huzpsb/JavaObfuscatorTest

Lack of warmup in calc test

terminalsin opened this issue · 2 comments

terminalsin commented

Issue
Calc test is consistent due to a lack of warmup

Solution
Add a call for each method before it is iterated. Then, iterate the methods as done previously. Then repeat the class 10x times, and only then take their mean and median and measure those. This would be a much more accurate way at measuring performance respective to the obfuscator.

huzpsb commented

Hmm...
[s]Can I say this check is made alike this simply because I want to count the warm-up time in? With things like MyJ2C+VMP, the warm up time can be up to 200ms per class and I found that annoying.[/s]
Maybe I would add a seperate flag for warm-up and modify this check accoardingly. Thank you for your suggestion.

terminalsin commented

Hmm... [s]Can I say this check is made alike this simply because I want to count the warm-up time in? With things like MyJ2C+VMP, the warm up time can be up to 200ms per class and I found that annoying.[/s] Maybe I would add a seperate flag for warm-up and modify this check accoardingly. Thank you for your suggestion.

Yeah but once executed and cached, JIT assists and makes execution flawless. Performance should measuring consistent calls in an application, not one time loading schemes.