search

quickfix-j / quickfixj

QuickFIX/J is a full featured messaging engine for the FIX protocol. - This is the official project repository.
http://www.quickfixj.org
Other
955 stars 611 forks source link

Remove "if else" statements which is slower by nearly 7 times compared to `HashMap` in generated `MessageCracker` class #577

Closed busy-spin closed 2 months ago

busy-spin commented 1 year ago

Describe the bug MessageCracker generated class contains a lengthy if else statement to resolve the method call for the respective message type. Current logic is about 7 times slower than using HashMap key'd by message type and value as the method reference.

I wrote a new MessageCracker.xsl which use the HashMap to resolve the method respective to message type.

Following is the JMH benchmarking results.

Before:

benchmarkCrackingGenerated - use the current MessageCracker.xls generated class benchmarkCrackingWithHashMap use custom MessageCracker class which use HashMaps to resolve methods

Benchmark Mode Cnt Score Error Units MessageCrackerBenchmark.benchmarkCrackingGenerated thrpt 5 10923194.106 ± 2861197.934 ops/s MessageCrackerBenchmark.benchmarkCrackingWithHashMap thrpt 5 69723897.515 ± 3242125.694 ops/s

After:

benchmarkCrackingGenerated - use the new MessageCracker.xls generated class benchmarkCrackingWithHashMap use custom MessageCracker class which use HashMaps to resolve methods

Benchmark Mode Cnt Score Error Units MessageCrackerBenchmark.benchmarkCrackingGenerated thrpt 5 71897498.754 ± 19147681.365 ops/s MessageCrackerBenchmark.benchmarkCrackingWithHashMap thrpt 5 72165054.010 ± 4905711.116 ops/s

To Reproduce Run my JMH benchmark code in my forked repo for QFJ

Before

Before Code Benchmark

After

After Code Benchmark

Expected behavior N/A

system information:

Additional context N/A

busy-spin commented 1 year ago

I made a PR for this , https://github.com/quickfix-j/quickfixj/pull/578