From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id E2172C433EF for ; Wed, 30 Mar 2022 10:44:25 +0000 (UTC) Received: from localhost ([::1]:41932 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1nZVoP-0001oz-18 for qemu-devel@archiver.kernel.org; Wed, 30 Mar 2022 06:44:25 -0400 Received: from eggs.gnu.org ([209.51.188.92]:42202) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1nZVNN-0004lI-67 for qemu-devel@nongnu.org; Wed, 30 Mar 2022 06:16:34 -0400 Received: from us-smtp-delivery-124.mimecast.com ([170.10.133.124]:20237) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1nZVNE-0000AN-Mp for qemu-devel@nongnu.org; Wed, 30 Mar 2022 06:16:27 -0400 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1648635362; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=9uc1lYQ33j9cBoLRQW+p4+VRHTtQiLeRFHoxSa4LHys=; b=I27wV7XNlU6a9lN5LAzMz77GyOaiJPAStYT2DbD3zGUNx13+eo6Z66icD4wovWV2Y0NNQ2 ESDSwk0DV/3XBNq6DklHzW3D6cp8aBHY9g66Y8xpUa2zDshK4RspLijWqOqSJAtywVblP9 qL50d5kDiFk1t/Yj8ORe5zYIyraAbL4= Received: from mimecast-mx02.redhat.com (mimecast-mx02.redhat.com [66.187.233.88]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-376-gAcJizOhMpeXg23OLpmUww-1; Wed, 30 Mar 2022 06:16:01 -0400 X-MC-Unique: gAcJizOhMpeXg23OLpmUww-1 Received: from smtp.corp.redhat.com (int-mx02.intmail.prod.int.rdu2.redhat.com [10.11.54.2]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id 8CCBE85A5BE; Wed, 30 Mar 2022 10:16:00 +0000 (UTC) Received: from [10.33.192.183] (dhcp-192-183.str.redhat.com [10.33.192.183]) by smtp.corp.redhat.com (Postfix) with ESMTP id AE85340D2821; Wed, 30 Mar 2022 10:15:59 +0000 (UTC) Message-ID: Date: Wed, 30 Mar 2022 12:15:59 +0200 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101 Thunderbird/91.6.0 Subject: Re: [PATCH 1/2] target/s390x: Fix determination of overflow condition code after addition From: Thomas Huth To: David Hildenbrand , Richard Henderson , qemu-devel@nongnu.org References: <20220323162621.139313-1-thuth@redhat.com> <20220323162621.139313-2-thuth@redhat.com> <2b82de5e-a259-576c-5ea5-eb5c10e6bbeb@redhat.com> <36106411-4cf1-5eaf-b63f-c331380e087b@redhat.com> <6c73160b-787c-0f64-aabc-25bd943d8ffd@redhat.com> <5930e000-35d2-64ec-e301-9305fa09db39@redhat.com> In-Reply-To: X-Scanned-By: MIMEDefang 2.84 on 10.11.54.2 Authentication-Results: relay.mimecast.com; auth=pass smtp.auth=CUSA124A263 smtp.mailfrom=thuth@redhat.com X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Language: en-US Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit Received-SPF: pass client-ip=170.10.133.124; envelope-from=thuth@redhat.com; helo=us-smtp-delivery-124.mimecast.com X-Spam_score_int: -28 X-Spam_score: -2.9 X-Spam_bar: -- X-Spam_report: (-2.9 / 5.0 requ) BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.082, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, NICE_REPLY_A=-0.001, RCVD_IN_DNSWL_LOW=-0.7, RCVD_IN_MSPIKE_H5=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_SCC_BODY_TEXT_LINE=-0.01 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: qemu-s390x@nongnu.org, Bruno Haible Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: "Qemu-devel" On 30/03/2022 12.12, Thomas Huth wrote: > On 30/03/2022 11.47, David Hildenbrand wrote: >> On 30.03.22 11:42, Thomas Huth wrote: >>> On 30/03/2022 11.34, David Hildenbrand wrote: >>>> On 30.03.22 11:29, Thomas Huth wrote: >>>>> On 30/03/2022 10.52, David Hildenbrand wrote: >>>>>> On 23.03.22 17:26, Thomas Huth wrote: >>>>>>> This program currently prints different results when run with TCG >>>>>>> instead >>>>>>> of running on real s390x hardware: >>>>>>> >>>>>>>     #include >>>>>>> >>>>>>>     int overflow_32 (int x, int y) >>>>>>>     { >>>>>>>       int sum; >>>>>>>       return ! __builtin_add_overflow (x, y, &sum); >>>>>>>     } >>>>>>> >>>>>>>     int overflow_64 (long long x, long long y) >>>>>>>     { >>>>>>>       long sum; >>>>>>>       return ! __builtin_add_overflow (x, y, &sum); >>>>>>>     } >>>>>>> >>>>>>>     int a1 = -2147483648; >>>>>>>     int b1 = -2147483648; >>>>>>>     long long a2 = -9223372036854775808L; >>>>>>>     long long b2 = -9223372036854775808L; >>>>>>> >>>>>>>     int main () >>>>>>>     { >>>>>>>       { >>>>>>>         int a = a1; >>>>>>>         int b = b1; >>>>>>>         printf ("a = 0x%x, b = 0x%x\n", a, b); >>>>>>>         printf ("no_overflow = %d\n", overflow_32 (a, b)); >>>>>>>       } >>>>>>>       { >>>>>>>         long long a = a2; >>>>>>>         long long b = b2; >>>>>>>         printf ("a = 0x%llx, b = 0x%llx\n", a, b); >>>>>>>         printf ("no_overflow = %d\n", overflow_64 (a, b)); >>>>>>>       } >>>>>>>     } >>>>>>> >>>>>>> Resolves: https://gitlab.com/qemu-project/qemu/-/issues/616 >>>>>>> Suggested-by: Bruno Haible >>>>>>> Signed-off-by: Thomas Huth >>>>>>> --- >>>>>>>     target/s390x/tcg/cc_helper.c | 4 ++-- >>>>>>>     1 file changed, 2 insertions(+), 2 deletions(-) >>>>>>> >>>>>>> diff --git a/target/s390x/tcg/cc_helper.c b/target/s390x/tcg/cc_helper.c >>>>>>> index 8d04097f78..e11cdb745d 100644 >>>>>>> --- a/target/s390x/tcg/cc_helper.c >>>>>>> +++ b/target/s390x/tcg/cc_helper.c >>>>>>> @@ -136,7 +136,7 @@ static uint32_t cc_calc_subu(uint64_t borrow_out, >>>>>>> uint64_t result) >>>>>>>     static uint32_t cc_calc_add_64(int64_t a1, int64_t a2, int64_t ar) >>>>>>>     { >>>>>>> -    if ((a1 > 0 && a2 > 0 && ar < 0) || (a1 < 0 && a2 < 0 && ar > 0)) { >>>>>>> +    if ((a1 > 0 && a2 > 0 && ar < 0) || (a1 < 0 && a2 < 0 && ar >= >>>>>>> 0)) { >>>>>> >>>>>> >>>>>> Intuitively, I'd have checked for any overflow/underflow by comparing >>>>>> with one of the input variables: >>>>>> >>>>>> a) Both numbers are positive >>>>>> >>>>>> Adding to positive numbers has to result in something that's bigger than >>>>>> the input parameters. >>>>>> >>>>>> "a1 > 0 && a2 > 0 && ar < a1" >>>>> >>>>> I think it doesn't really matter whether we compare ar with a1 or 0 >>>>> here. If >>>>> an overflow happens, what's the biggest number that we can get? AFAICT >>>>> it's >>>>> with a1 = 0x7fffffffffffffff and a2 = 0x7fffffffffffffff. You then get: >>>>> >>>>>     0x7fffffffffffffff + 0x7fffffffffffffff = 0xFFFFFFFFFFFFFFFE >>>>> >>>>> and that's still < 0 if treated as a signed value. I don't see a way where >>>>> ar could be in the range between 0 and a1. >>>>> >>>>> (OTOH, checking for ar < a1 instead of ar < 0 wouldn't hurt either, I >>>>> guess). >>>>> >>>>>> b) Both numbers are negative >>>>>> >>>>>> Adding to negative numbers has to result in something that's smaller >>>>>> than the input parameters. >>>>>> >>>>>> "a1 < 0 && a2 < 0 && ar > a1" >>>>> >>>>> What about if the uppermost bit gets lost in 64-bit mode: >>>>> >>>>>     0x8000000000000000 + 0x8000000000000000 = 0x0000000000000000 >>>>> >>>>> ar > a1 does not work here anymore, does it? >>>> >>>> >>>> 0 > -9223372036854775808, no? >>> >>> current coffe level < correct coffee level >>> >>> ... sorry, never mind, you're right of course. >>> >>> Anyway, 0 is the lowest number we can get for an underflow, so comparing >>> with >= 0 should be fine (but comparing with a1 wouldn't hurt either). >> >> At least for me it takes more brainpower to understand that than >> comparing against one of both parameters as we usually do, e.g., for >> unsigned values > Maybe we should simply switch the code to use __builtin_add_overflow and > __builtin_sub_overflow and let the compiler figure out the details... Never mind again, that doesn't work in this context either ... /me finally goes fetching another coffee... Thomas