From mboxrd@z Thu Jan 1 00:00:00 1970 X-Msuck: nntp://news.gmane.org/gmane.linux.lib.musl.general/3629 Path: news.gmane.org!not-for-mail From: Rich Felker Newsgroups: gmane.linux.lib.musl.general Subject: Re: Thinking about release Date: Thu, 11 Jul 2013 08:49:03 -0400 Message-ID: <20130711124903.GP29800@brightrain.aerifal.cx> References: <20130613012517.GA5859@brightrain.aerifal.cx> <20130613014314.GC29800@brightrain.aerifal.cx> <20130709053711.GO29800@brightrain.aerifal.cx> <20130711033754.GL29800@brightrain.aerifal.cx> Reply-To: musl@lists.openwall.com NNTP-Posting-Host: plane.gmane.org Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit X-Trace: ger.gmane.org 1373546958 13391 80.91.229.3 (11 Jul 2013 12:49:18 GMT) X-Complaints-To: usenet@ger.gmane.org NNTP-Posting-Date: Thu, 11 Jul 2013 12:49:18 +0000 (UTC) To: musl@lists.openwall.com Original-X-From: musl-return-3633-gllmg-musl=m.gmane.org@lists.openwall.com Thu Jul 11 14:49:21 2013 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 1UxGJ9-0008PV-SS for gllmg-musl@plane.gmane.org; Thu, 11 Jul 2013 14:49:15 +0200 Original-Received: (qmail 1393 invoked by uid 550); 11 Jul 2013 12:49:15 -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 1384 invoked from network); 11 Jul 2013 12:49:15 -0000 Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.21 (2010-09-15) Xref: news.gmane.org gmane.linux.lib.musl.general:3629 Archived-At: On Thu, Jul 11, 2013 at 07:27:11AM +0200, Daniel Cegiełka wrote: > 2013/7/11 Rich Felker : > > On Thu, Jul 11, 2013 at 10:44:16AM +1200, Andre Renaud wrote: > > > Yes, I suspect so. One slightly crazy idea I had was to write the > > function in C with just inline asm for the inner ldm/stm loop. > > A bit of useful code (x86): > > http://dpdk.org/browse/dpdk/tree/lib/librte_eal/common/include/rte_memcpy.h On modern x86 (32-bit), this is slower than even the naive "rep movsb" version. Some x86 chips have problems with rep movsb, so the version in musl does a little bit more work (possibly more than it needs to) to use "rep movsd". On x86_64, there _may_ be faster approaches than the "rep movsq" we have right now, but so far my impression is that they don't work on baseline x86_64 (only later variants) and don't gain much. Rich