From patchwork Fri Mar 14 13:53:11 2014 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Adhemerval Zanella Netto X-Patchwork-Id: 87 X-Patchwork-Delegate: azanella@linux.vnet.ibm.com Return-Path: X-Original-To: siddhesh@wilcox.dreamhost.com Delivered-To: siddhesh@wilcox.dreamhost.com Received: from homiemail-mx23.g.dreamhost.com (caibbdcaabja.dreamhost.com [208.113.200.190]) by wilcox.dreamhost.com (Postfix) with ESMTP id BDB2B3600AB for ; Fri, 14 Mar 2014 06:53:23 -0700 (PDT) Received: by homiemail-mx23.g.dreamhost.com (Postfix, from userid 14307373) id 72A6A6366F097; Fri, 14 Mar 2014 06:53:23 -0700 (PDT) X-Original-To: glibc@patchwork.siddhesh.in Delivered-To: x14307373@homiemail-mx23.g.dreamhost.com Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by homiemail-mx23.g.dreamhost.com (Postfix) with ESMTPS id 4DE9B6365AD3E for ; Fri, 14 Mar 2014 06:53:23 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; c=nofws; d=sourceware.org; h=list-id :list-unsubscribe:list-subscribe:list-archive:list-post :list-help:sender:message-id:date:from:mime-version:to:subject :content-type:content-transfer-encoding; q=dns; s=default; b=reM 2L9bZE1cbIquesw3Iye967D6DxMPyPjvCGMs66N+W3Ft6WWL1Dr9uQNHtEQ0U9cc U4qdT4Tg7q468JV4xJfiaiNfwUNk/eXIXFlxtibK9q/4fIYniGpNBCXwkmrEhjvY xgRvYmrv2d9d1w6onaMBdu+mRyRUOMS/MOLDoNGk= DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=sourceware.org; h=list-id :list-unsubscribe:list-subscribe:list-archive:list-post :list-help:sender:message-id:date:from:mime-version:to:subject :content-type:content-transfer-encoding; s=default; bh=vlzu3abrk LYRhdoafQjUGL/swTg=; b=r45O0YqnT+EYAxdxuZwGo/y4ksjGTCfcebFGE85Bp vWs7rQbXquRVdb1lBR7NOynqqetWPj3DLbyzQl94pHgJIp5CO//AGYdT3p1uOiXY cg74xQ/fv7VkakoShsqJkM45pjqVr2IHD94FamCq5mxElKYiryz54LDnUPYpMksk p4= Received: (qmail 8537 invoked by alias); 14 Mar 2014 13:53:21 -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 8527 invoked by uid 89); 14 Mar 2014 13:53:20 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-2.8 required=5.0 tests=AWL, BAYES_00, T_RP_MATCHES_RCVD autolearn=ham version=3.3.2 X-HELO: e24smtp04.br.ibm.com Message-ID: <532309C7.6070303@linux.vnet.ibm.com> Date: Fri, 14 Mar 2014 10:53:11 -0300 From: Adhemerval Zanella User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:24.0) Gecko/20100101 Thunderbird/24.3.0 MIME-Version: 1.0 To: "GNU C. Library" Subject: [PATCH 2/3] PowerPC: remove wrong nearbyintl implementation for, PowerPC64 X-TM-AS-MML: No X-Content-Scanned: Fidelis XPS MAILER x-cbid: 14031413-1820-0000-0000-0000001B2306 X-DH-Original-To: glibc@patchwork.siddhesh.in The nearbyintl assembly implementation (sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S) returns wrong results for some inputs where first double is a exact integer and the precision is determined by second long double. Checking on implementation comments and history, I am very confident the assembly implementation was based on a version before commit 5c68d401698a58cf7da150d9cce769fa6679ba5f that fixes BZ#2423 (Errors in long double (ldbl-128ibm) rounding functions in glibc-2.4). By just removing the implementation and make the build select sysdeps/ieee754/ldbl-128ibm/s_nearbyintl.c instead fixes the failing math. Fixes BZ#16706. Test on PPC64 and PPC64LE. If no one opposes it, I'll commit in a couple of hours. --- 2014-03-14 Adhemerval Zanella [BZ #16706] * sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S: Remove wrong implementation. * math/libm-test.inc (nearbyint_test_data): Add more tests. --- diff --git a/NEWS b/NEWS index 16a17c8..c9bfcb8 100644 --- a/NEWS +++ b/NEWS @@ -11,7 +11,7 @@ Version 2.20 15347, 15804, 15894, 16447, 16532, 16545, 16574, 16600, 16609, 16610, 16611, 16613, 16623, 16632, 16639, 16642, 16670, 16674, 16677, 16683, - 16689, 16695, 16701. + 16689, 16695, 16701, 16706. * The am33 port, which had not worked for several years, has been removed from ports. diff --git a/math/libm-test.inc b/math/libm-test.inc index 59cfe61..ee0a353 100644 --- a/math/libm-test.inc +++ b/math/libm-test.inc @@ -9519,6 +9519,14 @@ static const struct test_f_f_data nearbyint_test_data[] = TEST_f_f (nearbyint, -562949953421312.75, -562949953421313.0, NO_INEXACT_EXCEPTION), TEST_f_f (nearbyint, -1125899906842624.75, -1125899906842625.0, NO_INEXACT_EXCEPTION), #endif +#ifdef TEST_LDOUBLE + /* Check cases where first double is a exact integer higher than 2^52 and + the precision is