PEP 817: Final set of pre-split updates

[mgorny]

• Change is either:
  • To a Draft PEP
  • To an Accepted or Final PEP, with Steering Council approval
  • To fix an editorial issue (markup, typo, link, header, etc)
• PR title prefixed with PEP number (e.g. PEP 123: Summary of changes)

This is a set of updates to PEP 817 that were prepared prior to the decision to split it into smaller PEPs. We do not intend to pursue PEP 817 in this updated form. However, we would like to get the changes committed to the repository so that the history of changes is not lost.

The changes include refocusing of the security details, some clarifications, tool behavior outlines and update for the images to respect dark theme.

We are not marking the PEP as withdrawn, as our current plan is to eventually replace the specification part with an index listing the split PEPs.

---

📚 Documentation preview 📚: https://pep-previews--4867.org.readthedocs.build/

[rgommers]

This was reviewed in detail by multiple co-authors before opening this PR. It contains important clarifications, as well as the diagram that multiple reviewers found really helpful.

PEP editors, can this please be merged?

[hugovk]

This shows up in a blockquote:

Unindent the block to fix:

.. figure:: pep-0817/avx512_gromacs_benchmark.svg
    :alt: A bar graph comparing GROMACS performance (in ns/day) with
          various targets. The first two bars are labeled "yum
          (2018.8)" and "generic (SSE2)", reach about 1.0 ns/day and
          are both marked as "SSE2". The next bar is labeled
          "ivybridge" ("AVX") and reaches almost 1.5 ns/day. Two
          following bars are labeled "haswell" and "broadwell" (both
          "AVX2") and exceed 1.5 ns/day slightly. The last two bars
          are labeled "skylake_avx512" and "cascadelake" (both
          "AVX512") and reach almost 2.0 ns/day.

    Performance of GROMACS 2020.1 built for different generations of
    CPUs. Vertical axis shows performance expressed in ns/day, a
    GROMACS-specific measure of simulation speed (higher is better).

Compiling GROMACS_ for architectures that can exploit the AVX-512
instructions supported by the Intel Cascade Lake microarchitecture
gives an additional 18% performance improvement relative to using
AVX2 instructions, with a speedup of about 70% compared to a generic
GROMACS installation with only SSE2.

— `archspec: A library for detecting, labeling, and reasoning about
microarchitectures
<https://tgamblin.github.io/pubs/archspec-canopie-hpc-2020.pdf>`__

[rgommers]

I'll note that that's how it came out of the last review, rather than being introduced here: https://peps.python.org/pep-0817/#runtime-cpu-dispatching.

Might be improve-able indeed though, worth trying.

[mgorny]

It is a blockquote.

[hugovk]

Ah right, then I suggest just moving the source ("—archspec: A library for detecting, labeling, and reasoning about microarchitectures") bit outside the blockquote.

[hugovk]

Please add alt text.

[rgommers]

Thanks for the quick response @hugovk. We'll update.

[hugovk]

Unindent to remove blockquote:

Suggested change

	.. figure:: pep-0817/conceptual_diagram_installers.svg
	:target: _images/conceptual_diagram_installers.svg
	:alt: TODO.
	A conceptual diagram of installing a wheel.
	.. figure:: pep-0817/conceptual_diagram_installers.svg
	:target: _images/conceptual_diagram_installers.svg
	:alt: TODO.
	A conceptual diagram of installing a wheel.

[hugovk]

This looks okay in dark mode:

But the lines and a lot of text are invisible or hard to read in light:

[rgommers]

@mgorny I thought we fixed the light/dark mode behavior?