Zanie Blue

It looks like they might use -O3 instead of -O2 https://github.com/conda-forge/python-feedstock/blob/50a4e2d4203f05082fcbb93e14541180de3aa8ac/recipe/build_base.sh#L98C1-L104C3 edit: This is already the default in CPython https://github.com/python/cpython/blob/94a7279fbdcfc2ccb6e7bb3c1cec863cce30a591/configure#L9600

Our v3 builds are a bit better, but that's not the bulk of it (geometric mean: 1.20x slower)

3.14 3.13 doesn't feel like a fair comparison since we're using the tail calling interpreter.

Should be fixed in the next release ``` ❯ uv run -p 3.13.2 bench.py 3.13.2 (main, Feb 12 2025, 14:59:08) [Clang 19.1.6 ] Number of long_calcul per second: 54.34 ❯...

And using hyperfine ``` Benchmark 1: uv 3.13.2 Time (mean ± σ): 279.1 ms ± 2.8 ms [User: 275.3 ms, System: 2.8 ms] Range (min … max): 274.9 ms …...

Yeah, it was the LLVM 20 upgrade with the backport for the LLVM 19 regression (conda-forge is on LLVM 18 on macOS and gcc on Linux) https://github.com/llvm/llvm-project/pull/114990. This was merged...

I did a comprehensive benchmark and I'm seeing about a 6% difference ``` ❯ uvx pyperf compare_to conda-forge-313.json pbs-313.json Benchmarks with tag 'apps': =========================== 2to3: Mean +- std dev: [conda-forge-313]...

I can reproduce some of the difference you're reporting for that microbenchmark ``` ❯ hyperfine "/home/zb/.local/share/uv/python/cpython-3.13.2-linux-x86_64-gnu/bin/python3.13 bench.py" "/home/zb/miniforge3/envs/env-py3.13/bin/python bench.py" Benchmark 1: /home/zb/.local/share/uv/python/cpython-3.13.2-linux-x86_64-gnu/bin/python3.13 bench.py Time (mean ± σ): 2.992 s ±...

I got the `x86_64-unknown-linux-musl-noopt` build passing on my machine and gave it a quick test in an alpine container and it looks like I didn't package it right. ``` /test...

`ORIGIN` is supported for RPATH, but seemingly not for NEEDED (ref https://git.musl-libc.org/cgit/musl/tree/ldso/dynlink.c#n897) — switching to that for now.