youmuliu/stgsharp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Daily update on 20250721-223448 by Nitload Youmu. Additional Message : Rename StgSHarp.C to StgSharp.Native. Add SDL to repo.

Browse Source 
Signed-off-by: Nitload <YoumuLiu_NSI@whv0t.onmicrosoft.com>
This commit is contained in:
Nitload
2025-07-21 22:34:48 +08:00
parent c624a8ca38
commit 28ea5d1d1e
1215 changed files with 628924 additions and 114 deletions
Download Patch File Download Diff File Expand all files Collapse all files
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb.h
+33
View File
@@ -0,0 +1,33 @@
#ifndef YUV_RGB_H_
#define YUV_RGB_H_
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
// Provide optimized functions to convert images from 8bits yuv420 to rgb24 format
// There are a few slightly different variations of the YCbCr color space with different parameters that
// change the conversion matrix.
// The three most common YCbCr color space, defined by BT.601, BT.709 and JPEG standard are implemented here.
// See the respective standards for details
// The matrix values used are derived from http://www.equasys.de/colorconversion.html
// YUV420 is stored as three separate channels, with U and V (Cb and Cr) subsampled by a 2 factor
// For conversion from yuv to rgb, no interpolation is done, and the same UV value are used for 4 rgb pixels. This
// is suboptimal for image quality, but by far the fastest method.
// For all methods, width and height should be even, if not, the last row/column of the result image won't be affected.
// For sse methods, if the width if not divisable by 32, the last (width%32) pixels of each line won't be affected.
/*#include <stdint.h>*/
// yuv to rgb, standard c implementation
#include "yuv_rgb_std.h"
// yuv to rgb, sse2 implementation
#include "yuv_rgb_sse.h"
// yuv to rgb, lsx implementation
#include "yuv_rgb_lsx.h"
#endif /* YUV_RGB_H_ */
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_common.h
+15
View File
@@ -0,0 +1,15 @@
#ifndef YUV_RGB_COMMON_H_
#define YUV_RGB_COMMON_H_
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
typedef enum
{
YCBCR_601_FULL,
YCBCR_601_LIMITED,
YCBCR_709_FULL,
YCBCR_709_LIMITED,
YCBCR_2020_NCL_FULL,
} YCbCrType;
#endif /* YUV_RGB_COMMON_H_ */
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_internal.h
+85
View File
@@ -0,0 +1,85 @@
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
#include "yuv_rgb.h"
#define PRECISION 6
#define PRECISION_FACTOR (1<<PRECISION)
typedef struct
{
uint8_t y_shift;
int16_t matrix[3][3];
} RGB2YUVParam;
// |Y| |y_shift| |matrix[0][0] matrix[0][1] matrix[0][2]| |R|
// |U| = | 128 | + 1/PRECISION_FACTOR * |matrix[1][0] matrix[1][1] matrix[1][2]| * |G|
// |V| | 128 | |matrix[2][0] matrix[2][1] matrix[2][2]| |B|
typedef struct
{
uint8_t y_shift;
int16_t y_factor;
int16_t v_r_factor;
int16_t u_g_factor;
int16_t v_g_factor;
int16_t u_b_factor;
} YUV2RGBParam;
// |R| |y_factor 0 v_r_factor| |Y-y_shift|
// |G| = 1/PRECISION_FACTOR * |y_factor u_g_factor v_g_factor| * | U-128 |
// |B| |y_factor u_b_factor 0 | | V-128 |
#ifdef _MSC_VER
#pragma warning(push)
#pragma warning(disable : 26451)
#endif
#define V(value) (int16_t)((value*PRECISION_FACTOR)+0.5)
// for ITU-T T.871, values can be found in section 7
// for ITU-R BT.601-7 values are derived from equations in sections 2.5.1-2.5.3, assuming RGB is encoded using full range ([0-1]<->[0-255])
// for ITU-R BT.709-6 values are derived from equations in sections 3.2-3.4, assuming RGB is encoded using full range ([0-1]<->[0-255])
// for ITU-R BT.2020 values are assuming RGB is encoded using full 10-bit range ([0-1]<->[0-1023])
// all values are rounded to the fourth decimal
static const YUV2RGBParam YUV2RGB[] = {
// ITU-T T.871 (JPEG)
{/*.y_shift=*/ 0, /*.y_factor=*/ V(1.0), /*.v_r_factor=*/ V(1.402), /*.u_g_factor=*/ -V(0.3441), /*.v_g_factor=*/ -V(0.7141), /*.u_b_factor=*/ V(1.772)},
// ITU-R BT.601-7
{/*.y_shift=*/ 16, /*.y_factor=*/ V(1.1644), /*.v_r_factor=*/ V(1.596), /*.u_g_factor=*/ -V(0.3918), /*.v_g_factor=*/ -V(0.813), /*.u_b_factor=*/ V(2.0172)},
// ITU-R BT.709-6 full range
{/*.y_shift=*/ 0, /*.y_factor=*/ V(1.0), /*.v_r_factor=*/ V(1.581), /*.u_g_factor=*/ -V(0.1881), /*.v_g_factor=*/ -V(0.47), /*.u_b_factor=*/ V(1.8629)},
// ITU-R BT.709-6
{/*.y_shift=*/ 16, /*.y_factor=*/ V(1.1644), /*.v_r_factor=*/ V(1.7927), /*.u_g_factor=*/ -V(0.2132), /*.v_g_factor=*/ -V(0.5329), /*.u_b_factor=*/ V(2.1124)},
// ITU-R BT.2020 10-bit full range
{/*.y_shift=*/ 0, /*.y_factor=*/ V(1.0), /*.v_r_factor=*/ V(1.4760), /*.u_g_factor=*/ -V(0.1647), /*.v_g_factor=*/ -V(0.5719), /*.u_b_factor=*/ V(1.8832) }
};
static const RGB2YUVParam RGB2YUV[] = {
// ITU-T T.871 (JPEG)
{/*.y_shift=*/ 0, /*.matrix=*/ {{V(0.299), V(0.587), V(0.114)}, {-V(0.1687), -V(0.3313), V(0.5)}, {V(0.5), -V(0.4187), -V(0.0813)}}},
// ITU-R BT.601-7
{/*.y_shift=*/ 16, /*.matrix=*/ {{V(0.2568), V(0.5041), V(0.0979)}, {-V(0.1482), -V(0.291), V(0.4392)}, {V(0.4392), -V(0.3678), -V(0.0714)}}},
// ITU-R BT.709-6 full range
{/*.y_shift=*/ 0, /*.matrix=*/ {{V(0.2126), V(0.7152), V(0.0722)}, {-V(0.1141), -V(0.3839), V(0.498)}, {V(0.498), -V(0.4524), -V(0.0457)}}},
// ITU-R BT.709-6
{/*.y_shift=*/ 16, /*.matrix=*/ {{V(0.1826), V(0.6142), V(0.062)}, {-V(0.1006), -V(0.3386), V(0.4392)}, {V(0.4392), -V(0.3989), -V(0.0403)}}},
// ITU-R BT.2020 10-bit full range
{/*.y_shift=*/ 0, /*.matrix=*/ {{V(0.2627), V(0.6780), V(0.0593)}, {-V(0.1395), -V(0.3600), V(0.4995)}, {V(0.4995), -V(0.4593), -V(0.0402)}}},
};
#ifdef _MSC_VER
#pragma warning(pop)
#endif
/* The various layouts of YUV data we support */
#define YUV_FORMAT_420 1
#define YUV_FORMAT_422 2
#define YUV_FORMAT_NV12 3
/* The various formats of RGB pixel that we support */
#define RGB_FORMAT_RGB565 1
#define RGB_FORMAT_RGB24 2
#define RGB_FORMAT_RGBA 3
#define RGB_FORMAT_BGRA 4
#define RGB_FORMAT_ARGB 5
#define RGB_FORMAT_ABGR 6
#define RGB_FORMAT_XBGR2101010 7
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_lsx.c
+43
View File
@@ -0,0 +1,43 @@
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
#include "SDL_internal.h"
#ifdef SDL_HAVE_YUV
#include "yuv_rgb_lsx.h"
#include "yuv_rgb_internal.h"
#ifdef SDL_LSX_INTRINSICS
#define LSX_FUNCTION_NAME yuv420_rgb24_lsx
#define STD_FUNCTION_NAME yuv420_rgb24_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGB24
#include "yuv_rgb_lsx_func.h"
#define LSX_FUNCTION_NAME yuv420_rgba_lsx
#define STD_FUNCTION_NAME yuv420_rgba_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGBA
#include "yuv_rgb_lsx_func.h"
#define LSX_FUNCTION_NAME yuv420_bgra_lsx
#define STD_FUNCTION_NAME yuv420_bgra_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_BGRA
#include "yuv_rgb_lsx_func.h"
#define LSX_FUNCTION_NAME yuv420_argb_lsx
#define STD_FUNCTION_NAME yuv420_argb_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ARGB
#include "yuv_rgb_lsx_func.h"
#define LSX_FUNCTION_NAME yuv420_abgr_lsx
#define STD_FUNCTION_NAME yuv420_abgr_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ABGR
#include "yuv_rgb_lsx_func.h"
#endif // SDL_LSX_INTRINSICS
#endif // SDL_HAVE_YUV
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_lsx.h
+36
View File
@@ -0,0 +1,36 @@
#ifdef SDL_LSX_INTRINSICS
#include "yuv_rgb_common.h"
//yuv420 to bgra, lsx implementation
void yuv420_rgb24_lsx(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_rgba_lsx(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *v, const uint8_t *u, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_bgra_lsx(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *v, const uint8_t *u, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_argb_lsx(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *v, const uint8_t *u, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_abgr_lsx(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *v, const uint8_t *u, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
#endif //SDL_LSX_INTRINSICS
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_lsx_func.h
+372
View File
@@ -0,0 +1,372 @@
// Copyright 2016 Adrien Descamps
// // Distributed under BSD 3-Clause License
#include <lsxintrin.h>
#if YUV_FORMAT == YUV_FORMAT_420
#define READ_Y(y_ptr) \
y = __lsx_vld(y_ptr, 0); \
#define READ_UV \
u_temp = __lsx_vld(u_ptr, 0); \
v_temp = __lsx_vld(v_ptr, 0); \
#else
#error READ_UV unimplemented
#endif
#define PACK_RGBA_32(R1, R2, G1, G2, B1, B2, A1, A2, RGB1, RGB2, \
RGB3, RGB4, RGB5, RGB6, RGB7, RGB8) \
{ \
__m128i ab_l, ab_h, gr_l, gr_h; \
ab_l = __lsx_vilvl_b(B1, A1); \
ab_h = __lsx_vilvh_b(B1, A1); \
gr_l = __lsx_vilvl_b(R1, G1); \
gr_h = __lsx_vilvh_b(R1, G1); \
RGB1 = __lsx_vilvl_h(gr_l, ab_l); \
RGB2 = __lsx_vilvh_h(gr_l, ab_l); \
RGB3 = __lsx_vilvl_h(gr_h, ab_h); \
RGB4 = __lsx_vilvh_h(gr_h, ab_h); \
ab_l = __lsx_vilvl_b(B2, A2); \
ab_h = __lsx_vilvh_b(B2, A2); \
gr_l = __lsx_vilvl_b(R2, G2); \
gr_h = __lsx_vilvh_b(R2, G2); \
RGB5 = __lsx_vilvl_h(gr_l, ab_l); \
RGB6 = __lsx_vilvh_h(gr_l, ab_l); \
RGB7 = __lsx_vilvl_h(gr_h, ab_h); \
RGB8 = __lsx_vilvh_h(gr_h, ab_h); \
}
#define PACK_RGB24_32_STEP(R, G, B, RGB1, RGB2, RGB3) \
RGB1 = __lsx_vilvl_b(G, R); \
RGB1 = __lsx_vshuf_b(B, RGB1, mask1); \
RGB2 = __lsx_vshuf_b(B, G, mask2); \
RGB2 = __lsx_vshuf_b(R, RGB2, mask3); \
RGB3 = __lsx_vshuf_b(R, B, mask4); \
RGB3 = __lsx_vshuf_b(G, RGB3, mask5); \
#define PACK_RGB24_32(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
