2015-06-21 17:33:46 +02:00
|
|
|
/*
|
|
|
|
Simple DirectMedia Layer
|
2020-01-17 05:49:25 +01:00
|
|
|
Copyright (C) 1997-2020 Sam Lantinga <slouken@libsdl.org>
|
2015-06-21 17:33:46 +02:00
|
|
|
|
|
|
|
This software is provided 'as-is', without any express or implied
|
|
|
|
warranty. In no event will the authors be held liable for any damages
|
|
|
|
arising from the use of this software.
|
|
|
|
|
|
|
|
Permission is granted to anyone to use this software for any purpose,
|
|
|
|
including commercial applications, and to alter it and redistribute it
|
|
|
|
freely, subject to the following restrictions:
|
|
|
|
|
|
|
|
1. The origin of this software must not be misrepresented; you must not
|
|
|
|
claim that you wrote the original software. If you use this software
|
|
|
|
in a product, an acknowledgment in the product documentation would be
|
|
|
|
appreciated but is not required.
|
|
|
|
2. Altered source versions must be plainly marked as such, and must not be
|
|
|
|
misrepresented as being the original software.
|
|
|
|
3. This notice may not be removed or altered from any source distribution.
|
|
|
|
*/
|
|
|
|
#include "../SDL_internal.h"
|
|
|
|
|
|
|
|
/* This is the software implementation of the YUV texture support */
|
|
|
|
|
|
|
|
#include "SDL_assert.h"
|
|
|
|
|
2017-11-13 07:51:12 +01:00
|
|
|
#include "SDL_yuv_sw_c.h"
|
2015-06-21 17:33:46 +02:00
|
|
|
|
|
|
|
|
|
|
|
SDL_SW_YUVTexture *
|
|
|
|
SDL_SW_CreateYUVTexture(Uint32 format, int w, int h)
|
|
|
|
{
|
|
|
|
SDL_SW_YUVTexture *swdata;
|
|
|
|
|
|
|
|
switch (format) {
|
|
|
|
case SDL_PIXELFORMAT_YV12:
|
|
|
|
case SDL_PIXELFORMAT_IYUV:
|
|
|
|
case SDL_PIXELFORMAT_YUY2:
|
|
|
|
case SDL_PIXELFORMAT_UYVY:
|
|
|
|
case SDL_PIXELFORMAT_YVYU:
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
case SDL_PIXELFORMAT_NV12:
|
|
|
|
case SDL_PIXELFORMAT_NV21:
|
2015-06-21 17:33:46 +02:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
SDL_SetError("Unsupported YUV format");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
swdata = (SDL_SW_YUVTexture *) SDL_calloc(1, sizeof(*swdata));
|
|
|
|
if (!swdata) {
|
|
|
|
SDL_OutOfMemory();
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
swdata->format = format;
|
|
|
|
swdata->target_format = SDL_PIXELFORMAT_UNKNOWN;
|
|
|
|
swdata->w = w;
|
|
|
|
swdata->h = h;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
{
|
|
|
|
const int sz_plane = w * h;
|
|
|
|
const int sz_plane_chroma = ((w + 1) / 2) * ((h + 1) / 2);
|
|
|
|
const int sz_plane_packed = ((w + 1) / 2) * h;
|
|
|
|
int dst_size = 0;
|
|
|
|
switch(format)
|
|
|
|
{
|
|
|
|
case SDL_PIXELFORMAT_YV12: /**< Planar mode: Y + V + U (3 planes) */
|
|
|
|
case SDL_PIXELFORMAT_IYUV: /**< Planar mode: Y + U + V (3 planes) */
|
|
|
|
dst_size = sz_plane + sz_plane_chroma + sz_plane_chroma;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case SDL_PIXELFORMAT_YUY2: /**< Packed mode: Y0+U0+Y1+V0 (1 plane) */
|
|
|
|
case SDL_PIXELFORMAT_UYVY: /**< Packed mode: U0+Y0+V0+Y1 (1 plane) */
|
|
|
|
case SDL_PIXELFORMAT_YVYU: /**< Packed mode: Y0+V0+Y1+U0 (1 plane) */
|
|
|
|
