CPU scaling benchmark

W=0 W=1 W=2 W=3 W=4 W=6 W=8 W=9 W=10 W=12 W=16 iters: 100M 500M 1B 2B
workers
8 +1 main
iters total
500M
55555555/stream
elapsed
1152.33 ms
total CPU used
9378.16 ms
speedup
8.14×
vs serial
efficiency
90.4%
of 9× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 15.71 15.72 1053.54 1037.82 0
1 2.189 2.21 25.82 1127.92 1102.1 76.29
2 1.682 3.91 29.46 953.01 923.55 0.12
3 1.651 5.58 31.34 1088.27 1056.93 40.36
4 1.704 7.3 36.7 1027.55 990.85 0.17
5 1.599 8.92 36.78 1149.25 1112.47 95.83
6 1.721 10.66 37.24 1109.07 1071.83 60.82
7 1.639 12.31 57.27 1079.23 1021.96 28.71
8 3.361 15.69 63.61 1124.26 1060.65 70.85
main
w1
w2
w3
w4
w5
w6
w7
w8
    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.