From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from zen.linaro.local ([81.128.185.34]) by smtp.gmail.com with ESMTPSA id g5-v6sm6103762wrw.97.2018.11.06.06.23.26 (version=TLS1_2 cipher=ECDHE-RSA-CHACHA20-POLY1305 bits=256/256); Tue, 06 Nov 2018 06:23:26 -0800 (PST) Received: from zen (localhost [127.0.0.1]) by zen.linaro.local (Postfix) with ESMTPS id AB49D3E0245; Tue, 6 Nov 2018 14:23:25 +0000 (GMT) References: User-agent: mu4e 1.1.0; emacs 26.1.50 From: Alex =?utf-8?Q?Benn=C3=A9e?= To: Peter Maydell Cc: QEMU Developers , qemu-arm , Richard Henderson Subject: Re: aarch32 acquire/release vs mttcg In-reply-to: Date: Tue, 06 Nov 2018 14:23:25 +0000 Message-ID: <87o9b2e142.fsf@linaro.org> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable X-TUID: iyOx7VO5QCXh Peter Maydell writes: > Looking through code I noticed that in target/arm/translate.c there > is the following comment in the decode for the load-acquire/store-release > instructions: > > /* Since the emulation does not have barriers, > the acquire/release semantics need no special > handling */ > > This is out of date now we have MTTCG, isn't it? Yes > The equivalent code in translate-a64.c has some calls to tcg_gen_mb() > to emit barriers...presumably we need to fix the a32 code to do > the same. Sounds right. Are they the same semantics as their bigger 64 bit cousins? > > (We haven't noticed this because I think in practice nobody > much is compiling Armv8-specific AArch32 code.) > > thanks > -- PMM -- Alex Benn=C3=A9e From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:50489) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gK2GU-0000Or-9t for qemu-devel@nongnu.org; Tue, 06 Nov 2018 09:23:35 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1gK2GO-0003mD-OY for qemu-devel@nongnu.org; Tue, 06 Nov 2018 09:23:33 -0500 Received: from mail-wm1-x32c.google.com ([2a00:1450:4864:20::32c]:32808) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1gK2GO-0003kG-I8 for qemu-devel@nongnu.org; Tue, 06 Nov 2018 09:23:28 -0500 Received: by mail-wm1-x32c.google.com with SMTP id f19-v6so9057860wmb.0 for ; Tue, 06 Nov 2018 06:23:28 -0800 (PST) References: From: Alex =?utf-8?Q?Benn=C3=A9e?= In-reply-to: Date: Tue, 06 Nov 2018 14:23:25 +0000 Message-ID: <87o9b2e142.fsf@linaro.org> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] aarch32 acquire/release vs mttcg List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Peter Maydell Cc: QEMU Developers , qemu-arm , Richard Henderson Peter Maydell writes: > Looking through code I noticed that in target/arm/translate.c there > is the following comment in the decode for the load-acquire/store-release > instructions: > > /* Since the emulation does not have barriers, > the acquire/release semantics need no special > handling */ > > This is out of date now we have MTTCG, isn't it? Yes > The equivalent code in translate-a64.c has some calls to tcg_gen_mb() > to emit barriers...presumably we need to fix the a32 code to do > the same. Sounds right. Are they the same semantics as their bigger 64 bit cousins? > > (We haven't noticed this because I think in practice nobody > much is compiling Armv8-specific AArch32 code.) > > thanks > -- PMM -- Alex Benn=C3=A9e