On 12/29/25 7:09 PM, Derrick Stolee wrote:

On 12/26/2025 2:44 AM, René Scharfe wrote:

quoted

That number is not limited, though, and in pathological cases like the
one in p6010 we see a sizable improvement:

Test                      v2.52.0           HEAD
------------------------------------------------------------------
6010.4: git show-branch   2.19(2.19+0.00)   0.03(0.02+0.00) -98.6%

I love to see improvements like this, even if the construction is
unlikely to exist in reality. I do think it's likely to be valuable
for some large repos with many parallel branches.

Indeed, I tested this patch against a monorepo with lots of merges
with hyperfine, getting this output:

Benchmark 1: old
  Time (mean ± σ):      3.303 s ±  0.146 s    [User: 0.058 s, System: 0.069 s]
  Range (min … max):    3.162 s …  3.631 s    10 runs

Benchmark 2: new
  Time (mean ± σ):     141.7 ms ±   3.2 ms    [User: 30.5 ms, System: 93.1 ms]
  Range (min … max):   137.5 ms … 149.4 ms    19 runs

Summary
  new ran
   23.31 ± 1.15 times faster than old

Woah, the perf test gets a speedup by factor 46 in a repository
purpose-built to highlight this very difference, and here you get half
of that in the wild!  Interesting to see that there are real commit
histories out there with such a taxing topology.

And thanks for your review!

René