CPU scaling benchmark
workers
3 +1 main
iters total
500M
125000000/stream
elapsed
1110.21 ms
total CPU used
4345.77 ms
speedup
3.91×
vs serial
efficiency
97.8%
of 4× ideal
| stream | spawn ms | spawned@ | work start@ | work end@ | work ms | reap wait ms |
|---|---|---|---|---|---|---|
| 0 (main) | 0 | 4.45 | 4.46 | 1083.92 | 1079.46 | 0 |
| 1 | 1.571 | 1.58 | 13.83 | 1105.92 | 1092.09 | 24.27 |
| 2 | 1.411 | 3.01 | 15.12 | 1102.61 | 1087.49 | 18.85 |
| 3 | 1.407 | 4.44 | 16.16 | 1102.89 | 1086.73 | 26.23 |
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.