From: Jiufu Guo <guojiufu@linux.ibm.com>
To: Segher Boessenkool <segher@kernel.crashing.org>
Cc: gcc-patches@gcc.gnu.org, dje.gcc@gmail.com, linkw@gcc.gnu.org
Subject: Re: [PATCH V4] rs6000: Optimize cmp on rotated 16bits constant
Date: Thu, 25 Aug 2022 20:11:31 +0800 [thread overview]
Message-ID: <7eedx4h6a4.fsf@pike.rch.stglabs.ibm.com> (raw)
In-Reply-To: <20220824140722.GZ25951@gate.crashing.org> (Segher Boessenkool's message of "Wed, 24 Aug 2022 09:07:22 -0500")
Hi,
Segher Boessenkool <segher@kernel.crashing.org> writes:
> On Wed, Aug 24, 2022 at 03:48:49PM +0800, Jiufu Guo wrote:
>> Segher Boessenkool <segher@kernel.crashing.org> writes:
>> >> + "TARGET_POWERPC64 && !reload_completed && can_create_pseudo_p ()
>> >
>> > reload_completed in splitters is almost always wrong. It isn't any
>> > better if it is in the insn condition of a define_insn_and_split :-)
>> >
>> Thanks, 'can_create_pseudo_p' would be ok for this patch.
>> Or just FAIL, if !can_create_pseudo_p()?
>
> You usually can split fine if you cannot create new pseudos, by reusing
> existing registers.
>
> FAIL will cause an ICE: the RTL instruction does match, but will fail
> when trying to generate machine code for it.
>
Previous patch is using "gen_reg_rtx (DImode)" to generate a pseudo for
the rotated result to prevent orignal one being changed accidently.
So, an 'assert (can_create_pseudo_p ())' would catch it in after RA.
To enable this splitter works after RA, we may need to reserve one
register (clobber would be ok). Such as below:
[(set (pc)
(if_then_else (eqne (match_operand:DI 1 "gpc_reg_operand" "r")
(match_operand:DI 2 "const_int_operand" "n"))
(label_ref (match_operand 0 ""))
(pc)))
(clobber (match_scratch:DI 3 "=r"))
(clobber (match_scratch:CCUNS 4 "=y"))]
"TARGET_POWERPC64 && num_insns_constant (operands[2], DImode) > 1
&& compare_rotate_immediate_p (UINTVAL (operands[2]))"
"#"
"&& 1"
[(pc)]
>> >> + && num_insns_constant (operands[2], DImode) > 1
>> >> + && (rotate_from_leading_zeros_const (~UINTVAL (operands[2]), 49) > 0
>> >> + || rotate_from_leading_zeros_const (UINTVAL (operands[2]), 48) > 0)"
>> > There must be a better way to describe this.
>> Will update this. I'm thinking to replace this with a meaning function,
>> maybe 'compare_rotate_immediate_p'.
>
> Thanks!
>
>> > Why is this doing a conditional branch at all? Unpredictable
>> > conditional branches are extremely costly.
>> This optimization needs to check whether the comparison code is ne/eq or
>> not. To get the comparison code, we need to check the parent insn of
>> the 'cmp' insn. This is why conditional branch patterns in used here.
>>
>> This patch should not change the information (about prediction) of the
>> branch insn. I'm thinking of updating the patch to keep the 'note info
>> REG_BR_PROB' for the branch instruction.
>
> Ah, good. Explain a bit about that? In a code comment or in the commit
> message, whichever works best here.
>
Thanks! will add a comment for this.
BR,
Jeff(Jiufu)
> Thanks!
>
>
> Segher
next prev parent reply other threads:[~2022-08-25 12:11 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-25 13:29 Jiufu Guo
2022-08-17 3:12 ` Jiufu Guo
2022-08-23 22:18 ` Segher Boessenkool
2022-08-24 7:48 ` Jiufu Guo
2022-08-24 14:07 ` Segher Boessenkool
2022-08-25 12:11 ` Jiufu Guo [this message]
2022-08-25 12:34 ` Segher Boessenkool
2022-08-26 9:28 ` Jiufu Guo
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=7eedx4h6a4.fsf@pike.rch.stglabs.ibm.com \
--to=guojiufu@linux.ibm.com \
--cc=dje.gcc@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=linkw@gcc.gnu.org \
--cc=segher@kernel.crashing.org \
/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).