determined by second long double for IBM long double. */ + TEST_f_f (nearbyint, 34503599627370498.515625L, 34503599627370499.0L), + TEST_f_f (nearbyint, 1192568192774434123539907640624.484375L, 1192568192774434123539907640624.0L), + TEST_f_f (nearbyint, -34503599627370498.515625L, -34503599627370499.0L), + TEST_f_f (nearbyint, -1192568192774434123539907640624.484375L, -1192568192774434123539907640624.0L), +#endif }; static void diff --git a/sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S b/sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S deleted file mode 100644 index acd95da..0000000 --- a/sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S +++ /dev/null @@ -1,113 +0,0 @@ -/* nearbyint long double. - IBM extended format long double version. - Copyright (C) 2004-2014 Free Software Foundation, Inc. - This file is part of the GNU C Library. - - The GNU C Library is free software; you can redistribute it and/or - modify it under the terms of the GNU Lesser General Public - License as published by the Free Software Foundation; either - version 2.1 of the License, or (at your option) any later version. - - The GNU C Library is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - Lesser General Public License for more details. - - You should have received a copy of the GNU Lesser General Public - License along with the GNU C Library; if not, see - . */ - -#include -#include - - .section ".toc","aw" -.LC0: /* 2**52 */ - .tc FD_43300000_0[TC],0x4330000000000000 - .section ".text" - -/* long double [fp1,fp2] nearbyintl (long double x [fp1,fp2]) - IEEE 1003.1 nearbyintl function. nearbyintl is similar to the rintl - but does raise the "inexact" exception. This implementation is - based on rintl but explicitly masks the inexact exception on entry - and clears any pending inexact before restoring the exception mask - on exit. - - PowerPC64 long double uses the IBM extended format which is - represented two 64-floating point double values. The values are - non-overlapping giving an effective precision of 106 bits. The first - double contains the high order bits of mantissa and is always rounded - to represent a normal rounding of long double to double. Since the - long double value is sum of the high and low values, the low double - normally has the opposite sign to compensate for the this rounding. - - For long double there are two cases: - 1) |x| < 2**52, all the integer bits are in the high double. - floor the high double and set the low double to -0.0. - 2) |x| >= 2**52, Rounding involves both doubles. - See the comment before label .L2 for details. - */ -ENTRY (__nearbyintl) - mffs fp11 /* Save current FPSCR. */ - lfd fp13,.LC0@toc(2) - fabs fp0,fp1 - mtfsb0 28 /* Disable "inexact" exceptions. */ - fsub fp12,fp13,fp13 /* generate 0.0 */ - fabs fp9,fp2 - fcmpu cr7,fp0,fp13 /* if (fabs(x) > TWO52) */ - fcmpu cr6,fp1,fp12 /* if (x > 0.0) */ - bnl- cr7,.L2 - fmr fp2,fp12 - bng- cr6,.L4 - fadd fp1,fp1,fp13 /* x+= TWO52; */ - fsub fp1,fp1,fp13 /* x-= TWO52; */ - b .L9 -.L4: - bnl- cr6,.L9 /* if (x < 0.0) */ - fsub fp1,fp13,fp1 /* x = TWO52 - x; */ - fsub fp0,fp1,fp13 /* x = - (x - TWO52); */ - fneg fp1,fp0 -.L9: - mtfsb0 6 /* Clear any pending "inexact" exceptions. */ - mtfsf 0x01,fp11 /* restore exception mask. */ - blr - -/* The high double is > TWO52 so we need to round the low double and - perhaps the high double. This gets a bit tricky so we use the - following algorithm: - - tau = floor(x_high/TWO52); - x0 = x_high - tau; - x1 = x_low + tau; - r1 = nearbyint(x1); - y_high = x0 + r1; - y_low = r1 - tau; - return y; */ -.L2: - fcmpu cr7,fp9,fp13 /* if (|x_low| > TWO52) */ - fcmpu cr0,fp9,fp12 /* || (|x_low| == 0.0) */ - bge- cr7,.L9 /* return x; */ - beq- cr0,.L9 - fdiv fp8,fp1,fp13 /* x_high/TWO52 */ - fctidz fp0,fp8 - fcfid fp8,fp0 /* tau = floor(x_high/TWO52); */ - fsub fp3,fp1,fp8 /* x0 = x_high - tau; */ - fadd fp4,fp2,fp8 /* x1 = x_low + tau; */ - - fcmpu cr6,fp4,fp12 /* if (x1 > 0.0) */ - bng- cr6,.L8 - fadd fp5,fp4,fp13 /* r1 = x1 + TWO52; */ - fsub fp5,fp5,fp13 /* r1 = r1 - TWO52; */ - b .L6 -.L8: - fmr fp5,fp4 - bge- cr6,.L6 /* if (x1 < 0.0) */ - fsub fp5,fp13,fp4 /* r1 = TWO52 - x1; */ - fsub fp0,fp5,fp13 /* r1 = - (r1 - TWO52); */ - fneg fp5,fp0 -.L6: - fadd fp1,fp3,fp5 /* y_high = x0 + r1; */ - fsub fp2,fp5,fp8 /* y_low = r1 - tau; */ - b .L9 -END (__nearbyintl) - -long_double_symbol (libm, __nearbyintl, nearbyintl)