Feedback after mini-workshop on 2024.09.26

[ashwinvis]

Timing notes

• Opening note on ENCCS can take a few minutes, but I was unprepared.
• Installation took a good 15 minutes - unsure.
• Intro and perf. fundamentals episodes took a long time, but more on that later
• Start to end of profiling episode was around 1 hour 20 minutes.
• Optimization episode was done quickly without a demo and it still took 15 minutes.
• Parallelizing was skipped.

Possible improvements

• Spread theory part (intro and perf. fundamentals) into discussion points along at the end of hands-on episodes. This will allow us to get to the practical aspects quicker and appreciate the context of the theory.

Technical problems

• Move pip install instruction from footer to the main page along with conda - most learners had either Linux or macOS.
• Remove mpi4py - we are not using it and installing it with pip requires a working mpi library.

[ashwinvis]

• Python Tutor demo does not add any value, it can be removed
• Also in the code, change f.readlines() to f, since using readlines would create a list of the whole file and the latter is an iterator. Otherwise the discussion topic is confusing.

import io

def rms_from_text_file(f):
    """Compute root-mean-square of comma-separated values."""
    rms = 0
    n_samples = 0
    
    for line in f.readlines():
        fields = line.split(",")
        x, y, z = float(fields[0]), float(fields[1]), float(fields[2])
        # compute root-mean-square value
        rms += ((x**2 + y**2 + z**2) / 3) ** 0.5
        n_samples += 1
        
    return rms / n_samples


fake_file = io.StringIO("""\
0.27194615,0.85939776,0.76905204
0.51586611,0.59174447,0.06501842
0.23109192,0.8260391,0.08045166
""")

avg_rms = rms_from_text_file(fake_file)

[ashwinvis]

• Different levels of optimization, here https://enccs.github.io/python-perf/perf-fundamentals/#structured-approach-towards-optimization could be moved to the optimize section
• Word count problem is hard to optimize. Can we identify an easier problem, but still optimizable?

[ashwinvis]

• Output from %run -p magic is strange. It is best to avoid this and stick to python -m cProfile