PACK_RGB24_32_STEP(R1, G1, B1, RGB1, RGB2, RGB3); \
PACK_RGB24_32_STEP(R2, G2, B2, RGB4, RGB5, RGB6); \
#if RGB_FORMAT == RGB_FORMAT_RGB24
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6; \
__m128i rgb_7, rgb_8, rgb_9, rgb_10, rgb_11, rgb_12; \
PACK_RGB24_32(r_8_11, r_8_12, g_8_11, g_8_12, b_8_11, b_8_12, \
rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6) \
PACK_RGB24_32(r_8_21, r_8_22, g_8_21, g_8_22, b_8_21, b_8_22, \
rgb_7, rgb_8, rgb_9, rgb_10, rgb_11, rgb_12) \
#elif RGB_FORMAT == RGB_FORMAT_RGBA
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = __lsx_vldi(0xFF); \
PACK_RGBA_32(r_8_11, r_8_12, g_8_11, g_8_12, b_8_11, b_8_12, a, a, \
rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
PACK_RGBA_32(r_8_21, r_8_22, g_8_21, g_8_22, b_8_21, b_8_22, a, a, \
rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#elif RGB_FORMAT == RGB_FORMAT_BGRA
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = __lsx_vldi(0xFF); \
PACK_RGBA_32(b_8_11, b_8_12, g_8_11, g_8_12, r_8_11, r_8_12, a, a, \
rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
PACK_RGBA_32(b_8_21, b_8_22, g_8_21, g_8_22, r_8_21, r_8_22, a, a, \
rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#elif RGB_FORMAT == RGB_FORMAT_ARGB
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = __lsx_vldi(0xFF); \
PACK_RGBA_32(a, a, r_8_11, r_8_12, g_8_11, g_8_12, b_8_11, b_8_12, \
rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
PACK_RGBA_32(a, a, r_8_21, r_8_22, g_8_21, g_8_22, b_8_21, b_8_22, \
rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#elif RGB_FORMAT == RGB_FORMAT_ABGR
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = __lsx_vldi(0xFF); \
PACK_RGBA_32(a, a, b_8_11, b_8_12, g_8_11, g_8_12, r_8_11, r_8_12, \
rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
PACK_RGBA_32(a, a, b_8_21, b_8_22, g_8_21, g_8_22, r_8_21, r_8_22, \
rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#else
#error PACK_PIXEL unimplemented
#endif
#define LSX_ST_UB2(in0, in1, pdst, stride) \
{ \
__lsx_vst(in0, pdst, 0); \
__lsx_vst(in1, pdst + stride, 0); \
}
#if RGB_FORMAT == RGB_FORMAT_RGB24 \
#define SAVE_LINE1 \
LSX_ST_UB2(rgb_1, rgb_2, rgb_ptr1, 16); \
LSX_ST_UB2(rgb_3, rgb_4, rgb_ptr1 + 32, 16); \
LSX_ST_UB2(rgb_5, rgb_6, rgb_ptr1 + 64, 16); \
#define SAVE_LINE2 \
LSX_ST_UB2(rgb_7, rgb_8, rgb_ptr2, 16); \
LSX_ST_UB2(rgb_9, rgb_10, rgb_ptr2 + 32, 16); \
LSX_ST_UB2(rgb_11, rgb_12, rgb_ptr2 + 64, 16); \
#elif RGB_FORMAT == RGB_FORMAT_RGBA || RGB_FORMAT == RGB_FORMAT_BGRA || \
RGB_FORMAT == RGB_FORMAT_ARGB || RGB_FORMAT == RGB_FORMAT_ABGR \
#define SAVE_LINE1 \
LSX_ST_UB2(rgb_1, rgb_2, rgb_ptr1, 16); \
LSX_ST_UB2(rgb_3, rgb_4, rgb_ptr1 + 32, 16); \
LSX_ST_UB2(rgb_5, rgb_6, rgb_ptr1 + 64, 16); \
LSX_ST_UB2(rgb_7, rgb_8, rgb_ptr1 + 96, 16); \
#define SAVE_LINE2 \
LSX_ST_UB2(rgb_9, rgb_10, rgb_ptr2, 16); \
LSX_ST_UB2(rgb_11, rgb_12, rgb_ptr2 + 32, 16); \
LSX_ST_UB2(rgb_13, rgb_14, rgb_ptr2 + 64, 16); \
LSX_ST_UB2(rgb_15, rgb_16, rgb_ptr2 + 96, 16); \
#else
#error SAVE_LINE unimplemented
#endif
// = u*vr g=u*ug+v*vg b=u*ub
#define UV2RGB_16(U, V, R1, G1, B1, R2, G2, B2) \
r_temp = __lsx_vmul_h(V, v2r); \
g_temp = __lsx_vmul_h(U, u2g); \
g_temp = __lsx_vmadd_h(g_temp, V, v2g); \
b_temp = __lsx_vmul_h(U, u2b); \
R1 = __lsx_vilvl_h(r_temp, r_temp); \
G1 = __lsx_vilvl_h(g_temp, g_temp); \
B1 = __lsx_vilvl_h(b_temp, b_temp); \
R2 = __lsx_vilvh_h(r_temp, r_temp); \
G2 = __lsx_vilvh_h(g_temp, g_temp); \
B2 = __lsx_vilvh_h(b_temp, b_temp); \
// Y=(Y-shift)*shift R=(Y+R)>>6,G=(Y+G)>>6,B=(B+Y)>>6
#define ADD_Y2RGB_16(Y1, Y2, R1, G1, B1, R2, G2, B2) \
Y1 = __lsx_vsub_h(Y1, shift); \
Y2 = __lsx_vsub_h(Y2, shift); \
Y1 = __lsx_vmul_h(Y1, yf); \
Y2 = __lsx_vmul_h(Y2, yf); \
R1 = __lsx_vadd_h(R1, Y1); \
G1 = __lsx_vadd_h(G1, Y1); \
B1 = __lsx_vadd_h(B1, Y1); \
R2 = __lsx_vadd_h(R2, Y2); \
G2 = __lsx_vadd_h(G2, Y2); \
B2 = __lsx_vadd_h(B2, Y2); \
R1 = __lsx_vsrai_h(R1, PRECISION); \
G1 = __lsx_vsrai_h(G1, PRECISION); \
B1 = __lsx_vsrai_h(B1, PRECISION); \
R2 = __lsx_vsrai_h(R2, PRECISION); \
G2 = __lsx_vsrai_h(G2, PRECISION); \
B2 = __lsx_vsrai_h(B2, PRECISION); \
#define CLIP(in0, in1, in2, in3, in4, in5) \
{ \
in0 = __lsx_vmaxi_h(in0, 0); \
in1 = __lsx_vmaxi_h(in1, 0); \
in2 = __lsx_vmaxi_h(in2, 0); \
in3 = __lsx_vmaxi_h(in3, 0); \
in4 = __lsx_vmaxi_h(in4, 0); \
in5 = __lsx_vmaxi_h(in5, 0); \
in0 = __lsx_vsat_hu(in0, 7); \
in1 = __lsx_vsat_hu(in1, 7); \
in2 = __lsx_vsat_hu(in2, 7); \
in3 = __lsx_vsat_hu(in3, 7); \
in4 = __lsx_vsat_hu(in4, 7); \
in5 = __lsx_vsat_hu(in5, 7); \
}
#define YUV2RGB_32 \
__m128i y, u_temp, v_temp; \
__m128i r_8_11, g_8_11, b_8_11, r_8_21, g_8_21, b_8_21; \
__m128i r_8_12, g_8_12, b_8_12, r_8_22, g_8_22, b_8_22; \
__m128i u, v, r_temp, g_temp, b_temp; \
__m128i r_1, g_1, b_1, r_2, g_2, b_2; \
__m128i y_1, y_2; \
__m128i r_uv_1, g_uv_1, b_uv_1, r_uv_2, g_uv_2, b_uv_2; \
\
READ_UV \
\
/* process first 16 pixels of first line */ \
u = __lsx_vilvl_b(zero, u_temp); \
v = __lsx_vilvl_b(zero, v_temp); \
u = __lsx_vsub_h(u, bias); \
v = __lsx_vsub_h(v, bias); \
UV2RGB_16(u, v, r_1, g_1, b_1, r_2, g_2, b_2); \
r_uv_1 = r_1; g_uv_1 = g_1; b_uv_1 = b_1; \
r_uv_2 = r_2; g_uv_2 = g_2; b_uv_2 = b_2; \
READ_Y(y_ptr1) \
y_1 = __lsx_vilvl_b(zero, y); \
y_2 = __lsx_vilvh_b(zero, y); \
ADD_Y2RGB_16(y_1, y_2, r_1, g_1, b_1, r_2, g_2, b_2) \
CLIP(r_1, g_1, b_1, r_2, g_2, b_2); \
r_8_11 = __lsx_vpickev_b(r_2, r_1); \
g_8_11 = __lsx_vpickev_b(g_2, g_1); \
b_8_11 = __lsx_vpickev_b(b_2, b_1); \
\
/* process first 16 pixels of second line */ \
r_1 = r_uv_1; g_1 = g_uv_1; b_1 = b_uv_1; \
r_2 = r_uv_2; g_2 = g_uv_2; b_2 = b_uv_2; \
\
READ_Y(y_ptr2) \
y_1 = __lsx_vilvl_b(zero, y); \
y_2 = __lsx_vilvh_b(zero, y); \
ADD_Y2RGB_16(y_1, y_2, r_1, g_1, b_1, r_2, g_2, b_2) \
CLIP(r_1, g_1, b_1, r_2, g_2, b_2); \
r_8_21 = __lsx_vpickev_b(r_2, r_1); \
g_8_21 = __lsx_vpickev_b(g_2, g_1); \
b_8_21 = __lsx_vpickev_b(b_2, b_1); \
\
/* process last 16 pixels of first line */ \
u = __lsx_vilvh_b(zero, u_temp); \
v = __lsx_vilvh_b(zero, v_temp); \
u = __lsx_vsub_h(u, bias); \
v = __lsx_vsub_h(v, bias); \
UV2RGB_16(u, v, r_1, g_1, b_1, r_2, g_2, b_2); \
r_uv_1 = r_1; g_uv_1 = g_1; b_uv_1 = b_1; \
r_uv_2 = r_2; g_uv_2 = g_2; b_uv_2 = b_2; \
READ_Y(y_ptr1 + 16 * y_pixel_stride) \
y_1 = __lsx_vilvl_b(zero, y); \
y_2 = __lsx_vilvh_b(zero, y); \
ADD_Y2RGB_16(y_1, y_2, r_1, g_1, b_1, r_2, g_2, b_2) \
CLIP(r_1, g_1, b_1, r_2, g_2, b_2); \
r_8_12 = __lsx_vpickev_b(r_2, r_1); \
g_8_12 = __lsx_vpickev_b(g_2, g_1); \
b_8_12 = __lsx_vpickev_b(b_2, b_1); \
\
/* process last 16 pixels of second line */ \
r_1 = r_uv_1; g_1 = g_uv_1; b_1 = b_uv_1; \
r_2 = r_uv_2; g_2 = g_uv_2; b_2 = b_uv_2; \
\
READ_Y(y_ptr2 + 16 * y_pixel_stride) \
y_1 = __lsx_vilvl_b(zero, y); \
y_2 = __lsx_vilvh_b(zero, y); \
ADD_Y2RGB_16(y_1, y_2, r_1, g_1, b_1, r_2, g_2, b_2) \
CLIP(r_1, g_1, b_1, r_2, g_2, b_2); \
r_8_22 = __lsx_vpickev_b(r_2, r_1); \
g_8_22 = __lsx_vpickev_b(g_2, g_1); \
b_8_22 = __lsx_vpickev_b(b_2, b_1); \
\
void LSX_FUNCTION_NAME(uint32_t width, uint32_t height, const uint8_t *Y,
const uint8_t *U, const uint8_t *V, uint32_t Y_stride,
uint32_t UV_stride, uint8_t *RGB, uint32_t RGB_stride,
YCbCrType yuv_type)
{
const YUV2RGBParam *const param = &(YUV2RGB[yuv_type]);
#if YUV_FORMAT == YUV_FORMAT_420
const int y_pixel_stride = 1;
const int uv_pixel_stride = 1;
const int uv_x_sample_interval = 2;
const int uv_y_sample_interval = 2;
#endif
#if RGB_FORMAT == RGB_FORMAT_RGB565
const int rgb_pixel_stride = 2;
#elif RGB_FORMAT == RGB_FORMAT_RGB24
const int rgb_pixel_stride = 3;
__m128i mask1 = {0x0504110302100100, 0x0A14090813070612};
__m128i mask2 = {0x1808170716061505, 0x00000000000A1909};
__m128i mask3 = {0x0504170302160100, 0x0A1A090819070618};
__m128i mask4 = {0x1E0D1D0C1C0B1B0A, 0x00000000000F1F0E};
__m128i mask5 = {0x05041C03021B0100, 0x0A1F09081E07061D};
#elif RGB_FORMAT == RGB_FORMAT_RGBA || RGB_FORMAT_BGRA || \
RGB_FORMAT == RGB_FORMAT_ARGB || RGB_FORMAT_ABGR
const int rgb_pixel_stride = 4;
#else
#error Unknown RGB pixel size
#endif
uint32_t xpos, ypos;
__m128i v2r = __lsx_vreplgr2vr_h(param->v_r_factor);
__m128i v2g = __lsx_vreplgr2vr_h(param->v_g_factor);
__m128i u2g = __lsx_vreplgr2vr_h(param->u_g_factor);
__m128i u2b = __lsx_vreplgr2vr_h(param->u_b_factor);
__m128i bias = __lsx_vreplgr2vr_h(128);
__m128i shift = __lsx_vreplgr2vr_h(param->y_shift);
__m128i yf = __lsx_vreplgr2vr_h(param->y_factor);
__m128i zero = __lsx_vldi(0);
if (width >= 32) {
for (ypos = 0; ypos < (height - (uv_y_sample_interval - 1)); ypos += uv_y_sample_interval) {
const uint8_t *y_ptr1 = Y + ypos * Y_stride,
*y_ptr2 = Y + (ypos + 1) * Y_stride,
*u_ptr = U + (ypos/uv_y_sample_interval) * UV_stride,
*v_ptr = V + (ypos/uv_y_sample_interval) * UV_stride;
uint8_t *rgb_ptr1 = RGB + ypos * RGB_stride,
*rgb_ptr2 = RGB + (ypos + 1) * RGB_stride;
for (xpos = 0; xpos < (width - 31); xpos += 32){
YUV2RGB_32
{
PACK_PIXEL
SAVE_LINE1
if (uv_y_sample_interval > 1)
{
SAVE_LINE2
}
}
y_ptr1 += 32 * y_pixel_stride;
y_ptr2 += 32 * y_pixel_stride;
u_ptr += 32 * uv_pixel_stride/uv_x_sample_interval;
v_ptr += 32 * uv_pixel_stride/uv_x_sample_interval;
rgb_ptr1 += 32 * rgb_pixel_stride;
rgb_ptr2 += 32 * rgb_pixel_stride;
}
}
if (uv_y_sample_interval == 2 && ypos == (height - 1)) {
const uint8_t *y_ptr = Y + ypos * Y_stride,
*u_ptr = U + (ypos/uv_y_sample_interval) * UV_stride,
*v_ptr = V + (ypos/uv_y_sample_interval) * UV_stride;
