From patchwork Thu Sep 15 11:21:33 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Prathamesh Kulkarni X-Patchwork-Id: 76273 Delivered-To: patch@linaro.org Received: by 10.140.106.72 with SMTP id d66csp2383877qgf; Thu, 15 Sep 2016 04:22:40 -0700 (PDT) X-Received: by 10.66.43.164 with SMTP id x4mr13513890pal.11.1473938560806; Thu, 15 Sep 2016 04:22:40 -0700 (PDT) Return-Path: Received: from sourceware.org (server1.sourceware.org. [209.132.180.131]) by mx.google.com with ESMTPS id we7si2149115pab.149.2016.09.15.04.22.21 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Thu, 15 Sep 2016 04:22:21 -0700 (PDT) Received-SPF: pass (google.com: domain of gcc-patches-return-435970-patch=linaro.org@gcc.gnu.org designates 209.132.180.131 as permitted sender) client-ip=209.132.180.131; Authentication-Results: mx.google.com; dkim=pass header.i=@gcc.gnu.org; spf=pass (google.com: domain of gcc-patches-return-435970-patch=linaro.org@gcc.gnu.org designates 209.132.180.131 as permitted sender) smtp.mailfrom=gcc-patches-return-435970-patch=linaro.org@gcc.gnu.org; dmarc=fail (p=NONE dis=NONE) header.from=linaro.org DomainKey-Signature: a=rsa-sha1; c=nofws; d=gcc.gnu.org; h=list-id :list-unsubscribe:list-archive:list-post:list-help:sender :mime-version:in-reply-to:references:from:date:message-id :subject:to:content-type; q=dns; s=default; b=Jt7RJ6XGmzk1j0Bh1i YceR0rYBLHgbxEmhfYcRgI5kzy47vcIsSQxWLk40n3iFRdTnZIUYt3YfnzcePODQ +aqJYJ5PK8qAl7sstQWy+VeURtkVsUC9Q2c0FOuapTb5PL7UkqRruidCRR5Jd6ct nr3W0sNFd04BO5TBQtfOwsSZg= DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=gcc.gnu.org; h=list-id :list-unsubscribe:list-archive:list-post:list-help:sender :mime-version:in-reply-to:references:from:date:message-id :subject:to:content-type; s=default; bh=WlR7xD67eZ1b0j1hSDXFBIPy Bmo=; b=EeBLeIyoJNQi3Vy+vbsdN5jM4Yjgf+BljO4030n11eeUwQS0TnKMONy5 kMj04MPcL14UTmKXLPpJ+4UMzVuLiMpaaqL5RSyPJcXJ1yy/b8v3HILxkU0CxA7i xKRSw1jnI7c0fQO9KaDXzb2jt3/vAJBAwTwDLjYG2okBn6vyFac= Received: (qmail 55616 invoked by alias); 15 Sep 2016 11:21:48 -0000 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Delivered-To: mailing list gcc-patches@gcc.gnu.org Received: (qmail 55591 invoked by uid 89); 15 Sep 2016 11:21:47 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-0.3 required=5.0 tests=AWL, BAYES_00, LIKELY_SPAM_BODY, RCVD_IN_DNSWL_LOW, RCVD_IN_SORBS_SPAM, SPF_PASS autolearn=no version=3.3.2 spammy=profit X-HELO: mail-it0-f52.google.com Received: from mail-it0-f52.google.com (HELO mail-it0-f52.google.com) (209.85.214.52) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Thu, 15 Sep 2016 11:21:37 +0000 Received: by mail-it0-f52.google.com with SMTP id n143so69791006ita.1 for ; Thu, 15 Sep 2016 04:21:37 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:mime-version:in-reply-to:references:from:date :message-id:subject:to; bh=gAAvx4yiaETmSGapSJKxXN9STRWQ6uZkYuCbuLq1FQo=; b=FxTKxl1lJ56BVqlRxV0rHGIWSwXUwNzGKo1U9jgQlY+Lh4euA6W1pzWj3P5a9EHKmU t8VcHVL9OLoRsLgnOuWEGi1JLsInKrJ0yaA6uzZYDvWTWouFksRM/eiLD0A0lnijydRI Shsq690YB/qFOAHefQ/ZqWdZRqcoN79ELfFszU+k/QywvsM3CwqY0vS9ndR9cCQaefel SRXkSHB42GnZO2VNPO5DA0qjhBv302XqbpGmvRhmilNteygRiF4TZx1daNO22dxvJMXs ra2wMD3zyzBNlfaDNhkrizLh190MHLbVXHshoV8LvF2DkA+1jueKpA0st9uLerjBfHLV 8mYw== X-Gm-Message-State: