1 Movit 1.6.1, January 31st, 2018
3 - Loosen up some restrictions on strong one-to-one-effects. In particular,
4 this fixes a bug with MixEffect and compute shaders that hit some
8 Movit 1.6.0, January 24th, 2018
10 - Support for effects that work as compute shaders. Compute shaders are
11 generally slower than fragment shaders for the same algorithm,
12 but allow some forms of communication between shader invocations
13 and have more flexible output, which can enable more efficient algorithms.
14 See effect.h for more details. Note that the fastest rendering API on
15 EffectChain is now to a texture if possible, not to an FBO. This will
16 only matter if the last effect is a compute shader.
18 - Movit now includes a compute shader implementation of DeinterlaceEffect,
19 which is automatically used instead of the fragment shader implementation
20 if your GPU and OpenGL driver supports it (in practice, this means on
21 all platforms except on macOS). The compute shader version is typically
22 20–80% faster than the fragment shader version, depending on your GPU
25 A compute shader implementation of ResampleEffect was written but
26 ultimately failed to be faster, and so is not included.
28 - Support for microbenchmarks of effects through the Google microbenchmarking
29 framework (optional). Currently, DeinterlaceEffect and ResampleEffect has
30 benchmarks; enable them by running the unit test with --benchmark (also try
33 - Effects can now explicitly request _not_ to have mipmaps, which means they
34 can do so without needing to request bounce and fiddling with the sampler
35 state. Note that this is an API change for effects.
37 - Movit now requires C++11, both to build and to #include the header files.
38 Support for SDL1 has been dropped; unit tests and the demo program now need
41 - Various smaller bugfixes and optimizations.
44 Movit 1.5.3, August 10th, 2017
49 Movit 1.5.2, July 5th, 2017
51 - The texture generation for ResampleEffect (when the size is changed)
54 - Work around an NVIDIA driver threading bug by aggressively caching VAOs.
57 Movit 1.5.1, May 29th, 2017
59 - Allow YCbCrInput to change input format after finalize.
61 - Some minor YCbCrInput bugfixes.
64 Movit 1.5.0, March 21st, 2017
66 - Support interleaved Y'CbCr input (4:4:4 in a single texture).
68 - Support 10-bit and 12-bit Y'CbCr, both for input and output. For planar,
69 these are supported packed in 16-bit ints; for interleaved, 10:10:10:2 is
70 supported. (Efficient conversion to and from v210, ie. 10-bit 4:2:2,
71 is possible using compute shaders, but Movit does not include support
72 for them at the current point.) Note that this now means the num_levels
73 flag in YCbCrFormat actually matters, although 0 will be interpreted
74 as 256 (8-bit) for the benefit of older applications.
76 - Limited support for having multiple Y'CbCr outputs from a chain.
78 - Allow changing the Y'CbCr output coefficients runtime, ie., after finalize.
80 - Fix an issue where the last pass would have been rendered with the sRGB
81 flag set, which confused Qt applications running in certain NVIDIA
85 Movit 1.4.0, November 5th, 2016
87 - Allow setting the intermediate format for chains, instead of hardcoding
88 it at 16-bit RGBA; advanced users can use this to e.g. ask for 8-bit
89 sRGB intermediates, reducing the amount of memory bandwidth needed
90 at the cost of reduced precision. Whether this is a good tradeoff or not
91 depends on the exact chain and your requirements.
93 - Fix an issue where a (cached) shader program could be used from multiple
94 threads at a time, causing the uniforms to contain unpredictable values.
96 - Make the error printed on check_error() slightly friendlier: Include the
97 enum if possible, and print it to stderr instead of stdout.
100 Movit 1.3.2, February 23rd, 2016
102 - Fix an issue with initialization in certain locales. Patch from
103 Jean-Baptiste Mardelle.
106 Movit 1.3.1, February 15th, 2016
108 - Fix an issue where certain effect chains (particularly involving
109 out-of-tree effects that return only a constant color) could cause
110 texture coordinates not to be set properly. (The new code probably
111 also has slightly lower OpenGL driver overhead.) Reported by
112 Christophe Thommeret.
115 Movit 1.3.0, January 31st, 2016
117 - Movit now requires GLSL 1.30 (so a driver from 2008 or later);
118 before, it claimed to support 1.10, but actually used 1.30 features.
119 Note that some OpenGL drivers, in particular on OS X, only supports
120 GLSL 1.30 (actually, 1.50) if you have an OpenGL core context.
122 - Add a deinterlacer, based on YADIF.
124 - Allow parallel output to RGBA and Y'CbCr textures at the same time.
126 - Make FlatInput and YCbCrInput support taking in external OpenGL
127 textures. Also allow them to change width/height after instantiation.