dst_size = 4 * sz_plane_packed;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case SDL_PIXELFORMAT_NV12: /**< Planar mode: Y + U/V interleaved (2 planes) */
|
|
|
|
case SDL_PIXELFORMAT_NV21: /**< Planar mode: Y + V/U interleaved (2 planes) */
|
|
|
|
dst_size = sz_plane + sz_plane_chroma + sz_plane_chroma;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
SDL_assert(0 && "We should never get here (caught above)");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
swdata->pixels = (Uint8 *) SDL_malloc(dst_size);
|
2017-11-13 07:51:12 +01:00
|
|
|
if (!swdata->pixels) {
|
|
|
|
SDL_SW_DestroyYUVTexture(swdata);
|
|
|
|
SDL_OutOfMemory();
|
|
|
|
return NULL;
|
|
|
|
}
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
|
2017-11-13 07:51:12 +01:00
|
|
|
/* Find the pitch and offset values for the texture */
|
2015-06-21 17:33:46 +02:00
|
|
|
switch (format) {
|
|
|
|
case SDL_PIXELFORMAT_YV12:
|
|
|
|
case SDL_PIXELFORMAT_IYUV:
|
|
|
|
swdata->pitches[0] = w;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
swdata->pitches[1] = (swdata->pitches[0] + 1) / 2;
|
|
|
|
swdata->pitches[2] = (swdata->pitches[0] + 1) / 2;
|
2015-06-21 17:33:46 +02:00
|
|
|
swdata->planes[0] = swdata->pixels;
|
|
|
|
swdata->planes[1] = swdata->planes[0] + swdata->pitches[0] * h;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
swdata->planes[2] = swdata->planes[1] + swdata->pitches[1] * ((h + 1) / 2);
|
2015-06-21 17:33:46 +02:00
|
|
|
break;
|
|
|
|
case SDL_PIXELFORMAT_YUY2:
|
|
|
|
case SDL_PIXELFORMAT_UYVY:
|
|
|
|
case SDL_PIXELFORMAT_YVYU:
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
swdata->pitches[0] = ((w + 1) / 2) * 4;
|
2015-06-21 17:33:46 +02:00
|
|
|
swdata->planes[0] = swdata->pixels;
|
|
|
|
break;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
|
|
|
|
case SDL_PIXELFORMAT_NV12:
|
|
|
|
case SDL_PIXELFORMAT_NV21:
|
|
|
|
swdata->pitches[0] = w;
|
|
|
|
swdata->pitches[1] = 2 * ((swdata->pitches[0] + 1) / 2);
|
|
|
|
swdata->planes[0] = swdata->pixels;
|
|
|
|
swdata->planes[1] = swdata->planes[0] + swdata->pitches[0] * h;
|
|
|
|
break;
|
|
|
|
|
2015-06-21 17:33:46 +02:00
|
|
|
default:
|
|
|
|
SDL_assert(0 && "We should never get here (caught above)");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* We're all done.. */
|
|
|
|
return (swdata);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
SDL_SW_QueryYUVTexturePixels(SDL_SW_YUVTexture * swdata, void **pixels,
|
|
|
|
int *pitch)
|
|
|
|
{
|
|
|
|
*pixels = swdata->planes[0];
|
|
|
|
*pitch = swdata->pitches[0];
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
SDL_SW_UpdateYUVTexture(SDL_SW_YUVTexture * swdata, const SDL_Rect * rect,
|
|
|
|
const void *pixels, int pitch)
|
|
|
|
{
|
|
|
|
switch (swdata->format) {
|
|
|
|
case SDL_PIXELFORMAT_YV12:
|
|
|
|
case SDL_PIXELFORMAT_IYUV:
|
|
|
|
if (rect->x == 0 && rect->y == 0 &&
|
|
|
|
rect->w == swdata->w && rect->h == swdata->h) {
|
|
|
|
SDL_memcpy(swdata->pixels, pixels,
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
(swdata->h * swdata->w) + 2* ((swdata->h + 1) /2) * ((swdata->w + 1) / 2));
|
2015-06-21 17:33:46 +02:00
|
|
|
} else {
|
|
|
|
Uint8 *src, *dst;
|
|
|
|
