Files
test/source/blender/draw/intern/draw_view.cc

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

271 lines
9.8 KiB
C++
Raw Normal View History

DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
/* SPDX-License-Identifier: GPL-2.0-or-later
* Copyright 2022 Blender Foundation. */
/** \file
* \ingroup draw
*/
#include "BLI_math_geom.h"
#include "GPU_compute.h"
#include "GPU_debug.h"
#include "draw_debug.hh"
#include "draw_shader.h"
#include "draw_view.hh"
namespace blender::draw {
void View::sync(const float4x4 &view_mat, const float4x4 &win_mat, int view_id)
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
{
data_[view_id].viewmat = view_mat;
data_[view_id].viewinv = view_mat.inverted();
data_[view_id].winmat = win_mat;
data_[view_id].wininv = win_mat.inverted();
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
is_inverted_ = (is_negative_m4(view_mat.ptr()) == is_negative_m4(win_mat.ptr()));
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
BoundBox &bound_box = *reinterpret_cast<BoundBox *>(&culling_[view_id].corners);
BoundSphere &bound_sphere = *reinterpret_cast<BoundSphere *>(&culling_[view_id].bound_sphere);
frustum_boundbox_calc(bound_box, view_id);
frustum_culling_planes_calc(view_id);
frustum_culling_sphere_calc(bound_box, bound_sphere, view_id);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
dirty_ = true;
}
void View::frustum_boundbox_calc(BoundBox &bbox, int view_id)
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
{
/* Extract the 8 corners from a Projection Matrix. */
#if 0 /* Equivalent to this but it has accuracy problems. */
BKE_boundbox_init_from_minmax(&bbox, float3(-1.0f),float3(1.0f));
for (int i = 0; i < 8; i++) {
mul_project_m4_v3(data_.wininv.ptr(), bbox.vec[i]);
}
#endif
float left, right, bottom, top, near, far;
bool is_persp = data_[view_id].winmat[3][3] == 0.0f;
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
projmat_dimensions(data_[view_id].winmat.ptr(), &left, &right, &bottom, &top, &near, &far);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
bbox.vec[0][2] = bbox.vec[3][2] = bbox.vec[7][2] = bbox.vec[4][2] = -near;
bbox.vec[0][0] = bbox.vec[3][0] = left;
bbox.vec[4][0] = bbox.vec[7][0] = right;
bbox.vec[0][1] = bbox.vec[4][1] = bottom;
bbox.vec[7][1] = bbox.vec[3][1] = top;
/* Get the coordinates of the far plane. */
if (is_persp) {
float sca_far = far / near;
left *= sca_far;
right *= sca_far;
bottom *= sca_far;
top *= sca_far;
}
bbox.vec[1][2] = bbox.vec[2][2] = bbox.vec[6][2] = bbox.vec[5][2] = -far;
bbox.vec[1][0] = bbox.vec[2][0] = left;
bbox.vec[6][0] = bbox.vec[5][0] = right;
bbox.vec[1][1] = bbox.vec[5][1] = bottom;
bbox.vec[2][1] = bbox.vec[6][1] = top;
/* Transform into world space. */
for (int i = 0; i < 8; i++) {
mul_m4_v3(data_[view_id].viewinv.ptr(), bbox.vec[i]);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
}
}
void View::frustum_culling_planes_calc(int view_id)
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
{
float4x4 persmat = data_[view_id].winmat * data_[view_id].viewmat;
planes_from_projmat(persmat.ptr(),
culling_[view_id].planes[0],
culling_[view_id].planes[5],
culling_[view_id].planes[1],
culling_[view_id].planes[3],
culling_[view_id].planes[4],
culling_[view_id].planes[2]);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
/* Normalize. */
for (int p = 0; p < 6; p++) {
culling_[view_id].planes[p].w /= normalize_v3(culling_[view_id].planes[p]);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
}
}
void View::frustum_culling_sphere_calc(const BoundBox &bbox, BoundSphere &bsphere, int view_id)
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
{
/* Extract Bounding Sphere */
if (data_[view_id].winmat[3][3] != 0.0f) {
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
/* Orthographic */
/* The most extreme points on the near and far plane. (normalized device coords). */
const float *nearpoint = bbox.vec[0];
const float *farpoint = bbox.vec[6];
/* just use median point */
mid_v3_v3v3(bsphere.center, farpoint, nearpoint);
bsphere.radius = len_v3v3(bsphere.center, farpoint);
}
else if (data_[view_id].winmat[2][0] == 0.0f && data_[view_id].winmat[2][1] == 0.0f) {
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
/* Perspective with symmetrical frustum. */
/* We obtain the center and radius of the circumscribed circle of the
* isosceles trapezoid composed by the diagonals of the near and far clipping plane */
/* center of each clipping plane */
