This benchmark is using the Mandelbrot set as its task.

The actual algorithm code is not mine and was written by contributors to The Computer Language Benchmarks Game:

• C
• PHP

The code in the game is all licenced under a Revised BSD license.

To run the builds you will need to first ensure you have built the following from source:

• PHP
• HHVM
• Zephir
• QB

It is best to build PHP from source instead of using a precompiled package to level the playing field. You must build HHVM from source otherwise the HHVM extension won't be installable.

So that the build script knows where Zephir and HHVM are installed you'll need to specify the following to environment variables. Run the next two commands in your console (after adjusting the paths!):

export HPHP_HOME=/path/to/hhvm
export ZPHP_HOME=/path/to/zephir

In the root directory of this project running the following command (turn on unlimited scrollback in your console first):

./build.sh

Keep an eye out for any build errors that will need to be rectified before continuing.

Tests are run a number of times so that an average can be taken (by default this is 20 times). Additionally there is a seed value - this is the value that is used to seed the Mandelbrot set calculations. The higher the seed number the more complex the calculation that the programmes must complete.

If you want to ensure that your tests are given favourable scheduling for proc time then review the nice manual page.

So that the exec script knows where Zephir and HHVM are installed you'll need to specify the following to environment variables. Run the next two commands in your console (after adjusting the paths!):

export HPHP_HOME=/path/to/hhvm
export ZPHP_HOME=/path/to/zephir

You'll only need to run this once per ssh connection/console session.

This can take some time to complete as it does a number of iterations with a static seed value (currently defaults to 1000).

./exec.sh

You can specify some options at execution time.

This is the value that is used to seed the Mandelbrot set calculations. The higher the seed number the more complex the calculation that the programmes must complete.

./exec.sh 16000

See the time man page for more information on the tokens you can use.

./exec.sh 1600 "Time in seconds: %e CPU: %P"

This will be quite slow as it does a number of iterations with a number of different seeds:

• iterations x seed
• 20 x 100
• 20 x 200
• 20 x 1000
• 20 x 5000
• 20 x 16000

./reporting_exec.sh

This will output a csv file for each item listed above and additionally a parsed (averaged and prettified) csv for each test.

You can specify some options at execution time.

Tests are run a number of times so that an average can be taken (by default this is 20 times). The following would run all the tests 100 times.

./reporting_exec.sh 100

This package falls under the 3-clause BSD licence. See LICENCE.markdown for more.