fix: maintain results_size and top_k_size manually

Calling size() is only faster if it is cached which because we call pop, it is not.

https://stackoverflow.com/questions/2753381/performance-of-vectorsize-is-it-as-fast-as-reading-a-variable

Author: aminya

src/filter.cc

@@ -25,6 +25,7 @@ void filter_internal(const std::vector<CandidateString> &candidates,
   size_t max_results,
   CandidateScorePriorityQueue &results) {
     const auto scoreProvider = options.usePathScoring ? path_scorer_score : scorer_score;
+    auto results_size = results.size();
     for (size_t i = 0, len = candidates.size(); i < len; i++) {
         const auto &candidate = candidates[i];
         if (candidate.empty()) {
@@ -33,8 +34,10 @@ void filter_internal(const std::vector<CandidateString> &candidates,
         auto score = scoreProvider(candidate, query, options);
         if (score > 0) {
             results.emplace(score, start_index + i);
-            if (results.size() > max_results) {
+            ++results_size;// maintain size manually rather than calling results.size() every time
+            if (results_size > max_results) {
                 results.pop();
+                --results_size;
             }
         }
     }
@@ -86,13 +89,18 @@ std::vector<CandidateIndex> filter(const vector<std::vector<CandidateString>> &c
     // Do the work for first thread.
     filter_internal(candidates[0], 0, query, options, max_results, top_k);//inbounds (candidate_size >= 1)
     // Wait for threads to complete and merge the results.
+
+    auto top_k_size = 0u;// maintain size manually rather than calling top_k.size() every time
     for (size_t i = 1; i < candidates_size; i++) {
         threads[i - 1].join();//inbounds
         while (!results[i].empty()) {
             top_k.emplace(results[i].top());
             results[i].pop();
-            if (top_k.size() > max_results) {
+
+            ++top_k_size;
+            if (top_k_size > max_results) {
                 top_k.pop();
+                --top_k_size;
             }
         }
     }