This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFA:] change back name of initial rtl dump suffix to ".rtl".


> Date: Mon, 21 Feb 2005 11:57:25 +0100 (CET)
> From: Steven Bosscher <stevenb@suse.de>

> On Feb 21, 2005 11:29 AM, Hans-Peter Nilsson <hans-peter.nilsson@axis.com> wrote:
> > IMHO, it's generally just irritating to have different layers of
> > dump information.
> 
> All tree dumps have this.  I think it's very useful because
> it gives you more insight of what is going on internally. Is
> that not what the debug dumps exist for?

The dumps exists to give that information, not a subset and a
need to figure out more options.

> I don't see your point.  The -da option still exists, and
> those folks that want to see more details can use the -fdump*
> options.  Are you trying to deny the latter group this new
> ability of the compiler?

Why the suggestive question?  I want *all* the information to be
there, right from the start, with a single option that says
"produce a dump from this pass" (or all passes).

> We've killed the old tree-to-rtl pass and replaced it with
> an entirely new pass that we have named "expand", which is
> the best name we could think of to refect what the pass does:
> Expand trees to RTL.  It's quite obvious for anyone working
> on GCC for some time that "expand" can only mean one thing.

It's quite obvious that renaming "expand" to "rtl" isn't adding
any information; just as obvious to new and old GCC hackers.  Or
maybe just to old GCC hackers?

> As I said, anyone who's worked on GCC for a while knows
> that "expand" can only mean tree-to-RTL.

Why do you repeat that?  I've worked on GCC for quite a while
and I don't think "00.expand" is more revealing than "00.rtl".

> > I don't agree, they rather reflect whatever transformation is
> > done.
> 
> Then you contradict yourself.  The ".rtl" name could be *any*
> RTL dump, the name does not reflect what transformation is
> done at all.

There's no contradiction.  It's "initial *rtl*", perhaps
"transformed from tree".

> > That's not really a compelling reason to make it optional; we're
> > talking dumps that are supposed to be maximally informational!
> 
> No, they're supposed to provide just the information that is
> useful to whoever is looking at the dumps.  Sometimes the more
> elaborate dumps are useful, most of the time they are not.

When the elaborate info can sometimes be unhelpful, it's better
to not look at the extra information than to have to invoke the
compiler again.  In any event, for information purporting to be
the initial tree-to-rtl dump, seeing the trees would likely be
more helpful than redundant.  When debugging GCC, it's not rare
that re-starting cc1 with other options isn't what you want at
times.  That's obvious to anyone who's worked on GCC for a
while. 8-)

brgds, H-P


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]