Add bm_base64 covering common base64 module APIs

[gpshead]

Motivation python/cpython#143262 but also in general this wasn't covered and @serhiy-storchaka is also doing work in this area that'll become relevant such as python/cpython#143216.

[serhiy-storchaka]

What are the results and the total time?

My suggestions:

• Test an ASCII string input for decoding.
• Balance encoding and decoding.

[gpshead]

Example results comparing before and after python/cpython#143262.

main-before-opt-b64.json ======================== Performance version: 1.14.0 Python version: 3.15.0a3+ (64-bit) revision 6b9a6c6ec3b Report on Linux-6.17.0-8-generic-x86_64-with-glibc2.42 Number of logical CPUs: 15 Start date: 2026-01-02 06:47:53.863590 End date: 2026-01-02 06:54:17.985677 main-after-opt-b64.json ======================= Performance version: 1.14.0 Python version: 3.15.0a3+ (64-bit) revision 61fc72a4a43 Report on Linux-6.17.0-8-generic-x86_64-with-glibc2.42 Number of logical CPUs: 15 Start date: 2026-01-02 06:30:08.791666 End date: 2026-01-02 06:36:26.047033 ### ascii85_large ### Mean +- std dev: 1.04 sec +- 0.01 sec -> 1.03 sec +- 0.01 sec: 1.01x faster Not significant ### ascii85_small ### Mean +- std dev: 20.7 ms +- 0.3 ms -> 20.6 ms +- 0.4 ms: 1.00x faster Not significant ### base16_large ### Mean +- std dev: 6.39 ms +- 0.06 ms -> 6.36 ms +- 0.04 ms: 1.00x faster Not significant ### base16_small ### Mean +- std dev: 376 us +- 3 us -> 372 us +- 3 us: 1.01x faster Not significant ### base32_large ### Mean +- std dev: 585 ms +- 8 ms -> 586 ms +- 7 ms: 1.00x slower Not significant ### base32_small ### Mean +- std dev: 11.4 ms +- 0.1 ms -> 11.5 ms +- 0.1 ms: 1.00x slower Not significant ### base64_large ### Mean +- std dev: 4.66 ms +- 0.03 ms -> 2.06 ms +- 0.01 ms: 2.26x faster Significant (t=749.12) ### base64_small ### Mean +- std dev: 314 us +- 5 us -> 253 us +- 5 us: 1.24x faster Significant (t=64.73) ### base85_large ### Mean +- std dev: 432 ms +- 4 ms -> 431 ms +- 4 ms: 1.00x faster Not significant ### base85_small ### Mean +- std dev: 8.37 ms +- 0.08 ms -> 8.39 ms +- 0.12 ms: 1.00x slower Not significant ### urlsafe_base64_small ### Mean +- std dev: 511 us +- 5 us -> 445 us +- 4 us: 1.15x faster Significant (t=80.03) 

[serhiy-storchaka]

I am going to add also ignorechars to support decoding multi-line data. So it may be worth to separate "raw" and "multi-line" cases, if we want to compare the effect of optimization. We could also separate encoding and decoding. But then we will have too much tests. So this is up to you.

[gpshead]

yeah i'd initially started out with separate encode/decode but the matrix of benchmark cases got large and pyperformance is not really intended as a microbenchmark. improvements in any of these will be reflected in the overall scores.