From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1NxIzg-0007Tl-JO for qemu-devel@nongnu.org; Thu, 01 Apr 2010 07:55:28 -0400 Received: from [140.186.70.92] (port=44698 helo=eggs.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1NxIzf-0007T9-A9 for qemu-devel@nongnu.org; Thu, 01 Apr 2010 07:55:28 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.69) (envelope-from ) id 1NxIzc-0000Xn-SP for qemu-devel@nongnu.org; Thu, 01 Apr 2010 07:55:26 -0400 Received: from mx20.gnu.org ([199.232.41.8]:39375) by eggs.gnu.org with esmtp (Exim 4.69) (envelope-from ) id 1NxIzc-0000Xi-QZ for qemu-devel@nongnu.org; Thu, 01 Apr 2010 07:55:24 -0400 Received: from mail.codesourcery.com ([38.113.113.100]) by mx20.gnu.org with esmtp (Exim 4.60) (envelope-from ) id 1NxIzb-0000Mr-BF for qemu-devel@nongnu.org; Thu, 01 Apr 2010 07:55:23 -0400 From: Paul Brook Subject: Re: [Qemu-devel] GSoC projects about AHCI and S3 Trio Date: Thu, 1 Apr 2010 11:55:19 +0000 References: <201003301314.03044.paul@codesourcery.com> In-Reply-To: MIME-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-15" Content-Transfer-Encoding: 7bit Message-Id: <201004011255.19796.paul@codesourcery.com> List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Roland Elek > I will contact jai soon about the S3 Trio. I'll also skim through the > documentation so that I can set reasonable goals in my timeline. TBH I'm less convinced about the utility of emulating an S3 graphics card. You'd want to make sure you know what you're expecting from it. While it may have a bit wider OS support that the Cirrus Logic cards, it's still an extremely old card that AFAIK ceased production many years ago. In modern terms it also has extremely limited capabilities. Even the 2D engine is likely to be tricky accelerate because. The framebuffer is exposed to the guest, so you have to accomodate the guest reading back the results. Some models have rudimentary 3D acceleration, however in practice I doubt this is worth much. I did a bit of 3D graphics when these cards were common, and IIRC they're missing some common features (multitexturing, possibly even prespective correction) and [as real hardware] are generally slower than a good software renderer. If you want useful 3D acceleration, then you pretty much have to go for a paravirtual interface. The interface exposed by real hardware is closely tied to the implementation, which makes it hard (and slow) to emulate. This has been discussed a few times before, and amounts to feeding OpenGL commands (or equivalents such as Gallium pipes) over a remote transport. Paul