John Sadler wrote: >> How? There's much more to it than a simple "move.l is called mov on >> Intel". > In principle the same way as a RISC compiler handles higher level > instructions. > Though it may be that modern Intel or AMD chips do not have enough > registers.
Internally they have anything from a dozen to over 100, depending on brand and model. Viewable from the programmer's point of view are only a hand full general purpose ones. And there's more to it. Like different flag handling, different endianess etc. > Somewhere I got the impression that Athlon and AMD and the Pentium > chips were RISC chips with a hardware layer which converted Intel > instructions to RISC instructions, but that may have changed if it > existed. No, that's still true. The Pentium Pro was the first one. > If so could one access the RISC instructions and registers directly? No. > Does that mean to say QPC is an emulator in principle? Hm? Can't follow you here. It emulates a processor core and provides interfaces for the hardware like SER, PAR. Note that I say "provides interfaces" and not "emulates". Apart from the screen layout no original hardware is emulated. > Does that mean that QPC runs a lot slower because of the emulation? Like most emulators QPC is an interpreter, i.e. much of the time is wasted in constantly decoding commands, the execution itself is pretty fast. So the answer is "yes". If it was a JIT compiler an equivalent of a several hundred Mhz 68k could be achieved on a decent PC, I guess. > Which should come first for QPC? > JIT > 6020+ & FPU emulation I would estimate the development of a JIT compiler to be several hundred hours. I had the time to do this a few years ago, but probably not the knowledge and skills. Plus back then I thought that the overhead of the JIT eats up all the savings. Now I definitely don't have the time and I also think that the demand for an even faster QPC is not high enough to justify the time spent. Now 68020+ command set is different, though testing and debugging can be quite time consuming it is a much more reasonable amount of work. Question again is, is it really worth it? > I suppose the answer depends on what do your customers use the QL for? > Probably tinkering and hence the answer neither. Yes, I guess you're right for the majority. Marcel
