>     accepted? I've done only the C part but left out stuff like nested
>     functions. I didn't touch any other languages (not even C++). The
>     compiler bootstraps and the testsuite results are quite encouraging.
>     Is that enough?

I guess so.

>   - I can set up our T3E to run the tests once a week. Is that helpful and
>     how do I go about it?

Yes, that's definitely helpful. You could, for example, use the script
from the contrib subdirectory in the GCC CVS tree.

Gerald
-- 
Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/