From patchwork Tue Sep 13 15:42:49 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Paul E. Murphy" X-Patchwork-Id: 15586 Received: (qmail 89710 invoked by alias); 13 Sep 2016 15:43:02 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 89690 invoked by uid 89); 13 Sep 2016 15:43:01 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-1.5 required=5.0 tests=AWL, BAYES_00, KAM_LAZY_DOMAIN_SECURITY, RCVD_IN_DNSWL_LOW autolearn=no version=3.3.2 spammy=vd, murphy, subnormal, _float128 X-HELO: mx0a-001b2d01.pphosted.com X-IBM-Helo: d03dlp01.boulder.ibm.com X-IBM-MailFrom: murphyp@linux.vnet.ibm.com Subject: Re: [PATCHv4] ldbl-128: Use L(x) macro for long double constants To: Joseph Myers References: <997588c1-5bcf-6339-4c7f-1339d91e45cf@linux.vnet.ibm.com> <93742807-3c5e-0d3b-3f49-34e264afd855@linux.vnet.ibm.com> Cc: "libc-alpha@sourceware.org" From: "Paul E. Murphy" Date: Tue, 13 Sep 2016 10:42:49 -0500 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.2.0 MIME-Version: 1.0 In-Reply-To: X-TM-AS-GCONF: 00 X-Content-Scanned: Fidelis XPS MAILER x-cbid: 16091315-0020-0000-0000-000009C525EF X-IBM-SpamModules-Scores: X-IBM-SpamModules-Versions: BY=3.00005754; HX=3.00000240; KW=3.00000007; PH=3.00000004; SC=3.00000185; SDB=6.00757285; UDB=6.00359075; IPR=6.00530656; BA=6.00004715; NDR=6.00000001; ZLA=6.00000005; ZF=6.00000009; ZB=6.00000000; ZP=6.00000000; ZH=6.00000000; ZU=6.00000002; MB=3.00012661; XFM=3.00000011; UTC=2016-09-13 15:42:52 X-IBM-AV-DETECTION: SAVI=unused REMOTE=unused XFE=unused x-cbparentid: 16091315-0021-0000-0000-00005570A35D Message-Id: <0d7ac509-f338-fa7c-9cbd-2e0f5d845788@linux.vnet.ibm.com> X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10432:, , definitions=2016-09-13_09:, , signatures=0 X-Proofpoint-Spam-Details: rule=outbound_notspam policy=outbound score=0 spamscore=0 suspectscore=0 malwarescore=0 phishscore=0 adultscore=0 bulkscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.0.1-1609020000 definitions=main-1609130229 On 09/09/2016 09:14 AM, Joseph Myers wrote: > On Thu, 8 Sep 2016, Paul E. Murphy wrote: > >> On 09/08/2016 05:56 PM, Joseph Myers wrote: >>> On Fri, 2 Sep 2016, Paul E. Murphy wrote: >>> >>>> @@ -65,7 +65,7 @@ __ieee754_atan2l(_Float128 y, _Float128 x) >>>> if(((ix|((lx|-lx)>>63))>0x7fff000000000000LL)|| >>>> ((iy|((ly|-ly)>>63))>0x7fff000000000000LL)) /* x or y is NaN */ >>>> return x+y; >>>> - if(((hx-0x3fff000000000000LL)|lx)==0) return __atanl(y); /* x=1.0L */ >>>> + if(((hx-0x3fff000000000000LL)|lx)==0) return __atanl(y); /* x=1 */ >>> >>> Is this change of a comment deliberate? I thought you were avoiding >>> changing comments. >> >> This is a limitation of the conversion script. It only skips lines >> which start with a /* comment. This can be reverted via the fixup, >> patch or I can note the limitation in the commit message. > > The patch is OK with the comment change reverted. > I'll commit with the updated messages: ldbl-128: Use L(x) macro for long double constants This runs the attached sed script against these files using a regex which aggressively matches long double literals when not obviously part of a comment. Likewise, 5 digit or less integral constants are replaced with integral constants, excepting the two cases of 0 used in large tables, which are also the only integral values of the form x.0*E0L encountered within these converted files. Likewise, -L(x) is transformed into L(-x). Naturally, the script has a few minor hiccups which are more clearly remedied via the attached fixup patch. Such hiccups include, context-sensitive promotion to a real type, and munging constants inside harder to detect comment blocks. * sysdeps/ieee754/ldbl-128/e_acoshl.c: Wrap long double literals with L() and remove explicit suffix, convert real literals which can be equivalentally substituted with integers, and transform -L(x) to L(-x). * sysdeps/ieee754/ldbl-128/e_acosl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_asinl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_atan2l.c: Likewise. * sysdeps/ieee754/ldbl-128/e_atanhl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_coshl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_exp10l.c: Likewise. * sysdeps/ieee754/ldbl-128/e_expl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_gammal_r.c: Likewise. * sysdeps/ieee754/ldbl-128/e_hypotl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_j0l.c: Likewise. * sysdeps/ieee754/ldbl-128/e_j1l.c: Likewise. * sysdeps/ieee754/ldbl-128/e_jnl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_lgammal_r.c: Likewise. * sysdeps/ieee754/ldbl-128/e_log10l.c: Likewise. * sysdeps/ieee754/ldbl-128/e_log2l.c: Likewise. * sysdeps/ieee754/ldbl-128/e_logl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_powl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_rem_pio2l.c: