CPU scaling benchmark
workers
3 +1 main
iters total
500M
125000000/stream
elapsed
1146.5 ms
total CPU used
4390.95 ms
speedup
3.83×
vs serial
efficiency
95.8%
of 4× ideal
| stream | spawn ms | spawned@ | work start@ | work end@ | work ms | reap wait ms |
|---|---|---|---|---|---|---|
| 0 (main) | 0 | 5.34 | 5.35 | 1086.92 | 1081.57 | 0 |
| 1 | 2.064 | 2.08 | 16.46 | 1105.05 | 1088.59 | 18.3 |
| 2 | 1.564 | 3.67 | 16.71 | 1111.48 | 1094.77 | 24.65 |
| 3 | 1.632 | 5.32 | 17.69 | 1143.71 | 1126.02 | 56.91 |
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.