I was bored last night, so I added single-precision float arrays to
the VM. I called the class "sfloat-array", in a vocab "sfloat-arrays",
and used lowercase f{ } for literal sfloat-array brackets. They look
and act like good old double-precision float-arrays, including being
transparent to the C interface, but use half the memory and have half
the precision. Great for feeding video cards and vector units.
There were some bumps in the bootstrapping process that I had to sort
of weasel my way around, and I'd like to know if there's a "right way
to do it" I missed. In bootstrap.layouts, I say "IN: sfloat-arrays
DEFER: sfloat-array" to create the sfloat-array word, because the
layouts need to be defined before the builtin classes can be defined
in bootstrap.primitives, which would be where the word for the class
normally gets defined. Also, I had to avoid using f{ in sfloat-arrays,
because the new syntax word gets created in stage1 by bootstrap.syntax
but isn't actually defined until stage2 reloads the core syntax vocab.
And an unrelated question: why are byte-array, float-array, and bit-
array implemented as separate primitive types in the VM? It seems like
the latter two, and indeed any packed unboxed (c-type)-array, could be
implemented in pure Factor on top of byte-array. I'm guessing
efficiency has something to do with it.
-Joe
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Factor-talk mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/factor-talk
