mkgist-created gist

text.txt

String stats:
 - Questions:
   - How much memory did we save via deduplication?
   - How many strings did we save via deduplication?
   - What percent of ropes end up flattened?
   - How much memory does eager extensibility save?
   - What percent of rope nodes are directly referenced?
   - How many string compares are we doing?
   - How many bytes of string compares are we doing?
   - How many string hashes are we doing?
   - How many bytes of string hashes are we doing?
   - Would it make sense:
     - to deduplicate during tenuring?
     - to deduplicate and atomize all contents matching atoms?
     - to atomize all deduplicated strings?
     - to atomize all deduplicatable strings (during tenuring)?
     - to preserve hashes on all deduplicatable strings?
     - to deduplicate within tenured heap?
       - with or without atomizing?
 - Scenarios:
   - Some standard (AWSY?) load
   - Nailed-down deterministic load