Depth texture copy using glCopyTexSubImage2D is undefined behavior since
you cannot bind GL_DEPTH_ATTACHMENT to glReadBuffer.
Using glBlitFramebuffer as a fallback.
This new shader is able to emulate smooth wide lines drawing using a
geometry shader.
This shader needs viewportSize and lineWidth uniforms to be set.
There is multiple variants to replace the usage of wide lines for most
shaders.
This patch only fix the gizmo_types files and the navigation gizmo.
Other areas could be fixed afterward, I just limited the patch size.
Fix T57570.
Reviewed By: billreynish
Differential Revision: https://developer.blender.org/D7487
This allow to copy entire texture in a faster way than using framebuffer
blitting.
This uses ARB_copy_image extension if available and fallback to
glCopyTexSubImage2D for older gl version.
Both method should be as fast if not faster than the framebuffer blitting.
These values were hardcoded before Face Sets were enabled for Multires,
so enable the show_face_sets checks now.
Reviewed By: jbakker
Maniphest Tasks: T75329
Differential Revision: https://developer.blender.org/D7444
glAttributes also include `gl_` names. These don't have a location and
should be ignored during shader interface creation. Those internal names
received a location of -1 and therefore the bitmasking was undefined.
Users wouldn't notice this, but ASAN warned developers of this situation.
ASAN could quit making ASAN un-usable as most shaders have this issue.
Reviewed By: Clément Foucault`
Differential Revision: https://developer.blender.org/D7448
We used to have a single buffer that was shared between strict and
unstrict draw calls. This leads to many recreation events for the draw
buffers. This patch separates the Unstrict draw buffer from the strict
draw buffer.
This improves performance on Windows Intel 10th gen platform.
On a reference platfor before the patch I got 10 FPS, after this patch
it became 34fps. Note that the same test normally on a low end GPU can
get to 60fps so this does not solve all teh bottlenecks yet.
Reviewed By: Clément Foucault
Differential Revision: https://developer.blender.org/D7421
When drawing the viewport to the screen the draw calls were not batched.
This resulted in measurable slowdown on Windows Intel 10th gen
platforms.
This patch would cache the last draw calls per viewport. Our API does
support partial redrawing of the viewport, but that isn't used anywhere.
This patch does not include stereoscopy rendering. This still uses the
imm approach and would still be slow on certain hardware.
Reviewed By: Clément Foucault
Differential Revision: https://developer.blender.org/D7357
The root cause is that viewport can draw cached version of themself but
the scene can have been updated and the pointed curvemapping could have
been freed.
To workaround this we just keep a copy of the curvemap at the viewport
level.
This solution involves adding a uniform to each fragment shader that is
used by gizmo drawing and use the framebuffer state to set this uniform
accordingly.
This solution can also be carried to external shaders (addons).
A single line of code would then be enough to fix the issue.
The only trickery here is the dummy define:
`#define srgb_to_framebuffer_space(a)`
This is in order to avoid breaking other DRW shaders that use the same
fragment shader code but do not need the tranformation.
Related to T74139
Reviewed By: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D7261
Currently in fractal_noise functions, each subsequent octave doubles the
frequency and reduces the amplitude by half. This patch introduces Roughness
input to Noise and Wave nodes. This multiplier determines how quickly the
amplitudes of the subsequent octaves decrease.
Value of 0.5 will be the default, generating identical noise we had before.
Values above 0.5 will increase influence of each octave resulting in more
"rough" noise, most interesting pattern changes happen there. Values below
0.5 will result in more "smooth" noise.
Differential Revision: https://developer.blender.org/D7065
When drawing the viewport to the screen the draw calls were not batched.
This resulted in measurable slowdown on Windows Intel 10th gen
platforms.
This patch would cache the last draw calls per viewport. Our API does
support partial redrawing of the viewport, but that isn't used anywhere.
This patch does not include stereoscopy rendering. This still uses the
imm approach and would still be slow on certain hardware.
Reviewed By: Clément Foucault
Differential Revision: https://developer.blender.org/D7357
This reverts commit 58e20b432a.
Was calling discard twice, tsk.
Although for some reason it did quiet the leak.
Reverting because this is causing a crash.
