From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:59672) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1dPLvb-0005UZ-36 for qemu-devel@nongnu.org; Mon, 26 Jun 2017 00:47:11 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1dPLva-0000iX-4p for qemu-devel@nongnu.org; Mon, 26 Jun 2017 00:47:11 -0400 Received: from mail-it0-x22a.google.com ([2607:f8b0:4001:c0b::22a]:37918) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1dPLvZ-0000iH-W3 for qemu-devel@nongnu.org; Mon, 26 Jun 2017 00:47:10 -0400 Received: by mail-it0-x22a.google.com with SMTP id b205so33453564itg.1 for ; Sun, 25 Jun 2017 21:47:09 -0700 (PDT) MIME-Version: 1.0 Sender: philippe.mathieu.daude@gmail.com In-Reply-To: References: <20170620205121.26515-1-laurent@vivier.eu> <20170620205121.26515-4-laurent@vivier.eu> From: =?UTF-8?Q?Philippe_Mathieu=2DDaud=C3=A9?= Date: Mon, 26 Jun 2017 01:47:08 -0300 Message-ID: Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH v5 3/6] target-m68k: use floatx80 internally List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Richard Henderson Cc: Laurent Vivier , "qemu-devel@nongnu.org Developers" , Riku Voipio , Thomas Huth , Aurelien Jarno On Wed, Jun 21, 2017 at 1:37 PM, Richard Henderson wrote: > On 06/21/2017 09:18 AM, Philippe Mathieu-Daud=C3=A9 wrote: >> What an awful name... Anyway checking on "qemu/bswap.h" it seems there i= s >> some endianess issue with it if your host is little-endian. > > > There is no endian-ness issue because we do not attempt to read that > structure from memory as a whole. Instead, Laurent uses two big-endian > loads (with appropriate address arithmetic) and stores the result into th= is > host structure in host-endian order. Further, the host routines use the > structure members by name and do not assume any particular relationship > between them. Ok, thank your Richard for this clear explanation. ("Endian-ness" ok, Peter also corrected me, I'll give a try to some speller= ...) > >> Do you have a way to run Berkeley TestFloat? > > > As noted in Laurent's cover message, floatx80 isn't quite right -- that i= s > the x86 data type, and the proper m68k data type is slightly different. > > I would expect the results from using floatx80 to be Just Good Enough to > produce a working m68k user-land. It will produce correct results for > normal numbers in arithmetic such as 1.0 + 10.0. But I would expect many= of > the edge conditions that TestFloat would attempt (especially de-normals a= nd > un-normals) would fail. Ok. I'm asking thinking about cross unit-tests I can add to our continuous integration system, this might be a candidate. Regards, Phil.