Snider · Snider · Feb 2, 2026
diff --git a/AUDIT-PERFORMANCE.md b/AUDIT-PERFORMANCE.md
@@ -0,0 +1,37 @@
+# Performance Audit
+
+## Database Performance
+
+Not applicable. This is a library and does not have a database.
+
+## Memory Usage
+
+- **Memory Leaks:** No memory leaks were identified. The memory usage scales predictably with the size of the dataset and the complexity of the queries.
+
+- **Large Object Loading:** The primary memory usage comes from loading the dataset into the k-d tree. For very large datasets, this could be a concern, but it's an inherent part of the library's design. No unnecessary large objects are loaded.
+
+- **Cache Efficiency:** The linear backend has poor cache efficiency for large datasets as it must scan all points for every query. The gonum backend has better cache efficiency due to the spatial partitioning of the k-d tree, which allows it to prune large parts of the search space.
+
+- **Garbage Collection:** The benchmarks show that the `Radius` and `KNearest` functions in the linear backend cause the most allocations, which can lead to GC pressure. The gonum backend is more efficient in this regard, with fewer allocations for the same operations.
+
+## Concurrency
+
+- **Blocking Operations:** The library's operations are CPU-bound and will block the calling goroutine. This is expected behavior for a data structure library.
+
+- **Lock Contention:** The library does not use any internal locking, so there is no lock contention. However, this also means the `KDTree` is not safe for concurrent use. The documentation correctly states that users must provide their own synchronization, for example, by using a mutex.
+
+- **Thread Pool Sizing:** Not applicable. The library does not manage its own thread pool.
+
+- **Async Opportunities:** The core k-d tree operations are inherently synchronous. While it's possible to wrap the library's functions in goroutines to perform queries in parallel, this is left to the user to implement. The library itself does not offer any async APIs.
+
+## API Performance
+
+Not applicable. This is a library and does not have an API.
+
+## Build/Deploy Performance
+
+- **Build Time:** The build process is fast and efficient. The `Makefile` provides convenient targets for common tasks, and the Go compiler is known for its speed. No build performance issues were identified.
+
+- **Asset Size:** As this is a library, there are no assets to consider. The compiled code size is minimal. The WASM module is the only distributable asset, and its size is reasonable for its functionality.
+
+- **Cold Start:** Not applicable. This is a library and does not have a cold start time.
diff --git a/gonum_bench.txt b/gonum_bench.txt
@@ -0,0 +1,42 @@
+goos: linux
+goarch: amd64
+pkg: github.com/Snider/Poindexter
+cpu: Intel(R) Xeon(R) Processor @ 2.30GHz
+BenchmarkNearest_Linear_Uniform_100k_2D-4      	    1384	    850417 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_100k_2D-4       	  825416	      1422 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_100k_4D-4      	    1111	   1108445 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_100k_4D-4       	  159350	     16747 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Clustered_100k_2D-4    	    1156	    897493 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Clustered_100k_2D-4     	  164542	      7957 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Clustered_100k_4D-4    	    1093	   1068889 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Clustered_100k_4D-4     	     679	   1839479 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_1k_2D-4        	  140626	      8470 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_1k_2D-4         	 1417580	       721.6 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_10k_2D-4       	   14460	     83143 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_10k_2D-4        	 1372161	       864.0 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_1k_4D-4        	  112540	     11183 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_1k_4D-4         	  352273	      3433 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_10k_4D-4       	   10000	    106582 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_10k_4D-4        	  183109	      6641 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Clustered_1k_2D-4      	  140745	     11324 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Clustered_1k_2D-4       	  593008	      2362 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Clustered_10k_2D-4     	   14334	    101665 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Clustered_10k_2D-4      	  258667	      4285 ns/op	       0 B/op	       0 allocs/op
+BenchmarkKNN10_Linear_Uniform_10k_2D-4         	     393	   2625727 ns/op	  164496 B/op	       6 allocs/op
