RP2: Floats smaller than 1e-37 printed wrongly (and subnormals treated as 0.0)

In completing my function for showing the representation of floats for the case of subnormals:

def fp32_repr(x, bits=True):
    import struct
    a0 = int.from_bytes(struct.pack('f', x), 'little')
    if bits:
        return f'{(a0>>31)&1:01b} {((a0>>23) & 0xff):08b} {(a0 & 0x7fffff):023b}'
    else:
        sign = '-' if a0 >> 31 else ''
        exp = ((a0>>23) & 0xff)
        m = a0 & 0x7fffff
        if 255 > exp > 0:
            exp -= 150
            m |= 0x800000
        elif exp == 0:   # subnormal
            exp -= 149
        elif exp == 255: # Inf or NaN
            if m == 0:
                return f'{sign}Inf'
            else:
                return 'NaN'
        return f'{sign}{m} * 2**{exp}'

I noted that the RP2040 port (as of v1.22.0-preview.226.g0960d64d3 on 2023-12-06), after correction of the last float imprecision issue handles small floats differently than the other ports in that it treats subnormal representations as zero (0.0). This might be documented as a feature, not a bug:-), if correction seems too difficult.

A real bug is the printing of very small floats (< 1e-37, but still normal representation i.e. exponent bits > 0) as 1e-37.
This is demonstrated by the following code:

def irep2fp32(i):
    import struct
    return struct.unpack('f',i.to_bytes(4, 'little'))[0]

# With correct implementation of fp32: (including subnormal representation, where exp bits are 0)
# e.g. on Pyb 1.1 / Blackpill (stm32f411):
#
# int.from_bytes(struct.pack('f', 4.2e-41),'little')  -->    29972  subnormal
# int.from_bytes(struct.pack('f', 4.2e-39),'little')  -->  2997220  subnormal
# int.from_bytes(struct.pack('f', 3.2e-38),'little')  --> 19806589     normal
# int.from_bytes(struct.pack('f', 1.2e-37),'little')  --> 35870182     normal

print(irep2fp32(29972), irep2fp32(29972)*1000)
print(irep2fp32(2997220), irep2fp32(2997220)*1000)
print(irep2fp32(19806589), irep2fp32(19806589)*1000)
print(irep2fp32(35870182), irep2fp32(35870182)*1000)

print(3.2e-38)

On Blackpill and w600 boards (correct):

4.200112e-41 4.199972e-38
4.199999e-39 4.2e-36
3.2e-38 3.2e-35
1.2e-37 1.2e-34
3.2e-38

on esp8266 (30 bit):

4.20011e-41 4.19997e-38
4.2e-39 4.2e-36
3.2e-38 3.2e-35
1.2e-37 1.2e-34
3.2e-38

on Raspberry Pi Pico with RP2040 MicroPython v1.22.0-preview.226.g0960d64d3 on 2023-12-06

1e-37 0.0         <-- subnormal (4.2e-41), interpreted as 0.0 in calculation, but printed as 1e-37
1e-37 0.0         <-- subnormal (4.2e-39), interpreted as 0.0 in calculation, but printed as 1e-37
1e-37 3.2e-35     <-- first is incorrect, although not subnormal, probably only wrong print
1.2e-37 1.2e-34   <-- correct
1e-37                    <-- wrong, bug

Reproduce the supplied irep2fp32 and printing examples on the RP2040, comparing them with the Blackpill or w600 results. Use the fp32_repr normal and subnormal cases as the expected behavior, then locate the RP2040 float conversion or formatting entry point; done means very small normal floats print accurately and the subnormal behavior is resolved or documented.