Cycles: Remove round-up trickery for max closure in split OpenCL kernel
authorSergey Sharybin <sergey.vfx@gmail.com>
Sat, 6 Jun 2015 11:14:31 +0000 (13:14 +0200)
committerSergey Sharybin <sergey.vfx@gmail.com>
Mon, 8 Jun 2015 09:15:39 +0000 (11:15 +0200)
commit8c2750bc82533945022fe413844b8868dbdcd5e8
tree2ace99e4f2c52ba17bafbca9fa77897a9310f86f
parentc6c06285a7a248621134120a8a64b31d8b5f4f5d
Cycles: Remove round-up trickery for max closure in split OpenCL kernel

Round-up was only enabled for viewport render, which was for a long time hardcoded to
use 64 closures. This was done in order to avoid unnecessary kernel re-compilations
when tweaking the shader tree.

We could enable selective closure compilation in the viewport later if it'll give
measurable speed improvements, but even then round-up is to happen outside of the
device level,

This commit also removes early output which happened in cases when max closure did
not change. It was wrong because other requested kernel features might have been
changed.
intern/cycles/device/device_opencl.cpp