Article 71192 of comp.sys.cbm: Path: news.acns.nwu.edu!newsfeed.acns.nwu.edu!news.ece.nwu.edu!news.cse.psu.edu!uwm.edu!vixen.cso.uiuc.edu!howland.erols.net!europa.clark.net!news-feed.inet.tele.dk!uninett.no!news.uni-c.dk!diku.dk!alstrup From: alstrup@diku.dk (Asger K. Alstrup Nielsen) Newsgroups: comp.sys.cbm,comp.sys.sinclair Subject: Re: Elite line routine (was Re: spectrum faster in 3D) Date: 9 Jul 1997 12:43:23 GMT Organization: Department of Computer Science, U of Copenhagen Lines: 49 Sender: alstrup@tyr.diku.dk Message-ID: <5q011b$m0p@vidar.diku.dk> References: <33A9DD07.153E@skynet.be> <5pmm0e$1cm2@ds2.acs.ucalgary.ca> <5prb65$6i1@news.acns.nwu.edu> <33C21416.41C6@wi.leidenuniv.nl> <5pu309$96o@news.acns.nwu.edu> NNTP-Posting-Host: tyr.diku.dk X-Newsreader: NN version 6.5.0 #13 Xref: news.acns.nwu.edu comp.sys.cbm:71192 comp.sys.sinclair:41085 judd@merle.acns.nwu.edu (Stephen Judd) writes: >The weirdest line idea I ever heard was to use an interrupt to step >the pointer in the y-direction. >Anyways, I look forwards to your routines! It's not ridiculous to use the SID as line-plotter. In effect, the frequency modulator works with the Bresenham algorithm, and the triangle waveform is well suited for up to 8-bit precision lines. The problem is to map the slope to a frequency, but it should be workable. You'll get the best speed-up with a character-map, because the address-calculation is constant time with the right setup (i.e. just an .y index). So, in effect the line routine would be: set-up SID ldy $d41b ; or is it $d41c lda #$80 ora charmap,y sta charmap,y ldy $d41b lda #$40 ora charmap,y sta charmap,y ldy $d41b lda #$20 ora charmap,y sta charmap,y ... ldy $d41b lda #$01 ora charmap,y sta charmap,y ldy $d41b lda #$80 ora charmap+$80,y sta charmap+$80,y ... which works with a standard 16x16 character layout. This method takes 4+2+5(?)+5 = 16 cycles per pixel, regardless of slope. Greets, Asger Alstrup Article 71223 of comp.sys.cbm: Path: news.acns.nwu.edu!merle!judd From: judd@merle.acns.nwu.edu (Stephen Judd) Newsgroups: comp.sys.cbm,comp.sys.sinclair Subject: Re: Elite line routine (was Re: spectrum faster in 3D) Date: 10 Jul 1997 02:27:13 GMT Organization: Northwestern University, Evanston, IL Lines: 22 Message-ID: <5q1ha1$spl@news.acns.nwu.edu> References: <33A9DD07.153E@skynet.be> <33C21416.41C6@wi.leidenuniv.nl> <5pu309$96o@news.acns.nwu.edu> <5q011b$m0p@vidar.diku.dk> Reply-To: sjudd@nwu.edu (Stephen Judd) NNTP-Posting-Host: merle.acns.nwu.edu Xref: news.acns.nwu.edu comp.sys.cbm:71223 comp.sys.sinclair:41120 In article <5q011b$m0p@vidar.diku.dk>, Asger K. Alstrup Nielsen wrote: >judd@merle.acns.nwu.edu (Stephen Judd) writes: > >>The weirdest line idea I ever heard was to use an interrupt to step >>the pointer in the y-direction. > >It's not ridiculous to use the SID as line-plotter. In effect, the frequency An intriguing idea. >So, in effect the line routine would be: > > set-up SID How does one initialize the voice 3 waveform register? -S >Asger Alstrup Article 71334 of comp.sys.cbm: From: =?ISO-8859-1?Q?Marko_M=E4kel=E4?