This changes the effect of the multiple scattering approximation to be more
physically plausible (and easier to use) by making it conserve energy.
Previously, the multiple scattering could show wildly different results
based on the spread settings, often outputting much more light than was put
in (via lamps), which is physically incorrect and made it difficult to use
by requiring a lot of tweaking of the intensity value.
This fixes it to some extent with a simple normalization, where it scales
the light energy after multiple scattering to be the same as what was
previously there in the light cache via single scattering. This means that
using the default intensity of 1.0 should give good results by default,
although you can still use it to tweak the effect.
Note: This will render differently if you've already set up a .blend using
multiple scattering with the old code, so you'll need to tweak older files.
Setting the intensity back to the default 1.0 should be good though.
* Smaller thing - fixed the camera view vector stuff up a bit in light
cache, it now gives much more similar results to non-light cache when using
anisotropic scattering.
This adds a header to the voxel data texture's data file format, to
auto-fill the resolution settings upon loading a file. I don't have a data
file of the right format to test with, so I'll trust it works and wait for
confirmation!
It also adds a 'still frame' setting, to pause the voxel data sequence on a
specified frame, throughout the course of the rendered animation.
Assorted smaller fixes:
- Fix: modal keymaps for editmode in view3d were not set again
when you copy areas or go fullscreen.
- Improved "redo last op" (F6) to search back in history for
a redoable operator. Operator also used wrong pupmenu type.
- On creating new FCurve editor, the channel rainbow colors are
set correct.
- EditMesh: fixed code for Spin/Screw, correct props, init and
error reporting. (Spin hotkey ALT+R temporary)
- recompiled all to check for uninitialized variable warnings.
(compile flag should be -O for this). Fixed some proto's.
Safe method to move render results to the displayed image.
It now allocates a single image for display, and on each
refresh callback from render, it copies the refreshed
section over to this image, in 32 bits. While rendering
that image then only shows progress updates, as usual.
This also now works for scenes in composte and results
for composite.
This should solve reported crashes for MBlur or SSS.
Render back! And not only back, even full threaded now. :)
Current state is unfinished, but too much fun to not to
commit for review and test!
WARNING: because render is in a threaded job, it will
use data as can be edited in the UI. That'll crash in many
cases of course... the idea is to limit UI usage to viewing
stuff, especially for the Image Window to inspect layers
or zoom in/out.
What works now;
- F12 render (no anim)
- ESC from render
- ESC pushes back temporary Image Window
- Render to ImageWindow or full-screen.
- Executing composites, and edit composites after render.
Note that the UI is 100% responsive in a render, you can
switch screens, slide area dividers around, or even load
a new file during render. :) It's quite stable even.
I'll collect all crash reports especially to get a good
picture of where the protection is required at least.
Also added: XKey "Delete Objects", to get things crash...
unfortunately it didn't for me.
* Cleaned up UI_interface.h a bit, and added some comments to
organize things a bit and indicate what should be used when.
* uiMenu* functions can now be used to create menus for headers
too, this is done with a uiDefMenuBut, which takes a pointer
to a uiMenuCreateFunc, that will then call uiMenu* functions.
* Renamed uiMenuBegin/End to uiPupMenuBegin/End, as these are
specific to making popup menus. Will convert the other
conformation popup menu functions to use this too so we can
remove some code.
* Extended uiMenu functions, now there is is also:
BooleanO, FloatO, BooleanR, EnumR, LevelEnumR, Separator.
* Converted image window headers to use uiMenu functions, simplifies
menu code further here. Did not remove the uiDefMenu functions as
they are used in sequencer/view3d in some places now (will fix).
* Also tried to simplify and fix bounds computation a bit better
for popup menus. It tried to find out in advance what the size
of the menu was but this is difficult with keymap strings in
there, now uiPopupBoundsBlock can figure this out afterwards and
ensure the popup is within the window bounds. Will convert some
other functions to use this too.
This is mostly a contribution from Raul 'farsthary' Hernandez - an approximation for
multiple scattering within volumes. Thanks, Raul! Where single scattering considers
the path from the light to a point in the volume, and to the eye, multiple scattering
approximates the interactions of light as it bounces around randomly within the
volume, before eventually reaching the eye.
