[Kc] group research project idea

[David Nicol]

a while ago someone suggested that we use the group mailing list to
discuss ideas for YAS grants.  I have one, which would make sense
done as a group.

The idea is a research project into the difference it would make to the
speed of perl if all (or a significant number of them, anyway) of the
flags-based
dynamic differences in the basic operations on scalar variables were replaced
with an array of virtual function-pointer tables each designed to handle
a particular flag state.

We would create this perl by adding a field to the basic data types to
hold a pointer to the vtable, modifying all the code in the basic data access
macros (such as SvOK and SvIV -- everything mentioned in perldoc perlapi
and perldoc perlguts) to become calls to an offset in a vtable instead of
macros.  This would be a lot of work, but it could be made a couple orders
of magnitude smaller by using a naming scheme and conditional compilation
directives and including a file with function definitions very similar to what
already exists 64 times (if we are virtualizing on six flags)  with different
settings for the preprocessor defines.

Anyone want to work on this, YAS grant or no?  We could treat it like
a group project, with assigned tasks and dependency graphs and
everything, and it would at least give us something to talk about at
the meetings besides the quality of the wardriving in different neighborhoods.


David Nicol


-- 
Although efforts are under way to mitigate the problem, this message
may contain flippancy, hyperbole and/or confusing assertions.  Please
reply directly to fall2006sigfile at davidnicol.com for clarification of any points
appearing unclear, vague, cruel, frustrating, threatening, negative,
dilletantish or otherwise unprofessional before taking action based on
misintepretation or misconstruction of such points.