uint8_t *rgb_ptr = RGB + ypos * RGB_stride;
STD_FUNCTION_NAME(width, 1, y_ptr, u_ptr, v_ptr, Y_stride, UV_stride, rgb_ptr, RGB_stride, yuv_type);
}
}
{
int converted = (width & ~31);
if (converted != width)
{
const uint8_t *y_ptr = Y + converted * y_pixel_stride,
*u_ptr = U + converted * uv_pixel_stride / uv_x_sample_interval,
*v_ptr = V + converted * uv_pixel_stride / uv_x_sample_interval;
uint8_t *rgb_ptr = RGB + converted * rgb_pixel_stride;
STD_FUNCTION_NAME(width-converted, height, y_ptr, u_ptr, v_ptr, Y_stride, UV_stride, rgb_ptr, RGB_stride, yuv_type);
}
}
}
#undef LSX_FUNCTION_NAME
#undef STD_FUNCTION_NAME
#undef YUV_FORMAT
#undef RGB_FORMAT
#undef LSX_ALIGNED
#undef LSX_ST_UB2
#undef UV2RGB_16
#undef ADD_Y2RGB_16
#undef PACK_RGB24_32_STEP
#undef PACK_RGB24_32
#undef PACK_PIXEL
#undef PACK_RGBA_32
#undef SAVE_LINE1
#undef SAVE_LINE2
#undef READ_Y
#undef READ_UV
#undef YUV2RGB_32
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_sse.c
+460
View File
@@ -0,0 +1,460 @@
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
#include "SDL_internal.h"
#ifdef SDL_HAVE_YUV
#include "yuv_rgb_internal.h"
#ifdef SDL_SSE2_INTRINSICS
/* SDL doesn't use these atm and compiling them adds seconds onto the build. --ryan.
#define SSE_FUNCTION_NAME yuv420_rgb565_sse
#define STD_FUNCTION_NAME yuv420_rgb565_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGB565
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_rgb24_sse
#define STD_FUNCTION_NAME yuv420_rgb24_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGB24
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_rgba_sse
#define STD_FUNCTION_NAME yuv420_rgba_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGBA
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_bgra_sse
#define STD_FUNCTION_NAME yuv420_bgra_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_BGRA
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_argb_sse
#define STD_FUNCTION_NAME yuv420_argb_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ARGB
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_abgr_sse
#define STD_FUNCTION_NAME yuv420_abgr_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ABGR
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_rgb565_sse
#define STD_FUNCTION_NAME yuv422_rgb565_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGB565
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_rgb24_sse
#define STD_FUNCTION_NAME yuv422_rgb24_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGB24
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_rgba_sse
#define STD_FUNCTION_NAME yuv422_rgba_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGBA
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_bgra_sse
#define STD_FUNCTION_NAME yuv422_bgra_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_BGRA
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_argb_sse
#define STD_FUNCTION_NAME yuv422_argb_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_ARGB
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_abgr_sse
#define STD_FUNCTION_NAME yuv422_abgr_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_ABGR
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_rgb565_sse
#define STD_FUNCTION_NAME yuvnv12_rgb565_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGB565
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_rgb24_sse
#define STD_FUNCTION_NAME yuvnv12_rgb24_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGB24
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_rgba_sse
#define STD_FUNCTION_NAME yuvnv12_rgba_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGBA
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_bgra_sse
#define STD_FUNCTION_NAME yuvnv12_bgra_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_BGRA
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_argb_sse
#define STD_FUNCTION_NAME yuvnv12_argb_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_ARGB
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_abgr_sse
#define STD_FUNCTION_NAME yuvnv12_abgr_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_ABGR
#define SSE_ALIGNED
#include "yuv_rgb_sse_func.h"
*/
#define SSE_FUNCTION_NAME yuv420_rgb565_sseu
#define STD_FUNCTION_NAME yuv420_rgb565_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGB565
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_rgb24_sseu
#define STD_FUNCTION_NAME yuv420_rgb24_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGB24
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_rgba_sseu
#define STD_FUNCTION_NAME yuv420_rgba_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGBA
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_bgra_sseu
#define STD_FUNCTION_NAME yuv420_bgra_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_BGRA
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_argb_sseu
#define STD_FUNCTION_NAME yuv420_argb_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ARGB
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv420_abgr_sseu
#define STD_FUNCTION_NAME yuv420_abgr_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ABGR
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_rgb565_sseu
#define STD_FUNCTION_NAME yuv422_rgb565_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGB565
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_rgb24_sseu
#define STD_FUNCTION_NAME yuv422_rgb24_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGB24
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_rgba_sseu
#define STD_FUNCTION_NAME yuv422_rgba_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGBA
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_bgra_sseu
#define STD_FUNCTION_NAME yuv422_bgra_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_BGRA
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_argb_sseu
#define STD_FUNCTION_NAME yuv422_argb_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_ARGB
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuv422_abgr_sseu
#define STD_FUNCTION_NAME yuv422_abgr_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_ABGR
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_rgb565_sseu
#define STD_FUNCTION_NAME yuvnv12_rgb565_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGB565
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_rgb24_sseu
#define STD_FUNCTION_NAME yuvnv12_rgb24_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGB24
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_rgba_sseu
#define STD_FUNCTION_NAME yuvnv12_rgba_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGBA
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_bgra_sseu
#define STD_FUNCTION_NAME yuvnv12_bgra_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_BGRA
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_argb_sseu
#define STD_FUNCTION_NAME yuvnv12_argb_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_ARGB
#include "yuv_rgb_sse_func.h"
#define SSE_FUNCTION_NAME yuvnv12_abgr_sseu
#define STD_FUNCTION_NAME yuvnv12_abgr_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_ABGR
#include "yuv_rgb_sse_func.h"
/* SDL doesn't use these atm and compiling them adds seconds onto the build. --ryan.
#define UNPACK_RGB24_32_STEP1(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
R1 = _mm_unpacklo_epi8(RGB1, RGB4); \
R2 = _mm_unpackhi_epi8(RGB1, RGB4); \
G1 = _mm_unpacklo_epi8(RGB2, RGB5); \
G2 = _mm_unpackhi_epi8(RGB2, RGB5); \
B1 = _mm_unpacklo_epi8(RGB3, RGB6); \
B2 = _mm_unpackhi_epi8(RGB3, RGB6);
#define UNPACK_RGB24_32_STEP2(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
RGB1 = _mm_unpacklo_epi8(R1, G2); \
RGB2 = _mm_unpackhi_epi8(R1, G2); \
RGB3 = _mm_unpacklo_epi8(R2, B1); \
RGB4 = _mm_unpackhi_epi8(R2, B1); \
RGB5 = _mm_unpacklo_epi8(G1, B2); \
RGB6 = _mm_unpackhi_epi8(G1, B2); \
#define UNPACK_RGB24_32(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
UNPACK_RGB24_32_STEP1(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
UNPACK_RGB24_32_STEP2(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
UNPACK_RGB24_32_STEP1(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
UNPACK_RGB24_32_STEP2(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
UNPACK_RGB24_32_STEP1(RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, R1, R2, G1, G2, B1, B2) \
#define RGB2YUV_16(R, G, B, Y, U, V) \
Y = _mm_add_epi16(_mm_mullo_epi16(R, _mm_set1_epi16(param->matrix[0][0])), \
_mm_mullo_epi16(G, _mm_set1_epi16(param->matrix[0][1]))); \
Y = _mm_add_epi16(Y, _mm_mullo_epi16(B, _mm_set1_epi16(param->matrix[0][2]))); \
Y = _mm_add_epi16(Y, _mm_set1_epi16((param->y_shift)<<PRECISION)); \
Y = _mm_srai_epi16(Y, PRECISION); \
U = _mm_add_epi16(_mm_mullo_epi16(R, _mm_set1_epi16(param->matrix[1][0])), \
_mm_mullo_epi16(G, _mm_set1_epi16(param->matrix[1][1]))); \
U = _mm_add_epi16(U, _mm_mullo_epi16(B, _mm_set1_epi16(param->matrix[1][2]))); \
U = _mm_add_epi16(U, _mm_set1_epi16(128<<PRECISION)); \
U = _mm_srai_epi16(U, PRECISION); \
V = _mm_add_epi16(_mm_mullo_epi16(R, _mm_set1_epi16(param->matrix[2][0])), \
_mm_mullo_epi16(G, _mm_set1_epi16(param->matrix[2][1]))); \
V = _mm_add_epi16(V, _mm_mullo_epi16(B, _mm_set1_epi16(param->matrix[2][2]))); \
V = _mm_add_epi16(V, _mm_set1_epi16(128<<PRECISION)); \
V = _mm_srai_epi16(V, PRECISION);
*/
#if 0 // SDL doesn't use these atm and compiling them adds seconds onto the build. --ryan.
