On Fri, Jan 26, 2018 at 11:11:00PM +0000, STINNER Victor wrote:
> vstinner@apu$ ./python -m perf compare_to master.json pr5278.json 
> Mean +- std dev: [master] 1.86 us +- 0.03 us -> [pr5278] 2.27 us +- 0.04 us: 1.22x slower (+22%)
> 
> Note: master is the commit 29a7df78277447cf6b898dfa0b1b42f8da7abc0c and I rebased PR 5278 on top on this commit.

Thank you and Elvis for running the benchmarks. Yes, the exact version does seem
important -- I have been getting some differences based on the checkout.

> This is obvious the *worst* case: a *micro* benchmark using local contexts and modifying this local context. In this case, I understand that this microbenchmark basically measures the overhead of contextvars on modying a context.
> 
> The question here is if the bottleneck of applications using decimal is the code modifying the context or the code computing numbers (a+b, a*b, a/b, etc.).

Yes, that's the big question. In the generator discussions people were advised
to use "with" whenever possible, so I assume short blocks *will* be used.

I would use the context functions, which would not require PEP-567 at all.
This means that I'm somewhat okay with excessive with-statements being a
bit slower.

> vstinner@apu$ ./python -m perf compare_to telco_master.json telco_pr5278.json -v
> Mean +- std dev: [telco_master] 10.7 ms +- 0.4 ms -> [telco_pr5278] 10.7 ms +- 0.4 ms: 1.00x faster (-0%)
> Not significant!

Okay. In my above reference to telco, I ran the "telco.py full" command
from http://www.bytereef.org/mpdecimal/quickstart.html#telco-benchmark .

The numbers I posted weren't cooked, but I have a hard time reproducing
them myself now consistently with the latest revisions, so let's declare
telco.py and bench.py a tie.

This means that I no longer have any objections, so Yury, please go ahead
and merge the PR!

Stefan Krah