Re: [4/4] Make SMS schedule register moves

2011-10-11 Thread Richard Sandiford
Ayal Zaks writes: >>> The issue of assigning stages to reg-moves is mostly relevant for >>> prolog and epilog generation, which requires and receives special >>> attention -- handled very nicely by ps_num_consecutive_stages! Note >>> that currently a simple boolean indicator for (the exceptional c

Re: [4/4] Make SMS schedule register moves

2011-10-10 Thread Ayal Zaks
On Mon, Oct 10, 2011 at 1:57 PM, Richard Sandiford wrote: > Ayal Zaks writes: >>> I agree it's natural to schedule moves for intra-iteration dependencies >>> in the normal get_sched_window way.  But suppose we have a dependency: >>> >>>   A --(T,N,1)--> B >>> >>> that requires two moves M1 and M2

Re: [4/4] Make SMS schedule register moves

2011-10-10 Thread Richard Sandiford
Ayal Zaks writes: >> I agree it's natural to schedule moves for intra-iteration dependencies >> in the normal get_sched_window way.  But suppose we have a dependency: >> >>   A --(T,N,1)--> B >> >> that requires two moves M1 and M2.  If we think in terms of cycles >> (in the SCHED_TIME sense), the

Re: [4/4] Make SMS schedule register moves

2011-10-09 Thread Ayal Zaks
On Wed, Sep 28, 2011 at 4:49 PM, Richard Sandiford wrote: > Ayal Zaks writes: >>> >> +  /* The cyclic lifetime of move->new_reg starts and ends at move->def >>> >> +     (the instruction that defines move->old_reg). >>> > >>> > So instruction I_REG_MOVE (new_reg=reg) must be scheduled before the

Re: [4/4] Make SMS schedule register moves

2011-09-28 Thread Richard Sandiford
Ayal Zaks writes: >> >> +  /* The cyclic lifetime of move->new_reg starts and ends at move->def >> >> +     (the instruction that defines move->old_reg). >> > >> > So instruction I_REG_MOVE (new_reg=reg) must be scheduled before the >> > next I_MUST_FOLLOW move/original-def (due to anti-dependence

Re: [4/4] Make SMS schedule register moves

2011-09-22 Thread Ayal Zaks
o: gcc-patches@gcc.gnu.org > >> > Cc: Ayal Zaks/Haifa/IBM@IBMIL > >> > Date: 30/08/2011 03:29 PM > >> > Subject: [4/4] Make SMS schedule register moves > >> > > >> > This is the move-scheduling patch itself.  It should be fairly > >

Re: [4/4] Make SMS schedule register moves

2011-09-22 Thread Richard Sandiford
Thanks as always for the review. Ayal Zaks writes: >> Richard Sandiford wrote on 30/08/2011 >> 03:29:26 PM: >> >> > From: Richard Sandiford >> > To: gcc-patches@gcc.gnu.org >> > Cc: Ayal Zaks/Haifa/IBM@IBMIL >> > Date: 30/08/2011 03:29 P

Re: [4/4] Make SMS schedule register moves

2011-09-22 Thread Ayal Zaks
> Richard Sandiford wrote on 30/08/2011 > 03:29:26 PM: > > > From: Richard Sandiford > > To: gcc-patches@gcc.gnu.org > > Cc: Ayal Zaks/Haifa/IBM@IBMIL > > Date: 30/08/2011 03:29 PM > > Subject: [4/4] Make SMS schedule register moves > > > > This

Re: [4/4] Make SMS schedule register moves

2011-08-30 Thread Richard Sandiford
Richard Guenther writes: > On Tue, Aug 30, 2011 at 2:43 PM, Richard Sandiford > wrote: >> Richard Guenther writes: >>> On Tue, Aug 30, 2011 at 2:29 PM, Richard Sandiford >>> wrote: This is the move-scheduling patch itself.  It should be fairly self-explanatory.  Let me know if it isn'

Re: [4/4] Make SMS schedule register moves

2011-08-30 Thread Richard Guenther
On Tue, Aug 30, 2011 at 2:43 PM, Richard Sandiford wrote: > Richard Guenther writes: >> On Tue, Aug 30, 2011 at 2:29 PM, Richard Sandiford >> wrote: >>> This is the move-scheduling patch itself.  It should be fairly >>> self-explanatory.  Let me know if it isn't, and I'll try to improve >>> the

Re: [4/4] Make SMS schedule register moves

2011-08-30 Thread Richard Sandiford
Richard Guenther writes: > On Tue, Aug 30, 2011 at 2:29 PM, Richard Sandiford > wrote: >> This is the move-scheduling patch itself.  It should be fairly >> self-explanatory.  Let me know if it isn't, and I'll try to improve >> the commentary. > > Can you add some testcases? I don't think it's an

Re: [4/4] Make SMS schedule register moves

2011-08-30 Thread Richard Guenther
On Tue, Aug 30, 2011 at 2:29 PM, Richard Sandiford wrote: > This is the move-scheduling patch itself.  It should be fairly > self-explanatory.  Let me know if it isn't, and I'll try to improve > the commentary. Can you add some testcases? Thanks, Richard. > One potentially controversial change

[4/4] Make SMS schedule register moves

2011-08-30 Thread Richard Sandiford
This is the move-scheduling patch itself. It should be fairly self-explanatory. Let me know if it isn't, and I'll try to improve the commentary. One potentially controversial change is to the way we handle moves in the prologue and epilogue. The current code uses a conservative check to decide