#define RGB2YUV_32 \
__m128i r1, r2, b1, b2, g1, g2; \
__m128i r_16, g_16, b_16; \
__m128i y1_16, y2_16, u1_16, u2_16, v1_16, v2_16, y, u1, u2, v1, v2, u1_tmp, u2_tmp, v1_tmp, v2_tmp; \
__m128i rgb1 = LOAD_SI128((const __m128i*)(rgb_ptr1)), \
rgb2 = LOAD_SI128((const __m128i*)(rgb_ptr1+16)), \
rgb3 = LOAD_SI128((const __m128i*)(rgb_ptr1+32)), \
rgb4 = LOAD_SI128((const __m128i*)(rgb_ptr2)), \
rgb5 = LOAD_SI128((const __m128i*)(rgb_ptr2+16)), \
rgb6 = LOAD_SI128((const __m128i*)(rgb_ptr2+32)); \
/* unpack rgb24 data to r, g and b data in separate channels*/ \
UNPACK_RGB24_32(rgb1, rgb2, rgb3, rgb4, rgb5, rgb6, r1, r2, g1, g2, b1, b2) \
/* process pixels of first line */ \
r_16 = _mm_unpacklo_epi8(r1, _mm_setzero_si128()); \
g_16 = _mm_unpacklo_epi8(g1, _mm_setzero_si128()); \
b_16 = _mm_unpacklo_epi8(b1, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y1_16, u1_16, v1_16) \
r_16 = _mm_unpackhi_epi8(r1, _mm_setzero_si128()); \
g_16 = _mm_unpackhi_epi8(g1, _mm_setzero_si128()); \
b_16 = _mm_unpackhi_epi8(b1, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y2_16, u2_16, v2_16) \
y = _mm_packus_epi16(y1_16, y2_16); \
u1 = _mm_packus_epi16(u1_16, u2_16); \
v1 = _mm_packus_epi16(v1_16, v2_16); \
/* save Y values */ \
SAVE_SI128((__m128i*)(y_ptr1), y); \
/* process pixels of second line */ \
r_16 = _mm_unpacklo_epi8(r2, _mm_setzero_si128()); \
g_16 = _mm_unpacklo_epi8(g2, _mm_setzero_si128()); \
b_16 = _mm_unpacklo_epi8(b2, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y1_16, u1_16, v1_16) \
r_16 = _mm_unpackhi_epi8(r2, _mm_setzero_si128()); \
g_16 = _mm_unpackhi_epi8(g2, _mm_setzero_si128()); \
b_16 = _mm_unpackhi_epi8(b2, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y2_16, u2_16, v2_16) \
y = _mm_packus_epi16(y1_16, y2_16); \
u2 = _mm_packus_epi16(u1_16, u2_16); \
v2 = _mm_packus_epi16(v1_16, v2_16); \
/* save Y values */ \
SAVE_SI128((__m128i*)(y_ptr2), y); \
/* vertical subsampling of u/v values */ \
u1_tmp = _mm_avg_epu8(u1, u2); \
v1_tmp = _mm_avg_epu8(v1, v2); \
/* do the same again with next data */ \
rgb1 = LOAD_SI128((const __m128i*)(rgb_ptr1+48)); \
rgb2 = LOAD_SI128((const __m128i*)(rgb_ptr1+64)); \
rgb3 = LOAD_SI128((const __m128i*)(rgb_ptr1+80)); \
rgb4 = LOAD_SI128((const __m128i*)(rgb_ptr2+48)); \
rgb5 = LOAD_SI128((const __m128i*)(rgb_ptr2+64)); \
rgb6 = LOAD_SI128((const __m128i*)(rgb_ptr2+80)); \
/* unpack rgb24 data to r, g and b data in separate channels*/ \
UNPACK_RGB24_32(rgb1, rgb2, rgb3, rgb4, rgb5, rgb6, r1, r2, g1, g2, b1, b2) \
/* process pixels of first line */ \
r_16 = _mm_unpacklo_epi8(r1, _mm_setzero_si128()); \
g_16 = _mm_unpacklo_epi8(g1, _mm_setzero_si128()); \
b_16 = _mm_unpacklo_epi8(b1, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y1_16, u1_16, v1_16) \
r_16 = _mm_unpackhi_epi8(r1, _mm_setzero_si128()); \
g_16 = _mm_unpackhi_epi8(g1, _mm_setzero_si128()); \
b_16 = _mm_unpackhi_epi8(b1, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y2_16, u2_16, v2_16) \
y = _mm_packus_epi16(y1_16, y2_16); \
u1 = _mm_packus_epi16(u1_16, u2_16); \
v1 = _mm_packus_epi16(v1_16, v2_16); \
/* save Y values */ \
SAVE_SI128((__m128i*)(y_ptr1+16), y); \
/* process pixels of second line */ \
r_16 = _mm_unpacklo_epi8(r2, _mm_setzero_si128()); \
g_16 = _mm_unpacklo_epi8(g2, _mm_setzero_si128()); \
b_16 = _mm_unpacklo_epi8(b2, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y1_16, u1_16, v1_16) \
r_16 = _mm_unpackhi_epi8(r2, _mm_setzero_si128()); \
g_16 = _mm_unpackhi_epi8(g2, _mm_setzero_si128()); \
b_16 = _mm_unpackhi_epi8(b2, _mm_setzero_si128()); \
RGB2YUV_16(r_16, g_16, b_16, y2_16, u2_16, v2_16) \
y = _mm_packus_epi16(y1_16, y2_16); \
u2 = _mm_packus_epi16(u1_16, u2_16); \
v2 = _mm_packus_epi16(v1_16, v2_16); \
/* save Y values */ \
SAVE_SI128((__m128i*)(y_ptr2+16), y); \
/* vertical subsampling of u/v values */ \
u2_tmp = _mm_avg_epu8(u1, u2); \
v2_tmp = _mm_avg_epu8(v1, v2); \
/* horizontal subsampling of u/v values */ \
u1 = _mm_packus_epi16(_mm_srl_epi16(u1_tmp, _mm_cvtsi32_si128(8)), _mm_srl_epi16(u2_tmp, _mm_cvtsi32_si128(8))); \
v1 = _mm_packus_epi16(_mm_srl_epi16(v1_tmp, _mm_cvtsi32_si128(8)), _mm_srl_epi16(v2_tmp, _mm_cvtsi32_si128(8))); \
u2 = _mm_packus_epi16(_mm_and_si128(u1_tmp, _mm_set1_epi16(0xFF)), _mm_and_si128(u2_tmp, _mm_set1_epi16(0xFF))); \
v2 = _mm_packus_epi16(_mm_and_si128(v1_tmp, _mm_set1_epi16(0xFF)), _mm_and_si128(v2_tmp, _mm_set1_epi16(0xFF))); \
u1 = _mm_avg_epu8(u1, u2); \
v1 = _mm_avg_epu8(v1, v2); \
SAVE_SI128((__m128i*)(u_ptr), u1); \
SAVE_SI128((__m128i*)(v_ptr), v1);
#endif
/* SDL doesn't use these atm and compiling them adds seconds onto the build. --ryan.
void SDL_TARGETING("sse2") rgb24_yuv420_sse(uint32_t width, uint32_t height,
const uint8_t *RGB, uint32_t RGB_stride,
uint8_t *Y, uint8_t *U, uint8_t *V, uint32_t Y_stride, uint32_t UV_stride,
YCbCrType yuv_type)
{
#define LOAD_SI128 _mm_load_si128
#define SAVE_SI128 _mm_stream_si128
const RGB2YUVParam *const param = &(RGB2YUV[yuv_type]);
uint32_t xpos, ypos;
for(ypos=0; ypos<(height-1); ypos+=2)
{
const uint8_t *rgb_ptr1=RGB+ypos*RGB_stride,
*rgb_ptr2=RGB+(ypos+1)*RGB_stride;
uint8_t *y_ptr1=Y+ypos*Y_stride,
*y_ptr2=Y+(ypos+1)*Y_stride,
*u_ptr=U+(ypos/2)*UV_stride,
*v_ptr=V+(ypos/2)*UV_stride;
for(xpos=0; xpos<(width-31); xpos+=32)
{
RGB2YUV_32
rgb_ptr1+=96;
rgb_ptr2+=96;
y_ptr1+=32;
y_ptr2+=32;
u_ptr+=16;
v_ptr+=16;
}
}
#undef LOAD_SI128
#undef SAVE_SI128
}
void SDL_TARGETING("sse2") rgb24_yuv420_sseu(uint32_t width, uint32_t height,
const uint8_t *RGB, uint32_t RGB_stride,
uint8_t *Y, uint8_t *U, uint8_t *V, uint32_t Y_stride, uint32_t UV_stride,
YCbCrType yuv_type)
{
#define LOAD_SI128 _mm_loadu_si128
#define SAVE_SI128 _mm_storeu_si128
const RGB2YUVParam *const param = &(RGB2YUV[yuv_type]);
uint32_t xpos, ypos;
for(ypos=0; ypos<(height-1); ypos+=2)
{
const uint8_t *rgb_ptr1=RGB+ypos*RGB_stride,
*rgb_ptr2=RGB+(ypos+1)*RGB_stride;
uint8_t *y_ptr1=Y+ypos*Y_stride,
*y_ptr2=Y+(ypos+1)*Y_stride,
*u_ptr=U+(ypos/2)*UV_stride,
*v_ptr=V+(ypos/2)*UV_stride;
for(xpos=0; xpos<(width-31); xpos+=32)
{
RGB2YUV_32
rgb_ptr1+=96;
rgb_ptr2+=96;
y_ptr1+=32;
y_ptr2+=32;
u_ptr+=16;
v_ptr+=16;
}
}
#undef LOAD_SI128
#undef SAVE_SI128
}
*/
#endif // SDL_SSE2_INTRINSICS
#endif // SDL_HAVE_YUV
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_sse.h
+241
View File
@@ -0,0 +1,241 @@
#ifdef SDL_SSE2_INTRINSICS
#include "yuv_rgb_common.h"
// yuv to rgb, sse implementation
// pointers must be 16 byte aligned, and strides must be divisable by 16
void yuv420_rgb565_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_rgb24_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_rgba_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_bgra_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_argb_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_abgr_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgb565_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgb24_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgba_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_bgra_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_argb_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_abgr_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgb565_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgb24_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgba_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_bgra_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_argb_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_abgr_sse(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
// yuv to rgb, sse implementation
// pointers do not need to be 16 byte aligned
void yuv420_rgb565_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_rgb24_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_rgba_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_bgra_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_argb_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_abgr_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgb565_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgb24_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgba_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_bgra_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_argb_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_abgr_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgb565_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgb24_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgba_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_bgra_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_argb_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_abgr_sseu(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
