1 /*****************************************************************************
2 * video_yuvmmx.c: MMX YUV transformation functions
3 * Provides functions to perform the YUV conversion.
4 *****************************************************************************
5 * Copyright (C) 1999, 2000 VideoLAN
6 * $Id: video_yuvmmx.c,v 1.7 2001/03/21 13:42:34 sam Exp $
8 * Authors: Samuel Hocevar <sam@zoy.org>
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; either version 2 of the License, or
13 * (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 02111-1307, USA.
24 *****************************************************************************/
26 #define MODULE_NAME yuvmmx
27 #include "modules_inner.h"
29 /*****************************************************************************
31 *****************************************************************************/
34 #include <math.h> /* exp(), pow() */
35 #include <errno.h> /* ENOMEM */
36 #include <stdlib.h> /* free() */
37 #include <string.h> /* strerror() */
48 #include "video_output.h"
50 #include "video_common.h"
54 static int yuv_Probe ( probedata_t *p_data );
55 static int yuv_Init ( vout_thread_t *p_vout );
56 static int yuv_Reset ( vout_thread_t *p_vout );
57 static void yuv_End ( vout_thread_t *p_vout );
59 static void SetYUV ( vout_thread_t *p_vout );
61 /*****************************************************************************
62 * Functions exported as capabilities. They are declared as static so that
63 * we don't pollute the namespace too much.
64 *****************************************************************************/
65 void _M( yuv_getfunctions )( function_list_t * p_function_list )
67 p_function_list->pf_probe = yuv_Probe;
68 p_function_list->functions.yuv.pf_init = yuv_Init;
69 p_function_list->functions.yuv.pf_reset = yuv_Reset;
70 p_function_list->functions.yuv.pf_end = yuv_End;
73 /*****************************************************************************
74 * yuv_Probe: tests probe the audio device and return a score
75 *****************************************************************************
76 * This function tries to open the DSP and returns a score to the plugin
77 * manager so that it can choose the most appropriate one.
78 *****************************************************************************/
79 static int yuv_Probe( probedata_t *p_data )
81 /* Test for MMX support in the CPU */
82 if( TestCPU( CPU_CAPABILITY_MMX ) )
84 if( TestMethod( YUV_METHOD_VAR, "yuvmmx" ) )
99 /*****************************************************************************
100 * yuv_Init: allocate and initialize translations tables
101 *****************************************************************************
102 * This function will allocate memory to store translation tables, depending
103 * of the screen depth.
104 *****************************************************************************/
105 static int yuv_Init( vout_thread_t *p_vout )
107 size_t tables_size; /* tables size, in bytes */
109 /* Computes tables size for 8bbp only */
110 if( p_vout->i_bytes_per_pixel == 1 )
112 tables_size = sizeof( u8 )
113 * (p_vout->b_grayscale ? GRAY_TABLE_SIZE : PALETTE_TABLE_SIZE);
115 /* Allocate memory */
116 p_vout->yuv.p_base = malloc( tables_size );
117 if( p_vout->yuv.p_base == NULL )
119 intf_ErrMsg("error: %s", strerror(ENOMEM));
125 p_vout->yuv.p_base = NULL;
128 /* Allocate memory for conversion buffer and offset array */
129 p_vout->yuv.p_buffer = malloc( VOUT_MAX_WIDTH * p_vout->i_bytes_per_pixel );
130 if( p_vout->yuv.p_buffer == NULL )
132 intf_ErrMsg("error: %s", strerror(ENOMEM));
133 free( p_vout->yuv.p_base );
136 p_vout->yuv.p_offset = malloc( p_vout->i_width * sizeof( int ) );
137 if( p_vout->yuv.p_offset == NULL )
139 intf_ErrMsg("error: %s", strerror(ENOMEM));
140 free( p_vout->yuv.p_base );
141 free( p_vout->yuv.p_buffer );
145 /* Initialize tables */
150 /*****************************************************************************
151 * yuv_End: destroy translations tables
152 *****************************************************************************
153 * Free memory allocated by yuv_CCreate.
