From: "Kewen.Lin" <linkw@linux.ibm.com>
To: "Bin.Cheng" <amker.cheng@gmail.com>
Cc: GCC Patches <gcc-patches@gcc.gnu.org>,
"bin.cheng" <bin.cheng@linux.alibaba.com>,
Richard Guenther <rguenther@suse.de>,
Bill Schmidt <wschmidt@linux.ibm.com>,
Segher Boessenkool <segher@kernel.crashing.org>,
Richard Sandiford <richard.sandiford@arm.com>,
Jiufu Guo <guojiufu@linux.ibm.com>
Subject: Re: [PATCH 3/4 v3] ivopts: Consider cost_step on different forms during unrolling
Date: Wed, 2 Sep 2020 11:50:19 +0800 [thread overview]
Message-ID: <311f2a04-70a1-a871-b111-646a3cef8e82@linux.ibm.com> (raw)
In-Reply-To: <CAHFci2-dg0o9jweJdQCM36NiOQTd7btHDZCjL5PDXinyYhZdOg@mail.gmail.com>
Hi Bin,
>> 2) This case makes me think we should exclude ainc candidates in function
>> mark_reg_offset_candidates. The justification is that: ainc candidate
>> handles step update itself and when we calculate the cost for it against
>> its ainc_use, the cost_step has been reduced. When unrolling happens,
>> the ainc computations are replicated and it doesn't save step updates
>> like normal reg_offset_p candidates.
> Though auto-inc candidate embeds stepping operation into memory
> access, we might want to avoid it in case of unroll if there are many
> sequences of memory accesses, and if the unroll factor is big. The
> rationale is embedded stepping is a u-arch operation and does have its
> cost.
>
Thanks for the comments! Agree! Excluding them from reg_offset_p
candidates here is consistent with this expectation, it makes us
consider the unroll factor effect when checking the corresponding
step cost and the embedded stepping cost (in group/candidate cost,
minus step cost and use the cost from the address_cost hook).
>>
>> I've updated the patch to punt ainc_use candidates as below:
>>
>>> + /* Skip AINC candidate since it contains address update itself,
>>> + the replicated AINC computations when unrolling still have
>>> + updates, unlike reg_offset_p candidates can save step updates
>>> + when unrolling. */
>>> + if (cand->ainc_use)
>>> + continue;
>>> +
>>
>> For this new attached patch, it's bootstrapped and regress-tested without
>> explicit unrolling, while the only one failure has been identified as
>> rs6000 specific address_cost issue in bootstrapping and regression testing
>> with explicit unrolling.
>>
>> By the way, with above simple hack of address_cost, I also did one
>> bootstrapping and regression testing with explicit unrolling, the above
>> sms-4.c did pass as I expected but had two failures instead:
>>
>> PASS->FAIL: gcc.dg/sms-compare-debug-1.c (test for excess errors)
>> PASS->FAIL: gcc.dg/tree-ssa/ivopts-lt.c scan-tree-dump-times ivopts "PHI" 2
>>
>> By further investigation, the 2nd one is expected due to the adddress_cost hook
>> hack, while the 1st one one exposed -fcompare-debug issue in sms. The RTL
>> sequence starts to off from sms, just some NOTE_INSN_DELETED positions change.
>> I believe it's just exposed by this combination unluckily/luckily ;-) I will
>> send a patch separately for it once I got time to look into it, but it should
>> be unrelated to this patch series for sure.
> This is the kind of situation I intended to avoid before. IMHO, this
> isn't a neat change (it can't be given we are predicting the future
> transformation in compilation pipeline), accumulation of such changes
> could make IVOPTs break in one way or another. So as long as you make
> sure it doesn't have functional impact in case of no-rtl_unroll, I am
> fine.
Yeah, I admit it's not neat, but the proposals in the previous discussions
without predicting unroll factor can not work well for all scenarios with
different unroll factors, they could over-blame some kind of candidates.
For the case of no-rtl_unroll, unroll factor estimation should set
loop->estimated_unroll to zero, all these changes won't take effect. The
estimation function follows the same logics as that of RTL unroll factor
calculation, I did test with explicit unrolling disablement before, it
worked expectedly.
