SHORT Local and remote data transfers

EVENTSET
FIXC0 INSTR_RETIRED_ANY
FIXC1 CPU_CLK_UNHALTED_CORE
FIXC2 CPU_CLK_UNHALTED_REF
BBOX0C0 REQUESTS_READS_LOCAL
BBOX1C0 REQUESTS_READS_LOCAL
BBOX0C1 REQUESTS_READS_REMOTE
BBOX1C1 REQUESTS_READS_REMOTE
BBOX0C2 REQUESTS_WRITES_LOCAL
BBOX1C2 REQUESTS_WRITES_LOCAL
BBOX0C3 REQUESTS_WRITES_REMOTE
BBOX1C3 REQUESTS_WRITES_REMOTE

METRICS
Runtime (RDTSC) [s] time
Runtime unhalted [s] FIXC1*inverseClock
Clock [MHz]  1.E-06*(FIXC1/FIXC2)/inverseClock
CPI  FIXC1/FIXC0
Local bandwidth [MByte/s]  1.E-06*((BBOX0C0+BBOX1C0+BBOX0C2+BBOX1C2)*64)/time
Local data volume [GByte]  1.E-09*(BBOX0C0+BBOX1C0+BBOX0C2+BBOX1C2)*64
Remote bandwidth [MByte/s]  1.E-06*((BBOX0C1+BBOX1C1+BBOX0C3+BBOX1C3)*64)/time
Remote data volume [GByte]  1.E-09*(BBOX0C1+BBOX1C1+BBOX0C3+BBOX1C3)*64
Total bandwidth [MByte/s] 1.E-06*((BBOX0C0+BBOX1C0+BBOX0C2+BBOX1C2+BBOX0C1+BBOX1C1+BBOX0C3+BBOX1C3)*64)/time
Total data volume [GByte] 1.E-09*(BBOX0C0+BBOX1C0+BBOX0C2+BBOX1C2+BBOX0C1+BBOX1C1+BBOX0C3+BBOX1C3)*64


LONG
Formulas:
CPI = CPU_CLK_UNHALTED_CORE/INSTR_RETIRED_ANY
Local bandwidth [MByte/s] = 1.E-06*((SUM(REQUESTS_READS_LOCAL)+SUM(REQUESTS_WRITES_LOCAL))*64)/time
Local data volume [GByte] = 1.E-09*(SUM(REQUESTS_READS_LOCAL)+SUM(REQUESTS_WRITES_LOCAL))*64
Remote bandwidth [MByte/s] = 1.E-06*((SUM(REQUESTS_READS_REMOTE)+SUM(REQUESTS_WRITES_REMOTE))*64)/time
Remote data volume [GByte] = 1.E-09*(SUM(REQUESTS_READS_REMOTE)+SUM(REQUESTS_WRITES_REMOTE))*64
Total bandwidth [MByte/s] = 1.E-06*((SUM(REQUESTS_READS_LOCAL)+SUM(REQUESTS_WRITES_LOCAL)+SUM(REQUESTS_READS_REMOTE)+SUM(REQUESTS_WRITES_REMOTE))*64)/time
Total data volume [GByte] = 1.E-09*(SUM(REQUESTS_READS_LOCAL)+SUM(REQUESTS_WRITES_LOCAL)+SUM(REQUESTS_READS_REMOTE)+SUM(REQUESTS_WRITES_REMOTE))*64
--
This performance group measures the data traffic of CPU sockets to local and remote
CPU sockets. It uses the Home Agent for calculation. This may include also data from
other sources than the memory controllers.