Em Mon, Sep 02, 2019 at 06:13:17PM +0300, Alexey Budankov escreveu: > > On 02.09.2019 15:12, Jiri Olsa wrote: > > Adding new --per-numa option to aggregate counts per NUMA > > nodes for system-wide mode measurements. > > > > You can specify --per-numa in live mode: > > > > # perf stat -a -I 1000 -e cycles --per-numa > > # time numa cpus counts unit events > > It might probably better have 'node' instead of 'numa' as in the > option name '--per-node' as in the table header, like this:
Agreed > > # time node cpus counts unit events > 1.000542550 0 20 6,202,097 cycles > 1.000542550 1 20 639,559 cycles > 2.002040063 0 20 7,412,495 cycles > 2.002040063 1 20 2,185,577 cycles > 3.003451699 0 20 6,508,917 cycles > 3.003451699 1 20 765,607 cycles > ... > > BR, > Alexey -- - Arnaldo
