diff options
Diffstat (limited to 'src/video_dec/libmpeg2')
27 files changed, 14506 insertions, 0 deletions
diff --git a/src/video_dec/libmpeg2/Makefile.am b/src/video_dec/libmpeg2/Makefile.am new file mode 100644 index 000000000..f0dd252eb --- /dev/null +++ b/src/video_dec/libmpeg2/Makefile.am @@ -0,0 +1,33 @@ +include $(top_builddir)/misc/Makefile.plugins +include $(top_srcdir)/misc/Makefile.common + +AM_CFLAGS = $(DEFAULT_OCFLAGS) $(VISIBILITY_FLAG) +AM_LDFLAGS = $(xineplug_ldflags) + +noinst_HEADERS = vlc.h mpeg2.h xvmc.h xvmc_vld.h mpeg2_internal.h idct_mlib.h vis.h \ + libmpeg2_accel.h + +xineplug_LTLIBRARIES = xineplug_decode_mpeg2.la + +xineplug_decode_mpeg2_la_SOURCES = \ + cpu_state.c \ + decode.c \ + header.c \ + idct.c \ + idct_altivec.c \ + idct_mlib.c \ + idct_mmx.c \ + motion_comp.c \ + motion_comp_altivec.c \ + motion_comp_mmx.c \ + motion_comp_mlib.c \ + motion_comp_vis.c \ + slice.c \ + slice_xvmc.c \ + slice_xvmc_vld.c \ + stats.c \ + xine_mpeg2_decoder.c \ + libmpeg2_accel.c + +xineplug_decode_mpeg2_la_LIBADD = $(XINE_LIB) $(MLIB_LIBS) $(LTLIBINTL) $(AVUTIL_LIBS) -lm +xineplug_decode_mpeg2_la_CFLAGS = $(AM_CFLAGS) $(MLIB_CFLAGS) $(AVUTIL_CFLAGS) diff --git a/src/video_dec/libmpeg2/cpu_state.c b/src/video_dec/libmpeg2/cpu_state.c new file mode 100644 index 000000000..d1507eec1 --- /dev/null +++ b/src/video_dec/libmpeg2/cpu_state.c @@ -0,0 +1,184 @@ +/* + * cpu_state.c + * Copyright (C) 2000-2003 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#include <stdlib.h> +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include <xine/xineutils.h> +#include "xine_mmx.h" + +void (* mpeg2_cpu_state_save) (cpu_state_t * state) = NULL; +void (* mpeg2_cpu_state_restore) (cpu_state_t * state) = NULL; + +#if defined(ARCH_X86) || defined(ARCH_X86_64) +static void state_restore_mmx (cpu_state_t * state) +{ + emms (); +} +#endif + +#if defined (ARCH_PPC) && defined (ENABLE_ALTIVEC) + +#ifndef HOST_OS_DARWIN + +static void state_save_altivec (cpu_state_t * state) +{ + asm (" \n" + " li %r9, 16 \n" + " stvx %v20, 0, %r3 \n" + " li %r11, 32 \n" + " stvx %v21, %r9, %r3 \n" + " li %r9, 48 \n" + " stvx %v22, %r11, %r3 \n" + " li %r11, 64 \n" + " stvx %v23, %r9, %r3 \n" + " li %r9, 80 \n" + " stvx %v24, %r11, %r3 \n" + " li %r11, 96 \n" + " stvx %v25, %r9, %r3 \n" + " li %r9, 112 \n" + " stvx %v26, %r11, %r3 \n" + " li %r11, 128 \n" + " stvx %v27, %r9, %r3 \n" + " li %r9, 144 \n" + " stvx %v28, %r11, %r3 \n" + " li %r11, 160 \n" + " stvx %v29, %r9, %r3 \n" + " li %r9, 176 \n" + " stvx %v30, %r11, %r3 \n" + " stvx %v31, %r9, %r3 \n" + ); +} + +static void state_restore_altivec (cpu_state_t * state) +{ + asm (" \n" + " li %r9, 16 \n" + " lvx %v20, 0, %r3 \n" + " li %r11, 32 \n" + " lvx %v21, %r9, %r3 \n" + " li %r9, 48 \n" + " lvx %v22, %r11, %r3 \n" + " li %r11, 64 \n" + " lvx %v23, %r9, %r3 \n" + " li %r9, 80 \n" + " lvx %v24, %r11, %r3 \n" + " li %r11, 96 \n" + " lvx %v25, %r9, %r3 \n" + " li %r9, 112 \n" + " lvx %v26, %r11, %r3 \n" + " li %r11, 128 \n" + " lvx %v27, %r9, %r3 \n" + " li %r9, 144 \n" + " lvx %v28, %r11, %r3 \n" + " li %r11, 160 \n" + " lvx %v29, %r9, %r3 \n" + " li %r9, 176 \n" + " lvx %v30, %r11, %r3 \n" + " lvx %v31, %r9, %r3 \n" + ); +} + +#else /* HOST_OS_DARWIN */ + +#define LI(a,b) "li r" #a "," #b "\n\t" +#define STVX0(a,b,c) "stvx v" #a ",0,r" #c "\n\t" +#define STVX(a,b,c) "stvx v" #a ",r" #b ",r" #c "\n\t" +#define LVX0(a,b,c) "lvx v" #a ",0,r" #c "\n\t" +#define LVX(a,b,c) "lvx v" #a ",r" #b ",r" #c "\n\t" + +static void state_save_altivec (cpu_state_t * state) +{ + asm (LI (9, 16) + STVX0 (20, 0, 3) + LI (11, 32) + STVX (21, 9, 3) + LI (9, 48) + STVX (22, 11, 3) + LI (11, 64) + STVX (23, 9, 3) + LI (9, 80) + STVX (24, 11, 3) + LI (11, 96) + STVX (25, 9, 3) + LI (9, 112) + STVX (26, 11, 3) + LI (11, 128) + STVX (27, 9, 3) + LI (9, 144) + STVX (28, 11, 3) + LI (11, 160) + STVX (29, 9, 3) + LI (9, 176) + STVX (30, 11, 3) + STVX (31, 9, 3)); +} + +static void state_restore_altivec (cpu_state_t * state) +{ + asm (LI (9, 16) + LVX0 (20, 0, 3) + LI (11, 32) + LVX (21, 9, 3) + LI (9, 48) + LVX (22, 11, 3) + LI (11, 64) + LVX (23, 9, 3) + LI (9, 80) + LVX (24, 11, 3) + LI (11, 96) + LVX (25, 9, 3) + LI (9, 112) + LVX (26, 11, 3) + LI (11, 128) + LVX (27, 9, 3) + LI (9, 144) + LVX (28, 11, 3) + LI (11, 160) + LVX (29, 9, 3) + LI (9, 176) + LVX (30, 11, 3) + LVX (31, 9, 3)); +} +#endif /* HOST_OS_DARWIN */ + +#endif /* defined (ARCH_PPC) && defined (ENABLE_ALTIVEC) */ + +void mpeg2_cpu_state_init (uint32_t mm_accel) +{ +#if defined(ARCH_X86) || defined(ARCH_X86_64) + if (mm_accel & MM_ACCEL_X86_MMX) { + mpeg2_cpu_state_restore = state_restore_mmx; + } +#endif +#if defined (ARCH_PPC) && defined (ENABLE_ALTIVEC) + if (mm_accel & MM_ACCEL_PPC_ALTIVEC) { + mpeg2_cpu_state_save = state_save_altivec; + mpeg2_cpu_state_restore = state_restore_altivec; + } +#endif +} + diff --git a/src/video_dec/libmpeg2/decode.c b/src/video_dec/libmpeg2/decode.c new file mode 100644 index 000000000..848d111fc --- /dev/null +++ b/src/video_dec/libmpeg2/decode.c @@ -0,0 +1,1006 @@ +/* + * decode.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + * + * xine-specific version by G. Bartsch + * + */ + +#include "config.h" +#include <stdio.h> +#include <string.h> /* memcpy/memset, try to remove */ +#include <stdlib.h> +#include <inttypes.h> +#include <math.h> + +#define LOG_MODULE "decode" +#define LOG_VERBOSE +/* +#define LOG +*/ + +#ifdef HAVE_FFMPEG_AVUTIL_H +# include <mem.h> +#else +# include <libavutil/mem.h> +#endif + +#include <xine/xine_internal.h> +#include <xine/video_out.h> + +#include "mpeg2.h" +#include "mpeg2_internal.h" +#include <xine/xineutils.h> +#include "libmpeg2_accel.h" + +/* +#define LOG_PAN_SCAN +*/ + +/* #define BUFFER_SIZE (224 * 1024) */ +#define BUFFER_SIZE (1194 * 1024) /* new buffer size for mpeg2dec 0.2.1 */ + +static void process_userdata(mpeg2dec_t *mpeg2dec, uint8_t *buffer); + +void mpeg2_init (mpeg2dec_t * mpeg2dec, + xine_video_port_t * output) +{ + static int do_init = 1; + uint32_t mm_accel; + + if (do_init) { + do_init = 0; + mm_accel = xine_mm_accel(); + mpeg2_cpu_state_init (mm_accel); + mpeg2_idct_init (mm_accel); + mpeg2_mc_init (mm_accel); + libmpeg2_accel_scan(&mpeg2dec->accel, mpeg2_scan_norm, mpeg2_scan_alt); + } + + if( !mpeg2dec->chunk_buffer ) + mpeg2dec->chunk_buffer = av_mallocz(BUFFER_SIZE + 4); + if( !mpeg2dec->picture ) + mpeg2dec->picture = av_mallocz(sizeof(picture_t)); + + mpeg2dec->shift = 0xffffff00; + mpeg2dec->new_sequence = 0; + mpeg2dec->is_sequence_needed = 1; + mpeg2dec->is_wait_for_ip_frames = 2; + mpeg2dec->frames_to_drop = 0; + mpeg2dec->drop_frame = 0; + mpeg2dec->in_slice = 0; + mpeg2dec->output = output; + mpeg2dec->chunk_ptr = mpeg2dec->chunk_buffer; + mpeg2dec->code = 0xb4; + mpeg2dec->seek_mode = 0; + + /* initialize AFD storage */ + mpeg2dec->afd_value_seen = XINE_VIDEO_AFD_NOT_PRESENT; + mpeg2dec->afd_value_reported = (XINE_VIDEO_AFD_NOT_PRESENT - 1); + + /* initialize substructures */ + mpeg2_header_state_init (mpeg2dec->picture); + + if ( output->get_capabilities(output) & VO_CAP_XXMC) { + printf("libmpeg2: output port has XxMC capability\n"); + mpeg2dec->frame_format = XINE_IMGFMT_XXMC; + } else if( output->get_capabilities(output) & VO_CAP_XVMC_MOCOMP) { + printf("libmpeg2: output port has XvMC capability\n"); + mpeg2dec->frame_format = XINE_IMGFMT_XVMC; + } else { + mpeg2dec->frame_format = XINE_IMGFMT_YV12; + } +} + +static inline void get_frame_duration (mpeg2dec_t * mpeg2dec, vo_frame_t *frame) +{ + static const double durations[] = { + 0, /* invalid */ + 3753.75, /* 23.976 fps */ + 3750, /* 24 fps */ + 3600, /* 25 fps */ + 3003, /* 29.97 fps */ + 3000, /* 30 fps */ + 1800, /* 50 fps */ + 1501.5, /* 59.94 fps */ + 1500, /* 60 fps */ + }; + double duration = ((unsigned) mpeg2dec->picture->frame_rate_code > 8u) + ? 0 : durations[mpeg2dec->picture->frame_rate_code]; + + duration = duration * (mpeg2dec->picture->frame_rate_ext_n + 1.0) / + (mpeg2dec->picture->frame_rate_ext_d + 1.0); + + /* this should be used to detect any special rff pattern */ + mpeg2dec->rff_pattern = mpeg2dec->rff_pattern << 1; + mpeg2dec->rff_pattern |= !!frame->repeat_first_field; + + if( ((mpeg2dec->rff_pattern & 0xff) == 0xaa || + (mpeg2dec->rff_pattern & 0xff) == 0x55) && + !mpeg2dec->picture->progressive_sequence ) { + /* special case for ntsc 3:2 pulldown */ + duration *= 5.0 / 4.0; + } + else + { + if( frame->repeat_first_field ) { + if( !mpeg2dec->picture->progressive_sequence && + frame->progressive_frame ) { + /* decoder should output 3 fields, so adjust duration to + count on this extra field time */ + duration *= 3.0 / 2.0; + } else if( mpeg2dec->picture->progressive_sequence ) { + /* for progressive sequences the output should repeat the + frame 1 or 2 times depending on top_field_first flag. */ + duration *= (frame->top_field_first) ? 3 : 2; + } + } + } + + frame->duration = (int) ceil (duration); + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, frame->duration); + /*printf("mpeg2dec: rff=%u\n",frame->repeat_first_field);*/ +} + +static double get_aspect_ratio(mpeg2dec_t *mpeg2dec) +{ + double ratio; + picture_t * picture = mpeg2dec->picture; + double mpeg1_pel_ratio[16] = {1.0 /* forbidden */, + 1.0, 0.6735, 0.7031, 0.7615, 0.8055, 0.8437, 0.8935, 0.9157, + 0.9815, 1.0255, 1.0695, 1.0950, 1.1575, 1.2015, 1.0 /*reserved*/ }; + + /* TODO: For slower machines the value of this function should be computed + * once and cached! + */ + + if( !picture->mpeg1 ) { + /* these hardcoded values are defined on mpeg2 standard for + * aspect ratio. other values are reserved or forbidden. */ + switch(picture->aspect_ratio_information) { + case 2: + ratio = 4.0/3.0; + break; + case 3: + ratio = 16.0/9.0; + break; + case 4: + ratio = 2.11/1.0; + break; + case 1: + default: + ratio = (double)picture->coded_picture_width/(double)picture->coded_picture_height; + break; + } + } else { + /* mpeg1 constants refer to pixel aspect ratio */ + ratio = (double)picture->coded_picture_width/(double)picture->coded_picture_height; + ratio /= mpeg1_pel_ratio[picture->aspect_ratio_information]; + } + + return ratio; +} + +static void remember_metainfo (mpeg2dec_t *mpeg2dec) { + + picture_t * picture = mpeg2dec->picture; + + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_WIDTH, picture->display_width); + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_HEIGHT, picture->display_height); + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_RATIO, + ((double)10000 * get_aspect_ratio(mpeg2dec))); + + switch (mpeg2dec->picture->frame_rate_code) { + case 1: /* 23.976 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 3913); + break; + case 2: /* 24 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 3750); + break; + case 3: /* 25 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 3600); + break; + case 4: /* 29.97 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 3003); + break; + case 5: /* 30 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 3000); + break; + case 6: /* 50 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 1800); + break; + case 7: /* 59.94 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 1525); + break; + case 8: /* 60 fps */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 1509); + break; + default: + /* printf ("invalid/unknown frame rate code : %d \n", + frame->frame_rate_code); */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_FRAME_DURATION, 3000); + } + + _x_meta_info_set_utf8(mpeg2dec->stream, XINE_META_INFO_VIDEOCODEC, "MPEG (libmpeg2)"); +} + +static inline int parse_chunk (mpeg2dec_t * mpeg2dec, int code, + uint8_t * buffer, int next_code) +{ + picture_t * picture; + int is_frame_done; + double ratio; + + /* wait for sequence_header_code */ + if (mpeg2dec->is_sequence_needed) { + if (code != 0xb3) { + /* printf ("libmpeg2: waiting for sequence header\n"); */ + mpeg2dec->pts = 0; + return 0; + } + } + if (mpeg2dec->is_frame_needed) { + /* printf ("libmpeg2: waiting for frame start\n"); */ + mpeg2dec->pts = 0; + if (mpeg2dec->picture->current_frame) + mpeg2dec->picture->current_frame->bad_frame = 1; + } + + mpeg2_stats (code, buffer); + + picture = mpeg2dec->picture; + is_frame_done = mpeg2dec->in_slice && ((!code) || (code >= 0xb0)); + + if (is_frame_done) + mpeg2dec->in_slice = 0; + + if (is_frame_done && picture->current_frame != NULL) { + + libmpeg2_accel_frame_completion(&mpeg2dec->accel, mpeg2dec->frame_format, + picture, code); + + if (((picture->picture_structure == FRAME_PICTURE) || + (picture->second_field)) ) { + + if (mpeg2dec->drop_frame) + picture->current_frame->bad_frame = 1; + + if (picture->picture_coding_type == B_TYPE) { + if( picture->current_frame && !picture->current_frame->drawn ) { + + /* hack against wrong mpeg1 pts */ + if (picture->mpeg1) + picture->current_frame->pts = 0; + + get_frame_duration(mpeg2dec, picture->current_frame); + mpeg2dec->frames_to_drop = picture->current_frame->draw (picture->current_frame, mpeg2dec->stream); + picture->current_frame->drawn = 1; + } + } else if (picture->forward_reference_frame && !picture->forward_reference_frame->drawn) { + get_frame_duration(mpeg2dec, picture->forward_reference_frame); + mpeg2dec->frames_to_drop = picture->forward_reference_frame->draw (picture->forward_reference_frame, + mpeg2dec->stream); + picture->forward_reference_frame->drawn = 1; + } + } + } + + switch (code) { + case 0x00: /* picture_start_code */ + if (mpeg2_header_picture (picture, buffer)) { + fprintf (stderr, "bad picture header\n"); + abort(); + } + + mpeg2dec->is_frame_needed=0; + + if (!picture->second_field) { + /* find out if we want to skip this frame */ + mpeg2dec->drop_frame = 0; + + /* picture->skip_non_intra_dct = (mpeg2dec->frames_to_drop>0) ; */ + + switch (picture->picture_coding_type) { + case B_TYPE: + + lprintf ("B-Frame\n"); + + if (mpeg2dec->frames_to_drop>1) { + lprintf ("dropping b-frame because frames_to_drop==%d\n", + mpeg2dec->frames_to_drop); + mpeg2dec->drop_frame = 1; + } else if (!picture->forward_reference_frame || picture->forward_reference_frame->bad_frame + || !picture->backward_reference_frame || picture->backward_reference_frame->bad_frame) { +#ifdef LOG + printf ("libmpeg2: dropping b-frame because ref is bad ("); + if (picture->forward_reference_frame) + printf ("fw ref frame %d, bad %d;", picture->forward_reference_frame->id, + picture->forward_reference_frame->bad_frame); + else + printf ("fw ref frame not there;"); + if (picture->backward_reference_frame) + printf ("bw ref frame %d, bad %d)\n", picture->backward_reference_frame->id, + picture->backward_reference_frame->bad_frame); + else + printf ("fw ref frame not there)\n"); +#endif + mpeg2dec->drop_frame = 1; + } else if (mpeg2dec->is_wait_for_ip_frames > 0) { + lprintf("dropping b-frame because refs are invalid\n"); + mpeg2dec->drop_frame = 1; + } + break; + + case P_TYPE: + + lprintf ("P-Frame\n"); + + if (mpeg2dec->frames_to_drop>2) { + mpeg2dec->drop_frame = 1; + lprintf ("dropping p-frame because frames_to_drop==%d\n", + mpeg2dec->frames_to_drop); + } else if (!picture->backward_reference_frame || picture->backward_reference_frame->bad_frame) { + mpeg2dec->drop_frame = 1; +#ifdef LOG + if (!picture->backward_reference_frame) + printf ("libmpeg2: dropping p-frame because no ref frame\n"); + else + printf ("libmpeg2: dropping p-frame because ref %d is bad\n", picture->backward_reference_frame->id); +#endif + } else if (mpeg2dec->is_wait_for_ip_frames > 1) { + lprintf("dropping p-frame because ref is invalid\n"); + mpeg2dec->drop_frame = 1; + } else if (mpeg2dec->is_wait_for_ip_frames) + mpeg2dec->is_wait_for_ip_frames--; + + break; + + case I_TYPE: + lprintf ("I-Frame\n"); + /* for the sake of dvd menus, never drop i-frames + if (mpeg2dec->frames_to_drop>4) { + mpeg2dec->drop_frame = 1; + } + */ + + if (mpeg2dec->is_wait_for_ip_frames) + mpeg2dec->is_wait_for_ip_frames--; + + break; + } + } + + break; + + case 0xb2: /* user data code */ + process_userdata(mpeg2dec, buffer); + break; + + case 0xb3: /* sequence_header_code */ + if (mpeg2_header_sequence (picture, buffer)) { + fprintf (stderr, "bad sequence header\n"); + /* abort(); */ + break; + } + + /* reset AFD value to detect absence */ + mpeg2dec->afd_value_seen = XINE_VIDEO_AFD_NOT_PRESENT; + + /* according to ISO/IEC 13818-2, an extension start code will follow. + * Otherwise the stream follows ISO/IEC 11172-2 which means MPEG1 */ + picture->mpeg1 = (next_code != 0xb5); + + if (mpeg2dec->force_aspect) picture->aspect_ratio_information = mpeg2dec->force_aspect; + + if (mpeg2dec->is_sequence_needed ) { + mpeg2dec->new_sequence = 1; + } + + if (mpeg2dec->is_sequence_needed + || (picture->aspect_ratio_information != picture->saved_aspect_ratio) + || (picture->frame_width != picture->coded_picture_width) + || (picture->frame_height != picture->coded_picture_height)) { + xine_event_t event; + xine_format_change_data_t data; + + remember_metainfo (mpeg2dec); + event.type = XINE_EVENT_FRAME_FORMAT_CHANGE; + event.stream = mpeg2dec->stream; + event.data = &data; + event.data_length = sizeof(data); + data.width = picture->coded_picture_width; + data.height = picture->coded_picture_height; + data.aspect = picture->aspect_ratio_information; + data.pan_scan = mpeg2dec->force_pan_scan; + xine_event_send(mpeg2dec->stream, &event); + + _x_stream_info_set(mpeg2dec->stream,XINE_STREAM_INFO_VIDEO_WIDTH, + picture->display_width); + _x_stream_info_set(mpeg2dec->stream,XINE_STREAM_INFO_VIDEO_HEIGHT, + picture->display_height); + + if (picture->forward_reference_frame && + picture->forward_reference_frame != picture->current_frame && + picture->forward_reference_frame != picture->backward_reference_frame) + picture->forward_reference_frame->free (picture->forward_reference_frame); + + if (picture->backward_reference_frame && + picture->backward_reference_frame != picture->current_frame) + picture->backward_reference_frame->free (picture->backward_reference_frame); + + mpeg2dec->is_sequence_needed = 0; + picture->forward_reference_frame = NULL; + picture->backward_reference_frame = NULL; + + picture->frame_width = picture->coded_picture_width; + picture->frame_height = picture->coded_picture_height; + picture->saved_aspect_ratio = picture->aspect_ratio_information; + } + break; + + case 0xb5: /* extension_start_code */ + if (mpeg2_header_extension (picture, buffer)) { + fprintf (stderr, "bad extension\n"); + abort(); + } + break; + + case 0xb7: /* sequence end code */ + mpeg2_flush(mpeg2dec); + mpeg2dec->is_sequence_needed = 1; + break; + + case 0xb8: /* group of pictures start code */ + if (mpeg2_header_group_of_pictures (picture, buffer)) { + printf ("libmpeg2: bad group of pictures\n"); + abort(); + } + default: + if ((code >= 0xb9) && (code != 0xe4)) { + printf("Not multiplexed? 0x%x\n",code); + } + if (code >= 0xb0) + break; + + /* check for AFD change once per picture */ + if (mpeg2dec->afd_value_reported != mpeg2dec->afd_value_seen) { + /* AFD data should better be stored in current_frame to have it */ + /* ready and synchronous with other data like width or height. */ + /* An AFD change should then be detected when a new frame is emitted */ + /* from the decoder to report the AFD change in display order and not */ + /* in decoding order like it happens below for now. */ + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_AFD, mpeg2dec->afd_value_seen); + lprintf ("AFD changed from %d to %d\n", mpeg2dec->afd_value_reported, mpeg2dec->afd_value_seen); + mpeg2dec->afd_value_reported = mpeg2dec->afd_value_seen; + } + + if (!