mycpp Code Generation

Measure the speedup from mycpp, and the resource usage.

Source code: oil/mycpp/examples

User Time (milliseconds)

Lower ratios are better.

example name C++ Python C++ : Python
fib_iter 8 890 0.008
modules 2 174 0.010
loops 4 308 0.012
fib_recursive 11 879 0.012
asdl_generated 11 370 0.029
parse 27 770 0.034
scoped_resource 47 1,031 0.046
files 4 69 0.054
tuple_return_value 16 185 0.087
containers 12 119 0.099
classes 3 22 0.146
cartesian 74 336 0.220
length 45 202 0.222
escape 106 343 0.310
cgi 253 509 0.497
varargs 27 16 1.737
control_flow 208 107 1.947

Max Resident Set Size (MB)

Lower ratios are better. We use MB (powers of 10), not MiB (powers of 2).

example name C++ Python C++ : Python
classes 4.5 10.7 0.41
parse 3.8 7.6 0.50
cartesian 3.5 6.9 0.51
escape 3.5 6.9 0.51
fib_recursive 3.5 6.9 0.51
fib_iter 3.7 7.1 0.52
scoped_resource 3.7 7.1 0.52
asdl_generated 3.7 6.9 0.53
cgi 3.7 6.9 0.53
files 3.7 6.9 0.53
loops 3.8 7.1 0.54
control_flow 3.8 6.9 0.55
length 3.8 6.9 0.55
modules 3.8 6.9 0.55
tuple_return_value 3.9 6.9 0.57
containers 28.9 47.7 0.61
varargs 5.5 6.9 0.79

System Time (milliseconds)

Lower ratios are better.

example name C++ Python C++ : Python
asdl_generated 0 12 0.000
cgi 0 8 0.000
classes 0 9 0.000
control_flow 0 8 0.000
escape 0 8 0.000
fib_iter 0 4 0.000
fib_recursive 0 4 0.000
length 0 8 0.000
modules 0 8 0.000
scoped_resource 0 4 0.000
containers 4 20 0.199
files 4 8 0.458
parse 4 8 0.475
tuple_return_value 4 8 0.499
varargs 39 55 0.709
cartesian 16 0 inf
loops 0 0 NA

raw benchmark files