mlelstv@immd4.informatik.uni-erlangen.de (Michael van Elst ) (05/17/89)
bob@jacobs.CS.ORST.EDU (robert s. richardson) writes: >... Is it conceivable to utilize the >blitter as a math coprocessor with the proper code? Or is the >68000 faster in such areas? I know some simple binary operations >could probably be implemented, but what about floating point? I checked some simple (fixpoint) operations: In fact, the blitter may ADD and MULTIPLY faster than an 68000 if you choose a different number representation to avoid multiple carry operations AND use large vectors (about >500 numbers). BUT it's slower than an 68020 running at higher speeds !! But there are some operations not intended for graphics, that can make use of the blitter: -- There is a Game-of-Life program that produces about 10 generations per second on a 640x200 frame. -- A friend of mine rewrote the Lee-Algorithm (used for PCB routing) to work on the blitter. It's 14 times faster than a commercial router program. Michael van Elst E-mail: UUCP: ...uunet!unido!fauern!immd4!mlelstv
golden@cps3xx.UUCP (golden james) (05/18/89)
In article <261@medusa.informatik.uni-erlangen.de> mlelstv@immd4.informatik.uni-erlangen.de (Michael van Elst ) writes: >bob@jacobs.CS.ORST.EDU (robert s. richardson) writes: > >>... Is it conceivable to utilize the >>blitter as a math coprocessor with the proper code? Or is the > >I checked some simple (fixpoint) operations: >In fact, the blitter may ADD and MULTIPLY faster than an 68000 >if you choose a different number representation to avoid multiple What about using the blitter to implement a neural network, since they usually suffer from hundreds of simple integer calculations? Could you simply "blit" the network recursively to obtain a result? Mike Golden Physiology Undergraduate Michigan State University