Maybe. If you're just loading constants onto the stack, it seems like it would be pretty forgiving of optimizations and such, and even work across architectures (assuming the same endian-ness, at least).

I'd definitely look for init code for some calculation (as it appears the author has done here), rather than actual computation code, since, as you point out, optimizations could have a huge impact on the instructions emitted for those.