The issue was caused by the image save operator ignoring Z-Buf if the input
buffer due to the way how IMB_dupImBuf() worked. There's no reason to to
ignore z-buf on imbuf duplication, it only asks for issues in the future.
Windows-only bug, mmap & co are not threadsafe by default on this platform, so we have to add a dedicated
spinlock for them in win32.
Note that we may try to get rid of those mmap later, but not for 2.76!
To be backported to final 2.76...
ImBuf types were getting stored as bitflags in a 32bit integer which had
already run out of space. Solved the problem by separating file type to
an ftype enum, and file specific options to foptions.
Reviewed by Campbell, thanks a lot!
transform.
IMB_makeSingleUser makes a copy and destroys metadata. I am not sure if
this is the safest way to make a single user ImBuf (setting the refcount
to 0 is simpler and there's less, but no zero, risk of dangling
pointers) but I will leave this as is for now in case there is an actual
need for a copy here. The alternative approach should be tested at some
point.
Moral of the story: Make sure that size_t is used whenever pointer
arithmetic is involved. For images, that basically means whenever any
squared dimensions are involved. Casting an operand to size_t early in
the operation is usually sufficient to force the entire operation to
size_t.
There might still be places lurking where we don't support this
correctly. This has been tested with render pipeline, quite a few image
functions (meaning we can paint on such images now, albeit somewhat
slowly ;) ) and export to jpeg. Too many places in code to check so I
guess we'll be handling cases as they come.
Don't try this at home unless you have an immense ammount of RAM.
First GPixel render of suzanne in the multiverse can be found here:
http://download.blender.org/demo/test/suzanne-billion-pixel.jpg
Can be viewed from blender (takes about 3.3 GB after loading but may
take more during loading so 8GB might be more safe to try this).
Issue is likely caused by thread-unsafe nature of IMB_freeImBuf
which might lead to race condition in some circumstances.
Now made it thread-safe and from Sebastian's tests seems crash is
gone now, so hopefully the root of the issue is finally nailed down.
Replace old color pipeline which was supporting linear/sRGB color spaces
only with OpenColorIO-based pipeline.
This introduces two configurable color spaces:
- Input color space for images and movie clips. This space is used to convert
images/movies from color space in which file is saved to Blender's linear
space (for float images, byte images are not internally converted, only input
space is stored for such images and used later).
This setting could be found in image/clip data block settings.
- Display color space which defines space in which particular display is working.
This settings could be found in scene's Color Management panel.
When render result is being displayed on the screen, apart from converting image
to display space, some additional conversions could happen.
This conversions are:
- View, which defines tone curve applying before display transformation.
These are different ways to view the image on the same display device.
For example it could be used to emulate film view on sRGB display.
- Exposure affects on image exposure before tone map is applied.
- Gamma is post-display gamma correction, could be used to match particular
display gamma.
- RGB curves are user-defined curves which are applying before display
transformation, could be used for different purposes.
All this settings by default are only applying on render result and does not
affect on other images. If some particular image needs to be affected by this
transformation, "View as Render" setting of image data block should be set to
truth. Movie clips are always affected by all display transformations.
This commit also introduces configurable color space in which sequencer is
working. This setting could be found in scene's Color Management panel and
it should be used if such stuff as grading needs to be done in color space
different from sRGB (i.e. when Film view on sRGB display is use, using VD16
space as sequencer's internal space would make grading working in space
which is close to the space using for display).
Some technical notes:
- Image buffer's float buffer is now always in linear space, even if it was
created from 16bit byte images.
- Space of byte buffer is stored in image buffer's rect_colorspace property.
- Profile of image buffer was removed since it's not longer meaningful.
- OpenGL and GLSL is supposed to always work in sRGB space. It is possible
to support other spaces, but it's quite large project which isn't so
much important.
- Legacy Color Management option disabled is emulated by using None display.
It could have some regressions, but there's no clear way to avoid them.
- If OpenColorIO is disabled on build time, it should make blender behaving
in the same way as previous release with color management enabled.
More details could be found at this page (more details would be added soon):
http://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.64/Color_Management
--
Thanks to Xavier Thomas, Lukas Toene for initial work on OpenColorIO
integration and to Brecht van Lommel for some further development and code/
usecase review!
One important thing to keep in mind when using this feature is that you'll need to flip your textures vertically (both the GIMP and Photoshop DDS tools I've seen have support for this on export). This is a quirk in using a texture format originally made for DirectX/DirectDraw, and flipping the compressed data is a real headache. Another quick fix for this issue is to change the Y value for the Size in the Mapping panel in the Texture properties to -1 (default is 1).
Implementation of cache for general movie-related areas
such as sequencer and clip editor (in the future)
Some changes in limiter were necessary:
- Limiter counted mapped memory twice when was checking how many memory is used.
- It was using "global" memory usage not memory usage by cached elements.
It will cause big problems when there's large mesh or plenty of undo steps are
in memory nothing would be cached in sequencer.
- To solve this problem introduced "callback" to measure cached element size.
It could be not very accurate in general, but it works well for image buffers.
And if this callback isn't set old-school memory usage check would be used.
- The whole cache used to get freed when memory limit exceeded, now it'll drop only
as much elements as necessary to reduce memory usage.
Seqcache is switched to use this new cache code.
This patch adds:
* support for proxy building again (missing feature from Blender 2.49)
additionally to the way, Blender 2.49 worked, you can select several
strips at once and make Blender build proxies in the background (using
the job system)
Also a new thing: movie proxies are now build into AVI files, and
the proxy system is moved into ImBuf-library, so that other parts
of blender can also benefit from it.
* Timecode support: to fix seeking issues with files, that have
a) varying frame rates
b) very large GOP lengths
c) are broken inbetween
d) use different time code tracks
the proxy builder can now also build timecode indices, which are
used (optionally) for seeking.
For the first time, it is possible, to do frame exact seeking on
all file types.
* Support for different video-streams in one video file (can be
selected in sequencer, other parts of blender can also use it,
but UI has to be added accordingly)
* IMPORTANT: this patch *requires* ffmpeg 0.7 or newer, since
older versions don't support the pkt_pts field, that is essential
for building timecode indices.
Windows and Mac libs are already updated, Linux-users have to build
their own ffmpeg verions until distros keep up.
- fix mistake with grease pencil UI (&& was intended but & used).
- use (void) rather then () across _all_ blenders code.
- a few minor edits, don't shadow stack variables in roll calculation & avoid running memset() for VBO vertex map.
Adding a byte rect to float ImBuf was always freeing the mipmap levels.
Removed this convention since it crashes renders + image texture draw.
Proper ownership handling of ImBuf is high on the wish list :)
Three code fixes for 1 report. User experienced crashes while
painting on float buffer + having preview renders on.
- Texture Nodes: Image was re-allocated without using
proper thread lock
- Paint code: old convention to free the byte rect from
a float image as signal to re-create now is a proper
flag. This keeps image memory unchanged. Nice for render.
- Imbuf: call to make a byte rect from float was freeing
mipmaps unnecessary.