AE9vXwOH6+HBbVFrQBSjVSNyhYknDJnuKJAK4xpHMabnSlocWamFm1yOnpl9XADTxPVDtL26oi/B3XPhSWMikmX2 X-Received: by 10.107.132.157 with SMTP id o29mr18959156ioi.109.1473938494599; Thu, 15 Sep 2016 04:21:34 -0700 (PDT) MIME-Version: 1.0 Received: by 10.36.81.85 with HTTP; Thu, 15 Sep 2016 04:21:33 -0700 (PDT) In-Reply-To: <878tut77fg.fsf@e105548-lin.cambridge.arm.com> References: <87y42urtza.fsf@googlemail.com> <87twdirt7c.fsf@googlemail.com> <878tut77fg.fsf@e105548-lin.cambridge.arm.com> From: Prathamesh Kulkarni Date: Thu, 15 Sep 2016 16:51:33 +0530 Message-ID: Subject: Re: [PING] set libfunc entry for sdivmod_optab to NULL in optabs.def To: Prathamesh Kulkarni , gcc Patches , Richard Sandiford X-IsSubscribed: yes On 15 September 2016 at 16:31, Richard Sandiford wrote: > Prathamesh Kulkarni writes: >> On 15 September 2016 at 04:21, Richard Sandiford >> wrote: >>> Richard Sandiford writes: >>>> Prathamesh Kulkarni writes: >>>>> Hi, >>>>> I would like to ping the following patch: >>>>> https://gcc.gnu.org/ml/gcc-patches/2016-08/msg01015.html >>>>> >>>>> While implementing divmod transform: >>>>> https://gcc.gnu.org/ml/gcc-patches/2016-05/msg01757.html >>>>> >>>>> I ran into an issue with optab_libfunc(). >>>>> It appears optab_libfunc (sdivmod_optab, mode) returns >>>>> bogus libfunc for unsupported modes, for instance >>>>> on x86_64, optab_libfunc (sdivmod_optab, DImode) returns >>>>> a libfunc with name "__divmoddi4", even though such a libfunc >>>>> does not exist in libgcc. This happens because in optabs.def >>>>> the libfunc entry for sdivmod_optab has gen_int_libfunc, >>>>> and call to optab_libfunc (sdivmo_optab, DImode) lazily >>>>> creates a bogus libfunc "__divmoddi4" by calling gen_int_libfunc(). >>>>> >>>>> To work around this issue I set libfunc entry for sdivmod_optab to NULL >>>>> and verified that optab_libfunc (sdivmod_optab, DImode) returns NULL_RTX >>>>> instead of a bogus libfunc if it's not overriden by the target. >>>>> >>>>> Bootstrapped and tested on ppc64le-linux-gnu, x86_64-linux-gnu. >>>>> Cross tested on arm*-*-*, aarch64*-*-*. >>>>> OK for trunk ? >>>> >>>> I'm not a maintainer for this area, but: >>> >>> ...in https://gcc.gnu.org/ml/gcc-patches/2016-05/msg01757.html >>> you said that c6x follows the return-by-pointer convention. >>> I'm no c6x expert, but from a quick look, I think its divrem >>> function returns a div/mod pair in A4/A5, which matches the >>> ARM convention of returning both results by value. >>> >>> Does anyone know if the optab function registered by the SPU >>> backend is ever called directly? >>> >>> You mention that this is latent as far as expand_twoval_binop_libfunc >>> is concerned. AIUI expand_twoval_binop_libfunc implements the ARM/c6x >>> convention and expects the two values to be returned as a pair. >>> It then extracts one half of the pair and discards the other. >>> So my worry is that we're leaving the udivmod entry intact even though >>> the standard __udivmoddi4 doesn't do what expand_twoval_binop_libfunc >>> expects. >>> >>> Would it make sense to set both entries to null and treat __udivmoddi4 >>> as a non-optab function? ARM and c6x could then continue to register >>> their current optab functions and a non-null optab function would >>> indicate