Change Details

Weighting and algorithm are very dumb. Just take the interrupt counts and round-robin across cores in descending order. Open questions: * Is there another cheap metric that would be fairer? CPU time? * Weighted packing in-place of round-robin? * Topology- and NUMA-awareness; HT JeffR originally authored the patch; I've tried to clean it up a little bit for FreeBSD. I'm sure any bugs are ones I've introduced :-). Future work includes exporting this information in an MI manner, like intrcnts and intrnames, and exposing in vmstat -i.

Weighting and algorithm are very dumb. Just take the interrupt counts and round-robin across cores in descending order.

Weighting and algorithm are very dumb. Just take the interrupt counts and round-robin across cores in descending order. Open questions: * Is there another cheap metric that would be fairer? CPU time? * Weighted packing in-place of round-robin? * Topology- and NUMA-awareness; HT JeffR originally authored the patch; I've tried to clean it up a little bit for FreeBSD. I'm sure any bugs are ones I've introduced :-). Future work includes exporting this information in an MI manner, like intrcnts and intrnames, and exposing in vmstat -i.