movit
2 years agoRelease Movit 1.4.0. 1.4.0
Steinar H. Gunderson [Sat, 5 Nov 2016 10:45:46 +0000 (11:45 +0100)]
Release Movit 1.4.0.

2 years agoFix a typo.
Steinar H. Gunderson [Sat, 5 Nov 2016 10:41:46 +0000 (11:41 +0100)]
Fix a typo.

2 years agoFix an issue where a (cached) shader program could be used from multiple
Steinar H. Gunderson [Sat, 5 Nov 2016 10:32:56 +0000 (11:32 +0100)]
Fix an issue where a (cached) shader program could be used from multiple
threads at a time.

This isn't allowed, since uniforms belong to the program, not to the
context. Found by running Helgrind over Nageru.

3 years agoMake the error printed on check_error() slightly friendlier: Include the enum if...
Steinar H. Gunderson [Tue, 26 Jul 2016 13:30:56 +0000 (15:30 +0200)]
Make the error printed on check_error() slightly friendlier: Include the enum if possible, and print it to stderr instead of stdout.

3 years agoBe more defensive about width/height/pitch given to FlatInput and YCbCrInput.
Steinar H. Gunderson [Thu, 21 Jul 2016 22:40:50 +0000 (00:40 +0200)]
Be more defensive about width/height/pitch given to FlatInput and YCbCrInput.

3 years agoRemove a very old and outdated comment.
Steinar H. Gunderson [Fri, 1 Apr 2016 21:51:28 +0000 (23:51 +0200)]
Remove a very old and outdated comment.

3 years agoAdd some clarifying comments about the intermediate formats.
Steinar H. Gunderson [Fri, 1 Apr 2016 21:47:30 +0000 (23:47 +0200)]
Add some clarifying comments about the intermediate formats.

3 years agoFix some comment formatting.
Steinar H. Gunderson [Tue, 1 Mar 2016 00:00:51 +0000 (01:00 +0100)]
Fix some comment formatting.

3 years agoRework PaddingEffect alpha handling, which also fixes a long-standing assertion failure.
Steinar H. Gunderson [Mon, 29 Feb 2016 23:59:37 +0000 (00:59 +0100)]
Rework PaddingEffect alpha handling, which also fixes a long-standing assertion failure.

3 years agoAllow storing values in intermediate framebuffers as sqrt(x).
Steinar H. Gunderson [Sun, 28 Feb 2016 00:53:55 +0000 (01:53 +0100)]
Allow storing values in intermediate framebuffers as sqrt(x).

Together with GL_RGB10_A2, this would seem to be an even better tradeoff for
many chains than GL_SRGB8_ALPHA8 is, as long as you don't need intermediate
alpha. (We verify its accuracy with a unit test.)

This changes the API for specifying intermediate framebuffers, but that API
was never in a release, so it should be fine.

Also document a rather obscure problem where, if you can actually hold on to
non-linear values across a bounce buffer, you don't really want to store them
in sRGB encoding. (The square-root version actually avoids this problem.
I guess we could snoop on the type and do a similar thing if we see it's an
GL_SRGB* encoding, but it seems so obscure that we can ignore it for now.)

3 years agoHard-assert on something that has bitten me too many times now.
Steinar H. Gunderson [Sun, 28 Feb 2016 00:46:15 +0000 (01:46 +0100)]
Hard-assert on something that has bitten me too many times now.

3 years agoFix a typo.
Steinar H. Gunderson [Sun, 28 Feb 2016 00:26:47 +0000 (01:26 +0100)]
Fix a typo.

3 years agoAdd support for 10-bit RGB framebuffers.
Steinar H. Gunderson [Sat, 27 Feb 2016 23:37:01 +0000 (00:37 +0100)]
Add support for 10-bit RGB framebuffers.

3 years agoAdd some sRGB conversion functions to test_util.{h,cpp}.
Steinar H. Gunderson [Sat, 27 Feb 2016 23:28:31 +0000 (00:28 +0100)]
Add some sRGB conversion functions to test_util.{h,cpp}.

3 years agoBump version after new format support.
Steinar H. Gunderson [Sat, 27 Feb 2016 23:24:04 +0000 (00:24 +0100)]
Bump version after new format support.

3 years agoAdd support for some of the more esoteric minifloat formats to ResourcePool.
Steinar H. Gunderson [Wed, 24 Feb 2016 22:18:41 +0000 (23:18 +0100)]
Add support for some of the more esoteric minifloat formats to ResourcePool.