129 - Various compatibility and performance fixes. In particular,
130 fp16 conversion on older (non-Haswell) CPUs is much faster, due to
131 new conversion code by Fabian Giesen.
134 Movit 1.2.0, September 24th, 2015
136 - Movit now ships a version.h with a #define MOVIT_VERSION that
137 increases on every API change, including in git. (The number
138 itself carries no semantic meaning beyond this.) Movit 1.2.0
139 is not API or ABI compatible with 1.1.x, so there has been a
140 soname bump to libmovit.so.3.
142 - More flexible Y'CbCr input; in particular, a special input
143 for 4:2:2 interleaved data (UYVY), and support for semi-planar
144 input (Cb and Cr in the same texture, like in NV12). Note that
145 you should now also set the new “num_levels” member of
146 YCbYCbCrFormat to 256 (signifying 8-bit input), although
147 it is not used yet; this is a stepping stone towards supporting
148 10- and 12-bit Y'CbCr.
150 - Basic support for Y'CbCr _output_. Currently only 8-bit,
151 and only 4:4:4 (ie., you'll need to subsample yourself
152 afterwards). It is possible to split the output into multiple
153 textures, though, if you want a luma/chroma split (NV12-like)
156 - Support top-left origin for output. This is only really useful
157 if you are rendering directly into some memory area with top-left
158 origin; most users will get the expected behavior by using
159 bottom-left as before.
161 - Rework uniform handling for less OpenGL overhead. Note that
162 this means your effects now need to register uniforms in the
163 C++ code instead of declaring them in the .frag file.
165 - Make the PaddingEffect border subpixel-aware, and also support
166 an arbitrary (subpixel) border offset. This means that you can
167 compose PaddingEffect with an integral left/top offset
168 (use the new IntegralPaddingEffect for potentially more speed)
169 and ResampleEffect with a subpixel left/top offset (and zoom to
170 compensate) to get Lanczos interpolation for the pixels and a
171 simple bilinear interpolation for the border itself.
173 - Fix a bug that could cause very bad filter weights in
176 - Various performance improvements, in particular with regards to
177 CPU usage in ResampleEffect.
180 Movit 1.1.3, March 29th, 2015
182 - Fix accuracy issues in ResampleEffect, particularly when
183 zooming. Note that this might cost a few percent performance.
184 Reported by Christophe Thommeret.
186 - Make the number of BlurEffect taps configurable.
187 Based on patch by Christophe Thommeret.
189 - Multiple other bugfixes, some courtesy Dan Dennedy and
190 Christophe Thommeret. In particular, thread-safety fixes
191 related to locales and chain finalization.
194 Movit 1.1.2, August 12th, 2014
196 - Performance bugfix: Fix texture freelist behavior so that it's
197 LRU instead of the exact opposite. Patch by Christophe Thommeret.
199 - Performance bugfix: Correct the number of blur taps read in the
200 blur filter (it was reading about twice as many as it should).
201 Also found by Christophe Thommeret.
204 Movit 1.1.1, April 12th, 2014
206 - Fix an issue that could cause assertion failure in ResourcePool
207 when FBOs are reused, especially with NVIDIA's drivers.
210 Movit 1.1, April 10th, 2014
212 - NOTE: Movit now uses libepoxy rather than GLEW. This is needed because
213 of core context and GLES3 support (see below). Movit 1.1 should be
214 generally API-compatible with 1.0.3 (with the exception of the next
215 point), but is not ABI-compatible, so there has been a soname bump
218 - NOTE: The rules for using an EffectChain or ResourcePool in multiple
219 OpenGL contexts or threads have changed somewhat; see the comments in
222 - Movit will now work in OpenGL core contexts, and on GLES 3.0 or newer.
223 (GLES2 devices are not supported.) This also holds for all unit tests,
224 provided that you use SDL2 and not SDL1 (because SDL2 is needed to set
225 up such contexts). Note that the included demo application still requires
226 a classic OpenGL context.
228 - ResampleEffect can now do sub-pixel translate and/or zoom.
230 - LumaMixEffect now has an “invert” flag, as a convenience to e.g. change
231 a left-to-right wipe into a right-to-left one.
233 - Significant reduction in driver overhead, especially on NVIDIA drivers.
235 - Various smaller bugfixes and performance improvements.
238 Movit 1.0.3, March 16th, 2014
240 - Yet more build system tweaks mainly related to distribution packaging.
243 Movit 1.0.2, March 16th, 2014
245 - Make a few tweaks to “make install”, to make distributions' lives easier.
248 Movit 1.0.1, March 16th, 2014
250 - Fix so that shared libraries are built.
253 Movit 1.0, March 16th, 2014