oldk1331 wrote:
>
> My point is, base 2 and IEEE format can make sure
> DFLOAT in FriCAS is hardware accelerated, like
> MachineInteger, U32Vector, etc.
But DFLOAT is hardware accelerated _now_. We use whatever
underlying Lisp offers us and all supported Lisps
use machine arithmetic to perform operations
from DFLOAT.
Note: there are performance traps there, but they have
nothing to do with IEEE.
If you want to look at details try:
ellipticRF(1.0, 0.9, 0.8)
)lisp (disassemble #'|ELIPIDF;ellipticRF;4Df;3|)
This will show you disassembly of real version of ellipticRF
function. Most of time floating point numbers are kept
in machine registers (actually SSE registers) and operations
are performed just using machine operations. Speed is
decent, but C compiler can generate better code.
You may also compare this with complex version, which is
_much_ slower.
--
Waldek Hebisch
--
You received this message because you are subscribed to the Google Groups
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/fricas-devel.
For more options, visit https://groups.google.com/d/optout.
