Directional Blur node allows the users to do various blur operations on the input
image. It essentially offers three different kind of ways of blurring in one node.
It is possible to blur using a certain direction, spin and zoom. These three ways
can be used in conjunction.
The node contains following controls:
*Iterations, Wrap
*Center: X, Y
*Distance, Angle
*Spin
*Zoom
Iterations is used to determine the smoothness of the result. The more iterations,
the smoother result. Low values are good for preview.
Wrap means that the image is wrapped as if it was tiled on both x and y directions.
To see better what this means, try it with spin for instance.
Center values (X and Y) determine the location which is used as a pivot point for
the operations. It is center (0.5) of the image by default.
Distance and angle are used to adjust directional blur. The result can be described
as a sweep that varies based on given distance (bigger distance, longer sweep) and
angle. Angle is given in degrees.
Spin produces rotating blur based on given angle. Yet again it is in degrees. Also
negative values work.
Zoom causes the image to be zoomed towards set center point (Center values).
Thanks to Alfredo de Greef (eeshlo) for contribution.
Possible development ideas:
*Make an algorithm to extend image in case spin is used. Extend would temporarily
change the size of the canvas of the input image. Canvas would be filled based on
colors on the edges of the input image. After the blur operation has been done,
the image would be cropped back to normal size. The advantage of this would be nicer
result of spin (no problems with image size) on a computational cost.
*Make values animatable. This is something that is better solved on more general
level. ("everything is animatable" paradigm)
*Provide an option to calculate automatic value for iterations. A good value that
produces a smooth result could be calculated based on direction deltas. This would be
useful in conjuction of animatable values.
=================
Big commit, but little user visible changes.
- Dupliverts and duplifaces are now rendered as instances, instead
of storing all of the geometry for each dupli, now an instance is
created with a matrix transform refering to the source object.
This should allow us to render tree leaves more memory efficient.
- Radiosity and to some degree raytracing of such objects is not
really efficient still. For radiosity this is fundamentally hard
to solve, but raytracing an octree could be created for each object,
but the current octree code with it's fixed size doesn't allow this
efficiently.
- The regression tests survived, but with I expect that some bugs will
pop up .. hopefully not too many :).
Implementation Notes
====================
- Dupligroups and linked meshes are not rendered as instances yet,
since they can in fact be different due to various reasons,
instancing of these types of duplis that are the same can be added
for them at a later point.
- Each ObjectRen now stores it's own database, instead of there being
one big databases of faces, verts, .. . Which objects that are actually
rendered are defined by the list of ObjectRenInstances, which all refer
to an ObjectRen.
- Homogeneous coordinatess and clipping is now not stored in vertices
anymore, but instead computed on the fly. This couldn't work for
instances. That does mean some extra computation has to be done, but
memory lookups can be slow too, and this saves some memory. Overall
I didn't find a significant speed impact.
- OSA rendering for solid and ztransp now is different. Instead of e.g.
going 8 times over the databases times and rendering the z-buffer, it
now goes over the database once and renders each polygon 8 times. That
was necessary to keep instances efficient, and can also give some
performance improvement without instances.
- There was already instancing support in the yafray export code, now it
uses Blender's render instances for export.
- UV and color layer storage in the render was a bit messy before, now
should be easier to understand.
- convertblender.c was reorganized somewhat. Regular render, speedvector
and baking now use a single function to create the database, previously
there was code duplicated for it.
- Some of these changes were done with future multithreading of scene
and shadow buffer creation in mind, though especially for scene creation
much work remains to be done to make it threadsafe, since it also involves
a lot of code from blenkernel, and there is an ugly conflict with the way
dupli groups work here .. though in the render code itself it's almost there.
[#7660] Solaris 10 x86 support (Makefiles)
Hopefully it will not mess up anything for anyone else. I removed
some hardcoded static libs and made NAN_*_LIB definitions so they could be
overridden, to allow greater flexability.
Let me know if there are any problems/questions.
Kent
- update for new particle system
- made blenderplayer.exe compile again (DDS library missing)
- exchanged a few float math functions (logf, sinf, cosf, expf, powf) with their double counterparts in CMP_nodes
(MSVC chokes on them, because the compiler is not C99 compliant)
This option sets the relative scaling factor to the amount set in the
scene "100%/75%/50%/25%" buttons. It's useful when you've got a fixed
background image, and want to do preview renders at a lesser
percentage, so you don't have to go and change the scale node each
time you change the %.
Also removed unnecessary use of a global from texture node.
This commit adds a new node, crop, to the compositor. This node can be used to
crop input image. It has two modes of operation. It can either crop image
size (Crop Image Size option) or crop while retaining original size of the
image. This latter mode can be used to preview the crop.