// rgb to yuv, sse implementation
// pointers must be 16 byte aligned, and strides must be divisible by 16
void rgb24_yuv420_sse(
uint32_t width, uint32_t height,
const uint8_t *rgb, uint32_t rgb_stride,
uint8_t *y, uint8_t *u, uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
YCbCrType yuv_type);
// rgb to yuv, sse implementation
// pointers do not need to be 16 byte aligned
void rgb24_yuv420_sseu(
uint32_t width, uint32_t height,
const uint8_t *rgb, uint32_t rgb_stride,
uint8_t *y, uint8_t *u, uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
YCbCrType yuv_type);
#endif
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_sse_func.h
+529
View File
@@ -0,0 +1,529 @@
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
/* You need to define the following macros before including this file:
SSE_FUNCTION_NAME
STD_FUNCTION_NAME
YUV_FORMAT
RGB_FORMAT
*/
/* You may define the following macro, which affects generated code:
SSE_ALIGNED
*/
#ifdef SSE_ALIGNED
/* Unaligned instructions seem faster, even on aligned data? */
/*
#define LOAD_SI128 _mm_load_si128
#define SAVE_SI128 _mm_stream_si128
*/
#define LOAD_SI128 _mm_loadu_si128
#define SAVE_SI128 _mm_storeu_si128
#else
#define LOAD_SI128 _mm_loadu_si128
#define SAVE_SI128 _mm_storeu_si128
#endif
#define UV2RGB_16(U,V,R1,G1,B1,R2,G2,B2) \
r_tmp = _mm_mullo_epi16(V, _mm_set1_epi16(param->v_r_factor)); \
g_tmp = _mm_add_epi16( \
_mm_mullo_epi16(U, _mm_set1_epi16(param->u_g_factor)), \
_mm_mullo_epi16(V, _mm_set1_epi16(param->v_g_factor))); \
b_tmp = _mm_mullo_epi16(U, _mm_set1_epi16(param->u_b_factor)); \
R1 = _mm_unpacklo_epi16(r_tmp, r_tmp); \
G1 = _mm_unpacklo_epi16(g_tmp, g_tmp); \
B1 = _mm_unpacklo_epi16(b_tmp, b_tmp); \
R2 = _mm_unpackhi_epi16(r_tmp, r_tmp); \
G2 = _mm_unpackhi_epi16(g_tmp, g_tmp); \
B2 = _mm_unpackhi_epi16(b_tmp, b_tmp); \
#define ADD_Y2RGB_16(Y1,Y2,R1,G1,B1,R2,G2,B2) \
Y1 = _mm_mullo_epi16(_mm_sub_epi16(Y1, _mm_set1_epi16(param->y_shift)), _mm_set1_epi16(param->y_factor)); \
Y2 = _mm_mullo_epi16(_mm_sub_epi16(Y2, _mm_set1_epi16(param->y_shift)), _mm_set1_epi16(param->y_factor)); \
\
R1 = _mm_srai_epi16(_mm_add_epi16(R1, Y1), PRECISION); \
G1 = _mm_srai_epi16(_mm_add_epi16(G1, Y1), PRECISION); \
B1 = _mm_srai_epi16(_mm_add_epi16(B1, Y1), PRECISION); \
R2 = _mm_srai_epi16(_mm_add_epi16(R2, Y2), PRECISION); \
G2 = _mm_srai_epi16(_mm_add_epi16(G2, Y2), PRECISION); \
B2 = _mm_srai_epi16(_mm_add_epi16(B2, Y2), PRECISION); \
#define PACK_RGB565_32(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4) \
{ \
__m128i red_mask, tmp1, tmp2, tmp3, tmp4; \
\
red_mask = _mm_set1_epi16((unsigned short)0xF800); \
RGB1 = _mm_and_si128(_mm_unpacklo_epi8(_mm_setzero_si128(), R1), red_mask); \
RGB2 = _mm_and_si128(_mm_unpackhi_epi8(_mm_setzero_si128(), R1), red_mask); \
RGB3 = _mm_and_si128(_mm_unpacklo_epi8(_mm_setzero_si128(), R2), red_mask); \
RGB4 = _mm_and_si128(_mm_unpackhi_epi8(_mm_setzero_si128(), R2), red_mask); \
tmp1 = _mm_slli_epi16(_mm_srli_epi16(_mm_unpacklo_epi8(G1, _mm_setzero_si128()), 2), 5); \
tmp2 = _mm_slli_epi16(_mm_srli_epi16(_mm_unpackhi_epi8(G1, _mm_setzero_si128()), 2), 5); \
tmp3 = _mm_slli_epi16(_mm_srli_epi16(_mm_unpacklo_epi8(G2, _mm_setzero_si128()), 2), 5); \
tmp4 = _mm_slli_epi16(_mm_srli_epi16(_mm_unpackhi_epi8(G2, _mm_setzero_si128()), 2), 5); \
RGB1 = _mm_or_si128(RGB1, tmp1); \
RGB2 = _mm_or_si128(RGB2, tmp2); \
RGB3 = _mm_or_si128(RGB3, tmp3); \
RGB4 = _mm_or_si128(RGB4, tmp4); \
tmp1 = _mm_srli_epi16(_mm_unpacklo_epi8(B1, _mm_setzero_si128()), 3); \
tmp2 = _mm_srli_epi16(_mm_unpackhi_epi8(B1, _mm_setzero_si128()), 3); \
tmp3 = _mm_srli_epi16(_mm_unpacklo_epi8(B2, _mm_setzero_si128()), 3); \
tmp4 = _mm_srli_epi16(_mm_unpackhi_epi8(B2, _mm_setzero_si128()), 3); \
RGB1 = _mm_or_si128(RGB1, tmp1); \
RGB2 = _mm_or_si128(RGB2, tmp2); \
RGB3 = _mm_or_si128(RGB3, tmp3); \
RGB4 = _mm_or_si128(RGB4, tmp4); \
}
#define PACK_RGB24_32_STEP1(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
RGB1 = _mm_packus_epi16(_mm_and_si128(R1,_mm_set1_epi16(0xFF)), _mm_and_si128(R2,_mm_set1_epi16(0xFF))); \
RGB2 = _mm_packus_epi16(_mm_and_si128(G1,_mm_set1_epi16(0xFF)), _mm_and_si128(G2,_mm_set1_epi16(0xFF))); \
RGB3 = _mm_packus_epi16(_mm_and_si128(B1,_mm_set1_epi16(0xFF)), _mm_and_si128(B2,_mm_set1_epi16(0xFF))); \
RGB4 = _mm_packus_epi16(_mm_srli_epi16(R1,8), _mm_srli_epi16(R2,8)); \
RGB5 = _mm_packus_epi16(_mm_srli_epi16(G1,8), _mm_srli_epi16(G2,8)); \
RGB6 = _mm_packus_epi16(_mm_srli_epi16(B1,8), _mm_srli_epi16(B2,8)); \
#define PACK_RGB24_32_STEP2(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
R1 = _mm_packus_epi16(_mm_and_si128(RGB1,_mm_set1_epi16(0xFF)), _mm_and_si128(RGB2,_mm_set1_epi16(0xFF))); \
R2 = _mm_packus_epi16(_mm_and_si128(RGB3,_mm_set1_epi16(0xFF)), _mm_and_si128(RGB4,_mm_set1_epi16(0xFF))); \
G1 = _mm_packus_epi16(_mm_and_si128(RGB5,_mm_set1_epi16(0xFF)), _mm_and_si128(RGB6,_mm_set1_epi16(0xFF))); \
G2 = _mm_packus_epi16(_mm_srli_epi16(RGB1,8), _mm_srli_epi16(RGB2,8)); \
B1 = _mm_packus_epi16(_mm_srli_epi16(RGB3,8), _mm_srli_epi16(RGB4,8)); \
B2 = _mm_packus_epi16(_mm_srli_epi16(RGB5,8), _mm_srli_epi16(RGB6,8)); \
#define PACK_RGB24_32(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
PACK_RGB24_32_STEP1(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
PACK_RGB24_32_STEP2(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
PACK_RGB24_32_STEP1(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
PACK_RGB24_32_STEP2(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
PACK_RGB24_32_STEP1(R1, R2, G1, G2, B1, B2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6) \
#define PACK_RGBA_32(R1, R2, G1, G2, B1, B2, A1, A2, RGB1, RGB2, RGB3, RGB4, RGB5, RGB6, RGB7, RGB8) \
{ \
__m128i lo_ab, hi_ab, lo_gr, hi_gr; \
\
lo_ab = _mm_unpacklo_epi8( A1, B1 ); \
hi_ab = _mm_unpackhi_epi8( A1, B1 ); \
lo_gr = _mm_unpacklo_epi8( G1, R1 ); \
hi_gr = _mm_unpackhi_epi8( G1, R1 ); \
RGB1 = _mm_unpacklo_epi16( lo_ab, lo_gr ); \
RGB2 = _mm_unpackhi_epi16( lo_ab, lo_gr ); \
RGB3 = _mm_unpacklo_epi16( hi_ab, hi_gr ); \
RGB4 = _mm_unpackhi_epi16( hi_ab, hi_gr ); \
\
lo_ab = _mm_unpacklo_epi8( A2, B2 ); \
hi_ab = _mm_unpackhi_epi8( A2, B2 ); \
lo_gr = _mm_unpacklo_epi8( G2, R2 ); \
hi_gr = _mm_unpackhi_epi8( G2, R2 ); \
RGB5 = _mm_unpacklo_epi16( lo_ab, lo_gr ); \
RGB6 = _mm_unpackhi_epi16( lo_ab, lo_gr ); \
RGB7 = _mm_unpacklo_epi16( hi_ab, hi_gr ); \
RGB8 = _mm_unpackhi_epi16( hi_ab, hi_gr ); \
}
#if RGB_FORMAT == RGB_FORMAT_RGB565
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
\
PACK_RGB565_32(r_8_11, r_8_12, g_8_11, g_8_12, b_8_11, b_8_12, rgb_1, rgb_2, rgb_3, rgb_4) \
\
PACK_RGB565_32(r_8_21, r_8_22, g_8_21, g_8_22, b_8_21, b_8_22, rgb_5, rgb_6, rgb_7, rgb_8) \
#elif RGB_FORMAT == RGB_FORMAT_RGB24
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6; \
__m128i rgb_7, rgb_8, rgb_9, rgb_10, rgb_11, rgb_12; \
\
PACK_RGB24_32(r_8_11, r_8_12, g_8_11, g_8_12, b_8_11, b_8_12, rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6) \
\
PACK_RGB24_32(r_8_21, r_8_22, g_8_21, g_8_22, b_8_21, b_8_22, rgb_7, rgb_8, rgb_9, rgb_10, rgb_11, rgb_12) \
#elif RGB_FORMAT == RGB_FORMAT_RGBA
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = _mm_set1_epi8((unsigned char)0xFF); \
\
PACK_RGBA_32(r_8_11, r_8_12, g_8_11, g_8_12, b_8_11, b_8_12, a, a, rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
\
PACK_RGBA_32(r_8_21, r_8_22, g_8_21, g_8_22, b_8_21, b_8_22, a, a, rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#elif RGB_FORMAT == RGB_FORMAT_BGRA
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = _mm_set1_epi8((unsigned char)0xFF); \
\
PACK_RGBA_32(b_8_11, b_8_12, g_8_11, g_8_12, r_8_11, r_8_12, a, a, rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
\
PACK_RGBA_32(b_8_21, b_8_22, g_8_21, g_8_22, r_8_21, r_8_22, a, a, rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#elif RGB_FORMAT == RGB_FORMAT_ARGB
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = _mm_set1_epi8((unsigned char)0xFF); \
\
PACK_RGBA_32(a, a, r_8_11, r_8_12, g_8_11, g_8_12, b_8_11, b_8_12, rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
\
PACK_RGBA_32(a, a, r_8_21, r_8_22, g_8_21, g_8_22, b_8_21, b_8_22, rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#elif RGB_FORMAT == RGB_FORMAT_ABGR
#define PACK_PIXEL \
__m128i rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8; \
__m128i rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16; \
__m128i a = _mm_set1_epi8((unsigned char)0xFF); \
\
