On 07/17/2018 04:46 PM, Derek Martin wrote:
I'm sure you're aware the
reason it hasn't been fixed is because no one has come up with a
solution that does not cause non-threaded programs to take a
significant performance hit.
Isn't the Python GIL analogous to the Linux kernel's old "Big Kernel Lock"?
In both cases the fix is fine-grained locking, and that is work to get
right.
There is a man who has been working on a GILectomy for a few years, I
think pretty much on his own. He gives talk at PyCon (?) every year on
his progress. Sounds like he is getting impressive results. But no one
else seems to care very much, and breaking every binary library makes it
seem like it will never happen.
there are workarounds (like writing an
extension in C),
Or in Cython: I nice way to write glue (or actual code) that can be fast
if you define all your types. Cut all your ties to data Python thinks it
owns, and you can declare no GIL...
-kb
_______________________________________________
Discuss mailing list
[email protected]
http://lists.blu.org/mailman/listinfo/discuss
