waveform_hdrgen.py: Two minors bugs (mode filter and temperature range) and one proposal to support non 5 bpp waveforms in JSON

[dapeda42]

I think there are two minor bugs in waveform_hdrgen.py:

1. Bug: Default mode_filter is wrong
If no export_modes are given, the default mode_filter is set in line 64 as follows:
mode_filter = list(range(len(waveforms["modes"])))

With 5 modes in waveform, mode_filter is 0,1,2,3,4
However, mode_filter should list the exctual modes given in the waveform.
For example, if the waveform contains five modes: 1, 2, 5, 16, and 17 (like ED097OC4.json, generated with epdiy_waveform_gen.py), then mode_filter should be 1,2,5,16,17

This can be achieved by changing line 64 as follows:
mode_filter = [wm["mode"] for wm in waveforms["modes"]]

2. Bug: Temperature interval is too long if temperature range is limited
For example, the waveform contains five temperature ranges: 10-15, 15-20, 20-25, 25-30, and 30-35
If argument '--temperature-range' is set to '16,26' (i.e., tmin=16 and tmax=26), then
a) Waveforms (EpdWaveformPhases) are exported to the header file for the three requested temperature ranges only (15-20, 20-25, 25-30). This is ok.

b) ALL temperature ranges are exported, i.e., EpdWaveformTempInterval contains ALL five temperature ranges (10-15, 15-20, 20-25, 25-30, 30-35). This is presumably NOT ok.

In waveform_hdrgen.py the temperature ranges are created in lines 74 and 75:

for bounds in waveforms["temperature_ranges"]["range_bounds"]:
    temp_intervals.append(f"{{ .min = {bounds['from']}, .max = {bounds['to']} }}")

This code MUST consider tmin and tmax (like in lines 87 and 88):

for bounds in waveforms["temperature_ranges"]["range_bounds"]:
    if bounds["from"] < tmin or bounds["from"] > tmax:
        continue
    temp_intervals.append(f"{{ .min = {bounds['from']}, .max = {bounds['to']} }}")

3. Proposal to support non 5 bits per pixel waveforms in JSON
The current version of phase_to_c (function in waveform_hdrgen.py) supports 5 bits per pixel (bpp) waveforms in the JSON file only and gives 4 bpp waveforms in the header file.
To relax these requirements (e.g., support JSON files with 4 bpp waveforms), I propose the following version of phase_to_c:

def phase_to_c(phase,bits_per_pixel_c=4):
    global total_size
    
    N1 = len(phase)
    N2 = len(phase[0])
    N = 2**bits_per_pixel_c
    
    if N1%N != 0:
        raise ValueError(f"First dimension of phases is {N1}. Allowed are multiples of {N}")
    
    step1 = int(N1/N)
        
    if N2%N != 0:
        raise ValueError(f"Second dimension of phases is {N2}. Allowed are multiples of {N}")
    
    step2 = int(N2/N)

    targets = []
    for t in range(0, N1, step1):
        chunk = 0
        line = []
        i = 0
        for f in range(0, N2, step2):
            fr = phase[t][f]
            chunk = (chunk << 2) | fr
            i += 1
            if i == 4:
                i = 0
                line.append(chunk)
                chunk = 0
        targets.append(line)
        total_size += len(line)

    return targets

[vroland]

I haven't looked at it in a while, but this seems plausible. Could you do a pull request with your changes?