sandey/blender
1
0
Fork 0
forked from bartvdbraak/blender
Code Pull Requests Activity
5c34266383
blender/intern/cycles/kernel/svm/svm.h
George Kyriazis 7f4479da42 Cycles: OpenCL kernel split 
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
2015-05-09 19:52:40 +05:00

452 lines
13 KiB
C
Raw Blame History

/*
* 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.
*/
#ifndef __SVM_H__
#define __SVM_H__
/* Shader Virtual Machine
*
* A shader is a list of nodes to be executed. These are simply read one after
* the other and executed, using an node counter. Each node and it's associated
* data is encoded as one or more uint4's in a 1D texture. If the data is larger
* than an uint4, the node can increase the node counter to compensate for this.
* Floats are encoded as int and then converted to float again.
*
* Nodes write their output into a stack. All stack data in the stack is
* floats, since it's all factors, colors and vectors. The stack will be stored
* in local memory on the GPU, as it would take too many register and indexes in
* ways not known at compile time. This seems the only solution even though it
* may be slow, with two positive factors. If the same shader is being executed,
* memory access will be coalesced, and on fermi cards, memory will actually be
* cached.
*
* The result of shader execution will be a single closure. This means the
* closure type, associated label, data and weight. Sampling from multiple
* closures is supported through the mix closure node, the logic for that is
* mostly taken care of in the SVM compiler.
*/
#include "svm_types.h"
CCL_NAMESPACE_BEGIN
/* Stack */
ccl_device_inline float3 stack_load_float3(float *stack, uint a)
{
kernel_assert(a+2 < SVM_STACK_SIZE);
return make_float3(stack[a+0], stack[a+1], stack[a+2]);
}
ccl_device_inline void stack_store_float3(float *stack, uint a, float3 f)
{
kernel_assert(a+2 < SVM_STACK_SIZE);
stack[a+0] = f.x;
stack[a+1] = f.y;
stack[a+2] = f.z;
}
ccl_device_inline float stack_load_float(float *stack, uint a)
{
kernel_assert(a < SVM_STACK_SIZE);
return stack[a];
}
ccl_device_inline float stack_load_float_default(float *stack, uint a, uint value)
{
return (a == (uint)SVM_STACK_INVALID)? __uint_as_float(value): stack_load_float(stack, a);
}
ccl_device_inline void stack_store_float(float *stack, uint a, float f)
{
kernel_assert(a < SVM_STACK_SIZE);
stack[a] = f;
}
ccl_device_inline int stack_load_int(float *stack, uint a)
{
kernel_assert(a < SVM_STACK_SIZE);
return __float_as_int(stack[a]);
}
ccl_device_inline int stack_load_int_default(float *stack, uint a, uint value)
{
return (a == (uint)SVM_STACK_INVALID)? (int)value: stack_load_int(stack, a);
}
ccl_device_inline void stack_store_int(float *stack, uint a, int i)
{
kernel_assert(a < SVM_STACK_SIZE);
stack[a] = __int_as_float(i);
}
ccl_device_inline bool stack_valid(uint a)
{
return a != (uint)SVM_STACK_INVALID;
}
/* Reading Nodes */
ccl_device_inline uint4 read_node(KernelGlobals *kg, int *offset)
{
