Is there a "soft core" m68k for FPGAs?
Haveing the cpu in vhdl/verilog would allow to eliminate the need for a special Minimig board. And being able to use plain developer boards from Xilinx or other manufactors.
Have you seen the discussion at Open Cores?
The most finished 68000 in VHDL I know of is from the Atari ST in VHDL: Suska project : 68000 IP v2K7A
- Would it make any sense to replace the "FPGA chip + CPU chip" on the current Minimig, with a single-chip "FPGA + CPU chip" device? --DavidCary 21:28, 28 July 2007 (PDT)
TobiFlex 070826 13:39
This CPU Core is not finish. The Debuging must go on. Currently hangs up this CPU Core with Kickstartrom 2.04 at #$F81914.
TobiFlex 070826 14:06
I have adapted the minimig Core to the DE2 Board from Terasic with a ALTERA Cyclone 2C35. With an external CPU MC68HC000 runs the Minimig perfekt. I have change the RAM Timing from SRAM to SDRAM - so i can use the on Board SDRAM Chip. And now i use the Board to Debug Wolfgangs 68K Core.
mongo 070826 20:25
Wolfgang's 68K core will run up to about 17 MHz in a Spartan 3.
jkonstan 070826 21:23
external CBT16245 level shifters
- TobiFlex releases his TG68 MC68000 project source code in VHDL. With a size of 3000 lines.
- "The Kernal is 2x-3x faster then the 68000 with the same clock and you can clock them with 28MHz.
I think it is possible to make the minimig 8x-12x faster then the original."
- Is there any attempt to make the instruction timings the same as the original 68k?
- Most of the byte and word instructions are exact. Long and mulu/divu/rotation instructions are faster.
- The MC68EC000 is different from the 68000. I have my Core let run against a real MC68HC000 and can not :find software differences when the Kickstart or some programms are running. So I know all importand :Instructions works correct. About never used Instructions I can not say anything.
- Example: I have found no program for the AMIGA that use CHK or MOVEP Instructions.
- About the doc on the second link. My Core use no prefetch. My Core use classical Fetch/Execute last - :Decode - Fetch next/Execute - order.
- Thats why most Instrucions can execute in 2 clocks. The Bus wrapper expand this to 4 clocks.
- But for the minimig is the prefetch not so important. The core is fast enough to decode the opcode between as=high. So the timing is the same as the 68000 timing.
- On my DE2 Board with a real MC68EC000 and running ANARCHY1 DEMO I can see some rendering error but with the softcore I can not.
- I dont know what happens but I think my core has a high Betatested Level. (Upss I hope you can understood my - sorry for my poor english)
- OPCODE 0x0F0E MOVEP.W ..... My Validator stop here.
- There are some subtle undocumented 68k behaviour that you might be interested in:
- Testdisc known to be sensitive to 68000 incompatibility: http://amiga.nvg.org/warlock/adf/b/Birk+Sundell/100C64Tunes.adf.gz
- Quick compile using webpack 6.3.03i:
- Device utilization summary:
- Selected Device : 3s400pq208-4
- Number of Slices: 2627 out of 3584 73%
- Number of Slice Flip Flops: 1052 out of 7168 14%
- Number of 4 input LUTs: 4841 out of 7168 67%
- Number of bonded IOBs: 74 out of 141 52%
- Number of GCLKs: 1 out of 8 12%
- TIMING REPORT
- NOTE: THESE TIMING NUMBERS ARE ONLY A SYNTHESIS ESTIMATE. FOR ACCURATE TIMING INFORMATION PLEASE REFER TO THE TRACE REPORT GENERATED AFTER PLACE-and-ROUTE.
- Timing Summary:
- Speed Grade: -4
- Minimum period: 39.906ns (Maximum Frequency: 25.059MHz)
- Minimum input arrival time before clock: 16.900ns
- Maximum output required time after clock: 16.576ns
- Maximum combinational path delay: No path found
- I need a bigger FPGA :-)
- Days of Thunder uses the CHK instruction: http://eab.abime.net/showthread.php?t=1144