Proposal: Add "fma" to the RealFloat class

Levent Erkok erkokl at
Mon May 4 17:36:42 UTC 2015

Yitz: Thanks for taking over. I do agree that "fma" can just be added to
the Num class with all the ramifications and treated as an "optimization."
But that's a different proposal than what I had in mind, so I'm perfectly
happy you pursuing this version.

Just one comment: The name "FMA" is quite overloaded, and perhaps it should
be reserved to the true IEEE754 version. I think someone suggested
'mulAccum' as an alternative, which does make sense if one thinks about the
dot-product operation. Please be absolutely clear in the documentation that
this is not the IEEE754-fma; but rather a fused-multiply-add operation that
is used for the Num class, following some idealized notion of numbers. In
particular, the compiler should be free to substitute "a*b+c" with
"mulAccum a b c".

The latter (i.e., the IEEE754 variant) should be addressed in a different
proposal that I intend to work on separately.


On Mon, May 4, 2015 at 9:11 AM, Yitzchak Gale <gale at> wrote:

> Levent Erkok wrote:
> >> ...I think this proposal needs to be shelved for the time being.
> I wrote:
> > Nevertheless, I vote for doing it now.
> Edward Kmett wrote:
> > As this proposal has been withdrawn, the point is more or less
> > moot for now.
> OK, let me make myself more clear.
> I hereby propose the exact same proposal that Levant originally
> proposed in this thread and then withdrew, with the caveat that
> the scope of the proposal is explicitly orthogonal to any large
> scale change to the way we do floating point.
> Discussion period: 2 weeks, minus time spent so far in this
> thread since Levant's original proposal.
> Thanks,
> Yitz
> _______________________________________________
> Libraries mailing list
> Libraries at
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the Libraries mailing list