cc @mdboom in case there was a reason to omit them.

Ah, pypy3 failure seems relevant. decimal_factorial looks too slow on PyPy3 (with only pure-Python version of the decimal module).

@hugovk, anything else?

I think we can enable decimal_factorial test, but this will need a workaround for pure-Python decimal module. The presence of accelerated module we can detect, trying to import the _decimal module.

1. Either we can lower n and c.prec (we can't use MAX_PREC for pure-Python module) in bench_decimal_factorial()
2. or use different settings conditionally, for pure-Python and C modules

Ok, adapted version of the factorial benchmark works with pure-Python module. Timings (from CI): 287 ms on CPython 3.14 vs 2.36 sec on PyPy.

Edit: with the patch for python/cpython#140036 I got following on decimal_factorial benchmark with pure-Python class:

$ ./python Modules/_decimal/tests/bench.py

# ======================================================================
#                               Factorial
# ======================================================================

n = 100000

cdecimal:
calculation time: 26.301405s
conversion time: 0.000021s

int:
calculation time: 0.809898s
conversion time: 0.441830s


n = 1000000

cdecimal:
calculation time: 397.983149s
conversion time: 0.000115s

int:
calculation time: 39.675428s
conversion time: 10.546071s

So, probably we will have to adjust numbers anyway.

CC @vstinner

Sorry, I'm not working on pyperformance anymore. It's uneasy for me to judge if this benchmark would be a good addition and it would measure correctly the big integer/decimal peerformance.

The problem is that there is no decimal benchmarks at all.