What backends support packed native arrays at this point?  And what's
the performance like?

Native access to packed arrays is the big thing I'm looking for before I
port a pile of source filtered Perl 5 code to Perl 6.  It's a simple 3D
engine, so all of the libraries I need to work with want to work with
pointers to huge arrays of simple and structured C data types -- and I
need the Perl code to be able to directly manipulate those arrays.

In Perl 5 I'm using PDL for this, and for certain operations it's not
too bad, but it's REALLY slow to cross any of the Perl <-> PDL <-> C
borders (even "give me a pointer to the raw data inside a PDL object" is
a slow operation).

Apropos of the recent thread about specification lockdown:

To those wondering why I haven't started the Perl 6 port already, in
anticipation of this feature working:  I'm still adding features (and
even some major rewrites) in the Perl 5 version.  When I do the port, I
want to be able to cut over from one working code base to another.  I
don't want a period of bitrot in which the Perl 6 port is "done" but not
working, and continuously at risk of being out of sync with changes to
the working Perl 5 codebase.  I also want to be able to quickly graduate
from a "pure port" design to a colloquial Perl 6 design, which I can't
safely do and still keep syncing updates reasonable.

If I could expect packed arrays to be working at some *known* point in
the future, I could probably aim for that.  However, my recollection of
the Perl 6 history so far is that itch-scratching and bootstrapping are
the two forces driving the implementations, so I've no idea when I could
expect the feature to appear if it hasn't already -- and hence no desire
to start now and risk the aforementioned bitrot for an unknown and
possibly long period.

(I'm *NOT* complaining about the state of implementation -- the spec
lockdown thread just reminded me that the core team might want to know
why someone who is very excited about Perl 6 isn't jumping in with both
feet.  My gut feeling is that the details may change, but my hesitation
is probably not unusual.)


-'f


Reply via email to