Files
blender/intern/cycles/util/util_task.cpp

291 lines
5.5 KiB
C++
Raw Normal View History

Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
/*
* Copyright 2011, Blender Foundation.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*/
#include "util_debug.h"
#include "util_foreach.h"
#include "util_system.h"
#include "util_task.h"
CCL_NAMESPACE_BEGIN
/* Task Pool */
TaskPool::TaskPool()
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
{
num = 0;
do_cancel = false;
}
TaskPool::~TaskPool()
{
stop();
}
void TaskPool::push(Task *task, bool front)
{
Performance fix for Cycles: Don't wait in the main UI thread when resetting devices. When the scene is updated Cycles resets the renderer device, cancelling all existing tasks. The main thread would wait for all running tasks to finish before continuing. This is ok when tasks can actually cancel in a timely fashion. For OSL however, this does not work, since the OSL shader group optimization takes quite a bit of time and can not be easily be cancelled once running (on my crappy machine in full debug mode: ~0.12 seconds for simple node trees). This would lead to very laggy UI behavior and make it difficult to accurately control elements such as sliders. This patch removes the wait condition from the device->task_cancel method. Instead it just sets the do_cancel flag and returns. To avoid backlog in the task pool of the device it will return early from the BlenderSession::sync function while the reset is going on (tested in Session::resetting). Once all existing tasks have finished the do_cancel flag is finally cleared again (checked in TaskPool::num_decrease). Care has to be taken to avoid race conditions on the do_cancel flag, since it can now be modified outside the TaskPool::cancel function itself. For this purpose the scope of the TaskPool::num_mutex locks has been extended, in most cases the mutex is now locked by the TaskPool itself before calling TaskScheduler methods, instead of only locking inside the num_increase/num_decrease functions themselves. The only occurrence of a lock outside of the TaskPool methods is in TaskScheduler::thread_run. This patch is most useful in combination with the OSL renderer mode, so it can probably wait until after the 2.64 release. SVM tasks tend to be cancelled quickly, so the effect is less noticeable.
2012-09-11 11:41:51 +00:00
thread_scoped_lock num_lock(num_mutex);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
TaskScheduler::Entry entry;
entry.task = task;
entry.pool = this;
TaskScheduler::push(entry, front);
}
void TaskPool::push(const TaskRunFunction& run, bool front)
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
{
push(new Task(run), front);
}
void TaskPool::wait_work()
{
thread_scoped_lock num_lock(num_mutex);
while(num != 0) {
num_lock.unlock();
thread_scoped_lock queue_lock(TaskScheduler::queue_mutex);
/* find task from this pool. if we get a task from another pool,
* we can get into deadlock */
TaskScheduler::Entry work_entry;
bool found_entry = false;
list<TaskScheduler::Entry>::iterator it;
for(it = TaskScheduler::queue.begin(); it != TaskScheduler::queue.end(); it++) {
TaskScheduler::Entry& entry = *it;
if(entry.pool == this) {
work_entry = entry;
found_entry = true;
TaskScheduler::queue.erase(it);
break;
}
}
queue_lock.unlock();
/* if found task, do it, otherwise wait until other tasks are done */
if(found_entry) {
/* run task */
work_entry.task->run();
/* delete task */
delete work_entry.task;
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
/* notify pool task was done */
num_decrease(1);
}
num_lock.lock();
if(num == 0)
break;
if(!found_entry)
num_cond.wait(num_lock);
}
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
void TaskPool::cancel()
{
Performance fix for Cycles: Don't wait in the main UI thread when resetting devices. When the scene is updated Cycles resets the renderer device, cancelling all existing tasks. The main thread would wait for all running tasks to finish before continuing. This is ok when tasks can actually cancel in a timely fashion. For OSL however, this does not work, since the OSL shader group optimization takes quite a bit of time and can not be easily be cancelled once running (on my crappy machine in full debug mode: ~0.12 seconds for simple node trees). This would lead to very laggy UI behavior and make it difficult to accurately control elements such as sliders. This patch removes the wait condition from the device->task_cancel method. Instead it just sets the do_cancel flag and returns. To avoid backlog in the task pool of the device it will return early from the BlenderSession::sync function while the reset is going on (tested in Session::resetting). Once all existing tasks have finished the do_cancel flag is finally cleared again (checked in TaskPool::num_decrease). Care has to be taken to avoid race conditions on the do_cancel flag, since it can now be modified outside the TaskPool::cancel function itself. For this purpose the scope of the TaskPool::num_mutex locks has been extended, in most cases the mutex is now locked by the TaskPool itself before calling TaskScheduler methods, instead of only locking inside the num_increase/num_decrease functions themselves. The only occurrence of a lock outside of the TaskPool methods is in TaskScheduler::thread_run. This patch is most useful in combination with the OSL renderer mode, so it can probably wait until after the 2.64 release. SVM tasks tend to be cancelled quickly, so the effect is less noticeable.