uint4 node = kernel_tex_fetch(__svm_nodes, *offset);
(*offset)++;
return node;
}
ccl_device_inline float4 read_node_float(KernelGlobals *kg, int *offset)
{
uint4 node = kernel_tex_fetch(__svm_nodes, *offset);
float4 f = make_float4(__uint_as_float(node.x), __uint_as_float(node.y), __uint_as_float(node.z), __uint_as_float(node.w));
(*offset)++;
return f;
}
ccl_device_inline float4 fetch_node_float(KernelGlobals *kg, int offset)
{
uint4 node = kernel_tex_fetch(__svm_nodes, offset);
return make_float4(__uint_as_float(node.x), __uint_as_float(node.y), __uint_as_float(node.z), __uint_as_float(node.w));
}
ccl_device_inline void decode_node_uchar4(uint i, uint *x, uint *y, uint *z, uint *w)
{
if(x) *x = (i & 0xFF);
if(y) *y = ((i >> 8) & 0xFF);
if(z) *z = ((i >> 16) & 0xFF);
if(w) *w = ((i >> 24) & 0xFF);
}
CCL_NAMESPACE_END
/* Nodes */
#include "svm_noise.h"
#include "svm_texture.h"
#include "svm_math_util.h"
#include "svm_attribute.h"
#include "svm_gradient.h"
#include "svm_blackbody.h"
#include "svm_closure.h"
#include "svm_noisetex.h"
#include "svm_convert.h"
#include "svm_displace.h"
#include "svm_fresnel.h"
#include "svm_wireframe.h"
#include "svm_wavelength.h"
#include "svm_camera.h"
#include "svm_geometry.h"
#include "svm_hsv.h"
#include "svm_image.h"
#include "svm_gamma.h"
#include "svm_brightness.h"
#include "svm_invert.h"
#include "svm_light_path.h"
#include "svm_magic.h"
#include "svm_mapping.h"
#include "svm_normal.h"
#include "svm_wave.h"
#include "svm_math.h"
#include "svm_mix.h"
#include "svm_ramp.h"
#include "svm_sepcomb_hsv.h"
#include "svm_sepcomb_vector.h"
#include "svm_musgrave.h"
#include "svm_sky.h"
#include "svm_tex_coord.h"
#include "svm_value.h"
#include "svm_voronoi.h"
#include "svm_checker.h"
#include "svm_brick.h"
#include "svm_vector_transform.h"
CCL_NAMESPACE_BEGIN
#define NODES_GROUP(group) ((group) <= __NODES_MAX_GROUP__)
#define NODES_FEATURE(feature) (__NODES_FEATURES__ & (feature) != 0)
/* Main Interpreter Loop */
ccl_device_noinline void svm_eval_nodes(KernelGlobals *kg, ShaderData *sd, ShaderType type, int path_flag)
{
float stack[SVM_STACK_SIZE];
int offset = ccl_fetch(sd, shader) & SHADER_MASK;
while(1) {
uint4 node = read_node(kg, &offset);
switch(node.x) {
#if NODES_GROUP(NODE_GROUP_LEVEL_0)
case NODE_SHADER_JUMP: {
if(type == SHADER_TYPE_SURFACE) offset = node.y;
else if(type == SHADER_TYPE_VOLUME) offset = node.z;
else if(type == SHADER_TYPE_DISPLACEMENT) offset = node.w;
else return;
break;
}
case NODE_CLOSURE_BSDF:
svm_node_closure_bsdf(kg, sd, stack, node, path_flag, &offset);
break;
case NODE_CLOSURE_EMISSION:
svm_node_closure_emission(sd, stack, node);
break;
case NODE_CLOSURE_BACKGROUND:
svm_node_closure_background(sd, stack, node);
break;
case NODE_CLOSURE_HOLDOUT:
svm_node_closure_holdout(sd, stack, node);
break;
case NODE_CLOSURE_AMBIENT_OCCLUSION:
svm_node_closure_ambient_occlusion(sd, stack, node);
break;
#if NODES_FEATURE(NODE_FEATURE_VOLUME)
case NODE_CLOSURE_VOLUME:
