New paper out – presented at Hipeac21

Our latest paper on the use of predictive models for accelerating GPU Convolution kernels on #ARM, #NVIDIA, and OpenCL based architectures is finally out. Read it here, free of charge. For a fast overview, check out Paolo’s presentation of the paper at Hipeac21!

https://www.youtube.com/watch?v=_xlqzVP3s6s&t