From: "Richard Guenther" <richard.guenther@gmail.com>
To: "Richard Kenner" <kenner@vlsi1.ultra.nyu.edu>
Cc: ebotcazou@adacore.com, gcc-patches@gcc.gnu.org, iant@google.com,
matz@suse.de
Subject: Re: [PATCH] Fix optimization regression in constant folder
Date: Mon, 01 Oct 2007 14:15:00 -0000 [thread overview]
Message-ID: <84fc9c000710010715q6604f8c4k193ff9963366e32c@mail.gmail.com> (raw)
In-Reply-To: <10710011319.AA12088@vlsi1.ultra.nyu.edu>
On 10/1/07, Richard Kenner <kenner@vlsi1.ultra.nyu.edu> wrote:
> > It matches undefined overflow && wrapping overflow, as for infinite precision
> > overflow really cannot happen.
>
> But it can happen: in the case of illegal programs when the programmer tries
> to make an object that's too big.
>
> > You say you need special semantics because otherwise optimization won't work.
>
> Actually, I didn't say that. I said that sizetypes *already have* and
> always have had different semantics than user-visible types and that trying
> to map those semantics onto a new set of semantics being defined for
> user-visible types is not correct.
>
> > This needs testcases.
>
> Eric and I have both given you an example of an optimization that should
> apply to sizetypes but which won't if you try to change their semantics to
> match those of a user-visible type.
>
> > Otherwise just mapping sizetype semantics to an _existing_ semantic makes
> > most sense as it reduces diversity of semantics and causes less
> > confusion.
>
> I'm not proposing changing the semantics of sizetypes! Quite the contrary:
> I'm saying that we keep them as they are and *not* try to map them to these
> new flags which are meant for user-visible types.
>
> > Look at all the places we _don't_ specialize sizetypes! You want to audit
> > all these?
>
> I don't follow. All these optimizations used to be done for sizetypes and
> I'm saying they need to continue to work. It's the proposal to use these
> new flags that's changing things. And I also don't understand what "audit"
> means in this context. The semantics of these various types (sizetypes and
> the different user-visible type semantics) only differ in overflow
> handling. So we only need to "audit" those cases where overflow tests are
> involved.
I see exactly 6 places in the middle-end that check TYPE_IS_SIZETYPE. All
of them in fold-const.c. What I say is that if sizetypes are worth to
be special
for optimization, why don't optimization passes like VRP or tree-ssa-reassoc.c
not check for TYPE_IS_SIZETYPE? I say, that TYPE_IS_SIZETYPE can't
be very important as its only used in _very_ few places. Instead I
ask for testcases
where it _matters_ that we do the optimization in the few places in fold-const.c
that check for TYPE_IS_SIZETYPE because no other optimization pass does
the required transformation. I argue we should fix optimization passes instead
of fold. Or if you rely on sizetype in a so different way in Ada than other
frontends maybe using fold () to simplify sizetype operations is not the best
way to ensure proper optimization?
Two of the 6 places are to check if sizetypes are sign-extended, one is to not
mark results TREE_OVERFLOW in int_const_binop for sizetypes. One is in
extract_multdiv_1 and it needs a comment - it's not exactly clear why sizetype
is special here. One is with widening which is somehow special for sizetypes,
one is for optimizing div/mult sequences. That's it.
You seem to have a testcase that shows that re-association is very important
for you for sizetypes. Why can't you produce something out of it that you
can show us?
Richard.