a return value pair. >> AFAIU, there are only three targets (c6x, spu, arm) that override >> optab_libfunc for udivmod_optab for following modes: >> ./c6x/c6x.c: set_optab_libfunc (udivmod_optab, SImode, "__c6xabi_divremu"); >> ./c6x/c6x.c: set_optab_libfunc (udivmod_optab, DImode, "__c6xabi_divremull"); >> ./arm/arm.c: set_optab_libfunc (udivmod_optab, DImode, "__aeabi_uldivmod"); >> ./arm/arm.c: set_optab_libfunc (udivmod_optab, SImode, "__aeabi_uidivmod"); >> ./spu/spu.c: set_optab_libfunc (udivmod_optab, DImode, "__udivmoddi4"); >> ./spu/spu.c: set_optab_libfunc (udivmod_optab, TImode, "__udivmodti4"); >> >> Out of these only the arm, and c6x have target-specific divmod libfuncs which >> return pair, while spu merely makes it point to the >> standard functions. >> >> So we could set libfunc entry for udivmod_optab to NULL, thus dropping >> support for generic >> divmod functions (__udivmoddi4, __udivmodti4). For targets that >> require standard divmod libfuncs like __udivmoddi4, >> they could explicitly override optab_libfunc and set it to >> __udivmoddi4, just as spu does. >> >> However this implies that non-null optab function doesn't necessarily >> follow arm/c6x convention. >> (i686-gcc for instance generates call to libgcc routines >> __udivdi3/__umoddi3 for DImode division/mod operations >> and could profit from divmod transform by calling __udivmoddi4). > > What I meant was that we shouldn't treat udivmoddi4 as an optab function > at all, but handle it with some on-the-side mechanism. That seems like > quite a natural fit if we handle the fused div/mod operation as an > internal function during gimple. Ah right, thanks for pointing out. So if optab function for [us]divmod_optab is defined, then it must follow the arm/c6x convention ? > > I think the current SPU code is wrong, but it looks like a latent bug. > (Like I say, does the udivmodti4 function that it registers ever > actually get used? It seems unlikely.) > > In that scenario no other targets should do what SPU does. I am testing the following patch which sets libfunc entries for both sdivmod_optab, udivmod_optab to NULL. This won't change the current (broken) behavior for SPU port since it explicitly overrides optab_libfunc for udivmod_optab and sets it to __udivmoddi4. Thanks, Prathamesh > > Thanks, > Richard diff --git a/gcc/optabs.def b/gcc/optabs.def index 8875e30..b37ac2e 100644 --- a/gcc/optabs.def +++ b/gcc/optabs.def @@ -116,8 +116,8 @@ OPTAB_NL(ssdiv_optab, "ssdiv$Q$a3", SS_DIV, "ssdiv", '3', gen_signed_fixed_libfu OPTAB_NL(udiv_optab, "udiv$I$a3", UDIV, "udiv", '3', gen_int_unsigned_fixed_libfunc) OPTAB_NX(udiv_optab, "udiv$Q$a3") OPTAB_NL(usdiv_optab, "usdiv$Q$a3", US_DIV, "usdiv", '3', gen_unsigned_fixed_libfunc) -OPTAB_NL(sdivmod_optab, "divmod$a4", UNKNOWN, "divmod", '4', gen_int_libfunc) -OPTAB_NL(udivmod_optab, "udivmod$a4", UNKNOWN, "udivmod", '4', gen_int_libfunc) +OPTAB_NL(sdivmod_optab, "divmod$a4", UNKNOWN, "divmod", '4', NULL) +OPTAB_NL(udivmod_optab, "udivmod$a4", UNKNOWN, "udivmod", '4', NULL) OPTAB_NL(smod_optab, "mod$a3", MOD, "mod", '3', gen_int_libfunc) OPTAB_NL(umod_optab, "umod$a3", UMOD, "umod", '3', gen_int_libfunc) OPTAB_NL(ftrunc_optab, "ftrunc$F$a2", UNKNOWN, "ftrunc", '2', gen_fp_libfunc)