BR,
Kewen
next prev parent reply other threads:[~2020-09-02 3:50 UTC|newest]
Thread overview: 64+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-28 12:17 [PATCH 0/4] IVOPTs consider step cost for different forms when unrolling Kewen.Lin
2020-05-28 12:19 ` [PATCH 1/4] unroll: Add middle-end unroll factor estimation Kewen.Lin
2020-08-31 5:49 ` PING " Kewen.Lin
2020-09-15 7:44 ` PING^2 " Kewen.Lin
2020-10-13 7:06 ` PING^3 " Kewen.Lin
2020-11-02 9:13 ` PING^4 " Kewen.Lin
2020-11-19 5:50 ` PING^5 " Kewen.Lin
2020-12-17 2:58 ` PING^6 " Kewen.Lin
2021-01-14 2:36 ` PING^7 " Kewen.Lin
2021-01-21 21:45 ` Segher Boessenkool
2021-01-22 12:50 ` Richard Sandiford
2021-01-22 13:47 ` Richard Biener
2021-01-22 21:37 ` Segher Boessenkool
2021-01-25 7:56 ` Richard Biener
2021-01-25 17:59 ` Richard Sandiford
2021-01-25 20:37 ` Segher Boessenkool
2021-01-26 8:53 ` Kewen.Lin
2021-01-26 17:31 ` Segher Boessenkool
2021-01-26 8:43 ` Kewen.Lin
2021-01-26 10:47 ` Richard Biener
2021-01-26 17:54 ` Segher Boessenkool
2021-01-26 8:36 ` Kewen.Lin
2021-01-26 10:53 ` Richard Biener
2021-01-27 9:43 ` Kewen.Lin
2021-03-01 2:45 ` Kewen.Lin
2020-05-28 12:23 ` [PATCH 2/4] param: Introduce one param to control ivopts reg-offset consideration Kewen.Lin
2020-05-28 12:24 ` [PATCH 3/4] ivopts: Consider cost_step on different forms during unrolling Kewen.Lin
2020-06-01 17:59 ` Richard Sandiford
2020-06-02 3:39 ` Kewen.Lin
2020-06-02 7:14 ` Richard Sandiford
2020-06-03 3:18 ` Kewen.Lin
2020-08-08 8:01 ` Bin.Cheng
2020-08-10 4:27 ` Kewen.Lin
2020-08-10 12:38 ` Bin.Cheng
2020-08-10 14:41 ` Kewen.Lin
2020-08-16 3:59 ` Bin.Cheng
2020-08-18 9:02 ` [PATCH 3/4 v2] " Kewen.Lin
2020-08-22 5:11 ` Bin.Cheng
2020-08-25 12:46 ` [PATCH 3/4 v3] " Kewen.Lin
2020-08-31 19:41 ` Segher Boessenkool
2020-09-02 3:16 ` Kewen.Lin
2020-09-02 10:25 ` Segher Boessenkool
2020-09-03 2:24 ` Kewen.Lin
2020-09-03 22:37 ` Segher Boessenkool
2020-09-04 8:27 ` Bin.Cheng
2020-09-04 13:53 ` Segher Boessenkool
2020-09-04 8:47 ` Kewen.Lin
2020-09-04 14:16 ` Segher Boessenkool
2020-09-04 15:47 ` Kewen.Lin
2020-09-17 23:12 ` Jeff Law
2020-09-17 23:46 ` Segher Boessenkool
2020-09-01 11:19 ` Bin.Cheng
2020-09-02 3:50 ` Kewen.Lin [this message]
2020-09-02 3:55 ` Bin.Cheng
2020-09-02 4:51 ` Kewen.Lin
2020-09-06 2:47 ` Hans-Peter Nilsson
2020-09-15 7:41 ` Kewen.Lin
2020-06-02 11:38 ` [PATCH 0/4] IVOPTs consider step cost for different forms when unrolling Richard Biener
2020-06-03 3:46 ` Kewen.Lin
2020-06-03 7:07 ` Richard Biener
2020-06-03 7:58 ` Kewen.Lin
2020-06-03 9:27 ` Richard Biener
2020-06-03 10:47 ` Kewen.Lin
2020-06-03 11:08 ` Richard Sandiford
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=311f2a04-70a1-a871-b111-646a3cef8e82@linux.ibm.com \
--to=linkw@linux.ibm.com \
--cc=amker.cheng@gmail.com \
--cc=bin.cheng@linux.alibaba.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=guojiufu@linux.ibm.com \
--cc=rguenther@suse.de \
--cc=richard.sandiford@arm.com \
--cc=segher@kernel.crashing.org \
--cc=wschmidt@linux.ibm.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).