Slow performance of compression.lzma and compression.bz2

[robquant]

While comparing Python version performance with https://github.com/lucianmarin/pybench I found that python-build-standalone versions show considerably slower compression/decompression performance than OS versions on Linux and macOS.

$ lsb_release -a
No LSB modules are available.
Distributor ID: Debian
Description:    Debian GNU/Linux 13 (trixie)
Release:        13
Codename:       trixie
~/pybench$ /usr/bin/python3 -c "import sys;print(sys.version)"
3.13.5 (main, Jun 25 2025, 18:55:22) [GCC 14.2.0]
~/pybench$ /usr/bin/python3 bench.py
Compress using BZ2 algorithm:
[========================================] 100.0% 0:00:09
Compress using LZMA algorithm:
[========================================] 100.0% 0:00:08
[...]
~/pybench$ uv run --python 3.13.5 --managed-python python bench.py
Compress using BZ2 algorithm:
[========================================] 100.0% 0:00:27
Compress using LZMA algorithm:
[========================================] 100.0% 0:00:31
[...]

I believe the cause is the same as in #761.
The CFLAGS in build-bzip2.sh and build-xz.sh overwrite the CFLAGS in the respective upstream Makefile. When an -O2 is added to the CFLAGS in each build script, the performance of the python-build-standalone version in in line with the system Python!

[robquant]

Might also be the cause of #327

[zanieb]

Here's the related comment in #761 (comment)

[jjhelmus]

I agree that this is likely caused by CFLAGS and a lack of an optimization flag to the compiler.

Adding optimization flags to the dependency specific scripts as was done in #768 is one solution to this but I think a better option would be a standardize set of optimization (and hardening, see #837) flags based upon the configuration being built.