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
16 +1 main
iters total
100M
5882352/stream
elapsed
308.08 ms
total CPU used
3171.21 ms
speedup
10.29×
vs serial
efficiency
60.5%
of 17× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 64.79 64.8 288.39 223.59 0
1 2.23 2.25 14.92 168.88 153.96 0.18
2 1.79 4.05 16.87 207.14 190.27 0.23
3 1.796 5.86 18.08 204.53 186.45 0.25
4 1.601 7.48 44.39 249.33 204.94 0.26
5 7.897 15.39 46.24 240.66 194.42 0.28
6 2.19 17.59 52.67 256.32 203.65 0.29
7 1.759 19.36 62.44 252.18 189.74 0.3
8 1.729 21.11 65.31 263.19 197.88 0.31
9 13.374 34.5 81.89 270.88 188.99 0.32
10 6.136 40.65 77.2 288.83 211.63 5.54
11 15.839 56.51 87.57 282.78 195.21 0.33
12 2.155 58.68 105.53 294.37 188.84 6.88
13 1.621 60.32 125.31 280.66 155.35 5.48
14 1.454 61.79 145.27 300.2 154.93 11.99
15 1.461 63.26 135.27 304.87 169.6 16.61
16 1.506 64.78 125.31 287.07 161.76 5.5
main
w1
w2
w3
w4
w5
w6
w7
w8
w9
w10
w11
w12
w13
w14
w15
w16
    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.