3 years agoDo not send NULL to glTexSubImage2D if there is no input data set; it is illegal...
Steinar H. Gunderson [Wed, 24 Feb 2016 00:32:49 +0000 (01:32 +0100)]
Do not send NULL to glTexSubImage2D if there is no input data set; it is illegal, and NVIDIA's drivers crash on it.

3 years agoMerge branch '1.3.x-release'
Steinar H. Gunderson [Tue, 23 Feb 2016 21:48:26 +0000 (22:48 +0100)]
Merge branch '1.3.x-release'

3 years agoRelease Movit 1.3.2. (From a branch, since I do not want to break ABI compatibility... 1.3.x-release 1.3.2
Steinar H. Gunderson [Tue, 23 Feb 2016 21:47:05 +0000 (22:47 +0100)]
Release Movit 1.3.2. (From a branch, since I do not want to break ABI compatibility at this point.)

3 years agoMake the sRGB intermediate test slightly more stringent (so that e.g. 0.0 will not...
Steinar H. Gunderson [Tue, 23 Feb 2016 21:41:33 +0000 (22:41 +0100)]
Make the sRGB intermediate test slightly more stringent (so that e.g. 0.0 will not work).

3 years agoMake sRGBIntermediate test less sensitive to the exact sRGB choices; fixes unit test...
Steinar H. Gunderson [Mon, 22 Feb 2016 20:34:05 +0000 (21:34 +0100)]
Make sRGBIntermediate test less sensitive to the exact sRGB choices; fixes unit test on NVIDIA.

3 years agoRequire OpenGL 3.0 unconditionally; this is a no-op, since we already required GLSL...
Steinar H. Gunderson [Sat, 20 Feb 2016 16:28:54 +0000 (17:28 +0100)]
Require OpenGL 3.0 unconditionally; this is a no-op, since we already required GLSL 1.30 (part of OpenGL 3.0).

3 years agoAllow setting the intermediate texture format; useful for reducing bandwidth at the...
Steinar H. Gunderson [Sat, 20 Feb 2016 16:13:48 +0000 (17:13 +0100)]
Allow setting the intermediate texture format; useful for reducing bandwidth at the expense of quality, and possibly future GLES2 support.

3 years agoMake timer query objects polled asynchronously, so that the CPU blocks less on the...
Steinar H. Gunderson [Sat, 20 Feb 2016 14:42:19 +0000 (15:42 +0100)]
Make timer query objects polled asynchronously, so that the CPU blocks less on the GPU when doing timing. ABI break.

3 years agoRemove an extern definition that no longer exists.
Steinar H. Gunderson [Sat, 20 Feb 2016 13:19:58 +0000 (14:19 +0100)]
Remove an extern definition that no longer exists.

3 years agoCorrect/update a comment on rounding.
Steinar H. Gunderson [Sat, 20 Feb 2016 13:19:19 +0000 (14:19 +0100)]
Correct/update a comment on rounding.

3 years agoMake gamma polynomial constants into an array; slightly fewer uniforms to set, and...
Steinar H. Gunderson [Sat, 20 Feb 2016 13:11:35 +0000 (14:11 +0100)]
Make gamma polynomial constants into an array; slightly fewer uniforms to set, and it makes sense overall, since they belong so much together.

3 years agoFix initialisation on locale with comma as numerical separator
Jean-Baptiste Mardelle [Wed, 17 Feb 2016 22:20:48 +0000 (23:20 +0100)]
Fix initialisation on locale with comma as numerical separator

3 years agoRelease Movit 1.3.1. 1.3.1
Steinar H. Gunderson [Tue, 16 Feb 2016 01:40:02 +0000 (17:40 -0800)]
Release Movit 1.3.1.

3 years agoAdd deinterlace_effect_test from .gitignore. Reported by Christophe Thommeret.
Steinar H. Gunderson [Tue, 9 Feb 2016 13:20:09 +0000 (05:20 -0800)]
Add deinterlace_effect_test from .gitignore. Reported by Christophe Thommeret.

3 years ago2016 README updates.
Steinar H. Gunderson [Tue, 9 Feb 2016 04:26:37 +0000 (05:26 +0100)]
2016 README updates.

