From: Jeff Law <law@redhat.com>
To: "Bin.Cheng" <amker.cheng@gmail.com>
Cc: Bin Cheng <Bin.Cheng@arm.com>,
"gcc-patches@gcc.gnu.org" <gcc-patches@gcc.gnu.org>,
nd <nd@arm.com>
Subject: Re: [PATCH PR69052]Check if loop inv can be propagated into mem ref with additional addr expr canonicalization
Date: Tue, 01 Mar 2016 23:44:00 -0000 [thread overview]
Message-ID: <56D62962.3010309@redhat.com> (raw)
In-Reply-To: <CAHFci28xEg2Q1mDmOajKk7OqBgX3sj1zAQ5kibviDoHRDc-K2g@mail.gmail.com>
On 03/01/2016 10:08 AM, Bin.Cheng wrote:
> On Thu, Feb 25, 2016 at 8:47 AM, Bin.Cheng <amker.cheng@gmail.com> wrote:
>> On Thu, Feb 25, 2016 at 6:39 AM, Jeff Law <law@redhat.com> wrote:
>>> On 02/22/2016 02:22 AM, Bin.Cheng wrote:
>>>
>>>>> My only question is why didn't you use FOR_EACH_SUBRTX_VRA from
>>>>> rtl-iter.h
>>>>> to walk the RTX expressions in collect_address_parts and
>>>>> canonicalize_address_mult?
>>>>
>>>> Hi Jeff,
>>>> Nothing special, just I haven't used this before, also
>>>> canonicalize_address_mult is alphabetically copied from fwprop.c. One
>>>> question is when rewriting SHIFT to MULT, we need to modify rtl
>>>> expression in place, does FOR_EACH_SUBRTX iterator support this? If
>>>> yes, what is the behavior for modified sub-expression?
>>>
>>> Hmm. The question of semantics when we change the underlying
>>> sub-expressions is an interesting one.
>>>
>>> While I think we're OK in practice using the iterators, I think that's more
>>> of an accident than by design -- if MULT/ASHIFT had a different underlying
>>> RTL structure then I'm much less sure using the iterators would be safe.
>> Hi Jeff,
>> Yes, I thought about this too and finally decided to skip sub rtxes in
>> modified MULT/ASHIFT expressions. I think this will make the patch
>> with FOR_EACH_SUBRTX iterator safe. Although it doesn't dive into
>> MULT/ASHIFT while the original one does, I don't think there is such
>> case in practice, after all we can't handle such complicated address
>> expression either.
>>>
>>> Let's go with your original patch that didn't use the iterators. Sorry for
>>> making you do the additional work/testing to build the iterator version.
>> Not even a problem.
>>> But after pondering the issue you raised, I think your original patch is
>>> safer.
>> So in conclusion, I think both versions should be safe, the iterator
>> one is definitely cleaner. It is your call which version I should
>> apply.
> Hi Jeff,
> I tend to apply the new patch with FOR_EACH_SUBRTX because it's
> clearer. Does it work for you? Of course if you do think it's not
> that safe I will fall back to the old one.
Sorry, things are just busy here.
I'm reasonably comfortable with both as long as we're not diving inside
the subexpressions in the ASHIFT/MULT case. If you'd prefer the
FOR_EACH_SUBRTX variant, that's OK with me.
jeff
next prev parent reply other threads:[~2016-03-01 23:44 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-02-09 11:08 Bin Cheng
2016-02-11 7:14 ` Jeff Law
2016-02-11 17:59 ` Bin.Cheng
2016-02-11 23:26 ` Jeff Law
2016-02-16 18:43 ` Bin Cheng
2016-02-19 22:24 ` Jeff Law
2016-02-22 9:22 ` Bin.Cheng
2016-02-25 6:39 ` Jeff Law
2016-02-25 8:47 ` Bin.Cheng
2016-03-01 17:08 ` Bin.Cheng
2016-03-01 23:44 ` Jeff Law [this message]
2016-02-23 15:10 ` Bin.Cheng
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=56D62962.3010309@redhat.com \
--to=law@redhat.com \
--cc=Bin.Cheng@arm.com \
--cc=amker.cheng@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=nd@arm.com \
/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).