next prev parent reply other threads:[~2007-10-01 14:15 UTC|newest]
Thread overview: 140+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-09-17 14:35 Eric Botcazou
2007-09-17 14:45 ` Richard Guenther
2007-09-17 14:56 ` Eric Botcazou
2007-09-17 15:13 ` Richard Guenther
2007-09-18 11:12 ` Eric Botcazou
2007-09-18 11:22 ` Richard Guenther
2007-09-18 13:17 ` Richard Kenner
2007-09-19 11:33 ` Eric Botcazou
2007-09-19 14:40 ` Richard Guenther
2007-09-19 14:50 ` Richard Kenner
2007-09-19 15:50 ` Eric Botcazou
2007-09-29 10:27 ` Eric Botcazou
2007-09-29 14:59 ` Richard Guenther
2007-09-29 15:12 ` Eric Botcazou
2007-09-29 15:38 ` Richard Guenther
2007-09-29 17:46 ` Richard Guenther
2007-09-29 18:43 ` Eric Botcazou
2007-09-29 20:52 ` Richard Guenther
2007-09-30 14:38 ` Eric Botcazou
2007-09-30 16:01 ` Eric Botcazou
2007-09-30 16:03 ` Richard Guenther
2007-09-30 16:36 ` Eric Botcazou
2007-09-30 17:37 ` Richard Guenther
2007-09-30 18:53 ` Eric Botcazou
2007-09-30 17:59 ` Richard Kenner
2007-09-30 18:38 ` Richard Guenther
2007-09-30 21:16 ` Richard Kenner
2007-09-30 22:10 ` Eric Botcazou
[not found] ` <m3r6kf3gyx.fsf@localhost.localdomain>
[not found] ` <200710010127.59677.ebotcazou@adacore.com>
2007-10-01 0:10 ` Ian Lance Taylor
2007-10-01 0:51 ` Richard Kenner
2007-10-01 1:54 ` Michael Matz
2007-10-01 2:53 ` Richard Kenner
2007-10-01 3:25 ` Michael Matz
2007-10-01 3:40 ` Richard Kenner
2007-10-01 9:12 ` Richard Guenther
2007-10-01 12:29 ` Richard Kenner
2007-10-01 12:40 ` Richard Guenther
2007-10-01 13:07 ` Eric Botcazou
2007-10-01 13:15 ` Richard Kenner
2007-10-01 14:15 ` Richard Guenther [this message]
2007-10-01 14:23 ` Richard Kenner
2007-10-01 14:29 ` Richard Guenther
2007-10-01 14:35 ` Richard Kenner
2007-10-01 17:51 ` Ian Lance Taylor
2007-10-01 19:32 ` Richard Kenner
2007-10-01 19:51 ` Ian Lance Taylor
2007-10-01 20:28 ` Richard Kenner
2007-10-02 13:58 ` Michael Matz
2007-10-03 6:41 ` Eric Botcazou
2007-10-03 7:00 ` Ian Lance Taylor
2007-10-03 7:10 ` Eric Botcazou
2007-10-03 14:18 ` Ian Lance Taylor
2007-10-03 15:53 ` Eric Botcazou
2007-10-04 5:20 ` Michael Matz
2007-10-10 17:31 ` Mark Mitchell
2007-10-10 22:43 ` Richard Kenner
2007-10-11 0:40 ` Mark Mitchell
2007-10-11 1:02 ` Richard Kenner
2007-10-11 15:06 ` Michael Matz
2007-10-11 15:10 ` Richard Guenther
2007-10-11 15:55 ` Michael Matz
2007-10-11 15:57 ` Richard Guenther
2007-10-11 16:02 ` Richard Kenner
2007-10-11 15:59 ` Richard Kenner
2007-10-11 16:02 ` Richard Guenther
2007-10-11 16:08 ` Richard Kenner
2007-10-11 16:14 ` Richard Guenther
2007-10-11 16:37 ` Richard Kenner
2007-10-11 16:44 ` Richard Guenther
2007-10-11 16:50 ` Richard Kenner
2007-10-11 15:32 ` Richard Kenner
2007-10-11 16:03 ` Michael Matz
2007-10-11 16:06 ` Richard Guenther
2007-10-11 16:19 ` Richard Kenner
2007-10-11 16:25 ` Michael Matz
2007-10-11 16:12 ` Richard Kenner
