Re: [PATCH 02/10] perf jevents: Sort BigCString entries alphabetically

[sashiko-bot@kernel.org]

> Currently BigCString partitions non-metric event strings before metric
> and metricgroup strings, and orders deduplicated entries within each
> partition by first-seen insertion order (insert_point). Because strings
> shared across multiple CPU models stay at their earliest insertion
> position while model-unique strings are appended later, later models are
> already fragmented across big_c_string, and identically prefixed event
> and metric definitions from different models are scattered far apart.
> 
> Drop insert_number/insert_point so non-folded BigCString entries are
> still partitioned with event strings before metric strings, and sorted
> lexicographically by their raw UTF-8 byte sequence (which starts with
> the NUL-terminated event or metric name) within each partition. Grouping
> identically prefixed event names, event encodings, descriptions, and
> metric expressions adjacent to one another prepares big_c_string for
> block-indexed string compression (splitting big_c_string into
> independently compressed ~8 KiB blocks decompressed on demand by block
> index and intra-block byte offset), shrinking the compressed blocks by
> ~20-23%.
> 
> Regenerate empty-pmu-events.c for the updated string offsets.
> 
> Signed-off-by: Ian Rogers <irogers@google.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261113210822.3112700-1-irogers@google.com?part=2