-rw-r--r-- | noncore/multimedia/opieplayer2/alphablend.c | 753 | ||||
-rw-r--r-- | noncore/multimedia/opieplayer2/alphablend.h | 57 | ||||
-rw-r--r-- | noncore/multimedia/opieplayer2/yuv2rgb.c | 3160 | ||||
-rw-r--r-- | noncore/multimedia/opieplayer2/yuv2rgb.h | 151 | ||||
-rw-r--r-- | noncore/multimedia/opieplayer2/yuv2rgb_mlib.c | 313 | ||||
-rw-r--r-- | noncore/multimedia/opieplayer2/yuv2rgb_mmx.c | 1047 |
6 files changed, 5481 insertions, 0 deletions
diff --git a/noncore/multimedia/opieplayer2/alphablend.c b/noncore/multimedia/opieplayer2/alphablend.c new file mode 100644 index 0000000..57f6013 --- a/dev/null +++ b/noncore/multimedia/opieplayer2/alphablend.c @@ -0,0 +1,753 @@ +//TOAST_SPU will define ALL spu entries - no matter the tranparency +//#define TOAST_SPU +/* #define PRIV_CLUT */ +/* Currently only blend_yuv(..) works */ +/* + * + * Copyright (C) James Courtier-Dutton James@superbug.demon.co.uk - July 2001 + * + * Copyright (C) 2000 Thomas Mirlacher + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * + * The author may be reached as <dent@linuxvideo.org> + * + *------------------------------------------------------------ + * + */ + +/* +#define LOG_BLEND_YUV +*/ + +#include <string.h> +#include <stdlib.h> +#include <stdio.h> +#include <inttypes.h> + +#include <xine/video_out.h> +#include "alphablend.h" + + +#define BLEND_COLOR(dst, src, mask, o) ((((src&mask)*o + ((dst&mask)*(0x0f-o)))/0xf) & mask) + +#define BLEND_BYTE(dst, src, o) (((src)*o + ((dst)*(0xf-o)))/0xf) + +static void mem_blend16(uint16_t *mem, uint16_t clr, uint8_t o, int len) { + uint16_t *limit = mem + len; + while (mem < limit) { + *mem = + BLEND_COLOR(*mem, clr, 0xf800, o) | + BLEND_COLOR(*mem, clr, 0x07e0, o) | + BLEND_COLOR(*mem, clr, 0x001f, o); + mem++; + } +} + +static void mem_blend24(uint8_t *mem, uint8_t r, uint8_t g, uint8_t b, + uint8_t o, int len) { + uint8_t *limit = mem + len*3; + while (mem < limit) { + *mem = BLEND_BYTE(*mem, r, o); + mem++; + *mem = BLEND_BYTE(*mem, g, o); + mem++; + *mem = BLEND_BYTE(*mem, b, o); + mem++; + } +} + +static void mem_blend24_32(uint8_t *mem, uint8_t r, uint8_t g, uint8_t b, + uint8_t o, int len) { + uint8_t *limit = mem + len*4; + while (mem < limit) { + *mem = BLEND_BYTE(*mem, r, o); + mem++; + *mem = BLEND_BYTE(*mem, g, o); + mem++; + *mem = BLEND_BYTE(*mem, b, o); + mem += 2; + } +} + +static void mem_blend32(uint8_t *mem, uint8_t *src, uint8_t o, int len) { + uint8_t *limit = mem + len*4; + while (mem < limit) { + *mem = BLEND_BYTE(*mem, src[0], o); + mem++; + *mem = BLEND_BYTE(*mem, src[1], o); + mem++; + *mem = BLEND_BYTE(*mem, src[2], o); + mem++; + *mem = BLEND_BYTE(*mem, src[3], o); + mem++; + } +} + + +/* + * Some macros for fixed point arithmetic. + * + * The blend_rgb* routines perform rle image scaling using + * scale factors that are expressed as integers scaled with + * a factor of 2**16. + * + * INT_TO_SCALED()/SCALED_TO_INT() converts from integer + * to scaled fixed point and back. + */ +#define SCALE_SHIFT 16 +#define SCALE_FACTOR (1<<SCALE_SHIFT) +#define INT_TO_SCALED(i) ((i) << SCALE_SHIFT) +#define SCALED_TO_INT(sc) ((sc) >> SCALE_SHIFT) + + +static rle_elem_t * +rle_img_advance_line(rle_elem_t *rle, rle_elem_t *rle_limit, int w) +{ + int x; + + for (x = 0; x < w && rle < rle_limit; ) { + x += rle->len; + rle++; + } + return rle; +} + + +void blend_rgb16 (uint8_t * img, vo_overlay_t * img_overl, + int img_width, int img_height, + int dst_width, int dst_height) +{ + uint8_t *trans; + clut_t* clut = (clut_t*) img_overl->clip_color; + + int src_width = img_overl->width; + int src_height = img_overl->height; + rle_elem_t *rle = img_overl->rle; + rle_elem_t *rle_limit = rle + img_overl->num_rle; + int x, y, x1_scaled, x2_scaled; + int dy, dy_step, x_scale; /* scaled 2**SCALE_SHIFT */ + int clip_right; + uint16_t *img_pix; + + dy_step = INT_TO_SCALED(dst_height) / img_height; + x_scale = INT_TO_SCALED(img_width) / dst_width; + + img_pix = (uint16_t *) img + + (img_overl->y * img_height / dst_height) * img_width + + (img_overl->x * img_width / dst_width); + + trans = img_overl->clip_trans; + + /* avoid wraping overlay if drawing to small image */ + if( (img_overl->x + img_overl->clip_right) < dst_width ) + clip_right = img_overl->clip_right; + else + clip_right = dst_width - 1 - img_overl->x; + + /* avoid buffer overflow */ + if( (src_height + img_overl->y) >= dst_height ) + src_height = dst_height - 1 - img_overl->y; + + for (y = dy = 0; y < src_height && rle < rle_limit;) { + int mask = !(img_overl->clip_top > y || img_overl->clip_bottom < y); + rle_elem_t *rle_start = rle; + + for (x = x1_scaled = 0; x < src_width;) { + uint8_t clr; + uint16_t o; + int rlelen; + + clr = rle->color; + o = trans[clr]; + rlelen = rle->len; + + if (o && mask) { + /* threat cases where clipping border is inside rle->len pixels */ + if ( img_overl->clip_left > x ) { + if( img_overl->clip_left < x + rlelen ) { + x1_scaled = SCALED_TO_INT( img_overl->clip_left * x_scale ); + rlelen -= img_overl->clip_left - x; + x += img_overl->clip_left - x; + } else { + o = 0; + } + } else if( clip_right < x + rlelen ) { + if( clip_right > x ) { + x2_scaled = SCALED_TO_INT( clip_right * x_scale); + mem_blend16(img_pix+x1_scaled, *((uint16_t *)&clut[clr]), o, + x2_scaled-x1_scaled); + o = 0; + } else { + o = 0; + } + } + } + + x2_scaled = SCALED_TO_INT((x + rlelen) * x_scale); + if (o && mask) { + mem_blend16(img_pix+x1_scaled, *((uint16_t *)&clut[clr]), o, x2_scaled-x1_scaled); + } + + x1_scaled = x2_scaled; + x += rlelen; + rle++; + if (rle >= rle_limit) break; + } + + img_pix += img_width; + dy += dy_step; + if (dy >= INT_TO_SCALED(1)) { + dy -= INT_TO_SCALED(1); + ++y; + while (dy >= INT_TO_SCALED(1)) { + rle = rle_img_advance_line(rle, rle_limit, src_width); + dy -= INT_TO_SCALED(1); + ++y; + } + } else { + rle = rle_start; /* y-scaling, reuse the last rle encoded line */ + } + } +} + +void blend_rgb24 (uint8_t * img, vo_overlay_t * img_overl, + int img_width, int img_height, + int dst_width, int dst_height) +{ + clut_t* clut = (clut_t*) img_overl->clip_color; + uint8_t *trans; + int src_width = img_overl->width; + int src_height = img_overl->height; + rle_elem_t *rle = img_overl->rle; + rle_elem_t *rle_limit = rle + img_overl->num_rle; + int x, y, x1_scaled, x2_scaled; + int dy, dy_step, x_scale; /* scaled 2**SCALE_SHIFT */ + int clip_right; + uint8_t *img_pix; + + dy_step = INT_TO_SCALED(dst_height) / img_height; + x_scale = INT_TO_SCALED(img_width) / dst_width; + + img_pix = img + 3 * ( (img_overl->y * img_height / dst_height) * img_width + + (img_overl->x * img_width / dst_width)); + + trans = img_overl->clip_trans; + + /* avoid wraping overlay if drawing to small image */ + if( (img_overl->x + img_overl->clip_right) < dst_width ) + clip_right = img_overl->clip_right; + else + clip_right = dst_width - 1 - img_overl->x; + + /* avoid buffer overflow */ + if( (src_height + img_overl->y) >= dst_height ) + src_height = dst_height - 1 - img_overl->y; + + for (dy = y = 0; y < src_height && rle < rle_limit; ) { + int mask = !(img_overl->clip_top > y || img_overl->clip_bottom < y); + rle_elem_t *rle_start = rle; + + for (x = x1_scaled = 0; x < src_width;) { + uint8_t clr; + uint16_t o; + int rlelen; + + clr = rle->color; + o = trans[clr]; + rlelen = rle->len; + + if (o && mask) { + /* threat cases where clipping border is inside rle->len pixels */ + if ( img_overl->clip_left > x ) { + if( img_overl->clip_left < x + rlelen ) { + x1_scaled = SCALED_TO_INT( img_overl->clip_left * x_scale ); + rlelen -= img_overl->clip_left - x; + x += img_overl->clip_left - x; + } else { + o = 0; + } + } else if( clip_right < x + rlelen ) { + if( clip_right > x ) { + x2_scaled = SCALED_TO_INT( clip_right * x_scale); + mem_blend24(img_pix + x1_scaled*3, clut[clr].cb, + clut[clr].cr, clut[clr].y, + o, x2_scaled-x1_scaled); + o = 0; + } else { + o = 0; + } + } + } + + x2_scaled = SCALED_TO_INT((x + rlelen) * x_scale); + if (o && mask) { + mem_blend24(img_pix + x1_scaled*3, clut[clr].cb, + clut[clr].cr, clut[clr].y, + o, x2_scaled-x1_scaled); + } + + x1_scaled = x2_scaled; + x += rlelen; + rle++; + if (rle >= rle_limit) break; + } + + img_pix += img_width * 3; + dy += dy_step; + if (dy >= INT_TO_SCALED(1)) { + dy -= INT_TO_SCALED(1); + ++y; + while (dy >= INT_TO_SCALED(1)) { + rle = rle_img_advance_line(rle, rle_limit, src_width); + dy -= INT_TO_SCALED(1); + ++y; + } + } else { + rle = rle_start; /* y-scaling, reuse the last rle encoded line */ + } + } +} + +void blend_rgb32 (uint8_t * img, vo_overlay_t * img_overl, + int img_width, int img_height, + int dst_width, int dst_height) +{ + clut_t* clut = (clut_t*) img_overl->clip_color; + uint8_t *trans; + int src_width = img_overl->width; + int src_height = img_overl->height; + rle_elem_t *rle = img_overl->rle; + rle_elem_t *rle_limit = rle + img_overl->num_rle; + int x, y, x1_scaled, x2_scaled; + int dy, dy_step, x_scale; /* scaled 2**SCALE_SHIFT */ + int clip_right; + uint8_t *img_pix; + + dy_step = INT_TO_SCALED(dst_height) / img_height; + x_scale = INT_TO_SCALED(img_width) / dst_width; + + img_pix = img + 4 * ( (img_overl->y * img_height / dst_height) * img_width + + (img_overl->x * img_width / dst_width)); + + trans = img_overl->clip_trans; + + /* avoid wraping overlay if drawing to small image */ + if( (img_overl->x + img_overl->clip_right) < dst_width ) + clip_right = img_overl->clip_right; + else + clip_right = dst_width - 1 - img_overl->x; + + /* avoid buffer overflow */ + if( (src_height + img_overl->y) >= dst_height ) + src_height = dst_height - 1 - img_overl->y; + + for (y = dy = 0; y < src_height && rle < rle_limit; ) { + int mask = !(img_overl->clip_top > y || img_overl->clip_bottom < y); + rle_elem_t *rle_start = rle; + + for (x = x1_scaled = 0; x < src_width;) { + uint8_t clr; + uint16_t o; + int rlelen; + + clr = rle->color; + o = trans[clr]; + rlelen = rle->len; + + if (o && mask) { + /* threat cases where clipping border is inside rle->len pixels */ + if ( img_overl->clip_left > x ) { + if( img_overl->clip_left < x + rlelen ) { + x1_scaled = SCALED_TO_INT( img_overl->clip_left * x_scale ); + rlelen -= img_overl->clip_left - x; + x += img_overl->clip_left - x; + } else { + o = 0; + } + } else if( clip_right < x + rlelen ) { + if( clip_right > x ) { + x2_scaled = SCALED_TO_INT( clip_right * x_scale); + mem_blend24_32(img_pix + x1_scaled*4, clut[clr].cb, + clut[clr].cr, clut[clr].y, + o, x2_scaled-x1_scaled); + o = 0; + } else { + o = 0; + } + } + } + + x2_scaled = SCALED_TO_INT((x + rlelen) * x_scale); + if (o && mask) { + mem_blend24_32(img_pix + x1_scaled*4, clut[clr].cb, + clut[clr].cr, clut[clr].y, + o, x2_scaled-x1_scaled); + } + + x1_scaled = x2_scaled; + x += rlelen; + rle++; + if (rle >= rle_limit) break; + } + + img_pix += img_width * 4; + dy += dy_step; + if (dy >= INT_TO_SCALED(1)) { + dy -= INT_TO_SCALED(1); + ++y; + while (dy >= INT_TO_SCALED(1)) { + rle = rle_img_advance_line(rle, rle_limit, src_width); + dy -= INT_TO_SCALED(1); + ++y; + } + } else { + rle = rle_start; /* y-scaling, reuse the last rle encoded line */ + } + } +} + +static void mem_blend8(uint8_t *mem, uint8_t val, uint8_t o, size_t sz) +{ + uint8_t *limit = mem + sz; + while (mem < limit) { + *mem = BLEND_BYTE(*mem, val, o); + mem++; + } +} + +void blend_yuv (uint8_t *dst_base[3], vo_overlay_t * img_overl, + int dst_width, int dst_height) +{ + clut_t *my_clut; + uint8_t *my_trans; + + int src_width = img_overl->width; + int src_height = img_overl->height; + rle_elem_t *rle = img_overl->rle; + rle_elem_t *rle_limit = rle + img_overl->num_rle; + int x_off = img_overl->x; + int y_off = img_overl->y; + int ymask,xmask; + int rle_this_bite; + int rle_remainder; + int rlelen; + int x, y; + int clip_right; + uint8_t clr=0; + + uint8_t *dst_y = dst_base[0] + dst_width * y_off + x_off; + uint8_t *dst_cr = dst_base[2] + + (y_off / 2) * (dst_width / 2) + (x_off / 2) + 1; + uint8_t *dst_cb = dst_base[1] + + (y_off / 2) * (dst_width / 2) + (x_off / 2) + 1; +#ifdef LOG_BLEND_YUV + printf("overlay_blend started x=%d, y=%d, w=%d h=%d\n",img_overl->x,img_overl->y,img_overl->width,img_overl->height); +#endif + my_clut = (clut_t*) img_overl->clip_color; + my_trans = img_overl->clip_trans; + + /* avoid wraping overlay if drawing to small image */ + if( (x_off + img_overl->clip_right) < dst_width ) + clip_right = img_overl->clip_right; + else + clip_right = dst_width - 1 - x_off; + + /* avoid buffer overflow */ + if( (src_height + y_off) >= dst_height ) + src_height = dst_height - 1 - y_off; + + rlelen=rle_remainder=0; + for (y = 0; y < src_height; y++) { + ymask = ((img_overl->clip_top > y) || (img_overl->clip_bottom < y)); + xmask = 0; +#ifdef LOG_BLEND_YUV + printf("X started ymask=%d y=%d src_height=%d\n",ymask, y, src_height); +#endif + + for (x = 0; x < src_width;) { + uint16_t o; +#ifdef LOG_BLEND_YUV + printf("1:rle_len=%d, remainder=%d, x=%d\n",rlelen, rle_remainder, x); +#endif + + if ((rlelen < 0) || (rle_remainder < 0)) { + printf("alphablend: major bug in blend_yuv < 0\n"); + } + if (rlelen == 0) { + rle_remainder = rlelen = rle->len; + clr = rle->color; + rle++; + } + if (rle_remainder == 0) { + rle_remainder = rlelen; + } + if ((rle_remainder + x) > src_width) { + /* Do something for long rlelengths */ + rle_remainder = src_width - x; + ; + } +#ifdef LOG_BLEND_YUV + printf("2:rle_len=%d, remainder=%d, x=%d\n",rlelen, rle_remainder, x); +#endif + + if (ymask == 0) { + if (x <= img_overl->clip_left) { + /* Starts outside clip area */ + if ((x + rle_remainder - 1) > img_overl->clip_left ) { +#ifdef LOG_BLEND_YUV + printf("Outside clip left %d, ending inside\n", img_overl->clip_left); +#endif + /* Cutting needed, starts outside, ends inside */ + rle_this_bite = (img_overl->clip_left - x + 1); + rle_remainder -= rle_this_bite; + rlelen -= rle_this_bite; + my_clut = (clut_t*) img_overl->color; + my_trans = img_overl->trans; + xmask = 0; + } else { +#ifdef LOG_BLEND_YUV + printf("Outside clip left %d, ending outside\n", img_overl->clip_left); +#endif + /* no cutting needed, starts outside, ends outside */ + rle_this_bite = rle_remainder; + rle_remainder = 0; + rlelen -= rle_this_bite; + my_clut = (clut_t*) img_overl->color; + my_trans = img_overl->trans; + xmask = 0; + } + } else if (x < clip_right) { + /* Starts inside clip area */ + if ((x + rle_remainder) > clip_right ) { +#ifdef LOG_BLEND_YUV + printf("Inside clip right %d, ending outside\n", clip_right); +#endif + /* Cutting needed, starts inside, ends outside */ + rle_this_bite = (clip_right - x); + rle_remainder -= rle_this_bite; + rlelen -= rle_this_bite; + my_clut = (clut_t*) img_overl->clip_color; + my_trans = img_overl->clip_trans; + xmask++; + } else { +#ifdef LOG_BLEND_YUV + printf("Inside clip right %d, ending inside\n", clip_right); +#endif + /* no cutting needed, starts inside, ends inside */ + rle_this_bite = rle_remainder; + rle_remainder = 0; + rlelen -= rle_this_bite; + my_clut = (clut_t*) img_overl->clip_color; + my_trans = img_overl->clip_trans; + xmask++; + } + } else if (x >= clip_right) { + /* Starts outside clip area, ends outsite clip area */ + if ((x + rle_remainder ) > src_width ) { +#ifdef LOG_BLEND_YUV + printf("Outside clip right %d, ending eol\n", clip_right); +#endif + /* Cutting needed, starts outside, ends at right edge */ + /* It should never reach here due to the earlier test of src_width */ + rle_this_bite = (src_width - x ); + rle_remainder -= rle_this_bite; + rlelen -= rle_this_bite; + my_clut = (clut_t*) img_overl->color; + my_trans = img_overl->trans; + xmask = 0; + } else { + /* no cutting needed, starts outside, ends outside */ +#ifdef LOG_BLEND_YUV + printf("Outside clip right %d, ending outside\n", clip_right); +#endif + rle_this_bite = rle_remainder; + rle_remainder = 0; + rlelen -= rle_this_bite; + my_clut = (clut_t*) img_overl->color; + my_trans = img_overl->trans; + xmask = 0; + } + } + } else { + /* Outside clip are due to y */ + /* no cutting needed, starts outside, ends outside */ + rle_this_bite = rle_remainder; + rle_remainder = 0; + rlelen -= rle_this_bite; + my_clut = (clut_t*) img_overl->color; + my_trans = img_overl->trans; + xmask = 0; + } + o = my_trans[clr]; +#ifdef LOG_BLEND_YUV + printf("Trans=%d clr=%d xmask=%d my_clut[clr]=%d\n",o, clr, xmask, my_clut[clr].y); +#endif + if (o) { + if(o >= 15) { + memset(dst_y + x, my_clut[clr].y, rle_this_bite); + if (y & 1) { + memset(dst_cr + (x >> 1), my_clut[clr].cr, (rle_this_bite+1) >> 1); + memset(dst_cb + (x >> 1), my_clut[clr].cb, (rle_this_bite+1) >> 1); + } + } else { + mem_blend8(dst_y + x, my_clut[clr].y, o, rle_this_bite); + if (y & 1) { + /* Blending cr and cb should use a different function, with pre -128 to each sample */ + mem_blend8(dst_cr + (x >> 1), my_clut[clr].cr, o, (rle_this_bite+1) >> 1); + mem_blend8(dst_cb + (x >> 1), my_clut[clr].cb, o, (rle_this_bite+1) >> 1); + } + } + + } +#ifdef LOG_BLEND_YUV + printf("rle_this_bite=%d, remainder=%d, x=%d\n",rle_this_bite, rle_remainder, x); +#endif + x += rle_this_bite; + if (rle >= rle_limit) { +#ifdef LOG_BLEND_YUV + printf("x-rle_limit\n"); +#endif + break; + } + } + if (rle >= rle_limit) { +#ifdef LOG_BLEND_YUV + printf("x-rle_limit\n"); +#endif + break; + } + + dst_y += dst_width; + + if (y & 1) { + dst_cr += (dst_width + 1) / 2; + dst_cb += (dst_width + 1) / 2; + } + } +#ifdef LOG_BLEND_YUV + printf("overlay_blend ended\n"); +#endif +} + +void blend_yuy2 (uint8_t * dst_img, vo_overlay_t * img_overl, + int dst_width, int dst_height) +{ + clut_t *my_clut; + uint8_t *my_trans; + + int src_width = img_overl->width; + int src_height = img_overl->height; + rle_elem_t *rle = img_overl->rle; + rle_elem_t *rle_limit = rle + img_overl->num_rle; + int x_off = img_overl->x; + int y_off = img_overl->y; + int mask; + int x, y; + int l; + int clip_right; + uint32_t yuy2; + + uint8_t *dst_y = dst_img + 2 * (dst_width * y_off + x_off); + uint8_t *dst; + + my_clut = (clut_t*) img_overl->clip_color; + my_trans = img_overl->clip_trans; + + /* avoid wraping overlay if drawing to small image */ + if( (x_off + img_overl->clip_right) < dst_width ) + clip_right = img_overl->clip_right; + else + clip_right = dst_width - 1 - x_off; + + /* avoid buffer overflow */ + if( (src_height + y_off) >= dst_height ) + src_height = dst_height - 1 - y_off; + + for (y = 0; y < src_height; y++) { + mask = !(img_overl->clip_top > y || img_overl->clip_bottom < y); + + dst = dst_y; + for (x = 0; x < src_width;) { + uint8_t clr; + uint16_t o; + int rlelen; + + clr = rle->color; + o = my_trans[clr]; + rlelen = rle->len; + + if (o && mask) { + /* threat cases where clipping border is inside rle->len pixels */ + if ( img_overl->clip_left > x ) { + if( img_overl->clip_left < x + rlelen ) { + rlelen -= img_overl->clip_left - x; + x += img_overl->clip_left - x; + } else { + o = 0; + } + } else if( clip_right < x + rlelen ) { + if( clip_right > x ) { + /* fixme: case not implemented */ + o = 0; + } else { + o = 0; + } + } + } + + + if (o && mask) { + l = rlelen>>1; + if( !(x & 1) ) { + yuy2 = my_clut[clr].y + (my_clut[clr].cb << 8) + + (my_clut[clr].y << 16) + (my_clut[clr].cr << 24); + } else { + yuy2 = my_clut[clr].y + (my_clut[clr].cr << 8) + + (my_clut[clr].y << 16) + (my_clut[clr].cb << 24); + } + + if (o >= 15) { + while(l--) { + *((uint32_t *)dst)++ = yuy2; + } + if(rlelen & 1) + *((uint16_t *)dst)++ = yuy2 & 0xffff; + } else { + if( l ) { + mem_blend32(dst, (uint8_t *)&yuy2, o, l); + dst += 4*l; + } + + if(rlelen & 1) { + *dst = BLEND_BYTE(*dst, *((uint8_t *)&yuy2), o); + dst++; + *dst = BLEND_BYTE(*dst, *((uint8_t *)&yuy2+1), o); + dst++; + } + } + } else { + dst += rlelen*2; + } + + x += rlelen; + rle++; + if (rle >= rle_limit) break; + } + if (rle >= rle_limit) break; + + dst_y += dst_width*2; + } +} diff --git a/noncore/multimedia/opieplayer2/alphablend.h b/noncore/multimedia/opieplayer2/alphablend.h new file mode 100644 index 0000000..7230f41 --- a/dev/null +++ b/noncore/multimedia/opieplayer2/alphablend.h @@ -0,0 +1,57 @@ +/* + * + * Copyright (C) 2000 Thomas Mirlacher + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * + * The author may be reached as <dent@linuxvideo.org> + * + *------------------------------------------------------------ + * + */ + +#ifndef __ALPHABLEND_H__ +#define __ALPHABLEND_H__ + +#include <xine/video_out.h> + +typedef struct { /* CLUT == Color LookUp Table */ + uint8_t cb : 8; + uint8_t cr : 8; + uint8_t y : 8; + uint8_t foo : 8; +} __attribute__ ((packed)) clut_t; + +void blend_rgb16 (uint8_t * img, vo_overlay_t * img_overl, + int img_width, int img_height, + int dst_width, int dst_height); + +void blend_rgb24 (uint8_t * img, vo_overlay_t * img_overl, + int img_width, int img_height, + int dst_width, int dst_height); + +void blend_rgb32 (uint8_t * img, vo_overlay_t * img_overl, + int img_width, int img_height, + int dst_width, int dst_height); + +void blend_yuv (uint8_t *dst_base[3], vo_overlay_t * img_overl, + int dst_width, int dst_height); + +void blend_yuy2 (uint8_t * dst_img, vo_overlay_t * img_overl, + int dst_width, int dst_height); + +void crop_overlay (vo_overlay_t * overlay); + +#endif diff --git a/noncore/multimedia/opieplayer2/yuv2rgb.c b/noncore/multimedia/opieplayer2/yuv2rgb.c new file mode 100644 index 0000000..d1d6627 --- a/dev/null +++ b/noncore/multimedia/opieplayer2/yuv2rgb.c @@ -0,0 +1,3160 @@ +/* + * yuv2rgb.c + * + * This file is part of xine, a unix video player. + * + * based on work from mpeg2dec: + * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * + * mpeg2dec is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * mpeg2dec is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + * $Id$ + */ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <inttypes.h> + +#include "yuv2rgb.h" +#include <xine/xineutils.h> + + +static int prof_scale_line = -1; + +static scale_line_func_t find_scale_line_func(int step); + + +const int32_t Inverse_Table_6_9[8][4] = { + {117504, 138453, 13954, 34903}, /* no sequence_display_extension */ + {117504, 138453, 13954, 34903}, /* ITU-R Rec. 709 (1990) */ + {104597, 132201, 25675, 53279}, /* unspecified */ + {104597, 132201, 25675, 53279}, /* reserved */ + {104448, 132798, 24759, 53109}, /* FCC */ + {104597, 132201, 25675, 53279}, /* ITU-R Rec. 624-4 System B, G */ + {104597, 132201, 25675, 53279}, /* SMPTE 170M */ + {117579, 136230, 16907, 35559} /* SMPTE 240M (1987) */ +}; + + +static void *my_malloc_aligned (size_t alignment, size_t size, void **chunk) { + + char *pMem; + + pMem = xine_xmalloc (size+alignment); + + *chunk = pMem; + + while ((int) pMem % alignment) + pMem++; + + return pMem; +} + + +static int yuv2rgb_configure (yuv2rgb_t *this, + int source_width, int source_height, + int y_stride, int uv_stride, + int dest_width, int dest_height, + int rgb_stride) { + /* + printf ("yuv2rgb setup (%d x %d => %d x %d)\n", source_width, source_height, + dest_width, dest_height); + */ + if (prof_scale_line == -1) + prof_scale_line = xine_profiler_allocate_slot("xshm scale line"); + + this->source_width = source_width; + this->source_height = source_height; + this->y_stride = y_stride; + this->uv_stride = uv_stride; + this->dest_width = dest_width; + this->dest_height = dest_height; + this->rgb_stride = rgb_stride; + + if (this->y_chunk) { + free (this->y_chunk); + this->y_buffer = this->y_chunk = NULL; + } + if (this->u_chunk) { + free (this->u_chunk); + this->u_buffer = this->u_chunk = NULL; + } + if (this->v_chunk) { + free (this->v_chunk); + this->v_buffer = this->v_chunk = NULL; + } + + + this->step_dx = source_width * 32768 / dest_width; + this->step_dy = source_height * 32768 / dest_height; + + this->scale_line = find_scale_line_func(this->step_dx); + + if ((source_width == dest_width) && (source_height == dest_height)) { + this->do_scale = 0; + + /* + * space for two y-lines (for yuv2rgb_mlib) + * u,v subsampled 2:1 + */ + this->y_buffer = my_malloc_aligned (16, 2*dest_width, &this->y_chunk); + if (!this->y_buffer) + return 0; + this->u_buffer = my_malloc_aligned (16, (dest_width+1)/2, &this->u_chunk); + if (!this->u_buffer) + return 0; + this->v_buffer = my_malloc_aligned (16, (dest_width+1)/2, &this->v_chunk); + if (!this->v_buffer) + return 0; + + } else { + this->do_scale = 1; + + /* + * space for two y-lines (for yuv2rgb_mlib) + * u,v subsampled 2:1 + */ + this->y_buffer = my_malloc_aligned (16, 2*dest_width, &this->y_chunk); + if (!this->y_buffer) + return 0; + this->u_buffer = my_malloc_aligned (16, (dest_width+1)/2, &this->u_chunk); + if (!this->u_buffer) + return 0; + this->v_buffer = my_malloc_aligned (16, (dest_width+1)/2, &this->v_chunk); + if (!this->v_buffer) + return 0; + } + return 1; +} + + +static void scale_line_gen (uint8_t *source, uint8_t *dest, + int width, int step) { + + /* + * scales a yuv source row to a dest row, with interpolation + * (good quality, but slow) + */ + int p1; + int p2; + int dx; + + xine_profiler_start_count(prof_scale_line); + + p1 = *source++; + p2 = *source++; + dx = 0; + + /* + * the following code has been optimized by Scott Smith <ssmith@akamai.com>: + * + * ok now I have a meaningful optimization for yuv2rgb.c:scale_line_gen. + * it removes the loop from within the while() loop by separating it out + * into 3 cases: where you are enlarging the line (<32768), where you are + * between 50% and 100% of the original line (<=65536), and where you are + * shrinking it by a lot. anyways, I went from 200 delivered / 100+ + * skipped to 200 delivered / 80 skipped for the enlarging case. I + * noticed when looking at the assembly that the compiler was able to + * unroll these while(width) loops, whereas before it was trying to + * unroll the while(dx>32768) loops. so the compiler is better able to + * deal with this code. + */ + + + if (step < 32768) { + while (width) { + *dest = p1 + (((p2-p1) * dx)>>15); + + dx += step; + if (dx > 32768) { + dx -= 32768; + p1 = p2; + p2 = *source++; + } + + dest ++; + width --; + } + } else if (step <= 65536) { + while (width) { + *dest = p1 + (((p2-p1) * dx)>>15); + + dx += step; + if (dx > 65536) { + dx -= 65536; + p1 = *source++; + p2 = *source++; + } else { + dx -= 32768; + p1 = p2; + p2 = *source++; + } + + dest ++; + width --; + } + } else { + while (width) { + int offs; + + *dest = p1 + (((p2-p1) * dx)>>15); + + dx += step; + offs=((dx-1)>>15); + dx-=offs<<15; + source+=offs-2; + p1=*source++; + p2=*source++; + dest ++; + width --; + } + } + xine_profiler_stop_count(prof_scale_line); + + + + +} + +/* + * Interpolates 16 output pixels from 15 source pixels using shifts. + * Useful for scaling a PAL mpeg2 dvd input source to 4:3 format on + * a monitor using square pixels. + * (720 x 576 ==> 768 x 576) + */ +static void scale_line_15_16 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 16) >= 0) { + p1 = source[0]; + dest[0] = p1; + p2 = source[1]; + dest[1] = (1*p1 + 7*p2) >> 3; + p1 = source[2]; + dest[2] = (1*p2 + 7*p1) >> 3; + p2 = source[3]; + dest[3] = (1*p1 + 3*p2) >> 2; + p1 = source[4]; + dest[4] = (1*p2 + 3*p1) >> 2; + p2 = source[5]; + dest[5] = (3*p1 + 5*p2) >> 3; + p1 = source[6]; + dest[6] = (3*p2 + 5*p1) >> 3; + p2 = source[7]; + dest[7] = (1*p1 + 1*p1) >> 1; + p1 = source[8]; + dest[8] = (1*p2 + 1*p1) >> 1; + p2 = source[9]; + dest[9] = (5*p1 + 3*p2) >> 3; + p1 = source[10]; + dest[10] = (5*p2 + 3*p1) >> 3; + p2 = source[11]; + dest[11] = (3*p1 + 1*p2) >> 2; + p1 = source[12]; + dest[12] = (3*p2 + 1*p1) >> 2; + p2 = source[13]; + dest[13] = (7*p1 + 1*p2) >> 3; + p1 = source[14]; + dest[14] = (7*p2 + 1*p1) >> 3; + dest[15] = p1; + source += 15; + dest += 16; + } + + if ((width += 16) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 7*source[1]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[1] + 7*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[2] + 3*source[3]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[3] + 3*source[4]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[4] + 5*source[5]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[5] + 5*source[6]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[6] + 1*source[7]) >> 1; + if (--width <= 0) goto done; + *dest++ = (1*source[7] + 1*source[8]) >> 1; + if (--width <= 0) goto done; + *dest++ = (5*source[8] + 3*source[9]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[9] + 3*source[10]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[10] + 1*source[11]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[11] + 1*source[12]) >> 2; + if (--width <= 0) goto done; + *dest++ = (7*source[12] + 