Likewise. * sysdeps/ieee754/ldbl-128/e_remainderl.c: Likewise. * sysdeps/ieee754/ldbl-128/e_sinhl.c: Likewise. * sysdeps/ieee754/ldbl-128/k_cosl.c: Likewise. * sysdeps/ieee754/ldbl-128/k_sincosl.c: Likewise. * sysdeps/ieee754/ldbl-128/k_sinl.c: Likewise. * sysdeps/ieee754/ldbl-128/k_tanl.c: Likewise. * sysdeps/ieee754/ldbl-128/lgamma_negl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_asinhl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_atanl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_cbrtl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_cosl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_erfl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_expm1l.c: Likewise. * sysdeps/ieee754/ldbl-128/s_fmal.c: Likewise. * sysdeps/ieee754/ldbl-128/s_frexpl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_llrintl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_llroundl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_log1pl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_lrintl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_lroundl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_nearbyintl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_nextafterl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_remquol.c: Likewise. * sysdeps/ieee754/ldbl-128/s_scalblnl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_scalbnl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_sincosl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_sinl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_tanhl.c: Likewise. * sysdeps/ieee754/ldbl-128/s_tanl.c: Likewise. * sysdeps/ieee754/ldbl-128/t_expl.h: Likewise. * sysdeps/ieee754/ldbl-128/t_sincosl.c: Likewise. * sysdeps/ieee754/ldbl-128/x2y2m1l.c: Likewise. With the updated fixup patch to revert the 1.0L -> 1 change in noted comment of e_atan2l.c. The script remains unchanged. diff --git a/sysdeps/ieee754/ldbl-128/e_atan2l.c b/sysdeps/ieee754/ldbl-128/e_atan2l.c index d51a971..faecd1a 100644 --- a/sysdeps/ieee754/ldbl-128/e_atan2l.c +++ b/sysdeps/ieee754/ldbl-128/e_atan2l.c @@ -65,7 +65,7 @@ __ieee754_atan2l(_Float128 y, _Float128 x) if(((ix|((lx|-lx)>>63))>0x7fff000000000000LL)|| ((iy|((ly|-ly)>>63))>0x7fff000000000000LL)) /* x or y is NaN */ return x+y; - if(((hx-0x3fff000000000000LL)|lx)==0) return __atanl(y); /* x=1 */ + if(((hx-0x3fff000000000000LL)|lx)==0) return __atanl(y); /* x=1.0L */ m = ((hy>>63)&1)|((hx>>62)&2); /* 2*sign(x)+sign(y) */ /* when y = 0 */ diff --git a/sysdeps/ieee754/ldbl-128/e_exp10l.c b/sysdeps/ieee754/ldbl-128/e_exp10l.c index 9a2c8b8..9432b18 100644 --- a/sysdeps/ieee754/ldbl-128/e_exp10l.c +++ b/sysdeps/ieee754/ldbl-128/e_exp10l.c @@ -39,7 +39,7 @@ __ieee754_exp10l (_Float128 arg) return 1; u.value = arg; - u.parts64sw &= 0xfe00000000000000LL; + u.parts64.lsw &= 0xfe00000000000000LL; arg_high = u.value; arg_low = arg - arg_high; exp_high = arg_high * log10_high; diff --git a/sysdeps/ieee754/ldbl-128/e_jnl.c b/sysdeps/ieee754/ldbl-128/e_jnl.c index 1f80b00..470631e 100644 --- a/sysdeps/ieee754/ldbl-128/e_jnl.c +++ b/sysdeps/ieee754/ldbl-128/e_jnl.c @@ -328,7 +328,7 @@ __ieee754_ynl (int n, _Float128 x) if (x <= 0) { if (x == 0) - return ((n < 0 && (n & 1) != 0) ? 1 : -1) / 0; + return ((n < 0 && (n & 1) != 0) ? 1 : -1) / L(0.0); if (se & 0x80000000) return zero / (zero * x); } diff --git a/sysdeps/ieee754/ldbl-128/lgamma_negl.c b/sysdeps/ieee754/ldbl-128/lgamma_negl.c index 601d8f8..0969dcc 100644 --- a/sysdeps/ieee754/ldbl-128/lgamma_negl.c +++ b/sysdeps/ieee754/ldbl-128/lgamma_negl.c @@ -450,7 +450,7 @@ __lgamma_negl (_Float128 x, int *signgamp) integers and determine the sign of the result. */ int i = __floorl (-2 * x); if ((i & 1) == 0 && i == -2 * x) - return 1 / 0; + return L(1.0) / L(0.0); _Float128 xn = ((i & 1) == 0 ? -i / 2 : (-i - 1) / 2); i -= 4; *signgamp = ((i & 2) == 0 ? -1 : 1); diff --git a/sysdeps/ieee754/ldbl-128/s_fmal.c b/sysdeps/ieee754/ldbl-128/s_fmal.c index 7b12441..eb1e27e 100644 --- a/sysdeps/ieee754/ldbl-128/s_fmal.c +++ b/sysdeps/ieee754/ldbl-128/s_fmal.c @@ -261,10 +261,10 @@ __fmal (_Float128 x, _Float128 y, _Float128 z) if (v.ieee.exponent > 228) return (a1 + u.d) * L(0x1p-228); /* If v.d * 0x1p-228L with round to zero is a subnormal above - or equal to LDBL_MIN / 2, then v.d * L(0x1p-228) shifts mantissa + or equal to LDBL_MIN / 2, then v.d * 0x1p-228L shifts mantissa down just by 1 bit, which means v.ieee.mantissa3 |= j would change the round bit, not sticky or guard bit. - v.d * L(0x1p-228) never normalizes by shifting up, + v.d * 0x1p-228L never normalizes by shifting up, so round bit plus sticky bit should be already enough for proper rounding. */ if (v.ieee.exponent == 228)