From: Kyrill Tkachov <kyrylo.tkachov@arm.com>
To: Wilco Dijkstra <wdijkstr@arm.com>,
'GCC Patches' <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH][AArch64] Fix aarch64_rtx_costs of PLUS/MINUS
Date: Mon, 27 Apr 2015 13:53:00 -0000 [thread overview]
Message-ID: <553E3F3C.7020201@arm.com> (raw)
In-Reply-To: <000501d080f0$235454a0$69fcfde0$@com>
Hi Wilco,
On 27/04/15 14:43, Wilco Dijkstra wrote:
> ping
>
>> -----Original Message-----
>> From: Wilco Dijkstra [mailto:wdijkstr@arm.com]
>> Sent: 04 March 2015 15:38
>> To: GCC Patches
>> Subject: [PATCH][AArch64] Fix aarch64_rtx_costs of PLUS/MINUS
>>
>> Include the cost of op0 and op1 in all cases in PLUS and MINUS in aarch64_rtx_costs.
>> Bootstrap & regression OK.
>>
>> ChangeLog:
>> 2015-03-04 Wilco Dijkstra <wdijkstr@arm.com>
>>
>> * gcc/config/aarch64/aarch64.c (aarch64_rtx_costs):
>> Calculate cost of op0 and op1 in PLUS and MINUS cases.
>>
>> ---
>> gcc/config/aarch64/aarch64.c | 12 +++++-------
>> 1 file changed, 5 insertions(+), 7 deletions(-)
>>
>> diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
>> index 39921a7..e22d72e 100644
>> --- a/gcc/config/aarch64/aarch64.c
>> +++ b/gcc/config/aarch64/aarch64.c
>> @@ -5794,6 +5794,8 @@ aarch64_rtx_costs (rtx x, int code, int outer ATTRIBUTE_UNUSED,
>> op1 = XEXP (x, 1);
>>
>> cost_minus:
>> + *cost += rtx_cost (op0, MINUS, 0, speed);
>> +
>> /* Detect valid immediates. */
>> if ((GET_MODE_CLASS (mode) == MODE_INT
>> || (GET_MODE_CLASS (mode) == MODE_CC
>> @@ -5801,13 +5803,10 @@ cost_minus:
>> && CONST_INT_P (op1)
>> && aarch64_uimm12_shift (INTVAL (op1)))
>> {
>> - *cost += rtx_cost (op0, MINUS, 0, speed);
>> -
>> if (speed)
>> /* SUB(S) (immediate). */
>> *cost += extra_cost->alu.arith;
>> return true;
>> -
>> }
>>
>> /* Look for SUB (extended register). */
>> @@ -5832,7 +5831,6 @@ cost_minus:
>> *cost += aarch64_rtx_mult_cost (new_op1, MULT,
>> (enum rtx_code) code,
>> speed);
>> - *cost += rtx_cost (op0, MINUS, 0, speed);
>> return true;
>> }
>>
>> @@ -5879,6 +5877,8 @@ cost_plus:
>> return true;
>> }
>>
>> + *cost += rtx_cost (op1, PLUS, 1, speed);
>> +
I don't think this is correct. In the code directly below
(when the aarch64_rtx_arith_op_extract_p condition is true)
we have a shift/extend operation by a constant, so we don't
want to take into account the cost of operand 1 (which is
the extend+shift rtx).
So, I looked at this code myself recently and found that the
wrong logic is addressed by my patch at:
https://gcc.gnu.org/ml/gcc-patches/2015-04/msg01617.html
which fixes the case where we forget to cost operand0 in the
MINUS case.
What do you think?
Kyrill
>> /* Look for ADD (extended register). */
>> if (aarch64_rtx_arith_op_extract_p (op0, mode))
>> {
>> @@ -5900,12 +5900,10 @@ cost_plus:
>> {
>> *cost += aarch64_rtx_mult_cost (new_op0, MULT, PLUS,
>> speed);
>> - *cost += rtx_cost (op1, PLUS, 1, speed);
>> return true;
>> }
>>
>> - *cost += (rtx_cost (new_op0, PLUS, 0, speed)
>> - + rtx_cost (op1, PLUS, 1, speed));
>> + *cost += rtx_cost (new_op0, PLUS, 0, speed);
>>
>> if (speed)
>> {
>> --
>> 1.9.1
>
next prev parent reply other threads:[~2015-04-27 13:53 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-04-27 13:43 Wilco Dijkstra
2015-04-27 13:53 ` Kyrill Tkachov [this message]
2015-04-27 14:17 ` Kyrill Tkachov
-- strict thread matches above, loose matches on Subject: below --
2015-03-04 15:38 Wilco Dijkstra
2015-05-01 10:22 ` Marcus Shawcroft
2015-05-01 12:53 ` Kyrill Tkachov
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=553E3F3C.7020201@arm.com \
--to=kyrylo.tkachov@arm.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=wdijkstr@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).