Re: [sensors] Don't allocate response objects to avoid triggering GC

@Maksims do you have any test set (data) which shows that current 
approach makes 60fps target unreachable? I quckly measured codepath 
for sensor reading update using current API and with modified codepath
 that matches proposal in this issue. The difference between two is 
neglectable: fluctuations on a level of 0.1-1 microseconds per reading
 update. Tested on Mi Pad2 device.


> Requirement 1) is extremely perf sensitive, and if our ability to 
deliver on it is hindered because we're triggering too much GC, we 
should definitely fix that.

Before changing the API, we need some data which proves that 
performance issues do exist, otherwise, it looks like a premature 

GitHub Notification of comment by alexshalamov
Please view or discuss this issue at using
 your GitHub account

Received on Friday, 9 December 2016 13:13:48 UTC