Re:Optimizing matrix multiplication in C
by tom_mellior on Monday August 29, 2016 @05:29AM
(#12380873)
The transpose "trick" should work the same way and have similar beneficial cache effects. Manually unrolling the sdot loop should probably help, too. Vectorized intrinsics are also available from .NET, so that should work as well: https://msdn.microsoft.com/en-us/library/hh977022.aspx
[ Parent
| on HN ]
There are no comments on this comment yet.