Path: chuka.playstation.co.uk!news From: "Martin Keates" Newsgroups: scee.yaroze.freetalk.english Subject: Re: negative scrolling on homebrew map routine - need help!! Date: Mon, 26 Mar 2001 02:06:46 +0100 Organization: PlayStation Net Yaroze (SCEE) Lines: 37 Message-ID: <99m4ju$roa2@www.netyaroze-europe.com> References: <3AAF7BB2.B427AED3@harbinger.com> <98o38v$ij81@www.netyaroze-europe.com> <98p3qp$rms2@www.netyaroze-europe.com> <98t840$6f5@www.netyaroze-europe.com> <99057g$3ok7@www.netyaroze-europe.com> <3AB6404F.65C9AF1C@harbinger.com> <9961fn$7jl4@www.netyaroze-europe.com> <99624q$7jl5@www.netyaroze-europe.com> <9963bb$7jl6@www.netyaroze-europe.com> <997j6l$7jl8@www.netyaroze-europe.com> <3AB774C7.6B4D8F95@harbinger.com> <998bjc$7jl11@www.netyaroze-europe.com> <3AB8A751.A5D36CCF@harbinger.com> <3AB8DFBA.7BDE2A4E@harbinger.com> <99b11c$cia1@www.netyaroze-europe.com> <99e5uf$mp51@www.netyaroze-europe.com> <99fb8j$mp52@www.netyaroze-europe.com> <99lnvq$roa1@www.netyaroze-europe.com> NNTP-Posting-Host: modem-250.helium.dialup.pol.co.uk X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2314.1300 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2314.1300 > - colour depth doesn't have much impact on performance (although it has > some) From my tests colour depth has a big effect. 8 bit is 25-40% faster than 16 bit. 4 bit is 50% quicker than 16 bit in some circumstances, but sometimes not quicker than 8 bit. Effective use of the texture cache makes a big difference too (you can double your rendering time with a poor use of texture caching). > - block/texture size is fairly critical to get max throughput with the GPU I ran your demos, and noted the hsyncs for a few of the same map positions with each one. I got about the same results for all the 16x16 demos, and about the same for all the 32x32 ones. This probably indicates that your program is CPU dominated, which wouldn't be surprising, because zoom-rotated sprites take a lot longer to sort than to draw. I don't think it's the GPU causing the difference between the 16x16 and 32x32 timings. Try timing the CPU and GPU separately (i.e. use one packet list at a time and do a VSync (0) after both the CPU code and the GPU code). (This will obviously make your program run (visually) half as fast...). > There's not really much scope for cheating either, I really need each block > to be zoomable/rotatable/rgb adjustable, so I can't turn those features off. > Mebbe sprites aren't really the way forward - would dynamic TMD make better > use of the hardware? I think it probably would be better (not that I've tried anything like that though). If you set up a bunch of polygons with your background textures on them you'd only need to set up the GTE once with your transformation matrix, rather than set up the zoom/rotate on a per-sprite basis (which I guess is what GsSortSprite does). Martin.