Mark Dufour

just started pushing commits for this (shedskin --float). the following benchmark becomes about 30% faster on my system: ``` for x in range(100): a = [1.1 for a in range(1000000)]...

running into issues with the use of va_arg, which has problems with floats (non-double). there is probably a nice way to replace the crappy va_arg stuff in modern C++.. (variadic...

tagging with 0.9.9 as 0.9.8 removes the va_arg stuff.

added --float32

of course, yes, the nice thing about shedskin is that you get to choose your int type. you can even choose different types for different (extension) modules. but let's be...

for 0.9.9 we should have --float32/float64/int32/int64/int128 options to override default double/int types. don't think having this issue open will make anyone ever start working on it (I won't), so closing.

the -x option was renamed to -t/--traceback in 0.9.6/0.9.7.

let's first see if we can get this to work under windows at all.

very interested related reading about how different languages use this: https://kipp.ly/blog/escape-analysis/

closing for more general issue.