This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: tiny tree.c update
- To: jh at suse dot cz (Jan Hubicka)
- Subject: Re: tiny tree.c update
- From: "John David Anglin" <dave at hiauly1 dot hia dot nrc dot ca>
- Date: Tue, 25 Sep 2001 09:54:51 -0400 (EDT)
- Cc: gcc-patches at gcc dot gnu dot org, jh at suse dot cz, rth at redhat dot com
> Can you describe the VAX problem in more detail?
The vax "ext" instructions generate illegal instruction faults
when the shift treated as unsigned is > 32 or when the position
specifier (again unsigned) is > 31.
The following code in fixup_unsigned_type causes problems when TI
nodes are built on the vax:
TYPE_MAX_VALUE (type)
= build_int_2 (precision - HOST_BITS_PER_WIDE_INT >= 0
? -1 : ((HOST_WIDE_INT) 1 << precision) - 1,
precision - HOST_BITS_PER_WIDE_INT > 0
? ((unsigned HOST_WIDE_INT) ~0
>> (HOST_BITS_PER_WIDE_INT
- (precision - HOST_BITS_PER_WIDE_INT)))
: 0);
The right shift in the above evaluates to -64 when HOST_BITS_PER_WIDE_INT
is 32: 32 - (128 - 32). The behaviour of the right shift operator is
undefined when the shift specified is negative and on the vax it
generates an illegal insn trap.
Dave
--
J. David Anglin dave.anglin@nrc.ca
National Research Council of Canada (613) 990-0752 (FAX: 952-6605)