3 years agoRemove GL_GLEXT_PROTOTYPES from some files, since it is irrelevant with epoxy.
Steinar H. Gunderson [Sun, 7 Feb 2016 12:05:39 +0000 (13:05 +0100)]
Remove GL_GLEXT_PROTOTYPES from some files, since it is irrelevant with epoxy.

3 years agoRemove the check for movit_shader_rounding_supported, as we now demand 1.30 unconditi...
Steinar H. Gunderson [Sun, 7 Feb 2016 11:15:07 +0000 (12:15 +0100)]
Remove the check for movit_shader_rounding_supported, as we now demand 1.30 unconditionally.

3 years agoDo not bother with unbinding vertex attributes; that is automatically done when we...
Steinar H. Gunderson [Sun, 7 Feb 2016 11:12:03 +0000 (12:12 +0100)]
Do not bother with unbinding vertex attributes; that is automatically done when we unbind the VAO.

3 years agoRemove a few unneeded shader program switches.
Steinar H. Gunderson [Sun, 7 Feb 2016 01:18:48 +0000 (02:18 +0100)]
Remove a few unneeded shader program switches.

3 years agoOptimize VAO/VBO usage for minimal state changes.
Steinar H. Gunderson [Sun, 7 Feb 2016 01:12:14 +0000 (02:12 +0100)]
Optimize VAO/VBO usage for minimal state changes.

This is similar to what we had earlier to just reuse the VAO,
but now with correct bindings no matter what vertex attributes
are assigned to what index, so that the (new) test passes.

This is actually slightly more efficient than what we had before,
since we don't look up the attributes by text anymore, and don't
reupload the VBO for each frame anymore. In practice, the effects
should be small.

3 years agoRevert "Reuse the VAO across all phases."
Steinar H. Gunderson [Sat, 6 Feb 2016 23:11:51 +0000 (00:11 +0100)]
Revert "Reuse the VAO across all phases."

The patch trickles a bug where, if the first phase doesn't need texture
coordinates, the rest of the phases don't get it either. (Or more generally,
if the vertex shader varying indices are not predictable, the patch does
the wrong thing.) Add a unit test and revert it for now; in time, we'll find a
way that's both low-overhead (the patch fixes a real problem) _and_ correct in
these cases.

This reverts patch 5e34f7a8969f4afc169f034d34fb908019b3a389.

Reported by Christophe Thommeret.

3 years agoRelease Movit 1.3.0. 1.3.0
Steinar H. Gunderson [Sun, 31 Jan 2016 12:37:24 +0000 (13:37 +0100)]
Release Movit 1.3.0.

3 years agoIn OverlayEffect, add support for swapping the inputs.
Steinar H. Gunderson [Sun, 31 Jan 2016 12:35:38 +0000 (13:35 +0100)]
In OverlayEffect, add support for swapping the inputs.

3 years agoUse ryg's much faster fp16 conversion code.
Steinar H. Gunderson [Mon, 25 Jan 2016 21:51:09 +0000 (22:51 +0100)]
Use ryg's much faster fp16 conversion code.

3 years agoMake all fp16 routines work with fp32 as input instead of fp64, since that is what...
Steinar H. Gunderson [Fri, 15 Jan 2016 23:57:00 +0000 (00:57 +0100)]
Make all fp16 routines work with fp32 as input instead of fp64, since that is what hardware supports anyway.

3 years agoUnbreak some effects that were broken by 0c821b2e.
Steinar H. Gunderson [Thu, 24 Dec 2015 21:31:47 +0000 (22:31 +0100)]
Unbreak some effects that were broken by 0c821b2e.

3 years agoMake shader generation more deterministic by removing a sort of pointers.
Steinar H. Gunderson [Thu, 24 Dec 2015 15:43:23 +0000 (16:43 +0100)]
Make shader generation more deterministic by removing a sort of pointers.

3 years agoIncrease version number after YADIF.
Steinar H. Gunderson [Thu, 24 Dec 2015 12:31:38 +0000 (13:31 +0100)]
Increase version number after YADIF.

3 years agoAdd a deinterlacer based on YADIF.
Steinar H. Gunderson [Sun, 22 Nov 2015 17:34:26 +0000 (18:34 +0100)]
Add a deinterlacer based on YADIF.

I tried a few different things before I finally settled on this, in particular
Weston's 3-field deinterlacer (w3fdif). It's not perfect (see .h comments),
but it works overall pretty well.

3 years agoAdd even long convenience overloads for add_effect().
Steinar H. Gunderson [Mon, 21 Dec 2015 13:52:34 +0000 (14:52 +0100)]
Add even long convenience overloads for add_effect().

