vectorized offset decoding benchmark

[julienledem]

The tests use the repetition levels example from here: https://blog.twitter.com/2013/dremel-made-simple-with-parquet

to run the benchmark: mvn clean package -pl parquet-benchmarks/ -am -DskipTests -Drat.skip

java -jar parquet-benchmarks/target/parquet-benchmarks.jar org.apache.parquet.benchmarks.VectorizedConversionBenchmark.* running the benchmark with defaults takes a while.

use -h to see benchmark options

[julienledem]

java -jar parquet-benchmarks/target/parquet-benchmarks.jar org.apache.parquet.benchmarks.VectorizedConversionBenchmark.*.* -wi 2 -i 2 -f 8

Benchmark                                               Mode  Samples     Score     Error  Units
o.a.p.b.VectorizedConversionBenchmark.t0_baseLine      thrpt       16  2658.709 ±  69.008  ops/s
o.a.p.b.VectorizedConversionBenchmark.t1_arrayInc      thrpt       16  1372.189 ±  21.640  ops/s
o.a.p.b.VectorizedConversionBenchmark.t2_arrayInc2     thrpt       16  6352.240 ± 123.574  ops/s
o.a.p.b.VectorizedConversionBenchmark.t3_binaryInc     thrpt       16  7341.081 ± 120.916  ops/s
o.a.p.b.VectorizedConversionBenchmark.t4_binaryInc1    thrpt       16  7175.026 ± 535.405  ops/s
o.a.p.b.VectorizedConversionBenchmark.t5_binaryInc2    thrpt       16  3784.155 ± 107.305  ops/s
o.a.p.b.VectorizedConversionBenchmark.t6_binaryInc3    thrpt       16  5031.790 ±  86.004  ops/s
o.a.p.b.VectorizedConversionBenchmark.t7_binaryIncU    thrpt       16  3877.526 ±  54.100  ops/s
o.a.p.b.VectorizedConversionBenchmark.t8_ternaryInc    thrpt       16  4185.769 ±  70.135  ops/s