Ohhhh, brainfuck again, srsly?

Why?

This is my attempt to build a JIT powered BF interpreter using RPython toolchain. In order to make it even faster I have build a simple compiler, that generates flat and slightly optimized version of original *.b program.

Preparation

First things first, start with building the VM itself, PyPy toolchain is required.

python pypy3-v5.5.0-src/rpython/bin/rpython -Ojit bin/wtf_runner.py 

This should produce native wtf_runner-c binary - BF VM, yes.

Run

• First build VM compatible representation:

$ wtf_compile -i src/mandelbrot.b -o mandelbrot.wtf

• Run it with wtf_runner-c

~/workspace/bf $ time ./wtf_runner-c hanoi.wtf > /dev/null
real	0m0.370s
user	0m0.344s
sys	0m0.024s

~/workspace/bf $ time ./wtf_runner-c mandelbrot.wtf > /dev/null
real	0m3.143s
user	0m3.120s
sys	0m0.020s

In comparison to one of the fastest optimizing interpreters (written in pure C, ofc) I have found on the Internet:

~/workspace/bf $ time ../bff/bff src/hanoi.b > /dev/null 
real	0m0.537s
user	0m0.532s
sys	0m0.004s

~/workspace/bf $ time ../bff/bff src/mandelbrot.b > /dev/null 
real	0m4.777s
user	0m4.776s
sys	0m0.000s

See other assets inside src folder. Have fun!