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
9 +1 main
iters total
500M
50000000/stream
elapsed
1153.05 ms
total CPU used
10432.83 ms
speedup
9.05×
vs serial
efficiency
90.5%
of 10× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 23.52 23.53 1090.75 1067.22 0
1 2.143 2.16 23.65 1034.46 1010.81 0.15
2 1.623 3.81 39.61 1132.79 1093.18 42.16
3 1.563 5.39 25.91 1138.15 1112.24 47.5
4 1.584 6.99 32.71 1041.97 1009.26 0.24
5 1.65 8.67 37.99 1042.58 1004.59 0.25
6 1.555 10.24 39.63 1129.65 1090.02 41.8
7 4.433 14.69 50.12 1097.16 1047.04 6.51
8 1.822 16.53 68.07 1150.56 1082.49 59.94
9 6.964 23.51 61.12 977.1 915.98 0.27
main
w1
w2
w3
w4
w5
w6
w7
w8
w9
    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.