This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFC 1/8] Language independent bits
On Tue, 17 Jan 2012 17:34:03 +0100, Doug Evans wrote:
> On Sun, Jan 15, 2012 at 12:34 PM, Jan Kratochvil <jan.kratochvil@redhat.com> wrote:
> Sure, but I think this is excessively nitpicky.
I agree with the part this was code moving, not new fields.
> Each element is (at least) 16 bytes in size and gdb has been using
> ints just fine so far.
Evaluating each memory size carefully whether it can or can not exceed 2GB is
too much burden on the human. Any such evaluation mistake then can mean
deployment failure like which happened to Red Hat customer(s). I do not find
the performance impact 4 -> 8 bytes measurable nowadays to justify such
optimization cost. It is and it will be already a lot of work to fix all the
existing wrong memory sizes as `int', trying to optimize them all properly for
4 vs. 8 bytes I do not find realistic.
It may make sense only in some special cases like struct partial_symbol or
struct main_type.
Thanks,
Jan