This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] CFG dump patch.
> In message <1076003823.12370.18.camel@p4>, Andrew MacLeod writes:
> >
> >Does anyone have any objections to this patch? I find that the pred/succ
> >information makes it hard for me to look at large listings sometimes
> >because there is a lot of cruft that I dont normally care about in the
> >pred/succ chains. All I did here was make the block print routine
> >summarize the pred/succ blocks when TDF_SLIM is passed as a flags.
> >
> >So instead of seeing:
> > # BLOCK 29
> > # PRED: 27 [50.0%] (false,exec) 26 [50.0%] (false,exec) 24 [50.0%]
Perhaps it may sense to avoid the probabilities from being printed in
PRED lists at all. These are outgoing probabilities, not incomming
probabilities so they don't have that good meaning in this context.
Honza
> >(false,exec) 25 [100.0%] (fallthru,exec) 28 [100.0%] (fallthru,exec)
> >
> >when I add 'slim', to the command line -fdump-tree-slim-...
> >I get:
> > # BLOCK 29
> > # PRED: 27 26 24 25 28
> >
> >Which makes it much easier to find the other information I am interested
> >in.
> >
> >Andrew
> >
> >
> > * tree-pretty-print.c (dump_bb_header): Allow TDF_SLIM printing.
> > (dump_bb_end): Allow TDF_SLIM printing.
> > (dump_generic_bb_buff): Add flags parameter to dump_bb_end.
> No objections from me. The more verbose output is at the moment interesting
> to me, but generally it hasn't been. Having it selectable would IMHO
> be a good thing.
>
> BTW, can you throw your latest coalescing changes my way. They're helpful
> to work with when analyzing what's going on as they make the tree-ssa
> output more closely resemble mainline in regards to how registers are
> used.
>
>
>
>
> jeff
>
>
>