2007-10-11 16:38 ` Michael Matz
2007-10-11 16:47 ` Richard Kenner
2007-10-11 17:04 ` Michael Matz
2007-10-11 17:17 ` Ian Lance Taylor
2007-10-11 18:41 ` Michael Matz
[not found] ` <10710111850.AA27 912@vlsi1.ultra.nyu.edu>
[not found] ` <10710111850.AA27 912@vlsi1.ultra.nyu.edu>
[not found] ` <10710111850.AA27 912@vlsi1.ultra.nyu.edu>
[not found] ` <10710111850.AA27 912@vlsi1.ultra.nyu.edu>
[not found] ` <10710111850.AA27 912@vlsi1.ultra.nyu.edu>
[not found] ` <10710111850.AA27 912@vlsi1.ultra.nyu.edu>
2007-10-11 18:45 ` Richard Kenner
2007-10-11 19:10 ` Michael Matz
2007-10-11 19:17 ` Richard Kenner
2007-10-11 19:30 ` Michael Matz
2007-10-11 20:06 ` Richard Kenner
2007-10-11 21:24 ` Ian Lance Taylor
2007-10-12 2:36 ` Richard Kenner
2007-10-12 4:40 ` Ian Lance Taylor
2007-10-12 4:56 ` Mark Mitchell
2007-10-12 12:30 ` Richard Kenner
2007-10-12 17:48 ` Mark Mitchell
2007-10-12 18:08 ` Richard Kenner
2007-10-12 18:37 ` Mark Mitchell
2007-10-12 18:51 ` Richard Kenner
2007-10-12 20:13 ` Richard Guenther
2007-10-14 11:09 ` Eric Botcazou
2007-10-12 11:42 ` Richard Kenner
2007-10-14 10:31 ` Eric Botcazou
2007-10-14 11:24 ` Richard Guenther
2007-10-14 11:58 ` Eric Botcazou
2007-10-14 12:16 ` Richard Kenner
2007-10-14 13:29 ` Richard Guenther
2007-10-15 2:12 ` Richard Kenner
2007-10-15 7:33 ` Michael Matz
2007-10-15 11:36 ` Richard Kenner
2007-10-15 13:27 ` Richard Guenther
2007-10-15 13:28 ` Richard Kenner
2007-10-14 16:07 ` Ian Lance Taylor
2007-10-14 16:27 ` Richard Guenther
2007-10-15 3:26 ` Richard Kenner
2007-10-15 9:12 ` Richard Guenther
2007-10-15 11:41 ` Richard Kenner
2007-10-15 13:15 ` Richard Guenther
2007-10-15 13:17 ` Richard Kenner
2007-10-15 3:29 ` Richard Kenner
2007-10-15 2:17 ` Richard Kenner
2007-10-15 5:21 ` Ian Lance Taylor
2007-10-15 11:01 ` Richard Kenner
2007-10-11 17:21 ` Richard Kenner
2007-10-11 16:16 ` Richard Kenner
2007-10-11 16:51 ` Michael Matz
2007-10-11 16:57 ` Richard Kenner
2007-10-11 17:06 ` Richard Guenther
2007-10-11 17:24 ` Richard Kenner
2007-10-11 17:31 ` Michael Matz
2007-10-11 17:54 ` Richard Kenner
2007-10-12 8:20 ` Richard Guenther
2007-10-12 11:47 ` Paolo Bonzini
2007-10-12 12:44 ` Richard Kenner
2007-10-12 12:49 ` Richard Guenther
2007-10-11 15:57 ` Mark Mitchell
2007-10-11 16:01 ` Richard Kenner
2007-10-11 16:03 ` Richard Guenther
2007-10-11 16:16 ` Richard Kenner
2007-10-11 16:11 ` Mark Mitchell
2007-10-11 16:29 ` Richard Kenner
2007-10-01 6:07 ` Eric Botcazou
2007-10-01 6:07 ` Eric Botcazou
2007-10-01 8:58 ` Richard Guenther
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=84fc9c000710010715q6604f8c4k193ff9963366e32c@mail.gmail.com \
--to=richard.guenther@gmail.com \
--cc=ebotcazou@adacore.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=iant@google.com \
--cc=kenner@vlsi1.ultra.nyu.edu \
--cc=matz@suse.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).