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
1147.87 ms
total CPU used
10468.26 ms
speedup
9.12×
vs serial
efficiency
91.2%
of 10× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 24.37 24.38 1128.17 1103.79 0
1 1.933 1.95 15.11 1108.12 1093.01 0.14
2 1.463 3.43 19 1096.82 1077.82 0.2
3 1.439 4.89 16.97 868.27 851.3 0.21
4 2.507 7.42 43.1 1057.11 1014.01 0.22
5 4.058 11.49 49.19 1118.28 1069.09 0.23
6 1.348 12.85 42.82 1126.68 1083.86 0.24
7 8.498 21.36 80.25 1144.72 1064.47 16.67
8 1.576 22.95 45.9 1085.88 1039.98 2.25
9 1.389 24.36 54.54 1125.47 1070.93 2.28
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.