Thanks to Eric Anholt I've finally understood (at least I hope) how to
count texture indirections and save up some. Texture sampling dependent
on the result of some math counts as an indirection phase. Grouped
texture lookups with no math involved count as a single indirection.
Math on the coordinates count as indirection.
So the best thing is to group all the math involving coordinates and
then do all the lookups.
This saves enough indirections to make glfilterblur and glow effect
work, albeit a bit slowly, on i915.
Remove unused uniforms from the laplacian filter. Also remove if
kernel[i] != 0 checks so that it compiles where IF is not available.
Again, big thanks to Eric Anholt for the hints.
Apparently assigning gl_TexCoord to a temp count as an indirection.
Using it directly avoids it and limits indirections to four not
exceeding i915 limit. Now xpro effect works on i915.
Get rid of polar coordinates in the twirl effect. The same can be done
using a rotation matrix, saving alu instructions and, most of all,
avoiding the use of the evil atan() function (which uses IF operators).
Calculate rotation angle in a saner, understandable way.
Works on i915! (Hope it still works elsewhere too as I'm not able to
test at the moment)
Get rid of polar coordinates in the tunnel effect as the same can easily
be done just clamping the radius and multiplying.
Remove the evil atan() call that uses branching and a lot of unneeded alu
instructions. Now works on i915!
Generate a normalized gaussian kernel with given size and standard
deviation on the fly.
Remove "norm_const" uniform from convolution shaders and provide a
normalized kernel instead. Remove norm_offset uniform as it was always
zero, will reintroduce it if really needed in the future. Thanks to Eric
Anholt for suggesting it.
Save some ALU instruction calculating directly the coordinate for
texture lookup instead of summing an offset.
Still exceed maximum indirect texture lookups on i915, the only solution
I see is using a 3x3 kernel.
Reduce the number of register calculating texture lookup offset on the
fly. It was just a simple sequence, no need to store it in a array.
Fixes maximum number of registers exceeded error with i915. Still
exceed maximum indirect texture lookups and maximum ALU instructions.
Maybe we should gave up some blur goodness and use lightly more little
kernels.
Apparently saving up some texture lookup for zero kernel elements is
definitely not worth the use of branching. This way convolution
fragment programs also work where IF operator is not supported (tested
on i915 and nouveau). See also discussion on bug #615696.
Thanks to Eric Anholt for spotting this.
Port blur filter to use the common convolution shaders in
gstgleffectssources.c. This reduces code duplication and, incidentally,
the shaders in the common file were already updated to not use array
constructor and to not depend on #version 120.
First step towards bug #615696 fixing.
Fix some crazy formatting caused by gst-indent previous runs and disable
the script for this file. The best would be to move shaders into
separate files and load them at runtime or hardcode them at compile
time.
For now only identity, mirror and squeeze effects are available.
Maybe some factorization is needed about compilation shader
before to put the other effects since only a copy/past is needed,
at least until effect number 9: heat.
The effects from 10:sepia to 15:glow require more work.
qglwtextureshare now works again. In this example,
the pipeline is src ! glupload ! fakesink.
So in this case the glupload element is a sink in
terms of gl chain.
But the problem is still there if the pipeline is
src ! glupload ! glfilter ! fakesink
(it's the case in sdlshare and cluttershare examples)
because since recent changes about how the gstgldisplay
is transmitted to the gl element, the context is usually
created by the sink in terms of gl chain.
A solution would be to also install this property on glfilter.
The background image needs to be scaled to fit current texture size.
Previously this was done by gdk_pixbuf_scale_simple but that's been
removed.
Create a texture from the background pixbuf with correct dimensions and
use interpolation shader to scale it to the right size. Interpolation
fragment shader doesn't have too much sense if all the textures don't
have the same size so this seemed the most natural place to do the
scaling. It could probably be done with some custom texture mapping
outside the shader but it involved more code.
Fixes bug #599883.
glmixer can be seen as a glfilter except it handles N requested
sink pads.
Each sink pad and the src pad are video/x-raw-gl.
glmixer is responsible for managing different framerates from inputs.
It uses OpenGL context sharing. It means that each input is in its
own OpenGL context shared together and shared with the OpenGL context
of the ouput gl chain.
Also add a glmosaic which is an example of implementation of glmixer.
For now glmosaic is a cube but it will be fixed in the next commits.
For now the glmixer has some weird behaviours in some configurations
but it will be improved in the next commits.
The autotools builds is temporarly broken since those changes
have been made on win32.
Before, a gstgldisplay was instancied by the gl src in terms of gl chain.
And then the next element got it through the first gstglbuffer.
Now, this is done though queries.
All glelements get their ref on a gstgldisplay in READY state.
This rewrite is mainly a first step to be able to share OpenGL context hold
by the gstgldisplay using more complex glelements.
For example, with a glvideomixer. The associated gstgldisplay of each gl chain
of the sink pads will share their OpenGL context.
Add a pkg-config check for opengl and if not found assume opengl-es. If user has
none of both one still get build error later on (there is no pkg-config for
opengl-es).
Add more files to EXTRA dist and build the opengles variant if selected.
Simmilar changes could be done for the winCE backend.
Greedyh operation implemented using OpenGL Shading Language.
We could add other operations later.
Does some good results but still not as expected.
That's why I do not add it yet to the build.
