Short: Compare your Amiga to a Vivid 24! Author: RTillery@crash.cts.com Uploader: RTillery crash cts com Architecture: m68k-amigaos /************************************************************************/ /* */ /* Vivid 24 to Amiga speed comparison - ©1993 Digital Micronics, Inc. */ /* */ /************************************************************************/ The accompanying executable will allow you to compare the speed of your Commodore Amiga computer to that of our Vivid 24 rendering engine when rendering a Mandelbrot. If you do not own a Vivid 24, you can still run the program to check the speed of your Amiga and the program will compare it's speed to the three co-processor-equipped versions of the Vivid 24. To use the program with your Vivid 24, you must put the file "EXTSRAM.RLM" into "SAGE:Exec", place "extSRAM.library" into "SAGE:Libs". Then put the "Mandel.xxxx" files along with the "MANDEL.ABS" file in a directory some- where together. You can then run any of the "Mandel.xxxx" files to see a comparison of your Amiga to your Vivid. To use the program without a Vivid 24, simply run any of the "Mandel.xxxx" programs to see a comparison of your Amiga to the Vivid 24. All test results used in the non-Vivid comparison are based on a 40Mhz Vivid 24. For those of you who are interested (and those who think this may be all smoke and mirrors) I've included the actual source code used to generate the Mandelbrots. The file called "AmigaMandel.c" contains the main routines used to set up the Vivid if it's there, time a rendering, set up and render on the Amiga and time it's rendering. You will note by looking at the "MakeFile" that the code was optimized for speed and each version was customized for the 68882, IEEE, or FFP as denoted by its file extension. All executables were compiled with SAS/C 6.3. The file "VividMANDEL.C" is the source for the TMS34082 math coprocessors. Each of the available coprocessors is loaded with this code. Additional code on the TMS34020 graphics processor is used to control these coprocessors. Each processor is given one line at a time to render. When finished, the coprocessor signals the 34020 and the data is copied from the 34082 memory (SRAM) to the screen. Then the 34082 is given another line to do if any remain. The 34020 then waits for a signal from another coprocessor. This technique is the simplest that will allow the same code to run on one, two or four math coprocessors. For a fair comparison, you'll notice the two source files are nearly identical. Although this is not the fastest algorithm for generating Mandelbrots, it does allow a RELATIVE comparison of speeds. As one last (marketing hype:) comment, remember that a speed of 30 times your Amiga means that one Vivid can replace 30 Amigas! One Amiga 4000 costs about $2500. 30 would cost $75,000. One Amiga and a Vivid 24 stocked with 4 coprocessors would cost you about $8000 or nearly one tenth the cost for the same rendering power. For more information on the Vivid 24, please contact Digital Micronics at (619) 931-8554. Rick Tillery (Digital Micronics, Inc.)