Adding a workaround in this case: we blit the depth buffer instead of the
stencil buffer and use the copy as the texture. This is slower but at
least it should work.
When calling glBlitFramebuffer on most (if not all) mac that have a GPU
from the Radeon Pro series, the data is not properly copied and only a
subset of the pixels are correctly copied.
This only happens when blitting the depth buffer if the depth buffer is
GL_DEPTH24_STENCIL8.
Changing the depth buffer format to GPU_DEPTH32F_STENCIL8 fixes the issue
but only works if blitting the depth componnent. The stencil componnent
still provoke issues when being copied.
This is caused by a driver bug that prevent us from rendering to (or even
binding) a texture mip level that is below GL_TEXTURE_MAX_LEVEL of the
target texture. This is fine in most drivers (and legal AFAIK) but not on
thoses Intels HDXXX + Windows.
As a fix we just put GL_TEXTURE_MAX_LEVEL lower (which is illegal because
it is undefined behaviour), but in practice it works ok and does not
trigger any warnings or errors.
This commit fixes most of the problems encountered on these GPUs (T56668).
The appearance is a bit different than 2.79 where the flame was just added
on top of the smoke without correct blending.
Now it's much more realistic and using volumetric integration. You can see
the smoke actually masking the flame.
The other difference is that the flame color was not using proper color
managed blending. Now with the use of filmic it shows bright yellow.
This could be adjusted and displayed as a user parameter in the future.
Includes the following fixes
- Fix smoke texture creation: data was interpreted as Byte instead of Floats.
- Fix Velocity texture not being free after draw: also was causing crashes.
- Fix display_thickness not being copied during COW.
- Fix Blending and general volume rendering algorithm.
- Add Volume Shadowing support.
NOTE: This commit only concern edit UVs and not the "shadow" mesh displayed
when texture painting. This will be address in a future commit.
We now cache the uv mesh in the mesh batch cache and only reupload data on
changes.
Update could be more granular (and a bit faster) but it's not our main
concern ATM.
This should fix problem caused by the IMM api used to draw large meshes.
This makes performance skyrocket compared to previous implementation.
There is still a big CPU bottleneck when not in sync selection mode but it
is not related to the drawing function directly.
Use the multiply blending mode for the weight paint overlay.
To support the opacity slider, we need a new shader. Otherwise this combination of multiplication and mixing does not seem to be supported by glBlendFunc.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3727
This implementation is a bit hacky but match cycles pretty close.
If pixel size is not enabled, it will use the geom shader to
compute distances between vertices. This will have a cost.
Implementation is a bit hacky in gpu_codegen to make the geom shader works
in an optional manner.
Seems like GLEW_ARB_texture_float is not defined in core profile on OSX
even if float textures are part of the 3.3 core spec.
So removing uneeded backup cases.
Terms get/set don't make much sense when casting values.
Name macros so the conversion is obvious,
use common prefix for easier completion.
- GET_INT_FROM_POINTER -> POINTER_AS_INT
- SET_INT_IN_POINTER -> POINTER_FROM_INT
- GET_UINT_FROM_POINTER -> POINTER_AS_UINT
- SET_UINT_IN_POINTER -> POINTER_FROM_UINT
This happens on NVidia GPUs, using more textures than the maximum allowed
by the gl will NOT trigger a linking issue (maybe because of bindless
texture implementation?).
So in this case we manually count the number of samplers per shader stage
and compare it against the GL limit. We discard the shader if the sampler
count is too high. This shows the user something is wrong with the shader.
Fix T56789: There was issue with certain driver with glGenerateMipmap and
GPU_DEPTH_COMPONENT24.
In this case we just create a complete texture with mipmaps manually
without downsampling / initializing the data.
Flat shaded grids are now rendered as a long triangle strip with
degenerate triangles.
Smooth shaded grids are not changed.
Also add back fast navigate. The only difference with 2.79 is that fast
navigate + flat shading will show slitly "incorrect" smooth face. But that's too
much effort to fix for what it is.
Didn't realize the index buffer is stored once in a BVH
and same pointer is reused. Surprisingly, simple files
were fixed with the previous fix.
Now disabled the optimization all together, and it was
simpler to just completely remove all residue of the
code. It is likely to be a different implementation
anyway, so no need to try to keep code in a semi-broken
state.
