[Libre-soc-dev] SVP64 Vectorised add-carry => big int add

lkcl luke.leighton at gmail.com
Tue Apr 19 13:00:13 BST 2022

On April 19, 2022 11:43:23 AM UTC, lkcl <luke.leighton at gmail.com> wrote:

>btw i am starting to wonder if this is what is in Power ISA 3.1 MMA.

took a look, it seems to be primarily matrix-focussed, mul-and-accumulate acc = x0*y0 + x1*y1 ... which of course is covered by Vectorised standard v3.0B madd

for mule/mulx, allowing RC-as-src to be separate from RC-as-dest is damn powerful, it allows either RC to be used as a scalar accumulator of the *upper* half of the large mul, or if RC is set vector it gets you all the computations as separate elements, allowing other algorithms beyond big-div to use it.

the more general purpose the better the chances of acceptance.


More information about the Libre-soc-dev mailing list