2012-09-11 11:41:51 +00:00
thread_scoped_lock num_lock(num_mutex);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
do_cancel = true;
TaskScheduler::clear(this);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
void TaskPool::stop()
{
Performance fix for Cycles: Don't wait in the main UI thread when resetting devices. When the scene is updated Cycles resets the renderer device, cancelling all existing tasks. The main thread would wait for all running tasks to finish before continuing. This is ok when tasks can actually cancel in a timely fashion. For OSL however, this does not work, since the OSL shader group optimization takes quite a bit of time and can not be easily be cancelled once running (on my crappy machine in full debug mode: ~0.12 seconds for simple node trees). This would lead to very laggy UI behavior and make it difficult to accurately control elements such as sliders. This patch removes the wait condition from the device->task_cancel method. Instead it just sets the do_cancel flag and returns. To avoid backlog in the task pool of the device it will return early from the BlenderSession::sync function while the reset is going on (tested in Session::resetting). Once all existing tasks have finished the do_cancel flag is finally cleared again (checked in TaskPool::num_decrease). Care has to be taken to avoid race conditions on the do_cancel flag, since it can now be modified outside the TaskPool::cancel function itself. For this purpose the scope of the TaskPool::num_mutex locks has been extended, in most cases the mutex is now locked by the TaskPool itself before calling TaskScheduler methods, instead of only locking inside the num_increase/num_decrease functions themselves. The only occurrence of a lock outside of the TaskPool methods is in TaskScheduler::thread_run. This patch is most useful in combination with the OSL renderer mode, so it can probably wait until after the 2.64 release. SVM tasks tend to be cancelled quickly, so the effect is less noticeable.
2012-09-11 11:41:51 +00:00
thread_scoped_lock num_lock(num_mutex);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
TaskScheduler::clear(this);
assert(num == 0);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
bool TaskPool::cancelled()
{
return do_cancel;
}
void TaskPool::num_decrease(int done)
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
{
num -= done;
assert(num >= 0);
Performance fix for Cycles: Don't wait in the main UI thread when resetting devices. When the scene is updated Cycles resets the renderer device, cancelling all existing tasks. The main thread would wait for all running tasks to finish before continuing. This is ok when tasks can actually cancel in a timely fashion. For OSL however, this does not work, since the OSL shader group optimization takes quite a bit of time and can not be easily be cancelled once running (on my crappy machine in full debug mode: ~0.12 seconds for simple node trees). This would lead to very laggy UI behavior and make it difficult to accurately control elements such as sliders. This patch removes the wait condition from the device->task_cancel method. Instead it just sets the do_cancel flag and returns. To avoid backlog in the task pool of the device it will return early from the BlenderSession::sync function while the reset is going on (tested in Session::resetting). Once all existing tasks have finished the do_cancel flag is finally cleared again (checked in TaskPool::num_decrease). Care has to be taken to avoid race conditions on the do_cancel flag, since it can now be modified outside the TaskPool::cancel function itself. For this purpose the scope of the TaskPool::num_mutex locks has been extended, in most cases the mutex is now locked by the TaskPool itself before calling TaskScheduler methods, instead of only locking inside the num_increase/num_decrease functions themselves. The only occurrence of a lock outside of the TaskPool methods is in TaskScheduler::thread_run. This patch is most useful in combination with the OSL renderer mode, so it can probably wait until after the 2.64 release. SVM tasks tend to be cancelled quickly, so the effect is less noticeable.
2012-09-11 11:41:51 +00:00
if(num == 0) {
do_cancel = false;
num_cond.notify_all();
Performance fix for Cycles: Don't wait in the main UI thread when resetting devices. When the scene is updated Cycles resets the renderer device, cancelling all existing tasks. The main thread would wait for all running tasks to finish before continuing. This is ok when tasks can actually cancel in a timely fashion. For OSL however, this does not work, since the OSL shader group optimization takes quite a bit of time and can not be easily be cancelled once running (on my crappy machine in full debug mode: ~0.12 seconds for simple node trees). This would lead to very laggy UI behavior and make it difficult to accurately control elements such as sliders. This patch removes the wait condition from the device->task_cancel method. Instead it just sets the do_cancel flag and returns. To avoid backlog in the task pool of the device it will return early from the BlenderSession::sync function while the reset is going on (tested in Session::resetting). Once all existing tasks have finished the do_cancel flag is finally cleared again (checked in TaskPool::num_decrease). Care has to be taken to avoid race conditions on the do_cancel flag, since it can now be modified outside the TaskPool::cancel function itself. For this purpose the scope of the TaskPool::num_mutex locks has been extended, in most cases the mutex is now locked by the TaskPool itself before calling TaskScheduler methods, instead of only locking inside the num_increase/num_decrease functions themselves. The only occurrence of a lock outside of the TaskPool methods is in TaskScheduler::thread_run. This patch is most useful in combination with the OSL renderer mode, so it can probably wait until after the 2.64 release. SVM tasks tend to be cancelled quickly, so the effect is less noticeable.