154 *****************************************************************************/
155 static void yuv_End( vout_thread_t *p_vout )
157 free( p_vout->yuv.p_base );
158 free( p_vout->yuv.p_buffer );
159 free( p_vout->yuv.p_offset );
162 /*****************************************************************************
163 * yuv_Reset: re-initialize translations tables
164 *****************************************************************************
165 * This function will initialize the tables allocated by vout_CreateTables and
166 * set functions pointers.
167 *****************************************************************************/
168 static int yuv_Reset( vout_thread_t *p_vout )
171 return( yuv_Init( p_vout ) );
174 /*****************************************************************************
175 * SetYUV: compute tables and set function pointers
176 *****************************************************************************/
177 static void SetYUV( vout_thread_t *p_vout )
179 int i_index; /* index in tables */
182 * Set pointers and build YUV tables
184 if( p_vout->b_grayscale )
186 /* Grayscale: build gray table */
187 if( p_vout->i_bytes_per_pixel == 1 )
189 u16 bright[256], transp[256];
191 for( i_index = 0; i_index < 256; i_index++)
193 bright[ i_index ] = i_index << 8;
194 transp[ i_index ] = 0;
196 /* the colors have been allocated, we can set the palette */
197 p_vout->pf_setpalette( p_vout, bright, bright, bright, transp );
198 p_vout->i_white_pixel = 0xff;
199 p_vout->i_black_pixel = 0x00;
200 p_vout->i_gray_pixel = 0x44;
201 p_vout->i_blue_pixel = 0x3b;
206 /* Color: build red, green and blue tables */
207 if( p_vout->i_bytes_per_pixel == 1 )
211 #define CLIP( x ) ( ((x < 0) ? 0 : (x > 255) ? 255 : x) << 8 )
213 #define U_GREEN_COEF ((int)(-0.391 * (1<<SHIFT) / 1.164))
214 #define U_BLUE_COEF ((int)(2.018 * (1<<SHIFT) / 1.164))
215 #define V_RED_COEF ((int)(1.596 * (1<<SHIFT) / 1.164))
216 #define V_GREEN_COEF ((int)(-0.813 * (1<<SHIFT) / 1.164))
222 u16 red[256], green[256], blue[256], transp[256];
223 unsigned char lookup[PALETTE_TABLE_SIZE];
225 p_vout->yuv.yuv.p_rgb8 = (u8 *)p_vout->yuv.p_base;
227 /* this loop calculates the intersection of an YUV box
228 * and the RGB cube. */
229 for ( y = 0; y <= 256; y += 16 )
231 for ( u = 0; u <= 256; u += 32 )
232 for ( v = 0; v <= 256; v += 32 )
234 uvr = (V_RED_COEF*(v-128)) >> SHIFT;
235 uvg = (U_GREEN_COEF*(u-128) + V_GREEN_COEF*(v-128)) >> SHIFT;
236 uvb = (U_BLUE_COEF*(u-128)) >> SHIFT;
241 if( r >= RGB_MIN && g >= RGB_MIN && b >= RGB_MIN
242 && r <= RGB_MAX && g <= RGB_MAX && b <= RGB_MAX )
244 /* this one should never happen unless someone fscked up my code */
245 if(j == 256) { intf_ErrMsg( "vout error: no colors left to build palette" ); break; }
247 /* clip the colors */
249 green[j] = CLIP( g );
255 p_vout->yuv.yuv.p_rgb8[i++] = j;
261 p_vout->yuv.yuv.p_rgb8[i++] = 0;
267 /* the colors have been allocated, we can set the palette */
268 /* there will eventually be a way to know which colors
269 * couldn't be allocated and try to find a replacement */
270 p_vout->pf_setpalette( p_vout, red, green, blue, transp );
272 p_vout->i_white_pixel = 0xff;
273 p_vout->i_black_pixel = 0x00;
274 p_vout->i_gray_pixel = 0x44;
275 p_vout->i_blue_pixel = 0x3b;
278 /* this loop allocates colors that got outside
280 for ( y = 0; y <= 256; y += 16 )
282 for ( u = 0; u <= 256; u += 32 )
284 for ( v = 0; v <= 256; v += 32 )
287 int dist, mindist = 100000000;
289 if( lookup[i] || y==0)
296 for( u2 = 0; u2 <= 256; u2 += 32 )
297 for( v2 = 0; v2 <= 256; v2 += 32 )
299 j = ((y>>4)<<7) + (u2>>5)*9 + (v2>>5);
300 dist = (u-u2)*(u-u2) + (v-v2)*(v-v2);
302 /* find the nearest color */
305 p_vout->yuv.yuv.p_rgb8[i] = p_vout->yuv.yuv.p_rgb8[j];
310 /* find the nearest color */
311 if( dist + 128 < mindist )
313 p_vout->yuv.yuv.p_rgb8[i] = p_vout->yuv.yuv.p_rgb8[j];
314 mindist = dist + 128;
326 * Set functions pointers
330 switch( p_vout->i_bytes_per_pixel)
333 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420YCbr8;
334 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422YCbr8;
