forked from bartvdbraak/blender
7f4479da42
This commit contains all the work related on the AMD megakernel split work which was mainly done by Varun Sundar, George Kyriazis and Lenny Wang, plus some help from Sergey Sharybin, Martijn Berger, Thomas Dinges and likely someone else which we're forgetting to mention. Currently only AMD cards are enabled for the new split kernel, but it is possible to force split opencl kernel to be used by setting the following environment variable: CYCLES_OPENCL_SPLIT_KERNEL_TEST=1. Not all the features are supported yet, and that being said no motion blur, camera blur, SSS and volumetrics for now. Also transparent shadows are disabled on AMD device because of some compiler bug. This kernel is also only implements regular path tracing and supporting branched one will take a bit. Branched path tracing is exposed to the interface still, which is a bit misleading and will be hidden there soon. More feature will be enabled once they're ported to the split kernel and tested. Neither regular CPU nor CUDA has any difference, they're generating the same exact code, which means no regressions/improvements there. Based on the research paper: https://research.nvidia.com/sites/default/files/publications/laine2013hpg_paper.pdf Here's the documentation: https://docs.google.com/document/d/1LuXW-CV-sVJkQaEGZlMJ86jZ8FmoPfecaMdR-oiWbUY/edit Design discussion of the patch: https://developer.blender.org/T44197 Differential Revision: https://developer.blender.org/D1200
129 lines
3.4 KiB
C
129 lines
3.4 KiB
C
/*
|
|
* Copyright 2011-2013 Blender Foundation
|
|
*
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
|
|
/* Constant Globals */
|
|
|
|
CCL_NAMESPACE_BEGIN
|
|
|
|
/* On the CPU, we pass along the struct KernelGlobals to nearly everywhere in
|
|
* the kernel, to access constant data. These are all stored as "textures", but
|
|
* these are really just standard arrays. We can't use actually globals because
|
|
* multiple renders may be running inside the same process. */
|
|
|
|
#ifdef __KERNEL_CPU__
|
|
|
|
#ifdef __OSL__
|
|
struct OSLGlobals;
|
|
struct OSLThreadData;
|
|
struct OSLShadingSystem;
|
|
#endif
|
|
|
|
#define MAX_BYTE_IMAGES 1024
|
|
#define MAX_FLOAT_IMAGES 1024
|
|
|
|
typedef struct KernelGlobals {
|
|
texture_image_uchar4 texture_byte_images[MAX_BYTE_IMAGES];
|
|
texture_image_float4 texture_float_images[MAX_FLOAT_IMAGES];
|
|
|
|
#define KERNEL_TEX(type, ttype, name) ttype name;
|
|
#define KERNEL_IMAGE_TEX(type, ttype, name)
|
|
#include "kernel_textures.h"
|
|
|
|
KernelData __data;
|
|
|
|
#ifdef __OSL__
|
|
/* On the CPU, we also have the OSL globals here. Most data structures are shared
|
|
* with SVM, the difference is in the shaders and object/mesh attributes. */
|
|
OSLGlobals *osl;
|
|
OSLShadingSystem *osl_ss;
|
|
OSLThreadData *osl_tdata;
|
|
#endif
|
|
|
|
} KernelGlobals;
|
|
|
|
#endif
|
|
|
|
/* For CUDA, constant memory textures must be globals, so we can't put them
|
|
* into a struct. As a result we don't actually use this struct and use actual
|
|
* globals and simply pass along a NULL pointer everywhere, which we hope gets
|
|
* optimized out. */
|
|
|
|
#ifdef __KERNEL_CUDA__
|
|
|
|
__constant__ KernelData __data;
|
|
typedef struct KernelGlobals {} KernelGlobals;
|
|
|
|
#ifdef __KERNEL_CUDA_TEX_STORAGE__
|
|
#define KERNEL_TEX(type, ttype, name) ttype name;
|
|
#else
|
|
#define KERNEL_TEX(type, ttype, name) const __constant__ __device__ type *name;
|
|
#endif
|
|
#define KERNEL_IMAGE_TEX(type, ttype, name) ttype name;
|
|
#include "kernel_textures.h"
|
|
|
|
#endif
|
|
|
|
/* OpenCL */
|
|
|
|
#ifdef __KERNEL_OPENCL__
|
|
|
|
typedef ccl_addr_space struct KernelGlobals {
|
|
ccl_constant KernelData *data;
|
|
|
|
#define KERNEL_TEX(type, ttype, name) \
|
|
ccl_global type *name;
|
|
#include "kernel_textures.h"
|
|
} KernelGlobals;
|
|
|
|
#endif
|
|
|
|
/* Interpolated lookup table access */
|
|
|
|
ccl_device float lookup_table_read(KernelGlobals *kg, float x, int offset, int size)
|
|
{
|
|
x = saturate(x)*(size-1);
|
|
|
|
int index = min(float_to_int(x), size-1);
|
|
int nindex = min(index+1, size-1);
|
|
float t = x - index;
|
|
|
|
float data0 = kernel_tex_fetch(__lookup_table, index + offset);
|
|
if(t == 0.0f)
|
|
return data0;
|
|
|
|
float data1 = kernel_tex_fetch(__lookup_table, nindex + offset);
|
|
return (1.0f - t)*data0 + t*data1;
|
|
}
|
|
|
|
ccl_device float lookup_table_read_2D(KernelGlobals *kg, float x, float y, int offset, int xsize, int ysize)
|
|
{
|
|
y = saturate(y)*(ysize-1);
|
|
|
|
int index = min(float_to_int(y), ysize-1);
|
|
int nindex = min(index+1, ysize-1);
|
|
float t = y - index;
|
|
|
|
float data0 = lookup_table_read(kg, x, offset + xsize*index, xsize);
|
|
if(t == 0.0f)
|
|
return data0;
|
|
|
|
float data1 = lookup_table_read(kg, x, offset + xsize*nindex, xsize);
|
|
return (1.0f - t)*data0 + t*data1;
|
|
}
|
|
|
|
CCL_NAMESPACE_END
|
|
|