medium10 pts · first answer countsgpu-computingoptimization

What kernel fusion saves

Three ops run one after another: y = x * scale; z = y + bias; out = relu(z). How does fusing them into one kernel help?

A.Intermediates stay in registers instead of going to HBMB.It reduces the number of FLOPs by two thirdsC.It moves the work to tensor coresD.It makes the output use less precision

Discussion 0

Join free to post in the discussion.

What kernel fusion saves

Discussion 0

Related questions