On Dec 30, 8:03 am, Volker Braun <[email protected]> wrote:
> No, if you mmap() with MAP_NORESERVE then it always succeeds (if the
> processor can address it, 16TB on x86_64). Reads always succeed. But
> whenever you write to the memory map, you can get a SIGSEGV.

In the face of Linux's overcommit strategy and its OOM-killer solution
isn't this a bit academic? I quote from the mmap manpage bugs section:

On Linux there are no guarantees like those suggested above under
MAP_NORESERVE. By default, any process can be killed at any moment
when the system runs out of memory.

In particular, it's not so clear there will even be a SIGSEGV to
recover from.

We have one example of a library for which memory extension was
improved upon integration with sage: libpari. However, there the
memory management was set up differently, already allowing for manual
extension. That was just automated. Doing this on libgap sounds like a
LOT of work. Especially if the gap people are already thinking about
redoing their memory management, it sounds like a bad investment.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
Visit this group at http://groups.google.com/group/sage-devel?hl=en.


Reply via email to