sim_ecore - an i960Sx emulator for microcontrollers to allow drop in replacement of i960Sx processors
This is an out cropping of all the work I have done in bringing up i960 based custom hardware. I have been successful in bringing up a real i960SB-10/SA-16 processor in a design I call type 1 with type2 on the way. However, the i960 processors themselves are a limited resource that will eventually run out. So originally sim3 was written to solve that little issue. However, it is very boring because it is a full emulation of an i960 design. So there is really no fun with connecting it up to physical memory chips or external devices. It also required that I maintain basically two copies of my chipset design which really killed momentum for me.
Recently, I have been experimenting with the EBI/EMI of AVR processors and found it to be exceptionally powerful.
I came up with an idea to map the address and data lines of a real i960 to the EBI of a proper chipset MCU to greatly improve performance. However, you still need a real i960 processor plus various support hardware. But what if we could just turn the chipset itself into the i960 processor?
Well with an EBI/EMI based microcontroller we can! The microcontroller core runs the simulator itself and accesses everything through the EBI/EMI (or just EBI from now on). Through bank switching we can even map the entire 32-bit memory space as well.
This idea also fits with something like the Feather-S2 with its onboard 8 Megs of PSRAM. All one would need to do is provide a large flash chip that is accessible over SPI and it could be easily done.
The only problems I can see with this design is lack of program flash to store the emulator. Thus I have forked sim3 so I can make the design as tight as possible. My initial target is an Arduino Mega 2560 (256k flash, 8k SRAM) so the bus will be 32-bit wide.
The only part that will be awkward is dealing with the EBI not mapping everything into RAM correctly. On SRAM, this isn't an issue but a FLASH chip will be much more awkward as it would require holes where things are not mapped... I'll have to come up with a solution to map everything without going crazy.
Regardless of the outcome, all of this will be very interesting to be sure :)