Hi,
do I have to run autotvm on the specific CPU I will use for the deployment or it’s enough to run it on a CPU having the same architecture? Example: run autotvm on my local PC with i5 CPU and deploy to xeon CPU (both x86_64).
Thanks.
For the best performance, I recommend re-tuning for your hardware architecture since i5 and xeon although they have the same ISA, they’ll have different architectural features (cache sizes, number of threads etc.)
Ok, thank you. I’ll give it a try.