CPU scaling benchmark
workers
2 +1 main
iters total
500M
166666666/stream
elapsed
1472.63 ms
total CPU used
4365.08 ms
speedup
2.96×
vs serial
efficiency
98.7%
of 3× ideal
| stream | spawn ms | spawned@ | work start@ | work end@ | work ms | reap wait ms |
|---|---|---|---|---|---|---|
| 0 (main) | 0 | 3.14 | 3.14 | 1471.31 | 1468.17 | 0 |
| 1 | 1.745 | 1.78 | 19.07 | 1470.31 | 1451.24 | 0.1 |
| 2 | 1.326 | 3.12 | 19.48 | 1465.15 | 1445.67 | 1.26 |
main
w1
w2
fork+handshake
CPU work
parent reap wait
what this measures
Each stream runs a tight integer LCG loop — working set is one CPU register, no memory access,
no shared data. Speedup = sum(stream CPU time) / wall-clock elapsed. Efficiency = speedup / (workers+1).
100% efficiency means perfect linear scaling; less than 100% is the cost of serial fork setup,
reap tail, SMT/core contention.