From 336416a6c4ac311b7e4193308798cfdcaf29d36d Mon Sep 17 00:00:00 2001
From: Sam Thorogood <sam.thorogood@gmail.com>
Date: Tue, 28 Apr 2020 15:28:53 +1000
Subject: [PATCH] clarify benchmark

---
 bench/README.md | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/bench/README.md b/bench/README.md
index 167a8b7..b337c71 100644
--- a/bench/README.md
+++ b/bench/README.md
@@ -1,4 +1,4 @@
-Benchmark code.
+Benchmark code for Node.
 Usage:
 
 ```bash
@@ -16,6 +16,14 @@ This is an odd number, but we're comparing the on-disk UTF-8 bytes (which optimi
 All Unicode code points can be represented as either one or two "lengths" of a JavaScript string, but each code point can be between 1-4 bytes in UTF-8.
 The possible ratios therefore range from 0.25 (e.g., all emoji) through 1.0 (e.g., ASCII).
 
+# Options
+
+By default, the benchmark tool disables and removes native-like implementations in Node.
+It removes `Buffer` plus the native `TextEncoder` and `TextDecoder` from the global scope.
+
+Use `--native` to enable support for them.
+This will speed up `fast-text-encoding`, as it uses `Buffer` when available.
+
 # Results
 
 For the suggested text on my test rig (macOS 3.6GHz i9), output looks like (snipped):