Continue work on benchmarking

benchmarks/README.md

@@ -6,9 +6,12 @@ Legend-state seeks to be the fastest React state management library.  Achieving
 - Micro-optimizations: Optimizations related to JavaScript primitives (iteration, Object types, etc.)
 - Array Optimizations: Optimizations related to the efficient rendering of large lists of data.
 
-## Running Benchmarks
 
-We use [hyperfine](https://github.com/sharkdp/hyperfine) for running benchmarks.  You will need this installed in order to run the benchmark script.
+## Benchmark Dependencies
+
+We use [hyperfine](https://github.com/sharkdp/hyperfine) for running benchmarks.
+
+## Running Benchmarks
 
 All benchmarked optimizations will have a directory within the relevant optimization category directory (i.e. `architecture-optimizations`, `micro-optimizations`, `array-optimizations`).  Each optimization directory will have a list of JavaScript files that each implement an approach and then a `bench.sh` that uses `hyperfine` to run the benchmark.
 

benchmarks/baseline/README.md

@@ -0,0 +1,5 @@
+# Establishing a baseline
+
+When benchmarking Javascript with different engines the startup time of the engine needs to be considered.  Some engines, such as Hermes, are optimized for very quick startup time.  To ensure that we are performing an apples to applies comparison of engine performance, and not startup, we start by establishing a baseline of how long each engine takes to startup.  We then adjust each benchmark by this startup time.
+
+To measure the startup time we have each engine run on a empty `.js` file.  The startup benchmarks are stored in `$ENGINEBaseline.json`.

benchmarks/baseline/bench.sh

@@ -0,0 +1,3 @@
+hyperfine --warmup 3 --runs 1000 --export-json hermesBaseline.json "hermes empty.js"
+
+hyperfine --warmup 3 --runs 1000 --export-json nodeBaseline.json "node empty.js"