This reduce the base size of the shaderinterface from 400 to 136 bytes.
Improves memory usage and cache coherency when querying a lot of uniforms
at once.
We set the default buffer size for strict buffers to 0. This increased
the performance on intel but reduced the performance on old AMD cards.
This patch sets back the default buffer size for strict buffers. After
testing I didn't detect a slow-down anymore.
Note that today new firmwares were pushed to the device what might fix
some issues.
This fixes a freeze when closing temporary windows with `AMD Radeon HD 7570M`
The performance is practically the same between calls (with a micro advantage for `GL_STATIC_DRAW`)
I couldn't check the difference in memory usage.
The ideal would be profile in different setups.
But due to the seriousness of the bug, these tests were postponed.
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
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
- Move gpuPush/Pop from GPU_draw.h into GPU_state.h
as this is for pushing/popping state.
- Add 'GPU_STANDALONE' define, to bypass use of user-preferences
for theme colors and pixelsize, as well as pbvh init/free functions.
Needed to get GHOST tests working again.
Instead of fixed size, `IMM_BUFFER_SIZE` is adjustable now. The internal buffer can expand if there is a need a bigger buffer. All other behaviors are still the same.
Reviewed By: fclem, #gpu_viewport
Differential Revision: https://developer.blender.org/D5570
BF-admins agree to remove header information that isn't useful,
to reduce noise.
- BEGIN/END license blocks
Developers should add non license comments as separate comment blocks.
No need for separator text.
- Contributors
This is often invalid, outdated or misleading
especially when splitting files.
It's more useful to git-blame to find out who has developed the code.
See P901 for script to perform these edits.
This is in order to make the API more multithread friendly inside the
draw manager.
GPU_shader_get_uniform will only serve to query the shader interface and
not do any GL call, making it threadsafe.
For now it only print a warning if the uniform was not queried before.
GPUFrameBuffers were being free when no context was attached or in the
wrong gl context. This make sure this does not happen again.
You can now safely free any gl resource from any thread (well as long as
it's not used anymore!).
This does not include all the struct and type renaming. Only files were
renamed.
gwn_batch.c/h was fusioned with GPU_batch.c/h
gwn_immediate.c/h was fusioned with GPU_immediate.c/h
gwn_imm_util.c/h was fusioned with GPU_immediate_util.c/h
This eliminates tons of glGetUniformLocation calls from the drawing loop. Vast majority of code can keep making the same function calls. They're just faster now!
- Batch_Uniform*
- immUniform*
- gpuBindMatrices
- and others
Core of the issue was that some of our Theme colors are RGB-only, but
were loaded as RGBA.
Note that tracking all possible cases is pretty impossible, so we'll
have to tackle those as they get reported am afraid.
Had to add a few utility functions to replace existing functions. Let me know if these are duplicates.
Reviewers: merwin
Reviewed By: merwin
Tags: #bf_blender_2.8
Maniphest Tasks: T49043
Differential Revision: https://developer.blender.org/D2434
- any shader program can use matrix state (not only built-in shaders)
- you can mix matrix & begin/end calls, and the bound shader will use the latest matrix state
Part of T49450 & T49043
Built-in shaders now use uniforms instead of legacy built-in matrices. So far I only hooked this up for new immediate mode.
We use the same matrix naming convention as OpenGL, but without the gl_ prefix, e.g. gl_ModelView becomes ModelView.
Right now it can skip the new matrix stack and use the legacy built-in matrices app-side. This will help us transition gradually from glMatrix functions to gpuMatrix functions.
Still some work to do in gpuBindMatrices. See TODO comments in gpu_matrix.c for specifics.
Was already done for immediate mode, but rearranged code to make a clean separation. Cleaned up #includes for code that uses this feature.
Added same for batched rendering.
Put Gawain source code in a subfolder to make the boundary between the
library and the rest of Blender clear.
Changed Gawain’s license from Apache to Mozilla Public License. Has
more essence of copyleft — closer to GPL but not as restrictive.
Split immediate.c into several files so parts can be reused (adding
more files soon…)
immBegin requires us to know how many vertices will be drawn. Most times this is fine, but sometimes it can be tricky. Do we make the effort to count everything in one pass, then draw it in a second?
immBeginAtMost makes this simple. Example: I'll draw at most 100 vertices. Supply only 6 verts and it draws only 6.
Any unused space is reclaimed and given to the next immBegin.
Was using GL_NONE to mean "no primitive" but GL_NONE and GL_POINTS are both defined as 0x0000.
Introducing PRIM_NONE = 0xF which does not clash with any primitive types.
