Turns out that passing dom::element by reference can be a performance killer. (simdjson#1086)

lemire · web-flow · commit 2f92a34bb734 · 2020-08-01T10:31:47.000-04:00
* Turns out that passing dom::element by reference can be a performance killer.

* Tweaking.
diff --git a/benchmark/parseandstatcompetition.cpp b/benchmark/parseandstatcompetition.cpp
@@ -131,7 +131,7 @@ struct Stat {
   size_t stringLength; // Number of code units in all strings
 };
 
-static void GenStatPlus(Stat &stat, const dom::element &v) {
+static void GenStatPlus(Stat &stat, const dom::element v) {
   switch (v.type()) {
   case dom::element_type::ARRAY:
     for (dom::element child : dom::array(v)) {
@@ -142,7 +142,8 @@ static void GenStatPlus(Stat &stat, const dom::element &v) {
     break;
   case dom::element_type::OBJECT:
     for (dom::key_value_pair kv : dom::object(v)) {
-      GenStatPlus(stat, dom::element(kv.value));
+      GenStatPlus(stat, kv.value);
+      stat.stringLength += kv.key.size();
       stat.memberCount++;
       stat.stringCount++;
     }
diff --git a/doc/performance.md b/doc/performance.md
@@ -12,7 +12,7 @@ are still some scenarios where tuning can enhance performance.
 * [Number parsing](#number-parsing)
 * [Visual Studio](#visual-studio)
 * [Downclocking](#downclocking)
-
+* [Best Use of the DOM API](#best-use-of-the-dom-api)
 
 Reusing the parser for maximum efficiency
 -----------------------------------------
@@ -181,3 +181,9 @@ On some Intel processors, using SIMD instructions in a sustained manner on the s
 The simdjson library does not currently support AVX-512 instructions and it does not make use of heavy 256-bit instructions. We do use vectorized multiplications, but only using 128-bit registers. Thus there should be no downclocking due to simdjson on recent processors. 
 
 You may still be worried about which SIMD instruction set is used by simdjson.  Thankfully,  [you can always determine and change which architecture-specific implementation is used](implementation-selection.md) by simdjson. Thus even if your CPU supports AVX2, you do not need to use AVX2. You are in control.
+
+Best Use of the DOM API
+-------------------------
+
+The simdjson API provides access to the JSON DOM (document-object-model) content as a tree of `dom::element` instances, each representing an object, an array or an atomic type (null, true, false, number). These `dom::element` instances are lightweight objects (e.g., spanning 16 bytes) and it might be advantageous to pass them by value, as opposed to passing them by reference or by pointer.
+
