diff --git a/performance.md b/performance.md
index fd627ac..73b2b81 100644
--- a/performance.md
+++ b/performance.md
@@ -878,6 +878,8 @@ Techniques specific to the architecture running the code
     counts[i & 1] ++
     "branch-free code", benchmark; not always faster, but frequently harder to read
 
+* sorting data can help improve performance via both cache locality and branch prediction, even taking into account the time it takes to sort
+   * use a specialized sort, such as radix sort, if possible
 * function call overhead
 * reduce data copies