From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on inbox.vuxu.org X-Spam-Level: X-Spam-Status: No, score=0.7 required=5.0 tests=MAILING_LIST_MULTI, RCVD_IN_MSPIKE_H2,URIBL_BLACK autolearn=no autolearn_force=no version=3.4.4 Received: (qmail 12549 invoked from network); 9 Dec 2022 21:30:31 -0000 Received: from second.openwall.net (193.110.157.125) by inbox.vuxu.org with ESMTPUTF8; 9 Dec 2022 21:30:31 -0000 Received: (qmail 9718 invoked by uid 550); 9 Dec 2022 21:30:28 -0000 Mailing-List: contact musl-help@lists.openwall.com; run by ezmlm Precedence: bulk List-Post: List-Help: List-Unsubscribe: List-Subscribe: List-ID: Reply-To: musl@lists.openwall.com Received: (qmail 9683 invoked from network); 9 Dec 2022 21:30:27 -0000 Date: Fri, 9 Dec 2022 22:30:15 +0100 From: Szabolcs Nagy To: gzh2233 Cc: musl@lists.openwall.com Message-ID: <20221209213015.GH98588@port70.net> Mail-Followup-To: gzh2233 , musl@lists.openwall.com References: <627bcd3b.2804.184f0bc7d37.Coremail.gzh1754344759@163.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <627bcd3b.2804.184f0bc7d37.Coremail.gzh1754344759@163.com> Subject: Re: [musl] ld128 math function support * gzh2233 [2022-12-08 15:56:35 +0800]: > I am trying to use ld128 in x86_64 musl and i have some problems. that has multiple issues. > Long double format is defined in arch/x86_64/bits/float.h and i am trying to > change long double relevant macros. However, many long double functions > still use double implementions(mostly "xxhl" functions). This is just a way to > make musl compile and can be a problem in some cases. Simply changing > the long double relevant macros can change ld80 into ld128 but this > problem still exists. > > > Is there any official support for long double math functions? musl does not support long double == float128 on x86_64. this is part of the abi. if you add a new abi variant where long double is float128, that works but many float128 math functions have low quality currently as you noticed. fixing this is significant effort. i don't have immediate plans to fix this. another option is to keep long double as is but add _Float128 support and f128 math functions (on all targets). such support would require additional work and increase the size of libc.so.