From: Richard Biener <richard.guenther@gmail.com>
To: Marc Glisse <marc.glisse@inria.fr>
Cc: "Bin.Cheng" <amker.cheng@gmail.com>,
gcc-patches List <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH GCC][1/2]Feed bound computation to folder in loop split
Date: Wed, 26 Jul 2017 11:46:00 -0000 [thread overview]
Message-ID: <CAFiYyc2Rw4jkd1c6egiHPgMfkJzsstV68=6wrkOizAnDPnX4Dg@mail.gmail.com> (raw)
In-Reply-To: <CAFiYyc39o8eb_hxzqw9ZCAHaY5gY+xNWX00X+ME0tKRdjYRHgA@mail.gmail.com>
On Wed, Jul 26, 2017 at 9:48 AM, Richard Biener
<richard.guenther@gmail.com> wrote:
> On Tue, Jul 25, 2017 at 7:45 PM, Marc Glisse <marc.glisse@inria.fr> wrote:
>> On Tue, 25 Jul 2017, Richard Biener wrote:
>>
>>>> I think we need Richard to say what the intent is for the valueization
>>>> function. It is used both to stop looking at defining stmt if the return
>>>> is
>>>> NULL, and to replace/optimize one SSA_NAME with another, but currently it
>>>> seems hard to prevent looking at the defining statement without
>>>> preventing
>>>> from looking at the SSA_NAME at all.
>>>
>>>
>>> Yeah, this semantic overloading is an issue. For gimple_build we have
>>> nothing
>>> to "valueize" but we only use it to tell genmatch that it may not look at
>>> the
>>> SSA_NAME_DEF_STMT.
>>>
>>>> I guess we'll need a fix in genmatch...
>>>
>>>
>>> I'll have a look tomorrow.
>>
>>
>> My impression yesterday was that we could replace the current do_valueize
>> wrapper by 2 wrappers (without touching the valueize callbacks):
>> - may_check_def_stmt, which returns a bool corresponding to the current
>> do_valueize != NULL_TREE
>> - maybe_valueize, which tries to valueize, but if it gets a NULL_TREE, it
>> returns its argument unchanged.
>>
>> Not very confident about it though.
>
> Note I've been there in the past (twice I think) but always ran into existing
> latent issues. So hopefully we've resolved those, I'm testing the following
> simplified variant of what I had back in time.
>
> It'll produce
>
> switch (TREE_CODE (op0))
> {
> case SSA_NAME:
> if (gimple *def_stmt = get_def (valueize, op0))
> {
> if (gassign *def = dyn_cast <gassign *> (def_stmt))
> switch (gimple_assign_rhs_code (def))
> {
> case MINUS_EXPR:
> {
> tree o20 = gimple_assign_rhs1 (def);
> o20 = do_valueize (valueize, o20);
> tree o21 = gimple_assign_rhs2 (def);
> o21 = do_valueize (valueize, o21);
> if (op1 == o21 || (operand_equal_p (op1, o21, 0) &&
> types_match (op1, o21)))
> {
>
> which also indents less which is nice.
>
> Bootstrap/regtest running on x86_64-unknown-linux-gnu.
Now applied with
* gcc/testsuite/gcc.dg/pr70920-2.c: Adjust for transform already
happening in ccp1.
* gcc/testsuite/gcc.dg/pr70920-4.c: Likewise.
that shows it's working (CCP is one of the passes eventually running
into this issue).
Richard.
> Richard.
>
> 2017-07-26 Richard Biener <rguenther@suse.de>
>
> * gimple-match-head.c (do_valueize): Return OP if valueize
> returns NULL_TREE.
> (get_def): New helper to get at the def stmt of a SSA name
> if valueize allows.
> * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
> do_valueize to get at the def stmt.
> (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
>
>
>> --
>> Marc Glisse
prev parent reply other threads:[~2017-07-26 11:46 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-06-14 13:07 Bin Cheng
2017-06-16 10:49 ` Richard Biener
2017-06-16 13:06 ` Bin.Cheng
2017-06-16 13:10 ` Richard Biener
2017-06-16 13:31 ` Bin.Cheng
2017-06-16 16:16 ` Richard Biener
2017-06-16 16:23 ` Bin.Cheng
2017-06-16 16:48 ` Marc Glisse
2017-06-16 16:58 ` Bin.Cheng
2017-06-16 17:04 ` Andrew Pinski
2017-07-24 11:45 ` Bin.Cheng
2017-07-24 12:16 ` Marc Glisse
2017-07-24 13:49 ` Bin.Cheng
2017-07-24 13:59 ` Marc Glisse
2017-07-24 14:06 ` Bin.Cheng
2017-07-24 14:31 ` Marc Glisse
2017-07-24 14:37 ` Bin.Cheng
2017-07-24 14:52 ` Marc Glisse
2017-07-25 14:32 ` Richard Biener
2017-07-25 17:45 ` Marc Glisse
2017-07-26 7:48 ` Richard Biener
2017-07-26 9:08 ` Richard Sandiford
2017-07-26 9:38 ` Marc Glisse
2017-07-26 9:45 ` Richard Sandiford
2017-07-26 9:57 ` Marc Glisse
2017-07-26 11:13 ` Richard Biener
2017-07-26 11:46 ` Richard Biener [this message]
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='CAFiYyc2Rw4jkd1c6egiHPgMfkJzsstV68=6wrkOizAnDPnX4Dg@mail.gmail.com' \
--to=richard.guenther@gmail.com \
--cc=amker.cheng@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=marc.glisse@inria.fr \
/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).