Interesting to see that we do have that now with Cooperative Vectors coming to Direct X.Don't really have a suitable api that would work across tensor and xmx etc. Hense would need them to implement it for their hardware, say cuda for Nvidia (assuming that's suitable)