float mid_min[3], mid_max[3];
mid_v3_v3v3(mid_min, bbox.vec[3], bbox.vec[4]);
mid_v3_v3v3(mid_max, bbox.vec[2], bbox.vec[5]);
/* square length of the diagonals of each clipping plane */
float a_sq = len_squared_v3v3(bbox.vec[3], bbox.vec[4]);
float b_sq = len_squared_v3v3(bbox.vec[2], bbox.vec[5]);
/* distance squared between clipping planes */
float h_sq = len_squared_v3v3(mid_min, mid_max);
float fac = (4 * h_sq + b_sq - a_sq) / (8 * h_sq);
/* The goal is to get the smallest sphere,
* not the sphere that passes through each corner */
CLAMP(fac, 0.0f, 1.0f);
interp_v3_v3v3(bsphere.center, mid_min, mid_max, fac);
/* distance from the center to one of the points of the far plane (1, 2, 5, 6) */
bsphere.radius = len_v3v3(bsphere.center, bbox.vec[1]);
}
else {
/* Perspective with asymmetrical frustum. */
/* We put the sphere center on the line that goes from origin
* to the center of the far clipping plane. */
/* Detect which of the corner of the far clipping plane is the farthest to the origin */
float nfar[4]; /* most extreme far point in NDC space */
float farxy[2]; /* far-point projection onto the near plane */
float farpoint[3] = {0.0f}; /* most extreme far point in camera coordinate */
float nearpoint[3]; /* most extreme near point in camera coordinate */
float farcenter[3] = {0.0f}; /* center of far clipping plane in camera coordinate */
float F = -1.0f, N; /* square distance of far and near point to origin */
float f, n; /* distance of far and near point to z axis. f is always > 0 but n can be < 0 */
float e, s; /* far and near clipping distance (<0) */
float c; /* slope of center line = distance of far clipping center
* to z axis / far clipping distance. */
float z; /* projection of sphere center on z axis (<0) */
/* Find farthest corner and center of far clip plane. */
float corner[3] = {1.0f, 1.0f, 1.0f}; /* in clip space */
for (int i = 0; i < 4; i++) {
float point[3];
mul_v3_project_m4_v3(point, data_[view_id].wininv.ptr(), corner);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
float len = len_squared_v3(point);
if (len > F) {
copy_v3_v3(nfar, corner);
copy_v3_v3(farpoint, point);
F = len;
}
add_v3_v3(farcenter, point);
/* rotate by 90 degree to walk through the 4 points of the far clip plane */
float tmp = corner[0];
corner[0] = -corner[1];
corner[1] = tmp;
}
/* the far center is the average of the far clipping points */
mul_v3_fl(farcenter, 0.25f);
/* the extreme near point is the opposite point on the near clipping plane */
copy_v3_fl3(nfar, -nfar[0], -nfar[1], -1.0f);
mul_v3_project_m4_v3(nearpoint, data_[view_id].wininv.ptr(), nfar);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
/* this is a frustum projection */
N = len_squared_v3(nearpoint);
e = farpoint[2];
s = nearpoint[2];
/* distance to view Z axis */
f = len_v2(farpoint);
/* get corresponding point on the near plane */
mul_v2_v2fl(farxy, farpoint, s / e);
/* this formula preserve the sign of n */
sub_v2_v2(nearpoint, farxy);
n = f * s / e - len_v2(nearpoint);
c = len_v2(farcenter) / e;
/* the big formula, it simplifies to (F-N)/(2(e-s)) for the symmetric case */
z = (F - N) / (2.0f * (e - s + c * (f - n)));
bsphere.center[0] = farcenter[0] * z / e;
bsphere.center[1] = farcenter[1] * z / e;
bsphere.center[2] = z;
/* For XR, the view matrix may contain a scale factor. Then, transforming only the center
* into world space after calculating the radius will result in incorrect behavior. */
mul_m4_v3(data_[view_id].viewinv.ptr(), bsphere.center); /* Transform to world space. */
mul_m4_v3(data_[view_id].viewinv.ptr(), farpoint);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
bsphere.radius = len_v3v3(bsphere.center, farpoint);
}
}
void View::bind()
{
if (dirty_) {
dirty_ = false;
data_.push_update();
culling_.push_update();
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
}
GPU_uniformbuf_bind(data_, DRW_VIEW_UBO_SLOT);
GPU_uniformbuf_bind(culling_, DRW_VIEW_CULLING_UBO_SLOT);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
}
void View::compute_visibility(ObjectBoundsBuf &bounds, uint resource_len, bool debug_freeze)
{
if (debug_freeze && frozen_ == false) {
data_freeze_[0] = static_cast<ViewMatrices>(data_[0]);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
data_freeze_.push_update();
culling_freeze_[0] = static_cast<ViewCullingData>(culling_[0]);
culling_freeze_.push_update();
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
}
