Compare with Cocotb

Coroutine

Cocotb wraps everything perfectly by its coroutine.

# Cocotb
# decorator
@coroutine
def Func():
    # yield an event
    # connected to Verilog by GPI
    yield clk_pos
# yield a coroutine
yield Func()

Nicotb, on the other hand, uses more Python features.

# Nicotb
def Func():
    # yield an event
    # In fact, events are just integers.
    # They can be registered from Verilog.
    # They can also be created in the testbench (maybe implicitly).
    yield clk_pos
# yield from a coroutine
yield from Func()

BTW, the yield from is introduced in Python 3.3. The Python 2 equivalent is

# Nicotb
def Func():
    yield clk_pos
for x in Func():
    yield x

. I use yield from just because it's cool.

Connect to Verilog

Compared with Cocotb, Nicotb use a Verilog file as the top module testbench. Therefore, Nicotb can be integerated to existing EDA tools easier and gives you better control over the Verilog code directly.

In Cocotb, the signals are accessed accordingly the hierarchy under the toplevel DUT.

# Cocotb
# read
print(int(dut.sig))
# write
dut.sub.nested_sig = 1
# write (immediate)
dut.sig.setimmediate(99)

In Nicotb, the signals are grouped as buses, and the grouping is defined by users. Buses are represented as Numpy arrays and the reading and writing should be called explicitly.

# Nicotb
bus = CreateBus((
    ("dut", "sig"),
    ("dut.sub", "nested_sig"),
));
# read (Verilog to Python)
bus.Read()
# read
print(bus.values[0][0]) # dut.sub
# write
bus.values[1][0] = 100 # dut.sub.nested_sig
numpy.copyto(bus.values[1][0], 100) # the same
# write (Python to Verilog)
bus.Write()
# write (Python to Verilog, immediate)
bus.Write(True)

Scoreboard

Nicotb has Scoreboard functionality, and can dump the results to a SQL DB. We also provide the Makefile to display the results. (TODO document)

High level verification

Currently Nicotb supports many common protocols such as vsync/hsync and AXI. (TODO document)