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 14:34:30 +00:00
|
|
|
/*
|
|
|
|
* Copyright 2011-2015 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.
|
|
|
|
*/
|
|
|
|
|
2015-05-21 12:40:04 +00:00
|
|
|
#include "kernel_split_common.h"
|
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 14:34:30 +00:00
|
|
|
|
2015-05-26 14:12:49 +00:00
|
|
|
/* Note on kernel_setup_next_iteration kernel.
|
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 14:34:30 +00:00
|
|
|
* This is the tenth kernel in the ray tracing logic. This is the ninth
|
|
|
|
* of the path iteration kernels. This kernel takes care of setting up
|
|
|
|
* Ray for the next iteration of path-iteration and accumulating radiance
|
|
|
|
* corresponding to AO and direct-lighting
|
|
|
|
*
|
|
|
|
* Ray state of rays that are terminated in this kernel are changed to RAY_UPDATE_BUFFER
|
|
|
|
*
|
|
|
|
* The input and output are as follows,
|
|
|
|
*
|
2015-05-21 12:40:04 +00:00
|
|
|
* rng_coop ---------------------------------------------|--- kernel_next_iteration_setup -|--- Queue_index (QUEUE_HITBG_BUFF_UPDATE_TOREGEN_RAYS)
|
|
|
|
* throughput_coop --------------------------------------| |--- Queue_data (QUEUE_HITBF_BUFF_UPDATE_TOREGEN_RAYS)
|
|
|
|
* PathRadiance_coop ------------------------------------| |--- throughput_coop
|
|
|
|
* PathState_coop ---------------------------------------| |--- PathRadiance_coop
|
2015-10-29 16:44:36 +00:00
|
|
|
* sd ---------------------------------------------------| |--- PathState_coop
|
2015-05-21 12:40:04 +00:00
|
|
|
* ray_state --------------------------------------------| |--- ray_state
|
|
|
|
* Queue_data (QUEUE_ACTIVE_AND_REGENERATD_RAYS) --------| |--- Ray_coop
|
|
|
|
* Queue_index (QUEUE_HITBG_BUFF_UPDATE_TOREGEN_RAYS) ---| |--- use_queues_flag
|
|
|
|
* Ray_coop ---------------------------------------------| |
|
2015-11-01 11:51:32 +00:00
|
|
|
* kg (globals) -----------------------------------------| |
|
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 14:34:30 +00:00
|
|
|
* LightRay_dl_coop -------------------------------------|
|
|
|
|
* ISLamp_coop ------------------------------------------|
|
|
|
|
* BSDFEval_coop ----------------------------------------|
|
|
|
|
* LightRay_ao_coop -------------------------------------|
|
|
|
|
* AOBSDF_coop ------------------------------------------|
|
|
|
|
* AOAlpha_coop -----------------------------------------|
|
|
|
|
*
|
|
|
|
* Note on queues,
|
|
|
|
* This kernel fetches rays from the queue QUEUE_ACTIVE_AND_REGENERATED_RAYS and processes only
|
|
|
|
* the rays of state RAY_ACTIVE.
|
|
|
|
* There are different points in this kernel where a ray may terminate and reach RAY_UPDATE_BUFF
|
|
|
|
* state. These rays are enqueued into QUEUE_HITBG_BUFF_UPDATE_TOREGEN_RAYS queue. These rays will
|
|
|
|
* still be present in QUEUE_ACTIVE_AND_REGENERATED_RAYS queue, but since their ray-state has been
|
|
|
|
* changed to RAY_UPDATE_BUFF, there is no problem.
|
|
|
|
*
|
|
|
|
* State of queues when this kernel is called :
|
|
|
|
* At entry,
|
|
|
|
* QUEUE_ACTIVE_AND_REGENERATED_RAYS will be filled with RAY_ACTIVE, RAY_REGENERATED, RAY_UPDATE_BUFFER rays.
|
|
|
|
* QUEUE_HITBG_BUFF_UPDATE_TOREGEN_RAYS will be filled with RAY_TO_REGENERATE and RAY_UPDATE_BUFFER rays
|
|
|
|
* At exit,
|
|
|
|
* QUEUE_ACTIVE_AND_REGENERATED_RAYS will be filled with RAY_ACTIVE, RAY_REGENERATED and more RAY_UPDATE_BUFFER rays.