int row;
|
|
|
|
size_t length;
|
|
|
|
|
|
|
|
/* Copy the Y plane */
|
|
|
|
src = (Uint8 *) pixels;
|
|
|
|
dst = swdata->pixels + rect->y * swdata->w + rect->x;
|
|
|
|
length = rect->w;
|
|
|
|
for (row = 0; row < rect->h; ++row) {
|
|
|
|
SDL_memcpy(dst, src, length);
|
|
|
|
src += pitch;
|
|
|
|
dst += swdata->w;
|
|
|
|
}
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
|
2015-06-21 17:33:46 +02:00
|
|
|
/* Copy the next plane */
|
|
|
|
src = (Uint8 *) pixels + rect->h * pitch;
|
|
|
|
dst = swdata->pixels + swdata->h * swdata->w;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
dst += rect->y/2 * ((swdata->w + 1) / 2) + rect->x/2;
|
|
|
|
length = (rect->w + 1) / 2;
|
|
|
|
for (row = 0; row < (rect->h + 1)/2; ++row) {
|
2015-06-21 17:33:46 +02:00
|
|
|
SDL_memcpy(dst, src, length);
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
src += (pitch + 1)/2;
|
|
|
|
dst += (swdata->w + 1)/2;
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy the next plane */
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
src = (Uint8 *) pixels + rect->h * pitch + ((rect->h + 1) / 2) * ((pitch + 1) / 2);
|
2015-06-21 17:33:46 +02:00
|
|
|
dst = swdata->pixels + swdata->h * swdata->w +
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
((swdata->h + 1)/2) * ((swdata->w+1) / 2);
|
|
|
|
dst += rect->y/2 * ((swdata->w + 1)/2) + rect->x/2;
|
|
|
|
length = (rect->w + 1) / 2;
|
|
|
|
for (row = 0; row < (rect->h + 1)/2; ++row) {
|
2015-06-21 17:33:46 +02:00
|
|
|
SDL_memcpy(dst, src, length);
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
src += (pitch + 1)/2;
|
|
|
|
dst += (swdata->w + 1)/2;
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case SDL_PIXELFORMAT_YUY2:
|
|
|
|
case SDL_PIXELFORMAT_UYVY:
|
|
|
|
case SDL_PIXELFORMAT_YVYU:
|
|
|
|
{
|
|
|
|
Uint8 *src, *dst;
|
|
|
|
int row;
|
|
|
|
size_t length;
|
|
|
|
|
|
|
|
src = (Uint8 *) pixels;
|
|
|
|
dst =
|
|
|
|
swdata->planes[0] + rect->y * swdata->pitches[0] +
|
|
|
|
rect->x * 2;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
length = 4 * ((rect->w + 1) / 2);
|
2015-06-21 17:33:46 +02:00
|
|
|
for (row = 0; row < rect->h; ++row) {
|
|
|
|
SDL_memcpy(dst, src, length);
|
|
|
|
src += pitch;
|
|
|
|
dst += swdata->pitches[0];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
case SDL_PIXELFORMAT_NV12:
|
|
|
|
case SDL_PIXELFORMAT_NV21:
|
|
|
|
{
|
|
|
|
if (rect->x == 0 && rect->y == 0 && rect->w == swdata->w && rect->h == swdata->h) {
|
|
|
|
SDL_memcpy(swdata->pixels, pixels,
|
|
|
|
(swdata->h * swdata->w) + 2* ((swdata->h + 1) /2) * ((swdata->w + 1) / 2));
|
|
|
|
} else {
|
|
|
|
|
|
|
|
Uint8 *src, *dst;
|
|
|
|
int row;
|
|
|
|
size_t length;
|
|
|
|
|
|
|
|
/* Copy the Y plane */
|
|
|
|
src = (Uint8 *) pixels;
|
|
|
|
dst = swdata->pixels + rect->y * swdata->w + rect->x;
|
|
|
|
length = rect->w;
|
|
|
|
for (row = 0; row < rect->h; ++row) {
|
|
|
|
SDL_memcpy(dst, src, length);
|
|
|
|
src += pitch;
|
|
|
|
dst += swdata->w;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy the next plane */
|
|
|
|
src = (Uint8 *) pixels + rect->h * pitch;
|
|
|
|
dst = swdata->pixels + swdata->h * swdata->w;
