Previous | Next --- Slide 43 of 60
Back to Lecture Thumbnails
schaganty

Observation from lecture: making use of SIMD instructions is energy-efficient because we are amortizing the cost of fetching+decoding each instruction across all the parallel data units

awu

SIMD with low vector utilization would be inefficient though, given that more register is needed consistently. See https://en.wikipedia.org/wiki/SIMD#Disadvantages

Please log in to leave a comment.