From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on inbox.vuxu.org X-Spam-Level: X-Spam-Status: No, score=-3.1 required=5.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,RCVD_IN_DNSWL_MED,RCVD_IN_MSPIKE_H4, RCVD_IN_MSPIKE_WL,T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.4 Received: from second.openwall.net (second.openwall.net [193.110.157.125]) by inbox.vuxu.org (Postfix) with SMTP id A40E32B5EA for ; Wed, 12 Jun 2024 16:12:48 +0200 (CEST) Received: (qmail 1497 invoked by uid 550); 12 Jun 2024 14:12:43 -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 1457 invoked from network); 12 Jun 2024 14:12:43 -0000 Date: Wed, 12 Jun 2024 10:12:58 -0400 From: Rich Felker To: Meng Zhuo Cc: musl@lists.openwall.com Message-ID: <20240612141256.GR10433@brightrain.aerifal.cx> References: <20240612090018.258573-2-mzh@mzh.io> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20240612090018.258573-2-mzh@mzh.io> User-Agent: Mutt/1.5.21 (2010-09-15) Subject: Re: [musl] [PATCH 2/2] math: add riscv64 round/roundf On Wed, Jun 12, 2024 at 05:00:18PM +0800, Meng Zhuo wrote: > --- > src/math/riscv64/round.c | 21 +++++++++++++++++++++ > src/math/riscv64/roundf.c | 21 +++++++++++++++++++++ > 2 files changed, 42 insertions(+) > create mode 100644 src/math/riscv64/round.c > create mode 100644 src/math/riscv64/roundf.c > > diff --git a/src/math/riscv64/round.c b/src/math/riscv64/round.c > new file mode 100644 > index 00000000..12e670a5 > --- /dev/null > +++ b/src/math/riscv64/round.c > @@ -0,0 +1,21 @@ > +#include > + > +#if __riscv_flen >= 64 > + > +double round(double x) > +{ > + if (isnan(x)) return x; > + if (x == INFINITY || x == -INFINITY) return x; > + double tmp; > + long long n; > + __asm__ ("fcvt.l.d %0, %1, rmm" : "=r"(n) : "f"(x)); > + __asm__ ("fcvt.d.l %0, %1" : "=f"(tmp) : "r"(n)); > + x = copysign(tmp, x); > + return x; > +} > + > +#else > + > +#include "../round.c" > + > +#endif > diff --git a/src/math/riscv64/roundf.c b/src/math/riscv64/roundf.c > new file mode 100644 > index 00000000..8da3c65a > --- /dev/null > +++ b/src/math/riscv64/roundf.c > @@ -0,0 +1,21 @@ > +#include > + > +#if __riscv_flen >= 32 > + > +float roundf(float x) > +{ > + if (isnan(x)) return x; > + if (x == INFINITY || x == -INFINITY) return x; > + float tmp; > + long n; > + __asm__ ("fcvt.w.s %0, %1, rmm" : "=r"(n) : "f"(x)); > + __asm__ ("fcvt.s.w %0, %1" : "=f"(tmp) : "r"(n)); > + x = copysign(tmp, x); > + return x; > +} > + > +#else > + > +#include "../roundf.c" > + > +#endif > -- > 2.39.2 These functions clearly have the wrong behavior for inputs like 0x1p1000. As a general principle, you can't implement round in terms of lround/llround, only the opposite direction. You may be able to do it by first performing a range check, though. For example, if (!(fabs(x) < 0x1p54)) return x; Note that this also covers the non-finite cases, but I'm not sure if it avoids raising exceptions it shouldn't. If not, or if you weren't doing this, you can collapse the multiple inf/nan checks you have to: if (!isfinite(x)) return x; You should also probably use copysignf rather than copysign for roundf, or do an inline version of it. Rich