|
|
|
|
dst += 2 * ((rect->y + 1)/2) * ((swdata->w + 1) / 2) + 2 * (rect->x/2);
|
|
|
|
length = 2 * ((rect->w + 1) / 2);
|
|
|
|
for (row = 0; row < (rect->h + 1)/2; ++row) {
|
|
|
|
SDL_memcpy(dst, src, length);
|
|
|
|
src += 2 * ((pitch + 1)/2);
|
|
|
|
dst += 2 * ((swdata->w + 1)/2);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
SDL_SW_UpdateYUVTexturePlanar(SDL_SW_YUVTexture * swdata, const SDL_Rect * rect,
|
|
|
|
const Uint8 *Yplane, int Ypitch,
|
|
|
|
const Uint8 *Uplane, int Upitch,
|
|
|
|
const Uint8 *Vplane, int Vpitch)
|
|
|
|
{
|
|
|
|
const Uint8 *src;
|
|
|
|
Uint8 *dst;
|
|
|
|
int row;
|
|
|
|
size_t length;
|
|
|
|
|
|
|
|
/* Copy the Y plane */
|
|
|
|
src = Yplane;
|
|
|
|
dst = swdata->pixels + rect->y * swdata->w + rect->x;
|
|
|
|
length = rect->w;
|
|
|
|
for (row = 0; row < rect->h; ++row) {
|
|
|
|
SDL_memcpy(dst, src, length);
|
|
|
|
src += Ypitch;
|
|
|
|
dst += swdata->w;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy the U plane */
|
|
|
|
src = Uplane;
|
|
|
|
if (swdata->format == SDL_PIXELFORMAT_IYUV) {
|
|
|
|
dst = swdata->pixels + swdata->h * swdata->w;
|
|
|
|
} else {
|
|
|
|
dst = swdata->pixels + swdata->h * swdata->w +
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
((swdata->h + 1) / 2) * ((swdata->w + 1) / 2);
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
dst += rect->y/2 * ((swdata->w + 1)/2) + rect->x/2;
|
|
|
|
length = (rect->w + 1) / 2;
|
|
|
|
for (row = 0; row < (rect->h + 1)/2; ++row) {
|
2015-06-21 17:33:46 +02:00
|
|
|
SDL_memcpy(dst, src, length);
|
|
|
|
src += Upitch;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
dst += (swdata->w + 1)/2;
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy the V plane */
|
|
|
|
src = Vplane;
|
|
|
|
if (swdata->format == SDL_PIXELFORMAT_YV12) {
|
|
|
|
dst = swdata->pixels + swdata->h * swdata->w;
|
|
|
|
} else {
|
|
|
|
dst = swdata->pixels + swdata->h * swdata->w +
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
((swdata->h + 1) / 2) * ((swdata->w + 1) / 2);
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
dst += rect->y/2 * ((swdata->w + 1)/2) + rect->x/2;
|
|
|
|
length = (rect->w + 1) / 2;
|
|
|
|
for (row = 0; row < (rect->h + 1)/2; ++row) {
|
2015-06-21 17:33:46 +02:00
|
|
|
SDL_memcpy(dst, src, length);
|
|
|
|
src += Vpitch;
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
dst += (swdata->w + 1)/2;
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
SDL_SW_LockYUVTexture(SDL_SW_YUVTexture * swdata, const SDL_Rect * rect,
|
|
|
|
void **pixels, int *pitch)
|
|
|
|
{
|
|
|
|
switch (swdata->format) {
|
|
|
|
case SDL_PIXELFORMAT_YV12:
|
|
|
|
case SDL_PIXELFORMAT_IYUV:
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
case SDL_PIXELFORMAT_NV12:
|
|
|
|
case SDL_PIXELFORMAT_NV21:
|
2015-06-21 17:33:46 +02:00
|
|
|
if (rect
|
|
|
|
&& (rect->x != 0 || rect->y != 0 || rect->w != swdata->w
|
|
|
|
|| rect->h != swdata->h)) {
|
|
|
|
return SDL_SetError
|
Fixed bug 3857 - SDL_ConvertPixels misses YUV conversions
Sylvain
Few issues with YUV on SDL2 when using odd dimensions, and missing conversions from/back to YUV formats.