|
|
|
|
* QUEUE_HITBG_BUFF_UPDATE_TOREGEN_RAYS will be filled with RAY_TO_REGENERATE and more RAY_UPDATE_BUFFER rays
|
|
|
|
*/
|
2015-05-26 14:12:49 +00:00
|
|
|
ccl_device char kernel_next_iteration_setup(
|
2015-10-29 16:44:36 +00:00
|
|
|
KernelGlobals *kg,
|
|
|
|
ShaderData *sd, /* Required for setting up ray for next iteration */
|
2015-05-26 14:12:49 +00:00
|
|
|
ccl_global uint *rng_coop, /* Required for setting up ray for next iteration */
|
|
|
|
ccl_global float3 *throughput_coop, /* Required for setting up ray for next iteration */
|
|
|
|
PathRadiance *PathRadiance_coop, /* Required for setting up ray for next iteration */
|
|
|
|
ccl_global Ray *Ray_coop, /* Required for setting up ray for next iteration */
|
|
|
|
ccl_global PathState *PathState_coop, /* Required for setting up ray for next iteration */
|
|
|
|
ccl_global Ray *LightRay_dl_coop, /* Required for radiance update - direct lighting */
|
|
|
|
ccl_global int *ISLamp_coop, /* Required for radiance update - direct lighting */
|
|
|
|
ccl_global BsdfEval *BSDFEval_coop, /* Required for radiance update - direct lighting */
|
|
|
|
ccl_global Ray *LightRay_ao_coop, /* Required for radiance update - AO */
|
|
|
|
ccl_global float3 *AOBSDF_coop, /* Required for radiance update - AO */
|
|
|
|
ccl_global float3 *AOAlpha_coop, /* Required for radiance update - AO */
|
|
|
|
ccl_global char *ray_state, /* Denotes the state of each ray */
|
|
|
|
ccl_global char *use_queues_flag, /* flag to decide if scene_intersect kernel should
|
|
|
|
* use queues to fetch ray index */
|
|
|
|
int ray_index)
|
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 14:34:30 +00:00
|
|
|
{
|
|
|
|
char enqueue_flag = 0;
|
|
|
|
|
2015-10-29 16:44:36 +00:00
|
|
|
/* Load ShaderData structure. */
|
|
|
|
PathRadiance *L = NULL;
|
|
|
|
ccl_global PathState *state = NULL;
|
2015-05-26 14:12:49 +00:00
|
|
|
|
2015-05-26 19:13:32 +00:00
|
|
|
/* Path radiance update for AO/Direct_lighting's shadow blocked. */
|
|
|
|
if(IS_FLAG(ray_state, ray_index, RAY_SHADOW_RAY_CAST_DL) ||
|
|
|
|
IS_FLAG(ray_state, ray_index, RAY_SHADOW_RAY_CAST_AO))
|
2016-02-03 14:00:55 +00:00
|
|
|
{
|
2015-05-26 14:12:49 +00:00
|
|
|
state = &PathState_coop[ray_index];
|
|
|
|
L = &PathRadiance_coop[ray_index];
|
|
|
|
float3 _throughput = throughput_coop[ray_index];
|
|
|
|
|
|
|
|
if(IS_FLAG(ray_state, ray_index, RAY_SHADOW_RAY_CAST_AO)) {
|
|
|
|
float3 shadow = LightRay_ao_coop[ray_index].P;
|
|
|
|
char update_path_radiance = LightRay_ao_coop[ray_index].t;
|
|
|
|
if(update_path_radiance) {
|
2015-05-26 19:13:32 +00:00
|
|
|
path_radiance_accum_ao(L,
|
|
|
|
_throughput,
|
|
|
|
AOAlpha_coop[ray_index],
|
|
|
|
AOBSDF_coop[ray_index],
|
|
|
|
shadow,
|
|
|
|
state->bounce);
|
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 14:34:30 +00:00
|
|
|
}
|
2015-05-26 14:12:49 +00:00
|
|
|
REMOVE_RAY_FLAG(ray_state, ray_index, RAY_SHADOW_RAY_CAST_AO);
|
|
|
|
}
|
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 14:34:30 +00:00
|
|
|
|
2015-05-26 14:12:49 +00:00
|
|
|
if(IS_FLAG(ray_state, ray_index, RAY_SHADOW_RAY_CAST_DL)) {
|
|
|
|
float3 shadow = LightRay_dl_coop[ray_index].P;
|
|
|
|
char update_path_radiance = LightRay_dl_coop[ray_index].t;
|
|
|
|
if(update_path_radiance) {
|
|
|
|
BsdfEval L_light = BSDFEval_coop[ray_index];
|
2015-05-26 19:13:32 +00:00
|
|
|
path_radiance_accum_light(L,
|
|
|
|
_throughput,
|
|
|
|
&L_light,
|
|
|
|
shadow,
|
|
|
|
1.0f,
|
|
|
|
state->bounce,
|
|
|
|
ISLamp_coop[ray_index]);
|
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 14:34:30 +00:00
|
|
|
}
|
2015-05-26 14:12:49 +00:00
|
|
|
REMOVE_RAY_FLAG(ray_state, ray_index, RAY_SHADOW_RAY_CAST_DL);
|
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 14:34:30 +00:00
|
|
|
}
|
2015-05-26 14:12:49 +00:00
|
|
|
}
|
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 14:34:30 +00:00
|
|
|
|
2015-05-26 14:12:49 +00:00
|
|
|
if(IS_STATE(ray_state, ray_index, RAY_ACTIVE)) {
|
|
|
|
ccl_global float3 *throughput = &throughput_coop[ray_index];
|
|
|
|
ccl_global Ray *ray = &Ray_coop[ray_index];
|
2016-09-19 14:04:45 +00:00
|
|
|
ccl_global RNG *rng = &rng_coop[ray_index];
|
2015-05-26 14:12:49 +00:00
|
|
|
state = &PathState_coop[ray_index];
|
|
|
|
L = &PathRadiance_coop[ray_index];
|
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 14:34:30 +00:00
|
|
|
|
2015-07-03 08:56:40 +00:00
|
|
|
/* Compute direct lighting and next bounce. */
|
2015-05-26 14:12:49 +00:00
|
|
|
if(!kernel_path_surface_bounce(kg, rng, sd, throughput, state, L, ray)) {
|
|
|
|
ASSIGN_RAY_STATE(ray_state, ray_index, RAY_UPDATE_BUFFER);
|
|
|
|
enqueue_flag = 1;
|
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 14:34:30 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-05-26 14:12:49 +00:00
|
|
|
return enqueue_flag;
|
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 14:34:30 +00:00
|
|
|
}
|