Thomas Harte
56111c75ae
Makes first efforts towards disassembly.
2021-01-26 19:52:30 -05:00
Thomas Harte
8789ffda15
Corrects performer storage, RMW/W confusion, implicit casts, port readback.
2021-01-24 22:30:42 -05:00
Thomas Harte
6e9ce50569
Corrects duration-based iteration.
2021-01-21 23:05:43 -05:00
Thomas Harte
adef2e9b4e
Starts formalising end conditions.
2021-01-21 22:36:44 -05:00
Thomas Harte
0fafbf5092
Completes M50740 instruction set.
2021-01-21 19:08:38 -05:00
Thomas Harte
e58608b25a
Gets as far as executing a first loop.
2021-01-20 18:15:24 -05:00
Thomas Harte
c78c121159
Succeeds at executing a single instruction.
2021-01-18 20:16:01 -05:00
Thomas Harte
a601ac0cab
Corrects performer population, lookup, calls.
2021-01-18 17:53:14 -05:00
Thomas Harte
9b92753e0a
In theory this should 'execute' up to the first unconditional branch.
...
Where execution means: do very little.
2021-01-18 17:11:11 -05:00
Thomas Harte
8b19c523cf
Starts to bend towards getting some performers in motion.
2021-01-18 16:45:52 -05:00
Thomas Harte
5ace61f9b9
Continues walking very slowly towards cached execution.
2021-01-18 11:20:45 -05:00
Thomas Harte
ad03858c6e
Switches performers to member functions. Very slightly starts work on M50740 performers.
2021-01-17 20:53:11 -05:00
Thomas Harte
54b26c7991
Bends to using 8-bit lookups for M50740 instructions.
2021-01-17 20:03:36 -05:00
Thomas Harte
17c3a3eb4b
Seeks to switch to maintaining a bank of performers.
...
My thinking here is that for really simple processors there'll be 256 or less, meaning that they can be stored by simple uint8_t; for every other processor I can currently think of it'll likely be uint16_t.
Either way, that's a much better outcome than using plain pointers, which on architectures I currently build for will always be 8 bytes. For the simple processors I can get eight times as much into the cache; for the others four times.
2021-01-17 19:38:23 -05:00
Thomas Harte
8860d0ff51
Starts to establish the CachingExecutor.
2021-01-16 22:06:16 -05:00