(mpeg2dec->in_slice)) { + mpeg2dec->in_slice = 1; + + if (picture->second_field) { + if (picture->current_frame) + picture->current_frame->field(picture->current_frame, + picture->picture_structure); + else + mpeg2dec->drop_frame = 1; + } else { + int flags = picture->picture_structure; + + if (!picture->mpeg1) flags |= VO_INTERLACED_FLAG; + if (mpeg2dec->force_pan_scan) flags |= VO_PAN_SCAN_FLAG; + if (mpeg2dec->new_sequence) flags |= VO_NEW_SEQUENCE_FLAG; + + if ( picture->current_frame && + picture->current_frame != picture->backward_reference_frame && + picture->current_frame != picture->forward_reference_frame ) { + picture->current_frame->free (picture->current_frame); + } + if (picture->picture_coding_type == B_TYPE) { + ratio = get_aspect_ratio(mpeg2dec); + picture->current_frame = + mpeg2dec->stream->video_out->get_frame (mpeg2dec->stream->video_out, + picture->coded_picture_width, + picture->coded_picture_height, + ratio, + mpeg2dec->frame_format, + flags); + libmpeg2_accel_new_frame( &mpeg2dec->accel, mpeg2dec->frame_format, + picture, ratio, flags); + } else { + ratio = get_aspect_ratio(mpeg2dec); + picture->current_frame = + mpeg2dec->stream->video_out->get_frame (mpeg2dec->stream->video_out, + picture->coded_picture_width, + picture->coded_picture_height, + ratio, + mpeg2dec->frame_format, + flags); + + libmpeg2_accel_new_frame( &mpeg2dec->accel, mpeg2dec->frame_format, + picture, ratio, flags); + + if (picture->forward_reference_frame && + picture->forward_reference_frame != picture->backward_reference_frame) + picture->forward_reference_frame->free (picture->forward_reference_frame); + + picture->forward_reference_frame = + picture->backward_reference_frame; + picture->backward_reference_frame = picture->current_frame; + } + + if(mpeg2dec->new_sequence) + mpeg2dec->new_sequence = + libmpeg2_accel_new_sequence(&mpeg2dec->accel, mpeg2dec->frame_format, + picture); + + picture->current_frame->bad_frame = 1; + picture->current_frame->drawn = 0; + picture->current_frame->pts = mpeg2dec->pts; + picture->current_frame->top_field_first = picture->top_field_first; + picture->current_frame->repeat_first_field = picture->repeat_first_field; + picture->current_frame->progressive_frame = picture->progressive_frame; + picture->current_frame->crop_right = picture->coded_picture_width - picture->display_width; + picture->current_frame->crop_bottom = picture->coded_picture_height - picture->display_height; + + switch( picture->picture_coding_type ) { + case I_TYPE: + picture->current_frame->picture_coding_type = XINE_PICT_I_TYPE; + break; + case P_TYPE: + picture->current_frame->picture_coding_type = XINE_PICT_P_TYPE; + break; + case B_TYPE: + picture->current_frame->picture_coding_type = XINE_PICT_B_TYPE; + break; + case D_TYPE: + picture->current_frame->picture_coding_type = XINE_PICT_D_TYPE; + break; + } + + lprintf ("decoding frame %d, type %s\n", + picture->current_frame->id, picture->picture_coding_type == I_TYPE ? "I" : + picture->picture_coding_type == P_TYPE ? "P" : "B"); + mpeg2dec->pts = 0; + /*printf("Starting to decode frame %d\n",picture->current_frame->id);*/ + } + } + + if (!mpeg2dec->drop_frame && picture->current_frame != NULL) { +#ifdef DEBUG_LOG + printf("slice target %08x past %08x future %08x\n",picture->current_frame,picture->forward_reference_frame,picture->backward_reference_frame); + fflush(stdout); +#endif + libmpeg2_accel_slice(&mpeg2dec->accel, picture, code, buffer, mpeg2dec->chunk_size, + mpeg2dec->chunk_buffer); + + if( picture->v_offset > picture->limit_y || + picture->v_offset + 16 > picture->display_height ) { + picture->current_frame->bad_frame = 0; + } + } + } + + /* printf ("libmpeg2: parse_chunk %d completed\n", code); */ + return is_frame_done; +} + +static inline int find_start_code (mpeg2dec_t * mpeg2dec, + uint8_t ** current, uint8_t * limit) +{ + uint8_t * p; + + if (*current >= limit) + return 0; + if (mpeg2dec->shift == 0x00000100) + return 1; + + mpeg2dec->shift = (mpeg2dec->shift | *(*current)++) << 8; + + if (*current >= limit) + return 0; + if (mpeg2dec->shift == 0x00000100) + return 1; + + mpeg2dec->shift = (mpeg2dec->shift | *(*current)++) << 8; + + if (*current >= limit) + return 0; + if (mpeg2dec->shift == 0x00000100) + return 1; + + limit--; + + if (*current >= limit) { + mpeg2dec->shift = (mpeg2dec->shift | *(*current)++) << 8; + return 0; + } + + p = *current; + + while (p < limit && (p = (uint8_t *)memchr(p, 0x01, limit - p))) { + if (p[-2] || p[-1]) + p += 3; + else { + *current = ++p; + return 1; + } + } + + *current = ++limit; + p = limit - 3; + mpeg2dec->shift = (mpeg2dec->shift | *p++) << 8; + mpeg2dec->shift = (mpeg2dec->shift | *p++) << 8; + mpeg2dec->shift = (mpeg2dec->shift | *p++) << 8; + + return 0; +} + +static inline uint8_t * copy_chunk (mpeg2dec_t * mpeg2dec, + uint8_t * current, uint8_t * end) +{ + uint8_t * limit; + uint8_t * data = current; + int found, bite; + + /* sequence end code 0xb7 doesn't have any data and there might be the case + * that no start code will follow this code for quite some time (e. g. in case + * of a still image. + * Therefore, return immediately with a chunk_size of 0. Setting code to 0xb4 + * will eat up any trailing garbage next time. + */ + if (mpeg2dec->code == 0xb7) { + mpeg2dec->code = 0xb4; + mpeg2dec->chunk_size = 0; + return current; + } + + limit = current + (mpeg2dec->chunk_buffer + BUFFER_SIZE - mpeg2dec->chunk_ptr); + if (limit > end) + limit = end; + + found = find_start_code(mpeg2dec, ¤t, limit); + bite = current - data; + if (bite) { + xine_fast_memcpy(mpeg2dec->chunk_ptr, data, bite); + mpeg2dec->chunk_ptr += bite; + } + + if (found) { + mpeg2dec->code = *current++; + mpeg2dec->chunk_size = mpeg2dec->chunk_ptr - mpeg2dec->chunk_buffer - 3; + mpeg2dec->chunk_ptr = mpeg2dec->chunk_buffer; + mpeg2dec->shift = 0xffffff00; + return current; + } + + if (current == end) + return NULL; + + /* we filled the chunk buffer without finding a start code */ + mpeg2dec->code = 0xb4; /* sequence_error_code */ + mpeg2dec->chunk_ptr = mpeg2dec->chunk_buffer; + return current; +} + +int mpeg2_decode_data (mpeg2dec_t * mpeg2dec, uint8_t * current, uint8_t * end, + uint64_t pts) +{ + int ret; + uint8_t code; + + ret = 0; + if (mpeg2dec->seek_mode) { + mpeg2dec->chunk_ptr = mpeg2dec->chunk_buffer; + mpeg2dec->code = 0xb4; + mpeg2dec->seek_mode = 0; + mpeg2dec->shift = 0xffffff00; + mpeg2dec->is_frame_needed = 1; + } + + if (pts) + mpeg2dec->pts = pts; + + while (current != end || mpeg2dec->code == 0xb7) { + code = mpeg2dec->code; + current = copy_chunk (mpeg2dec, current, end); + if (current == NULL) + break; + ret += parse_chunk (mpeg2dec, code, mpeg2dec->chunk_buffer, mpeg2dec->code); + } + + libmpeg2_accel_frame_completion(&mpeg2dec->accel, mpeg2dec->frame_format, + mpeg2dec->picture, 0xff); + + return ret; +} + +void mpeg2_discontinuity (mpeg2dec_t * mpeg2dec) { + picture_t *picture = mpeg2dec->picture; + + if( !picture ) + return; + + mpeg2dec->in_slice = 0; + mpeg2dec->pts = 0; + if ( picture->current_frame ) + picture->current_frame->pts = 0; + if ( picture->forward_reference_frame ) + picture->forward_reference_frame->pts = 0; + if ( picture->backward_reference_frame ) + picture->backward_reference_frame->pts = 0; + + libmpeg2_accel_discontinuity(&mpeg2dec->accel, mpeg2dec->frame_format, picture); +} + +void mpeg2_reset (mpeg2dec_t * mpeg2dec) { + + picture_t *picture = mpeg2dec->picture; + + if( !picture ) + return; + + mpeg2_discontinuity(mpeg2dec); + + if( !picture->mpeg1 ) { + mpeg2dec->is_wait_for_ip_frames = 2; + + /* mark current frames as bad so they won't make to screen */ + if ( picture->current_frame ) + picture->current_frame->bad_frame=1; + if (picture->forward_reference_frame ) + picture->forward_reference_frame->bad_frame=1; + if (picture->backward_reference_frame) + picture->backward_reference_frame->bad_frame=1; + + } else { + /* to free reference frames one also needs to fix slice.c to + * abort when they are NULL. unfortunately it seems to break + * DVD menus. + * + * ...so let's do this for mpeg-1 only :) + */ + if ( picture->current_frame && + picture->current_frame != picture->backward_reference_frame && + picture->current_frame != picture->forward_reference_frame ) + picture->current_frame->free (picture->current_frame); + picture->current_frame = NULL; + + if (picture->forward_reference_frame && + picture->forward_reference_frame != picture->backward_reference_frame) + picture->forward_reference_frame->free (picture->forward_reference_frame); + picture->forward_reference_frame = NULL; + + if (picture->backward_reference_frame) + picture->backward_reference_frame->free (picture->backward_reference_frame); + picture->backward_reference_frame = NULL; + } + + mpeg2dec->in_slice = 0; + mpeg2dec->seek_mode = 1; + +} + +void mpeg2_flush (mpeg2dec_t * mpeg2dec) { + + picture_t *picture = mpeg2dec->picture; + + if (!picture) + return; + + if (picture->current_frame && !picture->current_frame->drawn && + !picture->current_frame->bad_frame) { + + lprintf ("blasting out current frame %d on flush\n", + picture->current_frame->id); + + picture->current_frame->drawn = 1; + get_frame_duration(mpeg2dec, picture->current_frame); + + picture->current_frame->pts = 0; + picture->current_frame->draw(picture->current_frame, mpeg2dec->stream); + } + +} + +void mpeg2_close (mpeg2dec_t * mpeg2dec) +{ + picture_t *picture = mpeg2dec->picture; + + /* + { + static uint8_t finalizer[] = {0,0,1,0xb4}; + mpeg2_decode_data (mpeg2dec, finalizer, finalizer+4, 0); + } + */ + + /* + dont remove any picture->*->free() below. doing so will cause buffer + leak, and we only have about 15 of them. + */ + + if ( picture->current_frame ) { + if( !picture->current_frame->drawn ) { + lprintf ("blasting out current frame on close\n"); + picture->current_frame->pts = 0; + get_frame_duration(mpeg2dec, picture->current_frame); + picture->current_frame->draw (picture->current_frame, mpeg2dec->stream); + picture->current_frame->drawn = 1; + } + + if( picture->current_frame != picture->backward_reference_frame && + picture->current_frame != picture->forward_reference_frame ) { + picture->current_frame->free (picture->current_frame); + } + picture->current_frame = NULL; + } + + if (picture->forward_reference_frame && + picture->forward_reference_frame != picture->backward_reference_frame) { + picture->forward_reference_frame->free (picture->forward_reference_frame); + picture->forward_reference_frame = NULL; + } + + if (picture->backward_reference_frame) { + if( !picture->backward_reference_frame->drawn) { + lprintf ("blasting out backward reference frame on close\n"); + picture->backward_reference_frame->pts = 0; + get_frame_duration(mpeg2dec, picture->backward_reference_frame); + picture->backward_reference_frame->draw (picture->backward_reference_frame, mpeg2dec->stream); + picture->backward_reference_frame->drawn = 1; + } + picture->backward_reference_frame->free (picture->backward_reference_frame); + picture->backward_reference_frame = NULL; + } + + av_freep(&mpeg2dec->chunk_buffer); + av_freep(&mpeg2dec->picture_base); + + if ( mpeg2dec->cc_dec) { + /* dispose the closed caption decoder */ + mpeg2dec->cc_dec->dispose(mpeg2dec->cc_dec); + mpeg2dec->cc_dec = NULL; + } +} + +void mpeg2_find_sequence_header (mpeg2dec_t * mpeg2dec, + uint8_t * current, uint8_t * end){ + + uint8_t code, next_code; + picture_t *picture = mpeg2dec->picture; + + mpeg2dec->seek_mode = 1; + + while (current != end) { + code = mpeg2dec->code; + current = copy_chunk (mpeg2dec, current, end); + if (current == NULL) + return ; + next_code = mpeg2dec->code; + + /* printf ("looking for sequence header... %02x\n", code); */ + + mpeg2_stats (code, mpeg2dec->chunk_buffer); + + if (code == 0xb3) { /* sequence_header_code */ + if (mpeg2_header_sequence (picture, mpeg2dec->chunk_buffer)) { + printf ("libmpeg2: bad sequence header\n"); + continue; + } + + /* according to ISO/IEC 13818-2, an extension start code will follow. + * Otherwise the stream follows ISO/IEC 11172-2 which means MPEG1 */ + picture->mpeg1 = (next_code != 0xb5); + + if (mpeg2dec->force_aspect) picture->aspect_ratio_information = mpeg2dec->force_aspect; + + if (mpeg2dec->is_sequence_needed) { + xine_event_t event; + xine_format_change_data_t data; + + mpeg2dec->new_sequence = 1; + + mpeg2dec->is_sequence_needed = 0; + picture->frame_width = picture->coded_picture_width; + picture->frame_height = picture->coded_picture_height; + + remember_metainfo (mpeg2dec); + + event.type = XINE_EVENT_FRAME_FORMAT_CHANGE; + event.stream = mpeg2dec->stream; + event.data = &data; + event.data_length = sizeof(data); + data.width = picture->coded_picture_width; + data.height = picture->coded_picture_height; + data.aspect = picture->aspect_ratio_information; + data.pan_scan = mpeg2dec->force_pan_scan; + xine_event_send(mpeg2dec->stream, &event); + + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_WIDTH, + picture->display_width); + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_HEIGHT, + picture->display_height); + } + } else if (code == 0xb5) { /* extension_start_code */ + if (mpeg2_header_extension (picture, mpeg2dec->chunk_buffer)) { + printf ("libmpeg2: bad extension\n"); + continue ; + } + } + } +} + +/* Find the end of the userdata field in an MPEG-2 stream */ +static uint8_t *find_end(uint8_t *buffer) +{ + uint8_t *current = buffer; + while(1) { + if (current[0] == 0 && current[1] == 0 && current[2] == 1) + break; + current++; + } + return current; +} + +static void process_userdata(mpeg2dec_t *mpeg2dec, uint8_t *buffer) +{ + /* check if user data denotes closed captions */ + if (buffer[0] == 'C' && buffer[1] == 'C') { + + if (!mpeg2dec->cc_dec) { + xine_event_t event; + xine_format_change_data_t data; + + /* open the closed caption decoder first */ + mpeg2dec->cc_dec = _x_get_spu_decoder(mpeg2dec->stream, (BUF_SPU_CC >> 16) & 0xff); + + /* send a frame format event so that the CC decoder knows the initial image size */ + event.type = XINE_EVENT_FRAME_FORMAT_CHANGE; + event.stream = mpeg2dec->stream; + event.data = &data; + event.data_length = sizeof(data); + data.width = mpeg2dec->picture->coded_picture_width; + data.height = mpeg2dec->picture->coded_picture_height; + data.aspect = mpeg2dec->picture->aspect_ratio_information; + data.pan_scan = mpeg2dec->force_pan_scan; + xine_event_send(mpeg2dec->stream, &event); + + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_WIDTH, + mpeg2dec->picture->display_width); + _x_stream_info_set(mpeg2dec->stream, XINE_STREAM_INFO_VIDEO_HEIGHT, + mpeg2dec->picture->display_height); + } + + if (mpeg2dec->cc_dec) { + buf_element_t buf; + + buf.type = BUF_SPU_CC; + buf.content = &buffer[2]; + buf.pts = mpeg2dec->pts; + buf.size = find_end(buffer) - &buffer[2]; + buf.decoder_flags = 0; + + mpeg2dec->cc_dec->decode_data(mpeg2dec->cc_dec, &buf); + } + } + /* check Active Format Description ETSI TS 101 154 V1.5.1 */ + else if (buffer[0] == 0x44 && buffer[1] == 0x54 && buffer[2] == 0x47 && buffer[3] == 0x31) + mpeg2dec->afd_value_seen = (buffer[4] & 0x40) ? (buffer[5] & 0x0f) : XINE_VIDEO_AFD_NOT_PRESENT; +} diff --git a/src/video_dec/libmpeg2/header.c b/src/video_dec/libmpeg2/header.c new file mode 100644 index 000000000..0c2b76891 --- /dev/null +++ b/src/video_dec/libmpeg2/header.c @@ -0,0 +1,419 @@ +/* + * header.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +/* +#define LOG_PAN_SCAN +*/ + +#include "config.h" + +#include <stdio.h> /* For printf debugging */ +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include <xine/attributes.h> + +/* default intra quant matrix, in zig-zag order */ +static const uint8_t default_intra_quantizer_matrix[64] ATTR_ALIGN(16) = { + 8, + 16, 16, + 19, 16, 19, + 22, 22, 22, 22, + 22, 22, 26, 24, 26, + 27, 27, 27, 26, 26, 26, + 26, 27, 27, 27, 29, 29, 29, + 34, 34, 34, 29, 29, 29, 27, 27, + 29, 29, 32, 32, 34, 34, 37, + 38, 37, 35, 35, 34, 35, + 38, 38, 40, 40, 40, + 48, 48, 46, 46, + 56, 56, 58, + 69, 69, + 83 +}; + +uint8_t mpeg2_scan_norm[64] ATTR_ALIGN(16) = +{ + /* Zig-Zag scan pattern */ + 0, 1, 8,16, 9, 2, 3,10, + 17,24,32,25,18,11, 4, 5, + 12,19,26,33,40,48,41,34, + 27,20,13, 6, 7,14,21,28, + 35,42,49,56,57,50,43,36, + 29,22,15,23,30,37,44,51, + 58,59,52,45,38,31,39,46, + 53,60,61,54,47,55,62,63 +}; + +uint8_t mpeg2_scan_alt[64] ATTR_ALIGN(16) = +{ + /* Alternate scan pattern */ + 0,8,16,24,1,9,2,10,17,25,32,40,48,56,57,49, + 41,33,26,18,3,11,4,12,19,27,34,42,50,58,35,43, + 51,59,20,28,5,13,6,14,21,29,36,44,52,60,37,45, + 53,61,22,30,7,15,23,31,38,46,54,62,39,47,55,63 +}; + +/* count must be between 1 and 32 */ +static uint32_t get_bits(uint8_t *buffer, uint32_t count, uint32_t *bit_position) { + uint32_t byte_offset; + uint32_t bit_offset; + uint32_t bit_mask; + uint32_t bit_bite; + uint32_t result=0; + if (count == 0) return 0; + do { + byte_offset = *bit_position >> 3; /* Div 8 */ + bit_offset = 8 - (*bit_position & 0x7); /* Bits got 87654321 */ + bit_mask = ((1 << (bit_offset)) - 1); + bit_bite = bit_offset; + if (count < bit_offset) { + bit_mask ^= ((1 << (bit_offset-count)) - 1); + bit_bite = count; + } + /* + printf("Byte=0x%02x Bitmask=0x%04x byte_offset=%u bit_offset=%u bit_byte=%u count=%u\n",buffer[byte_offset], bit_mask, byte_offset, bit_offset, bit_bite,count); + */ + result = (result << bit_bite) | ((buffer[byte_offset] & bit_mask) >> (bit_offset-bit_bite)); + *bit_position+=bit_bite; + count-=bit_bite; + } while ((count > 0) && (byte_offset<50) ); + return result; +} + +static int32_t get_bits_signed(uint8_t *buffer, uint32_t count, uint32_t *bit_position) { + uint32_t value = get_bits(buffer, count, bit_position); + uint32_t sign_mask = (uint32_t)(-1 << (count - 1)); + if (value & sign_mask) + value |= sign_mask; /* sign-extend value */ + return (int32_t)value; +} + +void mpeg2_header_state_init (picture_t * picture) +{ + picture->scan = mpeg2_scan_norm; + picture->load_intra_quantizer_matrix = 1; + picture->load_non_intra_quantizer_matrix = 1; +} + +int mpeg2_header_sequence (picture_t * picture, uint8_t * buffer) +{ + int width, height; + int i; + + if ((buffer[6] & 0x20) != 0x20) + return 1; /* missing marker_bit */ + + height = (buffer[0] << 16) | (buffer[1] << 8) | buffer[2]; + + picture->display_width = width = (height >> 12); + picture->display_height = height = (height & 0xfff); + + width = (width + 15) & ~15; + height = (height + 15) & ~15; + + if ((width > 1920) || (height > 1152)) + return 1; /* size restrictions for MP@HL */ + + picture->coded_picture_width = width; + picture->coded_picture_height = height; + + /* this is not used by the decoder */ + picture->aspect_ratio_information = buffer[3] >> 4; + picture->frame_rate_code = buffer[3] & 15; + picture->bitrate = (buffer[4]<<10)|(buffer[5]<<2)|(buffer[6]>>6); + + if (buffer[7] & 2) { + for (i = 0; i < 64; i++) + picture->intra_quantizer_matrix[mpeg2_scan_norm[i]] = + (buffer[i+7] << 7) | (buffer[i+8] >> 1); + buffer += 64; + } else + for (i = 0; i < 64; i++) + picture->intra_quantizer_matrix[mpeg2_scan_norm[i]] = + default_intra_quantizer_matrix [i]; + + if (buffer[7] & 1) + for (i = 0; i < 64; i++) + picture->non_intra_quantizer_matrix[mpeg2_scan_norm[i]] = + buffer[i+8]; + else + for (i = 0; i < 64; i++) + picture->non_intra_quantizer_matrix[i] = 16; + picture->load_intra_quantizer_matrix = 1; + picture->load_non_intra_quantizer_matrix = 1; + /* MPEG1 - for testing only */ + picture->mpeg1 = 1; + picture->intra_dc_precision = 0; + picture->frame_pred_frame_dct = 1; + picture->q_scale_type = 0; + picture->concealment_motion_vectors = 0; + /* picture->alternate_scan = 0; */ + picture->picture_structure = FRAME_PICTURE; + /* picture->second_field = 0; */ + + return 0; +} + +static int sequence_extension (picture_t * picture, uint8_t * buffer) +{ + /* check chroma format, size extensions, marker bit */ + if (((buffer[1] & 0x07) != 0x02) || (buffer[2] & 0xe0) || + ((buffer[3] & 0x01) != 0x01)) + return 1; + + /* this is not used by the decoder */ + picture->progressive_sequence = (buffer[1] >> 3) & 1; + + picture->low_delay = buffer[5] & 0x80; + + if (!picture->progressive_sequence) + picture->coded_picture_height = + (picture->coded_picture_height + 31) & ~31; + + + /* printf ("libmpeg2: low_delay : %d\n", picture->low_delay); */ + +/* + printf ("libmpeg2: sequence extension+5 : %08x (%d)\n", + buffer[5], buffer[5] % 0x80); + */ + + picture->frame_rate_ext_n = buffer[5] & 0x31; + picture->frame_rate_ext_d = (buffer[5] >> 2) & 0x03; + + /* MPEG1 - for testing only */ + picture->mpeg1 = 0; + + return 0; +} + +static int quant_matrix_extension (picture_t * picture, uint8_t * buffer) +{ + int i; + + if (buffer[0] & 8) { + for (i = 0; i < 64; i++) + picture->intra_quantizer_matrix[mpeg2_scan_norm[i]] = + (buffer[i] << 5) | (buffer[i+1] >> 3); + buffer += 64; + } + + if (buffer[0] & 4) + for (i = 0; i < 64; i++) + picture->non_intra_quantizer_matrix[mpeg2_scan_norm[i]] = + (buffer[i] << 6) | (buffer[i+1] >> 2); + + return 0; +} + +static int picture_coding_extension (picture_t * picture, uint8_t * buffer) +{ + /* pre subtract 1 for use later in compute_motion_vector */ + picture->f_motion.f_code[0] = (buffer[0] & 15) - 1; + picture->f_motion.f_code[1] = (buffer[1] >> 4) - 1; + picture->b_motion.f_code[0] = (buffer[1] & 15) - 1; + picture->b_motion.f_code[1] = (buffer[2] >> 4) - 1; + + picture->intra_dc_precision = (buffer[2] >> 2) & 3; + picture->picture_structure = buffer[2] & 3; + picture->frame_pred_frame_dct = (buffer[3] >> 6) & 1; + picture->concealment_motion_vectors = (buffer[3] >> 5) & 1; + picture->q_scale_type = (buffer[3] >> 4) & 1; + picture->intra_vlc_format = (buffer[3] >> 3) & 1; + + if (buffer[3] & 4) /* alternate_scan */ + picture->scan = mpeg2_scan_alt; + else + picture->scan = mpeg2_scan_norm; + + /* these are not used by the decoder */ + picture->top_field_first = buffer[3] >> 7; + picture->repeat_first_field = (buffer[3] >> 1) & 1; + picture->progressive_frame = buffer[4] >> 7; + + return 0; +} + +static int sequence_display_extension (picture_t * picture, uint8_t * buffer) { + /* FIXME: implement. */ + uint32_t bit_position; + uint32_t padding; + + bit_position = 0; + padding = get_bits(buffer, 4, &bit_position); + picture->video_format = get_bits(buffer, 3, &bit_position); + picture->colour_description = get_bits(buffer, 1, &bit_position); + if(picture->colour_description) { + picture->colour_primatives = get_bits(buffer, 8, &bit_position); + picture->transfer_characteristics = get_bits(buffer, 8, &bit_position); + picture->matrix_coefficients = get_bits(buffer, 8, &bit_position); + } + picture->display_horizontal_size = get_bits(buffer, 14, &bit_position); + padding = get_bits(buffer, 1, &bit_position); + picture->display_vertical_size = get_bits(buffer, 14, &bit_position); + +#ifdef LOG_PAN_SCAN + printf("Sequence_display_extension\n"); + printf(" video_format: %u\n", picture->video_format); + printf(" colour_description: %u\n", picture->colour_description); + if(picture->colour_description) { + printf(" colour_primatives: %u\n", picture->colour_primatives); + printf(" transfer_characteristics %u\n", picture->transfer_characteristics); + printf(" matrix_coefficients %u\n", picture->matrix_coefficients); + } + printf(" display_horizontal_size %u\n", picture->display_horizontal_size); + printf(" display_vertical_size %u\n", picture->display_vertical_size); +#endif + + return 0; +} + +static int picture_display_extension (picture_t * picture, uint8_t * buffer) { + uint32_t bit_position; + uint32_t padding; + +#ifdef LOG_PAN_SCAN + printf ("libmpeg2: picture_display_extension\n"); +#endif + + bit_position = 0; + padding = get_bits(buffer, 4, &bit_position); + picture->frame_centre_horizontal_offset = get_bits_signed(buffer, 16, &bit_position); + padding = get_bits(buffer, 1, &bit_position); + picture->frame_centre_vertical_offset = get_bits_signed(buffer, 16, &bit_position); + padding = get_bits(buffer, 1, &bit_position); + +#ifdef LOG_PAN_SCAN + printf("Pan & Scan centre (x,y) = (%d, %d)\n", + picture->frame_centre_horizontal_offset, + picture->frame_centre_vertical_offset); +#endif + + return 0; +} + +int mpeg2_header_extension (picture_t * picture, uint8_t * buffer) +{ + switch (buffer[0] & 0xf0) { + case 0x00: /* reserved */ + return 0; + + case 0x10: /* sequence extension */ + return sequence_extension (picture, buffer); + + case 0x20: /* sequence display extension for Pan & Scan */ + return sequence_display_extension (picture, buffer); + + case 0x30: /* quant matrix extension */ + return quant_matrix_extension (picture, buffer); + + case 0x40: /* copyright extension */ + return 0; + + case 0x50: /* sequence scalable extension */ + return 0; + + case 0x60: /* reserved */ + return 0; + + case 0x70: /* picture display extension for Pan & Scan */ + return picture_display_extension (picture, buffer); + + case 0x80: /* picture coding extension */ + return picture_coding_extension (picture, buffer); + + case 0x90: /* picture spacial scalable extension */ + return 0; + + case 0xA0: /* picture temporal scalable extension */ + return 0; + + case 0xB0: /* camera parameters extension */ + return 0; + + case 0xC0: /* ITU-T extension */ + return 0; + + case 0xD0: /* reserved */ + return 0; + + case 0xE0: /* reserved */ + return 0; + + case 0xF0: /* reserved */ + return 0; + } + + return 0; +} + +int mpeg2_header_group_of_pictures (picture_t * picture, uint8_t * buffer) { + uint32_t bit_position; + uint32_t padding; + bit_position = 0; + + picture->drop_frame_flag = get_bits(buffer, 1, &bit_position); + picture->time_code_hours = get_bits(buffer, 5, &bit_position); + picture->time_code_minutes = get_bits(buffer, 6, &bit_position); + padding = get_bits(buffer, 1, &bit_position); + picture->time_code_seconds = get_bits(buffer, 6, &bit_position); + picture->time_code_pictures = get_bits(buffer, 6, &bit_position); + picture->closed_gop = get_bits(buffer, 1, &bit_position); + picture->broken_link = get_bits(buffer, 1, &bit_position); + +#ifdef LOG_PAN_SCAN + printf("Group of pictures\n"); + printf(" drop_frame_flag: %u\n", picture->drop_frame_flag); + printf(" time_code: HH:MM:SS:Pictures %02u:%02u:%02u:%02u\n", + picture->time_code_hours, + picture->time_code_minutes, + picture->time_code_seconds, + picture->time_code_pictures); + printf(" closed_gop: %u\n", picture->closed_gop); + printf(" bloken_link: %u\n", picture->broken_link); +#endif + + return 0; +} + +int mpeg2_header_picture (picture_t * picture, uint8_t * buffer) +{ + picture->picture_coding_type = (buffer [1] >> 3) & 7; + picture->vbv_delay = ((buffer[1] << 13) | (buffer[2] << 5) | + (buffer[3] >> 3)) & 0xffff; + + /* forward_f_code and backward_f_code - used in mpeg1 only */ + picture->f_motion.f_code[1] = (buffer[3] >> 2) & 1; + picture->f_motion.f_code[0] = + (((buffer[3] << 1) | (buffer[4] >> 7)) & 7) - 1; + picture->b_motion.f_code[1] = (buffer[4] >> 6) & 1; + picture->b_motion.f_code[0] = ((buffer[4] >> 3) & 7) - 1; + + /* move in header_process_picture_header */ + picture->second_field = + (picture->picture_structure != FRAME_PICTURE) && + !(picture->second_field); + + return 0; +} diff --git a/src/video_dec/libmpeg2/idct.c b/src/video_dec/libmpeg2/idct.c new file mode 100644 index 000000000..9f216db58 --- /dev/null +++ b/src/video_dec/libmpeg2/idct.c @@ -0,0 +1,348 @@ +/* + * idct.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * Portions of this code are from the MPEG software simulation group + * idct implementation. This code will be replaced with a new + * implementation soon. + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +/**********************************************************/ +/* inverse two dimensional DCT, Chen-Wang algorithm */ +/* (cf. IEEE ASSP-32, pp. 803-816, Aug. 1984) */ +/* 32-bit integer arithmetic (8 bit coefficients) */ +/* 11 mults, 29 adds per DCT */ +/* sE, 18.8.91 */ +/**********************************************************/ +/* coefficients extended to 12 bit for IEEE1180-1990 */ +/* compliance sE, 2.1.94 */ +/**********************************************************/ + +/* this code assumes >> to be a two's-complement arithmetic */ +/* right shift: (-2)>>1 == -1 , (-3)>>1 == -2 */ + +#include "config.h" + +#include <stdio.h> +#include <stdlib.h> +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include <xine/xineutils.h> + +#define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */ +#define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */ +#define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */ +#define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */ +#define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */ +#define W7 565 /* 2048*sqrt (2)*cos (7*pi/16) */ + +/* idct main entry points */ +void (* mpeg2_idct_copy) (int16_t * block, uint8_t * dest, int stride); +void (* mpeg2_idct_add) (int16_t * block, uint8_t * dest, int stride); +void (* mpeg2_idct) (int16_t * block); +void (* mpeg2_zero_block) (int16_t * block); + +static uint8_t clip_lut[1024]; +#define CLIP(i) ((clip_lut+384)[ (i)]) + +/* row (horizontal) IDCT + * + * 7 pi 1 + * dst[k] = sum c[l] * src[l] * cos ( -- * ( k + - ) * l ) + * l=0 8 2 + * + * where: c[0] = 128 + * c[1..7] = 128*sqrt (2) + */ + +static void inline idct_row (int16_t * block) +{ + int x0, x1, x2, x3, x4, x5, x6, x7, x8; + + x1 = block[4] << 11; + x2 = block[6]; + x3 = block[2]; + x4 = block[1]; + x5 = block[7]; + x6 = block[5]; + x7 = block[3]; + + /* shortcut */ + if (! (x1 | x2 | x3 | x4 | x5 | x6 | x7 )) { + block[0] = block[1] = block[2] = block[3] = block[4] = + block[5] = block[6] = block[7] = block[0]<<3; + return; + } + + x0 = (block[0] << 11) + 128; /* for proper rounding in the fourth stage */ + + /* first stage */ + x8 = W7 * (x4 + x5); + x4 = x8 + (W1 - W7) * x4; + x5 = x8 - (W1 + W7) * x5; + x8 = W3 * (x6 + x7); + x6 = x8 - (W3 - W5) * x6; + x7 = x8 - (W3 + W5) * x7; + + /* second stage */ + x8 = x0 + x1; + x0 -= x1; + x1 = W6 * (x3 + x2); + x2 = x1 - (W2 + W6) * x2; + x3 = x1 + (W2 - W6) * x3; + x1 = x4 + x6; + x4 -= x6; + x6 = x5 + x7; + x5 -= x7; + + /* third stage */ + x7 = x8 + x3; + x8 -= x3; + x3 = x0 + x2; + x0 -= x2; + x2 = (181 * (x4 + x5) + 128) >> 8; + x4 = (181 * (x4 - x5) + 128) >> 8; + + /* fourth stage */ + block[0] = (x7 + x1) >> 8; + block[1] = (x3 + x2) >> 8; + block[2] = (x0 + x4) >> 8; + block[3] = (x8 + x6) >> 8; + block[4] = (x8 - x6) >> 8; + block[5] = (x0 - x4) >> 8; + block[6] = (x3 - x2) >> 8; + block[7] = (x7 - x1) >> 8; +} + +/* column (vertical) IDCT + * + * 7 pi 1 + * dst[8*k] = sum c[l] * src[8*l] * cos ( -- * ( k + - ) * l ) + * l=0 8 2 + * + * where: c[0] = 1/1024 + * c[1..7] = (1/1024)*sqrt (2) + */ + +static void inline idct_col (int16_t *block) +{ + int x0, x1, x2, x3, x4, x5, x6, x7, x8; + + /* shortcut */ + x1 = block [8*4] << 8; + x2 = block [8*6]; + x3 = block [8*2]; + x4 = block [8*1]; + x5 = block [8*7]; + x6 = block [8*5]; + x7 = block [8*3]; + +#if 0 + if (! (x1 | x2 | x3 | x4 | x5 | x6 | x7 )) { + block[8*0] = block[8*1] = block[8*2] = block[8*3] = block[8*4] = + block[8*5] = block[8*6] = block[8*7] = (block[8*0] + 32) >> 6; + return; + } +#endif + + x0 = (block[8*0] << 8) + 8192; + + /* first stage */ + x8 = W7 * (x4 + x5) + 4; + x4 = (x8 + (W1 - W7) * x4) >> 3; + x5 = (x8 - (W1 + W7) * x5) >> 3; + x8 = W3 * (x6 + x7) + 4; + x6 = (x8 - (W3 - W5) * x6) >> 3; + x7 = (x8 - (W3 + W5) * x7) >> 3; + + /* second stage */ + x8 = x0 + x1; + x0 -= x1; + x1 = W6 * (x3 + x2) + 4; + x2 = (x1 - (W2 + W6) * x2) >> 3; + x3 = (x1 + (W2 - W6) * x3) >> 3; + x1 = x4 + x6; + x4 -= x6; + x6 = x5 + x7; + x5 -= x7; + + /* third stage */ + x7 = x8 + x3; + x8 -= x3; + x3 = x0 + x2; + x0 -= x2; + x2 = (181 * (x4 + x5) + 128) >> 8; + x4 = (181 * (x4 - x5) + 128) >> 8; + + /* fourth stage */ + block[8*0] = (x7 + x1) >> 14; + block[8*1] = (x3 + x2) >> 14; + block[8*2] = (x0 + x4) >> 14; + block[8*3] = (x8 + x6) >> 14; + block[8*4] = (x8 - x6) >> 14; + block[8*5] = (x0 - x4) >> 14; + block[8*6] = (x3 - x2) >> 14; + block[8*7] = (x7 - x1) >> 14; +} + +static void mpeg2_idct_copy_c (int16_t * block, uint8_t * dest, int stride) +{ + int i; + + for (i = 0; i < 8; i++) + idct_row (block + 8 * i); + + for (i = 0; i < 8; i++) + idct_col (block + i); + + i = 8; + do { + dest[0] = CLIP (block[0]); + dest[1] = CLIP (block[1]); + dest[2] = CLIP (block[2]); + dest[3] = CLIP (block[3]); + dest[4] = CLIP (block[4]); + dest[5] = CLIP (block[5]); + dest[6] = CLIP (block[6]); + dest[7] = CLIP (block[7]); + + block[0] = 0; block[1] = 0; block[2] = 0; block[3] = 0; + block[4] = 0; block[5] = 0; block[6] = 0; block[7] = 0; + + dest += stride; + block += 8; + } while (--i); +} + +static void mpeg2_idct_add_c (int16_t * block, uint8_t * dest, int stride) +{ + int i; + + for (i = 0; i < 8; i++) + idct_row (block + 8 * i); + + for (i = 0; i < 8; i++) + idct_col (block + i); + + i = 8; + do { + dest[0] = CLIP (block[0] + dest[0]); + dest[1] = CLIP (block[1] + dest[1]); + dest[2] = CLIP (block[2] + dest[2]); + dest[3] = CLIP (block[3] + dest[3]); + dest[4] = CLIP (block[4] + dest[4]); + dest[5] = CLIP (block[5] + dest[5]); + dest[6] = CLIP (block[6] + dest[6]); + dest[7] = CLIP (block[7] + dest[7]); + + block[0] = 0; block[1] = 0; block[2] = 0; block[3] = 0; + block[4] = 0; block[5] = 0; block[6] = 0; block[7] = 0; + + dest += stride; + block += 8; + } while (--i); +} + +static void mpeg2_idct_c (int16_t * block) +{ + int i; + + for (i = 0; i < 8; i++) + idct_row (block + 8 * i); + + for (i = 0; i < 8; i++) + idct_col (block + i); +} + +static void mpeg2_zero_block_c (int16_t * wblock) +{ + memset( wblock, 0, sizeof(int16_t) * 64 ); +} + +void mpeg2_idct_init (uint32_t mm_accel) +{ + mpeg2_zero_block = mpeg2_zero_block_c; + +#if defined(ARCH_X86) || defined(ARCH_X86_64) + if (mm_accel & MM_ACCEL_X86_MMXEXT) { +#ifdef LOG + fprintf (stderr, "Using MMXEXT for IDCT transform\n"); +#endif + mpeg2_idct_copy = mpeg2_idct_copy_mmxext; + mpeg2_idct_add = mpeg2_idct_add_mmxext; + mpeg2_idct = mpeg2_idct_mmxext; + mpeg2_zero_block = mpeg2_zero_block_mmx; + mpeg2_idct_mmx_init (); + } else if (mm_accel & MM_ACCEL_X86_MMX) { +#ifdef LOG + fprintf (stderr, "Using MMX for IDCT transform\n"); +#endif + mpeg2_idct_copy = mpeg2_idct_copy_mmx; + mpeg2_idct_add = mpeg2_idct_add_mmx; + mpeg2_idct = mpeg2_idct_mmx; + mpeg2_zero_block = mpeg2_zero_block_mmx; + mpeg2_idct_mmx_init (); + } else +#endif +#if defined (ARCH_PPC) && defined (ENABLE_ALTIVEC) + if (mm_accel & MM_ACCEL_PPC_ALTIVEC) { +#ifdef LOG + fprintf (stderr, "Using altivec for IDCT transform\n"); +#endif + mpeg2_idct_copy = mpeg2_idct_copy_altivec; + mpeg2_idct_add = mpeg2_idct_add_altivec; + mpeg2_idct_altivec_init (); + mpeg2_idct = mpeg2_idct_c; + } else +#endif +#ifdef LIBMPEG2_MLIB + if (mm_accel & MM_ACCEL_MLIB) { + char * env_var; + + env_var = getenv ("MLIB_NON_IEEE"); + + mpeg2_idct = mpeg2_idct_mlib; + if (env_var == NULL) { +#ifdef LOG + fprintf (stderr, "Using mlib for IDCT transform\n"); +#endif + mpeg2_idct_add = mpeg2_idct_add_mlib; + } else { + fprintf (stderr, "Using non-IEEE mlib for IDCT transform\n"); + mpeg2_idct_add = mpeg2_idct_add_mlib_non_ieee; + } + mpeg2_idct_copy = mpeg2_idct_copy_mlib_non_ieee; + } else +#endif + { + int i; + +#ifdef LOG + fprintf (stderr, "No accelerated IDCT transform found\n"); +#endif + mpeg2_idct_copy = mpeg2_idct_copy_c; + mpeg2_idct_add = mpeg2_idct_add_c; + mpeg2_idct = mpeg2_idct_c; + for (i = -384; i < 640; i++) + clip_lut[i+384] = (i < 0) ? 0 : ((i > 255) ? 255 : i); + } +} diff --git a/src/video_dec/libmpeg2/idct_altivec.c b/src/video_dec/libmpeg2/idct_altivec.c new file mode 100644 index 000000000..de396560b --- /dev/null +++ b/src/video_dec/libmpeg2/idct_altivec.c @@ -0,0 +1,233 @@ +/* + * idct_altivec.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#if defined (ARCH_PPC) && defined (ENABLE_ALTIVEC) + +#include <altivec.h> + +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include <xine/xineutils.h> + +#define vector_s16_t vector signed short +#define vector_u16_t vector unsigned short +#define vector_s8_t vector signed char +#define vector_u8_t vector unsigned char +#define vector_s32_t vector signed int +#define vector_u32_t vector unsigned int + +#define IDCT_HALF \ + /* 1st stage */ \ + t1 = vec_mradds (a1, vx7, vx1 ); \ + t8 = vec_mradds (a1, vx1, vec_subs (zero, vx7)); \ + t7 = vec_mradds (a2, vx5, vx3); \ + t3 = vec_mradds (ma2, vx3, vx5); \ + \ + /* 2nd stage */ \ + t5 = vec_adds (vx0, vx4); \ + t0 = vec_subs (vx0, vx4); \ + t2 = vec_mradds (a0, vx6, vx2); \ + t4 = vec_mradds (a0, vx2, vec_subs (zero,vx6)); \ + t6 = vec_adds (t8, t3); \ + t3 = vec_subs (t8, t3); \ + t8 = vec_subs (t1, t7); \ + t1 = vec_adds (t1, t7); \ + \ + /* 3rd stage */ \ + t7 = vec_adds (t5, t2); \ + t2 = vec_subs (t5, t2); \ + t5 = vec_adds (t0, t4); \ + t0 = vec_subs (t0, t4); \ + t4 = vec_subs (t8, t3); \ + t3 = vec_adds (t8, t3); \ + \ + /* 4th stage */ \ + vy0 = vec_adds (t7, t1); \ + vy7 = vec_subs (t7, t1); \ + vy1 = vec_mradds (c4, t3, t5); \ + vy6 = vec_mradds (mc4, t3, t5); \ + vy2 = vec_mradds (c4, t4, t0); \ + vy5 = vec_mradds (mc4, t4, t0); \ + vy3 = vec_adds (t2, t6); \ + vy4 = vec_subs (t2, t6); + +#define IDCT \ + vector_s16_t vx0, vx1, vx2, vx3, vx4, vx5, vx6, vx7; \ + vector_s16_t vy0, vy1, vy2, vy3, vy4, vy5, vy6, vy7; \ + vector_s16_t a0, a1, a2, ma2, c4, mc4, zero, bias; \ + vector_s16_t t0, t1, t2, t3, t4, t5, t6, t7, t8; \ + vector_u16_t shift; \ + \ + c4 = vec_splat (constants[0], 0); \ + a0 = vec_splat (constants[0], 1); \ + a1 = vec_splat (constants[0], 2); \ + a2 = vec_splat (constants[0], 3); \ + mc4 = vec_splat (constants[0], 4); \ + ma2 = vec_splat (constants[0], 5); \ + bias = (vector_s16_t)vec_splat ((vector_s32_t)constants[0], 3); \ + \ + zero = vec_splat_s16 (0); \ + shift = vec_splat_u16 (4); \ + \ + vx0 = vec_mradds (vec_sl (block[0], shift), constants[1], zero); \ + vx1 = vec_mradds (vec_sl (block[1], shift), constants[2], zero); \ + vx2 = vec_mradds (vec_sl (block[2], shift), constants[3], zero); \ + vx3 = vec_mradds (vec_sl (block[3], shift), constants[4], zero); \ + vx4 = vec_mradds (vec_sl (block[4], shift), constants[1], zero); \ + vx5 = vec_mradds (vec_sl (block[5], shift), constants[4], zero); \ + vx6 = vec_mradds (vec_sl (block[6], shift), constants[3], zero); \ + vx7 = vec_mradds (vec_sl (block[7], shift), constants[2], zero); \ + \ + IDCT_HALF \ + \ + vx0 = vec_mergeh (vy0, vy4); \ + vx1 = vec_mergel (vy0, vy4); \ + vx2 = vec_mergeh (vy1, vy5); \ + vx3 = vec_mergel (vy1, vy5); \ + vx4 = vec_mergeh (vy2, vy6); \ + vx5 = vec_mergel (vy2, vy6); \ + vx6 = vec_mergeh (vy3, vy7); \ + vx7 = vec_mergel (vy3, vy7); \ + \ + vy0 = vec_mergeh (vx0, vx4); \ + vy1 = vec_mergel (vx0, vx4); \ + vy2 = vec_mergeh (vx1, vx5); \ + vy3 = vec_mergel (vx1, vx5); \ + vy4 = vec_mergeh (vx2, vx6); \ + vy5 = vec_mergel (vx2, vx6); \ + vy6 = vec_mergeh (vx3, vx7); \ + vy7 = vec_mergel (vx3, vx7); \ + \ + vx0 = vec_adds (vec_mergeh (vy0, vy4), bias); \ + vx1 = vec_mergel (vy0, vy4); \ + vx2 = vec_mergeh (vy1, vy5); \ + vx3 = vec_mergel (vy1, vy5); \ + vx4 = vec_mergeh (vy2, vy6); \ + vx5 = vec_mergel (vy2, vy6); \ + vx6 = vec_mergeh (vy3, vy7); \ + vx7 = vec_mergel (vy3, vy7); \ + \ + IDCT_HALF \ + \ + shift = vec_splat_u16 (6); \ + vx0 = vec_sra (vy0, shift); \ + vx1 = vec_sra (vy1, shift); \ + vx2 = vec_sra (vy2, shift); \ + vx3 = vec_sra (vy3, shift); \ + vx4 = vec_sra (vy4, shift); \ + vx5 = vec_sra (vy5, shift); \ + vx6 = vec_sra (vy6, shift); \ + vx7 = vec_sra (vy7, shift); + +#if defined( __APPLE_CC__ ) && defined( __APPLE_ALTIVEC__ ) /* apple */ +#define VEC_S16(a,b,c,d,e,f,g,h) (vector_s16_t) (a, b, c, d, e, f, g, h) +#else /* gnu */ +#define VEC_S16(a,b,c,d,e,f,g,h) (vector_s16_t) {a, b, c, d, e, f, g, h} +#endif + +static vector_s16_t constants[5] = { + VEC_S16(23170, 13573, 6518, 21895, -23170, -21895, 32, 31), + VEC_S16(16384, 22725, 21407, 19266, 16384, 19266, 21407, 22725), + VEC_S16(22725, 31521, 29692, 26722, 22725, 26722, 29692, 31521), + VEC_S16(21407, 29692, 27969, 25172, 21407, 25172, 27969, 29692), + VEC_S16(19266, 26722, 25172, 22654, 19266, 22654, 25172, 26722) +}; + +void mpeg2_idct_copy_altivec (vector_s16_t * block, unsigned char * dest, + int stride) +{ + vector_u8_t tmp; + + IDCT + +#define COPY(dest,src) \ + tmp = vec_packsu (src, src); \ + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \ + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + + COPY (dest, vx0) dest += stride; + COPY (dest, vx1) dest += stride; + COPY (dest, vx2) dest += stride; + COPY (dest, vx3) dest += stride; + COPY (dest, vx4) dest += stride; + COPY (dest, vx5) dest += stride; + COPY (dest, vx6) dest += stride; + COPY (dest, vx7) + memset (block, 0, 64 * sizeof (signed short)); +} + +void mpeg2_idct_add_altivec (vector_s16_t * block, unsigned char * dest, + int stride) +{ + vector_u8_t tmp; + vector_s16_t tmp2, tmp3; + vector_u8_t perm0; + vector_u8_t perm1; + vector_u8_t p0, p1, p; + + IDCT + + p0 = vec_lvsl (0, dest); + p1 = vec_lvsl (stride, dest); + p = vec_splat_u8 (-1); + perm0 = vec_mergeh (p, p0); + perm1 = vec_mergeh (p, p1); + +#define ADD(dest,src,perm) \ + /* *(uint64_t *)&tmp = *(uint64_t *)dest; */ \ + tmp = vec_ld (0, dest); \ + tmp2 = (vector_s16_t)vec_perm (tmp, (vector_u8_t)zero, perm); \ + tmp3 = vec_adds (tmp2, src); \ + tmp = vec_packsu (tmp3, tmp3); \ + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \ + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + + ADD (dest, vx0, perm0) dest += stride; + ADD (dest, vx1, perm1) dest += stride; + ADD (dest, vx2, perm0) dest += stride; + ADD (dest, vx3, perm1) dest += stride; + ADD (dest, vx4, perm0) dest += stride; + ADD (dest, vx5, perm1) dest += stride; + ADD (dest, vx6, perm0) dest += stride; + ADD (dest, vx7, perm1) + memset (block, 0, 64 * sizeof (signed short)); +} + +void mpeg2_idct_altivec_init (void) +{ + int i, j; + + /* the altivec idct uses a transposed input, so we patch scan tables */ + for (i = 0; i < 64; i++) { + j = mpeg2_scan_norm[i]; + mpeg2_scan_norm[i] = (j >> 3) | ((j & 7) << 3); + j = mpeg2_scan_alt[i]; + mpeg2_scan_alt[i] = (j >> 3) | ((j & 7) << 3); + } +} + +#endif /* ARCH_PPC && ENABLED_ALTIVEC */ + diff --git a/src/video_dec/libmpeg2/idct_mlib.c b/src/video_dec/libmpeg2/idct_mlib.c new file mode 100644 index 000000000..e573c9790 --- /dev/null +++ b/src/video_dec/libmpeg2/idct_mlib.c @@ -0,0 +1,62 @@ +/* + * idct_mlib.c + * Copyright (C) 1999-2002 Håkan Hjort <d95hjort@dtek.chalmers.se> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#ifdef LIBMPEG2_MLIB + +#include <mlib_types.h> +#include <mlib_status.h> +#include <mlib_sys.h> +#include <mlib_video.h> +#include <string.h> +#include <inttypes.h> + +#include "mpeg2_internal.h" + +void mpeg2_idct_add_mlib (int16_t * block, uint8_t * dest, int stride) +{ + mlib_VideoIDCT_IEEE_S16_S16 (block, block); + mlib_VideoAddBlock_U8_S16 (dest, block, stride); + memset (block, 0, 64 * sizeof (uint16_t)); +} + +void mpeg2_idct_copy_mlib_non_ieee (int16_t * block, uint8_t * dest, + int stride) +{ + mlib_VideoIDCT8x8_U8_S16 (dest, block, stride); + memset (block, 0, 64 * sizeof (uint16_t)); +} + +void mpeg2_idct_add_mlib_non_ieee (int16_t * block, uint8_t * dest, int stride) +{ + mlib_VideoIDCT8x8_S16_S16 (block, block); + mlib_VideoAddBlock_U8_S16 (dest, block, stride); + memset (block, 0, 64 * sizeof (uint16_t)); +} + +void mpeg2_idct_mlib (int16_t * block) +{ + mlib_VideoIDCT_IEEE_S16_S16 (block, block); +} + +#endif diff --git a/src/video_dec/libmpeg2/idct_mlib.h b/src/video_dec/libmpeg2/idct_mlib.h new file mode 100644 index 000000000..1fb0787dd --- /dev/null +++ b/src/video_dec/libmpeg2/idct_mlib.h @@ -0,0 +1,25 @@ +/* + * idct_mlib.h + * + * Copyright (C) 1999, Håkan Hjort <d95hjort@dtek.chalmers.se> + * + * 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, 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 GNU Make; see the file COPYING. If not, write to + * the Free Software Foundation, + * + */ + +void idct_block_copy_mlib (int16_t * block, uint8_t * dest, int stride); +void idct_block_add_mlib (int16_t * block, uint8_t * dest, int stride); diff --git a/src/video_dec/libmpeg2/idct_mmx.c b/src/video_dec/libmpeg2/idct_mmx.c new file mode 100644 index 000000000..92ae365b4 --- /dev/null +++ b/src/video_dec/libmpeg2/idct_mmx.c @@ -0,0 +1,741 @@ +/* + * idct_mmx.