Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!usc!ucsd!ames!uhccux!munnari.oz.au!cluster!metro!natmlab.dap.csiro.au!ditsyda!macuni!mqccsunc!ifarqhar From: ifarqhar@mqccsunc.mqcc.mq.OZ (Ian Farquhar) Newsgroups: comp.sys.amiga Subject: Re: A new Amiga... Message-ID: <129@macuni.mqcc.mq.oz> Date: 19 Jan 90 12:20:49 GMT References: <4144@wehi.dn.mu.oz> <1051@sdrc.UUCP> <10302@microsoft.UUCP> Sender: news@macuni.mqcc.mq.oz Reply-To: ifarqhar@mqccsunc.mq.oz (Ian Farquhar) Organization: Macquarie University, Sydney Lines: 86 In article <10302@microsoft.UUCP> tom@microsoft.UUCP (Tom McConnell) writes: >In article <1051@sdrc.UUCP> evgabb@sdrc.UUCP (Rob Gabbard) writes: >>In article <4144@wehi.dn.mu.oz>, BAXTER_A@wehi.dn.mu.oz writes: >>> >>With the Amiga's custom graphics chips the CPU is freed up to do other things >>while in the PC/Mac world the CPU is all too often bogged down processing the >>graphics operations.That is why the 68000-based Amiga 1000 can still outperform >>a '486 VGA box on similiar graphics operations. I'm sure there are graphics >>co-processor boards that take alot of this burden off the Intel/Mac but compare >>the $$$$ to an Amiga. > >Ummm...Actually, how many '486 boxes have you seen? I've played around with >flight sim on a '486, and it _way_ outperforms the Amiga's ability to display >fast complex graphics. I mean we are talking about a '486 here! These things >are _incredibly_ fast! Even a '386 running at 20MZ is faster than an amiga >when running flight sim. But a '386 at 16mz, well, that's getting more in the >ballpark. Remove the floating point chip, and it's closer even more! You are absolutely correct. There is a lot of misunderstanding about the polygon filling capabilities of the Amiga. Under certain circumstances they can work well, but for a general 3D system they aren't too hot. The main problem arises when you get a polygon with a vertex that has a very small inner angle (say a very pointy isocelles triangle). The Amiga will actually fill it, but when the outline becomes one line, it will fill to the left to the line to the edge of the fill area! Most disturbing, needless to say. I have thought about this problem for a while, but have not produced any decent solutions. The closest I have gotten is to ensure that if such an angle is required, the top point (or wherever it is) has one of the lines offset by one pixel, thus ensuring that there is always a proper boundary for the fill. If anyone has a better solution, please do tell. If you want to do real fills, you can still use the blitter: you just use the line drawing capability. As usual, you scan convert the polygons and create a display list containing horizontal lines, which can be rendered by the blitter. This is an asynchronous operation, and line draw and scan conversion can proceed simultaneously. Your other statement about removing the FP chip is not quite correct. Only the most masochistic of programmers would ever write a 3D game (or indeed any system requiring real-time animation) using a FP chip, as they are rarely fast enough. It is far quicker to do fixed point integer multiplies than wait for the chip to finish. They also make dealing with rotations rather difficult, as they tend to force radians for trig functions when the average graphics programmer might want 65536 pseudo-degrees (map the top eight bits to the look-up table, and use the least significant byte to interpolate the extracted values). >I am a diehard Amiga fan, but let's not get carried away! Agreed. Stupid statements like the one above just make the Amiga look bad. It reminds me of the pretensious little brat who walked into a store near this university and announced that he wanted some IIGS software. The assitant - as they don't sell IIGS's - humored the kid who then proceded to compare his machine to the Amiga with th argument that his machine was better because it has a 65816 that ran at 500MHz! I am sure that you see my point. >I would like to see an amiga with a 32bit 68030 running at the same speed with >32bit video chips. Now _that_ would really be flying! :-) > Yes, it would. In those custom chips should be hardware rotation and scaling for blits, a real polygon filler, hardware rendering of bezier splines (cubic and triangular), and a 3 x 3 fixed point 64 bit matrix multiplier. Now that, kiddies, would be a mean 3D machine! Do I ask too much? I don't think so. All hail Saint Fubar, parton saint of computer programmers. +-----------------------------------+-------------------------------+ | Ian Farquhar | Phone : (02) 805-7420 (STD) | | Microcomputer Support | (612) 805-7420 (ISD) | | Office of Computing Services | Fax : (02) 805-7433 (STD) | | Macquarie University NSW 2109 | (612) 805-7433 (ISD) | | Australia | Also : 805-7205 | +-----------------------------------+-------------------------------+ | ACSNet ifarqhar@macuni.mqcc.mq.oz | | ifarqhar@mqccsuna.mqcc.mq.oz | +-------------------------------------------------------------------+ D