Re: [PATCH] describe: limit default ref iteration to tags

[Junio C Hamano <gitster@pobox.com>]

Tamir Duberstein <tamird@gmail.com> writes:

[jc: Removing Shawn from CC who passed away quite a while ago, RIP].

> Unless --all is given, get_name() rejects every ref outside refs/tags/.
> The rejection happens only after the ref backend has enumerated the ref,
> so repositories with many other refs spend most of a simple describe
> invocation visiting refs which cannot affect its result.
> ...
> Both revisions were built with -O3, -mcpu=native, and ThinLTO using
> Apple clang 21.0.0 on macOS 26.5. The machine was a MacBook Pro
> (Mac16,6) with a 16-core Apple M4 Max (12 performance and four
> efficiency cores) and 128 GB RAM.
>
> Signed-off-by: Tamir Duberstein <tamird@gmail.com>
> ---
>  builtin/describe.c       |  3 +++
>  t/perf/p6100-describe.sh | 20 ++++++++++++++++++++
>  2 files changed, 23 insertions(+)

Interesting.  How would this relate to and work well with
<20260601233727.43558-1-jacob.e.keller@intel.com>?

> +test_lazy_prereq PERF_REFFILES '
> +	test "$(git rev-parse --show-ref-format)" = files
> +'
> +
> +ref_count=10000
> +
>  # clear out old tags and give us a known state
>  test_expect_success 'set up tags' '
>  	git for-each-ref --format="delete %(refname)" refs/tags >to-delete &&
> @@ -27,4 +33,18 @@ test_perf 'describe HEAD with one tag' '
>  	git describe --match=new HEAD
>  '
>  
> +test_expect_success PERF_REFFILES 'set up many unrelated refs' '
> +	git tag -m tip tip HEAD &&
> +	for i in $(test_seq $ref_count)
> +	do
> +		printf "create refs/heads/describe-perf/%05d HEAD\n" $i ||
> +		return 1
> +	done >instructions &&
> +	git update-ref --stdin <instructions
> +'
> +
> +test_perf 'describe exact tag with many loose refs' --prereq PERF_REFFILES '
> +	git describe --exact-match HEAD
> +'
> +

Is there a strong reason to guard this new test behind
`PERF_REFFILES`?

Even though the penalty of enumerating 10,000 unrelated loose
references may be most pronounced in the `files` backend, skipping
unnecessary reference enumeration is an architectural win for other
backends (like `reftable` or a fully packed repository) as well.

If we drop `PERF_REFFILES` and retitle the test to "describe exact
tag with many unrelated refs", we could run it unconditionally to
benchmark the improvement across all storage formats.