2012-09-11 11:41:51 +00:00
}
}
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
void TaskPool::num_increase()
{
num++;
Performance fix for Cycles: Don't wait in the main UI thread when resetting devices. When the scene is updated Cycles resets the renderer device, cancelling all existing tasks. The main thread would wait for all running tasks to finish before continuing. This is ok when tasks can actually cancel in a timely fashion. For OSL however, this does not work, since the OSL shader group optimization takes quite a bit of time and can not be easily be cancelled once running (on my crappy machine in full debug mode: ~0.12 seconds for simple node trees). This would lead to very laggy UI behavior and make it difficult to accurately control elements such as sliders. This patch removes the wait condition from the device->task_cancel method. Instead it just sets the do_cancel flag and returns. To avoid backlog in the task pool of the device it will return early from the BlenderSession::sync function while the reset is going on (tested in Session::resetting). Once all existing tasks have finished the do_cancel flag is finally cleared again (checked in TaskPool::num_decrease). Care has to be taken to avoid race conditions on the do_cancel flag, since it can now be modified outside the TaskPool::cancel function itself. For this purpose the scope of the TaskPool::num_mutex locks has been extended, in most cases the mutex is now locked by the TaskPool itself before calling TaskScheduler methods, instead of only locking inside the num_increase/num_decrease functions themselves. The only occurrence of a lock outside of the TaskPool methods is in TaskScheduler::thread_run. This patch is most useful in combination with the OSL renderer mode, so it can probably wait until after the 2.64 release. SVM tasks tend to be cancelled quickly, so the effect is less noticeable.
2012-09-11 11:41:51 +00:00
num_cond.notify_all();
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
/* Task Scheduler */
thread_mutex TaskScheduler::mutex;
int TaskScheduler::users = 0;
vector<thread*> TaskScheduler::threads;
vector<int> TaskScheduler::thread_level;
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
volatile bool TaskScheduler::do_exit = false;
list<TaskScheduler::Entry> TaskScheduler::queue;
thread_mutex TaskScheduler::queue_mutex;
thread_condition_variable TaskScheduler::queue_cond;
void TaskScheduler::init(int num_threads)
{
thread_scoped_lock lock(mutex);
/* multiple cycles instances can use this task scheduler, sharing the same
2012-06-09 17:22:52 +00:00
* threads, so we keep track of the number of users. */
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
if(users == 0) {
do_exit = false;
/* launch threads that will be waiting for work */
if(num_threads == 0)
num_threads = system_cpu_thread_count();
threads.resize(num_threads);
thread_level.resize(num_threads);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
for(size_t i = 0; i < threads.size(); i++) {
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
threads[i] = new thread(function_bind(&TaskScheduler::thread_run, i));
thread_level[i] = 0;
}
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
users++;
}
void TaskScheduler::exit()
{
thread_scoped_lock lock(mutex);
users--;
if(users == 0) {
/* stop all waiting threads */
do_exit = true;
TaskScheduler::queue_cond.notify_all();
/* delete threads */
foreach(thread *t, threads) {
t->join();
delete t;
}
threads.clear();
thread_level.clear();
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
}
bool TaskScheduler::thread_wait_pop(Entry& entry)
{
thread_scoped_lock queue_lock(queue_mutex);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
while(queue.empty() && !do_exit)
queue_cond.wait(queue_lock);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
if(queue.empty()) {
assert(do_exit);
return false;
}
entry = queue.front();
queue.pop_front();
return true;
}
void TaskScheduler::thread_run(int thread_id)
{
Entry entry;
/* todo: test affinity/denormal mask */
/* keep popping off tasks */
while(thread_wait_pop(entry)) {
/* run task */
entry.task->run();
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
/* delete task */
delete entry.task;
/* notify pool task was done */
Performance fix for Cycles: Don't wait in the main UI thread when resetting devices. When the scene is updated Cycles resets the renderer device, cancelling all existing tasks. The main thread would wait for all running tasks to finish before continuing. This is ok when tasks can actually cancel in a timely fashion. For OSL however, this does not work, since the OSL shader group optimization takes quite a bit of time and can not be easily be cancelled once running (on my crappy machine in full debug mode: ~0.12 seconds for simple node trees). This would lead to very laggy UI behavior and make it difficult to accurately control elements such as sliders. This patch removes the wait condition from the device->task_cancel method. Instead it just sets the do_cancel flag and returns. To avoid backlog in the task pool of the device it will return early from the BlenderSession::sync function while the reset is going on (tested in Session::resetting). Once all existing tasks have finished the do_cancel flag is finally cleared again (checked in TaskPool::num_decrease). Care has to be taken to avoid race conditions on the do_cancel flag, since it can now be modified outside the TaskPool::cancel function itself. For this purpose the scope of the TaskPool::num_mutex locks has been extended, in most cases the mutex is now locked by the TaskPool itself before calling TaskScheduler methods, instead of only locking inside the num_increase/num_decrease functions themselves. The only occurrence of a lock outside of the TaskPool methods is in TaskScheduler::thread_run. This patch is most useful in combination with the OSL renderer mode, so it can probably wait until after the 2.64 release. SVM tasks tend to be cancelled quickly, so the effect is less noticeable.