PACK_RGBA_32(a, a, b_8_11, b_8_12, g_8_11, g_8_12, r_8_11, r_8_12, rgb_1, rgb_2, rgb_3, rgb_4, rgb_5, rgb_6, rgb_7, rgb_8) \
\
PACK_RGBA_32(a, a, b_8_21, b_8_22, g_8_21, g_8_22, r_8_21, r_8_22, rgb_9, rgb_10, rgb_11, rgb_12, rgb_13, rgb_14, rgb_15, rgb_16) \
#else
#error PACK_PIXEL unimplemented
#endif
#if RGB_FORMAT == RGB_FORMAT_RGB565
#define SAVE_LINE1 \
SAVE_SI128((__m128i*)(rgb_ptr1), rgb_1); \
SAVE_SI128((__m128i*)(rgb_ptr1+16), rgb_2); \
SAVE_SI128((__m128i*)(rgb_ptr1+32), rgb_3); \
SAVE_SI128((__m128i*)(rgb_ptr1+48), rgb_4); \
#define SAVE_LINE2 \
SAVE_SI128((__m128i*)(rgb_ptr2), rgb_5); \
SAVE_SI128((__m128i*)(rgb_ptr2+16), rgb_6); \
SAVE_SI128((__m128i*)(rgb_ptr2+32), rgb_7); \
SAVE_SI128((__m128i*)(rgb_ptr2+48), rgb_8); \
#elif RGB_FORMAT == RGB_FORMAT_RGB24
#define SAVE_LINE1 \
SAVE_SI128((__m128i*)(rgb_ptr1), rgb_1); \
SAVE_SI128((__m128i*)(rgb_ptr1+16), rgb_2); \
SAVE_SI128((__m128i*)(rgb_ptr1+32), rgb_3); \
SAVE_SI128((__m128i*)(rgb_ptr1+48), rgb_4); \
SAVE_SI128((__m128i*)(rgb_ptr1+64), rgb_5); \
SAVE_SI128((__m128i*)(rgb_ptr1+80), rgb_6); \
#define SAVE_LINE2 \
SAVE_SI128((__m128i*)(rgb_ptr2), rgb_7); \
SAVE_SI128((__m128i*)(rgb_ptr2+16), rgb_8); \
SAVE_SI128((__m128i*)(rgb_ptr2+32), rgb_9); \
SAVE_SI128((__m128i*)(rgb_ptr2+48), rgb_10); \
SAVE_SI128((__m128i*)(rgb_ptr2+64), rgb_11); \
SAVE_SI128((__m128i*)(rgb_ptr2+80), rgb_12); \
#elif RGB_FORMAT == RGB_FORMAT_RGBA || RGB_FORMAT == RGB_FORMAT_BGRA || \
RGB_FORMAT == RGB_FORMAT_ARGB || RGB_FORMAT == RGB_FORMAT_ABGR
#define SAVE_LINE1 \
SAVE_SI128((__m128i*)(rgb_ptr1), rgb_1); \
SAVE_SI128((__m128i*)(rgb_ptr1+16), rgb_2); \
SAVE_SI128((__m128i*)(rgb_ptr1+32), rgb_3); \
SAVE_SI128((__m128i*)(rgb_ptr1+48), rgb_4); \
SAVE_SI128((__m128i*)(rgb_ptr1+64), rgb_5); \
SAVE_SI128((__m128i*)(rgb_ptr1+80), rgb_6); \
SAVE_SI128((__m128i*)(rgb_ptr1+96), rgb_7); \
SAVE_SI128((__m128i*)(rgb_ptr1+112), rgb_8); \
#define SAVE_LINE2 \
SAVE_SI128((__m128i*)(rgb_ptr2), rgb_9); \
SAVE_SI128((__m128i*)(rgb_ptr2+16), rgb_10); \
SAVE_SI128((__m128i*)(rgb_ptr2+32), rgb_11); \
SAVE_SI128((__m128i*)(rgb_ptr2+48), rgb_12); \
SAVE_SI128((__m128i*)(rgb_ptr2+64), rgb_13); \
SAVE_SI128((__m128i*)(rgb_ptr2+80), rgb_14); \
SAVE_SI128((__m128i*)(rgb_ptr2+96), rgb_15); \
SAVE_SI128((__m128i*)(rgb_ptr2+112), rgb_16); \
#else
#error SAVE_LINE unimplemented
#endif
#if YUV_FORMAT == YUV_FORMAT_420
#define READ_Y(y_ptr) \
y = LOAD_SI128((const __m128i*)(y_ptr)); \
#define READ_UV \
u = LOAD_SI128((const __m128i*)(u_ptr)); \
v = LOAD_SI128((const __m128i*)(v_ptr)); \
#elif YUV_FORMAT == YUV_FORMAT_422
#define READ_Y(y_ptr) \
{ \
__m128i y1, y2; \
y1 = _mm_srli_epi16(_mm_slli_epi16(LOAD_SI128((const __m128i*)(y_ptr)), 8), 8); \
y2 = _mm_srli_epi16(_mm_slli_epi16(LOAD_SI128((const __m128i*)(y_ptr+16)), 8), 8); \
y = _mm_packus_epi16(y1, y2); \
}
#define READ_UV \
{ \
__m128i u1, u2, u3, u4, v1, v2, v3, v4; \
u1 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(u_ptr)), 24), 24); \
u2 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(u_ptr+16)), 24), 24); \
u3 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(u_ptr+32)), 24), 24); \
u4 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(u_ptr+48)), 24), 24); \
u = _mm_packus_epi16(_mm_packs_epi32(u1, u2), _mm_packs_epi32(u3, u4)); \
v1 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(v_ptr)), 24), 24); \
v2 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(v_ptr+16)), 24), 24); \
v3 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(v_ptr+32)), 24), 24); \
v4 = _mm_srli_epi32(_mm_slli_epi32(LOAD_SI128((const __m128i*)(v_ptr+48)), 24), 24); \
v = _mm_packus_epi16(_mm_packs_epi32(v1, v2), _mm_packs_epi32(v3, v4)); \
}
#elif YUV_FORMAT == YUV_FORMAT_NV12
#define READ_Y(y_ptr) \
y = LOAD_SI128((const __m128i*)(y_ptr)); \
#define READ_UV \
{ \
__m128i u1, u2, v1, v2; \
u1 = _mm_srli_epi16(_mm_slli_epi16(LOAD_SI128((const __m128i*)(u_ptr)), 8), 8); \
u2 = _mm_srli_epi16(_mm_slli_epi16(LOAD_SI128((const __m128i*)(u_ptr+16)), 8), 8); \
u = _mm_packus_epi16(u1, u2); \
v1 = _mm_srli_epi16(_mm_slli_epi16(LOAD_SI128((const __m128i*)(v_ptr)), 8), 8); \
v2 = _mm_srli_epi16(_mm_slli_epi16(LOAD_SI128((const __m128i*)(v_ptr+16)), 8), 8); \
v = _mm_packus_epi16(v1, v2); \
}
#else
#error READ_UV unimplemented
#endif
#define YUV2RGB_32 \
__m128i r_tmp, g_tmp, b_tmp; \
__m128i r_16_1, g_16_1, b_16_1, r_16_2, g_16_2, b_16_2; \
__m128i r_uv_16_1, g_uv_16_1, b_uv_16_1, r_uv_16_2, g_uv_16_2, b_uv_16_2; \
__m128i y_16_1, y_16_2; \
__m128i y, u, v, u_16, v_16; \
__m128i r_8_11, g_8_11, b_8_11, r_8_21, g_8_21, b_8_21; \
__m128i r_8_12, g_8_12, b_8_12, r_8_22, g_8_22, b_8_22; \
\
READ_UV \
\
/* process first 16 pixels of first line */\
u_16 = _mm_unpacklo_epi8(u, _mm_setzero_si128()); \
v_16 = _mm_unpacklo_epi8(v, _mm_setzero_si128()); \
u_16 = _mm_add_epi16(u_16, _mm_set1_epi16(-128)); \
v_16 = _mm_add_epi16(v_16, _mm_set1_epi16(-128)); \
\
UV2RGB_16(u_16, v_16, r_16_1, g_16_1, b_16_1, r_16_2, g_16_2, b_16_2) \
r_uv_16_1=r_16_1; g_uv_16_1=g_16_1; b_uv_16_1=b_16_1; \
r_uv_16_2=r_16_2; g_uv_16_2=g_16_2; b_uv_16_2=b_16_2; \
\
READ_Y(y_ptr1) \
y_16_1 = _mm_unpacklo_epi8(y, _mm_setzero_si128()); \
y_16_2 = _mm_unpackhi_epi8(y, _mm_setzero_si128()); \
\
ADD_Y2RGB_16(y_16_1, y_16_2, r_16_1, g_16_1, b_16_1, r_16_2, g_16_2, b_16_2) \
\
r_8_11 = _mm_packus_epi16(r_16_1, r_16_2); \
g_8_11 = _mm_packus_epi16(g_16_1, g_16_2); \
b_8_11 = _mm_packus_epi16(b_16_1, b_16_2); \
\
/* process first 16 pixels of second line */\
r_16_1=r_uv_16_1; g_16_1=g_uv_16_1; b_16_1=b_uv_16_1; \
r_16_2=r_uv_16_2; g_16_2=g_uv_16_2; b_16_2=b_uv_16_2; \
\
READ_Y(y_ptr2) \
y_16_1 = _mm_unpacklo_epi8(y, _mm_setzero_si128()); \
y_16_2 = _mm_unpackhi_epi8(y, _mm_setzero_si128()); \
\
ADD_Y2RGB_16(y_16_1, y_16_2, r_16_1, g_16_1, b_16_1, r_16_2, g_16_2, b_16_2) \
\
r_8_21 = _mm_packus_epi16(r_16_1, r_16_2); \
g_8_21 = _mm_packus_epi16(g_16_1, g_16_2); \
b_8_21 = _mm_packus_epi16(b_16_1, b_16_2); \
\
/* process last 16 pixels of first line */\
u_16 = _mm_unpackhi_epi8(u, _mm_setzero_si128()); \
v_16 = _mm_unpackhi_epi8(v, _mm_setzero_si128()); \
u_16 = _mm_add_epi16(u_16, _mm_set1_epi16(-128)); \
v_16 = _mm_add_epi16(v_16, _mm_set1_epi16(-128)); \
\
UV2RGB_16(u_16, v_16, r_16_1, g_16_1, b_16_1, r_16_2, g_16_2, b_16_2) \
r_uv_16_1=r_16_1; g_uv_16_1=g_16_1; b_uv_16_1=b_16_1; \
r_uv_16_2=r_16_2; g_uv_16_2=g_16_2; b_uv_16_2=b_16_2; \
\
READ_Y(y_ptr1+16*y_pixel_stride) \
y_16_1 = _mm_unpacklo_epi8(y, _mm_setzero_si128()); \
y_16_2 = _mm_unpackhi_epi8(y, _mm_setzero_si128()); \
\
ADD_Y2RGB_16(y_16_1, y_16_2, r_16_1, g_16_1, b_16_1, r_16_2, g_16_2, b_16_2) \
\
r_8_12 = _mm_packus_epi16(r_16_1, r_16_2); \
g_8_12 = _mm_packus_epi16(g_16_1, g_16_2); \
b_8_12 = _mm_packus_epi16(b_16_1, b_16_2); \
\
/* process last 16 pixels of second line */\
r_16_1=r_uv_16_1; g_16_1=g_uv_16_1; b_16_1=b_uv_16_1; \
r_16_2=r_uv_16_2; g_16_2=g_uv_16_2; b_16_2=b_uv_16_2; \
\
READ_Y(y_ptr2+16*y_pixel_stride) \
y_16_1 = _mm_unpacklo_epi8(y, _mm_setzero_si128()); \
y_16_2 = _mm_unpackhi_epi8(y, _mm_setzero_si128()); \
\
ADD_Y2RGB_16(y_16_1, y_16_2, r_16_1, g_16_1, b_16_1, r_16_2, g_16_2, b_16_2) \
\
r_8_22 = _mm_packus_epi16(r_16_1, r_16_2); \
g_8_22 = _mm_packus_epi16(g_16_1, g_16_2); \
b_8_22 = _mm_packus_epi16(b_16_1, b_16_2); \
\
void SDL_TARGETING("sse2") SSE_FUNCTION_NAME(uint32_t width, uint32_t height,
const uint8_t *Y, const uint8_t *U, const uint8_t *V, uint32_t Y_stride, uint32_t UV_stride,
uint8_t *RGB, uint32_t RGB_stride,
YCbCrType yuv_type)
{
const YUV2RGBParam *const param = &(YUV2RGB[yuv_type]);
#if YUV_FORMAT == YUV_FORMAT_420
const int y_pixel_stride = 1;
const int uv_pixel_stride = 1;
const int uv_x_sample_interval = 2;
const int uv_y_sample_interval = 2;
#elif YUV_FORMAT == YUV_FORMAT_422
const int y_pixel_stride = 2;
const int uv_pixel_stride = 4;
const int uv_x_sample_interval = 2;
const int uv_y_sample_interval = 1;
#elif YUV_FORMAT == YUV_FORMAT_NV12
const int y_pixel_stride = 1;
const int uv_pixel_stride = 2;
const int uv_x_sample_interval = 2;
const int uv_y_sample_interval = 2;
#endif
#if RGB_FORMAT == RGB_FORMAT_RGB565
const int rgb_pixel_stride = 2;
#elif RGB_FORMAT == RGB_FORMAT_RGB24
const int rgb_pixel_stride = 3;
#elif RGB_FORMAT == RGB_FORMAT_RGBA || RGB_FORMAT == RGB_FORMAT_BGRA || \
RGB_FORMAT == RGB_FORMAT_ARGB || RGB_FORMAT == RGB_FORMAT_ABGR
const int rgb_pixel_stride = 4;
#else
#error Unknown RGB pixel size
#endif
#if YUV_FORMAT == YUV_FORMAT_NV12
/* For NV12 formats (where U/V are interleaved)
* SSE READ_UV does an invalid read access at the very last pixel.
