Re: [patch 6/9] genautomata chainsaw pass 1

["Vladimir N. Makarov" <vmakarov at redhat dot com>]

Bernd Schmidt wrote:

> Zack Weinberg wrote:
>
> > Okay, this is entirely tangential cleanups in genautomata and the code
> > it generates.  MAX_DFA_ISSUE_RATE is used nowhere, so kill it; so
> > there is no point in emitting a dead_lock_vect full of zeroes for
> > automata that have none (which is the normal case).  Also,
> > length_unit_log and max_insn_queue_index are never modified, so make
> > them const.
> >
> > zw
> >
> >     * genautomata.c (process_state_longest_path, max_dfa_issue_rate)
> >     (output_dfa_max_issue_rate, locked_states_num): Delete.
> >     (struct automaton): Add locked_states field.
> >     (output_dead_lock_vect): Save number of locked states in
> >     automaton->locked_states.  If it is zero, don't print anything.
> >     (output_max_insn_queue_index_def): Define max_insn_queue_index 
> > const.
> >     (output_internal_dead_lock_func): Only test the dead_lock_vect
> >     for automata that have locked states.
> >     (output_statistics): Always print the number of locked states for
> >     each automaton.  Calculate total number of locked states here.
> >     (write_automata): Don't call output_dfa_max_issue_rate.
> >     * genattrtab.c: Define length_unit_log as a const variable.
> >     * genattr.c: Don't emit MAX_DFA_ISSUE_RATE.  Declare
> >     max_insn_queue_index const.
> >     * final.c (length_unit_log): Declare const.
> >     * doc/tm.texi: Remove all references to MAX_DFA_ISSUE_RATE.
>
>
> This and part 8 probably ought to be reviewed by Vlad.  If he's OK 
> with them they are approved.

All the patch is ok for me.

The patch is hard to read because there are no function names in diff.  
When I wrote the generation of MAX_DFA_ISSUE_RATE, I thought that it 
could be used in modulo scheduling but it is not used so it can be 
removed.  If somebody needs it, it can be restored again.  I like 
removing zero lock state tables a lot.

Thanks for the patch.

Vlad