CPU scaling benchmark
workers
3 +1 main
iters total
500M
125000000/stream
elapsed
1113.96 ms
total CPU used
4346.59 ms
speedup
3.9×
vs serial
efficiency
97.5%
of 4× ideal
| stream | spawn ms | spawned@ | work start@ | work end@ | work ms | reap wait ms |
|---|---|---|---|---|---|---|
| 0 (main) | 0 | 4.73 | 4.74 | 1083.03 | 1078.29 | 0 |
| 1 | 1.784 | 1.8 | 16.1 | 1107.1 | 1091 | 24.17 |
| 2 | 1.435 | 3.26 | 16.1 | 1100.28 | 1084.18 | 17.37 |
| 3 | 1.437 | 4.72 | 18.05 | 1111.17 | 1093.12 | 28.23 |
main
w1
w2
w3
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.