* As a workaround. Make sure not to decode the last column using assembly but with STD fallback path.
* see https://github.com/libsdl-org/SDL/issues/4841
*/
const int fix_read_nv12 = ((width & 31) == 0);
#else
const int fix_read_nv12 = 0;
#endif
#if YUV_FORMAT == YUV_FORMAT_422
/* Avoid invalid read on last line */
const int fix_read_422 = 1;
#else
const int fix_read_422 = 0;
#endif
if (width >= 32) {
uint32_t xpos, ypos;
for(ypos=0; ypos<(height-(uv_y_sample_interval-1)) - fix_read_422; ypos+=uv_y_sample_interval)
{
const uint8_t *y_ptr1=Y+ypos*Y_stride,
*y_ptr2=Y+(ypos+1)*Y_stride,
*u_ptr=U+(ypos/uv_y_sample_interval)*UV_stride,
*v_ptr=V+(ypos/uv_y_sample_interval)*UV_stride;
uint8_t *rgb_ptr1=RGB+ypos*RGB_stride,
*rgb_ptr2=RGB+(ypos+1)*RGB_stride;
for(xpos=0; xpos<(width-31) - fix_read_nv12; xpos+=32)
{
YUV2RGB_32
{
PACK_PIXEL
SAVE_LINE1
if (uv_y_sample_interval > 1)
{
SAVE_LINE2
}
}
y_ptr1+=32*y_pixel_stride;
y_ptr2+=32*y_pixel_stride;
u_ptr+=32*uv_pixel_stride/uv_x_sample_interval;
v_ptr+=32*uv_pixel_stride/uv_x_sample_interval;
rgb_ptr1+=32*rgb_pixel_stride;
rgb_ptr2+=32*rgb_pixel_stride;
}
}
if (fix_read_422) {
const uint8_t *y_ptr=Y+ypos*Y_stride,
*u_ptr=U+(ypos/uv_y_sample_interval)*UV_stride,
*v_ptr=V+(ypos/uv_y_sample_interval)*UV_stride;
uint8_t *rgb_ptr=RGB+ypos*RGB_stride;
STD_FUNCTION_NAME(width, 1, y_ptr, u_ptr, v_ptr, Y_stride, UV_stride, rgb_ptr, RGB_stride, yuv_type);
ypos += uv_y_sample_interval;
}
/* Catch the last line, if needed */
if (uv_y_sample_interval == 2 && ypos == (height-1))
{
const uint8_t *y_ptr=Y+ypos*Y_stride,
*u_ptr=U+(ypos/uv_y_sample_interval)*UV_stride,
*v_ptr=V+(ypos/uv_y_sample_interval)*UV_stride;
uint8_t *rgb_ptr=RGB+ypos*RGB_stride;
STD_FUNCTION_NAME(width, 1, y_ptr, u_ptr, v_ptr, Y_stride, UV_stride, rgb_ptr, RGB_stride, yuv_type);
}
}
/* Catch the right column, if needed */
{
uint32_t converted = (width & ~31);
if (fix_read_nv12) {
converted -= 32;
}
if (converted != width)
{
const uint8_t *y_ptr=Y+converted*y_pixel_stride,
*u_ptr=U+converted*uv_pixel_stride/uv_x_sample_interval,
*v_ptr=V+converted*uv_pixel_stride/uv_x_sample_interval;
uint8_t *rgb_ptr=RGB+converted*rgb_pixel_stride;
STD_FUNCTION_NAME(width-converted, height, y_ptr, u_ptr, v_ptr, Y_stride, UV_stride, rgb_ptr, RGB_stride, yuv_type);
}
}
}
#undef SSE_FUNCTION_NAME
#undef STD_FUNCTION_NAME
#undef YUV_FORMAT
#undef RGB_FORMAT
#undef SSE_ALIGNED
#undef LOAD_SI128
#undef SAVE_SI128
#undef UV2RGB_16
#undef ADD_Y2RGB_16
#undef PACK_RGB24_32_STEP1
#undef PACK_RGB24_32_STEP2
#undef PACK_RGB24_32
#undef PACK_RGBA_32
#undef PACK_PIXEL
#undef SAVE_LINE1
#undef SAVE_LINE2
#undef READ_Y
#undef READ_UV
#undef YUV2RGB_32
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_std.c
+200
View File
@@ -0,0 +1,200 @@
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
#include "SDL_internal.h"
#ifdef SDL_HAVE_YUV
#include "yuv_rgb_internal.h"
// divide by PRECISION_FACTOR and clamp to [0:255] interval
// input must be in the [-128*PRECISION_FACTOR:384*PRECISION_FACTOR] range
static uint8_t clampU8(int32_t v)
{
static const uint8_t lut[512] =
{0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,
47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,
91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,
126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,
159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,
192,193,194,195,196,197,198,199,200,201,202,203,204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219,220,221,222,223,224,
225,226,227,228,229,230,231,232,233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254,255,
255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,
255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255
};
return lut[((v+128*PRECISION_FACTOR)>>PRECISION)&511];
}
static uint16_t clamp10(int32_t v)
{
v >>= PRECISION;
if (v < 0) {
return 0;
} else if (v > 1023) {
return 1023;
} else {
return (uint16_t)v;
}
}
#define YUV_BITS 8
#define STD_FUNCTION_NAME yuv420_rgb565_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGB565
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv420_rgb24_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGB24
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv420_rgba_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_RGBA
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv420_bgra_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_BGRA
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv420_argb_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ARGB
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv420_abgr_std
#define YUV_FORMAT YUV_FORMAT_420
#define RGB_FORMAT RGB_FORMAT_ABGR
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv422_rgb565_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGB565
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv422_rgb24_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGB24
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv422_rgba_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_RGBA
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv422_bgra_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_BGRA
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv422_argb_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_ARGB
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuv422_abgr_std
#define YUV_FORMAT YUV_FORMAT_422
#define RGB_FORMAT RGB_FORMAT_ABGR
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuvnv12_rgb565_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGB565
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuvnv12_rgb24_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGB24
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuvnv12_rgba_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_RGBA
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuvnv12_bgra_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_BGRA
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuvnv12_argb_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_ARGB
#include "yuv_rgb_std_func.h"
#define STD_FUNCTION_NAME yuvnv12_abgr_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_ABGR
#include "yuv_rgb_std_func.h"
#undef YUV_BITS
#define YUV_BITS 10
#define STD_FUNCTION_NAME yuvp010_xbgr2101010_std
#define YUV_FORMAT YUV_FORMAT_NV12
#define RGB_FORMAT RGB_FORMAT_XBGR2101010
#include "yuv_rgb_std_func.h"
void rgb24_yuv420_std(
uint32_t width, uint32_t height,
const uint8_t *RGB, uint32_t RGB_stride,
uint8_t *Y, uint8_t *U, uint8_t *V, uint32_t Y_stride, uint32_t UV_stride,
YCbCrType yuv_type)
{
const RGB2YUVParam *const param = &(RGB2YUV[yuv_type]);
uint32_t x, y;
for(y=0; y<(height-1); y+=2)
{
const uint8_t *rgb_ptr1=RGB+y*RGB_stride,
*rgb_ptr2=RGB+(y+1)*RGB_stride;
uint8_t *y_ptr1=Y+y*Y_stride,
*y_ptr2=Y+(y+1)*Y_stride,
*u_ptr=U+(y/2)*UV_stride,
*v_ptr=V+(y/2)*UV_stride;
for(x=0; x<(width-1); x+=2)
{
// compute yuv for the four pixels, u and v values are summed
int32_t y_tmp, u_tmp, v_tmp;
y_tmp = param->matrix[0][0]*rgb_ptr1[0] + param->matrix[0][1]*rgb_ptr1[1] + param->matrix[0][2]*rgb_ptr1[2];
u_tmp = param->matrix[1][0]*rgb_ptr1[0] + param->matrix[1][1]*rgb_ptr1[1] + param->matrix[1][2]*rgb_ptr1[2];
v_tmp = param->matrix[2][0]*rgb_ptr1[0] + param->matrix[2][1]*rgb_ptr1[1] + param->matrix[2][2]*rgb_ptr1[2];
y_ptr1[0]=clampU8(y_tmp+((param->y_shift)<<PRECISION));
y_tmp = param->matrix[0][0]*rgb_ptr1[3] + param->matrix[0][1]*rgb_ptr1[4] + param->matrix[0][2]*rgb_ptr1[5];
u_tmp += param->matrix[1][0]*rgb_ptr1[3] + param->matrix[1][1]*rgb_ptr1[4] + param->matrix[1][2]*rgb_ptr1[5];
v_tmp += param->matrix[2][0]*rgb_ptr1[3] + param->matrix[2][1]*rgb_ptr1[4] + param->matrix[2][2]*rgb_ptr1[5];
y_ptr1[1]=clampU8(y_tmp+((param->y_shift)<<PRECISION));
y_tmp = param->matrix[0][0]*rgb_ptr2[0] + param->matrix[0][1]*rgb_ptr2[1] + param->matrix[0][2]*rgb_ptr2[2];
u_tmp += param->matrix[1][0]*rgb_ptr2[0] + param->matrix[1][1]*rgb_ptr2[1] + param->matrix[1][2]*rgb_ptr2[2];
v_tmp += param->matrix[2][0]*rgb_ptr2[0] + param->matrix[2][1]*rgb_ptr2[1] + param->matrix[2][2]*rgb_ptr2[2];
y_ptr2[0]=clampU8(y_tmp+((param->y_shift)<<PRECISION));
y_tmp = param->matrix[0][0]*rgb_ptr2[3] + param->matrix[0][1]*rgb_ptr2[4] + param->matrix[0][2]*rgb_ptr2[5];
u_tmp += param->matrix[1][0]*rgb_ptr2[3] + param->matrix[1][1]*rgb_ptr2[4] + param->matrix[1][2]*rgb_ptr2[5];
v_tmp += param->matrix[2][0]*rgb_ptr2[3] + param->matrix[2][1]*rgb_ptr2[4] + param->matrix[2][2]*rgb_ptr2[5];
y_ptr2[1]=clampU8(y_tmp+((param->y_shift)<<PRECISION));
u_ptr[0] = clampU8(u_tmp/4+(128<<PRECISION));
v_ptr[0] = clampU8(v_tmp/4+(128<<PRECISION));
rgb_ptr1 += 6;
rgb_ptr2 += 6;
y_ptr1 += 2;
y_ptr2 += 2;
u_ptr += 1;
v_ptr += 1;
}
}
}
#endif /* SDL_HAVE_YUV */
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_std.h
+143
View File
@@ -0,0 +1,143 @@
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
// Provide optimized functions to convert images from 8bits yuv420 to rgb24 format
// There are a few slightly different variations of the YCbCr color space with different parameters that
// change the conversion matrix.
// The three most common YCbCr color space, defined by BT.601, BT.709 and JPEG standard are implemented here.