svm_node_closure_volume(kg, sd, stack, node, path_flag);
break;
#endif /* NODES_FEATURE(NODE_FEATURE_VOLUME) */
case NODE_CLOSURE_SET_WEIGHT:
svm_node_closure_set_weight(sd, node.y, node.z, node.w);
break;
case NODE_CLOSURE_WEIGHT:
svm_node_closure_weight(sd, stack, node.y);
break;
case NODE_EMISSION_WEIGHT:
svm_node_emission_weight(kg, sd, stack, node);
break;
case NODE_MIX_CLOSURE:
svm_node_mix_closure(sd, stack, node);
break;
#endif /* NODES_GROUP(NODE_GROUP_LEVEL_0) */
case NODE_JUMP_IF_ZERO:
if(stack_load_float(stack, node.z) == 0.0f)
offset += node.y;
break;
case NODE_JUMP_IF_ONE:
if(stack_load_float(stack, node.z) == 1.0f)
offset += node.y;
break;
#ifdef __TEXTURES__
case NODE_TEX_IMAGE:
svm_node_tex_image(kg, sd, stack, node);
break;
case NODE_TEX_IMAGE_BOX:
svm_node_tex_image_box(kg, sd, stack, node);
break;
case NODE_TEX_ENVIRONMENT:
svm_node_tex_environment(kg, sd, stack, node);
break;
case NODE_TEX_SKY:
svm_node_tex_sky(kg, sd, stack, node, &offset);
break;
case NODE_TEX_GRADIENT:
svm_node_tex_gradient(sd, stack, node);
break;
case NODE_TEX_NOISE:
svm_node_tex_noise(kg, sd, stack, node, &offset);
break;
case NODE_TEX_VORONOI:
svm_node_tex_voronoi(kg, sd, stack, node, &offset);
break;
case NODE_TEX_MUSGRAVE:
svm_node_tex_musgrave(kg, sd, stack, node, &offset);
break;
case NODE_TEX_WAVE:
svm_node_tex_wave(kg, sd, stack, node, &offset);
break;
case NODE_TEX_MAGIC:
svm_node_tex_magic(kg, sd, stack, node, &offset);
break;
case NODE_TEX_CHECKER:
svm_node_tex_checker(kg, sd, stack, node);
break;
case NODE_TEX_BRICK:
svm_node_tex_brick(kg, sd, stack, node, &offset);
break;
#endif /* __TEXTURES__ */
case NODE_CAMERA:
svm_node_camera(kg, sd, stack, node.y, node.z, node.w);
break;
case NODE_GEOMETRY:
svm_node_geometry(kg, sd, stack, node.y, node.z);
break;
#ifdef __EXTRA_NODES__
case NODE_GEOMETRY_BUMP_DX:
svm_node_geometry_bump_dx(kg, sd, stack, node.y, node.z);
break;
case NODE_GEOMETRY_BUMP_DY:
svm_node_geometry_bump_dy(kg, sd, stack, node.y, node.z);
break;
case NODE_LIGHT_PATH:
svm_node_light_path(sd, stack, node.y, node.z, path_flag);
break;
case NODE_OBJECT_INFO:
svm_node_object_info(kg, sd, stack, node.y, node.z);
break;
case NODE_PARTICLE_INFO:
svm_node_particle_info(kg, sd, stack, node.y, node.z);
break;
#ifdef __HAIR__
# if NODES_FEATURE(NODE_FEATURE_HAIR)
case NODE_HAIR_INFO:
svm_node_hair_info(kg, sd, stack, node.y, node.z);
break;
# endif /* NODES_FEATURE(NODE_FEATURE_HAIR) */
#endif /* __HAIR__ */
#endif /* __EXTRA_NODES__ */
case NODE_CONVERT:
svm_node_convert(sd, stack, node.y, node.z, node.w);
break;
case NODE_VALUE_F:
svm_node_value_f(kg, sd, stack, node.y, node.z);
break;
case NODE_VALUE_V:
svm_node_value_v(kg, sd, stack, node.y, &offset);
break;
#ifdef __EXTRA_NODES__
case NODE_INVERT:
svm_node_invert(sd, stack, node.y, node.z, node.w);
break;
case NODE_GAMMA:
svm_node_gamma(sd, stack, node.y, node.z, node.w);
break;
case NODE_BRIGHTCONTRAST:
svm_node_brightness(sd, stack, node.y, node.z, node.w);
break;
case NODE_MIX:
