chocolate_bar2410/lunarbench

lunarbench

lunarbench is a benchmarking package based off of deno bench.

lunar bench not only will test benchmarks but will also have the option to print it out as a nice table or as a CSV.

set bench cases, and compare results. bench cases can be grouped together.

methods:

• lunarbench.bench() - creates a new benchcase
• lunarbench.print_result() - prints the rule
• lunarbench.evaluate_as_csv() - returns the result as a CSV
• lunarbench.set_printoptions() - configures the print output

A bench context is passed into every benchcase on evaluation. You can use the bench context's start and stop methods to declaratively mark the start and stop times. Otherwise lunarbench will handle it for you.

example: for loop benchmarking

local lunarbench = require("@lunarbench")

lunarbench.bench("for loop", function(bench_context)
	local tab = table.create(1e7, 1)

	bench_context.start()
	for i = 1, #tab do
	end

end, { baseline = true })

lunarbench.bench("for each", function(bench_context)
	local tab = table.create(1e7, 1)

	bench_context.start()
	for i, v in tab do
	end

end)

lunarbench.bench("ipairs", function(bench_context)
	local tab = table.create(1e7, 1)

	bench_context.start()
	for i, v in ipairs(tab) do
	end

end)

lunarbench.set_printoptions({ column_seperator = " " })

lunarbench.print_result()

the result

 benchmark time/iter (avg) iter/s       (min .. max)    p75    p99   p995 
 --------- --------------- ------ ------------------ ------ ------ ------ 
  for loop          20.1μs   49.7 (19.6μs .. 20.8μs) 20.2μs 20.8μs 20.8μs 
  for each          27.6μs   36.3 (25.9μs .. 32.2μs) 27.8μs 31.9μs 32.0μs 
    ipairs          27.9μs   35.8 (26.1μs .. 32.1μs) 28.4μs 31.2μs 31.7μs 

summary
    for loop
        0.73x faster than for each
        0.72x faster than ipairs