// See the respective standards for details
// The matrix values used are derived from http://www.equasys.de/colorconversion.html
// YUV420 is stored as three separate channels, with U and V (Cb and Cr) subsampled by a 2 factor
// For conversion from yuv to rgb, no interpolation is done, and the same UV value are used for 4 rgb pixels. This
// is suboptimal for image quality, but by far the fastest method.
// For all methods, width and height should be even, if not, the last row/column of the result image won't be affected.
// For sse methods, if the width if not divisable by 32, the last (width%32) pixels of each line won't be affected.
/*#include <stdint.h>*/
#include "yuv_rgb_common.h"
// yuv to rgb, standard c implementation
void yuv420_rgb565_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_rgb24_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_rgba_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_bgra_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_argb_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv420_abgr_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgb565_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgb24_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_rgba_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_bgra_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_argb_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuv422_abgr_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgb565_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgb24_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_rgba_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_bgra_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_argb_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvnv12_abgr_std(
uint32_t width, uint32_t height,
const uint8_t *y, const uint8_t *u, const uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
void yuvp010_xbgr2101010_std(
uint32_t width, uint32_t height,
const uint16_t *y, const uint16_t *u, const uint16_t *v, uint32_t y_stride, uint32_t uv_stride,
uint8_t *rgb, uint32_t rgb_stride,
YCbCrType yuv_type);
// rgb to yuv, standard c implementation
void rgb24_yuv420_std(
uint32_t width, uint32_t height,
const uint8_t *rgb, uint32_t rgb_stride,
uint8_t *y, uint8_t *u, uint8_t *v, uint32_t y_stride, uint32_t uv_stride,
YCbCrType yuv_type);
StgSharp.Native/sdl-src/video/yuv2rgb/yuv_rgb_std_func.h
+271
View File
@@ -0,0 +1,271 @@
// Copyright 2016 Adrien Descamps
// Distributed under BSD 3-Clause License
/* You need to define the following macros before including this file:
STD_FUNCTION_NAME
YUV_FORMAT
RGB_FORMAT
*/
#if RGB_FORMAT == RGB_FORMAT_RGB565
#define PACK_PIXEL(rgb_ptr) \
*(Uint16 *)rgb_ptr = \
((((Uint16)clampU8(y_tmp+r_tmp)) << 8 ) & 0xF800) | \
((((Uint16)clampU8(y_tmp+g_tmp)) << 3) & 0x07E0) | \
(((Uint16)clampU8(y_tmp+b_tmp)) >> 3); \
rgb_ptr += 2; \
#elif RGB_FORMAT == RGB_FORMAT_RGB24
#define PACK_PIXEL(rgb_ptr) \
rgb_ptr[0] = clampU8(y_tmp+r_tmp); \
rgb_ptr[1] = clampU8(y_tmp+g_tmp); \
rgb_ptr[2] = clampU8(y_tmp+b_tmp); \
rgb_ptr += 3; \
#elif RGB_FORMAT == RGB_FORMAT_RGBA
#define PACK_PIXEL(rgb_ptr) \
*(Uint32 *)rgb_ptr = \
(((Uint32)clampU8(y_tmp+r_tmp)) << 24) | \
(((Uint32)clampU8(y_tmp+g_tmp)) << 16) | \
(((Uint32)clampU8(y_tmp+b_tmp)) << 8) | \
0x000000FF; \
rgb_ptr += 4; \
#elif RGB_FORMAT == RGB_FORMAT_BGRA
#define PACK_PIXEL(rgb_ptr) \
*(Uint32 *)rgb_ptr = \
(((Uint32)clampU8(y_tmp+b_tmp)) << 24) | \
(((Uint32)clampU8(y_tmp+g_tmp)) << 16) | \
(((Uint32)clampU8(y_tmp+r_tmp)) << 8) | \
0x000000FF; \
rgb_ptr += 4; \
#elif RGB_FORMAT == RGB_FORMAT_ARGB
#define PACK_PIXEL(rgb_ptr) \
*(Uint32 *)rgb_ptr = \
0xFF000000 | \
(((Uint32)clampU8(y_tmp+r_tmp)) << 16) | \
(((Uint32)clampU8(y_tmp+g_tmp)) << 8) | \
(((Uint32)clampU8(y_tmp+b_tmp)) << 0); \
rgb_ptr += 4; \
#elif RGB_FORMAT == RGB_FORMAT_ABGR
#define PACK_PIXEL(rgb_ptr) \
*(Uint32 *)rgb_ptr = \
0xFF000000 | \
(((Uint32)clampU8(y_tmp+b_tmp)) << 16) | \
(((Uint32)clampU8(y_tmp+g_tmp)) << 8) | \
(((Uint32)clampU8(y_tmp+r_tmp)) << 0); \
rgb_ptr += 4; \
#elif RGB_FORMAT == RGB_FORMAT_XBGR2101010
#define PACK_PIXEL(rgb_ptr) \
*(Uint32 *)rgb_ptr = \
0xC0000000 | \
(((Uint32)clamp10(y_tmp+b_tmp)) << 20) | \
(((Uint32)clamp10(y_tmp+g_tmp)) << 10) | \
(((Uint32)clamp10(y_tmp+r_tmp)) << 0); \
rgb_ptr += 4; \
#else
#error PACK_PIXEL unimplemented
#endif
#ifdef _MSC_VER /* Visual Studio analyzer can't tell that we're building this with different constants */
#pragma warning(push)
#pragma warning(disable : 6239)
#endif
#undef YUV_TYPE
#if YUV_BITS > 8
#define YUV_TYPE uint16_t
#else
#define YUV_TYPE uint8_t
#endif
#undef UV_OFFSET
#define UV_OFFSET (1 << ((YUV_BITS)-1))
#undef GET
#if YUV_BITS == 10
#define GET(X) ((X) >> 6)
#else
#define GET(X) (X)
#endif
void STD_FUNCTION_NAME(
uint32_t width, uint32_t height,
const YUV_TYPE *Y, const YUV_TYPE *U, const YUV_TYPE *V, uint32_t Y_stride, uint32_t UV_stride,
uint8_t *RGB, uint32_t RGB_stride,
YCbCrType yuv_type)
{
const YUV2RGBParam *const param = &(YUV2RGB[yuv_type]);
#if YUV_FORMAT == YUV_FORMAT_420
#define y_pixel_stride 1
#define uv_pixel_stride 1
#define uv_x_sample_interval 2
#define uv_y_sample_interval 2
#elif YUV_FORMAT == YUV_FORMAT_422
#define y_pixel_stride 2
#define uv_pixel_stride 4
#define uv_x_sample_interval 2
#define uv_y_sample_interval 1
#elif YUV_FORMAT == YUV_FORMAT_NV12
#define y_pixel_stride 1
#define uv_pixel_stride 2
#define uv_x_sample_interval 2
#define uv_y_sample_interval 2
#endif
Y_stride /= sizeof(YUV_TYPE);
UV_stride /= sizeof(YUV_TYPE);
uint32_t x, y;
for(y=0; y<(height-(uv_y_sample_interval-1)); y+=uv_y_sample_interval)
{
const YUV_TYPE *y_ptr1=Y+y*Y_stride,
*u_ptr=U+(y/uv_y_sample_interval)*UV_stride,
*v_ptr=V+(y/uv_y_sample_interval)*UV_stride;
#if uv_y_sample_interval > 1
const YUV_TYPE *y_ptr2=Y+(y+1)*Y_stride;
#endif
uint8_t *rgb_ptr1=RGB+y*RGB_stride;
#if uv_y_sample_interval > 1
uint8_t *rgb_ptr2=RGB+(y+1)*RGB_stride;
#endif
for(x=0; x<(width-(uv_x_sample_interval-1)); x+=uv_x_sample_interval)
{
// Compute U and V contributions, common to the four pixels
int32_t u_tmp = (GET(*u_ptr)-UV_OFFSET);
int32_t v_tmp = (GET(*v_ptr)-UV_OFFSET);
int32_t r_tmp = (v_tmp*param->v_r_factor);
int32_t g_tmp = (u_tmp*param->u_g_factor + v_tmp*param->v_g_factor);
int32_t b_tmp = (u_tmp*param->u_b_factor);
// Compute the Y contribution for each pixel
int32_t y_tmp = (GET(y_ptr1[0]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr1);
y_tmp = (GET(y_ptr1[y_pixel_stride]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr1);
#if uv_y_sample_interval > 1
y_tmp = (GET(y_ptr2[0]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr2);
y_tmp = (GET(y_ptr2[y_pixel_stride]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr2);
#endif
y_ptr1+=2*y_pixel_stride;
#if uv_y_sample_interval > 1
y_ptr2+=2*y_pixel_stride;
#endif
u_ptr+=2*uv_pixel_stride/uv_x_sample_interval;
v_ptr+=2*uv_pixel_stride/uv_x_sample_interval;
}
/* Catch the last pixel, if needed */
if (uv_x_sample_interval == 2 && x == (width-1))
{
// Compute U and V contributions, common to the four pixels
int32_t u_tmp = (GET(*u_ptr)-UV_OFFSET);
int32_t v_tmp = (GET(*v_ptr)-UV_OFFSET);
int32_t r_tmp = (v_tmp*param->v_r_factor);
int32_t g_tmp = (u_tmp*param->u_g_factor + v_tmp*param->v_g_factor);
int32_t b_tmp = (u_tmp*param->u_b_factor);
// Compute the Y contribution for each pixel
int32_t y_tmp = (GET(y_ptr1[0]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr1);
#if uv_y_sample_interval > 1
y_tmp = (GET(y_ptr2[0]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr2);
#endif
}
}
/* Catch the last line, if needed */
if (uv_y_sample_interval == 2 && y == (height-1))
{
const YUV_TYPE *y_ptr1=Y+y*Y_stride,
*u_ptr=U+(y/uv_y_sample_interval)*UV_stride,
*v_ptr=V+(y/uv_y_sample_interval)*UV_stride;
uint8_t *rgb_ptr1=RGB+y*RGB_stride;
for(x=0; x<(width-(uv_x_sample_interval-1)); x+=uv_x_sample_interval)
{
// Compute U and V contributions, common to the four pixels
int32_t u_tmp = (GET(*u_ptr)-UV_OFFSET);
int32_t v_tmp = (GET(*v_ptr)-UV_OFFSET);
int32_t r_tmp = (v_tmp*param->v_r_factor);
int32_t g_tmp = (u_tmp*param->u_g_factor + v_tmp*param->v_g_factor);
int32_t b_tmp = (u_tmp*param->u_b_factor);
// Compute the Y contribution for each pixel
int32_t y_tmp = (GET(y_ptr1[0]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr1);
y_tmp = (GET(y_ptr1[y_pixel_stride]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr1);
y_ptr1+=2*y_pixel_stride;
u_ptr+=2*uv_pixel_stride/uv_x_sample_interval;
v_ptr+=2*uv_pixel_stride/uv_x_sample_interval;
}
/* Catch the last pixel, if needed */
if (uv_x_sample_interval == 2 && x == (width-1))
{
// Compute U and V contributions, common to the four pixels
int32_t u_tmp = (GET(*u_ptr)-UV_OFFSET);
int32_t v_tmp = (GET(*v_ptr)-UV_OFFSET);
int32_t r_tmp = (v_tmp*param->v_r_factor);
int32_t g_tmp = (u_tmp*param->u_g_factor + v_tmp*param->v_g_factor);
int32_t b_tmp = (u_tmp*param->u_b_factor);
// Compute the Y contribution for each pixel
int32_t y_tmp = (GET(y_ptr1[0]-param->y_shift)*param->y_factor);
PACK_PIXEL(rgb_ptr1);
}
}
#undef y_pixel_stride
#undef uv_pixel_stride
#undef uv_x_sample_interval
#undef uv_y_sample_interval
}
#ifdef _MSC_VER
#pragma warning(pop)
#endif
#undef STD_FUNCTION_NAME
#undef YUV_FORMAT
#undef RGB_FORMAT
#undef PACK_PIXEL