Blender Git Loki

Use OpenCLCache for megakernel

Change class hierarchy in device_opencl.cpp and refactor

Class hierarchy change :
The new class hierarchy is as follows,
OpenCLDeviceBase (inherits Device) - Contains error-handlers, texture functiions and other
common functions between MegaKernel and SplitKernel
OpenCLDeviceSplitKernel (inherits OpenCLDeviceBase) - Contains variables and functions specific to
split kernel implementation
OpenCLDeviceMegaKernel (inherits OpenCLDeviceBase) - Contains variables and functions specific to
megakernel
Refactor :
Change "if " to "if"

Support__KERNEL_DEBUG__ flag

Dont use OpenCLCache with split kernel

Remove SPLIT_KERNEL_CLOSURE_COUNT macro

Refactor : Change variable name

Refactor : Add/Remove comments/spaces

Bug fix : Don't display sample count in progress bar when using split kernel

Refactor(cosmetics) : Modify ShaderData declaration

Bypass selective-compilation-of-svm-nodes for viewport

By-passing "selective-compilation-of-svm-nodes" optimization
will limit kernel re-compilations during viewport render

Remove un-used subsurface kernel

Refactor : Remove un-necessary code/comment

Remove __KERNEL_OPENCL__ definition in device_opencl.cpp

Restore original state of kernel_types.h

Prevent un-necessary megakernel compilation inside splitkernel

Bug fix : Remove __SPLIT_KERNEL__ build option for kernel.cl in OpenCLDeviceSplitKernel

Refactor : Move some build option setting into load_kernels for split_kernel

Refactor : Remove un-necessary SPLIT_KERNEL macros

Remove un-necessary code from OpenCLDevice class

Add OpenCLDeviceSplitKernel class