= Newsgroups: comp.sys.cbm Subject: Re: Elite line routine (was Re: spectrum faster in 3D) Date: 11 Jul 1997 00:28:33 +0300 Organization: Helsinki University of Technology, CS lab Distribution: world Message-ID: <87u3i2vbpa.fsf@sranje.tky.hut.fi> References: <33A9DD07.153E@skynet.be> <33C21416.41C6@wi.leidenuniv.nl> <5pu309$96o@news.acns.nwu.edu> <5q011b$m0p@vidar.diku.dk> <5q1ha1$spl@news.acns.nwu.edu> NNTP-Posting-Host: sranje.tky.hut.fi Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Original-Sender: postmaster@[127.0.0.1] X-Newsreader: Gnus v5.3/Emacs 19.34 Path: news.acns.nwu.edu!newsfeed.acns.nwu.edu!news.luc.edu!chi-news.cic.net!newsrelay.netins.net!mr.net!europa.clark.net!dispatch.news.demon.net!demon!bullseye.news.demon.net!demon!news2.euro.net!news.euro.net!news3.euro.net!news.bru.tfi.be!nntp.inet.fi!news.funet.fi!news.cs.hut.fi!not-for-mail Lines: 19 >>>>> "Steve" == Stephen Judd writes: Steve> How does one initialize the voice 3 waveform register? Hmm, I think that I have an example of that somewhere (how to write a desired value to the OSC3 register by starting the sawtooth waveform and turning the frequency to 0 when the correct value is reached), but I can't find it. Anyway, you have to set the waveform register of voice 3 to $20 (or $21, the gate bit shouldn't matter). That's it! You don't need to set the ADSR or anything, since you don't need (or want) to hear any sound. Anyway, I think that this idea that Asger presented is very cool. Was it actually used in any demo or game? Marko -- Marko Mäkelä < Marko.Makela @ HUT.FI >, http://www.hut.fi/~msmakela/ Helsinki University of Technology, dept. of Computer Science Article 71369 of comp.sys.cbm: Path: news.acns.nwu.edu!merle!judd From: judd@merle.acns.nwu.edu (Stephen Judd) Newsgroups: comp.sys.cbm Subject: Re: Elite line routine (was Re: spectrum faster in 3D) Date: 12 Jul 1997 21:02:31 GMT Organization: Northwestern University, Evanston, IL Lines: 30 Message-ID: <5q8rd7$k01@news.acns.nwu.edu> References: <33A9DD07.153E@skynet.be> <5q011b$m0p@vidar.diku.dk> <5q1ha1$spl@news.acns.nwu.edu> <87u3i2vbpa.fsf@sranje.tky.hut.fi> Reply-To: sjudd@nwu.edu (Stephen Judd) NNTP-Posting-Host: merle.acns.nwu.edu In article <87u3i2vbpa.fsf@sranje.tky.hut.fi>, =?ISO-8859-1?Q?Marko_M=E4kel=E4?= wrote: >>>>>> "Steve" == Stephen Judd writes: > >Steve> How does one initialize the voice 3 waveform register? > >Hmm, I think that I have an example of that somewhere (how to write a >desired value to the OSC3 register by starting the sawtooth waveform >and turning the frequency to 0 when the correct value is reached), but That's what I was thinking -- busy-loop wait until the correct value appears, and pray. Any idea on how long one would have to wait? Also, is it at all clear that any line slope can be computed in this way? That is, if each "loop iteration" (LDA #$40 STA blah,y) takes six cycles, doesn't that limit the slopes to some factor of six? Or, to put it another way, it seems to me that each cycle the waveform counter increases by a certain amount, x c=c+x So after every six cycles we've increased by 6x, so a very specific family of sequences is generated. Also, clearly a different method is needed for lines which move in the y-direction (i.e. slopes >1). -S Article 71370 of comp.sys.cbm: Path: news.acns.nwu.edu!merle!judd From: judd@merle.acns.nwu.edu (Stephen Judd) Newsgroups: comp.sys.cbm,comp.sys.sinclair Subject: Re: Line routine in hardware Date: 12 Jul 1997 21:18:56 GMT Organization: Northwestern University, Evanston, IL Lines: 45 Message-ID: <5q8sc0$kc8@news.acns.nwu.edu> References: <33A9DD07.153E@skynet.be> <5q011b$m0p@vidar.diku.dk> <5q1ha1$spl@news.acns.nwu.edu> <33C6849F.4D9A@chebucto.ns.ca> Reply-To: sjudd@nwu.edu (Stephen Judd) NNTP-Posting-Host: merle.acns.nwu.edu Xref: news.acns.nwu.edu comp.sys.cbm:71370 comp.sys.sinclair:41295 In article <33C6849F.4D9A@chebucto.ns.ca>, George Taylor wrote: >> >judd@merle.acns.nwu.edu (Stephen Judd) writes: >> > >> >>The weirdest line idea I ever heard was to use an interrupt to step >> >>the pointer in the y-direction. >> > >> >It's not ridiculous to use the SID as line-plotter. In effect, the frequency >(Asger alstrup comments:) >> An intriguing idea. >> >> >So, in effect the line routine would be: >> > >> > set-up SID >> >> How does one initialize the voice 3 waveform register? >> >> -S >> >> >Asger Alstrup >Stephen Judd is referring to an idea I have mentioned to him, >about using hardware tricks to draw lines. Quite right. But, I have a simple policy concerning ideas: until I see an implementation, I assume they don't work. :) >And finally, in terms of elite line routines, take a look Actually the title is "Elite line routine", i.e. originally it was the routine from "Elite", not 3l1+3 lines -- just wanted to get that straight :). >at the last traktor III. In the vector part, there is a line >routine which takes 21/24 cycles. There is a big unrolled loop, Does it handle all lines or only lines of particular slopes, and how much memory does it use? BTW, I just finished up my new polygon renderer, and man oh maneschewitz is it cool (and only 5k, too) -- it might even be faster now than the polygonamy renderer. It draws to the bitmap and can render full-screen, off-screen, ice-cream... sssssssmokin'! -S Article 71418 of comp.sys.cbm: Path: news.acns.nwu.edu!newsfeed.acns.nwu.edu!news.ece.nwu.edu!news.cse.psu.edu!rutgers!news.sgi.com!newsfeed.nacamar.de!news.apfel.de!fu-berlin.de!cs.tu-berlin.de!uni-duisburg.de!RRZ.Uni-Koeln.DE!news.rhrz.uni-bonn.de!usenet From: klee@sunrise.informatik.uni-bonn.de (Oliver Klee) Newsgroups: comp.sys.cbm,comp.sys.sinclair Subject: Re: Line routine in hardware Date: Sun, 13 Jul 1997 19:22:40 GMT Organization: RHRZ - University of Bonn (Germany) Lines: 25 Message-ID: <33c921c5.5900464@news.informatik.uni-bonn.de> References: <33A9DD07.153E@skynet.be> <33C21416.41C6@wi.leidenuniv.nl> <5pu309$96o@news.acns.nwu.edu> <5q011b$m0p@vidar.diku.dk> <5q1ha1$spl@news.acns.nwu.edu> <33C6849F.4D9A@chebucto.ns.ca> NNTP-Posting-Host: rhrz-isdn2-p24.rhrz.uni-bonn.de X-Newsreader: Forte Free Agent 1.1/32.230 Xref: news.acns.nwu.edu comp.sys.cbm:71418 comp.sys.sinclair:41351 On Fri, 11 Jul 1997 16:08:15 -0300, George Taylor wrote: >I also thought of using the sid waveform registers to do lines, >once you set the correct frequency, which is synchronized >to the line routine, simply lda waveform3 and no need to >increment a register. However, this isn't useful either, >because there are better routines. Another problem would be the following: a) You need a very exact timing for the routine, and it shouldn't get interrupted. b) The VIC takes 40 cycles at each char line when it's on. c) To have the drawn lines visible, the VIC has to be turned on. So this routine only works outside the screen or between the memory accesses by the VIC. . Oliver Klee / Master's Design Group .. musician, journalist & programmer ... klee"AT"sunrise.informatik.uni-bonn.de .... ^^ (replace with "@")