From owner-freebsd-multimedia@FreeBSD.ORG Mon Feb 21 10:43:02 2005 Return-Path: Delivered-To: freebsd-multimedia@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id A7B6416A4CE for ; Mon, 21 Feb 2005 10:43:02 +0000 (GMT) Received: from sohara.org (sohara.org [192.220.64.179]) by mx1.FreeBSD.org (Postfix) with ESMTP id 6B1EA43D55 for ; Mon, 21 Feb 2005 10:43:02 +0000 (GMT) (envelope-from steve@sohara.org) Received: (qmail 17831 invoked by uid 16563); 21 Feb 2005 10:43:01 -0000 Received: from unknown (HELO SOHARA) ([217.12.14.195]) (envelope-sender ) by 192.220.64.179 (qmail-ldap-1.03) with SMTP for ; 21 Feb 2005 10:43:01 -0000 Date: Mon, 21 Feb 2005 10:44:23 +0000 From: Steve O'Hara-Smith To: Paul Chvostek Message-Id: <20050221104423.24956bf7.steve@sohara.org> In-Reply-To: <20050218145654.GA31961@it.ca> References: <20050207032841.GA33816@it.ca> <20050207100521.544ed9bc.steve@sohara.org> <20050209180336.GA28606@it.ca> <20050210095713.3b155ce6.steve@sohara.org> <20050213182120.GT40151@it.ca> <20050214102331.0380d1b8.steve@sohara.org> <20050218145654.GA31961@it.ca> X-Mailer: Sylpheed version 1.9.2 (GTK+ 2.4.1; i586-pc-interix3) X-Face: %]+HVL}K`P8>+8ZcY-WGHP6j@&mxMo9JH6_WdgIgUGH)JX/usO0%jy7T~IVgqjumD^OBqX,Kv^-GM6mlw(fI^$"QRKyZ$?xx/ Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit cc: freebsd-multimedia@freebsd.org Subject: Re: ffmpeg at half speed ... sort of. X-BeenThere: freebsd-multimedia@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Multimedia discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 21 Feb 2005 10:43:02 -0000 On Fri, 18 Feb 2005 09:56:55 -0500 Paul Chvostek wrote: > So ... if the same video signal and same hardware shows these wildly > inconsistent framerates encoding with ffmpeg in FreeBSD but not in > Fedora, it's likely a problem in the bktr driver? No I think it's in the low level timekeeping that usleep uses rather than anything specific to bktr. There isn't much around that I know of other than video capture that is particularly sensitive to accurate usleep though. I rather suspect ACPI is screwing up fine timekeeping. The Linux grab code also uses a timer fallback for missed sync signals (it's a pretty common problem) but is less noisy about it than my BSD grab code in ffmpeg. > I get the same behaviour with multiple video sources (both analog and > digital satellite video decoders, a handycam, a VCR, various cables) and > multiple encoder cards (AVerMedia cx878 and an old Intel bt848) ... and > of course, as I said, the problem seems exclusive to FreeBSD (tested > with both 4.11 and 5.3). Do both 4.11 and 5.3 crash with ACPI disabled ? > > kern.timecounter.hardware should be the one - from the looks of it. Try setting > > it to TSC or i8254 (probably TSC will do a better job). > > Tried both, to no avail; the symptoms remain. In a word ... bugger! -- C:>WIN | Directable Mirror Arrays The computer obeys and wins. | A better way to focus the sun You lose and Bill collects. | licences available see | http://www.sohara.org/