>Could you explain what do you want to make?

We are using a C library (BOP, link here) that implements OpenMP-like directives, using processes instead of=20 threads to guarantee correctness. BOP uses forks to achieve this, so at a point (at the start of forks) we have different tasks with the exact=20 same address space. Because of how the library is implemented, these=20 tasks cannot modify to the same virtual address. Thus, we want the=20 tasks to allocate a new heap page using our custom memory allocator=20 (which replaces libc malloc calls in ruby). To force this new heap page=20 allocation, we want to use the method in the last message, as we are=20 trying to minimize the changes to ruby itself. I hope this is a bit more clear. I am trying to not go too much into the whole BOP library in itself= , for the sake of simplicity. If you're curious/interested you can read= the blog post (linked above) and/or the papers attached in there. Thanks a= gain!

-Cesar De Hoyos

On = Tue, Aug 4, 2015 at 1:54 PM, SASADA Koichi <ko1@atdot.net> wrote= :

On 2015/08/05 0:08, Cesar De Hoyos wrot= e:
> process-based support for parallelization

Could you explain what do you want to make? At least I can't understand=
what do you want. So that I can't explain how to solve.

--
// SASADA Koichi at atdot dot net

--089e01228c70807e88051c80bf93--