+- Cuberite Forum (https://forum.cuberite.org)
+-- Forum: Cuberite (https://forum.cuberite.org/forum-4.html)
+--- Forum: Development (https://forum.cuberite.org/forum-13.html)
+--- Thread: r245 question (/thread-336.html)
r245 question - tbar - 02-10-2012
In revision 245 the log message states:
Quote:Author: madmaxoft: MTRand class is not created in each tick, therefore much improving tick-thread time (now uses ~5 % CPU instead of one full core)
This sounds like a massive improvement but is it? How should I interpret the message above?
RE: r245 question - xoft - 02-10-2012
I'ts not really massive, but still is quite a lot
On my QuadCore, the tick thread was hogging one core all the time (-> 25 % total CPU) before the change. After the change, the thread is using only about 5 % of the total CPU. Both values were measured with 1 player on the server, 17 chunks viewdistance.
This means that it now runs at least 5 times faster, maybe even more, because a thread can't run faster than one-core-at-a-time.
Note, however, that this was a debug build. In a release build this needn't be so noticable.
RE: r245 question - FakeTruth - 02-10-2012
It was even worse on Linux before, because the Linux version of MTRand reads a file each time it's created
RE: r245 question - tbar - 02-10-2012
That _does_ sound massive