The 10g Intel/Win driver doesn't work well with our emulated
intermediate mode. This patch alters the drawing of the drag widget of
the panels to reduce unneeded drawing.
The previous method would draw 16 boxes per widget. This new way would
cache this drawing in a GPU batch and just move the matrix around.
Reviewed By: Clément Foucault
Differential Revision: https://developer.blender.org/D7345
Note this only changes cases where the variable was declared inside
the for loop. To handle it outside as well is a different challenge.
Differential Revision: https://developer.blender.org/D7320
This implements the Sculpt Mode API functions needed for Face Sets and
visibility management for PBVH_GRIDS. No major changes were needed in
the operators and the sculpt mode code. This implementation stores the
face sets in the base mesh, so faces created in higher subdivision
levels can't be modified individually. Also, we are not checking for
multiple face sets per vertex (that can be added in the future), so
relax tools don't work yet. The rest of the features (paint, undo,
visibility operators..) work as expected.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D7168
On some drivers, the default values is not respected correctly.
To workaround this we create a small VBO that contains only 1 vec4 worth of
data and just bind it using glBindVertexBuffer to ensure 0 stride.
This fixes T75069 Instances not rendered correctly by workbench.
Using a float to store and render the mask seems like a waste of memory
without any noticeable difference in the viewport for its use case.
After this commit, the mask and the face sets combined should take the
same amount of GPU memory than only the mask in previous versions.
Reviewed By: jbakker
Differential Revision: https://developer.blender.org/D7148
SCULPT_FACE_SET_NONE default value is 0 and it is rendered hidden, so
the invert sign operation to show it was not working. Now the show all
function sets this face set to ID 1 before setting its sign.
I also refactored this check in gpu_buffers.
Not related to the reported issue, but the mesh in attached contains non
manifold geometry with hidden loose vertices, so the visibility state
was not syncing correctly to those vertices. Now the toggle operators
checks the current visibility only on the face sets, so no manifold
vertices are ignored (as they are in the rest of operations in sculpt
mode).
Reviewed By: jbakker
Maniphest Tasks: T74780
Differential Revision: https://developer.blender.org/D7188
The default face set color is white, so we can skip drawing the default
face set. This allows to enable again the optimization of not drawing
overlays in nodes where the mask is empty.
This will still slow down the viewport when a new face set is created
for the whole mesh or when inverting the mask, like in previous
versions.
I also renamed the function to make more clear that now it is checking
for both mask and face sets.
Reviewed By: brecht
Maniphest Tasks: T74692
Differential Revision: https://developer.blender.org/D7207
```
...\gpu_texture.c(466,7): warning C4555: result of expression not used
...\gpu_texture.c(559,7): warning C4555: result of expression not used
...\gpu_texture.c:1205:72: warning: pointer targets in passing argument 4 of ‘glGetTexLevelParameteriv’ differ in signedness [-Wpointer-sign]
```
This commit adds a `mipmaps` member to the `GPUTexture` struct and also
computes to the memory used by these mipmaps and the memory used for
textures that are created from an external bindcode.
So it solves the following inconsistencies:
- The memory value for mipmaps was not being computed.
- As `GPU_texture_from_bindcode` didn't call
`gpu_texture_memory_footprint_add`, it brought inconsistencies to the
value of the used memory, especially when the texture is freed.
Differential Revision: https://developer.blender.org/D3554
This patch adds ability to set up colors and size of background
(transparency) checkerboard pattern in viewport and 2d editors. No new
backgrounds, only changing colors in existing ones.
This is not the background of the viewport, it is a transparency
checkerboard that is turned on only in render mode, when the
transparency mode is on. And also in 2D-editors, (image, sequencer,
etc).
Reviewed By: Pablo Vazquez, Julian Eisel
Differential Revision: https://developer.blender.org/D6791
On some platforms the stereo viewport rendering was not working. The
issue was that the fragment shader and vertex shaded didn't match. Some
platforms will remove the non-matching in/out parameters and blender
needs to provide only the optimal set of parameters. Other platform
still want to receive data for the parameters that aren't used.
This fix uses the correct vertex shader that matches the fragment shader
making both platforms render the same result.
