7 Sep
2011
7 Sep
'11
4:12 p.m.
On Wed, 7 Sep 2011, Bill Gribble wrote:
So far iteration on plain floats seems to be the best I can come up with, but HADDPS is tantalizingly close to what I want to do. Any hints?
Once I thought that with some commutativity you could speed things up like this :
(f0+f1+f2+f3)+(f4+f5+f6+f7)+...
can be rearranged as :
(f0+f4+...)+(f1+f5+...)+(f2+f6+...)+(f3+f7+...)
But I don't remember whether I tried it or not.
That's a speedup without even using MMX/SSE... theoretically, it can double the speed of a summation like this, and you can apply this boost to sum-of-products to get a certain amount of speedup too.
| Mathieu Bouchard ---- tél: +1.514.383.3801 ---- Villeray, Montréal, QC