1) The big part is that SDL_ConvertPixels() does not convert to/from YUV in most cases. This now works with any format and also with odd dimensions,
by adding two internal functions SDL_ConvertPixels_YUV_to_ARGB8888 and SDL_ConvertPixels_ARGB8888_to_YUV (could it be XRGB888 ?).
The target format is hard coded to ARGB888 (which is the default in the internal of the software renderer).
In case of different YUV conversion, it will do an intermediate conversion to a ARGB8888 buffer.
SDL_ConvertPixels_YUV_to_ARGB8888 is somehow redundant with all the "Color*Dither*Mod*".
But it allows some completeness of SDL_ConvertPixels to handle all YUV format.
It also works with odd dimensions.
Moreover, I did some benchmark(SDL_ConvertPixel vs Color32DitherYV12Mod1X and Color32DitherYUY2Mod1X).
gcc-6.3 and clang-4.0. gcc performs better than clang. And, with gcc, SDL_ConvertPixels() performs better (20%) than the two C function Color32Dither*().
For instance, to convert 10 times a 3888x2592 image, it takes ~195 ms with SDL_ConvertPixels and ~235 ms with Color32Dither*().
Especially because of gcc vectorize feature that optimises all conversion loops (-ftree-loop-vectorize).
Nb: I put no image pitch for the YUV buffers. because it complexify a little bit the code and the API :
There would be some ambiguity when setting the pitch exactly to image width:
would it a be pitch of image width (for luma and chroma). or just contiguous data ? (could set pitch=0 for the later).
2) Small issues with odd dimensions:
If width "w" is odd, luma plane width is still "w" whereas chroma planes will be "(w + 1)/2". Almost the same for odd h.
Solution is to strategically substitute "w" by "(w+1)/2" at the good places ...
- In the repository, SDL_ConvertPixels() handles YUV only if yuv source format is exactly the same as YUV destination format.
It basically does a memcpy of pixels, but it's done incorrectly when width or height is odd (wrong size of chroma planes). This is fixed.
- SDL Renderers don't support odd width/height for YUV textures.
This is fixed for software, opengl, opengles2. (opengles 1 does not support it and fallback to software rendering).
This is *not* fixed for D3D and D3D11 ... (and others, psp ?)
Only *two* Dither function are fixed ... not sure if others are really used.
- This is not possible to create a NV12/NV12 texture with the software renderer, whereas other renderers allow it.
This is fixed, by using SDL_ConvertPixels underneath.
- It was not possible to SDL_UpdateTexture() of format NV12/NV21 with the software renderer. this is fixed.
Here's also two testcases:
- that do all combination of conversion.