1*source[13]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[13] + 1*source[14]) >> 3; + done: + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Interpolates 53 output pixels from 45 source pixels using shifts. + * Useful for scaling a NTSC mpeg2 dvd input source to 16:9 display + * resulution + * fullscreen resolution, or to 16:9 format on a monitor using square + * pixels. + * (720 x 480 ==> 848 x 480) + */ +static void scale_line_45_53 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 53) >= 0) { + p1 = source[0]; + p2 = source[1]; + dest[0] = p1; + dest[1] = (1*p1 + 7*p2) >> 3; + p1 = source[2]; + dest[2] = (1*p2 + 3*p1) >> 2; + p2 = source[3]; + dest[3] = (1*p1 + 1*p2) >> 1; + p1 = source[4]; + dest[4] = (5*p2 + 3*p1) >> 3; + p2 = source[5]; + dest[5] = (3*p1 + 1*p2) >> 2; + p1 = source[6]; + dest[6] = (7*p2 + 1*p1) >> 3; + dest[7] = p1; + p2 = source[7]; + dest[8] = (1*p1 + 3*p2) >> 2; + p1 = source[8]; + dest[9] = (3*p2 + 5*p1) >> 3; + p2 = source[9]; + dest[10] = (1*p1 + 1*p2) >> 1; + p1 = source[10]; + dest[11] = (5*p2 + 3*p1) >> 3; + p2 = source[11]; + dest[12] = (3*p1 + 1*p2) >> 2; + p1 = source[12]; + dest[13] = p2; + dest[14] = (1*p2 + 7*p1) >> 3; + p2 = source[13]; + dest[15] = (1*p1 + 3*p2) >> 2; + p1 = source[14]; + dest[16] = (3*p2 + 5*p1) >> 3; + p2 = source[15]; + dest[17] = (5*p1 + 3*p2) >> 3; + p1 = source[16]; + dest[18] = (3*p2 + 1*p1) >> 2; + p2 = source[17]; + dest[19] = (7*p1 + 1*p2) >> 3; + dest[20] = p2; + p1 = source[18]; + dest[21] = (1*p2 + 7*p1) >> 3; + p2 = source[19]; + dest[22] = (3*p1 + 5*p2) >> 3; + p1 = source[20]; + dest[23] = (1*p2 + 1*p1) >> 1; + p2 = source[21]; + dest[24] = (5*p1 + 3*p2) >> 3; + p1 = source[22]; + dest[25] = (3*p2 + 1*p1) >> 2; + p2 = source[23]; + dest[26] = (7*p1 + 1*p2) >> 3; + dest[27] = (1*p1 + 7*p2) >> 3; + p1 = source[24]; + dest[28] = (1*p2 + 3*p1) >> 2; + p2 = source[25]; + dest[29] = (3*p1 + 5*p2) >> 3; + p1 = source[26]; + dest[30] = (1*p2 + 1*p1) >> 1; + p2 = source[27]; + dest[31] = (5*p1 + 3*p2) >> 3; + p1 = source[28]; + dest[32] = (7*p2 + 1*p1) >> 3; + p2 = source[29]; + dest[33] = p1; + dest[34] = (1*p1 + 7*p2) >> 3; + p1 = source[30]; + dest[35] = (1*p2 + 3*p1) >> 2; + p2 = source[31]; + dest[36] = (3*p1 + 5*p2) >> 3; + p1 = source[32]; + dest[37] = (5*p2 + 3*p1) >> 3; + p2 = source[33]; + dest[38] = (3*p1 + 1*p2) >> 2; + p1 = source[34]; + dest[39] = (7*p2 + 1*p1) >> 3; + dest[40] = p1; + p2 = source[35]; + dest[41] = (1*p1 + 3*p2) >> 2; + p1 = source[36]; + dest[42] = (3*p2 + 5*p1) >> 3; + p2 = source[37]; + dest[43] = (1*p1 + 1*p2) >> 1; + p1 = source[38]; + dest[44] = (5*p2 + 3*p1) >> 3; + p2 = source[39]; + dest[45] = (3*p1 + 1*p2) >> 2; + p1 = source[40]; + dest[46] = p2; + dest[47] = (1*p2 + 7*p1) >> 3; + p2 = source[41]; + dest[48] = (1*p1 + 3*p2) >> 2; + p1 = source[42]; + dest[49] = (3*p2 + 5*p1) >> 3; + p2 = source[43]; + dest[50] = (1*p1 + 1*p2) >> 1; + p1 = source[44]; + dest[51] = (3*p2 + 1*p1) >> 2; + p2 = source[45]; + dest[52] = (7*p1 + 1*p2) >> 3; + source += 45; + dest += 53; + } + + if ((width += 53) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 7*source[1]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[1] + 3*source[2]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[2] + 1*source[3]) >> 1; + if (--width <= 0) goto done; + *dest++ = (5*source[3] + 3*source[4]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[4] + 1*source[5]) >> 2; + if (--width <= 0) goto done; + *dest++ = (7*source[5] + 1*source[6]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[6]; + if (--width <= 0) goto done; + *dest++ = (1*source[6] + 3*source[7]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[7] + 5*source[8]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[8] + 1*source[9]) >> 1; + if (--width <= 0) goto done; + *dest++ = (5*source[9] + 3*source[10]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[10] + 1*source[11]) >> 2; + if (--width <= 0) goto done; + *dest++ = source[11]; + if (--width <= 0) goto done; + *dest++ = (1*source[11] + 7*source[12]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[12] + 3*source[13]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[13] + 5*source[14]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[14] + 3*source[15]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[15] + 1*source[16]) >> 2; + if (--width <= 0) goto done; + *dest++ = (7*source[16] + 1*source[17]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[17]; + if (--width <= 0) goto done; + *dest++ = (1*source[17] + 7*source[18]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[18] + 5*source[19]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[19] + 1*source[20]) >> 1; + if (--width <= 0) goto done; + *dest++ = (5*source[20] + 3*source[21]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[21] + 1*source[22]) >> 2; + if (--width <= 0) goto done; + *dest++ = (7*source[22] + 1*source[23]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[22] + 7*source[23]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[23] + 3*source[24]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[24] + 5*source[25]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[25] + 1*source[26]) >> 1; + if (--width <= 0) goto done; + *dest++ = (5*source[26] + 3*source[27]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[27] + 1*source[28]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[28]; + if (--width <= 0) goto done; + *dest++ = (1*source[28] + 7*source[29]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[29] + 3*source[30]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[30] + 5*source[31]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[31] + 3*source[32]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[32] + 1*source[33]) >> 2; + if (--width <= 0) goto done; + *dest++ = (7*source[33] + 1*source[34]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[34]; + if (--width <= 0) goto done; + *dest++ = (1*source[34] + 3*source[35]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[35] + 5*source[36]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[36] + 1*source[37]) >> 1; + if (--width <= 0) goto done; + *dest++ = (5*source[37] + 3*source[38]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[38] + 1*source[39]) >> 2; + if (--width <= 0) goto done; + *dest++ = source[39]; + if (--width <= 0) goto done; + *dest++ = (1*source[39] + 7*source[40]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[40] + 3*source[41]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[41] + 5*source[42]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[42] + 1*source[43]) >> 1; + if (--width <= 0) goto done; + *dest++ = (3*source[43] + 1*source[44]) >> 2; + done: + + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Interpolates 64 output pixels from 45 source pixels using shifts. + * Useful for scaling a PAL mpeg2 dvd input source to 1024x768 + * fullscreen resolution, or to 16:9 format on a monitor using square + * pixels. + * (720 x 576 ==> 1024 x 576) + */ +static void scale_line_45_64 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 64) >= 0) { + p1 = source[0]; + p2 = source[1]; + dest[0] = p1; + dest[1] = (1*p1 + 3*p2) >> 2; + p1 = source[2]; + dest[2] = (5*p2 + 3*p1) >> 3; + p2 = source[3]; + dest[3] = (7*p1 + 1*p2) >> 3; + dest[4] = (1*p1 + 3*p2) >> 2; + p1 = source[4]; + dest[5] = (1*p2 + 1*p1) >> 1; + p2 = source[5]; + dest[6] = (3*p1 + 1*p2) >> 2; + dest[7] = (1*p1 + 7*p2) >> 3; + p1 = source[6]; + dest[8] = (3*p2 + 5*p1) >> 3; + p2 = source[7]; + dest[9] = (5*p1 + 3*p2) >> 3; + p1 = source[8]; + dest[10] = p2; + dest[11] = (1*p2 + 3*p1) >> 2; + p2 = source[9]; + dest[12] = (5*p1 + 3*p2) >> 3; + p1 = source[10]; + dest[13] = (7*p2 + 1*p1) >> 3; + dest[14] = (1*p2 + 7*p1) >> 3; + p2 = source[11]; + dest[15] = (1*p1 + 1*p2) >> 1; + p1 = source[12]; + dest[16] = (3*p2 + 1*p1) >> 2; + dest[17] = p1; + p2 = source[13]; + dest[18] = (3*p1 + 5*p2) >> 3; + p1 = source[14]; + dest[19] = (5*p2 + 3*p1) >> 3; + p2 = source[15]; + dest[20] = p1; + dest[21] = (1*p1 + 3*p2) >> 2; + p1 = source[16]; + dest[22] = (1*p2 + 1*p1) >> 1; + p2 = source[17]; + dest[23] = (7*p1 + 1*p2) >> 3; + dest[24] = (1*p1 + 7*p2) >> 3; + p1 = source[18]; + dest[25] = (3*p2 + 5*p1) >> 3; + p2 = source[19]; + dest[26] = (3*p1 + 1*p2) >> 2; + dest[27] = p2; + p1 = source[20]; + dest[28] = (3*p2 + 5*p1) >> 3; + p2 = source[21]; + dest[29] = (5*p1 + 3*p2) >> 3; + p1 = source[22]; + dest[30] = (7*p2 + 1*p1) >> 3; + dest[31] = (1*p2 + 3*p1) >> 2; + p2 = source[23]; + dest[32] = (1*p1 + 1*p2) >> 1; + p1 = source[24]; + dest[33] = (3*p2 + 1*p1) >> 2; + dest[34] = (1*p2 + 7*p1) >> 3; + p2 = source[25]; + dest[35] = (3*p1 + 5*p2) >> 3; + p1 = source[26]; + dest[36] = (3*p2 + 1*p1) >> 2; + p2 = source[27]; + dest[37] = p1; + dest[38] = (1*p1 + 3*p2) >> 2; + p1 = source[28]; + dest[39] = (5*p2 + 3*p1) >> 3; + p2 = source[29]; + dest[40] = (7*p1 + 1*p2) >> 3; + dest[41] = (1*p1 + 7*p2) >> 3; + p1 = source[30]; + dest[42] = (1*p2 + 1*p1) >> 1; + p2 = source[31]; + dest[43] = (3*p1 + 1*p2) >> 2; + dest[44] = (1*p1 + 7*p2) >> 3; + p1 = source[32]; + dest[45] = (3*p2 + 5*p1) >> 3; + p2 = source[33]; + dest[46] = (5*p1 + 3*p2) >> 3; + p1 = source[34]; + dest[47] = p2; + dest[48] = (1*p2 + 3*p1) >> 2; + p2 = source[35]; + dest[49] = (1*p1 + 1*p2) >> 1; + p1 = source[36]; + dest[50] = (7*p2 + 1*p1) >> 3; + dest[51] = (1*p2 + 7*p1) >> 3; + p2 = source[37]; + dest[52] = (1*p1 + 1*p2) >> 1; + p1 = source[38]; + dest[53] = (3*p2 + 1*p1) >> 2; + dest[54] = p1; + p2 = source[39]; + dest[55] = (3*p1 + 5*p2) >> 3; + p1 = source[40]; + dest[56] = (5*p2 + 3*p1) >> 3; + p2 = source[41]; + dest[57] = (7*p1 + 1*p2) >> 3; + dest[58] = (1*p1 + 3*p2) >> 2; + p1 = source[42]; + dest[59] = (1*p2 + 1*p1) >> 1; + p2 = source[43]; + dest[60] = (7*p1 + 1*p2) >> 3; + dest[61] = (1*p1 + 7*p2) >> 3; + p1 = source[44]; + dest[62] = (3*p2 + 5*p1) >> 3; + p2 = source[45]; + dest[63] = (3*p1 + 1*p2) >> 2; + source += 45; + dest += 64; + } + + if ((width += 64) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 3*source[1]) >> 2; + if (--width <= 0) goto done; + *dest++ = (5*source[1] + 3*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[2] + 1*source[3]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[2] + 3*source[3]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[3] + 1*source[4]) >> 1; + if (--width <= 0) goto done; + *dest++ = (3*source[4] + 1*source[5]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[4] + 7*source[5]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[5] + 5*source[6]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[6] + 3*source[7]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[7]; + if (--width <= 0) goto done; + *dest++ = (1*source[7] + 3*source[8]) >> 2; + if (--width <= 0) goto done; + *dest++ = (5*source[8] + 3*source[9]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[9] + 1*source[10]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[9] + 7*source[10]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[10] + 1*source[11]) >> 1; + if (--width <= 0) goto done; + *dest++ = (3*source[11] + 1*source[12]) >> 2; + if (--width <= 0) goto done; + *dest++ = source[12]; + if (--width <= 0) goto done; + *dest++ = (3*source[12] + 5*source[13]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[13] + 3*source[14]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[14]; + if (--width <= 0) goto done; + *dest++ = (1*source[14] + 3*source[15]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[15] + 1*source[16]) >> 1; + if (--width <= 0) goto done; + *dest++ = (7*source[16] + 1*source[17]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[16] + 7*source[17]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[17] + 5*source[18]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[18] + 1*source[19]) >> 2; + if (--width <= 0) goto done; + *dest++ = source[19]; + if (--width <= 0) goto done; + *dest++ = (3*source[19] + 5*source[20]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[20] + 3*source[21]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[21] + 1*source[22]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[21] + 3*source[22]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[22] + 1*source[23]) >> 1; + if (--width <= 0) goto done; + *dest++ = (3*source[23] + 1*source[24]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[23] + 7*source[24]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[24] + 5*source[25]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[25] + 1*source[26]) >> 2; + if (--width <= 0) goto done; + *dest++ = source[26]; + if (--width <= 0) goto done; + *dest++ = (1*source[26] + 3*source[27]) >> 2; + if (--width <= 0) goto done; + *dest++ = (5*source[27] + 3*source[28]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[28] + 1*source[29]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[28] + 7*source[29]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[29] + 1*source[30]) >> 1; + if (--width <= 0) goto done; + *dest++ = (3*source[30] + 1*source[31]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[30] + 7*source[31]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[31] + 5*source[32]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[32] + 3*source[33]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[33]; + if (--width <= 0) goto done; + *dest++ = (1*source[33] + 3*source[34]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[34] + 1*source[35]) >> 1; + if (--width <= 0) goto done; + *dest++ = (7*source[35] + 1*source[36]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[35] + 7*source[36]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[36] + 1*source[37]) >> 1; + if (--width <= 0) goto done; + *dest++ = (3*source[37] + 1*source[38]) >> 2; + if (--width <= 0) goto done; + *dest++ = source[38]; + if (--width <= 0) goto done; + *dest++ = (3*source[38] + 5*source[39]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[39] + 3*source[40]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[40] + 1*source[41]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[40] + 3*source[41]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[41] + 1*source[42]) >> 1; + if (--width <= 0) goto done; + *dest++ = (7*source[42] + 