From: Rask Ingemann Lambertsen <rask@sygehus.dk>
To: Pompapathi V Gadad <pompapathi@gmail.com>
Cc: "Gadad, Pompapathi V" <Pompapathi.V.Gadad@nsc.com>,
gcc-patches@gcc.gnu.org
Subject: Re: Re: [Patch] New: CR16 port
Date: Fri, 13 Jul 2007 15:23:00 -0000 [thread overview]
Message-ID: <20070713144426.GB5690@sygehus.dk> (raw)
In-Reply-To: <4697387B.20908@gmail.com>
On Fri, Jul 13, 2007 at 02:01:55PM +0530, Pompapathi V Gadad wrote:
> Rask Ingemann Lambertsen wrote:
> >>+(define_insn "mov<mode>_regs"
[4 mov patterns in total]
> >
> >These would normally be one pattern with five alternatives. A reg->reg
> >copy can turn into a reg->mem copy during reload, for example.
> >
> Well, I started with suggestion you had: Having 5 alternatives in a single pattern.
> It turns out that some of the pattern generated pass through predicate test but won't find
> an appropriate alternative. For example, in CR16 ISA, immediate memory store instruction
> does not have 32-bit store operation variant. Therefore, I had to break up the single pattern into
> multiple patterns. The store immediate pattern have "SHORT" (stands for QI and HI) macro
> expansion while others have ALLMT (stands for QI, HI, SI, SF)
You could define "LONG" to expand to SI and SF and use something like
this:
(define_insn "*mov<mode>"
[(set (match_operand:LONG 0 "nonimmediate_operand" "=r, r, r,m")
(match_operand:LONG 1 "general_operand" "r,<iF>,m,r"))]
"register_operand (operands[0], <MODE>mode)
|| register_operand (operands[1], <MODE>mode)"
"@
mov<tIsa>\t%1, %0
mov<tIsa>\t%1, %0
load<tIsa>\t%1, %0
stor<tIsa>\t%1, %0"
[(set_attr "length" "2,<lImmArith>,<lImmArith>,<lImmArith>")]
)
(define_insn "*mov<mode>"
[(set (match_operand:SHORT 0 "nonimmediate_operand" "=r, r, r,m, m")
(match_operand:SHORT 1 "general_operand" "r,<iF>,m,r,<LL>"))]
"!store_operand (operands[0], <MODE>mode)
|| register_operand (operands[1], <MODE>mode)
|| u4bits_operand (operands[1], <MODE>mode)"
"@
mov<tIsa>\t%1, %0
mov<tIsa>\t%1, %0
load<tIsa>\t%1, %0
stor<tIsa>\t%1, %0
stor<tIsaShort>\t%1, %0"
[(set_attr "length" "2,<lImmArith>,<lImmArith>,<lImmArith>,<lImmShort>")]
)
You may have to adjust the predicates a little.
--
Rask Ingemann Lambertsen
next prev parent reply other threads:[~2007-07-13 14:44 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-07-10 7:11 Gadad, Pompapathi V
2007-07-11 23:41 ` Rask Ingemann Lambertsen
2007-07-13 9:22 ` Pompapathi V Gadad
2007-07-13 15:23 ` Rask Ingemann Lambertsen [this message]
2007-07-14 4:00 ` Pompapathi V Gadad
2007-07-16 9:14 ` Pompapathi V Gadad
2007-07-16 12:25 ` 'Rask Ingemann Lambertsen'
2007-07-16 13:01 ` Pompapathi V Gadad
2007-07-16 13:08 ` Pompapathi V Gadad
2007-07-20 9:48 ` Pompapathi V Gadad
2007-07-21 14:33 ` 'Rask Ingemann Lambertsen'
2007-07-25 9:40 ` Pompapathi V Gadad
2007-07-30 8:33 ` 'Rask Ingemann Lambertsen'
2007-07-30 9:14 ` Pompapathi V Gadad
2007-08-03 13:03 ` 'Rask Ingemann Lambertsen'
2007-08-06 4:44 ` Pompapathi V Gadad
2007-08-06 10:43 ` 'Rask Ingemann Lambertsen'
2007-08-06 10:49 ` Pompapathi V Gadad
2007-07-21 12:20 ` 'Rask Ingemann Lambertsen'
2007-07-13 15:56 ` Rask Ingemann Lambertsen
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=20070713144426.GB5690@sygehus.dk \
--to=rask@sygehus.dk \
--cc=Pompapathi.V.Gadad@nsc.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=pompapathi@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).