[bad_user]

As a matter of fact loops that mutate state are not parallelizable because you can't analyze what they do in order to infer intent and thus some imposed ordering with which you could efficiently distribute the required work while at the same time guaranteeing correctness - solve that and you'd solve the Halting problem ;-)

Can't be done.

[cfallin]

Well, you can't solve it in the general case. But parallelizing compilers can and have been written that perform a conservative analysis -- they can prove some subset of all parallelizable loops as parallelizable, and for the ones they can't prove as such, they err toward 'remain serialized'.

For a simple example, code that performs regular array-based accesses, where index computations are simple affine functions (a*i + b) of a single iteration variable 'i', is a case that's received a lot of attention. See e.g. Polly for LLVM.