That was more a case of 'let's write it down before I forget it'.
On Mon, 8 Jun 2020, Brad Chamberlain wrote:
We definitely adopted the real(n)/int(n) syntax in Chapel with an eye
toward eventually supporting int(128) or real(128) or real(96)
I remember that.
or the like (which at the time we thought was much closer to being
important than it's turned out to be) or real(16), int(1), or the like
going the other way.
Yes - 128 bit floating point numbers in hardware have been a long time
coming.
To date, though, we haven't had compelling cases that have caused us to
chase after implementing such variations (either in terms of use cases
or, as you note, hardware), so they've languished.
I was not rushing you. I only raised it because the work for having
_Float16 & _Float128
in the C2X standard is happening now. Maybe schedule it for Chapel 1.30!!
But I don't think there's anything significant that would prevent us
from adding them to Chapel ultimately. It's definitely the case that
support from back-end compilers (which support in C would help with)
would ease the ability to support such features; otherwise, we have to
plumb them all the way through ourselves.
Hopefully there will be support in the back end for software versions of
int(128) and uint(128) as hardware instructions for these is non-existent.
I think support for real(96) and int(96) and uint(96) is too hard!
Regards - Damian
Pacific Engineering Systems International, 277-279 Broadway, Glebe NSW 2037
Ph:+61-2-8571-0847 .. Fx:+61-2-9692-9623 | unsolicited email not wanted here
Views & opinions here are mine and not those of any past or present employer
_______________________________________________
Chapel-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-developers