1*source[43]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[42] + 7*source[43]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[43] + 5*source[44]) >> 3; + done: + + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Interpolates 16 output pixels from 9 source pixels using shifts. + * Useful for scaling a PAL mpeg2 dvd input source to 1280x1024 fullscreen + * (720 x 576 ==> 1280 x XXX) + */ +static void scale_line_9_16 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 16) >= 0) { + p1 = source[0]; + p2 = source[1]; + dest[0] = p1; + dest[1] = (1*p1 + 1*p2) >> 1; + p1 = source[2]; + dest[2] = (7*p2 + 1*p1) >> 3; + dest[3] = (3*p2 + 5*p1) >> 3; + p2 = source[3]; + dest[4] = (3*p1 + 1*p2) >> 2; + dest[5] = (1*p1 + 3*p2) >> 2; + p1 = source[4]; + dest[6] = (5*p2 + 3*p1) >> 3; + dest[7] = (1*p2 + 7*p1) >> 3; + p2 = source[5]; + dest[8] = (1*p1 + 1*p2) >> 1; + p1 = source[6]; + dest[9] = p2; + dest[10] = (3*p2 + 5*p1) >> 3; + p2 = source[7]; + dest[11] = (7*p1 + 1*p2) >> 3; + dest[12] = (1*p1 + 3*p2) >> 2; + p1 = source[8]; + dest[13] = (3*p2 + 1*p1) >> 2; + dest[14] = (1*p2 + 7*p1) >> 3; + p2 = source[9]; + dest[15] = (5*p1 + 3*p2) >> 3; + source += 9; + dest += 16; + } + + if ((width += 16) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 1*source[1]) >> 1; + if (--width <= 0) goto done; + *dest++ = (7*source[1] + 1*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[1] + 5*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[2] + 1*source[3]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[2] + 3*source[3]) >> 2; + if (--width <= 0) goto done; + *dest++ = (5*source[3] + 3*source[4]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[3] + 7*source[4]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[4] + 1*source[5]) >> 1; + if (--width <= 0) goto done; + *dest++ = source[5]; + if (--width <= 0) goto done; + *dest++ = (3*source[5] + 5*source[6]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[6] + 1*source[7]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[6] + 3*source[7]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[7] + 1*source[8]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[7] + 7*source[8]) >> 3; +done: + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Interpolates 12 output pixels from 11 source pixels using shifts. + * Useful for scaling a PAL vcd input source to 4:3 display format. + */ +static void scale_line_11_12 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 12) >= 0) { + p1 = source[0]; + p2 = source[1]; + dest[0] = p1; + dest[1] = (1*p1 + 7*p2) >> 3; + p1 = source[2]; + dest[2] = (1*p2 + 7*p1) >> 3; + p2 = source[3]; + dest[3] = (1*p1 + 3*p2) >> 2; + p1 = source[4]; + dest[4] = (3*p2 + 5*p1) >> 3; + p2 = source[5]; + dest[5] = (3*p1 + 5*p2) >> 3; + p1 = source[6]; + dest[6] = (1*p2 + 1*p1) >> 1; + p2 = source[7]; + dest[7] = (5*p1 + 3*p2) >> 3; + p1 = source[8]; + dest[8] = (5*p2 + 3*p1) >> 3; + p2 = source[9]; + dest[9] = (3*p1 + 1*p2) >> 2; + p1 = source[10]; + dest[10] = (7*p2 + 1*p1) >> 3; + p2 = source[11]; + dest[11] = (7*p1 + 1*p2) >> 3; + source += 11; + dest += 12; + } + + if ((width += 12) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 7*source[1]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[1] + 7*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[2] + 3*source[3]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[3] + 5*source[4]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[4] + 5*source[5]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[5] + 1*source[6]) >> 1; + if (--width <= 0) goto done; + *dest++ = (5*source[6] + 3*source[7]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[7] + 3*source[8]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[8] + 1*source[9]) >> 2; + if (--width <= 0) goto done; + *dest++ = (7*source[9] + 1*source[10]) >> 3; +done: + + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Interpolates 24 output pixels from 11 source pixels using shifts. + * Useful for scaling a PAL vcd input source to 4:3 display format + * at 2*zoom. + */ +static void scale_line_11_24 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 24) >= 0) { + p1 = source[0]; + p2 = source[1]; + dest[0] = p1; + dest[1] = (1*p1 + 1*p2) >> 1; + dest[2] = (1*p1 + 7*p2) >> 3; + p1 = source[2]; + dest[3] = (5*p2 + 3*p1) >> 3; + dest[4] = (1*p2 + 7*p1) >> 3; + p2 = source[3]; + dest[5] = (3*p1 + 1*p2) >> 2; + dest[6] = (1*p1 + 3*p2) >> 2; + p1 = source[4]; + dest[7] = (3*p2 + 1*p1) >> 2; + dest[8] = (3*p2 + 5*p1) >> 3; + p2 = source[5]; + dest[9] = (7*p1 + 1*p2) >> 3; + dest[10] = (3*p1 + 5*p2) >> 3; + p1 = source[6]; + dest[11] = p2; + dest[12] = (1*p2 + 1*p1) >> 1; + dest[13] = p1; + p2 = source[7]; + dest[14] = (5*p1 + 3*p2) >> 3; + dest[15] = (1*p1 + 7*p2) >> 3; + p1 = source[8]; + dest[16] = (5*p2 + 3*p1) >> 3; + dest[17] = (1*p2 + 3*p1) >> 2; + p2 = source[9]; + dest[18] = (3*p1 + 1*p2) >> 2; + dest[19] = (1*p1 + 3*p2) >> 2; + p1 = source[10]; + dest[20] = (7*p2 + 1*p1) >> 3; + dest[21] = (3*p2 + 5*p1) >> 3; + p2 = source[11]; + dest[22] = (7*p1 + 1*p2) >> 3; + dest[23] = (1*p1 + 1*p2) >> 1; + source += 11; + dest += 24; + } + + if ((width += 24) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 1*source[1]) >> 1; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 7*source[1]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[1] + 3*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[1] + 7*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[2] + 1*source[3]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[2] + 3*source[3]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[3] + 1*source[4]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[3] + 5*source[4]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[4] + 1*source[5]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[4] + 5*source[5]) >> 3; + if (--width <= 0) goto done; + *dest++ = source[5]; + if (--width <= 0) goto done; + *dest++ = (1*source[5] + 1*source[6]) >> 1; + if (--width <= 0) goto done; + *dest++ = source[6]; + if (--width <= 0) goto done; + *dest++ = (5*source[6] + 3*source[7]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[6] + 7*source[7]) >> 3; + if (--width <= 0) goto done; + *dest++ = (5*source[7] + 3*source[8]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[7] + 3*source[8]) >> 2; + if (--width <= 0) goto done; + *dest++ = (3*source[8] + 1*source[9]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[8] + 3*source[9]) >> 2; + if (--width <= 0) goto done; + *dest++ = (7*source[9] + 1*source[10]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[9] + 5*source[10]) >> 3; + if (--width <= 0) goto done; + *dest++ = (7*source[10] + 1*source[11]) >> 3; +done: + + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Interpolates 8 output pixels from 5 source pixels using shifts. + * Useful for scaling a PAL svcd input source to 4:3 display format. + */ +static void scale_line_5_8 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 8) >= 0) { + p1 = source[0]; + p2 = source[1]; + dest[0] = p1; + dest[1] = (3*p1 + 5*p2) >> 3; + p1 = source[2]; + dest[2] = (3*p2 + 1*p1) >> 2; + dest[3] = (1*p2 + 7*p1) >> 3; + p2 = source[3]; + dest[4] = (1*p1 + 1*p2) >> 1; + p1 = source[4]; + dest[5] = (7*p2 + 1*p1) >> 3; + dest[6] = (1*p2 + 3*p1) >> 2; + p2 = source[5]; + dest[7] = (5*p1 + 3*p2) >> 3; + source += 5; + dest += 8; + } + + if ((width += 8) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (3*source[0] + 5*source[1]) >> 3; + if (--width <= 0) goto done; + *dest++ = (3*source[1] + 1*source[2]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[1] + 7*source[2]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[2] + 1*source[3]) >> 1; + if (--width <= 0) goto done; + *dest++ = (7*source[3] + 1*source[4]) >> 3; + if (--width <= 0) goto done; + *dest++ = (1*source[3] + 3*source[4]) >> 2; +done: + + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Interpolates 4 output pixels from 3 source pixels using shifts. + * Useful for scaling a NTSC svcd input source to 4:3 display format. + */ +static void scale_line_3_4 (uint8_t *source, uint8_t *dest, + int width, int step) { + + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + while ((width -= 4) >= 0) { + p1 = source[0]; + p2 = source[1]; + dest[0] = p1; + dest[1] = (1*p1 + 3*p2) >> 2; + p1 = source[2]; + dest[2] = (1*p2 + 1*p1) >> 1; + p2 = source[3]; + dest[3] = (3*p1 + 1*p2) >> 2; + source += 3; + dest += 4; + } + + if ((width += 4) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (1*source[0] + 3*source[1]) >> 2; + if (--width <= 0) goto done; + *dest++ = (1*source[1] + 1*source[2]) >> 1; +done: + + xine_profiler_stop_count(prof_scale_line); +} + + +/* Interpolate 2 output pixels from one source pixel. */ + +static void scale_line_1_2 (uint8_t *source, uint8_t *dest, + int width, int step) { + int p1, p2; + + xine_profiler_start_count(prof_scale_line); + + p1 = *source; + while ((width -= 4) >= 0) { + *dest++ = p1; + p2 = *++source; + *dest++ = (p1 + p2) >> 1; + *dest++ = p2; + p1 = *++source; + *dest++ = (p2 + p1) >> 1; + } + + if ((width += 4) <= 0) goto done; + *dest++ = source[0]; + if (--width <= 0) goto done; + *dest++ = (source[0] + source[1]) >> 1; + if (--width <= 0) goto done; + *dest++ = source[1]; + done: + + xine_profiler_stop_count(prof_scale_line); +} + + +/* + * Scale line with no horizontal scaling. For NTSC mpeg2 dvd input in + * 4:3 output format (720x480 -> 720x540) + */ +static void scale_line_1_1 (uint8_t *source, uint8_t *dest, + int width, int step) { + + xine_profiler_start_count(prof_scale_line); + xine_fast_memcpy(dest, source, width); + xine_profiler_stop_count(prof_scale_line); +} + + +static scale_line_func_t find_scale_line_func(int step) { + static struct { + int src_step; + int dest_step; + scale_line_func_t func; + char *desc; + } scale_line[] = { + { 15, 16, scale_line_15_16, "dvd 4:3(pal)" }, + { 45, 64, scale_line_45_64, "dvd 16:9(pal), fullscreen(1024x768)" }, + { 9, 16, scale_line_9_16, "dvd fullscreen(1280x1024)" }, + { 45, 53, scale_line_45_53, "dvd 16:9(ntsc)" }, + { 11, 12, scale_line_11_12, "vcd 4:3(pal)" }, + { 11, 24, scale_line_11_24, "vcd 4:3(pal) 2*zoom" }, + { 5, 8, scale_line_5_8, "svcd 4:3(pal)" }, + { 3, 4, scale_line_3_4, "svcd 4:3(ntsc)" }, + { 1, 2, scale_line_1_2, "2*zoom" }, + { 1, 1, scale_line_1_1, "non-scaled" }, + }; + int i; + + for (i = 0; i < sizeof(scale_line)/sizeof(scale_line[0]); i++) { + if (step == scale_line[i].src_step*32768/scale_line[i].dest_step) { + printf("yuv2rgb: using %s optimized scale_line\n", scale_line[i].desc); + return scale_line[i].func; + } + } + printf("yuv2rgb: using generic scale_line with interpolation\n"); + return scale_line_gen; + +} + + +static void scale_line_2 (uint8_t *source, uint8_t *dest, + int width, int step) { + int p1; + int p2; + int dx; + + p1 = *source; source+=2; + p2 = *source; source+=2; + dx = 0; + + while (width) { + + *dest = (p1 * (32768 - dx) + p2 * dx) / 32768; + + dx += step; + while (dx > 32768) { + dx -= 32768; + p1 = p2; + p2 = *source; + source+=2; + } + + dest ++; + width --; + } +} + +static void scale_line_4 (uint8_t *source, uint8_t *dest, + int width, int step) { + int p1; + int p2; + int dx; + + p1 = *source; source+=4; + p2 = *source; source+=4; + dx = 0; + + while (width) { + + *dest = (p1 * (32768 - dx) + p2 * dx) / 32768; + + dx += step; + while (dx > 32768) { + dx -= 32768; + p1 = p2; + p2 = *source; + source+=4; + } + + dest ++; + width --; + } +} + + +#define RGB(i) \ + U = pu[i]; \ + V = pv[i]; \ + r = this->table_rV[V]; \ + g = (void *) (((uint8_t *)this->table_gU[U]) + this->table_gV[V]); \ + b = this->table_bU[U]; + +#define DST1(i) \ + Y = py_1[2*i]; \ + dst_1[2*i] = r[Y] + g[Y] + b[Y]; \ + Y = py_1[2*i+1]; \ + dst_1[2*i+1] = r[Y] + g[Y] + b[Y]; + +#define DST2(i) \ + Y = py_2[2*i]; \ + dst_2[2*i] = r[Y] + g[Y] + b[Y]; \ + Y = py_2[2*i+1]; \ + dst_2[2*i+1] = r[Y] + g[Y] + b[Y]; + +#define DST1RGB(i) \ + Y = py_1[2*i]; \ + dst_1[6*i] = r[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = b[Y]; \ + Y = py_1[2*i+1]; \ + dst_1[6*i+3] = r[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = b[Y]; + +#define DST2RGB(i) \ + Y = py_2[2*i]; \ + dst_2[6*i] = r[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = b[Y]; \ + Y = py_2[2*i+1]; \ + dst_2[6*i+3] = r[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = b[Y]; + +#define DST1BGR(i) \ + Y = py_1[2*i]; \ + dst_1[6*i] = b[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = r[Y]; \ + Y = py_1[2*i+1]; \ + dst_1[6*i+3] = b[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = r[Y]; + +#define DST2BGR(i) \ + Y = py_2[2*i]; \ + dst_2[6*i] = b[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = r[Y]; \ + Y = py_2[2*i+1]; \ + dst_2[6*i+3] = b[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = r[Y]; + +#define DST1CMAP(i) \ + Y = py_1[2*i]; \ + dst_1[2*i] = this->cmap[r[Y] + g[Y] + b[Y]]; \ + Y = py_1[2*i+1]; \ + dst_1[2*i+1] = this->cmap[r[Y] + g[Y] + b[Y]]; + +#define DST2CMAP(i) \ + Y = py_2[2*i]; \ + dst_2[2*i] = this->cmap[r[Y] + g[Y] + b[Y]]; \ + Y = py_2[2*i+1]; \ + dst_2[2*i+1] = this->cmap[r[Y] + g[Y] + b[Y]]; + +static void yuv2rgb_c_32 (yuv2rgb_t *this, uint8_t * _dst, + uint8_t * _py, uint8_t * _pu, uint8_t * _pv) +{ + int U, V, Y; + uint8_t * py_1, * py_2, * pu, * pv; + uint32_t * r, * g, * b; + uint32_t * dst_1, * dst_2; + int width, height, dst_height; + int dy; + + if (this->do_scale) { + scale_line_func_t scale_line = this->scale_line; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + dst_height = this->dest_height; + + for (height = 0;; ) { + dst_1 = (uint32_t*)_dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1(0); + + RGB(1); + DST1(1); + + RGB(2); + DST1(2); + + RGB(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width*4); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + _py += this->y_stride; + + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + _pu += this->uv_stride; + _pv += this->uv_stride; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while( dy>=32768); + } + } else { + height = this->source_height >> 1; + do { + dst_1 = (uint32_t*)_dst; + dst_2 = (void*)( (uint8_t *)_dst + this->rgb_stride ); + py_1 = _py; + py_2 = _py + this->y_stride; + pu = _pu; + pv = _pv; + + width = this->source_width >> 3; + do { + RGB(0); + DST1(0); + DST2(0); + + RGB(1); + DST2(1); + DST1(1); + + RGB(2); + DST1(2); + DST2(2); + + RGB(3); + DST2(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + py_2 += 8; + dst_1 += 8; + dst_2 += 8; + } while (--width); + + _dst += 2 * this->rgb_stride; + _py += 2 * this->y_stride; + _pu += this->uv_stride; + _pv += this->uv_stride; + + } while (--height); + } +} + +/* This is very near from the yuv2rgb_c_32 code */ +static void yuv2rgb_c_24_rgb (yuv2rgb_t *this, uint8_t * _dst, + uint8_t * _py, uint8_t * _pu, uint8_t * _pv) +{ + int U, V, Y; + uint8_t * py_1, * py_2, * pu, * pv; + uint8_t * r, * g, * b; + uint8_t * dst_1, * dst_2; + int width, height, dst_height; + int dy; + + if (this->do_scale) { + + scale_line_func_t scale_line = this->scale_line; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + dst_height = this->dest_height; + + for (height = 0;; ) { + dst_1 = _dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1RGB(0); + + RGB(1); + DST1RGB(1); + + RGB(2); + DST1RGB(2); + + RGB(3); + DST1RGB(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 24; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, _dst-this->rgb_stride, this->dest_width*3); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + _py += this->y_stride; + + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + _pu += this->uv_stride; + _pv += this->uv_stride; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while (dy>=32768); + } + } else { + height = this->source_height >> 1; + do { + dst_1 = _dst; + dst_2 = (void*)( (uint8_t *)_dst + this->rgb_stride ); + py_1 = _py; + py_2 = _py + this->y_stride; + pu = _pu; + pv = _pv; + + width = this->source_width >> 3; + do { + RGB(0); + DST1RGB(0); + DST2RGB(0); + + RGB(1); + DST2RGB(1); + DST1RGB(1); + + RGB(2); + DST1RGB(2); + DST2RGB(2); + + RGB(3); + DST2RGB(3); + DST1RGB(3); + + pu += 4; + pv += 4; + py_1 += 8; + py_2 += 8; + dst_1 += 24; + dst_2 += 24; + } while (--width); + + _dst += 2 * this->rgb_stride; + _py += 2 * this->y_stride; + _pu += this->uv_stride; + _pv += this->uv_stride; + + } while (--height); + } +} + +/* only trivial mods from yuv2rgb_c_24_rgb */ +static void yuv2rgb_c_24_bgr (yuv2rgb_t *this, uint8_t * _dst, + uint8_t * _py, uint8_t * _pu, uint8_t * _pv) +{ + int U, V, Y; + uint8_t * py_1, * py_2, * pu, * pv; + uint8_t * r, * g, * b; + uint8_t * dst_1, * dst_2; + int width, height, dst_height; + int dy; + + if (this->do_scale) { + + scale_line_func_t scale_line = this->scale_line; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + dst_height = this->dest_height; + + for (height = 0;; ) { + dst_1 = _dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1BGR(0); + + RGB(1); + DST1BGR(1); + + RGB(2); + DST1BGR(2); + + RGB(3); + DST1BGR(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 24; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, _dst-this->rgb_stride, this->dest_width*3); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + _py += this->y_stride; + + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + _pu += this->uv_stride; + _pv += this->uv_stride; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while( dy>=32768 ); + } + + } else { + height = this->source_height >> 1; + do { + dst_1 = _dst; + dst_2 = (void*)( (uint8_t *)_dst + this->rgb_stride ); + py_1 = _py; + py_2 = _py + this->y_stride; + pu = _pu; + pv = _pv; + width = this->source_width >> 3; + do { + RGB(0); + DST1BGR(0); + DST2BGR(0); + + RGB(1); + DST2BGR(1); + DST1BGR(1); + + RGB(2); + DST1BGR(2); + DST2BGR(2); + + RGB(3); + DST2BGR(3); + DST1BGR(3); + + pu += 4; + pv += 4; + py_1 += 8; + py_2 += 8; + dst_1 += 24; + dst_2 += 24; + } while (--width); + + _dst += 2 * this->rgb_stride; + _py += 2 * this->y_stride; + _pu += this->uv_stride; + _pv += this->uv_stride; + + } while (--height); + } +} + +/* This is exactly the same code as yuv2rgb_c_32 except for the types of */ +/* r, g, b, dst_1, dst_2 */ +static void yuv2rgb_c_16 (yuv2rgb_t *this, uint8_t * _dst, + uint8_t * _py, uint8_t * _pu, uint8_t * _pv) +{ + int U, V, Y; + uint8_t * py_1, * py_2, * pu, * pv; + uint16_t * r, * g, * b; + uint16_t * dst_1, * dst_2; + int width, height, dst_height; + int dy; + + if (this->do_scale) { + scale_line_func_t scale_line = this->scale_line; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + dst_height = this->dest_height; + + for (height = 0;; ) { + dst_1 = (uint16_t*)_dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1(0); + + RGB(1); + DST1(1); + + RGB(2); + DST1(2); + + RGB(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width*2); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + _py += this->y_stride; + + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + _pu += this->uv_stride; + _pv += this->uv_stride; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while( dy>=32768); + } + } else { + height = this->source_height >> 1; + do { + dst_1 = (uint16_t*)_dst; + dst_2 = (void*)( (uint8_t *)_dst + this->rgb_stride ); + py_1 = _py; + py_2 = _py + this->y_stride; + pu = _pu; + pv = _pv; + width = this->source_width >> 3; + do { + RGB(0); + DST1(0); + DST2(0); + + RGB(1); + DST2(1); + DST1(1); + + RGB(2); + DST1(2); + DST2(2); + + RGB(3); + DST2(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + py_2 += 8; + dst_1 += 8; + dst_2 += 8; + } while (--width); + + _dst += 2 * this->rgb_stride; + _py += 2 * this->y_stride; + _pu += this->uv_stride; + _pv += this->uv_stride; + + } while (--height); + } +} + +/* This is exactly the same code as yuv2rgb_c_32 except for the types of */ +/* r, g, b, dst_1, dst_2 */ +static void yuv2rgb_c_8 (yuv2rgb_t *this, uint8_t * _dst, + uint8_t * _py, uint8_t * _pu, uint8_t * _pv) +{ + int U, V, Y; + uint8_t * py_1, * py_2, * pu, * pv; + uint8_t * r, * g, * b; + uint8_t * dst_1, * dst_2; + int width, height, dst_height; + int dy; + + if (this->do_scale) { + scale_line_func_t scale_line = this->scale_line; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + dst_height = this->dest_height; + + for (height = 0;; ) { + dst_1 = (uint8_t*)_dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1(0); + + RGB(1); + DST1(1); + + RGB(2); + DST1(2); + + RGB(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + _py += this->y_stride; + + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + _pu += this->uv_stride; + _pv += this->uv_stride; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while( dy>=32768 ); + } + } else { + height = this->source_height >> 1; + do { + dst_1 = (uint8_t*)_dst; + dst_2 = (void*)( (uint8_t *)_dst + this->rgb_stride ); + py_1 = _py; + py_2 = _py + this->y_stride; + pu = _pu; + pv = _pv; + + width = this->source_width >> 3; + do { + RGB(0); + DST1(0); + DST2(0); + + RGB(1); + DST2(1); + DST1(1); + + RGB(2); + DST1(2); + DST2(2); + + RGB(3); + DST2(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + py_2 += 8; + dst_1 += 8; + dst_2 += 8; + } while (--width); + + _dst += 2 * this->rgb_stride; + _py += 2 * this->y_stride; + _pu += this->uv_stride; + _pv += this->uv_stride; + + } while (--height); + } +} + +/* now for something different: 256 grayscale mode */ +static void yuv2rgb_c_gray (yuv2rgb_t *this, uint8_t * _dst, + uint8_t * _py, uint8_t * _pu, uint8_t * _pv) +{ + int height, dst_height; + int dy; + + if (this->do_scale) { + scale_line_func_t scale_line = this->scale_line; + + dy = 0; + dst_height = this->dest_height; + + for (;;) { + scale_line (_py, _dst, this->dest_width, this->step_dx); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + _py += this->y_stride*(dy>>15); + dy &= 32767; + /* dy -= 32768; + _py += this->y_stride; + */ + } + } else { + for (height = this->source_height; --height >= 0; ) { + xine_fast_memcpy(_dst, _py, this->dest_width); + _dst += this->rgb_stride; + _py += this->y_stride; + } + } +} + +/* now for something different: 256 color mode */ +static void yuv2rgb_c_palette (yuv2rgb_t *this, uint8_t * _dst, + uint8_t * _py, uint8_t * _pu, uint8_t * _pv) +{ + int U, V, Y; + uint8_t * py_1, * py_2, * pu, * pv; + uint16_t * r, * g, * b; + uint8_t * dst_1, * dst_2; + int width, height, dst_height; + int dy; + + if (this->do_scale) { + scale_line_func_t scale_line = this->scale_line; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + dst_height = this->dest_height; + + for (height = 0;; ) { + dst_1 = _dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1CMAP(0); + + RGB(1); + DST1CMAP(1); + + RGB(2); + DST1CMAP(2); + + RGB(3); + DST1CMAP(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + _py += this->y_stride; + + scale_line (_py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + _pu += this->uv_stride; + _pv += this->uv_stride; + + scale_line (_pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (_pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while( dy>=32768 ); + } + } else { + height = this->source_height >> 1; + do { + dst_1 = _dst; + dst_2 = _dst + this->rgb_stride; + py_1 = _py; + py_2 = _py + this->y_stride; + pu = _pu; + pv = _pv; + width = this->source_width >> 3; + do { + RGB(0); + DST1CMAP(0); + DST2CMAP(0); + + RGB(1); + DST2CMAP(1); + DST1CMAP(1); + + RGB(2); + DST1CMAP(2); + DST2CMAP(2); + + RGB(3); + DST2CMAP(3); + DST1CMAP(3); + + pu += 4; + pv += 4; + py_1 += 8; + py_2 += 8; + dst_1 += 8; + dst_2 += 8; + } while (--width); + + _dst += 2 * this->rgb_stride; + _py += 2 * this->y_stride; + _pu += this->uv_stride; + _pv += this->uv_stride; + + } while (--height); + } +} + +static int div_round (int dividend, int divisor) +{ + if (dividend > 0) + return (dividend + (divisor>>1)) / divisor; + else + return -((-dividend + (divisor>>1)) / divisor); +} + +static void yuv2rgb_setup_tables (yuv2rgb_factory_t *this, int mode, int swapped) +{ + int i; + uint8_t table_Y[1024]; + uint32_t * table_32 = 0; + uint16_t * table_16 = 0; + uint8_t * table_8 = 0; + int entry_size = 0; + void *table_r = 0, *table_g = 0, *table_b = 0; + int shift_r = 0, shift_g = 0, shift_b = 0; + + int crv = Inverse_Table_6_9[this->matrix_coefficients][0]; + int cbu = Inverse_Table_6_9[this->matrix_coefficients][1]; + int cgu = -Inverse_Table_6_9[this->matrix_coefficients][2]; + int cgv = -Inverse_Table_6_9[this->matrix_coefficients][3]; + + for (i = 0; i < 1024; i++) { + int j; + + j = (76309 * (i - 384 - 16) + 32768) >> 16; + j = (j < 0) ? 0 : ((j > 255) ? 255 : j); + table_Y[i] = j; + } + + switch (mode) { + case MODE_32_RGB: + case MODE_32_BGR: + table_32 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint32_t)); + + entry_size = sizeof (uint32_t); + table_r = table_32 + 197; + table_b = table_32 + 197 + 685; + table_g = table_32 + 197 + 2*682; + + if (swapped) { + switch (mode) { + case MODE_32_RGB: shift_r = 8; shift_g = 16; shift_b = 24; break; + case MODE_32_BGR: shift_r = 24; shift_g = 16; shift_b = 8; break; + } + } else { + switch (mode) { + case MODE_32_RGB: shift_r = 16; shift_g = 8; shift_b = 0; break; + case MODE_32_BGR: shift_r = 0; shift_g = 8; shift_b = 16; break; + } + } + + for (i = -197; i < 256+197; i++) + ((uint32_t *) table_r)[i] = table_Y[i+384] << shift_r; + for (i = -132; i < 256+132; i++) + ((uint32_t *) table_g)[i] = table_Y[i+384] << shift_g; + for (i = -232; i < 256+232; i++) + ((uint32_t *) table_b)[i] = table_Y[i+384] << shift_b; + break; + + case MODE_24_RGB: + case MODE_24_BGR: + table_8 = malloc ((256 + 2*232) * sizeof (uint8_t)); + + entry_size = sizeof (uint8_t); + table_r = table_g = table_b = table_8 + 232; + + for (i = -232; i < 256+232; i++) + ((uint8_t * )table_b)[i] = table_Y[i+384]; + break; + + case MODE_15_BGR: + case MODE_16_BGR: + case MODE_15_RGB: + case MODE_16_RGB: + table_16 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint16_t)); + + entry_size = sizeof (uint16_t); + table_r = table_16 + 197; + table_b = table_16 + 197 + 685; + table_g = table_16 + 197 + 2*682; + + if (swapped) { + switch (mode) { + case MODE_15_BGR: shift_r = 8; shift_g = 5; shift_b = 2; break; + case MODE_16_BGR: shift_r = 8; shift_g = 5; shift_b = 3; break; + case MODE_15_RGB: shift_r = 2; shift_g = 5; shift_b = 8; break; + case MODE_16_RGB: shift_r = 3; shift_g = 5; shift_b = 8; break; + } + } else { + switch (mode) { + case MODE_15_BGR: shift_r = 0; shift_g = 5; shift_b = 10; break; + case MODE_16_BGR: shift_r = 0; shift_g = 5; shift_b = 11; break; + case MODE_15_RGB: shift_r = 10; shift_g = 5; shift_b = 0; break; + case MODE_16_RGB: shift_r = 11; shift_g = 5; shift_b = 0; break; + } + } + + for (i = -197; i < 256+197; i++) + ((uint16_t *)table_r)[i] = (table_Y[i+384] >> 3) << shift_r; + + for (i = -132; i < 256+132; i++) { + int j = table_Y[i+384] >> (((mode==MODE_16_RGB) || (mode==MODE_16_BGR)) ? 