In https://github.com/rahulmutt/nofib, run using:
nofib-runner imaginary/queens --run --jmh="-wi 10 -i 10" --way="-O2"Queens Solution:
-- !!! count the number of solutions to the "n queens" problem.
-- (grabbed from LML dist)
import System.Environment
main = do
[arg] <- getArgs
print $ nsoln $ read arg
nsoln nq = length (gen nq)
where
safe :: Int -> Int -> [Int] -> Bool
safe x d [] = True
safe x d (q:l) = x /= q && x /= q+d && x /= q-d && safe x (d+1) l
gen :: Int -> [[Int]]
gen 0 = [[]]
gen n = [ (q:b) | b <- gen (n-1), q <- [1..nq], safe q 1 b]gives the following result:
# JMH 1.15 (released 115 days ago)
# VM version: JDK 1.8.0_102, VM 25.102-b14
# VM invoker: /Library/Java/JavaVirtualMachines/jdk1.8.0_102.jdk/Contents/Home/jre/bin/java
# VM options: <none>
# Warmup: 10 iterations, single-shot each
# Measurement: 10 iterations, single-shot each
# Timeout: 10 min per iteration
# Threads: 1 thread
# Benchmark mode: Single shot invocation time
# Benchmark: com.typelead.TestBenchmark.benchmark
# Parameters: (args = 10 +RTS)
# Run progress: 0.00% complete, ETA 00:00:00
# Fork: 1 of 1
# Warmup Iteration 1: 2192.577 ms/op
# Warmup Iteration 2: 491.270 ms/op
# Warmup Iteration 3: 439.554 ms/op
# Warmup Iteration 4: 402.247 ms/op
# Warmup Iteration 5: 407.231 ms/op
# Warmup Iteration 6: 150.034 ms/op
# Warmup Iteration 7: 169.938 ms/op
# Warmup Iteration 8: 156.885 ms/op
# Warmup Iteration 9: 152.766 ms/op
# Warmup Iteration 10: 146.529 ms/op
Iteration 1: 160.485 ms/op
Iteration 2: 154.438 ms/op
Iteration 3: 151.378 ms/op
Iteration 4: 144.522 ms/op
Iteration 5: 149.795 ms/op
Iteration 6: 149.698 ms/op
Iteration 7: 152.700 ms/op
Iteration 8: 143.453 ms/op
Iteration 9: 144.717 ms/op
Iteration 10: 153.089 ms/op
Result "benchmark":
N = 10
mean = 150.427 ±(99.9%) 7.927 ms/op
Histogram, ms/op:
[140.000, 142.500) = 0
[142.500, 145.000) = 3
[145.000, 147.500) = 0
[147.500, 150.000) = 2
[150.000, 152.500) = 1
[152.500, 155.000) = 3
[155.000, 157.500) = 0
[157.500, 160.000) = 0
[160.000, 162.500) = 1
[162.500, 165.000) = 0
[165.000, 167.500) = 0
Percentiles, ms/op:
p(0.0000) = 143.453 ms/op
p(50.0000) = 150.586 ms/op
p(90.0000) = 159.880 ms/op
p(95.0000) = 160.485 ms/op
p(99.0000) = 160.485 ms/op
p(99.9000) = 160.485 ms/op
p(99.9900) = 160.485 ms/op
p(99.9990) = 160.485 ms/op
p(99.9999) = 160.485 ms/op
p(100.0000) = 160.485 ms/op
# Run complete. Total time: 00:00:06
Benchmark (args) Mode Cnt Score Error Units
TestBenchmark.benchmark 10 +RTS ss 10 150.427 ± 7.927 ms/op
The first warmup iteration spends a LOT of time on classloading, something we should probably look into at some point.