This is the mail archive of the
mailing list for the GCC project.
RE: [PATCH] Improved boehm-gc support for s390/s390x
- From: "Ulrich Weigand" <Ulrich dot Weigand at de dot ibm dot com>
- To: "Boehm, Hans" <hans_boehm at hp dot com>
- Cc: gcc-patches at gcc dot gnu dot org, java-patches at gcc dot gnu dot org, boehm at acm dot org
- Date: Fri, 27 Sep 2002 20:13:11 +0200
- Subject: RE: [PATCH] Improved boehm-gc support for s390/s390x
Hans Boehm wrote:
>Is there a way to merge the S370 and S390 code? I think this patch still
>leaves some #defines to handle the S370/LINUX case, which is no longer
>possible? If it really makes sense to distinguish S370 (as in the old
>Amdahl UTS4 support) from S390, perhaps the LINUX OS case should only be
>handled for S390?
Actually, the patch removes the LINUX subcase from S370 (so that only
UTS4 remains), and handles LINUX only under the new S390 machine type.
>Unfortunately, I know very little about the differences between the
>architectures, and what naming makes sense.
I've no strong opinion either way; I wanted to make sure the inline
assemblies are only used on s390 because I'm not definite whether
the old S/370 hardware already has the COMPARE AND SWAP instruction
(I think some models had it and others not ...). [But as the assembly
is under an additional __GNUC__ guard, it wouldn't be used on UTS4
anyway, so this is no hard reason, but mostly cosmetic.]
In any case I've no access to such old hardware (nor to UTS4 for that
matter), and thus no way to test any changes, so maybe it's better to
just leave the old code alone ...
But if you prefer, I'll move the s390/s390x Linux support back to
the S370 machine type.
>Otherwise this looks good to me. Generally we've tried to keep my version
>and the one in gcc synchronized as much as possible. Things like this seem
>to get integrated into both at roughly the same time. Once you check it
>into the gcc tree, I can integrate it into my tree as well. (Currently I
>believe the two GC versions are a bit different, in that 6.1 has not been
>integrated into gcc. But they differ mostly in ways that are not relevant
So if you agree with the machine type issue, I'll go ahead and check it
in. Otherwise, I'll redo the patch.
Mit freundlichen Gruessen / Best Regards
Dr. Ulrich Weigand
Linux for S/390 Design & Development
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen
Phone: +49-7031/16-3727 --- Email: Ulrich.Weigand@de.ibm.com