Channels
#announcements
- m
michaelrocks
11/26/2015, 6:23 PMI have to check if null checks really affect performance and it’s not a profiler mistake👍 1 - c
cy
11/26/2015, 6:25 PMthese checks runs very quickly so profiler could be unable to measure such short times correctly, at last you need to configure properly for that - m
michaelrocks
11/26/2015, 6:29 PM@cy: True, but I don’t have many configuration options in Android profiler simple smile - a
apatrida
11/26/2015, 6:41 PMand you have an actual performance problem that this is the last thing on the list that can help it? - a
apatrida
11/26/2015, 6:41 PMI can’t imagine null checks causing an issue unless you have a tight loop doing it 1.000.000 times - m
michaelrocks
11/26/2015, 6:42 PMI have a performance problem where I can make some improvements that will save me a few percent of CPU time according to the profiler. So 5% on just null checks is a lot for me. - c
cedric
11/26/2015, 6:43 PMCan you post a profiler dump? I’m curious to look at it - m
michaelrocks
11/26/2015, 6:47 PMI have no idea how to dump the results - m
michaelrocks
11/26/2015, 6:49 PMOk, I’ll try - a
Andreas Sinz
11/26/2015, 6:52 PMhey guys! i've got a gradle project which i want to compile in different versions. something like a production release with logging enabled, another version for testing with logging enabled with a different URL. what would be an easy way to achieve that? - y
ylemoigne
11/26/2015, 6:54 PMIn gradle
and on command line gradle -Pmyswitch taskif project.hasAttribute(“myswitch”) - y
ylemoigne
11/26/2015, 6:54 PMsomething like that - m
michaelrocks
11/26/2015, 7:06 PMWhen doing trace-based profiling null checks are just 0.5% - c
cy
11/26/2015, 7:07 PMit is most likely correct - m
michaelrocks
11/26/2015, 7:13 PMAnyway I’ve already removed them with ProGuard simple smile - c
cy
11/26/2015, 7:14 PManyway it is bad idea - m
michaelrocks
11/26/2015, 7:14 PMWhy isn’t this method inline? - m
michaelrocks
11/26/2015, 7:14 PMAgree, not a good idea - c
cy
11/26/2015, 7:14 PMbecause in many cases it could show real issues at Kotlin <-> Java - m
michaelrocks
11/26/2015, 7:15 PMBut why isn’t it inlined instead of invokestatic? - c
cy
11/26/2015, 7:16 PMbecause there is no benefit in fact - c
cy
11/26/2015, 7:16 PMJVM will inline it anyway - m
michaelrocks
11/26/2015, 7:16 PMAndroid VM doesn’t as far as I know - m
michaelrocks
11/26/2015, 7:16 PMMaybe ART does, I’m not sure - m
michaelrocks
11/26/2015, 7:18 PMNo method inlining according to this answer http://stackoverflow.com/a/4930538/170842 - m
michaelrocks
11/26/2015, 7:18 PMBut it’s outdated - c
cy
11/26/2015, 7:18 PMyes, I read the same right now simple smile - m
michaelrocks
11/26/2015, 7:19 PMRedundant null check elimination (if such redundancy can be found in a basic block).
- m
michaelrocks
11/26/2015, 7:19 PMI think Android would benefit from inlining null checks - c
cy
11/26/2015, 7:19 PMthis depends on "basic block" definition...