3 years agoMake register_int call register_uniform_int, now that we require GLSL 1.30.
Steinar H. Gunderson [Mon, 21 Dec 2015 12:46:23 +0000 (13:46 +0100)]
Make register_int call register_uniform_int, now that we require GLSL 1.30.

3 years agoRevert "Add a hack to use #version 110 but keep using 130 features, for the benefit...
Steinar H. Gunderson [Mon, 14 Dec 2015 19:54:41 +0000 (20:54 +0100)]
Revert "Add a hack to use #version 110 but keep using 130 features, for the benefit of OS X."

This turned out not to work well on OS X after all.

This reverts commit e0811ddf51aeb50575fb5f7d9c6e32b92a6bac0d.

3 years agoFix a crash in a unit test if nb_NO.UTF-8 is not available.
Steinar H. Gunderson [Mon, 14 Dec 2015 01:10:00 +0000 (02:10 +0100)]
Fix a crash in a unit test if nb_NO.UTF-8 is not available.

3 years agoUse libpng instead of the older libpng12; seemingly fixes compilation on some systems.
Steinar H. Gunderson [Mon, 14 Dec 2015 00:45:46 +0000 (01:45 +0100)]
Use libpng instead of the older libpng12; seemingly fixes compilation on some systems.

3 years agoFix some stack overflows in unit tests; found with asan.
Steinar H. Gunderson [Sun, 13 Dec 2015 14:21:12 +0000 (15:21 +0100)]
Fix some stack overflows in unit tests; found with asan.

3 years agoFix a stack overflow in ResampleEffectTest.
Steinar H. Gunderson [Sun, 13 Dec 2015 13:56:15 +0000 (14:56 +0100)]
Fix a stack overflow in ResampleEffectTest.

3 years agoWork around a rounding precision issue that would cause spurious test failures on...
Steinar H. Gunderson [Sun, 13 Dec 2015 13:13:32 +0000 (14:13 +0100)]
Work around a rounding precision issue that would cause spurious test failures on AMD.

3 years agoFix a double scaling issue in Y'CbCr conversion.
Steinar H. Gunderson [Sun, 13 Dec 2015 12:26:01 +0000 (13:26 +0100)]
Fix a double scaling issue in Y'CbCr conversion.

We multiplied by 224/219 once too many, causing some small accuracy issues.
Furthermore, we also did this for full-range Y'CbCr, which obviously is wrong.
The issue was so small that the unit tests kept on passing (its investigation
was prompted by a test that failed on AMD cards, which is a separate issue).

After this, the Rec. 601 matrices match Wikipedia exactly, both for limited
and full range. Added unit tests for this.

3 years agoExplicitly bind fragment shader outputs in order.
Steinar H. Gunderson [Sat, 12 Dec 2015 13:31:40 +0000 (14:31 +0100)]
Explicitly bind fragment shader outputs in order.

Evidently ATI drivers use the freedom the standard gives them to assign
these in another order than they are specified in the shader source,
so we need to explicitly bind them, or YCbCrConversionEffectTest will fail
in the multi-output tests.

Originally reported by Iwan Gabovitch.

3 years agoAdd a hack to use #version 110 but keep using 130 features, for the benefit of OS X.
Steinar H. Gunderson [Sat, 12 Dec 2015 11:27:21 +0000 (12:27 +0100)]
Add a hack to use #version 110 but keep using 130 features, for the benefit of OS X.

3 years agoStop linking widgets.o into the shared library.
Steinar H. Gunderson [Sun, 22 Nov 2015 22:13:03 +0000 (23:13 +0100)]
Stop linking widgets.o into the shared library.

This was never intended to be there, and we don't install headers for it
(so no API/ABI break); it is actively harmful because it has a static
ResourcePool, which is attempted destroyed during shutdown (which causes
use of uninitialized memory as we try to get the current context).

3 years agoAdd the missing two array uniform types.
Steinar H. Gunderson [Sun, 22 Nov 2015 13:34:56 +0000 (14:34 +0100)]
Add the missing two array uniform types.

3 years agoAllow setting width/height on FlatInput and YCbCrInput after instantiation.
Steinar H. Gunderson [Sat, 21 Nov 2015 20:32:29 +0000 (21:32 +0100)]
Allow setting width/height on FlatInput and YCbCrInput after instantiation.

