Commit Graph

2 Commits

Author SHA1 Message Date
Timmy Welch
da54b3a454 Fix decoding hashes 2024-10-14 02:02:26 -07:00
Timmy Welch
0928ed6ccf Optimize memory usage
Add a basic map storage that does manual searches to conserve memory
Change saved hashes format to allow multiple hashes for a given ID
Add a vptree storage

Maps in Go take up a huge amount of space changing IDList to []ID took
  memory from over 1GB down to 200MB (note this was on aarch64 MacOS
  which for some reason uses less memory than aarch64 Linux).
  Exhaustive searches using slices took about 30 ms search now takes
  50-60 ms as it takes longer to iterate a map. Partial hashes will
  speed up searches to 8 ms at the cost of 700MB initial memory usage
  and 400MB idle (though this is on MacOS, which for some reason uses
  less memory that aarch64 Linux so probably more like
  900MB initial -> 600 MB idle on an RPI running Linux)
2024-09-07 14:51:18 -07:00