I think its something like:
old was: 4*nlogn + 3*(n*6)
new is: (2*nlogn + 3*(n*6)) * f, with f<1
Still missing changing the sorting function to an introsort instead of qsort
Other options like bucketing sort may be worth trying (for very large trees)
only for 2childs splits
with "worse case heuristic" means each child is considered to have a cost linear on the number of leafs
no termination criteria
number of BB test/hits expected to "reduced" by some factor
tree building is also expected to be slower as previous split was "object mean", which is quite fast to evaluate
point is that other structures like trees can then distiguish between other nodes or rayobject primitives
withouth needing any other variable.
(Note yet used but will reduce memory by a nice factor (linear to the number of primitives))
*Octree works nicely on hierarchic trees
*the old code was quite destructive at the Isect variable
changing isec->start, end, vec..now it only changes isec->labda (and hit results)
Currently rendering a BVH of all objects, where each object has it own octree.
*Adapted octree to a more generic raytrace API
*ray shadow works (other untested stuff disabled atm)
On the scene tested the user-cpu time got from 1:24 to 1:19/20
probably because of removed callbacks or sligtly diferente memory usage
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.
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.
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.
Index OB pass didn't support FSA for Ztransp.
Also made buttons to set black/white for non-RGBA images hide in Image Window,
the Curves color code only supports 4 channels atm.
proper results, should bake as if SSS was disabled.
- Fix for GLSL to handle failing shadow buffer creation better.
- Fix for sky/atmosphere version patch, was not doing files from 2.46
and newer.
due to jittering of the start position for antialiasing in a pixel.
Now it distributes the start position over the fixed osa sample
positions, instead of of random positions in space. The ugly bit is
that a custom ordering was defined for osa 8/11/16 to ensure that the
first 4 are distributed relatively fair for adaptive sampling to decide
if more samples need to be taken.
SunSky didn't include skycolor in raytrace.
Note: there seems to be an error in sunsky when looking straight down,
so this option requires raytracing stuff not in outer space. :)
- removed ugly clamping function (it was dividing XYZ based on max of
one of the values)
- added option to use Exposure, this only works for brightness (Y).
results look very pleasant, foggy and hazy results are possible.
with exposre==0, no exposure happens for HDR extreme range skies,
this is how yafray rendered it.
- added menu for choosing color spaces (CIE = modern lcds)
Please review! (and yes i know it's still not in World :)
- Added blending mode and factor option, so it's more clear and
controllable what happens with it. Also nice for crazy effects
of course!
- Preview render now shows preview for it too
On the todos:
- have this in World buttons (as well) for quicker sky setups
- review math of color clamping and scaling, this is definitely
not good... but a fix will make old files look very different.
Finally, after a long time new render candy for the non-game peoples! :)
Good doc is here: (url splits in two)
http://www.harkyman.com/2008/08/06/controllable-shadow-intensity-
and-color/
Note the colorpicker for shadow is in "Shadow and Spot" panel. A bit
hidden, could get more attention. For later. :)
the features that are needed to run the game. Compile tested with
scons, make, but not cmake, that seems to have an issue not related
to these changes. The changes include:
* GLSL support in the viewport and game engine, enable in the game
menu in textured draw mode.
* Synced and merged part of the duplicated blender and gameengine/
gameplayer drawing code.
* Further refactoring of game engine drawing code, especially mesh
storage changed a lot.
* Optimizations in game engine armatures to avoid recomputations.
* A python function to get the framerate estimate in game.
* An option take object color into account in materials.
* An option to restrict shadow casters to a lamp's layers.
* Increase from 10 to 18 texture slots for materials, lamps, word.
An extra texture slot shows up once the last slot is used.
* Memory limit for undo, not enabled by default yet because it
needs the .B.blend to be changed.
* Multiple undo for image painting.
* An offset for dupligroups, so not all objects in a group have to
be at the origin.
Baking would split non-planer quads in an unpredictable way, which is fine for rending but game engines often use a fixed order (0,1,2), (0,2,3) or (1,2,3) (1,3,0).
Added an option to use a fixed order when baking.
computers, probably due to slow multithreaded malloc. Now each render
thread keeps a list of qmc samplers that it fills as needed while
rendering (since it is hard to predict the actual amount needed in
advance due to ray recursion).
Fix for bug #7418: texture ipo's didn't show for textures in node materials.
Fix for part of bug #6758: node materials in other node materials could
miss texture coordinates.
but only when the UV's are connected. That fixes some artifacts when
baking and using tangent space normal maps. It does mean increased
memory usage because it now stores 4 tangents per face like UV's,
and increased processing time, but there's no simple way around that.
the renderer could lead to results being reused for unrelated
points, result was one wrong pixel or strand per part. Now instead
of setting the sample counter to 0 multiple times, it keeps a global
counter per thread for the whole render.
coordinates from the object as if they were in their original position,
untransformed by dupligroups or dupliverts. Otherwise the textures of
such objects would change depending on their position.
This adds a few settings to control global render quality, for faster
renders when tweaking lighting etc. The implementation is not so great,
and this should really be part of a proper render profile and preset
system. So for now it's a hidden Peach feature, enabled by setting rt
to 1. Before the next release, I'll either remove or improve it.
Settings are:
- Maximum subsurf level
- Child particles percentage
- Maximum shadow map samples
- AO and SSS quality factor
images. Found method that doesn't require image to be rendered larger.
Note: assembling pre-rendered parts that are result of FSA renders might
still give minor visible artefacts on edges; however, we should include
such render methods in the render pipeline, so multiple computers can
each render parts, save all samples, and have 1 computer assembling and
compositing all. This is for another project... :)
Now bounding boxes are computed per object, and checked first before
zbuffering objects. For strands, bounding boxes are computed per
original face in the mesh. Overall the speed improvement from this
is quite small (zbuffering is rarely the bottleneck), but it seems a
sensible thing to do anyway.
They now also store a list of samples per pixel, and then get
shaded together with the ztransp samples. This comes with a
slight speed hit, but mainly memory might be a concern. However,
testing some peach scenes I haven't problems.
This completes the pipeline make-over, as started in 2006. With this
option, during rendering, each sample for every layer and pass is being
saved on disk (looks like non-antialiased images). Then the composite
and color correction happens, then a clip to 0-1 range, and only in end
all samples get combined - using sampling filters such as gauss/mitch/catmul.
This results in artefact-free antialiased images. Even Z-combine or
ID masks now work perfect for it!
This is an unfinished commit btw; Brecht will finish this for strands.
Also Halo doesnt work yet.
To activate FSA: press "Save Buffers" and the new button next to it. :)
Problem: artist wants character to walk in grass, but still have all rendered
in seperate render-layers, for postpro effects and vblur. How to efficiently
create a mask image you can put *over* the character for the grass?
Solution has two parts; this commits allows any layer inside of the renderlayers
to become a Z-mask (Z values for solid gets filled in, but not rendered).
Second part of commit is render option "Only render stuff that's in front of
a zbuffer value that was filled in (saves render time)
of strands changing between frames, vector blur couldn't work. Now
speed vectors are interpolated from the surface. This also means
child particles don't have to be computed in the previous and next
frames, so saves time too.
