Mirror of the gdb mailing list
 help / color / mirror / Atom feed
* Re: push and pop of FPU stack ; record st registers
@ 2009-05-23  6:17 paawan oza
  2009-05-23 12:01 ` Hui Zhu
  0 siblings, 1 reply; 5+ messages in thread
From: paawan oza @ 2009-05-23  6:17 UTC (permalink / raw)
  To: Hui Zhu; +Cc: gdb ml


yes. I can record all st0 to st7 registers. that is not a problem.
but, at every insn which push/pop FPU stack, if we record all st0 to st7, we will be occupying too much of record memory. 
(memory usage will be more for our records)

I was just thinking of optimizing it with some cache.
but at the moment I think I can go ahead with plain implementation of recording st0 to st7.

Regards,
Oza.



--- On Sat, 5/23/09, Hui Zhu <teawater@gmail.com> wrote:

> From: Hui Zhu <teawater@gmail.com>
> Subject: Re: push and pop of FPU stack ; record st registers
> To: "paawan oza" <paawan1982@yahoo.com>
> Cc: "gdb ml" <gdb@sourceware.org>
> Date: Saturday, May 23, 2009, 9:34 AM
> You cannot record st0 to st7?
> 
> Hui
> 
> On Fri, May 22, 2009 at 23:43, paawan oza <paawan1982@yahoo.com>
> wrote:
> >
> > Hi,
> >
> > the instructions such as;
> > fld1, which pushes fpu stack and insn faddp pops
> registers stack,
> > we need to record all st0 to st7 registers. because
> all will be changed on push/pop.
> > as pushing/popping changes all the registers, at such
> insns (which does push/pop) we record all st0 to st7
> registers that may affect the performance.
> >
> > currently I can not think of any alternative
> > (may be we cache registers all the time and we record
> only those which may change) but by doing that we only save
> memory. performance is still question :(
> >
> > Does anybody have any other suggestion ?
> >
> > Regards,
> > Oza.
> >
> >
> >
> >
> 


      


^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: push and pop of FPU stack ; record st registers
@ 2009-05-23 13:11 paawan oza
  0 siblings, 0 replies; 5+ messages in thread
From: paawan oza @ 2009-05-23 13:11 UTC (permalink / raw)
  To: Hui Zhu; +Cc: gdb ml


sure, I will focus on make it work first : )

--- On Sat, 5/23/09, Hui Zhu <teawater@gmail.com> wrote:

> From: Hui Zhu <teawater@gmail.com>
> Subject: Re: push and pop of FPU stack ; record st registers
> To: "paawan oza" <paawan1982@yahoo.com>
> Cc: "gdb ml" <gdb@sourceware.org>
> Date: Saturday, May 23, 2009, 5:31 PM
> I think make the function can run is
> the first job,
> 
> Hui
> 
> On Sat, May 23, 2009 at 14:17, paawan oza <paawan1982@yahoo.com>
> wrote:
> >
> > yes. I can record all st0 to st7 registers. that is
> not a problem.
> > but, at every insn which push/pop FPU stack, if we
> record all st0 to st7, we will be occupying too much of
> record memory.
> > (memory usage will be more for our records)
> >
> > I was just thinking of optimizing it with some cache.
> > but at the moment I think I can go ahead with plain
> implementation of recording st0 to st7.
> >
> > Regards,
> > Oza.
> >
> >
> >
> > --- On Sat, 5/23/09, Hui Zhu <teawater@gmail.com>
> wrote:
> >
> >> From: Hui Zhu <teawater@gmail.com>
> >> Subject: Re: push and pop of FPU stack ; record st
> registers
> >> To: "paawan oza" <paawan1982@yahoo.com>
> >> Cc: "gdb ml" <gdb@sourceware.org>
> >> Date: Saturday, May 23, 2009, 9:34 AM
> >> You cannot record st0 to st7?
> >>
> >> Hui
> >>
> >> On Fri, May 22, 2009 at 23:43, paawan oza <paawan1982@yahoo.com>
> >> wrote:
> >> >
> >> > Hi,
> >> >
> >> > the instructions such as;
> >> > fld1, which pushes fpu stack and insn faddp
> pops
> >> registers stack,
> >> > we need to record all st0 to st7 registers.
> because
> >> all will be changed on push/pop.
> >> > as pushing/popping changes all the registers,
> at such
> >> insns (which does push/pop) we record all st0 to
> st7
> >> registers that may affect the performance.
> >> >
> >> > currently I can not think of any alternative
> >> > (may be we cache registers all the time and
> we record
> >> only those which may change) but by doing that we
> only save
> >> memory. performance is still question :(
> >> >
> >> > Does anybody have any other suggestion ?
> >> >
> >> > Regards,
> >> > Oza.
> >> >
> >> >
> >> >
> >> >
> >>
> >
> >
> >
> >
> 


      


^ permalink raw reply	[flat|nested] 5+ messages in thread
* push and pop of FPU stack ; record st registers
@ 2009-05-22 15:43 paawan oza
  2009-05-23  4:04 ` Hui Zhu
  0 siblings, 1 reply; 5+ messages in thread
From: paawan oza @ 2009-05-22 15:43 UTC (permalink / raw)
  To: Hui Zhu; +Cc: gdb ml


Hi,

the instructions such as;
fld1, which pushes fpu stack and insn faddp pops registers stack, 
we need to record all st0 to st7 registers. because all will be changed on push/pop.
as pushing/popping changes all the registers, at such insns (which does push/pop) we record all st0 to st7 registers that may affect the performance.

currently I can not think of any alternative 
(may be we cache registers all the time and we record only those which may change) but by doing that we only save memory. performance is still question :( 

Does anybody have any other suggestion ? 

Regards,
Oza.


      


^ permalink raw reply	[flat|nested] 5+ messages in thread
end of thread, other threads:[~2009-05-23 13:11 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-05-23  6:17 push and pop of FPU stack ; record st registers paawan oza
2009-05-23 12:01 ` Hui Zhu
  -- strict thread matches above, loose matches on Subject: below --
2009-05-23 13:11 paawan oza
2009-05-22 15:43 paawan oza
2009-05-23  4:04 ` Hui Zhu

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox