Optimize UnicodeHash lookup performance using std::unordered_map #30
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
modernizing, expanded tests, cleanup
cleaning up XmlTools.
update FileUtils to use filesystem::remove
updating enum_types handling.
reworking LogStream
C++ code quality (C++17)
kosloot
added a commit
that referenced
this pull request
Nov 22, 2025
Date: Sat Nov 22 16:07:08 2025 +0100
updated as suggested by #30
entering 2025
updated nlohmann json.hpp
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Pull Request: Optimize UnicodeHash lookup performance using std::unordered_map
Description
This PR optimizes the
Hash::UnicodeHashclass by replacing the internal custom Trie implementation (Tries::UniTrie) with a standardstd::unordered_map.Profiling of the dependent
timblapplication revealed thatUnicodeHash::lookupandUnicodeHash::hashwere significant performance bottlenecks, consuming over 50% of the CPU time during the learning phase on large datasets. This was due to the linear time complexity O(L) of the Trie structure for string lookups. switching to a hash map provides O(1) average time complexity.Changes
include/ticcutils/UniHash.h:Tries::UniTrie<UniInfo> _treewithstd::unordered_map<icu::UnicodeString, UniInfo*, UnicodeStringHash> _map.UnicodeStringHashstruct to supporticu::UnicodeStringkeys.src/UniHash.cxx:hash()andlookup()to usestd::unordered_mapAPI (find,insert).Normalizer2::isNormalizedto avoid expensive normalization if the input string is already in NFC format.~UnicodeHash()to explicitly deleteUniInfopointers stored in the map to prevent memory leaks.Performance Analysis
Benchmarks were run using
timblon theedufineweb_train_000001-100kdataset (~4.9M lines).Profiling Details (Top Functions)
Before (Original):
```text
% cumulative self name
38.44 101.33 101.33 Hash::UnicodeHash::lookup
16.66 145.26 43.93 Hash::UnicodeHash::hash
```
After (Optimized):
```text
% cumulative self name
1.92 96.51 2.19 Hash::UnicodeHash::lookup
1.05 101.13 1.20 Hash::UnicodeHash::hash
```
The bottleneck has been effectively eliminated, shifting the primary processing time to the core algorithm logic (
ClassDistribution::IncFreq).ticcutils_optimization_pr.zip