3 years agoForgot to increment version.h for bounce override; doing so now.
Steinar H. Gunderson [Sun, 1 Nov 2015 15:08:49 +0000 (16:08 +0100)]
Forgot to increment version.h for bounce override; doing so now.

3 years agoAdd a function to let non-input effects override texture bounce.
Steinar H. Gunderson [Sun, 1 Nov 2015 01:09:56 +0000 (02:09 +0100)]
Add a function to let non-input effects override texture bounce.

Definitely read the comment before using; it is not for the faint
of heart. Also make ResampleEffect tolerate this kind of abuse.

3 years agoAdd some earlier check_error() calls so that we do not get confusing behavior if...
Steinar H. Gunderson [Sun, 1 Nov 2015 01:02:10 +0000 (02:02 +0100)]
Add some earlier check_error() calls so that we do not get confusing behavior if there is already error on the entrance to render_to_fbo().

3 years agoInstall identity.frag; it is needed for ResizeEffect.
Steinar H. Gunderson [Thu, 8 Oct 2015 19:13:44 +0000 (21:13 +0200)]
Install identity.frag; it is needed for ResizeEffect.

3 years agoFix another #if issue, this time in dither_effect.frag. Reported by Dan Dennedy.
Steinar H. Gunderson [Wed, 7 Oct 2015 19:03:24 +0000 (21:03 +0200)]
Fix another #if issue, this time in dither_effect.frag. Reported by Dan Dennedy.

3 years agoInstall the new GLSL 1.50 shaders.
Steinar H. Gunderson [Wed, 7 Oct 2015 18:21:08 +0000 (20:21 +0200)]
Install the new GLSL 1.50 shaders.

3 years agoMake the demo program run with core contexts.
Steinar H. Gunderson [Wed, 7 Oct 2015 18:12:25 +0000 (20:12 +0200)]
Make the demo program run with core contexts.

Also, if SDL2 is in use, actually _ask_ for a core context.

3 years agoAdd separate shaders for GLSL 1.50.
Steinar H. Gunderson [Wed, 7 Oct 2015 17:24:19 +0000 (19:24 +0200)]
Add separate shaders for GLSL 1.50.

Seemingly, Apple's drivers (in OS X) do not support GLSL 1.30 in
core contexts, only 1.50. Reported by Dan Dennedy.

3 years agoFix GLSL compilation errors on some drivers.
Steinar H. Gunderson [Tue, 6 Oct 2015 18:08:05 +0000 (20:08 +0200)]
Fix GLSL compilation errors on some drivers.

Evidently #if FOO is illegal in GLSL if FOO is not defined,
unlike in C++. Reported by Dan Dennedy.

3 years agoMake get_current_context_identifier() understand EGL.
Steinar H. Gunderson [Mon, 5 Oct 2015 23:24:26 +0000 (01:24 +0200)]
Make get_current_context_identifier() understand EGL.

If we're using EGL and not GLX (typically because we're using GLES,
but also increasingly with desktop GL), we'd always return NULL.
This could FBOs to be confused between contexts.

3 years agoDisable dither explicitly per frame; fixes some weird artifacts I found.
Steinar H. Gunderson [Mon, 5 Oct 2015 22:06:14 +0000 (00:06 +0200)]
Disable dither explicitly per frame; fixes some weird artifacts I found.

3 years agoCall init_lanczos_table() once instead of checking for it all the time.
Steinar H. Gunderson [Mon, 5 Oct 2015 21:08:35 +0000 (23:08 +0200)]
Call init_lanczos_table() once instead of checking for it all the time.

3 years agoGet rid of a bunch of STL inefficiencies in FBO freelist handling.
Steinar H. Gunderson [Mon, 5 Oct 2015 20:43:59 +0000 (22:43 +0200)]
Get rid of a bunch of STL inefficiencies in FBO freelist handling.

3 years agoSupport GL_RGB565 targets.
Steinar H. Gunderson [Mon, 5 Oct 2015 19:04:29 +0000 (21:04 +0200)]
Support GL_RGB565 targets.

3 years agoBump version number after support for external OpenGL textures (it was forgotten).
Steinar H. Gunderson [Mon, 5 Oct 2015 18:49:10 +0000 (20:49 +0200)]
Bump version number after support for external OpenGL textures (it was forgotten).

