gnu compiler questions

[gaf at blu.org (Jerry Feldman)]

Good point Mike. One of the first things I did when I started my Polaroid 
contract a year or so ago was to build on my Alpha, which is 64 bit little 
endian. In the office they were using both Debian and Solaris on Intel 
hardware with the final target being Solaris/Sparc.  The portion of the 
product that I worked on was a btree database. One problems I ran into on 
the Sparc was not an endian problem, but a data alignment problem. On most 
RISC chips, data must be aligned on its natural boundary. A 64 bit integer 
must be aligned on an 8 byte boundary, 32 bits on a 4 bytes, etc. On the 
Alpha, the kernel was set to fixup alignment traps, but Solaris on the 
Sparc was not. The key data was set up in an array of key/offset pairs. If 
the key length was not a multiple of sizeof(long), accessing the offsets 
would result in a trap. I did not have time to restructure the key 
algorithms, so I had to revert to using memcpy instead.
Mike Bilow wrote:
> The tools work, but watch out for subtle machine dependencies.  Solaris is
> BSD-ish while AI/X is POSIX-ish.  Solaris usually runs on bigendian
> machines, while AI/X commonly runs on biendian machines.  Native word size
> also varies with platform, with AI/X sometimes on 32 or 64 bits, and
> occasionally on weirdo mainframe-legacy things like 24 or 48 bits.
-- 
Jerry Feldman <gaf at blu.org>
Boston Linux and Unix user group
http://www.blu.org


-
Subcription/unsubscription/info requests: send e-mail with
"subscribe", "unsubscribe", or "info" on the first line of the
message body to discuss-request at blu.org (Subject line is ignored).