- to test partial UpdateTexture
2017-10-07 01:50:24 +02:00
|
|
|
("YV12, IYUV, NV12, NV21 textures only support full surface locks");
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (rect) {
|
|
|
|
*pixels = swdata->planes[0] + rect->y * swdata->pitches[0] + rect->x * 2;
|
|
|
|
} else {
|
|
|
|
*pixels = swdata->planes[0];
|
|
|
|
}
|
|
|
|
*pitch = swdata->pitches[0];
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
SDL_SW_UnlockYUVTexture(SDL_SW_YUVTexture * swdata)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
SDL_SW_CopyYUVToRGB(SDL_SW_YUVTexture * swdata, const SDL_Rect * srcrect,
|
|
|
|
Uint32 target_format, int w, int h, void *pixels,
|
|
|
|
int pitch)
|
|
|
|
{
|
|
|
|
int stretch;
|
|
|
|
|
|
|
|
/* Make sure we're set up to display in the desired format */
|
2017-11-13 07:51:12 +01:00
|
|
|
if (target_format != swdata->target_format && swdata->display) {
|
|
|
|
SDL_FreeSurface(swdata->display);
|
|
|
|
swdata->display = NULL;
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
stretch = 0;
|
2017-11-13 07:51:12 +01:00
|
|
|
if (srcrect->x || srcrect->y || srcrect->w < swdata->w || srcrect->h < swdata->h) {
|
2015-06-21 17:33:46 +02:00
|
|
|
/* The source rectangle has been clipped.
|
|
|
|
Using a scratch surface is easier than adding clipped
|
|
|
|
source support to all the blitters, plus that would
|
|
|
|
slow them down in the general unclipped case.
|
|
|
|
*/
|
|
|
|
stretch = 1;
|
|
|
|
} else if ((srcrect->w != w) || (srcrect->h != h)) {
|
2017-11-13 07:51:12 +01:00
|
|
|
stretch = 1;
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
if (stretch) {
|
|
|
|
int bpp;
|
|
|
|
Uint32 Rmask, Gmask, Bmask, Amask;
|
|
|
|
|
|
|
|
if (swdata->display) {
|
|
|
|
swdata->display->w = w;
|
|
|
|
swdata->display->h = h;
|
|
|
|
swdata->display->pixels = pixels;
|
|
|
|
swdata->display->pitch = pitch;
|
|
|
|
} else {
|
|
|
|
/* This must have succeeded in SDL_SW_SetupYUVDisplay() earlier */
|
|
|
|
SDL_PixelFormatEnumToMasks(target_format, &bpp, &Rmask, &Gmask,
|
|
|
|
&Bmask, &Amask);
|
|
|
|
swdata->display =
|
|
|
|
SDL_CreateRGBSurfaceFrom(pixels, w, h, bpp, pitch, Rmask,
|
|
|
|
Gmask, Bmask, Amask);
|
|
|
|
if (!swdata->display) {
|
|
|
|
return (-1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (!swdata->stretch) {
|
|
|
|
/* This must have succeeded in SDL_SW_SetupYUVDisplay() earlier */
|
|
|
|
SDL_PixelFormatEnumToMasks(target_format, &bpp, &Rmask, &Gmask,
|
|
|
|
&Bmask, &Amask);
|
|
|
|
swdata->stretch =
|
|
|
|
SDL_CreateRGBSurface(0, swdata->w, swdata->h, bpp, Rmask,
|
|
|
|
Gmask, Bmask, Amask);
|
|
|
|
if (!swdata->stretch) {
|
|
|
|
return (-1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
pixels = swdata->stretch->pixels;
|
|
|
|
pitch = swdata->stretch->pitch;
|
|
|
|
}
|
2017-11-13 07:51:12 +01:00
|
|
|
if (SDL_ConvertPixels(swdata->w, swdata->h, swdata->format,
|
|
|
|
swdata->planes[0], swdata->pitches[0],
|
|
|
|
target_format, pixels, pitch) < 0) {
|
|
|
|
return -1;
|
2015-06-21 17:33:46 +02:00
|
|
|
}
|
|
|
|
if (stretch) {
|
|
|
|
SDL_Rect rect = *srcrect;
|
|
|
|
SDL_SoftStretch(swdata->stretch, &rect, swdata->display, NULL);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
SDL_SW_DestroyYUVTexture(SDL_SW_YUVTexture * swdata)
|
|
|
|
{
|
|
|
|
if (swdata) {
|
|
|
|
SDL_free(swdata->pixels);
|
|
|
|
SDL_FreeSurface(swdata->stretch);
|
|
|
|
SDL_FreeSurface(swdata->display);
|
|
|
|
SDL_free(swdata);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* vi: set ts=4 sw=4 expandtab: */
|