3 years agoMake FlatInput and YCbCrInput support taking in external OpenGL textures.
Steinar H. Gunderson [Mon, 5 Oct 2015 17:49:51 +0000 (19:49 +0200)]
Make FlatInput and YCbCrInput support taking in external OpenGL textures.

3 years agoUnbreak make install after the last changes.
Steinar H. Gunderson [Sun, 4 Oct 2015 00:44:13 +0000 (02:44 +0200)]
Unbreak make install after the last changes.

3 years agoSome small cleanups after we got rid of GLSL 1.10; we can now unify 1.30 and ES 3...
Steinar H. Gunderson [Sun, 4 Oct 2015 00:43:02 +0000 (02:43 +0200)]
Some small cleanups after we got rid of GLSL 1.10; we can now unify 1.30 and ES 3.00 some places.

3 years agoAllow dual Y'CbCr/RGBA outputs.
Steinar H. Gunderson [Sun, 4 Oct 2015 00:37:56 +0000 (02:37 +0200)]
Allow dual Y'CbCr/RGBA outputs.

The intended use case is to have Y'CbCr for encoding output but keep
RGBA around for easier preview. This causes a few effects to need to
send arrays around; it's a bit ugly to special-case them like this,
but I'm concerned about going generic wrt. how good various shader
compilers are to optimize if we went full multi-model everywhere
(without having tested, though).

ABI break due to changed EffectChain size.

3 years agoRemove support for GLSL 1.10.
Steinar H. Gunderson [Sun, 4 Oct 2015 00:05:33 +0000 (02:05 +0200)]
Remove support for GLSL 1.10.

In practice, we haven't _actually_ supported this since we used integers
in ResampleEffect (and ResampleEffect is a pretty central effect),
so let's be honest with ourselves. (Also, we will soon start using arrays
in some cases, which are cumbersome pre-1.30.) I don't know of any drivers
that support all the other stuff we want but not GLSL 1.30 anyway;
it came with OpenGL 3.0, in 2008.

This actually isn't an ABI break, at least not on the C++ level.

3 years agoIn ycbcr_conversion_effect_test, use a non-float framebuffer.
Steinar H. Gunderson [Fri, 25 Sep 2015 23:33:29 +0000 (01:33 +0200)]
In ycbcr_conversion_effect_test, use a non-float framebuffer.

This way, we let the card convert float-to-int, which we have reasonable
control over, as opposed to glReadPixels(), which is rather unpredictable.
Fixes unit test failures on Broadwell on Linux (Mesa 10.1).

3 years agoFix a buffer overflow in ycbcr_conversion_effect_test.
Steinar H. Gunderson [Fri, 25 Sep 2015 23:29:40 +0000 (01:29 +0200)]
Fix a buffer overflow in ycbcr_conversion_effect_test.

3 years agoRelease Movit 1.2.0. 1.2.0
Steinar H. Gunderson [Thu, 24 Sep 2015 16:44:01 +0000 (18:44 +0200)]
Release Movit 1.2.0.

3 years agoIn ResampleEffect, precompute the Lanczos function into a table.
Steinar H. Gunderson [Thu, 24 Sep 2015 00:12:40 +0000 (02:12 +0200)]
In ResampleEffect, precompute the Lanczos function into a table.

A 2048-element table (with linear interpolation between the elements)
is seemingly enough to get down to beyond float epsilon, and this
saves a lot of CPU time when computing large filter kernels.

3 years agoFix a bug where combined fp16 weights would be horribly wrong.
Steinar H. Gunderson [Wed, 23 Sep 2015 23:59:47 +0000 (01:59 +0200)]
Fix a bug where combined fp16 weights would be horribly wrong.

Seemingly weights were always returned as float, and then cast
to fp16_int_t -- without proper conversion! And sum_sq_error
would be calculated based on the correct value, not the broken-
casted one.

It's a small miracle the unit tests didn't catch this; they didn't
until I started introducing small errors for another reason.
Most real-world testing seems to have hit fp32, and thus this
wasn't caught there either.

Also make fp16_int_t a struct so that it is not implicitly
convertible to/from numeric types, so this never ever can happen again.

3 years agoMicrooptimization in ResampleEffect.
Steinar H. Gunderson [Tue, 22 Sep 2015 23:34:58 +0000 (01:34 +0200)]
Microoptimization in ResampleEffect.

3 years agoAdd a mode for YCbCrInput where Cb and Cr are in the same texture.
Steinar H. Gunderson [Thu, 17 Sep 2015 23:48:10 +0000 (01:48 +0200)]
Add a mode for YCbCrInput where Cb and Cr are in the same texture.

