Here's a quick way to test if the lack of entropy (whatever the available value is) produces lag or not, WITHOUT having a process running each second - to increase entropy or to measure it - which some might argue might affect the CPU throttling:
- boot normally without any /dev/random modification active
- use if for a while to "let it settle" and get the feeling of the speed
- in a terminal or adb shell, temporarily point the /dev/random device to /dev/urandom, which doesn't block at all:
- use it from then on for a while and see if it makes any real differenceCode:su mv /dev/random /dev/random_ori; ln -s /dev/urandom /dev/random
- after testing, simply reboot the device to have everything as before or run this in the terminal:
Code:su rm /dev/random; mv /dev/random_ori /dev/random
With the symbolic link there will be no active code feeding entropy into the "random" device and all consumers will be reading from "urandom" instead without any lag.
I don't know whether the code using random data will validate if it's random enough and potentially spending more time calculating random stuff until it's good enough for e.g. SSL, though, so you'll need to interpret the outcome with a grain of salt...
Tested on HTC Sensation / CM10 (Bruce's excellent rom, JB 4.1.2), it does seem to have some effect on Trebuchet's micro-lag. This Rom is already very responsive, so that's not earth shattering, but there seems to be a perceptual change. Placebo ? There's been reports of improved gmaps rendering though, i don't see any difference, as expected somewhat.