Profiling/Optimizing a Lua application

[Rena <hyperhacker@...>]

I've been working on a Game Boy emulator, which is written mostly as a
collection of C modules (CPU emulator, memory, renderer, etc) which
are glued together by a Lua startup script. The script loads the
modules, connects them together as necessary, loads a game, and sets
the CPU off running. The frontend uses GTK 3, with C modules
responsible for pushing images to the screen and reading the
keyboard/joypad, and Lua (via lgi) running the main loop.

The gist of the main application loop is:
local function main()
	repeat
		gb.cpu:run_frames(frameskip + 1) --execute one or more frames
		coroutine.yield()
	until shutting_down
	Gtk.main_quit()
end
local main_thread = coroutine.create(main)
glib.idle_add(glib.PRIORITY_HIGH_IDLE + 25, --just below redrawing (or
else it won't redraw)
	function() coroutine.resume(main_thread) return true end)
Gtk.main()

I was told by a Gnome developer that this is the ideal way to work
with GTK, versus calling main_loop_do() every frame. This means for
every emulated frame (60/sec), GTK (C) calls this code (Lua) which
calls a coroutine (Lua) which calls the CPU core (C). I'm curious if
this constant switching between Lua and C code, and/or calling and
yielding a coroutine, is a significant bottleneck. I'm also wondering
what's a good way to go about profiling this program to find what
needs optimization.

-- 
Sent from my (emulated) Game Boy.