It works as a diffusion process that effectively blurs the lighting information
that's already stored within the light cache.
A cloudy sky setup, with single scattering, and multiple scattering:
http://mke3.net/blender/devel/rendering/volumetrics/vol_sky_ss_ms.jpghttp://mke3.net/blender/devel/rendering/volumetrics/sky_ms.blend
To enable it, there is a menu in the volume panel (which needs a bit of cleanup, for
later), that lets you choose between self-shading methods:
* None: No attenuation of the light source by the volume - light passes straight
through at full strength
* Single Scattering: (same as previously, with 'self-shading' enabled)
* Multiple Scattering: Uses multiple scattering only for shading information
* Single + Multiple: Adds the multiple scattering lighting on top of the existing
single scattered light - this can be useful to tweak the strength of the effect,
while still retaining details in the lighting.
An example of how the different scattering methods affect the visual result:
http://mke3.net/blender/devel/rendering/volumetrics/ss_ms_comparison.jpghttp://mke3.net/blender/devel/rendering/volumetrics/ss_ms_comparison.blend
The multiple scattering methods introduce 3 new controls when enabled:
* Blur: A factor blending between fully diffuse/blurred lighting, and sharper
* Spread: The range that the diffuse blurred lighting spreads over - similar to a
blur width. The higher the spread, the slower the processing time.
* Intensity: A multiplier for the multiple scattering light brightness
Here's the effect of multiple scattering on a tight beam (similar to a laser). The
effect of the 'spread' value is pretty clear here:
http://mke3.net/blender/devel/rendering/volumetrics/ms_spread_laser.jpg
Unlike the rest of the system so far, this part of the volume rendering engine isn't
physically based, and currently it's not unusual to get non-physical results (i.e.
much more light being scattered out then goes in via lamps or emit). To counter this,
you can use the intensity slider to tweak the brightness - on the todo, perhaps there is a more automatic method we can work on for this later on. I'd also like to check
on speeding this up further with threading too.
Added WM Jobs manager
- WM can manage threaded jobs for you; just provide a couple
of components to get it work:
- customdata, free callback for it
- timer step, notifier code
- start callback, update callback
- Once started, each job runs an own timer, and will for
every time step check necessary updates, or close the
job when ready.
- No drawing happens in jobs, that's for notifiers!
- Every job stores an owner pointer, and based on this owner
it will prevent multiple jobs to enter the stack.
Instead it will re-use a running job, signal it to stop
and allow caller to re-initialize it even.
- Check new wm_jobs.c for more explanation. Jobs API is still
under construction.
Fun: BLI_addtail(&wm->jobs, steve); :)
Put Node shader previews back using wmJobs
- Preview calculating is now fully threaded (1 thread still)
- Thanks to new event system + notifiers, you can see
previews update even while dragging sliders!
- Currently it only starts when you change a node setting.
Warning: the thread render shares Node data, so don't delete
nodes while it renders! This topic is on the todo to make safe.
Also:
- bug in region initialize (do_versions) showed channel list in
node editor wrong.
- flagged the channel list 'hidden' now, it was really in the
way! This is for later to work on anyway.
- recoded Render API callbacks so it gets handlers passed on,
no globals to use anymore, remember?
- previewrender code gets now so much nicer! Will remove a lot
of stuff from code soon.
Finally, here is the basic (functional) prototype of the new animation system which will allow for the infamous "everything is animatable", and which also addresses several of the more serious shortcomings of the old system. Unfortunately, this will break old animation files (especially right now, as I haven't written the version patching code yet), however, this is for the future.
Highlights of the new system:
* Scrapped IPO-Curves/IPO/(Action+Constraint-Channels)/Action system, and replaced it with F-Curve/Action.
- F-Curves (animators from other packages will feel at home with this name) replace IPO-Curves.
- The 'new' Actions, act as the containers for F-Curves, so that they can be reused. They are therefore more akin to the old 'IPO' blocks, except they do not have the blocktype restriction, so you can store materials/texture/geometry F-Curves in the same Action as Object transforms, etc.
