Re: RFC: Optimization to write_register_bytes()

[Fernando Nasser <fnasser at cygnus dot com>]

Andrew Cagney wrote:
> 
> Fernando Nasser wrote:
> >
> > Andrew Cagney wrote:
> > >
> > > Fernando Nasser wrote:
> > > >
> > > > We can save some time (and fix a problem I have) if we make a simple assumption in
> > > > write_register_bytes().
> > >
> > > Would the problem go away if the problematic code didn't use
> > > write_register_bytes?
> > >

I suspect now that we are talking about different things.

I did no write any calls to write_register_bytes().  The code
that calls it could probably, for this specific architecture, call
write_register_gen() instead.  But that code is generic and I guess
the author used write_register_bytes() because of some other targets.

So, the answer is yes, if someone (where someone != me) wants to be
adventurous and do some cleanups on that code and get rid of the call
to write_register_bytes() and replace it to call(s) to
write_register_gen() I believe it would solve my problem.

Until someone gets rid of all calls to write_register_bytes()  in gdb
(if this is at all possible), I don't see why we cannot improve it
and solve my immediate problem.  It is a shame that gdb cannot yet show
MMX registers.

So, IMO, we should fix/optimize it until such date where someone can
find time to undertake the write_register_bytes() cleanup.






-- 
Fernando Nasser
Cygnus Solutions (a Red Hat company)    E-Mail:  fnasser@cygnus.com
2323 Yonge Street, Suite #300           Tel:  416-482-2661 ext. 311
Toronto, Ontario   M4P 2C9              Fax:  416-482-6299