ParentFull threadthe_svd_doctor·FFT is memory bound (it’s N log N flops for N bytes, so little arithmetic). GPU HBM is much faster than DRAM, so generally it’s much faster on GPU.View on HN