From: "Ulrich Weigand" <uweigand@de.ibm.com>
To: mutazilah@gmail.com (Paul Edwards)
Cc: gcc@gcc.gnu.org
Subject: Re: i370 port
Date: Fri, 06 Apr 2012 18:16:00 -0000 [thread overview]
Message-ID: <201204061815.q36IFjJ4021718@d06av02.portsmouth.uk.ibm.com> (raw)
In-Reply-To: <CE93949ADBB94262AEF9A2328F3A0594@PaulHP> from "Paul Edwards" at Apr 06, 2012 10:48:50 PM
Paul Edwards wrote:
> I've managed to isolate the problem to a small test program.
>
> Any suggestions on how to debug this?
> bug27.c:28: error: unrecognizable insn:
> (insn 116 34 35 2 (set (reg:SI 5 5)
> (plus:SI (plus:SI (reg:SI 2 2 [orig:54 i ] [54])
> (reg/f:SI 13 13))
> (const_int 104 [0x68]))) -1 (nil)
> (nil))
Ah, yes. The problem is that reload assumes any valid address
can be loaded into a register with a single instruction, and
it will thus simply generate such instructions unconditionally
-- and if the target then doesn't actually provide such a pattern,
it will fail with "unrecognizable insn".
The "LA" pattern you showed accepts only "(certain) single register
+ constant", which doesn't match the pattern above "register +
register + constant".
The difficulty is now that while "LA" *does* actually support adding
base + index + displacement, you cannot simply add a pattern
expressing that, because LA only does a 31-bit add, so it must only
be used to add addresses, not when adding general 32-bit SImode
values.
The way I handle this in the s390 port is to provide a "forced"
LA instruction pattern using a magic marker that prevents the
pattern from being matched by regular instructions, and then add
a secondary reload to emit that "forced" LA when reload needs
to load an address.
Here's the relevant snippets from the 3.4 s390 back-end:
/* We need a secondary reload when loading a PLUS which is
not a valid operand for LOAD ADDRESS. */
#define SECONDARY_INPUT_RELOAD_CLASS(CLASS, MODE, IN) \
s390_secondary_input_reload_class ((CLASS), (MODE), (IN))
(define_insn "*la_31"
[(set (match_operand:SI 0 "register_operand" "=d,d")
(match_operand:QI 1 "address_operand" "U,W"))]
"!TARGET_64BIT && legitimate_la_operand_p (operands[1])"
"@
la\t%0,%a1
lay\t%0,%a1"
[(set_attr "op_type" "RX,RXY")
(set_attr "type" "la")])
(define_insn "force_la_31"
[(set (match_operand:SI 0 "register_operand" "=d,d")
(match_operand:QI 1 "address_operand" "U,W"))
(use (const_int 0))]
"!TARGET_64BIT"
"@
la\t%0,%a1
lay\t%0,%a1"
[(set_attr "op_type" "RX")
(set_attr "type" "la")])
(define_expand "reload_insi"
[(parallel [(match_operand:SI 0 "register_operand" "=a")
(match_operand:SI 1 "s390_plus_operand" "")
(match_operand:SI 2 "register_operand" "=&a")])]
"!TARGET_64BIT"
{
s390_expand_plus_operand (operands[0], operands[1], operands[2]);
DONE;
})
(and of course the functions in s390.c called by these.)
