RE: Note handling in haifa-sched.

["Virgil Palanciuc" <Virgil dot Palanciuc at cs dot pub dot ro>]

> On Fri, May 12, 2000 at 12:25:17PM +0300, Dragos Badea wrote:
> > 1.  We must admit that the machine description might not be
> quite correct;
> > it was not written by us. Anyhow, should this influence the scheduler?
>
> Not really.  However, it's certainly not possible to track
> down the problem with the information given.
> r~

     I don't really know what other information we should give.We whacked
our local copy of gcc and it is different some aspects of what you know. But
I don't think it's practical to send you the entire haifa-sched file, along
with the machine description and the additional code to exploit our custom
extentions to this file. I don't think that you have the time or are willing
to solve our problem.
     I understand that we might have re-done in a different manner something
already implemented . Regardless of the ineficiency of our approach, the
question still stands: why isn't it working? In fact, the question is : what
code in gcc is responsible for moving our notes?

      I re-attach the code snippipt:

# [....]
#
# while(sched_target_n_insns<target_n_insns){
#   int b1;
#   clock_var++;   /* This part is unchanged, I wrote it to give you a hint
as
# to where
# 				the code that emits the notes is placed */
# [...]
#
#   for(....){
#    [....]
#   } /* end of for */
#
#    if(reload_completed)
#
last_emitted_note=emit_note_after(NOTE_INSN_SET_END,last_sched_insn);
# /* don't worry about the value of last_sched_insn and last_emitted_note -
# initially,
# the note is emitted in the proper position - I checked that (using a
printf)
# */
#
# [...]
#   } /* end while */
#
#   After the 'while' loop, the notes are moved (I checked that, too). The
# problem is that I cannot understand what code in the while loop is moving
my
# notes.
#