Compare benchmarks for pull requests

[akaIDIOT]

No description provided.

[akaIDIOT]

Beginning to wonder whether running this on GitHub actions is even useful; preliminary results between this PR and its target, which currently changes only workflow configuration:

table of information now posted from github actions, see comment(s) below

benchmark	3.10	3.11	3.12	3.13	3.14	PyPy 3.10	PyPy 3.11
get	+1%	+0%	-1%	-10%	-3%	+29%	-2%
get_dotted	+3%	+1%	+1%	+1%	+2%	+29%	-1%
getattr	+0%	+1%	+5%	-12%	-0%	+21%	-3%
getitem	+0%	-1%	-1%	-9%	-2%	+16%	-2%
getitem_dotted	+3%	+1%	+0%	+1%	+1%	+26%	-2%
init_full_overlap	+1%	+2%	+4%	-2%	+1%	-2%	+0%
init_no_overlap	+1%	+1%	+4%	-1%	+2%	+2%	+1%
init_partial_overlap	+1%	+2%	+3%	-3%	-0%	-1%	-1%
reference_chain	+1%	-0%	+2%	-0%	-0%	+17%	+1%
unpack_args	+0%	+0%	+2%	-1%	+0%	+62%	-1%
unpack_kwargs	+2%	+0%	+1%	+0%	+2%	+59%	+1%
unpack_tuple	+2%	-2%	+1%	-1%	+1%	+71%	+0%

This is currently comparing the median stat for each benchmark, ops shows a similar issue: these readings should be within a margin of error. While it makes sense that a GitHub action runner has more things to do, this makes it hard to indicate whether a pull request would change the performance characteristics of the things being measured...

[github-actions]

Comparing median (lower is better) metric of benchmarks between this PR's target (e12dd7a) and the HEAD of this PR (f82a3f8):

	3.10	3.11	3.12	3.13	3.14	PyPy 3.10	PyPy 3.11
test_benchmark_get	-3%	-0%	+2%	-0%	+3%	-11%	+5%
test_benchmark_get_dotted	-1%	-1%	+1%	+0%	-1%	-10%	+2%
test_benchmark_getattr	+0%	+0%	-0%	-1%	+6%	-12%	-1%
test_benchmark_getitem	-3%	+1%	+2%	-4%	+1%	-14%	-0%
test_benchmark_getitem_dotted	-0%	+0%	+1%	+0%	-2%	-8%	+0%
test_benchmark_init_full_overlap	-2%	+0%	-1%	+1%	+0%	+5%	-1%
test_benchmark_init_no_overlap	-2%	+1%	-0%	+0%	+0%	+5%	-1%
test_benchmark_init_partial_overlap	-2%	+0%	-1%	+1%	-0%	+5%	-1%
test_benchmark_reference_chain	-4%	-0%	+0%	-1%	+2%	-10%	+1%
test_benchmark_splat_args	-0%	+2%	-1%	-0%	+3%	-20%	-2%
test_benchmark_spread_kwargs	-3%	-1%	+2%	-3%	+1%	-10%	-0%

(This comment will be updated on subsequent pushes)

[akaIDIOT]

Comparing median metric between base 032195b and proposed 8954907

Great, it works \o/

Metrics for PyPy 3.10 seems oddly jittery, but the rest almost looks usable 🎉

[akaIDIOT]

PDM on PyPy is currently broken, see pdm-project/pdm#3708

Let's wait on that to get released before merging things...

[akaIDIOT]

this PR's target (e12dd7a) and the HEAD of this PR (f82a3f8)

The HEAD of this PR points to a magical merge that likely will never show up as something that makes sense to anyone, but at least it's a state that will make sense to benchmark.

PDM's been fixed for PyPy, comparing locally works too, I'm thinking this is good enough for now, let's see what it brings to the table for future changes.