On Tue, May 31, 2022 at 6:14 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> However, now that I've corrected that mistaken image ... I wonder if
> it could make sense to redefine relptr as self-relative? That ought
> to provide some notational savings since you'd only need to carry
> around the relptr's own address not that plus a base address.
> Probably not something to consider for v15 though.
I think that would be pretty hard to make work, since copying around a
relative pointer would change its meaning. Code like "relptr_foo x =
*y" would be broken, for example, but the compiler would not complain.
Or maybe I misunderstand your idea?
Also keep in mind that the major use case here is DSM segments, which
can be mapped at different addresses in different processes. Mainly,
we expect to store relative pointers in the segment to other things in
the same segment. Sometimes, we might read the values from there into
local variables - or maybe global variables - in code that is
accessing those DSM segments for some purpose.
There is little use for a relative pointer that can access all of the
address space that exists. For that, it is better to just as a regular
pointer.
--
Robert Haas
EDB: http://www.enterprisedb.com