From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-patches-return-493928-listarch-gcc-patches=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 59610 invoked by alias); 12 Jan 2019 13:28:16 -0000
Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-patches.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-patches/>
List-Post: <mailto:gcc-patches@gcc.gnu.org>
List-Help: <mailto:gcc-patches-help@gcc.gnu.org>
Sender: gcc-patches-owner@gcc.gnu.org
Received: (qmail 59600 invoked by uid 89); 12 Jan 2019 13:28:16 -0000
Authentication-Results: sourceware.org; auth=none
X-Spam-SWARE-Status: No, score=-25.1 required=5.0 tests=BAYES_00,GIT_PATCH_0,GIT_PATCH_1,GIT_PATCH_2,GIT_PATCH_3,KAM_COUK,KAM_LAZY_DOMAIN_SECURITY autolearn=ham version=3.3.2 spammy=nn, sk:unspecv, sk:UNSPECV, Hx-spam-relays-external:sk:smtp.wa
X-HELO: smtp2.wavenetuk.net
Received: from smtp.wavenetuk.net (HELO smtp2.wavenetuk.net) (195.26.37.10) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Sat, 12 Jan 2019 13:28:14 +0000
Received: from [192.168.1.212] (host81-138-1-83.in-addr.btopenworld.com [81.138.1.83])	by smtp2.wavenetuk.net (Postfix) with ESMTPA id 94D8E600135;	Sat, 12 Jan 2019 13:28:11 +0000 (GMT)
From: Iain Sandoe <iain@sandoe.co.uk>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Mime-Version: 1.0 (Mac OS X Mail 10.3 \(3273\))
Subject: [PATCH, powerpc] Fix speculation barrier and group nop to emit target register names.
Message-Id: <8E1C034C-7935-4D83-810C-08A600310FB0@sandoe.co.uk>
Date: Sat, 12 Jan 2019 13:28:00 -0000
Cc: Segher Boessenkool <segher@kernel.crashing.org>
To: gcc-patches <gcc-patches@gcc.gnu.org>
X-SW-Source: 2019-01/txt/msg00697.txt.bz2

Hi,

The current implementation of =E2=80=9Cspeculation_barrier=E2=80=9D and =E2=
=80=9Cgroup_end_nop=E2=80=9D insns emit hard-wired register names which cau=
ses tests using them to fail on Darwin, at least, which uses =E2=80=9CrNN=
=E2=80=9D instead of =E2=80=9CNN=E2=80=9D.

The patch makes the register names for these insns use the operand output m=
echanism to substitute the appropriate variant when needed.

tested on powerpc-darwin9 and powerpc64-linux.

OK for trunk?
Iain

gcc/

	* config/rs6000/rs6000.md (group_end_nop): Emit
	insn register names using operand format, rather than
	hard-wired.  (speculation_barrier): Likewise.


diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md
index 56364e0..86badc2 100644
--- a/gcc/config/rs6000/rs6000.md
+++ b/gcc/config/rs6000/rs6000.md
@@ -12494,15 +12494,18 @@
   [(unspec [(const_int 0)] UNSPEC_GRP_END_NOP)]
   ""
 {
-  if (rs6000_tune =3D=3D PROCESSOR_POWER6)
-    return "ori 1,1,0";
-  return "ori 2,2,0";
+  operands[0] =3D gen_rtx_REG (Pmode,
+			     rs6000_tune =3D=3D PROCESSOR_POWER6 ? 1 : 2);
+  return "ori %0,%0,0";
 })
=20
 (define_insn "speculation_barrier"
   [(unspec_volatile:BLK [(const_int 0)] UNSPECV_SPEC_BARRIER)]
   ""
-  "ori 31,31,0")
+{
+  operands[0] =3D gen_rtx_REG (Pmode, 31);
+  return "ori %0,%0,0";
+})
 =0C
 ;; Define the subtract-one-and-jump insns, starting with the template
 ;; so loop.c knows what to generate.