Closed Rodrigodd closed 7 months ago
Replaced many inouts with input or output: also not sure if it was needed.
The idea was to make it obvious from any module that inout is a bus. But if some modules drive it and some use it (input), I think it's no big deal. Whoever needs it will figure it out :)
As for everything else, it's fine. Thanks :)
As for the slow simulation - I am also very sad about it and was sad for a few days :) I think we need to choose another strategy for core verification, specifically to test each opcode individually with a small program. Maybe even cut blargg tests for this purpose. Waiting for days for results is not fun at all :)
I'll hold a little PR for aftershocks, I'll run it a little later today if nothing comes on top. Thank you!
This PR contains the following changes:
.bat
files with aMakefile
: the.bat
file never worked for me, due to the order of the files in the glob expansion, so I replaced them with a Makefile. Not sure if you like this change.Bogus_HW
.#0
delays: not sure if they were actually needed, let me know if they were.inout
s withinput
oroutput
: also not sure if it was needed.These changes were made to allow debugging
cpu_instrs
more easily, or in an attempt to make Verilator work. If you are not happy with any of the changes, please let me know.I managed to run
cpu_instrs
once, which resulted in an error code07
in the first test, and then printed the start message again, but I didn't investigate further. Thevvp
is running the simulation at the speed of around 8 real hours per simulated second. This isn't great, considering that the full test takes 60s to run; this would take 20 days to run.So I tried to make Verilator work, but without success (its emulation was running at 12min per emulated second, although completely wrong).
dmg-sim
, without this CPU, takes "only" 1h15min per simulated second, if I remember correctly. Maybe there is a way to speed up the simulation.