svm_node_mix(kg, sd, stack, node.y, node.z, node.w, &offset);
break;
case NODE_SEPARATE_VECTOR:
svm_node_separate_vector(sd, stack, node.y, node.z, node.w);
break;
case NODE_COMBINE_VECTOR:
svm_node_combine_vector(sd, stack, node.y, node.z, node.w);
break;
case NODE_SEPARATE_HSV:
svm_node_separate_hsv(kg, sd, stack, node.y, node.z, node.w, &offset);
break;
case NODE_COMBINE_HSV:
svm_node_combine_hsv(kg, sd, stack, node.y, node.z, node.w, &offset);
break;
case NODE_HSV:
svm_node_hsv(kg, sd, stack, node.y, node.z, node.w, &offset);
break;
#endif /* __EXTRA_NODES__ */
case NODE_ATTR:
svm_node_attr(kg, sd, stack, node);
break;
#ifdef __EXTRA_NODES__
case NODE_ATTR_BUMP_DX:
svm_node_attr_bump_dx(kg, sd, stack, node);
break;
case NODE_ATTR_BUMP_DY:
svm_node_attr_bump_dy(kg, sd, stack, node);
break;
#endif /* __EXTRA_NODES__ */
case NODE_FRESNEL:
svm_node_fresnel(sd, stack, node.y, node.z, node.w);
break;
case NODE_LAYER_WEIGHT:
svm_node_layer_weight(sd, stack, node);
break;
#ifdef __EXTRA_NODES__
case NODE_WIREFRAME:
svm_node_wireframe(kg, sd, stack, node);
break;
case NODE_WAVELENGTH:
svm_node_wavelength(sd, stack, node.y, node.z);
break;
case NODE_BLACKBODY:
svm_node_blackbody(kg, sd, stack, node.y, node.z);
break;
case NODE_SET_DISPLACEMENT:
svm_node_set_displacement(sd, stack, node.y);
break;
case NODE_SET_BUMP:
svm_node_set_bump(kg, sd, stack, node);
break;
case NODE_MATH:
svm_node_math(kg, sd, stack, node.y, node.z, node.w, &offset);
break;
case NODE_VECTOR_MATH:
svm_node_vector_math(kg, sd, stack, node.y, node.z, node.w, &offset);
break;
case NODE_VECTOR_TRANSFORM:
svm_node_vector_transform(kg, sd, stack, node);
break;
case NODE_NORMAL:
svm_node_normal(kg, sd, stack, node.y, node.z, node.w, &offset);
break;
#endif /* __EXTRA_NODES__ */
case NODE_MAPPING:
svm_node_mapping(kg, sd, stack, node.y, node.z, &offset);
break;
case NODE_MIN_MAX:
svm_node_min_max(kg, sd, stack, node.y, node.z, &offset);
break;
case NODE_TEX_COORD:
svm_node_tex_coord(kg, sd, path_flag, stack, node, &offset);
break;
#ifdef __EXTRA_NODES__
case NODE_TEX_COORD_BUMP_DX:
svm_node_tex_coord_bump_dx(kg, sd, path_flag, stack, node, &offset);
break;
case NODE_TEX_COORD_BUMP_DY:
svm_node_tex_coord_bump_dy(kg, sd, path_flag, stack, node, &offset);
break;
case NODE_CLOSURE_SET_NORMAL:
svm_node_set_normal(kg, sd, stack, node.y, node.z );
break;
case NODE_RGB_RAMP:
svm_node_rgb_ramp(kg, sd, stack, node, &offset);
break;
case NODE_RGB_CURVES:
svm_node_rgb_curves(kg, sd, stack, node, &offset);
break;
case NODE_VECTOR_CURVES:
svm_node_vector_curves(kg, sd, stack, node, &offset);
break;
case NODE_LIGHT_FALLOFF:
svm_node_light_falloff(sd, stack, node);
break;
#endif /* __EXTRA_NODES__ */
case NODE_TANGENT:
svm_node_tangent(kg, sd, stack, node);
break;
case NODE_NORMAL_MAP:
svm_node_normal_map(kg, sd, stack, node);
break;
case NODE_END:
default:
return;
}
}
}
#undef NODES_GROUP
#undef NODES_FEATURE
CCL_NAMESPACE_END
#endif /* __SVM_H__ */
View Git Blame Copy Permalink