This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] Add Ada tasking support for hppa-unknown-linux-gnu
- From: Robert Dewar <dewar at adacore dot com>
- To: John David Anglin <dave at hiauly1 dot hia dot nrc dot ca>
- Cc: Arnaud Charlet <charlet at adacore dot com>, gcc-patches at gcc dot gnu dot org, laurent at guerby dot net, bosch at adacore dot com
- Date: Sun, 10 Apr 2005 12:34:50 -0400
- Subject: Re: [patch] Add Ada tasking support for hppa-unknown-linux-gnu
- References: <200504071719.j37HJJqw011140@hiauly1.hia.nrc.ca>
John David Anglin wrote:
* system-hpux.ads: Define Signed_Zeros to be True.
I'll let Geert and/or Robert comment on this part, but I believe that it
would not be correct to make this change.
Do you know why? The PA floating point co-processor is fully compatible
with ANSI/IEEE 754-1985. See page 8-7 of the PA-RISC 2.0 Architecture
book for the zero representation. The hardware supports signed zeros.
We certainly obvserved failures previously, we don't have to set the
value True if it is not supported properly. I do not know if these
failures were due to old versions of PA-RISC, (how early) or bad
code generation.
It's worth a try changing it and setting it true to see if we still
get failures I suppose.
Earlier versions of the PA architecture didn't support fneg in
hardware. There was a problem with the implementation of neg[sd]f2
on the PA. These patterns didn't handle signed zeros correctly.
However, this problem has been fixed for some time.
As I said, this change on hppa-unknown-linux-gnu fixed the failured
of the signed zero test in cxg1005. The test machine had a PA 2.0
processor.
no signed zero test should fail if signed zeros is false ???
Dave