Open szilard opened 6 years ago
This is from 2016 (probably earliest account of this problem): https://github.com/szilard/ml-x1 https://github.com/szilard/ml-x1/issues/1
Besides the scaling by number of cores, also note what happens when a server has 2 (or more) CPU sockets (serious NUMA issues/slowdown).
Some newer results here: https://github.com/szilard/GBM-multicore
This is from 2016 (probably earliest account of this problem): https://github.com/szilard/ml-x1 https://github.com/szilard/ml-x1/issues/1
Besides the scaling by number of cores, also note what happens when a server has 2 (or more) CPU sockets (serious NUMA issues/slowdown).
Some newer results here: https://github.com/szilard/GBM-multicore