#ifdef DEBUG
if (debug_freeze) {
float4x4 persmat = data_freeze_[0].winmat * data_freeze_[0].viewmat;
drw_debug_matrix_as_bbox(persmat.inverted(), float4(0, 1, 0, 1));
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
}
#endif
frozen_ = debug_freeze;
GPU_debug_group_begin("View.compute_visibility");
/* TODO(fclem): Early out if visibility hasn't changed. */
uint word_per_draw = this->visibility_word_per_draw();
/* Switch between tightly packed and set of whole word per instance. */
uint words_len = (view_len_ == 1) ? divide_ceil_u(resource_len, 32) :
resource_len * word_per_draw;
words_len = ceil_to_multiple_u(max_ii(1, words_len), 4);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
/* TODO(fclem): Resize to nearest pow2 to reduce fragmentation. */
visibility_buf_.resize(words_len);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
uint32_t data = 0xFFFFFFFFu;
GPU_storagebuf_clear(visibility_buf_, GPU_R32UI, GPU_DATA_UINT, &data);
if (do_visibility_) {
GPUShader *shader = DRW_shader_draw_visibility_compute_get();
GPU_shader_bind(shader);
GPU_shader_uniform_1i(shader, "resource_len", resource_len);
GPU_shader_uniform_1i(shader, "view_len", view_len_);
GPU_shader_uniform_1i(shader, "visibility_word_per_draw", word_per_draw);
DRWManager: New implementation. This is a new implementation of the draw manager using modern rendering practices and GPU driven culling. This only ports features that are not considered deprecated or to be removed. The old DRW API is kept working along side this new one, and does not interfeer with it. However this needed some more hacking inside the draw_view_lib.glsl. At least the create info are well separated. The reviewer might start by looking at `draw_pass_test.cc` to see the API in usage. Important files are `draw_pass.hh`, `draw_command.hh`, `draw_command_shared.hh`. In a nutshell (for a developper used to old DRW API): - `DRWShadingGroups` are replaced by `Pass<T>::Sub`. - Contrary to DRWShadingGroups, all commands recorded inside a pass or sub-pass (even binds / push_constant / uniforms) will be executed in order. - All memory is managed per object (except for Sub-Pass which are managed by their parent pass) and not from draw manager pools. So passes "can" potentially be recorded once and submitted multiple time (but this is not really encouraged for now). The only implicit link is between resource lifetime and `ResourceHandles` - Sub passes can be any level deep. - IMPORTANT: All state propagate from sub pass to subpass. There is no state stack concept anymore. Ensure the correct render state is set before drawing anything using `Pass::state_set()`. - The drawcalls now needs a `ResourceHandle` instead of an `Object *`. This is to remove any implicit dependency between `Pass` and `Manager`. This was a huge problem in old implementation since the manager did not know what to pull from the object. Now it is explicitly requested by the engine. - The pases need to be submitted to a `draw::Manager` instance which can be retrieved using `DRW_manager_get()` (for now). Internally: - All object data are stored in contiguous storage buffers. Removing a lot of complexity in the pass submission. - Draw calls are sorted and visibility tested on GPU. Making more modern culling and better instancing usage possible in the future. - Unit Tests have been added for regression testing and avoid most API breakage. - `draw::View` now contains culling data for all objects in the scene allowing caching for multiple views. - Bounding box and sphere final setup is moved to GPU. - Some global resources locations have been hardcoded to reduce complexity. What is missing: - ~~Workaround for lack of gl_BaseInstanceARB.~~ Done - ~~Object Uniform Attributes.~~ Done (Not in this patch) - Workaround for hardware supporting a maximum of 8 SSBO. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D15817
2022-09-02 18:30:48 +02:00
GPU_storagebuf_bind(bounds, GPU_shader_get_ssbo(shader, "bounds_buf"));
GPU_storagebuf_bind(visibility_buf_, GPU_shader_get_ssbo(shader, "visibility_buf"));
GPU_uniformbuf_bind((frozen_) ? data_freeze_ : data_, DRW_VIEW_UBO_SLOT);
GPU_compute_dispatch(shader, divide_ceil_u(resource_len, DRW_VISIBILITY_GROUP_SIZE), 1, 1);
GPU_memory_barrier(GPU_BARRIER_SHADER_STORAGE);
}
if (frozen_) {
/* Bind back the non frozen data. */
GPU_uniformbuf_bind(data_, DRW_VIEW_UBO_SLOT);
}
GPU_debug_group_end();
}
} // namespace blender::draw