Skip to content
Arnaud Roger edited this page Jul 11, 2016 · 3 revisions

Latest benchmark results JMH code

number of seconds to read http://www.maxmind.com/download/worldcities/worldcitiespop.txt.gz

Quotes row read the same file with added quotes around the cell. Parallel uses the ParallelReader to load read and decode the file in a separate thread. Concurrent use the concurrent reading of univocity. The percentage is the cost difference compare to the same sfm configuration (quotes/noquotes, singlethreaded/parallel)

SfmCsvParserBenchmark.parseCsv                       0.792 s/op
SfmCsvParserBenchmark.parseCsvQuotes                 1.199 s/op
SfmParallelCsvParserBenchmark.parseCsv               0.661 s/op
SfmParallelCsvParserBenchmark.parseCsvQuotes         0.870 s/op
JacksonCsvParserBenchmark.parseCsv                   1.134 s/op +43%
JacksonCsvParserBenchmark.parseCsvQuotes             1.479 s/op +23%
JacksonParallelCsvParserBenchmark.parseCsv           1.120 s/op +69%
JacksonParallelCsvParserBenchmark.parseCsvQuotes     1.233 s/op +42%
UnivocityCsvParserBenchmark.parseCsv                 1.053 s/op +33%
UnivocityCsvParserBenchmark.parseCsvQuotes           1.739 s/op +45%
UnivocityConcurrentCsvParserBenchmark.parseCsv       0.891 s/op +35%
UnivocityConcurrentCsvParserBenchmark.parseCsvQuotes 1.280 s/op +47%
Clone this wiki locally