3 years agoIn ResampleEffect, be more aggressive about giving up on saving bilinear samples.
Steinar H. Gunderson [Thu, 17 Sep 2015 18:05:43 +0000 (20:05 +0200)]
In ResampleEffect, be more aggressive about giving up on saving bilinear samples.

It turns out that for some kinds of loads, we can't use bilinearity at all
to our benefit, so we spend almost all of our time trying to go through
each line to see how much we can save. Simply send in the minimum number
so far when doing this evaluation to begin with, which means we'll effectively
short-circuit the entire thing pretty fast once we find a line that can save
nothing.

3 years agoReduce the amount of computation in combine_two_samples().
Steinar H. Gunderson [Thu, 17 Sep 2015 17:28:43 +0000 (19:28 +0200)]
Reduce the amount of computation in combine_two_samples().

Mostly microoptimization, but seemingly this function is somewhat expensive.

3 years agoAdd support for overriding the output origin.
Steinar H. Gunderson [Wed, 16 Sep 2015 22:39:43 +0000 (00:39 +0200)]
Add support for overriding the output origin.

For me, this was needed when I wanted to render directly into
VA-API's encoder buffers, which are always top-left origin (and FBOs
are always bottom-left origin).

3 years agoAdd support for Y'CbCr output split between multiple textures.
Steinar H. Gunderson [Wed, 16 Sep 2015 21:51:30 +0000 (23:51 +0200)]
Add support for Y'CbCr output split between multiple textures.

This is useful primarily for avoiding copies in later stages;
e.g., when rendering directly into a video encoder buffer.
We support both full planar and NV12-style interleaved Cb+Cr.
You still have to subsample chroma yourself, though; we don't
really support chains that diverge except in the final output node
(and changing resolution would definitely need a bounce;
and even worse, one in a non-fp16 intermediate format).

3 years agoMake EffectChainTester capable of testing chains with multiple outputs.
Steinar H. Gunderson [Wed, 16 Sep 2015 21:49:45 +0000 (23:49 +0200)]
Make EffectChainTester capable of testing chains with multiple outputs.

3 years agoIn EffectChainTester, there is no point in making the FBO and texture persistent.
Steinar H. Gunderson [Wed, 16 Sep 2015 21:04:45 +0000 (23:04 +0200)]
In EffectChainTester, there is no point in making the FBO and texture persistent.

3 years agoSmall refactoring in EffectChainTester.
Steinar H. Gunderson [Wed, 16 Sep 2015 21:00:10 +0000 (23:00 +0200)]
Small refactoring in EffectChainTester.

3 years agoSupport multi-texture FBOs in ResourcePool.
Steinar H. Gunderson [Wed, 16 Sep 2015 20:33:09 +0000 (22:33 +0200)]
Support multi-texture FBOs in ResourcePool.

4 years agoAdd some check_error() for shaders miscompiling.
Steinar H. Gunderson [Wed, 16 Sep 2015 18:02:30 +0000 (20:02 +0200)]
Add some check_error() for shaders miscompiling.

4 years agoReuse the VAO across all phases.
Steinar H. Gunderson [Sun, 13 Sep 2015 23:18:38 +0000 (01:18 +0200)]
Reuse the VAO across all phases.

4 years agoHelp the compiler out a tiny bit.
Steinar H. Gunderson [Sun, 13 Sep 2015 23:12:43 +0000 (01:12 +0200)]
Help the compiler out a tiny bit.

4 years agoReduce the boilerplate around uniforms a bit.
Steinar H. Gunderson [Sun, 13 Sep 2015 21:15:37 +0000 (23:15 +0200)]
Reduce the boilerplate around uniforms a bit.

4 years agoCleanup: Make uniforms for RTT samplers like all other uniforms.
Steinar H. Gunderson [Sun, 13 Sep 2015 18:40:58 +0000 (20:40 +0200)]
Cleanup: Make uniforms for RTT samplers like all other uniforms.

This also removes an ugly special-casing where one single place
in the entire code would call glUniform1i directly.

4 years agoHandle sampler2D uniforms specially.
Steinar H. Gunderson [Sun, 13 Sep 2015 18:15:29 +0000 (20:15 +0200)]
Handle sampler2D uniforms specially.

We're going to need this soon, since sampler uniforms are special
in that they cannot be in a uniform block.