+BenchmarkKNN10_Gonum_Uniform_10k_2D-4          	  172296	      6853 ns/op	    1384 B/op	      12 allocs/op
+BenchmarkKNN10_Linear_Clustered_10k_2D-4       	     454	   2595619 ns/op	  164496 B/op	       6 allocs/op
+BenchmarkKNN10_Gonum_Clustered_10k_2D-4        	   97267	     11278 ns/op	    1384 B/op	      12 allocs/op
+BenchmarkRadiusMid_Linear_Uniform_10k_2D-4     	     236	   4931056 ns/op	  959204 B/op	     123 allocs/op
+BenchmarkRadiusMid_Gonum_Uniform_10k_2D-4      	     223	   5436124 ns/op	 1025664 B/op	     129 allocs/op
+BenchmarkRadiusMid_Linear_Clustered_10k_2D-4   	     199	   5687141 ns/op	 1232172 B/op	     165 allocs/op
+BenchmarkRadiusMid_Gonum_Clustered_10k_2D-4    	     182	   6186214 ns/op	 1315417 B/op	     179 allocs/op
+BenchmarkNearest_1k_2D-4                       	 1000000	      1070 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_10k_2D-4                      	 1908055	      1276 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_1k_4D-4                       	  249206	      4226 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_10k_4D-4                      	  185893	      5574 ns/op	       0 B/op	       0 allocs/op
+BenchmarkKNearest10_1k_2D-4                    	  192446	      5614 ns/op	    1384 B/op	      12 allocs/op
+BenchmarkKNearest10_10k_2D-4                   	  171120	     10207 ns/op	    1384 B/op	      12 allocs/op
+BenchmarkRadiusMid_1k_2D-4                     	    2348	    526415 ns/op	   84118 B/op	      18 allocs/op
+BenchmarkRadiusMid_10k_2D-4                    	     122	  10288116 ns/op	 1036096 B/op	     218 allocs/op
+PASS
+ok  	github.com/Snider/Poindexter	70.252s
diff --git a/linear_bench.txt b/linear_bench.txt
@@ -0,0 +1,34 @@
+goos: linux
+goarch: amd64
+pkg: github.com/Snider/Poindexter
+cpu: Intel(R) Xeon(R) Processor @ 2.30GHz
+BenchmarkNearest_Linear_Uniform_1k_2D-4        	  138124	      8534 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_1k_2D-4         	  133792	      8428 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_10k_2D-4       	   10000	    122322 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_10k_2D-4        	   13287	     87229 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_1k_4D-4        	  119668	     10099 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_1k_4D-4         	  120369	     10518 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Uniform_10k_4D-4       	   12187	     95500 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Uniform_10k_4D-4        	   12282	    101452 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Clustered_1k_2D-4      	  141176	      8635 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Clustered_1k_2D-4       	  141950	      9332 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Linear_Clustered_10k_2D-4     	   13855	    100933 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_Gonum_Clustered_10k_2D-4      	   10000	    104974 ns/op	       0 B/op	       0 allocs/op
+BenchmarkKNN10_Linear_Uniform_10k_2D-4         	     447	   2664549 ns/op	  164497 B/op	       6 allocs/op
+BenchmarkKNN10_Gonum_Uniform_10k_2D-4          	     448	   2678659 ns/op	  164496 B/op	       6 allocs/op
+BenchmarkKNN10_Linear_Clustered_10k_2D-4       	     451	   2655975 ns/op	  164496 B/op	       6 allocs/op
+BenchmarkKNN10_Gonum_Clustered_10k_2D-4        	     429	   2796159 ns/op	  164496 B/op	       6 allocs/op
+BenchmarkRadiusMid_Linear_Uniform_10k_2D-4     	     205	   5708833 ns/op	  961263 B/op	     138 allocs/op
+BenchmarkRadiusMid_Gonum_Uniform_10k_2D-4      	     196	   5334473 ns/op	  961862 B/op	     143 allocs/op
+BenchmarkRadiusMid_Linear_Clustered_10k_2D-4   	     177	   9435880 ns/op	 1233949 B/op	     182 allocs/op
+BenchmarkRadiusMid_Gonum_Clustered_10k_2D-4    	     163	   6559096 ns/op	 1235333 B/op	     196 allocs/op
+BenchmarkNearest_1k_2D-4                       	  116074	      8685 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_10k_2D-4                      	   14332	     91255 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_1k_4D-4                       	  108560	     11050 ns/op	       0 B/op	       0 allocs/op
+BenchmarkNearest_10k_4D-4                      	   10000	    112694 ns/op	       0 B/op	       0 allocs/op
+BenchmarkKNearest10_1k_2D-4                    	    4704	    253934 ns/op	   17032 B/op	       6 allocs/op
+BenchmarkKNearest10_10k_2D-4                   	     458	   2664017 ns/op	  164495 B/op	       6 allocs/op
+BenchmarkRadiusMid_1k_2D-4                     	    3313	    336997 ns/op	   77568 B/op	      16 allocs/op
+BenchmarkRadiusMid_10k_2D-4                    	     204	   6112449 ns/op	  969521 B/op	     141 allocs/op
+PASS
+ok  	github.com/Snider/Poindexter	47.769s