[PD] pd on MacPro intel Xeon 64 bit
martin.peach at sympatico.ca
Sat Aug 4 17:50:12 CEST 2007
Hans-Christoph Steiner wrote:
> Just curious, why wouldn't there ever be a 64-bit version of GEM?
> I've been thinking that a 64-bit build of Pd-extended would be nice to
> have, especially because using 64-bit floats as Pd's native format
> would make it much more flexible. You could represent UNIX timestamps
> properly, for example.
Max/msp already uses doubles for its 'float' type, even on 32-bit
versions. As Chris said, in a 64-bit build the floats stay 32 bit. You
would have to declare them all as double instead.
In pd the t_float would probably have to be changed to a
pointer-to-double since the double is too big to fit in an atom
(assuming that atoms must all be the same size).
Perhaps a script could be written to change all the "float" to "double*"
and prefix all the float assignments with "*".
Also all the places where gcc complains about "type-punning", where a
string of four bytes is interpreted as a float, would need to be looked
at. And everywhere (especially in externals) float variables have been
used instead of t_floats. And anywhere the guts of the float are
manipulated directly (in osc~ this is done using a double).
More information about the Pd-list