[ruby-core:63088] Re: [RFC] vm_method.c (rb_method_entry_make): avoid freed me in m_tbl

From: SASADA Koichi <ko1@...>
Date: 2014-06-11 01:36:05 UTC
List: ruby-core #63088
(2014/06/11 10:13), Eric Wong wrote:
> I'm also thinking about a crazy idea to use giant VM-wide hashes for
> all methods and constants.  With a per-class list_head to:
> 
> a) preserve ordering (compatibility)
> b) ensure deletion from the giant hash in case a class goes away
> 
> This might reduce space waste from small classes.
> 

In this case, what is the key? A pair of [class, name_id]?

> [1] one "other project" involves a lock-free hash table from URCU
>     <http://lttng.org/urcu>.  Unfortunately Ruby cannot depend on
>     LGPLv2.1+ libraries, but maybe it can be optional like gmp.

Interesting. I want to know about idea of *userland* RCU, but I can't
find.  My understanding is that RCU can be achieved because kernel can
dominates shceduler.

-- 
// SASADA Koichi at atdot dot net

In This Thread