* F-Curves use RNA-paths for Data Access, hence allowing "every" (where sensible/editable that is) user-accessible setting from RNA to be animated.
* Drivers are no longer mixed with Animation Data, so rigs will not be that easily broken and several dependency problems can be eliminated. (NOTE: drivers haven't been hooked up yet, but the code is in place)
* F-Curve modifier system allows useful 'large-scale' manipulation of F-Curve values, including (I've only included implemented ones here): envelope deform (similar to lattices to allow broad-scale reshaping of curves), curve generator (polynomial or py-expression), cycles (replacing the old cyclic extrapolation modes, giving more control over this). (NOTE: currently this cannot be tested, as there's not access to them, but the code is all in place)
* NLA system with 'tracks' (i.e. layers), and multiple strips per track. (NOTE: NLA system is not yet functional, as it's only partially coded still)
There are more nice things that I will be preparing some nice docs for soon, but for now, check for more details:
http://lists.blender.org/pipermail/bf-taskforce25/2009-January/000260.html
So, what currently works:
* I've implemented two basic operators for the 3D-view only to Insert and Delete Keyframes. These are tempolary ones only that will be replaced in due course with 'proper' code.
* Object Loc/Rot/Scale can be keyframed. Also, the colour of the 'active' material (Note: this should really be for nth material instead, but that doesn't work yet in RNA) can also be keyframed into the same datablock.
* Standard animation refresh (i.e. animation resulting from NLA and Action evaluation) is now done completely separate from drivers before anything else is done after a frame change. Drivers are handled after this in a separate pass, as dictated by depsgraph flags, etc.
Notes:
* Drivers haven't been hooked up yet
* Only objects and data directly linked to objects can be animated.
* Depsgraph will need further tweaks. Currently, I've only made sure that it will update some things in the most basic cases (i.e. frame change).
* Animation Editors are currently broken (in terms of editing stuff). This will be my next target (priority to get Dopesheet working first, then F-Curve editor - i.e. old IPO Editor)
* I've had to put in large chunks of XXX sandboxing for old animation system code all around the place. This will be cleaned up in due course, as some places need special review.
In particular, the particles and sequencer code have far too many manual calls to calculate + flush animation info, which is really bad (this is a 'please explain yourselves' call to Physics coders!).
soc-2008-nicholasbishop branch.
Note: any old code with multires_test() or multires_level1_test() can
just be deleted, not needed by the multires modifier.
Think global, act local!
The old favorite G.scene gone! Man... that took almost 2 days.
Also removed G.curscreen and G.edbo.
Not everything could get solved; here's some notes.
- modifiers now store current scene in ModifierData. This is not
meant for permanent, but it can probably stick there until we
cleaned the anim system and depsgraph to cope better with
timing issues.
- Game engine G.scene should become an argument for staring it.
Didn't solve this yet.
- Texture nodes should get scene cfra, but the current implementation
is too tightly wrapped to do it easily.
From the anti-globalization department:
G.obedit terminated!
Wherever possible, use CTX_data_edit_object(C) to get this
now. It's stored in scene now, and the screen context has
it defined.
* Fixed an old problem where if both the camera and a solid surface were
inside a volume, the volume wouldn't attenuate in front of the surface (was
visible in the blender conference art festival clouds animation:
http://mke3.net/blender/devel/rendering/volumetrics/vol_shade_cam_inside.mov
* Initial support for refracting solids inside volumes. I might be able to
make this work better, but not sure at the moment. It's a bit dodgy,
limited by the code that does the recursive ray shading - it's really not
set up for this kind of thing and could use a refactor very much.
* Multithreaded volume light cache
While the render process itself is multithreaded, the light cache pre-process
previously wasn't (painfully noticed this the other week rendering on some
borrowed octocore nodes!). This commit adds threading, similar to the tiled render -
it divides the light cache's voxel grid into 3d parts and renders them with the
available threads.
This makes the most significant difference on shots where the light cache pre-
process is the bottleneck, so shots with either several lights, or a high res light
cache, or both. On this file (3 lights, light cache res 120), on my Core 2 Duo it now
renders in 27 seconds compared to 49 previously.
http://mke3.net/blender/devel/rendering/volumetrics/threaded_cache.jpg
Cleanup
- for portablity we can keep the old ugly defines for retrieving
active object, cfra and so on. But, they will use 'scene' not
G.scene.
