From mboxrd@z Thu Jan 1 00:00:00 1970 X-Msuck: nntp://news.gmane.org/gmane.linux.lib.musl.general/701 Path: news.gmane.org!not-for-mail From: Rich Felker Newsgroups: gmane.linux.lib.musl.general Subject: Re: float scanner status, upcoming release Date: Tue, 10 Apr 2012 09:02:00 -0400 Message-ID: <20120410130200.GC7281@brightrain.aerifal.cx> References: <20120409191712.GA5269@brightrain.aerifal.cx> <20120410004819.GB7281@brightrain.aerifal.cx> <20120410121515.GU5728@port70.net> Reply-To: musl@lists.openwall.com NNTP-Posting-Host: plane.gmane.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Trace: dough.gmane.org 1334062809 27517 80.91.229.3 (10 Apr 2012 13:00:09 GMT) X-Complaints-To: usenet@dough.gmane.org NNTP-Posting-Date: Tue, 10 Apr 2012 13:00:09 +0000 (UTC) To: musl@lists.openwall.com Original-X-From: musl-return-702-gllmg-musl=m.gmane.org@lists.openwall.com Tue Apr 10 15:00:09 2012 Return-path: Envelope-to: gllmg-musl@plane.gmane.org Original-Received: from mother.openwall.net ([195.42.179.200]) by plane.gmane.org with smtp (Exim 4.69) (envelope-from ) id 1SHag4-0001uk-4A for gllmg-musl@plane.gmane.org; Tue, 10 Apr 2012 15:00:08 +0200 Original-Received: (qmail 30124 invoked by uid 550); 10 Apr 2012 13:00:07 -0000 Mailing-List: contact musl-help@lists.openwall.com; run by ezmlm Precedence: bulk List-Post: List-Help: List-Unsubscribe: List-Subscribe: Original-Received: (qmail 30116 invoked from network); 10 Apr 2012 13:00:07 -0000 Content-Disposition: inline In-Reply-To: <20120410121515.GU5728@port70.net> User-Agent: Mutt/1.5.21 (2010-09-15) Xref: news.gmane.org gmane.linux.lib.musl.general:701 Archived-At: On Tue, Apr 10, 2012 at 02:15:15PM +0200, Szabolcs Nagy wrote: > * Rich Felker [2012-04-09 20:48:19 -0400]: > > Revised version of the code, in preparation to integrate with musl. > > Still needs hex float support, and I want to eliminate the fscanf > > dependency. Comments welcome. > > > > this one fails if the number starts with a . > ../floatscan .3 This is a bug in the wrapper code before it gets to decfloat() and it's easily fixed. > large exponent is slow: > ../floatscan 1e+1000000 > ../floatscan 1e-1000000 I added code that collapses large and tiny exponents to infinity and zero, but the max LD exponent is still rather slow... > subnormal tests still fail > > here are some test cases where many digits are needed for correct rounding: Indeed. Thanks for the tests. I'll see what I can make of it. Rich