public inbox for gcc-prs@sourceware.org
help / color / mirror / Atom feed
From: Alan Modra <amodra@bigpond.net.au>
To: nobody@gcc.gnu.org
Cc: gcc-prs@gcc.gnu.org,
Subject: Re: target/7282: powerpc64 SImode in FPR
Date: Sat, 13 Jul 2002 22:46:00 -0000 [thread overview]
Message-ID: <20020714054602.18565.qmail@sources.redhat.com> (raw)
The following reply was made to PR target/7282; it has been noted by GNATS.
From: Alan Modra <amodra@bigpond.net.au>
To: David Edelsohn <dje@watson.ibm.com>
Cc: gcc-gnats@gcc.gnu.org, gcc-patches@gcc.gnu.org
Subject: Re: target/7282: powerpc64 SImode in FPR
Date: Sun, 14 Jul 2002 15:15:21 +0930
On Sat, Jul 13, 2002 at 10:18:54AM -0400, David Edelsohn wrote:
> This looks okay except for
>
> + [(set (match_operand:DF 0 "gpc_reg_operand" "=f")
> + (float:DF (match_operand:SI 1 "gpc_reg_operand" "*f")))
> + (clobber (match_operand:DI 2 "memory_operand" "=o"))
> + (clobber (match_operand:DI 3 "gpc_reg_operand" "=r"))
> + (clobber (match_operand:DI 4 "gpc_reg_operand" "=f"))]
>
> The input SImode operand should have constraint "r", not "*f". The whole
> point of this pattern is to move the SImode operand from the GPR to the
> FPR because GCC sometimes gets confused when asked to do this itself.
> SImode is not allowed in FPRs, so the "*f" constraint is contradictory.
Revised patch follows, incorporating your floatunssidf2 suggestion too.
PR target/7282
* config/rs6000/rs6000.md (floatsidf2): Enable for POWERPC64.
(floatunssidf2): Likewise.
(floatsidf_ppc64): New insn_and_split.
(floatunssidf_ppc64): Likewise.
--
Alan Modra
IBM OzLabs - Linux Technology Centre
Index: gcc/config/rs6000/rs6000.md
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/rs6000/rs6000.md,v
retrieving revision 1.192
diff -u -p -r1.192 rs6000.md
--- gcc/config/rs6000/rs6000.md 3 Jul 2002 14:41:22 -0000 1.192
+++ gcc/config/rs6000/rs6000.md 14 Jul 2002 05:38:17 -0000
@@ -5350,9 +5350,18 @@
(clobber (match_dup 4))
(clobber (match_dup 5))
(clobber (match_dup 6))])]
- "! TARGET_POWERPC64 && TARGET_HARD_FLOAT"
+ "TARGET_HARD_FLOAT"
"
{
+ if (TARGET_POWERPC64)
+ {
+ rtx mem = assign_stack_temp (DImode, GET_MODE_SIZE (DImode), 0);
+ rtx t1 = gen_reg_rtx (DImode);
+ rtx t2 = gen_reg_rtx (DImode);
+ emit_insn (gen_floatsidf_ppc64 (operands[0], operands[1], mem, t1, t2));
+ DONE;
+ }
+
operands[2] = force_reg (SImode, GEN_INT (0x43300000));
operands[3] = force_reg (DFmode, CONST_DOUBLE_ATOF (\"4503601774854144\", DFmode));
operands[4] = assign_stack_temp (DFmode, GET_MODE_SIZE (DFmode), 0);
@@ -5417,9 +5426,19 @@
(use (match_dup 3))
(clobber (match_dup 4))
(clobber (match_dup 5))])]
- "! TARGET_POWERPC64 && TARGET_HARD_FLOAT"
+ "TARGET_HARD_FLOAT"
"
{
+ if (TARGET_POWERPC64)
+ {
+ rtx mem = assign_stack_temp (DImode, GET_MODE_SIZE (DImode), 0);
+ rtx t1 = gen_reg_rtx (DImode);
+ rtx t2 = gen_reg_rtx (DImode);
+ emit_insn (gen_floatunssidf_ppc64 (operands[0], operands[1], mem,
+ t1, t2));
+ DONE;
+ }
+
operands[2] = force_reg (SImode, GEN_INT (0x43300000));
operands[3] = force_reg (DFmode, CONST_DOUBLE_ATOF (\"4503599627370496\", DFmode));
operands[4] = assign_stack_temp (DFmode, GET_MODE_SIZE (DFmode), 0);
@@ -5533,6 +5552,38 @@
(float:DF (match_operand:DI 1 "gpc_reg_operand" "*f")))]
"TARGET_POWERPC64 && TARGET_HARD_FLOAT"
"fcfid %0,%1"
+ [(set_attr "type" "fp")])
+
+(define_insn_and_split "floatsidf_ppc64"
+ [(set (match_operand:DF 0 "gpc_reg_operand" "=f")
+ (float:DF (match_operand:SI 1 "gpc_reg_operand" "r")))
+ (clobber (match_operand:DI 2 "memory_operand" "=o"))
+ (clobber (match_operand:DI 3 "gpc_reg_operand" "=r"))
+ (clobber (match_operand:DI 4 "gpc_reg_operand" "=f"))]
+ "TARGET_POWERPC64 && TARGET_HARD_FLOAT"
+ "#"
+ ""
+ [(set (match_dup 3) (sign_extend:DI (match_dup 1)))
+ (set (match_dup 2) (match_dup 3))
+ (set (match_dup 4) (match_dup 2))
+ (set (match_dup 0) (float:DF (match_dup 4)))]
+ ""
+ [(set_attr "type" "fp")])
+
+(define_insn_and_split "floatunssidf_ppc64"
+ [(set (match_operand:DF 0 "gpc_reg_operand" "=f")
+ (unsigned_float:DF (match_operand:SI 1 "gpc_reg_operand" "r")))
+ (clobber (match_operand:DI 2 "memory_operand" "=o"))
+ (clobber (match_operand:DI 3 "gpc_reg_operand" "=r"))
+ (clobber (match_operand:DI 4 "gpc_reg_operand" "=f"))]
+ "TARGET_POWERPC64 && TARGET_HARD_FLOAT"
+ "#"
+ ""
+ [(set (match_dup 3) (zero_extend:DI (match_dup 1)))
+ (set (match_dup 2) (match_dup 3))
+ (set (match_dup 4) (match_dup 2))
+ (set (match_dup 0) (float:DF (match_dup 4)))]
+ ""
[(set_attr "type" "fp")])
(define_insn "fix_truncdfdi2"
next reply other threads:[~2002-07-14 5:46 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2002-07-13 22:46 Alan Modra [this message]
-- strict thread matches above, loose matches on Subject: below --
2002-07-13 7:26 David Edelsohn
2002-07-13 4:36 Alan Modra
2002-07-12 0:16 Alan Modra
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=20020714054602.18565.qmail@sources.redhat.com \
--to=amodra@bigpond.net.au \
--cc=gcc-prs@gcc.gnu.org \
--cc=nobody@gcc.gnu.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).