Use X1, Y1, X2, Y2 controls to manage the area to be cropped.
Note that I added a check for image preview min and max values to node_update.
This is because it could give inappropriate values in certain cases when
Crop Image Size option was toggled (values such as x1=0, y1=0, x2=60, y2=0 would
result in eternal loop due to bad min and max (min bigger than max!)). The check
makes sure that min and max values are always valid.
Previously if the texture node was not connected directly as the
second input to a node like Mix, the size of the generated image
was initialised to 140x140, which is not that useful. It now is
initialised to the size of the render result if it exists, and if
not, the size of the scene render sizeX/sizeY.
This alleviates bug #7068, which is caused by the displace node
trying to get pixels from a texture node, where the texture node
doesn't contain pixels to cover the entire image (only 140x140)
but it's not really a bulletproof solution...
This commit ports Hue Saturation Value node to the material node system.
Unlike in composite version, each value is an input. Composite node version
will be updated later to use the same scheme.
of the nodes, I realized I had strayed from the path of enlightened blending
by causing the math node to create an output the size of the larger of the
two inputs.
It has been corrected create the output the size of the
first image, and in its abscense the second image. In the event of
nether input containing image data the node does not function. I also
added some early out checks at the beginning of the function to speed it
up a tad in these cases and commented the code a bit more.
* Geometry node: Front/back output
This is used as a mask for determining whether you're looking at the front side or back side of a mesh, useful for blending materials, my practical need was giving different materials to the pages of a magazine: http://mke3.net/blender/etc/frontback-h264.mov
Give 1.0 if it's the front side, and 0.0 if it's the back side.
* Extended material node
This is the same as the material node, but gives more available inputs and outputs, (basically just connecting up more of ShadeInput and ShadeResult to the node). I didn't want to add it to the normal simple Material node since you don't always need all that stuff, and it would make the node huge, but when you do need it, it's nice to have it.
== Comp nodes ==
* Invert node
Inverting is something that happens all the time in a node setup, and this makes it easier. It's been possible to invert previously by adding a mix node and subtracting the input from 1.0, but it's not the best way of doing it. This node:
- makes it a lot faster to set up, rather than all the clicking required with the mix node
- is a lot more usable amidst a complex comp setup, when you're looking at a node tree, it's very helpful to be able to see at a glance what's going on. Using subtract for inverting is easily mixed up with other nodes in which you are actually subtracting, not inverting, and looks very similar to all the other mix nodes that usually litter a comp tree.
- has options to invert the RGB channels, the Alpha channel, or both. This saves adding lots of extra nodes (separate RGBA, subtract, set alpha) when you want to do something simple like invert an alpha channel. I'd like to add this option to other nodes too.
There's also a shader node version too.
* Also a few fixes that I committed ages ago, but seems to have been overwritten in Bob's node refactor:
- adding new compbufs to the set alpha and alphaover nodes when you have only one noodle connected to the lower input
- making the fac value on RGB curves still work when there's nothing connected to it
node_composite.c file was split into separate files for each node, it didn't
include a recent defocus node commit.
For the other half of the problem, see comments in code.
* refactor copying and freeing of node->storage by handlerizing them.
- freestoragefunc
- copystoragefunc
- node_util.c/h have generic handlers for these.
Crashes in composite; caused by bad bad bug in freeing "passed on" buffers,
only happens with option "Free unused" set though.
Appeared to be mixup of variable names. Code changes with 2 bytes... :)
Works simple; just check for
if(node->exec & NODE_BREAK)
break;
The main process (node processor) sets such a flag, checking for esc
20 times per second. That means you can check for ESC while doing image
processing without much cpu overhead.
Currently only added in blur nodes and defocus. Needs to be added all over,
nice for others... needs careful tests too.
What we now could do is even calling ESC on editing commands or mouseclicks
in composite editor? Could give user feeling of interactive app :) Further,
finished nodes are kept in memory anyway.
This will allow python or plugin defined nodes to work as well.
(And fixes compile issues with MSVC in yesterdays commit for nodes)
Code provided by Nathan L.
Fixes in his code:
- free_nodesystem() was called too late (after guarded alloc was closed)
- free_nodesystem() was freeing nodes that were not malloced even
- free_nodesystem was using free, not freeN :)
- the typedefs needed to be malloced yes, to allow duplicate nodes like
group but also for dynamic nodes.
- removed all UI .h include stuff! (Not allowed outside of src/)
- Makefile had typo error, compilation failed
Now it survives make, but it does not enter the two the subdirectories for
shader and compo nodes. So a linking gives unrecovered stuff in the end.
How does that work? I need help!
