From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([140.186.70.92]:36734)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <jcmvbkbc@gmail.com>) id 1QNLjD-0007c7-3v
	for qemu-devel@nongnu.org; Fri, 20 May 2011 05:10:39 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <jcmvbkbc@gmail.com>) id 1QNLj8-0007Gs-FW
	for qemu-devel@nongnu.org; Fri, 20 May 2011 05:10:39 -0400
Received: from mail-qy0-f180.google.com ([209.85.216.180]:61744)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <jcmvbkbc@gmail.com>) id 1QNLj8-0007Gl-8P
	for qemu-devel@nongnu.org; Fri, 20 May 2011 05:10:34 -0400
Received: by qyk10 with SMTP id 10so2057203qyk.4
	for <qemu-devel@nongnu.org>; Fri, 20 May 2011 02:10:33 -0700 (PDT)
MIME-Version: 1.0
In-Reply-To: <201105201125.16972.jcmvbkbc@gmail.com>
References: <1305671572-5899-1-git-send-email-jcmvbkbc@gmail.com>
	<1305671572-5899-20-git-send-email-jcmvbkbc@gmail.com>
	<4DD590D4.4070307@twiddle.net>
	<201105201125.16972.jcmvbkbc@gmail.com>
Date: Fri, 20 May 2011 13:10:33 +0400
Message-ID: <BANLkTimQSPxhtS4nU9kTRWCWWtiw8wCpwg@mail.gmail.com>
From: Max Filippov <jcmvbkbc@gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Subject: Re: [Qemu-devel] [PATCH 19/26] target-xtensa: implement loop option
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Richard Henderson <rth@twiddle.net>
Cc: qemu-devel@nongnu.org

>> > + =A0 =A0if (env->sregs[LEND] !=3D v) {
>> > + =A0 =A0 =A0 =A0tb_invalidate_phys_page_range(
>> > + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0env->sregs[LEND] - 1, env->sregs[LEND=
], 0);
>> > + =A0 =A0 =A0 =A0env->sregs[LEND] =3D v;
>> > + =A0 =A0 =A0 =A0tb_invalidate_phys_page_range(
>> > + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0env->sregs[LEND] - 1, env->sregs[LEND=
], 0);
>> > + =A0 =A0}
>>
>> Why are you invalidating twice?
>
> TB at the old LEND and at the new. Although it will work correctly withou=
t first invalidation.
>
>> > +static void gen_check_loop_end(DisasContext *dc, int slot)
>> > +{
>> > + =A0 =A0if (option_enabled(dc, XTENSA_OPTION_LOOP) &&
>> > + =A0 =A0 =A0 =A0 =A0 =A0!(dc->tb->flags & XTENSA_TBFLAG_EXCM) &&
>> > + =A0 =A0 =A0 =A0 =A0 =A0dc->next_pc =3D=3D dc->lend) {
>> > + =A0 =A0 =A0 =A0int label =3D gen_new_label();
>> > +
>> > + =A0 =A0 =A0 =A0tcg_gen_brcondi_i32(TCG_COND_NE, cpu_SR[LEND], dc->ne=
xt_pc, label);
>> > + =A0 =A0 =A0 =A0tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_SR[LCOUNT], 0, l=
abel);
>> > + =A0 =A0 =A0 =A0tcg_gen_subi_i32(cpu_SR[LCOUNT], cpu_SR[LCOUNT], 1);
>> > + =A0 =A0 =A0 =A0gen_jump(dc, cpu_SR[LBEG]);
>> > + =A0 =A0 =A0 =A0gen_set_label(label);
>> > + =A0 =A0 =A0 =A0gen_jumpi(dc, dc->next_pc, slot);
>>
>> If you're going to pretend that LEND is a constant, you might as well
>> pretend that LBEG is also a constant, so that you get to chain the TB's
>> around the loop.
>
> But there may be three exits from TB at the LEND if its last command is a=
 branch: to the LBEG, to the branch target and to the next insn.

Ok, I guess that I need to add gen_wsr_lbeg that invalidates TB at the
current LEND, pretend that LBEG is constant and use given slot to jump
to it. And also to get rid of tcg_gen_brcondi_i32(TCG_COND_NE,
cpu_SR[LEND], dc->next_pc, label);

--=20
Thanks.
-- Max