68060 accelerator cards for Mac: Would you be willing? — #31

Post #31 by Bunsen

Source Forum	68kMLA
Category	68k
Post Date	Sat, 8 Sep 2007 - 19:23
Original URL	https://68kmla.org/bb/threads/68060-accelerator-cards-for-mac-would-you-be-willing.8459/

Post

68K instructions which are not supported on the Coldfire, should generate an exception, which can then be handled by executing the proper code on the Coldfire to emulate the excepted 68K instruction. Very neat. It's the same mechanism which the Mac Toolbox uses to execute its routines.

I wonder ... if this is a PDS card, could we redirect the excepted instructions back over to the original 680x0 for execution? Or build the card with a 68k co-processor onboard?

What is the performance like?
There is quite a significant overhead incurred when the ColdFire processor encounters an unimplemented instruction which has to be handled by CF68KLib. As a result, the effective performance which you will see will depend very much on whether your application triggers lots of exceptions within performance-critical loops. For best performance of production code, we recommend either recompiling for ColdFire (if the code was written in C or another high-level language), or translating using PortAsm/68K for ColdFire (if the code was written in assembly language).

In short, one wonders if a fast 68k copro would help ...

Does CF68KLib implement floating-point instructions?
No.

... and an FPU.

Another approach would be to design a 68030 chip into an FPGA. ...Beyond properly emulating each instruction, one must make sure that all the control registers, interrupt functionality and supervisory systems operate as expected. That's a lot of stuff.

As well, if we're going to use this an accelerator, it has to be able to interface to the Mac at its original CPU bus speed, and cache instructions and data between accesses.

Again:

You might also be interested in this PPCMLA discussion

and the follow-one, more technical discussion at Applefritter.