335 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444YCbr8;
339 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420YCbr16;
340 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422YCbr16;
341 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444YCbr16;
345 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420YCbr24;
346 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422YCbr24;
347 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444YCbr24;
351 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420YCbr32;
352 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422YCbr32;
353 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444YCbr32;
357 else if( p_vout->b_grayscale )
360 switch( p_vout->i_bytes_per_pixel )
363 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertY4Gray8;
364 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertY4Gray8;
365 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertY4Gray8;
368 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertY4Gray16;
369 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertY4Gray16;
370 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertY4Gray16;
373 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420RGB24;
374 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertY4Gray24;
375 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertY4Gray24;
378 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420RGB32;
379 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertY4Gray32;
380 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertY4Gray32;
387 switch( p_vout->i_bytes_per_pixel )
390 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420RGB8;
391 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422RGB8;
392 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444RGB8;
395 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420RGB16;
396 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422RGB16;
397 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444RGB16;
400 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420RGB24;
401 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422RGB24;
402 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444RGB24;
405 p_vout->yuv.pf_yuv420 = (vout_yuv_convert_t *) ConvertYUV420RGB32;
406 p_vout->yuv.pf_yuv422 = (vout_yuv_convert_t *) ConvertYUV422RGB32;
407 p_vout->yuv.pf_yuv444 = (vout_yuv_convert_t *) ConvertYUV444RGB32;
413 /*****************************************************************************
414 * SetOffset: build offset array for conversion functions
415 *****************************************************************************
416 * This function will build an offset array used in later conversion functions.
417 * It will also set horizontal and vertical scaling indicators. If b_double
418 * is set, the p_offset structure has interleaved Y and U/V offsets.
419 *****************************************************************************/
420 void SetOffset( int i_width, int i_height, int i_pic_width, int i_pic_height,
421 boolean_t *pb_h_scaling, int *pi_v_scaling,
422 int *p_offset, boolean_t b_double )
424 int i_x; /* x position in destination */
425 int i_scale_count; /* modulo counter */
428 * Prepare horizontal offset array
430 if( i_pic_width - i_width == 0 )
432 /* No horizontal scaling: YUV conversion is done directly to picture */
435 else if( i_pic_width - i_width > 0 )
437 /* Prepare scaling array for horizontal extension */
439 i_scale_count = i_pic_width;
440 for( i_x = i_width; i_x--; )
442 while( (i_scale_count -= i_width) > 0 )
447 i_scale_count += i_pic_width;
450 else /* if( i_pic_width - i_width < 0 ) */
452 /* Prepare scaling array for horizontal reduction */
454 i_scale_count = i_width;
455 for( i_x = i_pic_width; i_x--; )
458 while( (i_scale_count -= i_pic_width) > 0 )
463 i_scale_count += i_width;
468 * Set vertical scaling indicator
470 if( i_pic_height - i_height == 0 )
474 else if( i_pic_height - i_height > 0 )
478 else /* if( i_pic_height - i_height < 0 ) */