2012-09-11 11:41:51 +00:00
{
/* not called from TaskPool, have to explicitly lock the mutex here */
thread_scoped_lock num_lock(entry.pool->num_mutex);
entry.pool->num_decrease(1);
}
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
}
void TaskScheduler::push(Entry& entry, bool front)
{
entry.pool->num_increase();
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
/* add entry to queue */
TaskScheduler::queue_mutex.lock();
if(front)
TaskScheduler::queue.push_front(entry);
else
TaskScheduler::queue.push_back(entry);
TaskScheduler::queue_cond.notify_one();
TaskScheduler::queue_mutex.unlock();
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
void TaskScheduler::clear(TaskPool *pool)
{
thread_scoped_lock queue_lock(TaskScheduler::queue_mutex);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
/* erase all tasks from this pool from the queue */
list<Entry>::iterator it = queue.begin();
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
int done = 0;
while(it != queue.end()) {
Entry& entry = *it;
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
if(entry.pool == pool) {
done++;
delete entry.task;
it = queue.erase(it);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
else
it++;
}
queue_lock.unlock();
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
/* notify done */
pool->num_decrease(done);
Cycles: merging features from tomato branch. === BVH build time optimizations === * BVH building was multithreaded. Not all building is multithreaded, packing and the initial bounding/splitting is still single threaded, but recursive splitting is, which was the main bottleneck. * Object splitting now uses binning rather than sorting of all elements, using code from the Embree raytracer from Intel. http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/ * Other small changes to avoid allocations, pack memory more tightly, avoid some unnecessary operations, ... These optimizations do not work yet when Spatial Splits are enabled, for that more work is needed. There's also other optimizations still needed, in particular for the case of many low poly objects, the packing step and node memory allocation. BVH raytracing time should remain about the same, but BVH build time should be significantly reduced, test here show speedup of about 5x to 10x on a dual core and 5x to 25x on an 8-core machine, depending on the scene. === Threads === Centralized task scheduler for multithreading, which is basically the CPU device threading code wrapped into something reusable. Basic idea is that there is a single TaskScheduler that keeps a pool of threads, one for each core. Other places in the code can then create a TaskPool that they can drop Tasks in to be executed by the scheduler, and wait for them to complete or cancel them early. === Normal ==== Added a Normal output to the texture coordinate node. This currently gives the object space normal, which is the same under object animation. In the future this might become a "generated" normal so it's also stable for deforming objects, but for now it's already useful for non-deforming objects. === Render Layers === Per render layer Samples control, leaving it to 0 will use the common scene setting. Environment pass will now render environment even if film is set to transparent. Exclude Layers" added. Scene layers (all object that influence the render, directly or indirectly) are shared between all render layers. However sometimes it's useful to leave out some object influence for a particular render layer. That's what this option allows you to do. === Filter Glossy === When using a value higher than 0.0, this will blur glossy reflections after blurry bounces, to reduce noise at the cost of accuracy. 1.0 is a good starting value to tweak. Some light paths have a low probability of being found while contributing much light to the pixel. As a result these light paths will be found in some pixels and not in others, causing fireflies. An example of such a difficult path might be a small light that is causing a small specular highlight on a sharp glossy material, which we are seeing through a rough glossy material. With path tracing it is difficult to find the specular highlight, but if we increase the roughness on the material the highlight gets bigger and softer, and so easier to find. Often this blurring will be hardly noticeable, because we are seeing it through a blurry material anyway, but there are also cases where this will lead to a loss of detail in lighting.
2012-04-28 08:53:59 +00:00
}
CCL_NAMESPACE_END