Am 26. Mai 2024 09:40:02 MESZ schrieb "Peter P." <[email protected]>:
>
>I suspect that externals have to be compiled for double precision as
>well?
Yes (well: there are some exceptions, but those are very niche)
> Will I have to install specific externals via Deken or apt then?
Via deken: only externals that are compatible will be shown. So when running
Pd64, you will only see externals that can deal with double precision (or are
precision agnostic, eg pure abstractions).
Via apt: *most* externals are built for both precisions, in a single binary
package (eg "pd-zexy" hat both pd32 and pd64).
Caveats:
- I'm currently afk and don't really know the state in Debian/stable (bookworm)
by heart (as I'm mostly working with testing/unstable)
- the packages that provide pd64 binaries have an (alternative) dependency on
"puredata64" (so they can be installed on a system that *only* has pd64)
- however there's no other marker that marks them as pd64 compatible (they
should probably have something like "Provides: pd64-zexy"
- the apt plugin for deken ("pd-deken-apt") does not handle pd64 externals at
all (and will happily install pd32 externals)
>Are all™ such externals available in both flavours?
>
No.
>From skimming Katja's article, I understand that memory usage of patches
>in pd64 will be larger than in single precision.
Mostly not. at least not on 64bit systems (eg the "amd64"/"x86_64" and "arm64"
archs).
On "i386" and "armhf" (old raspberries) the memory footprint will be doubled.
> Will cpu usage also
>increase?
Maybe.
With 32bit floating point values, there are more CPU registers available, so
the compiler could generate more optimized code.
Also, it's easier to hit cache limits.
In practice I wouldn't give much about such conjectures and simply try it out
with a specific patch.
Of course, if you do test it would be nice if you could share some (meaningful)
data...
mfg.sfg.jfd
IOhannes
_______________________________________________
[email protected] mailing list
UNSUBSCRIBE and account-management ->
https://lists.puredata.info/listinfo/pd-list
