Add a benchmark for rubyboy's headless emulator (#359)

The repo is a gem so we can just use it from our Gemfile.
The rom is included.
The current loop produces results similar to our optcarrot benchmark.

cruby: ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +PRISM [arm64-darwin23]
yjit: ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +YJIT +PRISM [arm64-darwin23]

---------  ----------  ----------  ---------  ----------  ------------  ----------
bench      cruby (ms)  stddev (%)  yjit (ms)  stddev (%)  yjit 1st itr  cruby/yjit
optcarrot  3100.0      0.6         800.8      1.7         3.599         3.871
rubyboy    2764.1      0.5         991.9      1.0         2.606         2.787
---------  ----------  ----------  ---------  ----------  ------------  ----------
Legend:
- yjit 1st itr: ratio of cruby/yjit time for the first benchmarking iteration.
- cruby/yjit: ratio of cruby/yjit time. Higher is better for yjit. Above 1 represents a speedup.

Author: rwstauner

benchmarks.yml

@@ -71,6 +71,8 @@ rack:
   desc: test the performance of the Rack framework with barely any routing.
 ruby-json:
   desc: an optimized version of the json_pure gem's pure Ruby JSON parser.
+rubyboy:
+  desc: Rubyboy is a functional headless GameBoy emulator, run on a specific game cartridge for a specific number of frames.
 rubykon:
   desc: Ruby solver for Go (the boardgame.) Runs many iterations forward from an initial starting board.
 tinygql:

benchmarks/rubyboy/Gemfile

@@ -0,0 +1,5 @@
+# frozen_string_literal: true
+
+source "https://rubygems.org"
+
+gem "rubyboy", git: "https://github.com/sacckey/rubyboy"

benchmarks/rubyboy/Gemfile.lock

@@ -0,0 +1,40 @@
+GIT
+  remote: https://github.com/sacckey/rubyboy
+  revision: e6c7d1d64ed7c6edb0ec6bae25ae3e7ec4cc9319
+  specs:
+    rubyboy (1.5.0)
+      ffi (~> 1.16, >= 1.16.3)
+
+GEM
+  remote: https://rubygems.org/
+  specs:
+    ffi (1.17.1)
+    ffi (1.17.1-aarch64-linux-gnu)
+    ffi (1.17.1-aarch64-linux-musl)
+    ffi (1.17.1-arm-linux-gnu)
+    ffi (1.17.1-arm-linux-musl)
+    ffi (1.17.1-arm64-darwin)
+    ffi (1.17.1-x86-linux-gnu)
+    ffi (1.17.1-x86-linux-musl)
+    ffi (1.17.1-x86_64-darwin)
+    ffi (1.17.1-x86_64-linux-gnu)
+    ffi (1.17.1-x86_64-linux-musl)
+
+PLATFORMS
+  aarch64-linux-gnu
+  aarch64-linux-musl
+  arm-linux-gnu
+  arm-linux-musl
+  arm64-darwin
+  ruby
+  x86-linux-gnu
+  x86-linux-musl
+  x86_64-darwin
+  x86_64-linux-gnu
+  x86_64-linux-musl
+
+DEPENDENCIES
+  rubyboy!
+
+BUNDLED WITH
+   2.4.19

benchmarks/rubyboy/benchmark.rb

@@ -0,0 +1,28 @@
+# frozen_string_literal: true
+
+require_relative "../../harness/loader"
+
+Dir.chdir(__dir__)
+use_gemfile
+
+# Rubyboy has a Bench module which is just a loop around `emulator.step`.
+# https://github.com/sacckey/rubyboy/blob/e6c7d1d64ed7c6edb0ec6bae25ae3e7ec4cc9319/lib/bench.rb
+
+require 'rubyboy/emulator_headless'
+
+# The rom is included in the gem in a sibling directory to the rubyboy code.
+rom_path = File.expand_path("../../roms/tobu.gb", $".detect { |x| x.end_with?("/rubyboy/emulator_headless.rb") })
+
+# A count of 500 produces results similar to our optcarrot benchmark.
+# It's possible there is a number that produces a consistent benchmark without
+# needing to re-initialize but not sure how to determine that.
+count = 500
+
+run_benchmark(200) do
+  # Results are much more consistent if we re-initialize each time.
+  # Reusing the same eumlator increases stddev by 65x.
+  emulator = Rubyboy::EmulatorHeadless.new(rom_path)
+  count.times do
+    emulator.step
+  end
+end