We might not need to cache parsed URLs at all

[alanorozco]

#31594 removes the parsed URL cache on module builds. This is because URL() without cache performs similarly to using <a> with a result cache.

The cache implementation takes about ~250B of a compressed build.

This explores whether we require a cache for <a> as shipped previously on nomodule builds.

Hit rate on real documents

Documents mostly hover on the 65%-75% range.

Methodology / Results

Cache performance at different hit rates

Benchmark / Results

• The cache performs better only for hit rates over around 95%. On Firefox it barely performs better in this range.
• The cache performs worse in most cases (see "dip" in the middle.)

Chrome

Safari

Firefox

[kristoferbaxter]

Thank for the in-depth analysis. Worth taking this to a deep dive to share the methodology more?

[samouri]

Love how deep you went here! Whats the next step? Remove the cache?

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.