Performance comparison of ChaiScript w/ several other scripting languages

ScriptPerformanceComparisons.txt

Summary for simple looped calculation (details and code below):
    - on OS X 10.10.3, Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5

Language	Tested Version		Best Time	Relative Perf	Notes
========	==============		=========	=============	=====
Node.js		v0.10.33		0.0028		1
Lua		v5.3.0			0.0098		3.5
SuperCollider	v3.6.6			0.0138		4.9
Ruby		v2.0.0p481		0.0143		5.1
Python		v2.7.6			0.0413		14.8
ChaiScript	v5.6.0+ (fdcc595)	0.1898		67.8		-DCHAISCRIPT_NO_THREADS -ltcmalloc
ChaiScript	v5.6.0+ (ebc6468)	0.2246		80.2		-DCHAISCRIPT_NO_THREADS -ltcmalloc
ChaiScript	v5.6.0+ (fdcc595)	0.2493		89.0		-DCHAISCRIPT_NO_THREADS
ChaiScript	v5.6.0+ (ebc6468)	0.3325		118.8		-DCHAISCRIPT_NO_THREADS
ChaiScript	v5.6.0+ (fdcc595)	0.3599		128.5		-ltcmalloc
ChaiScript	v5.6.0+ (fdcc595)	0.4206		150.2
ChaiScript	v5.6.0+ (8fc61bf)	0.4949		176.8


////////////////////////////////////////////////////////////
// ChaiScript:
//
// There is a performance unit test that runs this code; you can run it several
// times to get the best run, like this (from the cmake build directory):
//     $ (for i in {1..25}; do echo "Testing $i" 1>&2; ctest -R performance -V | grep Elapsed; done) | sort | head -1
//
// now() is implemented in C++ as follows:
//
// double now()
// {
//   using namespace std::chrono;
//   auto now = high_resolution_clock::now();
//   return duration_cast<duration<double>>(now.time_since_epoch()).count();
// }
//
// It is exposed as follows:
//
//     mChai.add(chaiscript::fun(&now), "now");

var sum = 0.0
var start = now()
for (var i = 1; i <= 100000; ++i) {
	if (i % 2 == 0) {
		sum += 1.0 / i;
	}
	else {
		sum += 1.0 / (double(i) * i);
	}
}
var end = now()
print("Elapsed time: " + to_string(end - start) + " sum: " + to_string(sum))

//
// ChaiScript v5.6.0+ (8fc61bf) Release (measured April 30, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
//     Elapsed time: 0.494943 sum: 6.9322
//
// ChaiScript v5.6.0+ (ebc6468) Release (measured April 30, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
// (compiled w/ -DCHAISCRIPT_NO_THREADS)
//     Elapsed time: 0.332631 sum: 6.9322
//
// ChaiScript v5.6.0+ (ebc6468) Release (measured May 4, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
// (compiled w/ -DCHAISCRIPT_NO_THREADS and linked with TCMalloc: -ltcmalloc_minimal)
// (TCMalloc is from gperftools v2.4)
//     Elapsed time: 0.224560 sum: 6.9322
//
// ChaiScript v5.6.0+ (fdcc595) Release (measured May 5, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
// (compiled w/ -DCHAISCRIPT_NO_THREADS; this commit has a 25% speed improvement compared to ebc6468
//     Elapsed time: 0.249339 sum: 6.9322
//
// ChaiScript v5.6.0+ (fdcc595) Release (measured May 5, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
// (compiled w/ -DCHAISCRIPT_NO_THREADS and linked with -ltcmalloc_minimal; 15% improvement compared to ebc6468)
//     Elapsed time: 0.18984 sum: 6.9322
//
// ChaiScript v5.6.0+ (fdcc595) Release (measured May 5, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
// (linked with -ltcmalloc_minimal)
//     Elapsed time: 0.359919 sum: 6.9322
//
// ChaiScript v5.6.0+ (fdcc595) Release (measured May 5, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
//     Elapsed time: 0.420597 sum: 6.9322



////////////////////////////////////////////////////////////
// SuperCollider:

(
var sum = 0.0;
var dur;
dur = {
	(1..100000).do { arg i;
		if (i % 2 == 0) {
			sum = sum + (1.0 / i);
		} {
			sum = sum + (1.0 / (i * i));
		}
	};
}.bench(false);
"Elapsed time: % sum: %".format(dur.round(1e-6), sum.round(1e-4)).postln;
""
)

//
// SuperCollider v3.6.6 (measured April 30, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
//     Elapsed time: 0.013837 sum: 6.9322



////////////////////////////////////////////////////////////
// Python:

from timeit import default_timer as timer

sum = 0.0
start = timer()
for i in range(1, 100001):
    if i % 2 == 0:
        sum += 1.0 / float(i)
    else:
        sum += 1.0 / (float(i) * i)

end = timer()
print("Elapsed time: {:.6f} sum: {:.4f}".format(end - start, sum))

//
// Python v2.7.6 (measured April 30, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
//     Elapsed time: 0.041317 sum: 6.9322



////////////////////////////////////////////////////////////
// Lua:

sum = 0.0
startTime = os.clock()
for i = 1, 100000 do
    if i % 2 == 0 then
        sum = sum + (1.0 / i)
    else
        sum = sum + (1.0 / (i * i))
    end
end
endTime = os.clock()
print(string.format("Elapsed time: %.6f sum: %.4f", endTime-startTime, sum))

//
// Lua v5.3.0 (measured April 30, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
//     Elapsed time: 0.009793 sum: 6.9322



////////////////////////////////////////////////////////////
// Node.js (Javascript):

var sum = 0.0;
var start = process.hrtime();
for (var i = 1; i <= 100000; ++i) {
    if (i % 2 == 0) {
        sum += 1.0 / i;
    }
    else {
        sum += 1.0 / (i * i);
    }
}
var elapsed = process.hrtime(start);
var elapsedSecs = elapsed[0] + elapsed[1] * 1e-9;
console.log('Elapsed time: %d sum: %d', elapsedSecs.toFixed(6), sum.toFixed(4));

//
// Node.js v0.10.33 (measured April 30, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
//     Elapsed time: 0.002819 sum: 6.9322



////////////////////////////////////////////////////////////
// Ruby:

require "benchmark"
sum = 0.0
elapsed = Benchmark.realtime do
    for i in (1..100000)
        if i % 2 == 0 then
            sum = sum + 1.0 / i
        else
            sum = sum + 1.0 / (i * i)
        end
    end
end
puts "Elapsed time: %.6f sum: %.4f" % [elapsed, sum]

//
// Ruby v2.0.0p481 (measured May 5, 2015 on Mac Pro: 3,7 GHz Quad-Core Intel Xeon E5):
//     Elapsed time: 0.014348 sum: 6.9322