Improve CompiledSimulation Iteration over Memory

[mdko]

The observed behavior of iterating over an inspected memory changes depending on whether it's a normal simulation (Simulation or FastSimulation) or a compiled simulation.

Observe the following program:

import pyrtl

m = pyrtl.MemBlock(32, 32, 'm')
addr = pyrtl.Input(32, 'addr')
out = pyrtl.Output(32, 'out')
w = m[addr]
out <<= w
m[addr] <<= pyrtl.as_wires(w + 10).truncate(32)

Normal Simulation

Given the above, if you simulate it via Simulation and then iterate over the inspected memory, it will only print the key-value pairs of memory addresses it touched during simulation (with the implication that all other memory locations are at their initial value, whatever that may be):

sim = pyrtl.Simulation()
sim.step_multiple({
    'addr': [0x0, 0x100, 0x102, 0x2345, 0x100, 0x102]
})
mv = sim.inspect_mem(m)
for addr, value in mv.items():
    print(f"{hex(addr)}: {value}")

Output:

> python ex1.py
0x0: 10
0x100: 20
0x102: 20
0x2345: 10

Compiled Simulation

sim = pyrtl.CompiledSimulation()
sim.step_multiple({
    'addr': [0x0, 0x100, 0x102, 0x2345, 0x100, 0x102]
})
mv = sim.inspect_mem(m)
for addr, value in mv.items():
    print(f"{hex(addr)}: {value}")

Output:

> python ex2.py
0x0: 10
0x1: 0
0x2: 0
0x3: 0
0x4: 0
0x5: 0
0x6: 0
0x7: 0
0x8: 0
0x9: 0
0xa: 0
0xb: 0
0xc: 0
0xd: 0
0xe: 0
0xf: 0
0x10: 0
0x11: 0
0x12: 0
0x13: 0
0x14: 0
....<etc. up to 2^32 address>

I think the first version makes much more sense, as it's easier to see through the noise. This means needing to update Compiled Simulation to be more judicious in what key-value pairs are returned during iteration. I've started adding another C function that can be called when the memory is being iterated, which will search for present elements in the hash map that currently represents memory in the C-made DLL.