Short: PoV-Ray2.2, '040 optimized with gcc2.6.1 Uploader: hobold informatik uni-bremen de (Holger Bettag) Type: gfx/3d Architecture: m68k-amigaos Authors: The PoV-Ray team. Nearly no modifications by uploader. Description: ------------ Did you ever wonder, where the horsepower of your 68040 was hiding from you? Well, I found out: It hides in the gnu c compiler! Recently I compiled PoV-Ray for the Sparcs here at the University of Bremen. I used gcc, naturally, and I wanted gcc to deliver good work. So I read quite a few manpages and stumbled over a fat load of options for optimization. At the end of this section there were also processor-specific options. I saw the magic number: 68040. I realized that I (not really able to program in C) NEEDED gcc! After some hours of ftp'ing to good ol' AMINET, I had both gcc 2.6.1 and the source of PoV-Ray 2.2, and I even managed to install gcc at first attempt that evening. Only a few nights of constant compiling later, I had an executable of povray that lacked nifty display features and even had a disgusting flaw (see section Bugs!), BUT: IT RENDERED CHESS.POV 44% FASTER THAN THE OFFICIAL 020+881 EXECUTABLE!!! So here I am now, thinking about all those trace-o-holics wasting CPU-cycles without even knowing it, and I can't resist uploading the executable. Despite all the Bugs that might be in there (see section Bugs!), despite the need for ixemul.library (see section Requirements), despite the small number of actual 68040 Amigas... HERE IT IS: THE HORSEPOWER OF YOUR 68040! For some more information on compiling, read "compiling". Requirements: ------------- You will need ixemul.library version 40 or above. I think ixemul.library works only on 2.04 Kickstarts and higher versions. You can find it on AMINET in "dev/gcc/", including it here would be a waste of AMINET diskspace... "My" povray has not really been tested so far, but runs nicely (except the Bugs, see there) on an Amiga 4000/040 with: - Kickstart and Workbench 3.1 - 12 MByte of Static-Column FastRAM; RAM access is about 20-30% faster than with Page-Mode RAMs, but raytracing is typically not sped up. (shameless plug: ask Omega Datentechnik, Oldenburg, Germany for more info, they sell it) - Original Commo CPU-Board with 68040 @ 25 MHz Bugs: ----- There is but one disgusting flaw in "my" povray: the second scanline of the picture is not output to disk unless you specify an I/O buffer with the "+Bnn" option (either on the commandline, in "povray.def" or in the POVRAYOPT environment variable). This behaviour is not affected by - what scene is rendered - the resolution it is rendered at - the output file format - the compiler options Sorry, I cannot fix this bug. I loaded tons of "printf"s into the source, only to find out that the proper functions are called at the right time with correct arguments passed... I have definitely no clue. I did not encounter other bugs so far. Restrictions: ------------- You already know that you have to use "+B" for every trace (see section Bugs). Now let me tell you that you also have to use "-d" for every trace. No preview display, sorry. The machine-specific "amiga.c" was too much for gcc. As I am not really familiar with C, I will not be able to fix this. There is one more disadvantage that is related to the Bug (see there): all versions of povray I checked, have trouble continuing traces that were done with an I/O buffer. Maybe a tool that removed a partial last line from the output file could fix this. Acknowledgements: ----------------- I would like to thank the PoV Team for their great work. I am especially glad that they decided to provide source code. Who knows, maybe I'll contribute some patches, too, now that I can compile povray :-) I would like to express my respect for the talent and attitude of the people involved in the gnu project. I also want to thank the people that managed to port gcc to the Amiga. Whatever happens to Commodore, with people like you out there, my Amiga will never die! The Uploader: ------------- I am a mere math student that likes to play around with povray on a Sparc 20... If you want to contact me, feel free to mail to the address mentioned at the top of this readme, or if you don't have access to the net, send your mail to Holger Bettag Kilianstrasse 44 67373 Dudenhofen Germany Happy tracing!