From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <080126488105ed00ad89215c1995194d@quintile.net> From: "Steve Simon" Date: Mon, 22 Dec 2008 11:02:16 +0000 To: 9fans@9fans.net In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Subject: Re: [9fans] graphics scaling Topicbox-Message-UUID: 6bd61bfc-ead4-11e9-9d60-3106f5b1d025 Re: fusion would 9vx running under linux be an interesting option? re: resample If you are experimenting with interploation a zoneplate is one of the best test signals to check your work on, though its interpretation can be tricky. /n/sources/contrib/steve/zoneplate.c http://downloads.bbc.co.uk/rd/pubs/reports/1980-09.pdf http://downloads.bbc.co.uk/rd/pubs/reports/1978-23.pdf I thought I remembered page(1) had a simpler interpolation algorithm which was much faster but much more prone to aliasing, however looking at the code it seems to use the same kaiser window method. maybe somone else will remember that code? for a long time the problem of using a graphics co-processor as a general purpose Graphics CPU is the speed at which the processed image can be read back from teh card. I believe there are cards with dual pci express buses which allow symetric access to the frame buffer which sidestep this problem (see discussion on 9fans a month or so ago). If you are resizing the image only for display like page(1) does this is a non-problem of course. -Steve