You ought to be able to do something similar in your backend ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
Ulrich.Weigand@de.ibm.com
next prev parent reply other threads:[~2012-04-06 18:16 UTC|newest]
Thread overview: 92+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-04-06 12:49 Paul Edwards
2012-04-06 18:16 ` Ulrich Weigand [this message]
2012-04-07 4:12 ` Paul Edwards
-- strict thread matches above, loose matches on Subject: below --
2014-02-13 4:23 Paul Edwards
2012-04-07 5:45 Paul Edwards
2012-04-08 17:43 ` Ulrich Weigand
2014-02-11 17:01 ` Paul Edwards
2012-04-06 5:51 Paul Edwards
2011-08-20 12:15 Paul Edwards
2011-08-22 12:23 ` Ulrich Weigand
2012-04-05 13:32 ` Paul Edwards
2012-04-06 18:13 ` Ulrich Weigand
2011-08-20 10:09 Paul Edwards
2011-08-20 7:44 Paul Edwards
2009-09-22 12:31 Paul Edwards
2009-09-09 22:33 Paul Edwards
2009-09-14 15:42 ` Ulrich Weigand
2009-09-15 12:59 ` Paul Edwards
2009-09-15 13:51 ` Ulrich Weigand
2009-09-17 13:00 ` Paul Edwards
2009-09-17 17:55 ` Ulrich Weigand
2009-09-18 0:35 ` Paul Edwards
2009-09-18 12:06 ` Ulrich Weigand
2009-09-18 12:23 ` Paul Edwards
2009-09-18 13:27 ` Ulrich Weigand
2009-09-18 13:42 ` Paul Edwards
2009-09-18 16:08 ` Ulrich Weigand
2009-09-19 12:57 ` Paul Edwards
2009-09-25 10:19 ` Paul Edwards
2009-09-25 15:20 ` Ulrich Weigand
2009-11-04 5:21 ` Paul Edwards
2009-11-04 16:47 ` Ulrich Weigand
2009-11-09 14:55 ` Paul Edwards
2009-11-09 15:57 ` Ian Lance Taylor
2009-11-09 23:10 ` Paul Edwards
2009-11-10 14:58 ` Paul Edwards
2009-11-10 15:36 ` Ian Lance Taylor
2009-11-10 15:51 ` Paul Edwards
2009-11-10 15:56 ` Ian Lance Taylor
2009-12-02 22:03 ` Paul Edwards
2011-08-13 8:34 ` Paul Edwards
2011-08-15 14:32 ` Ulrich Weigand
2011-08-15 15:26 ` Paul Edwards
2011-08-15 17:23 ` Ulrich Weigand
2011-08-16 11:20 ` Paul Edwards
2011-08-16 13:26 ` Ulrich Weigand
2011-08-18 12:15 ` Paul Edwards
2011-08-18 13:14 ` Ulrich Weigand
2011-08-18 14:18 ` Paul Edwards
2009-08-23 8:50 Paul Edwards
2009-08-26 22:13 ` Henrik Sorensen
2009-06-05 12:45 Paul Edwards
2009-06-05 14:33 ` Joseph S. Myers
2009-06-05 14:57 ` Paul Edwards
2009-06-05 15:03 ` Joseph S. Myers
2009-06-05 15:24 ` Paul Edwards
2009-06-05 15:47 ` Joseph S. Myers
2017-03-31 10:34 ` Paul Edwards
2009-09-12 12:41 ` Paul Edwards
2009-06-05 15:21 ` Ulrich Weigand
2009-06-05 15:39 ` Paul Edwards
2009-06-05 15:49 ` Daniel Jacobowitz
2009-06-05 15:57 ` Paul Edwards
2009-06-05 20:20 ` Joseph S. Myers
2009-06-05 20:45 ` Paul Edwards
2009-06-06 15:00 ` Paul Edwards
2009-06-15 17:46 ` Ulrich Weigand
2009-06-19 0:06 ` Paul Edwards
2009-06-19 12:28 ` Ulrich Weigand
2009-07-18 11:28 ` Paul Edwards
2009-07-20 14:27 ` Ulrich Weigand
2009-08-08 12:04 ` Paul Edwards
2009-08-10 21:25 ` Ulrich Weigand
2009-08-11 0:34 ` Paul Edwards
2009-08-11 15:21 ` Ulrich Weigand
2009-08-12 11:52 ` Paul Edwards
2009-08-12 15:27 ` Paolo Bonzini
2009-08-12 16:35 ` Ulrich Weigand
2009-08-12 17:27 ` Paul Edwards
2009-08-12 17:56 ` Paolo Bonzini
2009-08-12 19:46 ` Ulrich Weigand
2009-08-12 20:31 ` Paul Edwards
2009-08-19 12:07 ` Paul Edwards
2009-08-19 12:27 ` Paolo Bonzini
2009-08-20 12:49 ` Paul Edwards
2009-08-20 22:48 ` Ulrich Weigand
2009-08-21 2:37 ` Paul Edwards
2009-08-21 16:46 ` Ulrich Weigand
2009-06-05 15:44 ` Joseph S. Myers
2009-06-05 15:52 ` Paul Edwards
2009-09-08 15:55 ` Paul Edwards
2009-09-14 15:32 ` Ulrich Weigand
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=201204061815.q36IFjJ4021718@d06av02.portsmouth.uk.ibm.com \
--to=uweigand@de.ibm.com \
--cc=gcc@gcc.gnu.org \
--cc=mutazilah@gmail.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).