2 : 3); + if (swapped) + ((uint16_t *)table_g)[i] = (j&7) << 13 | (j>>3); + else + ((uint16_t *)table_g)[i] = j << 5; + } + for (i = -232; i < 256+232; i++) + ((uint16_t *)table_b)[i] = (table_Y[i+384] >> 3) << shift_b; + + break; + + case MODE_8_RGB: + case MODE_8_BGR: + table_8 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint8_t)); + + entry_size = sizeof (uint8_t); + table_r = table_8 + 197; + table_b = table_8 + 197 + 685; + table_g = table_8 + 197 + 2*682; + + switch (mode) { + case MODE_8_RGB: shift_r = 5; shift_g = 2; shift_b = 0; break; + case MODE_8_BGR: shift_r = 0; shift_g = 3; shift_b = 6; break; + } + + for (i = -197; i < 256+197; i++) + ((uint8_t *) table_r)[i] = (table_Y[i+384] >> 5) << shift_r; + for (i = -132; i < 256+132; i++) + ((uint8_t *) table_g)[i] = (table_Y[i+384] >> 5) << shift_g; + for (i = -232; i < 256+232; i++) + ((uint8_t *) table_b)[i] = (table_Y[i+384] >> 6) << shift_b; + break; + + case MODE_8_GRAY: + return; + + case MODE_PALETTE: + table_16 = malloc ((197 + 2*682 + 256 + 132) * sizeof (uint16_t)); + + entry_size = sizeof (uint16_t); + table_r = table_16 + 197; + table_b = table_16 + 197 + 685; + table_g = table_16 + 197 + 2*682; + + shift_r = 10; + shift_g = 5; + shift_b = 0; + + for (i = -197; i < 256+197; i++) + ((uint16_t *)table_r)[i] = (table_Y[i+384] >> 3) << 10; + + for (i = -132; i < 256+132; i++) + ((uint16_t *)table_g)[i] = (table_Y[i+384] >> 3) << 5; + + for (i = -232; i < 256+232; i++) + ((uint16_t *)table_b)[i] = (table_Y[i+384] >> 3) << 0; + + break; + + + default: + fprintf (stderr, "mode %d not supported by yuv2rgb\n", mode); + abort(); + } + + for (i = 0; i < 256; i++) { + this->table_rV[i] = (((uint8_t *) table_r) + + entry_size * div_round (crv * (i-128), 76309)); + this->table_gU[i] = (((uint8_t *) table_g) + + entry_size * div_round (cgu * (i-128), 76309)); + this->table_gV[i] = entry_size * div_round (cgv * (i-128), 76309); + this->table_bU[i] = (((uint8_t *)table_b) + + entry_size * div_round (cbu * (i-128), 76309)); + } + this->gamma = 0; + this->entry_size = entry_size; +} + +static uint32_t yuv2rgb_single_pixel_32 (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v) +{ + uint32_t * r, * g, * b; + + r = this->table_rV[v]; + g = (void *) (((uint8_t *)this->table_gU[u]) + this->table_gV[v]); + b = this->table_bU[u]; + + return r[y] + g[y] + b[y]; +} + +static uint32_t yuv2rgb_single_pixel_24_rgb (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v) +{ + uint8_t * r, * g, * b; + + r = this->table_rV[v]; + g = (void *) (((uint8_t *)this->table_gU[u]) + this->table_gV[v]); + b = this->table_bU[u]; + + return (uint32_t) r[y] + + ((uint32_t) g[y] << 8) + + ((uint32_t) b[y] << 16); +} + +static uint32_t yuv2rgb_single_pixel_24_bgr (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v) +{ + uint8_t * r, * g, * b; + + r = this->table_rV[v]; + g = (void *) (((uint8_t *)this->table_gU[u]) + this->table_gV[v]); + b = this->table_bU[u]; + + return (uint32_t) b[y] + + ((uint32_t) g[y] << 8) + + ((uint32_t) r[y] << 16); +} + +static uint32_t yuv2rgb_single_pixel_16 (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v) +{ + uint16_t * r, * g, * b; + + r = this->table_rV[v]; + g = (void *) (((uint8_t *)this->table_gU[u]) + this->table_gV[v]); + b = this->table_bU[u]; + + return r[y] + g[y] + b[y]; +} + +static uint32_t yuv2rgb_single_pixel_8 (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v) +{ + uint8_t * r, * g, * b; + + r = this->table_rV[v]; + g = (void *) (((uint8_t *)this->table_gU[u]) + this->table_gV[v]); + b = this->table_bU[u]; + + return r[y] + g[y] + b[y]; +} + +static uint32_t yuv2rgb_single_pixel_gray (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v) +{ + return y; +} + +static uint32_t yuv2rgb_single_pixel_palette (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v) +{ + uint16_t * r, * g, * b; + + r = this->table_rV[v]; + g = (void *) (((uint8_t *)this->table_gU[u]) + this->table_gV[v]); + b = this->table_bU[u]; + + return this->cmap[r[y] + g[y] + b[y]]; +} + + +static void yuv2rgb_c_init (yuv2rgb_factory_t *this) +{ + switch (this->mode) { + case MODE_32_RGB: + case MODE_32_BGR: + this->yuv2rgb_fun = yuv2rgb_c_32; + break; + + case MODE_24_RGB: + case MODE_24_BGR: + this->yuv2rgb_fun = + (this->mode==MODE_24_RGB && !this->swapped) || (this->mode==MODE_24_BGR && this->swapped) + ? yuv2rgb_c_24_rgb + : yuv2rgb_c_24_bgr; + break; + + case MODE_15_BGR: + case MODE_16_BGR: + case MODE_15_RGB: + case MODE_16_RGB: + this->yuv2rgb_fun = yuv2rgb_c_16; + break; + + case MODE_8_RGB: + case MODE_8_BGR: + this->yuv2rgb_fun = yuv2rgb_c_8; + break; + + case MODE_8_GRAY: + this->yuv2rgb_fun = yuv2rgb_c_gray; + break; + + case MODE_PALETTE: + this->yuv2rgb_fun = yuv2rgb_c_palette; + break; + + default: + printf ("yuv2rgb: mode %d not supported by yuv2rgb\n", this->mode); + abort(); + } + +} + +static void yuv2rgb_single_pixel_init (yuv2rgb_factory_t *this) { + + switch (this->mode) { + case MODE_32_RGB: + case MODE_32_BGR: + this->yuv2rgb_single_pixel_fun = yuv2rgb_single_pixel_32; + break; + + case MODE_24_RGB: + case MODE_24_BGR: + this->yuv2rgb_single_pixel_fun = + (this->mode==MODE_24_RGB && !this->swapped) || (this->mode==MODE_24_BGR && this->swapped) + ? yuv2rgb_single_pixel_24_rgb + : yuv2rgb_single_pixel_24_bgr; + break; + + case MODE_15_BGR: + case MODE_16_BGR: + case MODE_15_RGB: + case MODE_16_RGB: + this->yuv2rgb_single_pixel_fun = yuv2rgb_single_pixel_16; + break; + + case MODE_8_RGB: + case MODE_8_BGR: + this->yuv2rgb_single_pixel_fun = yuv2rgb_single_pixel_8; + break; + + case MODE_8_GRAY: + this->yuv2rgb_single_pixel_fun = yuv2rgb_single_pixel_gray; + break; + + case MODE_PALETTE: + this->yuv2rgb_single_pixel_fun = yuv2rgb_single_pixel_palette; + break; + + default: + printf ("yuv2rgb: mode %d not supported by yuv2rgb\n", this->mode); + abort(); + } +} + + +/* + * yuy2 stuff + */ + +static void yuy22rgb_c_32 (yuv2rgb_t *this, uint8_t * _dst, uint8_t * _p) +{ + int U, V, Y; + uint8_t * py_1, * pu, * pv; + uint32_t * r, * g, * b; + uint32_t * dst_1; + int width, height; + int dy; + + /* FIXME: implement unscaled version */ + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + height = this->dest_height; + + for (;;) { + dst_1 = (uint32_t*)_dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + + RGB(0); + DST1(0); + + RGB(1); + DST1(1); + + RGB(2); + DST1(2); + + RGB(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width*4); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (height <= 0) + break; + + _p += this->y_stride*2*(dy>>15); + dy &= 32767; + /* + dy -= 32768; + _p += this->y_stride*2; + */ + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + } +} + +static void yuy22rgb_c_24_rgb (yuv2rgb_t *this, uint8_t * _dst, uint8_t * _p) +{ + int U, V, Y; + uint8_t * py_1, * pu, * pv; + uint8_t * r, * g, * b; + uint8_t * dst_1; + int width, height; + int dy; + + /* FIXME: implement unscaled version */ + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + height = this->dest_height; + + for (;;) { + dst_1 = _dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1RGB(0); + + RGB(1); + DST1RGB(1); + + RGB(2); + DST1RGB(2); + + RGB(3); + DST1RGB(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 24; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width*3); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (height <= 0) + break; + + _p += this->y_stride*2*(dy>>15); + dy &= 32767; + /* + dy -= 32768; + _p += this->y_stride*2; + */ + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + } +} + +static void yuy22rgb_c_24_bgr (yuv2rgb_t *this, uint8_t * _dst, uint8_t * _p) +{ + int U, V, Y; + uint8_t * py_1, * pu, * pv; + uint8_t * r, * g, * b; + uint8_t * dst_1; + int width, height; + int dy; + + /* FIXME: implement unscaled version */ + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + height = this->dest_height; + + for (;;) { + dst_1 = _dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1BGR(0); + + RGB(1); + DST1BGR(1); + + RGB(2); + DST1BGR(2); + + RGB(3); + DST1BGR(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 24; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width*3); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (height <= 0) + break; + + _p += this->y_stride*2*(dy>>15); + dy &= 32767; + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + } +} + +static void yuy22rgb_c_16 (yuv2rgb_t *this, uint8_t * _dst, uint8_t * _p) +{ + int U, V, Y; + uint8_t * py_1, * pu, * pv; + uint16_t * r, * g, * b; + uint16_t * dst_1; + int width, height; + int dy; + + /* FIXME: implement unscaled version */ + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + height = this->dest_height; + + for (;;) { + dst_1 = (uint16_t*)_dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1(0); + + RGB(1); + DST1(1); + + RGB(2); + DST1(2); + + RGB(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width*2); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (height <= 0) + break; + + _p += this->y_stride*2*(dy>>15); + dy &= 32767; + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + } +} + +static void yuy22rgb_c_8 (yuv2rgb_t *this, uint8_t * _dst, uint8_t * _p) +{ + int U, V, Y; + uint8_t * py_1, * pu, * pv; + uint8_t * r, * g, * b; + uint8_t * dst_1; + int width, height; + int dy; + + /* FIXME: implement unscaled version */ + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + height = this->dest_height; + + for (;;) { + dst_1 = _dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1(0); + + RGB(1); + DST1(1); + + RGB(2); + DST1(2); + + RGB(3); + DST1(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (height <= 0) + break; + + _p += this->y_stride*2*(dy>>15); + dy &= 32767; + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + } +} + +static void yuy22rgb_c_gray (yuv2rgb_t *this, uint8_t * _dst, uint8_t * _p) +{ + int width, height; + int dy; + uint8_t * dst; + uint8_t * y; + + if (this->do_scale) { + dy = 0; + height = this->dest_height; + + for (;;) { + scale_line_2 (_p, _dst, this->dest_width, this->step_dx); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (height <= 0) + break; + + _p += this->y_stride*2*(dy>>15); + dy &= 32767; + } + } else { + for (height = this->source_height; --height >= 0; ) { + dst = _dst; + y = _p; + for (width = this->source_width; --width >= 0; ) { + *dst++ = *y; + y += 2; + } + _dst += this->rgb_stride; + _p += this->y_stride*2; + } + } +} + +static void yuy22rgb_c_palette (yuv2rgb_t *this, uint8_t * _dst, uint8_t * _p) +{ + int U, V, Y; + uint8_t * py_1, * pu, * pv; + uint16_t * r, * g, * b; + uint8_t * dst_1; + int width, height; + int dy; + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + + dy = 0; + height = this->dest_height; + + for (;;) { + dst_1 = _dst; + py_1 = this->y_buffer; + pu = this->u_buffer; + pv = this->v_buffer; + + width = this->dest_width >> 3; + + do { + RGB(0); + DST1CMAP(0); + + RGB(1); + DST1CMAP(1); + + RGB(2); + DST1CMAP(2); + + RGB(3); + DST1CMAP(3); + + pu += 4; + pv += 4; + py_1 += 8; + dst_1 += 8; + } while (--width); + + dy += this->step_dy; + _dst += this->rgb_stride; + + while (--height > 0 && dy < 32768) { + + xine_fast_memcpy (_dst, (uint8_t*)_dst-this->rgb_stride, this->dest_width); + + dy += this->step_dy; + _dst += this->rgb_stride; + } + + if (height <= 0) + break; + + _p += this->y_stride*2*(dy>>15); + dy &= 32767; + + scale_line_4 (_p+1, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_4 (_p+3, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line_2 (_p, this->y_buffer, + this->dest_width, this->step_dx); + } +} + +static void yuy22rgb_c_init (yuv2rgb_factory_t *this) +{ + switch (this->mode) { + case MODE_32_RGB: + case MODE_32_BGR: + this->yuy22rgb_fun = yuy22rgb_c_32; + break; + + case MODE_24_RGB: + case MODE_24_BGR: + this->yuy22rgb_fun = + (this->mode==MODE_24_RGB && !this->swapped) || (this->mode==MODE_24_BGR && this->swapped) + ? yuy22rgb_c_24_rgb + : yuy22rgb_c_24_bgr; + break; + case MODE_15_BGR: + case MODE_16_BGR: + case MODE_15_RGB: + case MODE_16_RGB: + this->yuy22rgb_fun = yuy22rgb_c_16; + break; + + case MODE_8_RGB: + case MODE_8_BGR: + this->yuy22rgb_fun = yuy22rgb_c_8; + break; + + case MODE_8_GRAY: + this->yuy22rgb_fun = yuy22rgb_c_gray; + break; + + case MODE_PALETTE: + this->yuy22rgb_fun = yuy22rgb_c_palette; + break; + + default: + printf ("yuv2rgb: mode %d not supported for yuy2\n", this->mode); + } +} + +yuv2rgb_t *yuv2rgb_create_converter (yuv2rgb_factory_t *factory) { + + yuv2rgb_t *this = xine_xmalloc (sizeof (yuv2rgb_t)); + + this->cmap = factory->cmap; + + this->y_chunk = this->y_buffer = NULL; + this->u_chunk = this->u_buffer = NULL; + this->v_chunk = this->v_buffer = NULL; + + this->table_rV = factory->table_rV; + this->table_gU = factory->table_gU; + this->table_gV = factory->table_gV; + this->table_bU = factory->table_bU; + + this->yuv2rgb_fun = factory->yuv2rgb_fun; + this->yuy22rgb_fun = factory->yuy22rgb_fun; + this->yuv2rgb_single_pixel_fun = factory->yuv2rgb_single_pixel_fun; + + this->configure = yuv2rgb_configure; + return this; +} + +/* + * factory functions + */ + +void yuv2rgb_set_gamma (yuv2rgb_factory_t *this, int gamma) { + + int i; + + for (i = 0; i < 256; i++) { + (uint8_t *)this->table_rV[i] += this->entry_size*(gamma - this->gamma); + (uint8_t *)this->table_gU[i] += this->entry_size*(gamma - this->gamma); + (uint8_t *)this->table_bU[i] += this->entry_size*(gamma - this->gamma); + } +#ifdef ARCH_X86 + mmx_yuv2rgb_set_gamma(gamma); +#endif + this->gamma = gamma; +} + +int yuv2rgb_get_gamma (yuv2rgb_factory_t *this) { + + return this->gamma; +} + +yuv2rgb_factory_t* yuv2rgb_factory_init (int mode, int swapped, + uint8_t *cmap) { + + yuv2rgb_factory_t *this; + +#ifdef ARCH_X86 + uint32_t mm = xine_mm_accel(); +#endif + + this = malloc (sizeof (yuv2rgb_factory_t)); + + this->mode = mode; + this->swapped = swapped; + this->cmap = cmap; + this->create_converter = yuv2rgb_create_converter; + this->set_gamma = yuv2rgb_set_gamma; + this->get_gamma = yuv2rgb_get_gamma; + this->matrix_coefficients = 6; + + + yuv2rgb_setup_tables (this, mode, swapped); + + /* + * auto-probe for the best yuv2rgb function + */ + + this->yuv2rgb_fun = NULL; +#ifdef ARCH_X86 + if ((this->yuv2rgb_fun == NULL) && (mm & MM_ACCEL_X86_MMXEXT)) { + + yuv2rgb_init_mmxext (this); + + if (this->yuv2rgb_fun != NULL) + printf ("yuv2rgb: using MMXEXT for colorspace transform\n"); + } + + if ((this->yuv2rgb_fun == NULL) && (mm & MM_ACCEL_X86_MMX)) { + + yuv2rgb_init_mmx (this); + + if (this->yuv2rgb_fun != NULL) + printf ("yuv2rgb: using MMX for colorspace transform\n"); + } +#endif +#if HAVE_MLIB + if (this->yuv2rgb_fun == NULL) { + + yuv2rgb_init_mlib (this); + + if (this->yuv2rgb_fun != NULL) + printf ("yuv2rgb: using medialib for colorspace transform\n"); + } +#endif + if (this->yuv2rgb_fun == NULL) { + printf ("yuv2rgb: no accelerated colorspace conversion found\n"); + yuv2rgb_c_init (this); + } + + /* + * auto-probe for the best yuy22rgb function + */ + + /* FIXME: implement mmx/mlib functions */ + yuy22rgb_c_init (this); + + /* + * set up single pixel function + */ + + yuv2rgb_single_pixel_init (this); + + return this; +} + diff --git a/noncore/multimedia/opieplayer2/yuv2rgb.h b/noncore/multimedia/opieplayer2/yuv2rgb.h new file mode 100644 index 0000000..5b9c3f6 --- a/dev/null +++ b/noncore/multimedia/opieplayer2/yuv2rgb.h @@ -0,0 +1,151 @@ + +#ifndef HAVE_YUV2RGB_H +#define HAVE_YUV2RGB_h + +#include <inttypes.h> + +typedef struct yuv2rgb_s yuv2rgb_t; + +typedef struct yuv2rgb_factory_s yuv2rgb_factory_t; + +/* + * function types for functions which can be replaced + * by hardware-accelerated versions + */ + +/* internal function use to scale yuv data */ +typedef void (*scale_line_func_t) (uint8_t *source, uint8_t *dest, int width, int step); + +typedef void (*yuv2rgb_fun_t) (yuv2rgb_t *this, uint8_t * image, uint8_t * py, uint8_t * pu, uint8_t * pv) ; + +typedef void (*yuy22rgb_fun_t) (yuv2rgb_t *this, uint8_t * image, uint8_t * p); + +typedef uint32_t (*yuv2rgb_single_pixel_fun_t) (yuv2rgb_t *this, uint8_t y, uint8_t u, uint8_t v); + + +/* + * modes supported - feel free to implement yours + */ + +#define MODE_8_RGB 1 +#define MODE_8_BGR 2 +#define MODE_15_RGB 3 +#define MODE_15_BGR 4 +#define MODE_16_RGB 5 +#define MODE_16_BGR 6 +#define MODE_24_RGB 7 +#define MODE_24_BGR 8 +#define MODE_32_RGB 9 +#define MODE_32_BGR 10 +#define MODE_8_GRAY 11 +#define MODE_PALETTE 12 + +struct yuv2rgb_s { + + /* + * configure converter for scaling factors + */ + int (*configure) (yuv2rgb_t *this, + int source_width, int source_height, + int y_stride, int uv_stride, + int dest_width, int dest_height, + int rgb_stride); + + /* + * this is the function to call for the yuv2rgb and scaling process + */ + yuv2rgb_fun_t yuv2rgb_fun; + + /* + * this is the function to call for the yuy2->rgb and scaling process + */ + yuy22rgb_fun_t yuy22rgb_fun; + + /* + * this is the function to call for the yuv2rgb for a single pixel + * (used for converting clut colors) + */ + + yuv2rgb_single_pixel_fun_t yuv2rgb_single_pixel_fun; + + /* private stuff below */ + + int source_width, source_height; + int y_stride, uv_stride; + int dest_width, dest_height; + int rgb_stride; + int step_dx, step_dy; + int do_scale; + + uint8_t *y_buffer; + uint8_t *u_buffer; + uint8_t *v_buffer; + void *y_chunk; + void *u_chunk; + void *v_chunk; + + void **table_rV; + void **table_gU; + int *table_gV; + void **table_bU; + + uint8_t *cmap; + scale_line_func_t scale_line; + +} ; + +/* + * convenience class to easily create a lot of converters + */ + +struct yuv2rgb_factory_s { + + yuv2rgb_t* (*create_converter) (yuv2rgb_factory_t *this); + + /* + * adjust gamma (-100 to 100 looks fine) + * for all converters produced by this factory + */ + void (*set_gamma) (yuv2rgb_factory_t *this, int gamma); + + /* + * get gamma value + */ + int (*get_gamma) (yuv2rgb_factory_t *this); + + /* private data */ + + int mode; + int swapped; + uint8_t *cmap; + + int gamma; + int entry_size; + + uint32_t matrix_coefficients; + + void *table_rV[256]; + void *table_gU[256]; + int table_gV[256]; + void *table_bU[256]; + + /* preselected functions for mode/swap/hardware */ + yuv2rgb_fun_t yuv2rgb_fun; + yuy22rgb_fun_t yuy22rgb_fun; + yuv2rgb_single_pixel_fun_t yuv2rgb_single_pixel_fun; + +}; + +yuv2rgb_factory_t *yuv2rgb_factory_init (int mode, int swapped, uint8_t *colormap); + + +/* + * internal stuff below this line + */ + +void mmx_yuv2rgb_set_gamma(int gamma); +void yuv2rgb_init_mmxext (yuv2rgb_factory_t *this); +void yuv2rgb_init_mmx (yuv2rgb_factory_t *this); +void yuv2rgb_init_mlib (yuv2rgb_factory_t *this); + +#endif diff --git a/noncore/multimedia/opieplayer2/yuv2rgb_mlib.c b/noncore/multimedia/opieplayer2/yuv2rgb_mlib.c new file mode 100644 index 0000000..908b439 --- a/dev/null +++ b/noncore/multimedia/opieplayer2/yuv2rgb_mlib.c @@ -0,0 +1,313 @@ +/* + * yuv2rgb_mlib.c + * Copyright (C) 2000-2001 Silicon Integrated System Corp. + * All Rights Reserved. + * + * Author: Juergen Keil <jk@tools.de> + * + * This file is part of xine, a free unix video player. + * + * xine is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * xine is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + + +#if HAVE_MLIB + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <inttypes.h> +#include <mlib_video.h> + +#include "attributes.h" +#include "yuv2rgb.h" + + +static void scale_line (uint8_t *source, uint8_t *dest, + int width, int step) { + + unsigned p1; + unsigned p2; + int dx; + + p1 = *source++; + p2 = *source++; + dx = 0; + + while (width) { + + /* + printf ("scale_line, width = %d\n", width); + printf ("scale_line, dx = %d, p1 = %d, p2 = %d\n", dx, p1, p2); + */ + + *dest = (p1 * (32768 - dx) + p2 * dx) / 32768; + + dx += step; + while (dx > 32768) { + dx -= 32768; + p1 = p2; + p2 = *source++; + } + + dest ++; + width --; + } +} + + + +static void mlib_yuv420_rgb24 (yuv2rgb_t *this, + uint8_t * image, uint8_t * py, + uint8_t * pu, uint8_t * pv) +{ + int dst_height; + int dy; + mlib_status mlib_stat; + + if (this->do_scale) { + dy = 0; + dst_height = this->dest_height; + + for (;;) { + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + pu += this->uv_stride; + + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + pv += this->uv_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + py += this->y_stride; + scale_line (py, this->y_buffer + this->dest_width, + this->dest_width, this->step_dx); + py += this->y_stride; + + mlib_stat = mlib_VideoColorYUV2RGB420(image, + this->y_buffer, + this->u_buffer, + this->v_buffer, + this->dest_width & ~1, 2, + this->rgb_stride, + this->dest_width, + this->dest_width >> 1); + dy += this->step_dy; + image += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + memcpy (image, (uint8_t*)image-this->rgb_stride, this->dest_width*6); + dy += this->step_dy; + image += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + dy -= 32768; + + dy += this->step_dy; + image += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + memcpy (image, (uint8_t*)image-this->rgb_stride, this->dest_width*3); + dy += this->step_dy; + image += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + dy -= 32768; + } + } else { + mlib_stat = mlib_VideoColorYUV2RGB420(image, py, pu, pv, + this->source_width, + this->source_height, + this->rgb_stride, + this->y_stride, + this->uv_stride); + } +} + +static void mlib_yuv420_argb32 (yuv2rgb_t *this, + uint8_t * image, uint8_t * py, + uint8_t * pu, uint8_t * pv) +{ + int dst_height; + int dy; + mlib_status mlib_stat; + + if (this->do_scale) { + dy = 0; + dst_height = this->dest_height; + + for (;;) { + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + pu += this->uv_stride; + + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + pv += this->uv_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + py += this->y_stride; + scale_line (py, this->y_buffer + this->dest_width, + this->dest_width, this->step_dx); + py += this->y_stride; + + mlib_stat = mlib_VideoColorYUV2ARGB420(image, + this->y_buffer, + this->u_buffer, + this->v_buffer, + this->dest_width & ~1, 2, + this->rgb_stride, + this->dest_width, + this->dest_width >> 1); + dy += this->step_dy; + image += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + memcpy (image, (uint8_t*)image-this->rgb_stride, this->dest_width*8); + dy += this->step_dy; + image += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + dy -= 32768; + + dy += this->step_dy; + image += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + memcpy (image, (uint8_t*)image-this->rgb_stride, this->dest_width*4); + dy += this->step_dy; + image += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + dy -= 32768; + } + } else { + mlib_stat = mlib_VideoColorYUV2ARGB420(image, py, pu, pv, + this->source_width, + this->source_height, + this->rgb_stride, + this->y_stride, + this->uv_stride); + } +} + +static void mlib_yuv420_abgr32 (yuv2rgb_t *this, + uint8_t * image, uint8_t * py, + uint8_t * pu, uint8_t * pv) +{ + int dst_height; + int dy; + mlib_status mlib_stat; + + if (this->do_scale) { + dy = 0; + dst_height = this->dest_height; + + for (;;) { + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + pu += this->uv_stride; + + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + pv += this->uv_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + py += this->y_stride; + scale_line (py, this->y_buffer + this->dest_width, + this->dest_width, this->step_dx); + py += this->y_stride; + + mlib_stat = mlib_VideoColorYUV2ABGR420(image, + this->y_buffer, + this->u_buffer, + this->v_buffer, + this->dest_width & ~1, 2, + this->rgb_stride, + this->dest_width, + this->dest_width >> 1); + dy += this->step_dy; + image += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + memcpy (image, (uint8_t*)image-this->rgb_stride, this->dest_width*8); + dy += this->step_dy; + image += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + dy -= 32768; + + dy += this->step_dy; + image += this->rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + memcpy (image, (uint8_t*)image-this->rgb_stride, this->dest_width*4); + dy += this->step_dy; + image += this->rgb_stride; + } + + if (dst_height <= 0) + break; + + dy -= 32768; + } + } else { + mlib_stat = mlib_VideoColorYUV2ABGR420(image, py, pu, pv, + this->source_width, + this->source_height, + this->rgb_stride, + this->y_stride, + this->uv_stride); + } +} + + +void yuv2rgb_init_mlib (yuv2rgb_factory_t *this) { + + if (this->swapped) return; /*no swapped pixel output upto now*/ + + switch (this->mode) { + case MODE_24_RGB: + this->yuv2rgb_fun = mlib_yuv420_rgb24; + break; + case MODE_32_RGB: + this->yuv2rgb_fun = mlib_yuv420_argb32; + break; + case MODE_32_BGR: + this->yuv2rgb_fun = mlib_yuv420_abgr32; + break; + } +} + + +#endif /* HAVE_MLIB */ diff --git a/noncore/multimedia/opieplayer2/yuv2rgb_mmx.c b/noncore/multimedia/opieplayer2/yuv2rgb_mmx.c new file mode 100644 index 0000000..f092e6f --- a/dev/null +++ b/noncore/multimedia/opieplayer2/yuv2rgb_mmx.c @@ -0,0 +1,1047 @@ +/* + * yuv2rgb_mmx.c + * Copyright (C) 2000-2001 Silicon Integrated System Corp. + * All Rights Reserved. + * + * Author: Olie Lho <ollie@sis.com.tw> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * + * mpeg2dec is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * mpeg2dec is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + + +#ifdef ARCH_X86 + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <inttypes.h> + +#include "yuv2rgb.h" +#include "xineutils.h" + +#define CPU_MMXEXT 0 +#define CPU_MMX 1 + +/* CPU_MMXEXT/CPU_MMX adaptation layer */ + +#define movntq(src,dest) \ +do { \ + if (cpu == CPU_MMXEXT) \ + movntq_r2m (src, dest); \ + else \ + movq_r2m (src, dest); \ +} while (0) + +static mmx_t mmx_subYw = {0x1010101010101010}; +static mmx_t mmx_addYw = {0x0000000000000000}; + +void mmx_yuv2rgb_set_gamma(int gamma) +{ +int a,s,i; + + if( gamma <= 16 ) { + a = 0; + s = 16 - gamma; + } else { + a = gamma - 16; + s = 0; + } + + for( i = 0; i < 8; i++ ) { + *((unsigned char *)&mmx_subYw + i) = s; + *((unsigned char *)&mmx_addYw + i) = a; + } +} + +static inline void mmx_yuv2rgb (uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + static mmx_t mmx_80w = {0x0080008000800080}; + static mmx_t mmx_U_green = {0xf37df37df37df37d}; + static mmx_t mmx_U_blue = {0x4093409340934093}; + static mmx_t mmx_V_red = {0x3312331233123312}; + static mmx_t mmx_V_green = {0xe5fce5fce5fce5fc}; + static mmx_t mmx_00ffw = {0x00ff00ff00ff00ff}; + static mmx_t mmx_Y_coeff = {0x253f253f253f253f}; + + movq_m2r (*py, mm6); // mm6 = Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 + pxor_r2r (mm4, mm4); // mm4 = 0 + + psubusb_m2r (mmx_subYw, mm6); // Y -= 16 + paddusb_m2r (mmx_addYw, mm6); + + movd_m2r (*pu, mm0); // mm0 = 00 00 00 00 u3 u2 u1 u0 + movq_r2r (mm6, mm7); // mm7 = Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 + + pand_m2r (mmx_00ffw, mm6); // mm6 = Y6 Y4 Y2 Y0 + psrlw_i2r (8, mm7); // mm7 = Y7 Y5 Y3 Y1 + + movd_m2r (*pv, mm1); // mm1 = 00 00 00 00 v3 v2 v1 v0 + psllw_i2r (3, mm6); // promote precision + + pmulhw_m2r (mmx_Y_coeff, mm6); // mm6 = luma_rgb even + psllw_i2r (3, mm7); // promote precision + + punpcklbw_r2r (mm4, mm0); // mm0 = u3 u2 u1 u0 + + psubsw_m2r (mmx_80w, mm0); // u -= 128 + punpcklbw_r2r (mm4, mm1); // mm1 = v3 v2 v1 v0 + + pmulhw_m2r (mmx_Y_coeff, mm7); // mm7 = luma_rgb odd + psllw_i2r (3, mm0); // promote precision + + psubsw_m2r (mmx_80w, mm1); // v -= 128 + movq_r2r (mm0, mm2); // mm2 = u3 u2 u1 u0 + + psllw_i2r (3, mm1); // promote precision + + movq_r2r (mm1, mm4); // mm4 = v3 v2 v1 v0 + + pmulhw_m2r (mmx_U_blue, mm0); // mm0 = chroma_b + + + // slot + + + // slot + + + pmulhw_m2r (mmx_V_red, mm1); // mm1 = chroma_r + movq_r2r (mm0, mm3); // mm3 = chroma_b + + paddsw_r2r (mm6, mm0); // mm0 = B6 B4 B2 B0 + paddsw_r2r (mm7, mm3); // mm3 = B7 B5 B3 B1 + + packuswb_r2r (mm0, mm0); // saturate to 0-255 + + + pmulhw_m2r (mmx_U_green, mm2); // mm2 = u * u_green + + + packuswb_r2r (mm3, mm3); // saturate to 0-255 + + + punpcklbw_r2r (mm3, mm0); // mm0 = B7 B6 B5 B4 B3 B2 B1 B0 + + + pmulhw_m2r (mmx_V_green, mm4); // mm4 = v * v_green + + + // slot + + + // slot + + + paddsw_r2r (mm4, mm2); // mm2 = chroma_g + movq_r2r (mm2, mm5); // mm5 = chroma_g + + + movq_r2r (mm1, mm4); // mm4 = chroma_r + paddsw_r2r (mm6, mm2); // mm2 = G6 G4 G2 G0 + + + packuswb_r2r (mm2, mm2); // saturate to 0-255 + paddsw_r2r (mm6, mm1); // mm1 = R6 R4 R2 R0 + + packuswb_r2r (mm1, mm1); // saturate to 0-255 + paddsw_r2r (mm7, mm4); // mm4 = R7 R5 R3 R1 + + packuswb_r2r (mm4, mm4); // saturate to 0-255 + paddsw_r2r (mm7, mm5); // mm5 = G7 G5 G3 G1 + + + packuswb_r2r (mm5, mm5); // saturate to 0-255 + + + punpcklbw_r2r (mm4, mm1); // mm1 = R7 R6 R5 R4 R3 R2 R1 R0 + + + punpcklbw_r2r (mm5, mm2); // mm2 = G7 G6 G5 G4 G3 G2 G1 G0 +} + +// basic opt +static inline void mmx_unpack_16rgb (uint8_t * image, int cpu) +{ + static mmx_t mmx_bluemask = {0xf8f8f8f8f8f8f8f8}; + static mmx_t mmx_greenmask = {0xfcfcfcfcfcfcfcfc}; + static mmx_t mmx_redmask = {0xf8f8f8f8f8f8f8f8}; + + /* + * convert RGB plane to RGB 16 bits + * mm0 -> B, mm1 -> R, mm2 -> G + * mm4 -> GB, mm5 -> AR pixel 4-7 + * mm6 -> GB, mm7 -> AR pixel 0-3 + */ + + pand_m2r (mmx_bluemask, mm0); // mm0 = b7b6b5b4b3______ + pxor_r2r (mm4, mm4); // mm4 = 0 + + pand_m2r (mmx_greenmask, mm2); // mm2 = g7g6g5g4g3g2____ + psrlq_i2r (3, mm0); // mm0 = ______b7b6b5b4b3 + + movq_r2r (mm2, mm7); // mm7 = g7g6g5g4g3g2____ + movq_r2r (mm0, mm5); // mm5 = ______b7b6b5b4b3 + + pand_m2r (mmx_redmask, mm1); // mm1 = r7r6r5r4r3______ + punpcklbw_r2r (mm4, mm2); + + punpcklbw_r2r (mm1, mm0); + + psllq_i2r (3, mm2); + + punpckhbw_r2r (mm4, mm7); + por_r2r (mm2, mm0); + + psllq_i2r (3, mm7); + + movntq (mm0, *image); + punpckhbw_r2r (mm1, mm5); + + por_r2r (mm7, mm5); + + // U + // V + + movntq (mm5, *(image+8)); +} + +static inline void mmx_unpack_15rgb (uint8_t * image, int cpu) +{ + static mmx_t mmx_bluemask = {0xf8f8f8f8f8f8f8f8}; + static mmx_t mmx_greenmask = {0xf8f8f8f8f8f8f8f8}; + static mmx_t mmx_redmask = {0xf8f8f8f8f8f8f8f8}; + + /* + * convert RGB plane to RGB 15 bits + * mm0 -> B, mm1 -> R, mm2 -> G + * mm4 -> GB, mm5 -> AR pixel 4-7 + * mm6 -> GB, mm7 -> AR pixel 0-3 + */ + + pand_m2r (mmx_bluemask, mm0); // mm0 = b7b6b5b4b3______ + pxor_r2r (mm4, mm4); // mm4 = 0 + + pand_m2r (mmx_greenmask, mm2); // mm2 = g7g6g5g4g3g2____ + psrlq_i2r (3, mm0); // mm0 = ______b7b6b5b4b3 + + movq_r2r (mm2, mm7); // mm7 = g7g6g5g4g3g2____ + movq_r2r (mm0, mm5); // mm5 = ______b7b6b5b4b3 + + pand_m2r (mmx_redmask, mm1); // mm1 = r7r6r5r4r3______ + punpcklbw_r2r (mm4, mm2); + + psrlq_i2r (1, mm1); + punpcklbw_r2r (mm1, mm0); + + psllq_i2r (2, mm2); + + punpckhbw_r2r (mm4, mm7); + por_r2r (mm2, mm0); + + psllq_i2r (2, mm7); + + movntq (mm0, *image); + punpckhbw_r2r (mm1, mm5); + + por_r2r (mm7, mm5); + + // U + // V + + movntq (mm5, *(image+8)); +} + +static inline void mmx_unpack_32rgb (uint8_t * image, int cpu) +{ + /* + * convert RGB plane to RGB packed format, + * mm0 -> B, mm1 -> R, mm2 -> G, mm3 -> 0, + * mm4 -> GB, mm5 -> AR pixel 4-7, + * mm6 -> GB, mm7 -> AR pixel 0-3 + */ + + pxor_r2r (mm3, mm3); + movq_r2r (mm0, mm6); + + punpcklbw_r2r (mm2, mm6); + movq_r2r (mm1, mm7); + + punpcklbw_r2r (mm3, mm7); + movq_r2r (mm0, mm4); + + punpcklwd_r2r (mm7, mm6); + movq_r2r (mm1, mm5); + + /* scheduling: this is hopeless */ + movntq (mm6, *image); + movq_r2r (mm0, mm6); + punpcklbw_r2r (mm2, mm6); + punpckhwd_r2r (mm7, mm6); + movntq (mm6, *(image+8)); + punpckhbw_r2r (mm2, mm4); + punpckhbw_r2r (mm3, mm5); + punpcklwd_r2r (mm5, mm4); + movntq (mm4, *(image+16)); + movq_r2r (mm0, mm4); + punpckhbw_r2r (mm2, mm4); + punpckhwd_r2r (mm5, mm4); + movntq (mm4, *(image+24)); +} + +static inline void mmx_unpack_32bgr (uint8_t * image, int cpu) +{ + /* + * convert RGB plane to RGB packed format, + * mm0 -> B, mm1 -> R, mm2 -> G, mm3 -> 0, + * mm4 -> GB, mm5 -> AR pixel 4-7, + * mm6 -> GB, mm7 -> AR pixel 0-3 + */ + + pxor_r2r (mm3, mm3); + movq_r2r (mm1, mm6); + + punpcklbw_r2r (mm2, mm6); + movq_r2r (mm0, mm7); + + punpcklbw_r2r (mm3, mm7); + movq_r2r (mm1, mm4); + + punpcklwd_r2r (mm7, mm6); + movq_r2r (mm0, mm5); + + /* scheduling: this is hopeless */ + movntq (mm6, *image); + movq_r2r (mm0, mm6); + punpcklbw_r2r (mm2, mm6); + punpckhwd_r2r (mm7, mm6); + movntq (mm6, *(image+8)); + punpckhbw_r2r (mm2, mm4); + punpckhbw_r2r (mm3, mm5); + punpcklwd_r2r (mm5, mm4); + movntq (mm4, *(image+16)); + movq_r2r (mm0, mm4); + punpckhbw_r2r (mm2, mm4); + punpckhwd_r2r (mm5, mm4); + movntq (mm4, *(image+24)); +} + +static inline void mmx_unpack_24rgb (uint8_t * image, int cpu) +{ + /* + * convert RGB plane to RGB packed format, + * mm0 -> B, mm1 -> R, mm2 -> G, mm3 -> 0, + * mm4 -> GB, mm5 -> AR pixel 4-7, + * mm6 -> GB, mm7 -> AR pixel 0-3 + */ + + pxor_r2r (mm3, mm3); + movq_r2r (mm0, mm6); + + punpcklbw_r2r (mm2, mm6); + movq_r2r (mm1, mm7); + + punpcklbw_r2r (mm3, mm7); + movq_r2r (mm0, mm4); + + punpcklwd_r2r (mm7, mm6); + movq_r2r (mm1, mm5); + + /* scheduling: this is hopeless */ + movntq (mm6, *image); + movq_r2r (mm0, mm6); + punpcklbw_r2r (mm2, mm6); + punpckhwd_r2r (mm7, mm6); + movntq (mm6, *(image+8)); + punpckhbw_r2r (mm2, mm4); + punpckhbw_r2r (mm3, mm5); + punpcklwd_r2r (mm5, mm4); + movntq (mm4, *(image+16)); +} + +static inline void yuv420_rgb16 (yuv2rgb_t *this, + uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv, + int cpu) +{ + int i; + int rgb_stride = this->rgb_stride; + int y_stride = this->y_stride; + int uv_stride = this->uv_stride; + int width = this->source_width; + int height = this->source_height; + int dst_height = this->dest_height; + uint8_t *img; + + width >>= 3; + + if (!this->do_scale) { + y_stride -= 8 * width; + uv_stride -= 4 * width; + + do { + + i = width; img = image; + do { + mmx_yuv2rgb (py, pu, pv); + mmx_unpack_16rgb (img, cpu); + py += 8; + pu += 4; + pv += 4; + img += 16; + } while (--i); + + py += y_stride; + image += rgb_stride; + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + } else { + pu -= 4 * width; + pv -= 4 * width; + } + } while (--height); + + } else { + + scale_line_func_t scale_line = this->scale_line; + uint8_t *y_buf, *u_buf, *v_buf; + int dy = 0; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + for (height = 0;; ) { + + y_buf = this->y_buffer; + u_buf = this->u_buffer; + v_buf = this->v_buffer; + + i = this->dest_width >> 3; img = image; + do { + /* printf ("i : %d\n",i); */ + + mmx_yuv2rgb (y_buf, u_buf, v_buf); + mmx_unpack_16rgb (img, cpu); + y_buf += 8; + u_buf += 4; + v_buf += 4; + img += 16; + } while (--i); + + dy += this->step_dy; + image += rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (image, image-rgb_stride, this->dest_width*2); + + dy += this->step_dy; + image += rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + + py += y_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while( dy>=32768); + } + } +} + +static inline void yuv420_rgb15 (yuv2rgb_t *this, + uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv, + int cpu) +{ + int i; + int rgb_stride = this->rgb_stride; + int y_stride = this->y_stride; + int uv_stride = this->uv_stride; + int width = this->source_width; + int height = this->source_height; + int dst_height = this->dest_height; + uint8_t *img; + + width >>= 3; + + if (!this->do_scale) { + y_stride -= 8 * width; + uv_stride -= 4 * width; + + do { + + i = width; img = image; + do { + mmx_yuv2rgb (py, pu, pv); + mmx_unpack_15rgb (img, cpu); + py += 8; + pu += 4; + pv += 4; + img += 16; + } while (--i); + + py += y_stride; + image += rgb_stride; + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + } else { + pu -= 4 * width; + pv -= 4 * width; + } + } while (--height); + + } else { + + scale_line_func_t scale_line = this->scale_line; + uint8_t *y_buf, *u_buf, *v_buf; + int dy = 0; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + for (height = 0;; ) { + + y_buf = this->y_buffer; + u_buf = this->u_buffer; + v_buf = this->v_buffer; + + i = this->dest_width >> 3; img = image; + do { + /* printf ("i : %d\n",i); */ + + mmx_yuv2rgb (y_buf, u_buf, v_buf); + mmx_unpack_15rgb (img, cpu); + y_buf += 8; + u_buf += 4; + v_buf += 4; + img += 16; + } while (--i); + + dy += this->step_dy; + image += rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (image, image-rgb_stride, this->dest_width*2); + + dy += this->step_dy; + image += rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + py += y_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + + } + height++; + } while( dy>=32768 ); + } + } +} + +static inline void yuv420_rgb24 (yuv2rgb_t *this, + uint8_t * image, uint8_t * py, + uint8_t * pu, uint8_t * pv, int cpu) +{ + int i; + int rgb_stride = this->rgb_stride; + int y_stride = this->y_stride; + int uv_stride = this->uv_stride; + int width = this->source_width; + int height = this->source_height; + int dst_height = this->dest_height; + uint8_t *img; + + /* rgb_stride -= 4 * this->dest_width; */ + width >>= 3; + + if (!this->do_scale) { + y_stride -= 8 * width; + uv_stride -= 4 * width; + + do { + i = width; img = image; + do { + mmx_yuv2rgb (py, pu, pv); + mmx_unpack_24rgb (img, cpu); + py += 8; + pu += 4; + pv += 4; + img += 24; + } while (--i); + + py += y_stride; + image += rgb_stride; + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + } else { + pu -= 4 * width; + pv -= 4 * width; + } + } while (--height); + } else { + + scale_line_func_t scale_line = this->scale_line; + uint8_t *y_buf, *u_buf, *v_buf; + int dy = 0; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + for (height = 0;; ) { + + y_buf = this->y_buffer; + u_buf = this->u_buffer; + v_buf = this->v_buffer; + + + i = this->dest_width >> 3; img=image; + do { + /* printf ("i : %d\n",i); */ + + mmx_yuv2rgb (y_buf, u_buf, v_buf); + mmx_unpack_24rgb (img, cpu); + y_buf += 8; + u_buf += 4; + v_buf += 4; + img += 24; + } while (--i); + + dy += this->step_dy; + image += rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (image, image-rgb_stride, this->dest_width*3); + + dy += this->step_dy; + image += rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + py += y_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + } + height++; + } while( dy>=32768 ); + + } + + } +} + +static inline void yuv420_argb32 (yuv2rgb_t *this, + uint8_t * image, uint8_t * py, + uint8_t * pu, uint8_t * pv, int cpu) +{ + int i; + int rgb_stride = this->rgb_stride; + int y_stride = this->y_stride; + int uv_stride = this->uv_stride; + int width = this->source_width; + int height = this->source_height; + int dst_height = this->dest_height; + uint8_t *img; + + /* rgb_stride -= 4 * this->dest_width; */ + width >>= 3; + + if (!this->do_scale) { + y_stride -= 8 * width; + uv_stride -= 4 * width; + + do { + i = width; img = image; + do { + mmx_yuv2rgb (py, pu, pv); + mmx_unpack_32rgb (img, cpu); + py += 8; + pu += 4; + pv += 4; + img += 32; + } while (--i); + + py += y_stride; + image += rgb_stride; + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + } else { + pu -= 4 * width; + pv -= 4 * width; + } + } while (--height); + } else { + + scale_line_func_t scale_line = this->scale_line; + uint8_t *y_buf, *u_buf, *v_buf; + int dy = 0; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + for (height = 0;; ) { + + y_buf = this->y_buffer; + u_buf = this->u_buffer; + v_buf = this->v_buffer; + + + i = this->dest_width >> 3; img=image; + do { + /* printf ("i : %d\n",i); */ + + mmx_yuv2rgb (y_buf, u_buf, v_buf); + mmx_unpack_32rgb (img, cpu); + y_buf += 8; + u_buf += 4; + v_buf += 4; + img += 32; + } while (--i); + + dy += this->step_dy; + image += rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (image, image-rgb_stride, this->dest_width*4); + + dy += this->step_dy; + image += rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + py += y_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + } + height++; + } while( dy>=32768 ); + } + + } +} + +static inline void yuv420_abgr32 (yuv2rgb_t *this, + uint8_t * image, uint8_t * py, + uint8_t * pu, uint8_t * pv, int cpu) +{ + int i; + int rgb_stride = this->rgb_stride; + int y_stride = this->y_stride; + int uv_stride = this->uv_stride; + int width = this->source_width; + int height = this->source_height; + int dst_height = this->dest_height; + uint8_t *img; + + /* rgb_stride -= 4 * this->dest_width; */ + width >>= 3; + + if (!this->do_scale) { + y_stride -= 8 * width; + uv_stride -= 4 * width; + + do { + i = width; img = image; + do { + mmx_yuv2rgb (py, pu, pv); + mmx_unpack_32bgr (img, cpu); + py += 8; + pu += 4; + pv += 4; + img += 32; + } while (--i); + + py += y_stride; + image += rgb_stride; + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + } else { + pu -= 4 * width; + pv -= 4 * width; + } + } while (--height); + } else { + + scale_line_func_t scale_line = this->scale_line; + uint8_t *y_buf, *u_buf, *v_buf; + int dy = 0; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + for (height = 0;; ) { + + y_buf = this->y_buffer; + u_buf = this->u_buffer; + v_buf = this->v_buffer; + + + i = this->dest_width >> 3; img=image; + do { + /* printf ("i : %d\n",i); */ + + mmx_yuv2rgb (y_buf, u_buf, v_buf); + mmx_unpack_32bgr (img, cpu); + y_buf += 8; + u_buf += 4; + v_buf += 4; + img += 32; + } while (--i); + + dy += this->step_dy; + image += rgb_stride; + + while (--dst_height > 0 && dy < 32768) { + + xine_fast_memcpy (image, image-rgb_stride, this->dest_width*4); + + dy += this->step_dy; + image += rgb_stride; + } + + if (dst_height <= 0) + break; + + do { + dy -= 32768; + py += y_stride; + + scale_line (py, this->y_buffer, + this->dest_width, this->step_dx); + + if (height & 1) { + pu += uv_stride; + pv += uv_stride; + + scale_line (pu, this->u_buffer, + this->dest_width >> 1, this->step_dx); + scale_line (pv, this->v_buffer, + this->dest_width >> 1, this->step_dx); + } + height++; + } while( dy>=32768 ); + + } + + } +} + +static void mmxext_rgb15 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_rgb15 (this, image, py, pu, pv, CPU_MMXEXT); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmxext_rgb16 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_rgb16 (this, image, py, pu, pv, CPU_MMXEXT); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmxext_rgb24 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_rgb24 (this, image, py, pu, pv, CPU_MMXEXT); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmxext_argb32 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_argb32 (this, image, py, pu, pv, CPU_MMXEXT); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmxext_abgr32 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_abgr32 (this, image, py, pu, pv, CPU_MMXEXT); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmx_rgb15 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_rgb15 (this, image, py, pu, pv, CPU_MMX); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmx_rgb16 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_rgb16 (this, image, py, pu, pv, CPU_MMX); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmx_rgb24 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_rgb24 (this, image, py, pu, pv, CPU_MMX); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmx_argb32 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_argb32 (this, image, py, pu, pv, CPU_MMX); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +static void mmx_abgr32 (yuv2rgb_t *this, uint8_t * image, + uint8_t * py, uint8_t * pu, uint8_t * pv) +{ + yuv420_abgr32 (this, image, py, pu, pv, CPU_MMX); + emms(); /* re-initialize x86 FPU after MMX use */ +} + +void yuv2rgb_init_mmxext (yuv2rgb_factory_t *this) { + + if (this->swapped) + return; /*no swapped pixel output upto now*/ + + switch (this->mode) { + case MODE_15_RGB: + this->yuv2rgb_fun = mmxext_rgb15; + break; + case MODE_16_RGB: + this->yuv2rgb_fun = mmxext_rgb16; + break; + case MODE_24_RGB: + this->yuv2rgb_fun = mmxext_rgb24; + break; + case MODE_32_RGB: + this->yuv2rgb_fun = mmxext_argb32; + break; + case MODE_32_BGR: + this->yuv2rgb_fun = mmxext_abgr32; + break; + } +} + +void yuv2rgb_init_mmx (yuv2rgb_factory_t *this) { + + if (this->swapped) + return; /*no swapped pixel output upto now*/ + + switch (this->mode) { + case MODE_15_RGB: + this->yuv2rgb_fun = mmx_rgb15; + break; + case MODE_16_RGB: + this->yuv2rgb_fun = mmx_rgb16; + break; + case MODE_24_RGB: + this->yuv2rgb_fun = mmx_rgb24; + break; + case MODE_32_RGB: + this->yuv2rgb_fun = mmx_argb32; + break; + case MODE_32_BGR: + this->yuv2rgb_fun = mmx_abgr32; + break; + } +} + + +#endif |