- fixed code that uses those defines.
- some unused variables/functions removed
* Fixed a typo in the patch that made trilinear interpolation really slow
* Replaced the patch's trilinear interpolation code with the existing trilinear code from pbrt / light cache (first stage of unifying this voxel interpolation code)
This commit introduces a new texture ('Voxel Data'), used to load up saved voxel
data sets for rendering, contributed by Raúl 'farsthary' Fernández Hernández
with some additional tweaks. Thanks, Raúl!
The texture works similar to the existing point density texture, currently it
only provides intensity information, which can then be mapped (for example) to
density in a volume material. This is an early version, intended to read the
voxel format saved by Raúl's command line simulators, in future revisions
there's potential for making a more full-featured 'Blender voxel file format',
and also for supporting other formats too.
Note: Due to some subtleties in Raúl's existing released simulators, in order
to load them correctly the voxel data texture, you'll need to raise the
'resolution' value by 2. So if you baked out the simulation at resolution 50,
enter 52 for the resolution in the texture panel. This can possibly be fixed in
the simulator later on.
Right now, the way the texture is mapped is just in the space 0,0,0 <-> 1,1,1
and it can appear rotated 90 degrees incorrectly. This will be tackled, for now,
probably the easiest way to map it is with and empty, using Map Input -> Object.
Smoke test: http://www.vimeo.com/2449270
One more note, trilinear interpolation seems a bit slow at the moment, we'll
look into this.
For curiosity, while testing/debugging this, I made a script that exports a mesh
to voxel data. Here's a test of grogan (www.kajimba.com) converted to voxels,
rendered as a volume: http://www.vimeo.com/2512028
The script is available here: http://mke3.net/projects/bpython/export_object_voxeldata.py
* Another smaller thing, brought back early ray termination (was disabled
previously for debugging) and made it user configurable. It now appears as a new
value in the volume material: 'Depth Cutoff'. For some background info on what
this does, check:
http://farsthary.wordpress.com/2008/12/11/cutting-down-render-times/
* Also some disabled work-in-progess code for light cache
The render generated from Freestyle's data is currently stored in the original scene's render structure ( as 'freestyle_render'): when the render database is generated, the scene's geometrical data is first imported into Freestyle and strokes are calculated. The generated strokes are used to create a Blender scene, rendered independently. The render result is used in the rendering loop.
The final rendering is performed the same way edge rendering is, in a function ('freestyle_enhance_add') operating on each individual render part. Freestyle strokes are only included if the toggle button "Freestyle" (in the 'Output' panel) is active and if the "Freestyle" render layer is also selected. Freestyle's panel appears when the toggle button 'Freestyle' is active.
IMPORTANT: as of now, rendering ONLY works when OSA is disabled and when Xparts = Yparts = 1. If these settings are not set, a bogus image will be created.
To make the render happen, many modifications had to be made:
- the Canvas::Draw and Operators::create methods no longer render strokes. They only generate shading and locational information.
- a BlenderStrokeRenderer class was added to turn Freestyle's strokes into a Blender scene. Basically, the scene consists of strokes in their projected image 2D coordinates and an orthographic camera centered in the middle of the corresponding canvas. The scene is rendered using vertex colors, in shadeless mode (therefore, no lamp is needed). BlenderStrokeRenderer uses the old GLTextureManager to load textures (as required by the StrokeRenderer class), even though stroke textures are probably not supported (not tested). After the scene is rendered, it is safely and automatically discarded.
- AppCanvas' code was greatly reduced to the bare minimum. The former AppCanvas would use an OpenGL-based back buffer and z buffer to determine the scene's color and depth information. In the future, this data will be determined from the corresponding render passes. Currently, the integration is not achieved so all style modules using depth/color information are sure to fail.
- before, Freestyle needed an OpenGL context to determine the camera's information and to compute the view map. As of now, the modelview and projection matrices are fully determined using data provided by Blender. This means both perspective and orthographic projections are supported. The AppGLWidget will very soon be removed completely.