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#if defined(ARCH_X86) || defined(ARCH_X86_64) + +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include <xine/xineutils.h> +#include "xine_mmx.h" + +#define ROW_SHIFT 11 +#define COL_SHIFT 6 + +#define round(bias) ((int)(((bias)+0.5) * (1<<ROW_SHIFT))) +#define rounder(bias) {round (bias), round (bias)} + + +#if 0 +/* C row IDCT - its just here to document the MMXEXT and MMX versions */ +static inline void idct_row (int16_t * row, int offset, + int16_t * table, int32_t * rounder) +{ + int C1, C2, C3, C4, C5, C6, C7; + int a0, a1, a2, a3, b0, b1, b2, b3; + + row += offset; + + C1 = table[1]; + C2 = table[2]; + C3 = table[3]; + C4 = table[4]; + C5 = table[5]; + C6 = table[6]; + C7 = table[7]; + + a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6] + *rounder; + a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6] + *rounder; + a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6] + *rounder; + a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6] + *rounder; + + b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7]; + b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7]; + b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7]; + b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7]; + + row[0] = (a0 + b0) >> ROW_SHIFT; + row[1] = (a1 + b1) >> ROW_SHIFT; + row[2] = (a2 + b2) >> ROW_SHIFT; + row[3] = (a3 + b3) >> ROW_SHIFT; + row[4] = (a3 - b3) >> ROW_SHIFT; + row[5] = (a2 - b2) >> ROW_SHIFT; + row[6] = (a1 - b1) >> ROW_SHIFT; + row[7] = (a0 - b0) >> ROW_SHIFT; +} +#endif + + +/* MMXEXT row IDCT */ + +#define mmxext_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, -c4, -c2, \ + c4, c6, c4, c6, \ + c1, c3, -c1, -c5, \ + c5, c7, c3, -c7, \ + c4, -c6, c4, -c6, \ + -c4, c2, c4, -c2, \ + c5, -c1, c3, -c1, \ + c7, c3, c7, -c5 } + +static inline void mmxext_row_head (int16_t * row, int offset, int16_t * table) +{ + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + + movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + + movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4 + pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2 + + pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4 +} + +static inline void mmxext_row (int16_t * table, int32_t * rounder) +{ + movq_m2r (*(table+8), mm1); // mm1 = -C5 -C1 C3 C1 + pmaddwd_r2r (mm2, mm4); // mm4 = C4*x0+C6*x2 C4*x4+C6*x6 + + pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x4-C6*x6 C4*x0-C6*x2 + pshufw_r2r (mm6, mm6, 0x4e); // mm6 = x3 x1 x7 x5 + + movq_m2r (*(table+12), mm7); // mm7 = -C7 C3 C7 C5 + pmaddwd_r2r (mm5, mm1); // mm1 = -C1*x5-C5*x7 C1*x1+C3*x3 + + paddd_m2r (*rounder, mm3); // mm3 += rounder + pmaddwd_r2r (mm6, mm7); // mm7 = C3*x1-C7*x3 C5*x5+C7*x7 + + pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x0-C2*x2 -C4*x4+C2*x6 + paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder + + pmaddwd_m2r (*(table+24), mm5); // mm5 = C3*x5-C1*x7 C5*x1-C1*x3 + movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder + + pmaddwd_m2r (*(table+28), mm6); // mm6 = C7*x1-C5*x3 C7*x5+C3*x7 + paddd_r2r (mm7, mm1); // mm1 = b1 b0 + + paddd_m2r (*rounder, mm0); // mm0 += rounder + psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder + + psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7 + paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder + + paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder + psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0 + + paddd_r2r (mm6, mm5); // mm5 = b3 b2 + movq_r2r (mm0, mm4); // mm4 = a3 a2 + rounder + + paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder + psubd_r2r (mm5, mm4); // mm4 = a3-b3 a2-b2 + rounder +} + +static inline void mmxext_row_tail (int16_t * row, int store) +{ + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + + psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5 + + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + + packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5 + + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4 + + /* slot */ + + movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4 +} + +static inline void mmxext_row_mid (int16_t * row, int store, + int offset, int16_t * table) +{ + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5 + + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + + packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4 + + movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4 + movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4 + + pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2 + + movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4 + pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4 +} + + +/* MMX row IDCT */ + +#define mmx_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, c4, c6, \ + c4, c6, -c4, -c2, \ + c1, c3, c3, -c7, \ + c5, c7, -c1, -c5, \ + c4, -c6, c4, -c2, \ + -c4, c2, c4, -c6, \ + c5, -c1, c7, -c5, \ + c7, c3, c3, -c1 } + +static inline void mmx_row_head (int16_t * row, int offset, int16_t * table) +{ + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + + movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + + punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0 + + movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4 + pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2 + + movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1 + punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4 +} + +static inline void mmx_row (int16_t * table, int32_t * rounder) +{ + pmaddwd_r2r (mm2, mm4); // mm4 = -C4*x4-C2*x6 C4*x4+C6*x6 + punpckldq_r2r (mm5, mm5); // mm5 = x3 x1 x3 x1 + + pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x0-C2*x2 C4*x0-C6*x2 + punpckhdq_r2r (mm6, mm6); // mm6 = x7 x5 x7 x5 + + movq_m2r (*(table+12), mm7); // mm7 = -C5 -C1 C7 C5 + pmaddwd_r2r (mm5, mm1); // mm1 = C3*x1-C7*x3 C1*x1+C3*x3 + + paddd_m2r (*rounder, mm3); // mm3 += rounder + pmaddwd_r2r (mm6, mm7); // mm7 = -C1*x5-C5*x7 C5*x5+C7*x7 + + pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x4-C6*x6 -C4*x4+C2*x6 + paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder + + pmaddwd_m2r (*(table+24), mm5); // mm5 = C7*x1-C5*x3 C5*x1-C1*x3 + movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder + + pmaddwd_m2r (*(table+28), mm6); // mm6 = C3*x5-C1*x7 C7*x5+C3*x7 + paddd_r2r (mm7, mm1); // mm1 = b1 b0 + + paddd_m2r (*rounder, mm0); // mm0 += rounder + psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder + + psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7 + paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder + + paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder + psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0 + + paddd_r2r (mm6, mm5); // mm5 = b3 b2 + movq_r2r (mm0, mm7); // mm7 = a3 a2 + rounder + + paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder + psubd_r2r (mm5, mm7); // mm7 = a3-b3 a2-b2 + rounder +} + +static inline void mmx_row_tail (int16_t * row, int store) +{ + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + + psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5 + + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + + packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5 + + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + movq_r2r (mm7, mm4); // mm4 = y6 y7 y4 y5 + + pslld_i2r (16, mm7); // mm7 = y7 0 y5 0 + + psrld_i2r (16, mm4); // mm4 = 0 y6 0 y4 + + por_r2r (mm4, mm7); // mm7 = y7 y6 y5 y4 + + /* slot */ + + movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4 +} + +static inline void mmx_row_mid (int16_t * row, int store, + int offset, int16_t * table) +{ + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5 + + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + + packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + movq_r2r (mm7, mm1); // mm1 = y6 y7 y4 y5 + + punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0 + psrld_i2r (16, mm7); // mm7 = 0 y6 0 y4 + + movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4 + pslld_i2r (16, mm1); // mm1 = y7 0 y5 0 + + movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4 + por_r2r (mm1, mm7); // mm7 = y7 y6 y5 y4 + + movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1 + punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4 + + movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4 + pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2 +} + + +#if 0 +// C column IDCT - its just here to document the MMXEXT and MMX versions +static inline void idct_col (int16_t * col, int offset) +{ +/* multiplication - as implemented on mmx */ +#define F(c,x) (((c) * (x)) >> 16) + +/* saturation - it helps us handle torture test cases */ +#define S(x) (((x)>32767) ? 32767 : ((x)<-32768) ? -32768 : (x)) + + int16_t x0, x1, x2, x3, x4, x5, x6, x7; + int16_t y0, y1, y2, y3, y4, y5, y6, y7; + int16_t a0, a1, a2, a3, b0, b1, b2, b3; + int16_t u04, v04, u26, v26, u17, v17, u35, v35, u12, v12; + + col += offset; + + x0 = col[0*8]; + x1 = col[1*8]; + x2 = col[2*8]; + x3 = col[3*8]; + x4 = col[4*8]; + x5 = col[5*8]; + x6 = col[6*8]; + x7 = col[7*8]; + + u04 = S (x0 + x4); + v04 = S (x0 - x4); + u26 = S (F (T2, x6) + x2); + v26 = S (F (T2, x2) - x6); + + a0 = S (u04 + u26); + a1 = S (v04 + v26); + a2 = S (v04 - v26); + a3 = S (u04 - u26); + + u17 = S (F (T1, x7) + x1); + v17 = S (F (T1, x1) - x7); + u35 = S (F (T3, x5) + x3); + v35 = S (F (T3, x3) - x5); + + b0 = S (u17 + u35); + b3 = S (v17 - v35); + u12 = S (u17 - u35); + v12 = S (v17 + v35); + u12 = S (2 * F (C4, u12)); + v12 = S (2 * F (C4, v12)); + b1 = S (u12 + v12); + b2 = S (u12 - v12); + + y0 = S (a0 + b0) >> COL_SHIFT; + y1 = S (a1 + b1) >> COL_SHIFT; + y2 = S (a2 + b2) >> COL_SHIFT; + y3 = S (a3 + b3) >> COL_SHIFT; + + y4 = S (a3 - b3) >> COL_SHIFT; + y5 = S (a2 - b2) >> COL_SHIFT; + y6 = S (a1 - b1) >> COL_SHIFT; + y7 = S (a0 - b0) >> COL_SHIFT; + + col[0*8] = y0; + col[1*8] = y1; + col[2*8] = y2; + col[3*8] = y3; + col[4*8] = y4; + col[5*8] = y5; + col[6*8] = y6; + col[7*8] = y7; +} +#endif + + +// MMX column IDCT +static inline void idct_col (int16_t * col, int offset) +{ +#define T1 13036 +#define T2 27146 +#define T3 43790 +#define C4 23170 + + static short _T1[] ATTR_ALIGN(8) = {T1,T1,T1,T1}; + static short _T2[] ATTR_ALIGN(8) = {T2,T2,T2,T2}; + static short _T3[] ATTR_ALIGN(8) = {T3,T3,T3,T3}; + static short _C4[] ATTR_ALIGN(8) = {C4,C4,C4,C4}; + + /* column code adapted from peter gubanov */ + /* http://www.elecard.com/peter/idct.shtml */ + + movq_m2r (*_T1, mm0); // mm0 = T1 + + movq_m2r (*(col+offset+1*8), mm1); // mm1 = x1 + movq_r2r (mm0, mm2); // mm2 = T1 + + movq_m2r (*(col+offset+7*8), mm4); // mm4 = x7 + pmulhw_r2r (mm1, mm0); // mm0 = T1*x1 + + movq_m2r (*_T3, mm5); // mm5 = T3 + pmulhw_r2r (mm4, mm2); // mm2 = T1*x7 + + movq_m2r (*(col+offset+5*8), mm6); // mm6 = x5 + movq_r2r (mm5, mm7); // mm7 = T3-1 + + movq_m2r (*(col+offset+3*8), mm3); // mm3 = x3 + psubsw_r2r (mm4, mm0); // mm0 = v17 + + movq_m2r (*_T2, mm4); // mm4 = T2 + pmulhw_r2r (mm3, mm5); // mm5 = (T3-1)*x3 + + paddsw_r2r (mm2, mm1); // mm1 = u17 + pmulhw_r2r (mm6, mm7); // mm7 = (T3-1)*x5 + + /* slot */ + + movq_r2r (mm4, mm2); // mm2 = T2 + paddsw_r2r (mm3, mm5); // mm5 = T3*x3 + + pmulhw_m2r (*(col+offset+2*8), mm4);// mm4 = T2*x2 + paddsw_r2r (mm6, mm7); // mm7 = T3*x5 + + psubsw_r2r (mm6, mm5); // mm5 = v35 + paddsw_r2r (mm3, mm7); // mm7 = u35 + + movq_m2r (*(col+offset+6*8), mm3); // mm3 = x6 + movq_r2r (mm0, mm6); // mm6 = v17 + + pmulhw_r2r (mm3, mm2); // mm2 = T2*x6 + psubsw_r2r (mm5, mm0); // mm0 = b3 + + psubsw_r2r (mm3, mm4); // mm4 = v26 + paddsw_r2r (mm6, mm5); // mm5 = v12 + + movq_r2m (mm0, *(col+offset+3*8)); // save b3 in scratch0 + movq_r2r (mm1, mm6); // mm6 = u17 + + paddsw_m2r (*(col+offset+2*8), mm2);// mm2 = u26 + paddsw_r2r (mm7, mm6); // mm6 = b0 + + psubsw_r2r (mm7, mm1); // mm1 = u12 + movq_r2r (mm1, mm7); // mm7 = u12 + + movq_m2r (*(col+offset+0*8), mm3); // mm3 = x0 + paddsw_r2r (mm5, mm1); // mm1 = u12+v12 + + movq_m2r (*_C4, mm0); // mm0 = C4/2 + psubsw_r2r (mm5, mm7); // mm7 = u12-v12 + + movq_r2m (mm6, *(col+offset+5*8)); // save b0 in scratch1 + pmulhw_r2r (mm0, mm1); // mm1 = b1/2 + + movq_r2r (mm4, mm6); // mm6 = v26 + pmulhw_r2r (mm0, mm7); // mm7 = b2/2 + + movq_m2r (*(col+offset+4*8), mm5); // mm5 = x4 + movq_r2r (mm3, mm0); // mm0 = x0 + + psubsw_r2r (mm5, mm3); // mm3 = v04 + paddsw_r2r (mm5, mm0); // mm0 = u04 + + paddsw_r2r (mm3, mm4); // mm4 = a1 + movq_r2r (mm0, mm5); // mm5 = u04 + + psubsw_r2r (mm6, mm3); // mm3 = a2 + paddsw_r2r (mm2, mm5); // mm5 = a0 + + paddsw_r2r (mm1, mm1); // mm1 = b1 + psubsw_r2r (mm2, mm0); // mm0 = a3 + + paddsw_r2r (mm7, mm7); // mm7 = b2 + movq_r2r (mm3, mm2); // mm2 = a2 + + movq_r2r (mm4, mm6); // mm6 = a1 + paddsw_r2r (mm7, mm3); // mm3 = a2+b2 + + psraw_i2r (COL_SHIFT, mm3); // mm3 = y2 + paddsw_r2r (mm1, mm4); // mm4 = a1+b1 + + psraw_i2r (COL_SHIFT, mm4); // mm4 = y1 + psubsw_r2r (mm1, mm6); // mm6 = a1-b1 + + movq_m2r (*(col+offset+5*8), mm1); // mm1 = b0 + psubsw_r2r (mm7, mm2); // mm2 = a2-b2 + + psraw_i2r (COL_SHIFT, mm6); // mm6 = y6 + movq_r2r (mm5, mm7); // mm7 = a0 + + movq_r2m (mm4, *(col+offset+1*8)); // save y1 + psraw_i2r (COL_SHIFT, mm2); // mm2 = y5 + + movq_r2m (mm3, *(col+offset+2*8)); // save y2 + paddsw_r2r (mm1, mm5); // mm5 = a0+b0 + + movq_m2r (*(col+offset+3*8), mm4); // mm4 = b3 + psubsw_r2r (mm1, mm7); // mm7 = a0-b0 + + psraw_i2r (COL_SHIFT, mm5); // mm5 = y0 + movq_r2r (mm0, mm3); // mm3 = a3 + + movq_r2m (mm2, *(col+offset+5*8)); // save y5 + psubsw_r2r (mm4, mm3); // mm3 = a3-b3 + + psraw_i2r (COL_SHIFT, mm7); // mm7 = y7 + paddsw_r2r (mm0, mm4); // mm4 = a3+b3 + + movq_r2m (mm5, *(col+offset+0*8)); // save y0 + psraw_i2r (COL_SHIFT, mm3); // mm3 = y4 + + movq_r2m (mm6, *(col+offset+6*8)); // save y6 + psraw_i2r (COL_SHIFT, mm4); // mm4 = y3 + + movq_r2m (mm7, *(col+offset+7*8)); // save y7 + + movq_r2m (mm3, *(col+offset+4*8)); // save y4 + + movq_r2m (mm4, *(col+offset+3*8)); // save y3 +} + + +static int32_t rounder0[] ATTR_ALIGN(8) = + rounder ((1 << (COL_SHIFT - 1)) - 0.5); +static int32_t rounder4[] ATTR_ALIGN(8) = rounder (0); +static int32_t rounder1[] ATTR_ALIGN(8) = + rounder (1.25683487303); /* C1*(C1/C4+C1+C7)/2 */ +static int32_t rounder7[] ATTR_ALIGN(8) = + rounder (-0.25); /* C1*(C7/C4+C7-C1)/2 */ +static int32_t rounder2[] ATTR_ALIGN(8) = + rounder (0.60355339059); /* C2 * (C6+C2)/2 */ +static int32_t rounder6[] ATTR_ALIGN(8) = + rounder (-0.25); /* C2 * (C6-C2)/2 */ +static int32_t rounder3[] ATTR_ALIGN(8) = + rounder (0.087788325588); /* C3*(-C3/C4+C3+C5)/2 */ +static int32_t rounder5[] ATTR_ALIGN(8) = + rounder (-0.441341716183); /* C3*(-C5/C4+C5-C3)/2 */ + + +#define declare_idct(idct,table,idct_row_head,idct_row,idct_row_tail,idct_row_mid) \ +static inline void idct (int16_t * block) \ +{ \ + static int16_t table04[] ATTR_ALIGN(16) = \ + table (22725, 21407, 19266, 16384, 12873, 8867, 4520); \ + static int16_t table17[] ATTR_ALIGN(16) = \ + table (31521, 29692, 26722, 22725, 17855, 12299, 6270); \ + static int16_t table26[] ATTR_ALIGN(16) = \ + table (29692, 27969, 25172, 21407, 16819, 11585, 5906); \ + static int16_t table35[] ATTR_ALIGN(16) = \ + table (26722, 25172, 22654, 19266, 15137, 10426, 5315); \ + \ + idct_row_head (block, 0*8, table04); \ + idct_row (table04, rounder0); \ + idct_row_mid (block, 0*8, 4*8, table04); \ + idct_row (table04, rounder4); \ + idct_row_mid (block, 4*8, 1*8, table17); \ + idct_row (table17, rounder1); \ + idct_row_mid (block, 1*8, 7*8, table17); \ + idct_row (table17, rounder7); \ + idct_row_mid (block, 7*8, 2*8, table26); \ + idct_row (table26, rounder2); \ + idct_row_mid (block, 2*8, 6*8, table26); \ + idct_row (table26, rounder6); \ + idct_row_mid (block, 6*8, 3*8, table35); \ + idct_row (table35, rounder3); \ + idct_row_mid (block, 3*8, 5*8, table35); \ + idct_row (table35, rounder5); \ + idct_row_tail (block, 5*8); \ + \ + idct_col (block, 0); \ + idct_col (block, 4); \ +} + + +#define COPY_MMX(offset,r0,r1,r2) \ +do { \ + movq_m2r (*(block+offset), r0); \ + dest += stride; \ + movq_m2r (*(block+offset+4), r1); \ + movq_r2m (r2, *dest); \ + packuswb_r2r (r1, r0); \ +} while (0) + +static void block_copy (int16_t * block, uint8_t * dest, int stride) +{ + movq_m2r (*(block+0*8), mm0); + movq_m2r (*(block+0*8+4), mm1); + movq_m2r (*(block+1*8), mm2); + packuswb_r2r (mm1, mm0); + movq_m2r (*(block+1*8+4), mm3); + movq_r2m (mm0, *dest); + packuswb_r2r (mm3, mm2); + COPY_MMX (2*8, mm0, mm1, mm2); + COPY_MMX (3*8, mm2, mm3, mm0); + COPY_MMX (4*8, mm0, mm1, mm2); + COPY_MMX (5*8, mm2, mm3, mm0); + COPY_MMX (6*8, mm0, mm1, mm2); + COPY_MMX (7*8, mm2, mm3, mm0); + movq_r2m (mm2, *(dest+stride)); +} + + +#define ADD_MMX(offset,r1,r2,r3,r4) \ +do { \ + movq_m2r (*(dest+2*stride), r1); \ + packuswb_r2r (r4, r3); \ + movq_r2r (r1, r2); \ + dest += stride; \ + movq_r2m (r3, *dest); \ + punpcklbw_r2r (mm0, r1); \ + paddsw_m2r (*(block+offset), r1); \ + punpckhbw_r2r (mm0, r2); \ + paddsw_m2r (*(block+offset+4), r2); \ +} while (0) + +static void block_add (int16_t * block, uint8_t * dest, int stride) +{ + movq_m2r (*dest, mm1); + pxor_r2r (mm0, mm0); + movq_m2r (*(dest+stride), mm3); + movq_r2r (mm1, mm2); + punpcklbw_r2r (mm0, mm1); + movq_r2r (mm3, mm4); + paddsw_m2r (*(block+0*8), mm1); + punpckhbw_r2r (mm0, mm2); + paddsw_m2r (*(block+0*8+4), mm2); + punpcklbw_r2r (mm0, mm3); + paddsw_m2r (*(block+1*8), mm3); + packuswb_r2r (mm2, mm1); + punpckhbw_r2r (mm0, mm4); + movq_r2m (mm1, *dest); + paddsw_m2r (*(block+1*8+4), mm4); + ADD_MMX (2*8, mm1, mm2, mm3, mm4); + ADD_MMX (3*8, mm3, mm4, mm1, mm2); + ADD_MMX (4*8, mm1, mm2, mm3, mm4); + ADD_MMX (5*8, mm3, mm4, mm1, mm2); + ADD_MMX (6*8, mm1, mm2, mm3, mm4); + ADD_MMX (7*8, mm3, mm4, mm1, mm2); + packuswb_r2r (mm4, mm3); + movq_r2m (mm3, *(dest+stride)); +} + +static inline void block_zero (int16_t * block) { + pxor_r2r (mm0, mm0); + movq_r2m (mm0, *(block+0*4)); + movq_r2m (mm0, *(block+1*4)); + movq_r2m (mm0, *(block+2*4)); + movq_r2m (mm0, *(block+3*4)); + movq_r2m (mm0, *(block+4*4)); + movq_r2m (mm0, *(block+5*4)); + movq_r2m (mm0, *(block+6*4)); + movq_r2m (mm0, *(block+7*4)); + movq_r2m (mm0, *(block+8*4)); + movq_r2m (mm0, *(block+9*4)); + movq_r2m (mm0, *(block+10*4)); + movq_r2m (mm0, *(block+11*4)); + movq_r2m (mm0, *(block+12*4)); + movq_r2m (mm0, *(block+13*4)); + movq_r2m (mm0, *(block+14*4)); + movq_r2m (mm0, *(block+15*4)); +} + +declare_idct (mmxext_idct, mmxext_table, + mmxext_row_head, mmxext_row, mmxext_row_tail, mmxext_row_mid) + +void mpeg2_idct_copy_mmxext (int16_t * block, uint8_t * dest, int stride) +{ + mmxext_idct (block); + block_copy (block, dest, stride); + block_zero (block); +} + +void mpeg2_idct_add_mmxext (int16_t * block, uint8_t * dest, int stride) +{ + mmxext_idct (block); + block_add (block, dest, stride); + block_zero (block); +} + +void mpeg2_idct_mmxext (int16_t * block) +{ + mmxext_idct (block); +} + +declare_idct (mmx_idct, mmx_table, + mmx_row_head, mmx_row, mmx_row_tail, mmx_row_mid) + +void mpeg2_idct_copy_mmx (int16_t * block, uint8_t * dest, int stride) +{ + mmx_idct (block); + block_copy (block, dest, stride); + block_zero (block); +} + +void mpeg2_idct_add_mmx (int16_t * block, uint8_t * dest, int stride) +{ + mmx_idct (block); + block_add (block, dest, stride); + block_zero (block); +} + +void mpeg2_idct_mmx (int16_t * block) +{ + mmx_idct (block); +} + +void mpeg2_zero_block_mmx (int16_t * block) +{ + block_zero (block); +} + +void mpeg2_idct_mmx_init (void) +{ + int i, j; + + /* the mmx/mmxext idct uses a reordered input, so we patch scan tables */ + + for (i = 0; i < 64; i++) { + j = mpeg2_scan_norm[i]; + mpeg2_scan_norm[i] = (j & 0x38) | ((j & 6) >> 1) | ((j & 1) << 2); + j = mpeg2_scan_alt[i]; + mpeg2_scan_alt[i] = (j & 0x38) | ((j & 6) >> 1) | ((j & 1) << 2); + } +} + +#endif diff --git a/src/video_dec/libmpeg2/libmpeg2_accel.c b/src/video_dec/libmpeg2/libmpeg2_accel.c new file mode 100644 index 000000000..92c0e280b --- /dev/null +++ b/src/video_dec/libmpeg2/libmpeg2_accel.c @@ -0,0 +1,223 @@ +/* + * libmpeg2_accel.c + * Copyright (C) 2004 The Unichrome Project. + * Copyright (C) 2005 Thomas Hellstrom. + * + * This file is part of xine, a free 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 + */ + +#include <xine/xine_internal.h> +#include "mpeg2.h" +#include "mpeg2_internal.h" +#include "xvmc_vld.h" +#include "libmpeg2_accel.h" + + +void +libmpeg2_accel_scan( mpeg2dec_accel_t *accel, uint8_t *scan_norm, uint8_t *scan_alt) +{ + xvmc_setup_scan_ptable(); +} + + +int +libmpeg2_accel_discontinuity(mpeg2dec_accel_t *accel, uint32_t frame_format, picture_t *picture) +{ + accel->xvmc_last_slice_code=-1; + if ( !picture->current_frame ) + return 0; + if (frame_format == XINE_IMGFMT_XXMC) { + xine_xxmc_t *xxmc = (xine_xxmc_t *) + picture->current_frame->accel_data; + switch(xxmc->acceleration) { + case XINE_XVMC_ACCEL_VLD: + case XINE_XVMC_ACCEL_IDCT: + case XINE_XVMC_ACCEL_MOCOMP: + xxmc->proc_xxmc_flush( picture->current_frame ); + break; + default: + break; + } + } + return 0; +} + +int +libmpeg2_accel_new_sequence(mpeg2dec_accel_t *accel, uint32_t frame_format, picture_t *picture) +{ + switch(frame_format) { + case XINE_IMGFMT_XXMC: + case XINE_IMGFMT_XVMC: { + xine_xvmc_t *xvmc = (xine_xvmc_t *) + picture->current_frame->accel_data; + picture->mc = xvmc->macroblocks; + return 0; + } + default: + break; + } + return 1; +} + +int +libmpeg2_accel_new_frame(mpeg2dec_accel_t *accel, uint32_t frame_format, + picture_t *picture, double ratio, uint32_t flags) +{ + if (picture->current_frame) { + if (XINE_IMGFMT_XXMC == frame_format) { + xine_xxmc_t *xxmc = (xine_xxmc_t *) + picture->current_frame->accel_data; + + /* + * Make a request for acceleration type and mpeg coding from + * the output plugin. + */ + + xxmc->fallback_format = XINE_IMGFMT_YV12; + xxmc->acceleration = XINE_XVMC_ACCEL_VLD| XINE_XVMC_ACCEL_IDCT + | XINE_XVMC_ACCEL_MOCOMP ; + + /* + * Standard MOCOMP / IDCT XvMC implementation for interlaced streams + * is buggy. The bug is inherited from the old XvMC driver. Don't use it until + * it has been fixed. (A volunteer ?) + */ + + if ( picture->picture_structure != 3 ) { + picture->top_field_first = (picture->picture_structure == 1); + xxmc->acceleration &= ~( XINE_XVMC_ACCEL_IDCT | XINE_XVMC_ACCEL_MOCOMP ); + } + + xxmc->mpeg = (picture->mpeg1) ? XINE_XVMC_MPEG_1:XINE_XVMC_MPEG_2; + xxmc->proc_xxmc_update_frame (picture->current_frame->driver, + picture->current_frame, + picture->coded_picture_width, + picture->coded_picture_height, + ratio, + XINE_IMGFMT_XXMC, flags); + } + } + return 0; +} + +void +libmpeg2_accel_frame_completion(mpeg2dec_accel_t * accel, uint32_t frame_format, picture_t *picture, + int code) +{ + + if ( !picture->current_frame ) return; + + if (frame_format == XINE_IMGFMT_XXMC) { + xine_xxmc_t *xxmc = (xine_xxmc_t *) + picture->current_frame->accel_data; + if (!xxmc->decoded) { + switch(picture->current_frame->format) { + case XINE_IMGFMT_XXMC: + switch(xxmc->acceleration) { + case XINE_XVMC_ACCEL_VLD: + mpeg2_xxmc_vld_frame_complete(accel, picture, code); + break; + case XINE_XVMC_ACCEL_IDCT: + case XINE_XVMC_ACCEL_MOCOMP: + xxmc->decoded = !picture->current_frame->bad_frame; + xxmc->proc_xxmc_flush( picture->current_frame ); + break; + default: + break; + } + default: + break; + } + } + } +} + + +int +libmpeg2_accel_slice(mpeg2dec_accel_t *accel, picture_t *picture, int code, char * buffer, + uint32_t chunk_size, uint8_t *chunk_buffer) +{ + /* + * Don't reference frames of other formats. They are invalid. This may happen if the + * xxmc plugin suddenly falls back to software decoding. + */ + + if (( picture->current_frame->picture_coding_type == XINE_PICT_P_TYPE ) || + ( picture->current_frame->picture_coding_type == XINE_PICT_B_TYPE )) { + if (! picture->forward_reference_frame) return 1; + if (picture->forward_reference_frame->format != picture->current_frame->format) { + picture->v_offset = 0; + return 1; + } + } + + if ( picture->current_frame->picture_coding_type == XINE_PICT_B_TYPE ) { + if (! picture->backward_reference_frame) return 1; + if (picture->backward_reference_frame->format != picture->current_frame->format) { + picture->v_offset = 0; + return 1; + } + } + + switch( picture->current_frame->format ) { + + case XINE_IMGFMT_XXMC: + { + xine_xxmc_t *xxmc = (xine_xxmc_t *) + picture->current_frame->accel_data; + + if ( xxmc->proc_xxmc_lock_valid( picture->current_frame, + picture->forward_reference_frame, + picture->backward_reference_frame, + picture->current_frame->picture_coding_type)) { + picture->v_offset = 0; + return 1; + } + + switch(picture->current_frame->format) { + case XINE_IMGFMT_XXMC: + switch(xxmc->acceleration) { + case XINE_XVMC_ACCEL_VLD: + mpeg2_xxmc_slice(accel, picture, code, buffer, chunk_size, chunk_buffer); + break; + case XINE_XVMC_ACCEL_IDCT: + case XINE_XVMC_ACCEL_MOCOMP: + mpeg2_xvmc_slice (accel, picture, code, buffer); + break; + default: + mpeg2_slice (picture, code, buffer); + break; + } + break; + default: + mpeg2_slice (picture, code, buffer); + break; + } + xxmc->proc_xxmc_unlock(picture->current_frame->driver); + break; + } + + case XINE_IMGFMT_XVMC: + mpeg2_xvmc_slice (accel, picture, code, buffer); + break; + + default: + mpeg2_slice (picture, code, buffer); + break; + } + return 0; +} diff --git a/src/video_dec/libmpeg2/libmpeg2_accel.h b/src/video_dec/libmpeg2/libmpeg2_accel.h new file mode 100644 index 000000000..5d0b37a78 --- /dev/null +++ b/src/video_dec/libmpeg2/libmpeg2_accel.h @@ -0,0 +1,48 @@ +/* + * libmpeg2_accel.h + * Copyright (C) 2004 The Unichrome Project. + * Copyright (C) 2005 Thomas Hellstrom. + * + * This file is part of xine, a free 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 + */ + +#ifndef LIBMPEG2_ACCEL_H +#define LIBMPEG2_ACCEL_H + +#include "mpeg2_internal.h" + +/* + * Internal context data type. + */ + +typedef struct { + int xvmc_last_slice_code; + int slices_per_row; + int row_slice_count; + unsigned xxmc_mb_pic_height; +} mpeg2dec_accel_t; + +extern int libmpeg2_accel_discontinuity(mpeg2dec_accel_t *accel, uint32_t frame_format, picture_t *picture); +extern int libmpeg2_accel_new_sequence(mpeg2dec_accel_t *accel, uint32_t frame_format, picture_t *picture); +extern int libmpeg2_accel_new_frame(mpeg2dec_accel_t *accel, uint32_t frame_format, picture_t *picture, double ratio, uint32_t flags); +extern void libmpeg2_accel_frame_completion(mpeg2dec_accel_t *accel, uint32_t frame_format, picture_t *picture, int code); + +extern int libmpeg2_accel_slice(mpeg2dec_accel_t *accel, picture_t *picture, int code, + char * buffer, uint32_t chunk_size, uint8_t *chunk_buffer); +extern void libmpeg2_accel_scan( mpeg2dec_accel_t *accel, uint8_t *scan_norm, uint8_t *scan_alt); + +#endif diff --git a/src/video_dec/libmpeg2/motion_comp.c b/src/video_dec/libmpeg2/motion_comp.c new file mode 100644 index 000000000..4a324b8ea --- /dev/null +++ b/src/video_dec/libmpeg2/motion_comp.c @@ -0,0 +1,154 @@ +/* + * motion_comp.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#include <stdio.h> +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include <xine/xineutils.h> + +mpeg2_mc_t mpeg2_mc; + +void mpeg2_mc_init (uint32_t mm_accel) +{ +#ifdef LIBMPEG2_MLIB + if (mm_accel & MM_ACCEL_MLIB) { +#ifdef LOG + fprintf (stderr, "Using mediaLib for motion compensation\n"); +#endif + mpeg2_mc = mpeg2_mc_mlib; + } +#endif + +#if defined(ARCH_X86) || defined(ARCH_X86_64) + if (mm_accel & MM_ACCEL_X86_MMXEXT) { +#ifdef LOG + fprintf (stderr, "Using MMXEXT for motion compensation\n"); +#endif + mpeg2_mc = mpeg2_mc_mmxext; + } else if (mm_accel & MM_ACCEL_X86_3DNOW) { +#ifdef LOG + fprintf (stderr, "Using 3DNOW for motion compensation\n"); +#endif + mpeg2_mc = mpeg2_mc_3dnow; + } else if (mm_accel & MM_ACCEL_X86_MMX) { +#ifdef LOG + fprintf (stderr, "Using MMX for motion compensation\n"); +#endif + mpeg2_mc = mpeg2_mc_mmx; + } else +#endif +#if defined (ARCH_PPC) && defined (ENABLE_ALTIVEC) + if (mm_accel & MM_ACCEL_PPC_ALTIVEC) { +#ifdef LOG + fprintf (stderr, "Using altivec for motion compensation\n"); +#endif + mpeg2_mc = mpeg2_mc_altivec; + } else +#endif +#if defined(ARCH_SPARC) && defined(ENABLE_VIS) + if (mm_accel & MM_ACCEL_SPARC_VIS) { +#ifdef LOG + fprintf (stderr, "Using VIS for motion compensation\n"); +#endif + mpeg2_mc = mpeg2_mc_vis; + } else +#endif + { +#ifdef LOG + fprintf (stderr, "No accelerated motion compensation found\n"); +#endif + mpeg2_mc = mpeg2_mc_c; + } +} + +#define avg2(a,b) ((a+b+1)>>1) +#define avg4(a,b,c,d) ((a+b+c+d+2)>>2) + +#define predict_o(i) (ref[i]) +#define predict_x(i) (avg2 (ref[i], ref[i+1])) +#define predict_y(i) (avg2 (ref[i], (ref+stride)[i])) +#define predict_xy(i) (avg4 (ref[i], ref[i+1], \ + (ref+stride)[i], (ref+stride)[i+1])) + +#define put(predictor,i) dest[i] = predictor (i) +#define avg(predictor,i) dest[i] = avg2 (predictor (i), dest[i]) + +/* mc function template */ + +#define MC_FUNC(op,xy) \ +static void MC_##op##_##xy##_16_c (uint8_t * dest, uint8_t * ref, \ + int stride, int height) \ +{ \ + do { \ + op (predict_##xy, 0); \ + op (predict_##xy, 1); \ + op (predict_##xy, 2); \ + op (predict_##xy, 3); \ + op (predict_##xy, 4); \ + op (predict_##xy, 5); \ + op (predict_##xy, 6); \ + op (predict_##xy, 7); \ + op (predict_##xy, 8); \ + op (predict_##xy, 9); \ + op (predict_##xy, 10); \ + op (predict_##xy, 11); \ + op (predict_##xy, 12); \ + op (predict_##xy, 13); \ + op (predict_##xy, 14); \ + op (predict_##xy, 15); \ + ref += stride; \ + dest += stride; \ + } while (--height); \ +} \ +static void MC_##op##_##xy##_8_c (uint8_t * dest, uint8_t * ref, \ + int stride, int height) \ +{ \ + do { \ + op (predict_##xy, 0); \ + op (predict_##xy, 1); \ + op (predict_##xy, 2); \ + op (predict_##xy, 3); \ + op (predict_##xy, 4); \ + op (predict_##xy, 5); \ + op (predict_##xy, 6); \ + op (predict_##xy, 7); \ + ref += stride; \ + dest += stride; \ + } while (--height); \ +} + +/* definitions of the actual mc functions */ + +MC_FUNC (put,o) +MC_FUNC (avg,o) +MC_FUNC (put,x) +MC_FUNC (avg,x) +MC_FUNC (put,y) +MC_FUNC (avg,y) +MC_FUNC (put,xy) +MC_FUNC (avg,xy) + +MPEG2_MC_EXTERN (c) diff --git a/src/video_dec/libmpeg2/motion_comp_altivec.c b/src/video_dec/libmpeg2/motion_comp_altivec.c new file mode 100644 index 000000000..99719b7fb --- /dev/null +++ b/src/video_dec/libmpeg2/motion_comp_altivec.c @@ -0,0 +1,2031 @@ +/* + * motion_comp_altivec.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#ifndef HOST_OS_DARWIN + +#if defined (ARCH_PPC) && defined (ENABLE_ALTIVEC) + +#include "mpeg2_internal.h" + +#include <inttypes.h> + +/* + * The asm code is generated with: + * + * gcc-2.95 -fvec -DHOST_OS_DARWIN -O9 -fomit-frame-pointer -mregnames -S + * motion_comp_altivec.c + * + * sed 's/.L/._L/g' motion_comp_altivec.s | + * awk '{args=""; len=split ($2, arg, ","); + * for (i=1; i<=len; i++) { a=arg[i]; if (i<len) a=a","; + * args = args sprintf ("%-6s", a) } + * printf ("\t\"\t%-16s%-24s\\n\"\n", $1, args) }' | + * unexpand -a + */ + +static void MC_put_o_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " srawi %r6, %r6, 1 \n" + " li %r9, 15 \n" + " addi %r6, %r6, -1 \n" + " lvsl %v12, 0, %r4 \n" + " mtctr %r6 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " add %r0, %r5, %r5 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + "._L6: \n" + " li %r9, 15 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " stvx %v13, %r5, %r3 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r0 \n" + " bdnz ._L6 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " stvx %v13, %r5, %r3 \n" + ); +} + +static void MC_put_o_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v12, 0, %r4 \n" + " lvsl %v1, %r5, %r4 \n" + " vmrghb %v12, %v12, %v12 \n" + " srawi %r6, %r6, 1 \n" + " li %r9, 7 \n" + " vmrghb %v1, %v1, %v1 \n" + " addi %r6, %r6, -1 \n" + " vpkuhum %v10, %v12, %v12 \n" + " lvx %v13, 0, %r4 \n" + " mtctr %r6 \n" + " vpkuhum %v11, %v1, %v1 \n" + " lvx %v0, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v12, %v13, %v0, %v10 \n" + "._L11: \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " stvewx %v12, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v13, %v0, %v11 \n" + " stvewx %v12, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v1, 0, %r3 \n" + " vperm %v12, %v13, %v0, %v10 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L11 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " stvewx %v12, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v13, %v0, %v11 \n" + " stvewx %v12, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v1, 0, %r3 \n" + " stvewx %v1, %r9, %r3 \n" + ); +} + +static void MC_put_x_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v11, 0, %r4 \n" + " vspltisb %v0, 1 \n" + " li %r9, 16 \n" + " lvx %v12, 0, %r4 \n" + " vaddubm %v10, %v11, %v0 \n" + " lvx %v13, %r9, %r4 \n" + " srawi %r6, %r6, 1 \n" + " addi %r6, %r6, -1 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " mtctr %r6 \n" + " add %r0, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + "._L16: \n" + " li %r9, 16 \n" + " lvx %v12, 0, %r4 \n" + " lvx %v13, %r9, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " lvx %v12, 0, %r4 \n" + " lvx %v13, %r9, %r4 \n" + " vavgub %v0, %v0, %v1 \n" + " stvx %v0, %r5, %r3 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " add %r3, %r3, %r0 \n" + " vavgub %v0, %v0, %v1 \n" + " bdnz ._L16 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " vavgub %v0, %v0, %v1 \n" + " stvx %v0, %r5, %r3 \n" + ); +} + +static void MC_put_x_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v0, 0, %r4 \n" + " vspltisb %v13, 1 \n" + " lvsl %v10, %r5, %r4 \n" + " vmrghb %v0, %v0, %v0 \n" + " li %r9, 8 \n" + " lvx %v11, 0, %r4 \n" + " vmrghb %v10, %v10, %v10 \n" + " vpkuhum %v8, %v0, %v0 \n" + " lvx %v12, %r9, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vpkuhum %v9, %v10, %v10 \n" + " vaddubm %v7, %v8, %v13 \n" + " addi %r6, %r6, -1 \n" + " vperm %v1, %v11, %v12, %v8 \n" + " mtctr %r6 \n" + " vaddubm %v13, %v9, %v13 \n" + " add %r4, %r4, %r5 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " vavgub %v0, %v1, %v0 \n" + "._L21: \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " stvewx %v0, %r9, %r3 \n" + " vperm %v0, %v11, %v12, %v9 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " vavgub %v10, %v0, %v1 \n" + " lvx %v11, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v10, 0, %r3 \n" + " vperm %v1, %v11, %v12, %v7 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " li %r9, 4 \n" + " stvewx %v10, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L21 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " stvewx %v0, %r9, %r3 \n" + " vperm %v0, %v11, %v12, %v9 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v10, %v0, %v1 \n" + " stvewx %v10, 0, %r3 \n" + " stvewx %v10, %r9, %r3 \n" + ); +} + +static void MC_put_y_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " li %r9, 15 \n" + " lvsl %v10, 0, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v12, %v13, %v1, %v10 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " addi %r6, %r6, -1 \n" + " vperm %v11, %v13, %v1, %v10 \n" + " mtctr %r6 \n" + " add %r0, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v12, %v11 \n" + "._L26: \n" + " li %r9, 15 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v12, %v13, %v1, %v10 \n" + " add %r4, %r4, %r5 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " vavgub %v0, %v12, %v11 \n" + " stvx %v0, %r5, %r3 \n" + " vperm %v11, %v13, %v1, %v10 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r0 \n" + " vavgub %v0, %v12, %v11 \n" + " bdnz ._L26 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v12, %v13, %v1, %v10 \n" + " vavgub %v0, %v12, %v11 \n" + " stvx %v0, %r5, %r3 \n" + ); +} + +static void MC_put_y_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v13, 0, %r4 \n" + " lvsl %v11, %r5, %r4 \n" + " vmrghb %v13, %v13, %v13 \n" + " li %r9, 7 \n" + " lvx %v12, 0, %r4 \n" + " vmrghb %v11, %v11, %v11 \n" + " lvx %v1, %r9, %r4 \n" + " vpkuhum %v9, %v13, %v13 \n" + " add %r4, %r4, %r5 \n" + " vpkuhum %v10, %v11, %v11 \n" + " vperm %v13, %v12, %v1, %v9 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v12, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " addi %r6, %r6, -1 \n" + " vperm %v11, %v12, %v1, %v10 \n" + " mtctr %r6 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v13, %v11 \n" + "._L31: \n" + " li %r9, 7 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v13, %v12, %v1, %v9 \n" + " stvewx %v0, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v13, %v11 \n" + " li %r9, 7 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v0, 0, %r3 \n" + " vperm %v11, %v12, %v1, %v10 \n" + " li %r9, 4 \n" + " stvewx %v0, %r9, %r3 \n" + " vavgub %v0, %v13, %v11 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L31 \n" + " li %r9, 7 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v13, %v12, %v1, %v9 \n" + " stvewx %v0, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v0, %v13, %v11 \n" + " stvewx %v0, 0, %r3 \n" + " stvewx %v0, %r9, %r3 \n" + ); +} + +static void MC_put_xy_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v5, 0, %r4 \n" + " vspltisb %v3, 1 \n" + " li %r9, 16 \n" + " lvx %v1, 0, %r4 \n" + " vaddubm %v4, %v5, %v3 \n" + " lvx %v0, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " srawi %r6, %r6, 1 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " addi %r6, %r6, -1 \n" + " lvx %v1, 0, %r4 \n" + " mtctr %r6 \n" + " lvx %v0, %r9, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r0, %r5, %r5 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vxor %v6, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + "._L36: \n" + " li %r9, 16 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " vavgub %v12, %v9, %v7 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vor %v13, %v8, %v6 \n" + " vxor %v0, %v9, %v7 \n" + " vxor %v6, %v11, %v10 \n" + " vand %v13, %v3, %v13 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v6 \n" + " vand %v13, %v13, %v0 \n" + " vxor %v0, %v9, %v7 \n" + " vand %v1, %v3, %v1 \n" + " vsububm %v13, %v12, %v13 \n" + " vand %v1, %v1, %v0 \n" + " stvx %v13, %r5, %r3 \n" + " vavgub %v0, %v9, %v7 \n" + " add %r3, %r3, %r0 \n" + " vsububm %v13, %v0, %v1 \n" + " bdnz ._L36 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vxor %v8, %v11, %v10 \n" + " vavgub %v9, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + " stvx %v13, %r5, %r3 \n" + ); +} + +static void MC_put_xy_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v4, 0, %r4 \n" + " vspltisb %v3, 1 \n" + " lvsl %v5, %r5, %r4 \n" + " vmrghb %v4, %v4, %v4 \n" + " li %r9, 16 \n" + " vmrghb %v5, %v5, %v5 \n" + " lvx %v1, 0, %r4 \n" + " vpkuhum %v4, %v4, %v4 \n" + " lvx %v0, %r9, %r4 \n" + " vpkuhum %v5, %v5, %v5 \n" + " add %r4, %r4, %r5 \n" + " vaddubm %v2, %v4, %v3 \n" + " vperm %v11, %v1, %v0, %v4 \n" + " srawi %r6, %r6, 1 \n" + " vaddubm %v19, %v5, %v3 \n" + " addi %r6, %r6, -1 \n" + " vperm %v10, %v1, %v0, %v2 \n" + " mtctr %r6 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v1, %v0, %v19 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vxor %v6, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + "._L41: \n" + " li %r9, 16 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvewx %v13, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v10, %v1, %v0, %v2 \n" + " stvewx %v13, %r9, %r3 \n" + " vperm %v11, %v1, %v0, %v4 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 16 \n" + " vavgub %v9, %v11, %v10 \n" + " lvx %v0, %r9, %r4 \n" + " vxor %v8, %v11, %v10 \n" + " lvx %v1, 0, %r4 \n" + " vavgub %v12, %v9, %v7 \n" + " vor %v13, %v8, %v6 \n" + " add %r3, %r3, %r5 \n" + " vperm %v10, %v1, %v0, %v19 \n" + " li %r9, 4 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vand %v13, %v3, %v13 \n" + " add %r4, %r4, %r5 \n" + " vxor %v0, %v9, %v7 \n" + " vxor %v6, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v6 \n" + " vand %v13, %v13, %v0 \n" + " vxor %v0, %v9, %v7 \n" + " vand %v1, %v3, %v1 \n" + " vsububm %v13, %v12, %v13 \n" + " vand %v1, %v1, %v0 \n" + " stvewx %v13, 0, %r3 \n" + " vavgub %v0, %v9, %v7 \n" + " stvewx %v13, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vsububm %v13, %v0, %v1 \n" + " bdnz ._L41 \n" + " li %r9, 16 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvewx %v13, 0, %r3 \n" + " vperm %v10, %v1, %v0, %v2 \n" + " li %r9, 4 \n" + " vperm %v11, %v1, %v0, %v4 \n" + " stvewx %v13, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vxor %v8, %v11, %v10 \n" + " vavgub %v9, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + " stvewx %v13, 0, %r3 \n" + " stvewx %v13, %r9, %r3 \n" + ); +} + +static void MC_avg_o_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " li %r9, 15 \n" + " lvx %v0, %r9, %r4 \n" + " lvsl %v11, 0, %r4 \n" + " lvx %v1, 0, %r4 \n" + " srawi %r6, %r6, 1 \n" + " addi %r6, %r6, -1 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " lvx %v13, 0, %r3 \n" + " mtctr %r6 \n" + " add %r9, %r5, %r5 \n" + " vavgub %v12, %v13, %v0 \n" + " add %r4, %r4, %r5 \n" + "._L46: \n" + " li %r11, 15 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r11, %r4 \n" + " lvx %v13, %r5, %r3 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " stvx %v12, 0, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v12, %v13, %v0 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r11, %r4 \n" + " lvx %v13, %r9, %r3 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " stvx %v12, %r5, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v12, %v13, %v0 \n" + " add %r3, %r3, %r9 \n" + " bdnz ._L46 \n" + " lvx %v0, %r11, %r4 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v13, %r5, %r3 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " stvx %v12, 0, %r3 \n" + " vavgub %v12, %v13, %v0 \n" + " stvx %v12, %r5, %r3 \n" + ); +} + +static void MC_avg_o_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v12, 0, %r4 \n" + " li %r9, 7 \n" + " vmrghb %v12, %v12, %v12 \n" + " lvsl %v1, %r5, %r4 \n" + " lvx %v13, 0, %r4 \n" + " vpkuhum %v9, %v12, %v12 \n" + " lvx %v0, %r9, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vmrghb %v1, %v1, %v1 \n" + " addi %r6, %r6, -1 \n" + " vperm %v0, %v13, %v0, %v9 \n" + " lvx %v11, 0, %r3 \n" + " mtctr %r6 \n" + " vpkuhum %v10, %v1, %v1 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v12, %v11, %v0 \n" + "._L51: \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v11, %r5, %r3 \n" + " stvewx %v12, 0, %r3 \n" + " vperm %v0, %v13, %v0, %v10 \n" + " li %r9, 4 \n" + " stvewx %v12, %r9, %r3 \n" + " vavgub %v1, %v11, %v0 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " add %r3, %r3, %r5 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v11, %r5, %r3 \n" + " stvewx %v1, 0, %r3 \n" + " vperm %v0, %v13, %v0, %v9 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " vavgub %v12, %v11, %v0 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L51 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v11, %r5, %r3 \n" + " stvewx %v12, 0, %r3 \n" + " vperm %v0, %v13, %v0, %v10 \n" + " li %r9, 4 \n" + " stvewx %v12, %r9, %r3 \n" + " vavgub %v1, %v11, %v0 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v1, 0, %r3 \n" + " stvewx %v1, %r9, %r3 \n" + ); +} + +static void MC_avg_x_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v8, 0, %r4 \n" + " vspltisb %v0, 1 \n" + " li %r9, 16 \n" + " lvx %v12, %r9, %r4 \n" + " vaddubm %v7, %v8, %v0 \n" + " lvx %v11, 0, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vperm %v1, %v11, %v12, %v7 \n" + " addi %r6, %r6, -1 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " lvx %v9, 0, %r3 \n" + " mtctr %r6 \n" + " add %r9, %r5, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v10, %v9, %v0 \n" + "._L56: \n" + " li %r11, 16 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v12, %r11, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " stvx %v10, 0, %r3 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " add %r4, %r4, %r5 \n" + " vperm %v1, %v11, %v12, %v8 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v12, %r11, %r4 \n" + " vavgub %v1, %v1, %v0 \n" + " add %r4, %r4, %r5 \n" + " vperm %v13, %v11, %v12, %v7 \n" + " vavgub %v10, %v9, %v1 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " lvx %v9, %r9, %r3 \n" + " stvx %v10, %r5, %r3 \n" + " vavgub %v0, %v0, %v13 \n" + " add %r3, %r3, %r9 \n" + " vavgub %v10, %v9, %v0 \n" + " bdnz ._L56 \n" + " lvx %v12, %r11, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " vperm %v1, %v11, %v12, %v7 \n" + " stvx %v10, 0, %r3 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " vavgub %v0, %v0, %v1 \n" + " vavgub %v10, %v9, %v0 \n" + " stvx %v10, %r5, %r3 \n" + ); +} + +static void MC_avg_x_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v10, 0, %r4 \n" + " vspltisb %v13, 1 \n" + " li %r9, 8 \n" + " vmrghb %v10, %v10, %v10 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v12, %r9, %r4 \n" + " vpkuhum %v7, %v10, %v10 \n" + " srawi %r6, %r6, 1 \n" + " lvsl %v10, %r5, %r4 \n" + " vaddubm %v6, %v7, %v13 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " addi %r6, %r6, -1 \n" + " vmrghb %v10, %v10, %v10 \n" + " lvx %v9, 0, %r3 \n" + " mtctr %r6 \n" + " vperm %v1, %v11, %v12, %v6 \n" + " add %r4, %r4, %r5 \n" + " vpkuhum %v8, %v10, %v10 \n" + " vavgub %v0, %v0, %v1 \n" + " vaddubm %v13, %v8, %v13 \n" + " vavgub %v10, %v9, %v0 \n" + "._L61: \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " stvewx %v10, 0, %r3 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " li %r9, 4 \n" + " stvewx %v10, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " vavgub %v10, %v9, %v0 \n" + " lvx %v11, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " vperm %v1, %v11, %v12, %v6 \n" + " lvx %v9, %r5, %r3 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " stvewx %v10, 0, %r3 \n" + " li %r9, 4 \n" + " vavgub %v0, %v0, %v1 \n" + " stvewx %v10, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v10, %v9, %v0 \n" + " bdnz ._L61 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " stvewx %v10, 0, %r3 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " li %r9, 4 \n" + " stvewx %v10, %r9, %r3 \n" + " vavgub %v0, %v0, %v1 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v10, %v9, %v0 \n" + " stvewx %v10, 0, %r3 \n" + " stvewx %v10, %r9, %r3 \n" + ); +} + +static void MC_avg_y_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " li %r9, 15 \n" + " lvx %v1, %r9, %r4 \n" + " lvsl %v9, 0, %r4 \n" + " lvx %v13, 0, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v11, %v13, %v1, %v9 \n" + " li %r11, 15 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vperm %v10, %v13, %v1, %v9 \n" + " addi %r6, %r6, -1 \n" + " lvx %v12, 0, %r3 \n" + " mtctr %r6 \n" + " vavgub %v0, %v11, %v10 \n" + " add %r9, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v12, %v0 \n" + "._L66: \n" + " li %r11, 15 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v12, %r5, %r3 \n" + " vperm %v11, %v13, %v1, %v9 \n" + " stvx %v0, 0, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v11, %v10 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " vavgub %v0, %v12, %v0 \n" + " add %r4, %r4, %r5 \n" + " lvx %v12, %r9, %r3 \n" + " vperm %v10, %v13, %v1, %v9 \n" + " stvx %v0, %r5, %r3 \n" + " vavgub %v0, %v11, %v10 \n" + " add %r3, %r3, %r9 \n" + " vavgub %v0, %v12, %v0 \n" + " bdnz ._L66 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v12, %r5, %r3 \n" + " vperm %v11, %v13, %v1, %v9 \n" + " stvx %v0, 0, %r3 \n" + " vavgub %v0, %v11, %v10 \n" + " vavgub %v0, %v12, %v0 \n" + " stvx %v0, %r5, %r3 \n" + ); +} + +static void MC_avg_y_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v12, 0, %r4 \n" + " lvsl %v9, %r5, %r4 \n" + " vmrghb %v12, %v12, %v12 \n" + " li %r9, 7 \n" + " lvx %v11, 0, %r4 \n" + " vmrghb %v9, %v9, %v9 \n" + " lvx %v13, %r9, %r4 \n" + " vpkuhum %v7, %v12, %v12 \n" + " add %r4, %r4, %r5 \n" + " vpkuhum %v8, %v9, %v9 \n" + " vperm %v12, %v11, %v13, %v7 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v13, %r9, %r4 \n" + " addi %r6, %r6, -1 \n" + " vperm %v9, %v11, %v13, %v8 \n" + " lvx %v10, 0, %r3 \n" + " mtctr %r6 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v12, %v9 \n" + " vavgub %v1, %v10, %v0 \n" + "._L71: \n" + " li %r9, 7 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v10, %r5, %r3 \n" + " stvewx %v1, 0, %r3 \n" + " vperm %v12, %v11, %v13, %v7 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " vavgub %v0, %v12, %v9 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 7 \n" + " vavgub %v1, %v10, %v0 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " vperm %v9, %v11, %v13, %v8 \n" + " lvx %v10, %r5, %r3 \n" + " stvewx %v1, 0, %r3 \n" + " vavgub %v0, %v12, %v9 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v1, %v10, %v0 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L71 \n" + " li %r9, 7 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v10, %r5, %r3 \n" + " vperm %v12, %v11, %v13, %v7 \n" + " stvewx %v1, 0, %r3 \n" + " li %r9, 4 \n" + " vavgub %v0, %v12, %v9 \n" + " stvewx %v1, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v1, %v10, %v0 \n" + " stvewx %v1, 0, %r3 \n" + " stvewx %v1, %r9, %r3 \n" + ); +} + +static void MC_avg_xy_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v4, 0, %r4 \n" + " vspltisb %v2, 1 \n" + " li %r9, 16 \n" + " lvx %v1, %r9, %r4 \n" + " vaddubm %v3, %v4, %v2 \n" + " lvx %v13, 0, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " li %r11, 16 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " addi %r6, %r6, -1 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " lvx %v6, 0, %r3 \n" + " mtctr %r6 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " add %r9, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vxor %v5, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v5 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v1, %v2, %v1 \n" + " vavgub %v0, %v9, %v7 \n" + " vand %v1, %v1, %v13 \n" + " vsububm %v0, %v0, %v1 \n" + " vavgub %v12, %v6, %v0 \n" + "._L76: \n" + " li %r11, 16 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v6, %r5, %r3 \n" + " stvx %v12, 0, %r3 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " add %r4, %r4, %r5 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " vavgub %v12, %v9, %v7 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " vor %v0, %v8, %v5 \n" + " vxor %v13, %v9, %v7 \n" + " vxor %v5, %v11, %v10 \n" + " vand %v0, %v2, %v0 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v5 \n" + " vand %v0, %v0, %v13 \n" + " vand %v1, %v2, %v1 \n" + " vxor %v13, %v9, %v7 \n" + " vsububm %v12, %v12, %v0 \n" + " vand %v1, %v1, %v13 \n" + " vavgub %v0, %v9, %v7 \n" + " vavgub %v12, %v6, %v12 \n" + " lvx %v6, %r9, %r3 \n" + " vsububm %v0, %v0, %v1 \n" + " stvx %v12, %r5, %r3 \n" + " vavgub %v12, %v6, %v0 \n" + " add %r3, %r3, %r9 \n" + " bdnz ._L76 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v6, %r5, %r3 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " stvx %v12, 0, %r3 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " vxor %v8, %v11, %v10 \n" + " vavgub %v9, %v11, %v10 \n" + " vor %v0, %v8, %v5 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v2, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v1, %v1, %v0 \n" + " vavgub %v12, %v6, %v1 \n" + " stvx %v12, %r5, %r3 \n" + ); +} + +static void MC_avg_xy_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v2, 0, %r4 \n" + " vspltisb %v19, 1 \n" + " lvsl %v3, %r5, %r4 \n" + " vmrghb %v2, %v2, %v2 \n" + " li %r9, 16 \n" + " vmrghb %v3, %v3, %v3 \n" + " lvx %v9, 0, %r4 \n" + " vpkuhum %v2, %v2, %v2 \n" + " lvx %v1, %r9, %r4 \n" + " vpkuhum %v3, %v3, %v3 \n" + " add %r4, %r4, %r5 \n" + " vaddubm %v18, %v2, %v19 \n" + " vperm %v11, %v9, %v1, %v2 \n" + " srawi %r6, %r6, 1 \n" + " vaddubm %v17, %v3, %v19 \n" + " addi %r6, %r6, -1 \n" + " vperm %v10, %v9, %v1, %v18 \n" + " lvx %v4, 0, %r3 \n" + " mtctr %r6 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v9, 0, %r4 \n" + " vavgub %v8, %v11, %v10 \n" + " vxor %v7, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v9, %v1, %v17 \n" + " vperm %v11, %v9, %v1, %v3 \n" + " vxor %v5, %v11, %v10 \n" + " vavgub %v6, %v11, %v10 \n" + " vor %v1, %v7, %v5 \n" + " vxor %v13, %v8, %v6 \n" + " vand %v1, %v19, %v1 \n" + " vavgub %v0, %v8, %v6 \n" + " vand %v1, %v1, %v13 \n" + " vsububm %v0, %v0, %v1 \n" + " vavgub %v13, %v4, %v0 \n" + "._L81: \n" + " li %r9, 16 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v9, 0, %r4 \n" + " lvx %v4, %r5, %r3 \n" + " stvewx %v13, 0, %r3 \n" + " vperm %v10, %v9, %v1, %v18 \n" + " vperm %v11, %v9, %v1, %v2 \n" + " li %r9, 4 \n" + " stvewx %v13, %r9, %r3 \n" + " vxor %v7, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 16 \n" + " vavgub %v8, %v11, %v10 \n" + " lvx %v1, %r9, %r4 \n" + " vor %v0, %v7, %v5 \n" + " lvx %v9, 0, %r4 \n" + " vxor %v12, %v8, %v6 \n" + " vand %v0, %v19, %v0 \n" + " add %r3, %r3, %r5 \n" + " vperm %v10, %v9, %v1, %v17 \n" + " vavgub %v13, %v8, %v6 \n" + " li %r9, 4 \n" + " vperm %v11, %v9, %v1, %v3 \n" + " vand %v0, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + " vxor %v5, %v11, %v10 \n" + " vavgub %v6, %v11, %v10 \n" + " vor %v1, %v7, %v5 \n" + " vsububm %v13, %v13, %v0 \n" + " vxor %v0, %v8, %v6 \n" + " vand %v1, %v19, %v1 \n" + " vavgub %v13, %v4, %v13 \n" + " vand %v1, %v1, %v0 \n" + " lvx %v4, %r5, %r3 \n" + " vavgub %v0, %v8, %v6 \n" + " stvewx %v13, 0, %r3 \n" + " stvewx %v13, %r9, %r3 \n" + " vsububm %v0, %v0, %v1 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v13, %v4, %v0 \n" + " bdnz ._L81 \n" + " li %r9, 16 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v9, 0, %r4 \n" + " lvx %v4, %r5, %r3 \n" + " vperm %v10, %v9, %v1, %v18 \n" + " stvewx %v13, 0, %r3 \n" + " vperm %v11, %v9, %v1, %v2 \n" + " li %r9, 4 \n" + " stvewx %v13, %r9, %r3 \n" + " vxor %v7, %v11, %v10 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v8, %v11, %v10 \n" + " vor %v0, %v7, %v5 \n" + " vxor %v13, %v8, %v6 \n" + " vand %v0, %v19, %v0 \n" + " vavgub %v1, %v8, %v6 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v1, %v1, %v0 \n" + " vavgub %v13, %v4, %v1 \n" + " stvewx %v13, 0, %r3 \n" + " stvewx %v13, %r9, %r3 \n" + ); +} + +MPEG2_MC_EXTERN (altivec) + +#endif /* ARCH_PPC */ + +#else /* HOST_OS_DARWIN */ + +#ifdef ENABLE_ALTIVEC + +#include "mpeg2_internal.h" + +#define vector_s16_t vector signed short +#define vector_u16_t vector unsigned short +#define vector_s8_t vector signed char +#define vector_u8_t vector unsigned char +#define vector_s32_t vector signed int +#define vector_u32_t vector unsigned int + +void MC_put_o_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp = vec_perm (ref0, ref1, perm); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, 0, dest); + tmp = vec_perm (ref0, ref1, perm); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_perm (ref0, ref1, perm); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + vec_st (tmp, 0, dest); + tmp = vec_perm (ref0, ref1, perm); + vec_st (tmp, stride, dest); +} + +void MC_put_o_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, ref0, ref1; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_put_x_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, tmp; + + permA = vec_lvsl (0, ref); + permB = vec_add (permA, vec_splat_u8 (1)); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, 0, dest); + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + vec_st (tmp, 0, dest); + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + vec_st (tmp, stride, dest); +} + +void MC_put_x_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ones, tmp0, tmp1, ref0, ref1; + + ones = vec_splat_u8 (1); + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0A = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + perm0B = vec_add (perm0A, ones); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1A = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + tmp0 = vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B)); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_put_y_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp0, tmp1, tmp; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + vec_st (tmp, stride, dest); +} + +void MC_put_y_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, tmp, ref0, ref1; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (tmp0, tmp1); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (tmp0, tmp1); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (tmp0, tmp1); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (tmp0, tmp1); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +void MC_put_xy_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, A, B, avg0, avg1, xor0, xor1, tmp; + vector_u8_t ones; + + ones = vec_splat_u8 (1); + permA = vec_lvsl (0, ref); + permB = vec_add (permA, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + vec_st (tmp, stride, dest); +} + +void MC_put_xy_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ref0, ref1, A, B; + vector_u8_t avg0, avg1, xor0, xor1, tmp, ones; + + ones = vec_splat_u8 (1); + perm0A = vec_lvsl (0, ref); + perm0A = vec_mergeh (perm0A, perm0A); + perm0A = vec_pack ((vector_u16_t)perm0A, (vector_u16_t)perm0A); + perm0B = vec_add (perm0A, ones); + perm1A = vec_lvsl (stride, ref); + perm1A = vec_mergeh (perm1A, perm1A); + perm1A = vec_pack ((vector_u16_t)perm1A, (vector_u16_t)perm1A); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +#if 0 +void MC_put_xy_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, A, B, C, D, tmp, zero, ones; + vector_u16_t splat2, temp; + + ones = vec_splat_u8 (1); + permA = vec_lvsl (0, ref); + permB = vec_add (permA, ones); + + zero = vec_splat_u8 (0); + splat2 = vec_splat_u16 (2); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + C = vec_perm (ref0, ref1, permA); + D = vec_perm (ref0, ref1, permB); + + temp = vec_add (vec_add ((vector_u16_t)vec_mergeh (zero, A), + (vector_u16_t)vec_mergeh (zero, B)), + vec_add ((vector_u16_t)vec_mergeh (zero, C), + (vector_u16_t)vec_mergeh (zero, D))); + temp = vec_sr (vec_add (temp, splat2), splat2); + tmp = vec_pack (temp, temp); + + vec_st (tmp, 0, dest); + dest += stride; + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + } while (--height); +} +#endif + +void MC_avg_o_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp, prev; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + vec_st (tmp, stride, dest); +} + +void MC_avg_o_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, ref0, ref1, prev; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp0 = vec_avg (prev, vec_perm (ref0, ref1, perm0)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_perm (ref0, ref1, perm1)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_avg (prev, vec_perm (ref0, ref1, perm0)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_perm (ref0, ref1, perm1)); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_avg_x_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, tmp, prev; + + permA = vec_lvsl (0, ref); + permB = vec_add (permA, vec_splat_u8 (1)); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (0, dest); + ref += stride; + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + vec_st (tmp, stride, dest); +} + +void MC_avg_x_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ones, tmp0, tmp1, ref0, ref1; + vector_u8_t prev; + + ones = vec_splat_u8 (1); + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0A = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + perm0B = vec_add (perm0A, ones); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1A = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + prev = vec_ld (0, dest); + ref += stride; + tmp0 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B))); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B))); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_avg_y_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp0, tmp1, tmp, prev; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + vec_st (tmp, stride, dest); +} + +void MC_avg_y_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, tmp, ref0, ref1, prev; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +void MC_avg_xy_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, A, B, avg0, avg1, xor0, xor1, tmp; + vector_u8_t ones, prev; + + ones = vec_splat_u8 (1); + permA = vec_lvsl (0, ref); + permB = vec_add (permA, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + vec_st (tmp, stride, dest); +} + +void MC_avg_xy_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ref0, ref1, A, B; + vector_u8_t avg0, avg1, xor0, xor1, tmp, ones, prev; + + ones = vec_splat_u8 (1); + perm0A = vec_lvsl (0, ref); + perm0A = vec_mergeh (perm0A, perm0A); + perm0A = vec_pack ((vector_u16_t)perm0A, (vector_u16_t)perm0A); + perm0B = vec_add (perm0A, ones); + perm1A = vec_lvsl (stride, ref); + perm1A = vec_mergeh (perm1A, perm1A); + perm1A = vec_pack ((vector_u16_t)perm1A, (vector_u16_t)perm1A); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (0, dest); + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +MPEG2_MC_EXTERN (altivec) + +#endif /* ENABLE_ALTIVEC */ + +#endif /* HOST_OS_DARWIN */ + diff --git a/src/video_dec/libmpeg2/motion_comp_mlib.c b/src/video_dec/libmpeg2/motion_comp_mlib.c new file mode 100644 index 000000000..1a37070ae --- /dev/null +++ b/src/video_dec/libmpeg2/motion_comp_mlib.c @@ -0,0 +1,181 @@ +/* + * motion_comp_mlib.c + * Copyright (C) 2000-2002 Håkan Hjort <d95hjort@dtek.chalmers.se> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#ifdef LIBMPEG2_MLIB + +#include <mlib_types.h> +#include <mlib_status.h> +#include <mlib_sys.h> +#include <mlib_video.h> +#include <inttypes.h> + +#include "mpeg2_internal.h" + +static void MC_put_o_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoCopyRef_U8_U8_16x16 (dest, ref, stride); + else + mlib_VideoCopyRef_U8_U8_16x8 (dest, ref, stride); +} + +static void MC_put_x_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoInterpX_U8_U8_16x16 (dest, ref, stride, stride); + else + mlib_VideoInterpX_U8_U8_16x8 (dest, ref, stride, stride); +} + +static void MC_put_y_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoInterpY_U8_U8_16x16 (dest, ref, stride, stride); + else + mlib_VideoInterpY_U8_U8_16x8 (dest, ref, stride, stride); +} + +static void MC_put_xy_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoInterpXY_U8_U8_16x16 (dest, ref, stride, stride); + else + mlib_VideoInterpXY_U8_U8_16x8 (dest, ref, stride, stride); +} + +static void MC_put_o_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoCopyRef_U8_U8_8x8 (dest, ref, stride); + else + mlib_VideoCopyRef_U8_U8_8x4 (dest, ref, stride); +} + +static void MC_put_x_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoInterpX_U8_U8_8x8 (dest, ref, stride, stride); + else + mlib_VideoInterpX_U8_U8_8x4 (dest, ref, stride, stride); +} + +static void MC_put_y_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoInterpY_U8_U8_8x8 (dest, ref, stride, stride); + else + mlib_VideoInterpY_U8_U8_8x4 (dest, ref, stride, stride); +} + +static void MC_put_xy_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoInterpXY_U8_U8_8x8 (dest, ref, stride, stride); + else + mlib_VideoInterpXY_U8_U8_8x4 (dest, ref, stride, stride); +} + +static void MC_avg_o_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoCopyRefAve_U8_U8_16x16 (dest, ref, stride); + else + mlib_VideoCopyRefAve_U8_U8_16x8 (dest, ref, stride); +} + +static void MC_avg_x_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoInterpAveX_U8_U8_16x16 (dest, ref, stride, stride); + else + mlib_VideoInterpAveX_U8_U8_16x8 (dest, ref, stride, stride); +} + +static void MC_avg_y_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoInterpAveY_U8_U8_16x16 (dest, ref, stride, stride); + else + mlib_VideoInterpAveY_U8_U8_16x8 (dest, ref, stride, stride); +} + +static void MC_avg_xy_16_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 16) + mlib_VideoInterpAveXY_U8_U8_16x16 (dest, ref, stride, stride); + else + mlib_VideoInterpAveXY_U8_U8_16x8 (dest, ref, stride, stride); +} + +static void MC_avg_o_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoCopyRefAve_U8_U8_8x8 (dest, ref, stride); + else + mlib_VideoCopyRefAve_U8_U8_8x4 (dest, ref, stride); +} + +static void MC_avg_x_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoInterpAveX_U8_U8_8x8 (dest, ref, stride, stride); + else + mlib_VideoInterpAveX_U8_U8_8x4 (dest, ref, stride, stride); +} + +static void MC_avg_y_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoInterpAveY_U8_U8_8x8 (dest, ref, stride, stride); + else + mlib_VideoInterpAveY_U8_U8_8x4 (dest, ref, stride, stride); +} + +static void MC_avg_xy_8_mlib (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + if (height == 8) + mlib_VideoInterpAveXY_U8_U8_8x8 (dest, ref, stride, stride); + else + mlib_VideoInterpAveXY_U8_U8_8x4 (dest, ref, stride, stride); +} + +MPEG2_MC_EXTERN (mlib) + +#endif diff --git a/src/video_dec/libmpeg2/motion_comp_mmx.c b/src/video_dec/libmpeg2/motion_comp_mmx.c new file mode 100644 index 000000000..9c5ab455d --- /dev/null +++ b/src/video_dec/libmpeg2/motion_comp_mmx.c @@ -0,0 +1,1013 @@ +/* + * motion_comp_mmx.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#if defined(ARCH_X86) || defined(ARCH_X86_64) + +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include <xine/xineutils.h> +#include "xine_mmx.h" + +#define CPU_MMXEXT 0 +#define CPU_3DNOW 1 + + +/* MMX code - needs a rewrite */ + +/* some rounding constants */ +static mmx_t round1 = {0x0001000100010001LL}; +static mmx_t round4 = {0x0002000200020002LL}; + +/* + * This code should probably be compiled with loop unrolling + * (ie, -funroll-loops in gcc)becuase some of the loops + * use a small static number of iterations. This was written + * with the assumption the compiler knows best about when + * unrolling will help + */ + +static inline void mmx_zero_reg () +{ + /* load 0 into mm0 */ + pxor_r2r (mm0, mm0); +} + +static inline void mmx_average_2_U8 (uint8_t * dest, + uint8_t * src1, uint8_t * src2) +{ + /* *dest = (*src1 + *src2 + 1)/ 2; */ + + movq_m2r (*src1, mm1); // load 8 src1 bytes + movq_r2r (mm1, mm2); // copy 8 src1 bytes + + movq_m2r (*src2, mm3); // load 8 src2 bytes + movq_r2r (mm3, mm4); // copy 8 src2 bytes + + punpcklbw_r2r (mm0, mm1); // unpack low src1 bytes + punpckhbw_r2r (mm0, mm2); // unpack high src1 bytes + + punpcklbw_r2r (mm0, mm3); // unpack low src2 bytes + punpckhbw_r2r (mm0, mm4); // unpack high src2 bytes + + paddw_r2r (mm3, mm1); // add lows to mm1 + paddw_m2r (round1, mm1); + psraw_i2r (1, mm1); // /2 + + paddw_r2r (mm4, mm2); // add highs to mm2 + paddw_m2r (round1, mm2); + psraw_i2r (1, mm2); // /2 + + packuswb_r2r (mm2, mm1); // pack (w/ saturation) + movq_r2m (mm1, *dest); // store result in dest +} + +static inline void mmx_interp_average_2_U8 (uint8_t * dest, + uint8_t * src1, uint8_t * src2) +{ + /* *dest = (*dest + (*src1 + *src2 + 1)/ 2 + 1)/ 2; */ + + movq_m2r (*dest, mm1); // load 8 dest bytes + movq_r2r (mm1, mm2); // copy 8 dest bytes + + movq_m2r (*src1, mm3); // load 8 src1 bytes + movq_r2r (mm3, mm4); // copy 8 src1 bytes + + movq_m2r (*src2, mm5); // load 8 src2 bytes + movq_r2r (mm5, mm6); // copy 8 src2 bytes + + punpcklbw_r2r (mm0, mm1); // unpack low dest bytes + punpckhbw_r2r (mm0, mm2); // unpack high dest bytes + + punpcklbw_r2r (mm0, mm3); // unpack low src1 bytes + punpckhbw_r2r (mm0, mm4); // unpack high src1 bytes + + punpcklbw_r2r (mm0, mm5); // unpack low src2 bytes + punpckhbw_r2r (mm0, mm6); // unpack high src2 bytes + + paddw_r2r (mm5, mm3); // add lows + paddw_m2r (round1, mm3); + psraw_i2r (1, mm3); // /2 + + paddw_r2r (mm6, mm4); // add highs + paddw_m2r (round1, mm4); + psraw_i2r (1, mm4); // /2 + + paddw_r2r (mm3, mm1); // add lows + paddw_m2r (round1, mm1); + psraw_i2r (1, mm1); // /2 + + paddw_r2r (mm4, mm2); // add highs + paddw_m2r (round1, mm2); + psraw_i2r (1, mm2); // /2 + + packuswb_r2r (mm2, mm1); // pack (w/ saturation) + movq_r2m (mm1, *dest); // store result in dest +} + +static inline void mmx_average_4_U8 (uint8_t * dest, + uint8_t * src1, uint8_t * src2, + uint8_t * src3, uint8_t * src4) +{ + /* *dest = (*src1 + *src2 + *src3 + *src4 + 2)/ 4; */ + + movq_m2r (*src1, mm1); // load 8 src1 bytes + movq_r2r (mm1, mm2); // copy 8 src1 bytes + + punpcklbw_r2r (mm0, mm1); // unpack low src1 bytes + punpckhbw_r2r (mm0, mm2); // unpack high src1 bytes + + movq_m2r (*src2, mm3); // load 8 src2 bytes + movq_r2r (mm3, mm4); // copy 8 src2 bytes + + punpcklbw_r2r (mm0, mm3); // unpack low src2 bytes + punpckhbw_r2r (mm0, mm4); // unpack high src2 bytes + + paddw_r2r (mm3, mm1); // add lows + paddw_r2r (mm4, mm2); // add highs + + /* now have partials in mm1 and mm2 */ + + movq_m2r (*src3, mm3); // load 8 src3 bytes + movq_r2r (mm3, mm4); // copy 8 src3 bytes + + punpcklbw_r2r (mm0, mm3); // unpack low src3 bytes + punpckhbw_r2r (mm0, mm4); // unpack high src3 bytes + + paddw_r2r (mm3, mm1); // add lows + paddw_r2r (mm4, mm2); // add highs + + movq_m2r (*src4, mm5); // load 8 src4 bytes + movq_r2r (mm5, mm6); // copy 8 src4 bytes + + punpcklbw_r2r (mm0, mm5); // unpack low src4 bytes + punpckhbw_r2r (mm0, mm6); // unpack high src4 bytes + + paddw_r2r (mm5, mm1); // add lows + paddw_r2r (mm6, mm2); // add highs + + /* now have subtotal in mm1 and mm2 */ + + paddw_m2r (round4, mm1); + psraw_i2r (2, mm1); // /4 + paddw_m2r (round4, mm2); + psraw_i2r (2, mm2); // /4 + + packuswb_r2r (mm2, mm1); // pack (w/ saturation) + movq_r2m (mm1, *dest); // store result in dest +} + +static inline void mmx_interp_average_4_U8 (uint8_t * dest, + uint8_t * src1, uint8_t * src2, + uint8_t * src3, uint8_t * src4) +{ + /* *dest = (*dest + (*src1 + *src2 + *src3 + *src4 + 2)/ 4 + 1)/ 2; */ + + movq_m2r (*src1, mm1); // load 8 src1 bytes + movq_r2r (mm1, mm2); // copy 8 src1 bytes + + punpcklbw_r2r (mm0, mm1); // unpack low src1 bytes + punpckhbw_r2r (mm0, mm2); // unpack high src1 bytes + + movq_m2r (*src2, mm3); // load 8 src2 bytes + movq_r2r (mm3, mm4); // copy 8 src2 bytes + + punpcklbw_r2r (mm0, mm3); // unpack low src2 bytes + punpckhbw_r2r (mm0, mm4); // unpack high src2 bytes + + paddw_r2r (mm3, mm1); // add lows + paddw_r2r (mm4, mm2); // add highs + + /* now have partials in mm1 and mm2 */ + + movq_m2r (*src3, mm3); // load 8 src3 bytes + movq_r2r (mm3, mm4); // copy 8 src3 bytes + + punpcklbw_r2r (mm0, mm3); // unpack low src3 bytes + punpckhbw_r2r (mm0, mm4); // unpack high src3 bytes + + paddw_r2r (mm3, mm1); // add lows + paddw_r2r (mm4, mm2); // add highs + + movq_m2r (*src4, mm5); // load 8 src4 bytes + movq_r2r (mm5, mm6); // copy 8 src4 bytes + + punpcklbw_r2r (mm0, mm5); // unpack low src4 bytes + punpckhbw_r2r (mm0, mm6); // unpack high src4 bytes + + paddw_r2r (mm5, mm1); // add lows + paddw_r2r (mm6, mm2); // add highs + + paddw_m2r (round4, mm1); + psraw_i2r (2, mm1); // /4 + paddw_m2r (round4, mm2); + psraw_i2r (2, mm2); // /4 + + /* now have subtotal/4 in mm1 and mm2 */ + + movq_m2r (*dest, mm3); // load 8 dest bytes + movq_r2r (mm3, mm4); // copy 8 dest bytes + + punpcklbw_r2r (mm0, mm3); // unpack low dest bytes + punpckhbw_r2r (mm0, mm4); // unpack high dest bytes + + paddw_r2r (mm3, mm1); // add lows + paddw_r2r (mm4, mm2); // add highs + + paddw_m2r (round1, mm1); + psraw_i2r (1, mm1); // /2 + paddw_m2r (round1, mm2); + psraw_i2r (1, mm2); // /2 + + /* now have end value in mm1 and mm2 */ + + packuswb_r2r (mm2, mm1); // pack (w/ saturation) + movq_r2m (mm1,*dest); // store result in dest +} + +/*-----------------------------------------------------------------------*/ + +static inline void MC_avg_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + mmx_zero_reg (); + + do { + mmx_average_2_U8 (dest, dest, ref); + + if (width == 16) + mmx_average_2_U8 (dest+8, dest+8, ref+8); + + dest += stride; + ref += stride; + } while (--height); +} + +static void MC_avg_o_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_mmx (16, height, dest, ref, stride); +} + +static void MC_avg_o_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_mmx (8, height, dest, ref, stride); +} + +/*-----------------------------------------------------------------------*/ + +static inline void MC_put_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + mmx_zero_reg (); + + do { + movq_m2r (* ref, mm1); // load 8 ref bytes + movq_r2m (mm1,* dest); // store 8 bytes at curr + + if (width == 16) + { + movq_m2r (* (ref+8), mm1); // load 8 ref bytes + movq_r2m (mm1,* (dest+8)); // store 8 bytes at curr + } + + dest += stride; + ref += stride; + } while (--height); +} + +static void MC_put_o_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_mmx (16, height, dest, ref, stride); +} + +static void MC_put_o_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_mmx (8, height, dest, ref, stride); +} + +/*-----------------------------------------------------------------------*/ + +/* Half pixel interpolation in the x direction */ +static inline void MC_avg_x_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + mmx_zero_reg (); + + do { + mmx_interp_average_2_U8 (dest, ref, ref+1); + + if (width == 16) + mmx_interp_average_2_U8 (dest+8, ref+8, ref+9); + + dest += stride; + ref += stride; + } while (--height); +} + +static void MC_avg_x_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_x_mmx (16, height, dest, ref, stride); +} + +static void MC_avg_x_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_x_mmx (8, height, dest, ref, stride); +} + +/*-----------------------------------------------------------------------*/ + +static inline void MC_put_x_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + mmx_zero_reg (); + + do { + mmx_average_2_U8 (dest, ref, ref+1); + + if (width == 16) + mmx_average_2_U8 (dest+8, ref+8, ref+9); + + dest += stride; + ref += stride; + } while (--height); +} + +static void MC_put_x_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_x_mmx (16, height, dest, ref, stride); +} + +static void MC_put_x_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_x_mmx (8, height, dest, ref, stride); +} + +/*-----------------------------------------------------------------------*/ + +static inline void MC_avg_xy_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + uint8_t * ref_next = ref+stride; + + mmx_zero_reg (); + + do { + mmx_interp_average_4_U8 (dest, ref, ref+1, ref_next, ref_next+1); + + if (width == 16) + mmx_interp_average_4_U8 (dest+8, ref+8, ref+9, + ref_next+8, ref_next+9); + + dest += stride; + ref += stride; + ref_next += stride; + } while (--height); +} + +static void MC_avg_xy_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_xy_mmx (16, height, dest, ref, stride); +} + +static void MC_avg_xy_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_xy_mmx (8, height, dest, ref, stride); +} + +/*-----------------------------------------------------------------------*/ + +static inline void MC_put_xy_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + uint8_t * ref_next = ref+stride; + + mmx_zero_reg (); + + do { + mmx_average_4_U8 (dest, ref, ref+1, ref_next, ref_next+1); + + if (width == 16) + mmx_average_4_U8 (dest+8, ref+8, ref+9, ref_next+8, ref_next+9); + + dest += stride; + ref += stride; + ref_next += stride; + } while (--height); +} + +static void MC_put_xy_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_xy_mmx (16, height, dest, ref, stride); +} + +static void MC_put_xy_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_xy_mmx (8, height, dest, ref, stride); +} + +/*-----------------------------------------------------------------------*/ + +static inline void MC_avg_y_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + uint8_t * ref_next = ref+stride; + + mmx_zero_reg (); + + do { + mmx_interp_average_2_U8 (dest, ref, ref_next); + + if (width == 16) + mmx_interp_average_2_U8 (dest+8, ref+8, ref_next+8); + + dest += stride; + ref += stride; + ref_next += stride; + } while (--height); +} + +static void MC_avg_y_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_y_mmx (16, height, dest, ref, stride); +} + +static void MC_avg_y_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg_y_mmx (8, height, dest, ref, stride); +} + +/*-----------------------------------------------------------------------*/ + +static inline void MC_put_y_mmx (int width, int height, + uint8_t * dest, uint8_t * ref, int stride) +{ + uint8_t * ref_next = ref+stride; + + mmx_zero_reg (); + + do { + mmx_average_2_U8 (dest, ref, ref_next); + + if (width == 16) + mmx_average_2_U8 (dest+8, ref+8, ref_next+8); + + dest += stride; + ref += stride; + ref_next += stride; + } while (--height); +} + +static void MC_put_y_16_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_y_mmx (16, height, dest, ref, stride); +} + +static void MC_put_y_8_mmx (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put_y_mmx (8, height, dest, ref, stride); +} + + +MPEG2_MC_EXTERN (mmx) + + + + + + + +/* CPU_MMXEXT/CPU_3DNOW adaptation layer */ + +#define pavg_r2r(src,dest) \ +do { \ + if (cpu == CPU_MMXEXT) \ + pavgb_r2r (src, dest); \ + else \ + pavgusb_r2r (src, dest); \ +} while (0) + +#define pavg_m2r(src,dest) \ +do { \ + if (cpu == CPU_MMXEXT) \ + pavgb_m2r (src, dest); \ + else \ + pavgusb_m2r (src, dest); \ +} while (0) + + +/* CPU_MMXEXT code */ + + +static inline void MC_put1_8 (int height, uint8_t * dest, uint8_t * ref, + int stride) +{ + do { + movq_m2r (*ref, mm0); + movq_r2m (mm0, *dest); + ref += stride; + dest += stride; + } while (--height); +} + +static inline void MC_put1_16 (int height, uint8_t * dest, uint8_t * ref, + int stride) +{ + do { + movq_m2r (*ref, mm0); + movq_m2r (*(ref+8), mm1); + ref += stride; + movq_r2m (mm0, *dest); + movq_r2m (mm1, *(dest+8)); + dest += stride; + } while (--height); +} + +static inline void MC_avg1_8 (int height, uint8_t * dest, uint8_t * ref, + int stride, int cpu) +{ + do { + movq_m2r (*ref, mm0); + pavg_m2r (*dest, mm0); + ref += stride; + movq_r2m (mm0, *dest); + dest += stride; + } while (--height); +} + +static inline void MC_avg1_16 (int height, uint8_t * dest, uint8_t * ref, + int stride, int cpu) +{ + do { + movq_m2r (*ref, mm0); + movq_m2r (*(ref+8), mm1); + pavg_m2r (*dest, mm0); + pavg_m2r (*(dest+8), mm1); + movq_r2m (mm0, *dest); + ref += stride; + movq_r2m (mm1, *(dest+8)); + dest += stride; + } while (--height); +} + +static inline void MC_put2_8 (int height, uint8_t * dest, uint8_t * ref, + int stride, int offset, int cpu) +{ + do { + movq_m2r (*ref, mm0); + pavg_m2r (*(ref+offset), mm0); + ref += stride; + movq_r2m (mm0, *dest); + dest += stride; + } while (--height); +} + +static inline void MC_put2_16 (int height, uint8_t * dest, uint8_t * ref, + int stride, int offset, int cpu) +{ + do { + movq_m2r (*ref, mm0); + movq_m2r (*(ref+8), mm1); + pavg_m2r (*(ref+offset), mm0); + pavg_m2r (*(ref+offset+8), mm1); + movq_r2m (mm0, *dest); + ref += stride; + movq_r2m (mm1, *(dest+8)); + dest += stride; + } while (--height); +} + +static inline void MC_avg2_8 (int height, uint8_t * dest, uint8_t * ref, + int stride, int offset, int cpu) +{ + do { + movq_m2r (*ref, mm0); + pavg_m2r (*(ref+offset), mm0); + pavg_m2r (*dest, mm0); + ref += stride; + movq_r2m (mm0, *dest); + dest += stride; + } while (--height); +} + +static inline void MC_avg2_16 (int height, uint8_t * dest, uint8_t * ref, + int stride, int offset, int cpu) +{ + do { + movq_m2r (*ref, mm0); + movq_m2r (*(ref+8), mm1); + pavg_m2r (*(ref+offset), mm0); + pavg_m2r (*(ref+offset+8), mm1); + pavg_m2r (*dest, mm0); + pavg_m2r (*(dest+8), mm1); + ref += stride; + movq_r2m (mm0, *dest); + movq_r2m (mm1, *(dest+8)); + dest += stride; + } while (--height); +} + +static mmx_t mask_one = {0x0101010101010101LL}; + +static inline void MC_put4_8 (int height, uint8_t * dest, uint8_t * ref, + int stride, int cpu) +{ + movq_m2r (*ref, mm0); + movq_m2r (*(ref+1), mm1); + movq_r2r (mm0, mm7); + pxor_r2r (mm1, mm7); + pavg_r2r (mm1, mm0); + ref += stride; + + do { + movq_m2r (*ref, mm2); + movq_r2r (mm0, mm5); + + movq_m2r (*(ref+1), mm3); + movq_r2r (mm2, mm6); + + pxor_r2r (mm3, mm6); + pavg_r2r (mm3, mm2); + + por_r2r (mm6, mm7); + pxor_r2r (mm2, mm5); + + pand_r2r (mm5, mm7); + pavg_r2r (mm2, mm0); + + pand_m2r (mask_one, mm7); + + psubusb_r2r (mm7, mm0); + + ref += stride; + movq_r2m (mm0, *dest); + dest += stride; + + movq_r2r (mm6, mm7); // unroll ! + movq_r2r (mm2, mm0); // unroll ! + } while (--height); +} + +static inline void MC_put4_16 (int height, uint8_t * dest, uint8_t * ref, + int stride, int cpu) +{ + do { + movq_m2r (*ref, mm0); + movq_m2r (*(ref+stride+1), mm1); + movq_r2r (mm0, mm7); + movq_m2r (*(ref+1), mm2); + pxor_r2r (mm1, mm7); + movq_m2r (*(ref+stride), mm3); + movq_r2r (mm2, mm6); + pxor_r2r (mm3, mm6); + pavg_r2r (mm1, mm0); + pavg_r2r (mm3, mm2); + por_r2r (mm6, mm7); + movq_r2r (mm0, mm6); + pxor_r2r (mm2, mm6); + pand_r2r (mm6, mm7); + pand_m2r (mask_one, mm7); + pavg_r2r (mm2, mm0); + psubusb_r2r (mm7, mm0); + movq_r2m (mm0, *dest); + + movq_m2r (*(ref+8), mm0); + movq_m2r (*(ref+stride+9), mm1); + movq_r2r (mm0, mm7); + movq_m2r (*(ref+9), mm2); + pxor_r2r (mm1, mm7); + movq_m2r (*(ref+stride+8), mm3); + movq_r2r (mm2, mm6); + pxor_r2r (mm3, mm6); + pavg_r2r (mm1, mm0); + pavg_r2r (mm3, mm2); + por_r2r (mm6, mm7); + movq_r2r (mm0, mm6); + pxor_r2r (mm2, mm6); + pand_r2r (mm6, mm7); + pand_m2r (mask_one, mm7); + pavg_r2r (mm2, mm0); + psubusb_r2r (mm7, mm0); + ref += stride; + movq_r2m (mm0, *(dest+8)); + dest += stride; + } while (--height); +} + +static inline void MC_avg4_8 (int height, uint8_t * dest, uint8_t * ref, + int stride, int cpu) +{ + do { + movq_m2r (*ref, mm0); + movq_m2r (*(ref+stride+1), mm1); + movq_r2r (mm0, mm7); + movq_m2r (*(ref+1), mm2); + pxor_r2r (mm1, mm7); + movq_m2r (*(ref+stride), mm3); + movq_r2r (mm2, mm6); + pxor_r2r (mm3, mm6); + pavg_r2r (mm1, mm0); + pavg_r2r (mm3, mm2); + por_r2r (mm6, mm7); + movq_r2r (mm0, mm6); + pxor_r2r (mm2, mm6); + pand_r2r (mm6, mm7); + pand_m2r (mask_one, mm7); + pavg_r2r (mm2, mm0); + psubusb_r2r (mm7, mm0); + movq_m2r (*dest, mm1); + pavg_r2r (mm1, mm0); + ref += stride; + movq_r2m (mm0, *dest); + dest += stride; + } while (--height); +} + +static inline void MC_avg4_16 (int height, uint8_t * dest, uint8_t * ref, + int stride, int cpu) +{ + do { + movq_m2r (*ref, mm0); + movq_m2r (*(ref+stride+1), mm1); + movq_r2r (mm0, mm7); + movq_m2r (*(ref+1), mm2); + pxor_r2r (mm1, mm7); + movq_m2r (*(ref+stride), mm3); + movq_r2r (mm2, mm6); + pxor_r2r (mm3, mm6); + pavg_r2r (mm1, mm0); + pavg_r2r (mm3, mm2); + por_r2r (mm6, mm7); + movq_r2r (mm0, mm6); + pxor_r2r (mm2, mm6); + pand_r2r (mm6, mm7); + pand_m2r (mask_one, mm7); + pavg_r2r (mm2, mm0); + psubusb_r2r (mm7, mm0); + movq_m2r (*dest, mm1); + pavg_r2r (mm1, mm0); + movq_r2m (mm0, *dest); + + movq_m2r (*(ref+8), mm0); + movq_m2r (*(ref+stride+9), mm1); + movq_r2r (mm0, mm7); + movq_m2r (*(ref+9), mm2); + pxor_r2r (mm1, mm7); + movq_m2r (*(ref+stride+8), mm3); + movq_r2r (mm2, mm6); + pxor_r2r (mm3, mm6); + pavg_r2r (mm1, mm0); + pavg_r2r (mm3, mm2); + por_r2r (mm6, mm7); + movq_r2r (mm0, mm6); + pxor_r2r (mm2, mm6); + pand_r2r (mm6, mm7); + pand_m2r (mask_one, mm7); + pavg_r2r (mm2, mm0); + psubusb_r2r (mm7, mm0); + movq_m2r (*(dest+8), mm1); + pavg_r2r (mm1, mm0); + ref += stride; + movq_r2m (mm0, *(dest+8)); + dest += stride; + } while (--height); +} + +static void MC_avg_o_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg1_16 (height, dest, ref, stride, CPU_MMXEXT); +} + +static void MC_avg_o_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg1_8 (height, dest, ref, stride, CPU_MMXEXT); +} + +static void MC_put_o_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put1_16 (height, dest, ref, stride); +} + +static void MC_put_o_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put1_8 (height, dest, ref, stride); +} + +static void MC_avg_x_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_16 (height, dest, ref, stride, 1, CPU_MMXEXT); +} + +static void MC_avg_x_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_8 (height, dest, ref, stride, 1, CPU_MMXEXT); +} + +static void MC_put_x_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_16 (height, dest, ref, stride, 1, CPU_MMXEXT); +} + +static void MC_put_x_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_8 (height, dest, ref, stride, 1, CPU_MMXEXT); +} + +static void MC_avg_y_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_16 (height, dest, ref, stride, stride, CPU_MMXEXT); +} + +static void MC_avg_y_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_8 (height, dest, ref, stride, stride, CPU_MMXEXT); +} + +static void MC_put_y_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_16 (height, dest, ref, stride, stride, CPU_MMXEXT); +} + +static void MC_put_y_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_8 (height, dest, ref, stride, stride, CPU_MMXEXT); +} + +static void MC_avg_xy_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg4_16 (height, dest, ref, stride, CPU_MMXEXT); +} + +static void MC_avg_xy_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg4_8 (height, dest, ref, stride, CPU_MMXEXT); +} + +static void MC_put_xy_16_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put4_16 (height, dest, ref, stride, CPU_MMXEXT); +} + +static void MC_put_xy_8_mmxext (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put4_8 (height, dest, ref, stride, CPU_MMXEXT); +} + + +MPEG2_MC_EXTERN (mmxext) + + + +static void MC_avg_o_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg1_16 (height, dest, ref, stride, CPU_3DNOW); +} + +static void MC_avg_o_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg1_8 (height, dest, ref, stride, CPU_3DNOW); +} + +static void MC_put_o_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put1_16 (height, dest, ref, stride); +} + +static void MC_put_o_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put1_8 (height, dest, ref, stride); +} + +static void MC_avg_x_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_16 (height, dest, ref, stride, 1, CPU_3DNOW); +} + +static void MC_avg_x_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_8 (height, dest, ref, stride, 1, CPU_3DNOW); +} + +static void MC_put_x_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_16 (height, dest, ref, stride, 1, CPU_3DNOW); +} + +static void MC_put_x_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_8 (height, dest, ref, stride, 1, CPU_3DNOW); +} + +static void MC_avg_y_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_16 (height, dest, ref, stride, stride, CPU_3DNOW); +} + +static void MC_avg_y_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg2_8 (height, dest, ref, stride, stride, CPU_3DNOW); +} + +static void MC_put_y_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_16 (height, dest, ref, stride, stride, CPU_3DNOW); +} + +static void MC_put_y_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put2_8 (height, dest, ref, stride, stride, CPU_3DNOW); +} + +static void MC_avg_xy_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg4_16 (height, dest, ref, stride, CPU_3DNOW); +} + +static void MC_avg_xy_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_avg4_8 (height, dest, ref, stride, CPU_3DNOW); +} + +static void MC_put_xy_16_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put4_16 (height, dest, ref, stride, CPU_3DNOW); +} + +static void MC_put_xy_8_3dnow (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + MC_put4_8 (height, dest, ref, stride, CPU_3DNOW); +} + + +MPEG2_MC_EXTERN (3dnow) + +#endif diff --git a/src/video_dec/libmpeg2/motion_comp_vis.c b/src/video_dec/libmpeg2/motion_comp_vis.c new file mode 100644 index 000000000..d0a6673d6 --- /dev/null +++ b/src/video_dec/libmpeg2/motion_comp_vis.c @@ -0,0 +1,2059 @@ +/* + * motion_comp_vis.c + * Copyright (C) 2003 David S. Miller <davem@redhat.com> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#if defined(ARCH_SPARC) && defined(ENABLE_VIS) + +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include "vis.h" + +/* The trick used in some of this file is the formula from the MMX + * motion comp code, which is: + * + * (x+y+1)>>1 == (x|y)-((x^y)>>1) + * + * This allows us to average 8 bytes at a time in a 64-bit FPU reg. + * We avoid overflows by masking before we do the shift, and we + * implement the shift by multiplying by 1/2 using mul8x16. So in + * VIS this is (assume 'x' is in f0, 'y' is in f2, a repeating mask + * of '0xfe' is in f4, a repeating mask of '0x7f' is in f6, and + * the value 0x80808080 is in f8): + * + * fxor f0, f2, f10 + * fand f10, f4, f10 + * fmul8x16 f8, f10, f10 + * fand f10, f6, f10 + * for f0, f2, f12 + * fpsub16 f12, f10, f10 + */ + +#define DUP4(x) {x, x, x, x} +#define DUP8(x) {x, x, x, x, x, x, x, x} +static const int16_t constants1[] ATTR_ALIGN(8) = DUP4 (1); +static const int16_t constants2[] ATTR_ALIGN(8) = DUP4 (2); +static const int16_t constants3[] ATTR_ALIGN(8) = DUP4 (3); +static const int16_t constants6[] ATTR_ALIGN(8) = DUP4 (6); +static const int8_t constants_fe[] ATTR_ALIGN(8) = DUP8 (0xfe); +static const int8_t constants_7f[] ATTR_ALIGN(8) = DUP8 (0x7f); +static const int8_t constants128[] ATTR_ALIGN(8) = DUP8 (128); +static const int16_t constants256_512[] ATTR_ALIGN(8) = + {256, 512, 256, 512}; +static const int16_t constants256_1024[] ATTR_ALIGN(8) = + {256, 1024, 256, 1024}; + +#define REF_0 0 +#define REF_0_1 1 +#define REF_2 2 +#define REF_2_1 3 +#define REF_4 4 +#define REF_4_1 5 +#define REF_6 6 +#define REF_6_1 7 +#define REF_S0 8 +#define REF_S0_1 9 +#define REF_S2 10 +#define REF_S2_1 11 +#define REF_S4 12 +#define REF_S4_1 13 +#define REF_S6 14 +#define REF_S6_1 15 +#define DST_0 16 +#define DST_1 17 +#define DST_2 18 +#define DST_3 19 +#define CONST_1 20 +#define CONST_2 20 +#define CONST_3 20 +#define CONST_6 20 +#define MASK_fe 20 +#define CONST_128 22 +#define CONST_256 22 +#define CONST_512 22 +#define CONST_1024 22 +#define TMP0 24 +#define TMP1 25 +#define TMP2 26 +#define TMP3 27 +#define TMP4 28 +#define TMP5 29 +#define ZERO 30 +#define MASK_7f 30 + +#define TMP6 32 +#define TMP8 34 +#define TMP10 36 +#define TMP12 38 +#define TMP14 40 +#define TMP16 42 +#define TMP18 44 +#define TMP20 46 +#define TMP22 48 +#define TMP24 50 +#define TMP26 52 +#define TMP28 54 +#define TMP30 56 +#define TMP32 58 + +static void MC_put_o_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int offset; + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 16 : 0; + do { /* 5 cycles */ + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, 8, TMP2); + + vis_ld64_2(ref, offset, TMP4); + ref += stride; + + vis_faligndata(TMP0, TMP2, REF_0); + vis_st64(REF_0, dest[0]); + + vis_faligndata(TMP2, TMP4, REF_2); + vis_st64_2(REF_2, dest, 8); + dest += stride; + } while (--height); +} + +static void MC_put_o_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int offset; + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 8 : 0; + do { /* 4 cycles */ + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, offset, TMP2); + ref += stride; + + /* stall */ + + vis_faligndata(TMP0, TMP2, REF_0); + vis_st64(REF_0, dest[0]); + dest += stride; + } while (--height); +} + + +static void MC_avg_o_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int stride_8 = stride + 8; + int offset; + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 16 : 0; + + vis_ld64(ref[0], TMP0); + + vis_ld64(ref[8], TMP2); + + vis_ld64_2(ref, offset, TMP4); + + vis_ld64(dest[0], DST_0); + + vis_ld64(dest[8], DST_2); + + vis_ld64(constants_fe[0], MASK_fe); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64(constants_7f[0], MASK_7f); + vis_faligndata(TMP2, TMP4, REF_2); + + vis_ld64(constants128[0], CONST_128); + + ref += stride; + height = (height >> 1) - 1; + + do { /* 24 cycles */ + vis_ld64(ref[0], TMP0); + vis_xor(DST_0, REF_0, TMP6); + + vis_ld64_2(ref, 8, TMP2); + vis_and(TMP6, MASK_fe, TMP6); + + vis_ld64_2(ref, offset, TMP4); + ref += stride; + vis_mul8x16(CONST_128, TMP6, TMP6); + vis_xor(DST_2, REF_2, TMP8); + + vis_and(TMP8, MASK_fe, TMP8); + + vis_or(DST_0, REF_0, TMP10); + vis_ld64_2(dest, stride, DST_0); + vis_mul8x16(CONST_128, TMP8, TMP8); + + vis_or(DST_2, REF_2, TMP12); + vis_ld64_2(dest, stride_8, DST_2); + + vis_ld64(ref[0], TMP14); + vis_and(TMP6, MASK_7f, TMP6); + + vis_and(TMP8, MASK_7f, TMP8); + + vis_psub16(TMP10, TMP6, TMP6); + vis_st64(TMP6, dest[0]); + + vis_psub16(TMP12, TMP8, TMP8); + vis_st64_2(TMP8, dest, 8); + + dest += stride; + vis_ld64_2(ref, 8, TMP16); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, offset, TMP18); + vis_faligndata(TMP2, TMP4, REF_2); + ref += stride; + + vis_xor(DST_0, REF_0, TMP20); + + vis_and(TMP20, MASK_fe, TMP20); + + vis_xor(DST_2, REF_2, TMP22); + vis_mul8x16(CONST_128, TMP20, TMP20); + + vis_and(TMP22, MASK_fe, TMP22); + + vis_or(DST_0, REF_0, TMP24); + vis_mul8x16(CONST_128, TMP22, TMP22); + + vis_or(DST_2, REF_2, TMP26); + + vis_ld64_2(dest, stride, DST_0); + vis_faligndata(TMP14, TMP16, REF_0); + + vis_ld64_2(dest, stride_8, DST_2); + vis_faligndata(TMP16, TMP18, REF_2); + + vis_and(TMP20, MASK_7f, TMP20); + + vis_and(TMP22, MASK_7f, TMP22); + + vis_psub16(TMP24, TMP20, TMP20); + vis_st64(TMP20, dest[0]); + + vis_psub16(TMP26, TMP22, TMP22); + vis_st64_2(TMP22, dest, 8); + dest += stride; + } while (--height); + + vis_ld64(ref[0], TMP0); + vis_xor(DST_0, REF_0, TMP6); + + vis_ld64_2(ref, 8, TMP2); + vis_and(TMP6, MASK_fe, TMP6); + + vis_ld64_2(ref, offset, TMP4); + vis_mul8x16(CONST_128, TMP6, TMP6); + vis_xor(DST_2, REF_2, TMP8); + + vis_and(TMP8, MASK_fe, TMP8); + + vis_or(DST_0, REF_0, TMP10); + vis_ld64_2(dest, stride, DST_0); + vis_mul8x16(CONST_128, TMP8, TMP8); + + vis_or(DST_2, REF_2, TMP12); + vis_ld64_2(dest, stride_8, DST_2); + + vis_ld64(ref[0], TMP14); + vis_and(TMP6, MASK_7f, TMP6); + + vis_and(TMP8, MASK_7f, TMP8); + + vis_psub16(TMP10, TMP6, TMP6); + vis_st64(TMP6, dest[0]); + + vis_psub16(TMP12, TMP8, TMP8); + vis_st64_2(TMP8, dest, 8); + + dest += stride; + vis_faligndata(TMP0, TMP2, REF_0); + + vis_faligndata(TMP2, TMP4, REF_2); + + vis_xor(DST_0, REF_0, TMP20); + + vis_and(TMP20, MASK_fe, TMP20); + + vis_xor(DST_2, REF_2, TMP22); + vis_mul8x16(CONST_128, TMP20, TMP20); + + vis_and(TMP22, MASK_fe, TMP22); + + vis_or(DST_0, REF_0, TMP24); + vis_mul8x16(CONST_128, TMP22, TMP22); + + vis_or(DST_2, REF_2, TMP26); + + vis_and(TMP20, MASK_7f, TMP20); + + vis_and(TMP22, MASK_7f, TMP22); + + vis_psub16(TMP24, TMP20, TMP20); + vis_st64(TMP20, dest[0]); + + vis_psub16(TMP26, TMP22, TMP22); + vis_st64_2(TMP22, dest, 8); +} + +static void MC_avg_o_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int offset; + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 8 : 0; + + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, offset, TMP2); + + vis_ld64(dest[0], DST_0); + + vis_ld64(constants_fe[0], MASK_fe); + + vis_ld64(constants_7f[0], MASK_7f); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64(constants128[0], CONST_128); + + ref += stride; + height = (height >> 1) - 1; + + do { /* 12 cycles */ + vis_ld64(ref[0], TMP0); + vis_xor(DST_0, REF_0, TMP4); + + vis_ld64_2(ref, offset, TMP2); + vis_and(TMP4, MASK_fe, TMP4); + + vis_or(DST_0, REF_0, TMP6); + vis_ld64_2(dest, stride, DST_0); + ref += stride; + vis_mul8x16(CONST_128, TMP4, TMP4); + + vis_ld64(ref[0], TMP12); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, offset, TMP2); + vis_xor(DST_0, REF_0, TMP0); + ref += stride; + + vis_and(TMP0, MASK_fe, TMP0); + + vis_and(TMP4, MASK_7f, TMP4); + + vis_psub16(TMP6, TMP4, TMP4); + vis_st64(TMP4, dest[0]); + dest += stride; + vis_mul8x16(CONST_128, TMP0, TMP0); + + vis_or(DST_0, REF_0, TMP6); + vis_ld64_2(dest, stride, DST_0); + + vis_faligndata(TMP12, TMP2, REF_0); + + vis_and(TMP0, MASK_7f, TMP0); + + vis_psub16(TMP6, TMP0, TMP4); + vis_st64(TMP4, dest[0]); + dest += stride; + } while (--height); + + vis_ld64(ref[0], TMP0); + vis_xor(DST_0, REF_0, TMP4); + + vis_ld64_2(ref, offset, TMP2); + vis_and(TMP4, MASK_fe, TMP4); + + vis_or(DST_0, REF_0, TMP6); + vis_ld64_2(dest, stride, DST_0); + vis_mul8x16(CONST_128, TMP4, TMP4); + + vis_faligndata(TMP0, TMP2, REF_0); + + vis_xor(DST_0, REF_0, TMP0); + + vis_and(TMP0, MASK_fe, TMP0); + + vis_and(TMP4, MASK_7f, TMP4); + + vis_psub16(TMP6, TMP4, TMP4); + vis_st64(TMP4, dest[0]); + dest += stride; + vis_mul8x16(CONST_128, TMP0, TMP0); + + vis_or(DST_0, REF_0, TMP6); + + vis_and(TMP0, MASK_7f, TMP0); + + vis_psub16(TMP6, TMP0, TMP4); + vis_st64(TMP4, dest[0]); +} + +static void MC_put_x_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + + ref = vis_alignaddr(ref); + + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, 8, TMP2); + + vis_ld64_2(ref, 16, TMP4); + + vis_ld64(constants_fe[0], MASK_fe); + + vis_ld64(constants_7f[0], MASK_7f); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64(constants128[0], CONST_128); + vis_faligndata(TMP2, TMP4, REF_4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + vis_faligndata(TMP2, TMP4, REF_6); + } else { + vis_src1(TMP2, REF_2); + vis_src1(TMP4, REF_6); + } + + ref += stride; + height = (height >> 1) - 1; + + do { /* 34 cycles */ + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP6); + + vis_ld64_2(ref, 8, TMP2); + vis_xor(REF_4, REF_6, TMP8); + + vis_ld64_2(ref, 16, TMP4); + vis_and(TMP6, MASK_fe, TMP6); + ref += stride; + + vis_ld64(ref[0], TMP14); + vis_mul8x16(CONST_128, TMP6, TMP6); + vis_and(TMP8, MASK_fe, TMP8); + + vis_ld64_2(ref, 8, TMP16); + vis_mul8x16(CONST_128, TMP8, TMP8); + vis_or(REF_0, REF_2, TMP10); + + vis_ld64_2(ref, 16, TMP18); + ref += stride; + vis_or(REF_4, REF_6, TMP12); + + vis_alignaddr_g0((void *)off); + + vis_faligndata(TMP0, TMP2, REF_0); + + vis_faligndata(TMP2, TMP4, REF_4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + vis_faligndata(TMP2, TMP4, REF_6); + } else { + vis_src1(TMP2, REF_2); + vis_src1(TMP4, REF_6); + } + + vis_and(TMP6, MASK_7f, TMP6); + + vis_and(TMP8, MASK_7f, TMP8); + + vis_psub16(TMP10, TMP6, TMP6); + vis_st64(TMP6, dest[0]); + + vis_psub16(TMP12, TMP8, TMP8); + vis_st64_2(TMP8, dest, 8); + dest += stride; + + vis_xor(REF_0, REF_2, TMP6); + + vis_xor(REF_4, REF_6, TMP8); + + vis_and(TMP6, MASK_fe, TMP6); + + vis_mul8x16(CONST_128, TMP6, TMP6); + vis_and(TMP8, MASK_fe, TMP8); + + vis_mul8x16(CONST_128, TMP8, TMP8); + vis_or(REF_0, REF_2, TMP10); + + vis_or(REF_4, REF_6, TMP12); + + vis_alignaddr_g0((void *)off); + + vis_faligndata(TMP14, TMP16, REF_0); + + vis_faligndata(TMP16, TMP18, REF_4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP14, TMP16, REF_2); + vis_faligndata(TMP16, TMP18, REF_6); + } else { + vis_src1(TMP16, REF_2); + vis_src1(TMP18, REF_6); + } + + vis_and(TMP6, MASK_7f, TMP6); + + vis_and(TMP8, MASK_7f, TMP8); + + vis_psub16(TMP10, TMP6, TMP6); + vis_st64(TMP6, dest[0]); + + vis_psub16(TMP12, TMP8, TMP8); + vis_st64_2(TMP8, dest, 8); + dest += stride; + } while (--height); + + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP6); + + vis_ld64_2(ref, 8, TMP2); + vis_xor(REF_4, REF_6, TMP8); + + vis_ld64_2(ref, 16, TMP4); + vis_and(TMP6, MASK_fe, TMP6); + + vis_mul8x16(CONST_128, TMP6, TMP6); + vis_and(TMP8, MASK_fe, TMP8); + + vis_mul8x16(CONST_128, TMP8, TMP8); + vis_or(REF_0, REF_2, TMP10); + + vis_or(REF_4, REF_6, TMP12); + + vis_alignaddr_g0((void *)off); + + vis_faligndata(TMP0, TMP2, REF_0); + + vis_faligndata(TMP2, TMP4, REF_4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + vis_faligndata(TMP2, TMP4, REF_6); + } else { + vis_src1(TMP2, REF_2); + vis_src1(TMP4, REF_6); + } + + vis_and(TMP6, MASK_7f, TMP6); + + vis_and(TMP8, MASK_7f, TMP8); + + vis_psub16(TMP10, TMP6, TMP6); + vis_st64(TMP6, dest[0]); + + vis_psub16(TMP12, TMP8, TMP8); + vis_st64_2(TMP8, dest, 8); + dest += stride; + + vis_xor(REF_0, REF_2, TMP6); + + vis_xor(REF_4, REF_6, TMP8); + + vis_and(TMP6, MASK_fe, TMP6); + + vis_mul8x16(CONST_128, TMP6, TMP6); + vis_and(TMP8, MASK_fe, TMP8); + + vis_mul8x16(CONST_128, TMP8, TMP8); + vis_or(REF_0, REF_2, TMP10); + + vis_or(REF_4, REF_6, TMP12); + + vis_and(TMP6, MASK_7f, TMP6); + + vis_and(TMP8, MASK_7f, TMP8); + + vis_psub16(TMP10, TMP6, TMP6); + vis_st64(TMP6, dest[0]); + + vis_psub16(TMP12, TMP8, TMP8); + vis_st64_2(TMP8, dest, 8); +} + +static void MC_put_x_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + + ref = vis_alignaddr(ref); + + vis_ld64(ref[0], TMP0); + + vis_ld64(ref[8], TMP2); + + vis_ld64(constants_fe[0], MASK_fe); + + vis_ld64(constants_7f[0], MASK_7f); + + vis_ld64(constants128[0], CONST_128); + vis_faligndata(TMP0, TMP2, REF_0); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + } else { + vis_src1(TMP2, REF_2); + } + + ref += stride; + height = (height >> 1) - 1; + + do { /* 20 cycles */ + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP4); + + vis_ld64_2(ref, 8, TMP2); + vis_and(TMP4, MASK_fe, TMP4); + ref += stride; + + vis_ld64(ref[0], TMP8); + vis_or(REF_0, REF_2, TMP6); + vis_mul8x16(CONST_128, TMP4, TMP4); + + vis_alignaddr_g0((void *)off); + + vis_ld64_2(ref, 8, TMP10); + ref += stride; + vis_faligndata(TMP0, TMP2, REF_0); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + } else { + vis_src1(TMP2, REF_2); + } + + vis_and(TMP4, MASK_7f, TMP4); + + vis_psub16(TMP6, TMP4, DST_0); + vis_st64(DST_0, dest[0]); + dest += stride; + + vis_xor(REF_0, REF_2, TMP12); + + vis_and(TMP12, MASK_fe, TMP12); + + vis_or(REF_0, REF_2, TMP14); + vis_mul8x16(CONST_128, TMP12, TMP12); + + vis_alignaddr_g0((void *)off); + vis_faligndata(TMP8, TMP10, REF_0); + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP8, TMP10, REF_2); + } else { + vis_src1(TMP10, REF_2); + } + + vis_and(TMP12, MASK_7f, TMP12); + + vis_psub16(TMP14, TMP12, DST_0); + vis_st64(DST_0, dest[0]); + dest += stride; + } while (--height); + + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP4); + + vis_ld64_2(ref, 8, TMP2); + vis_and(TMP4, MASK_fe, TMP4); + + vis_or(REF_0, REF_2, TMP6); + vis_mul8x16(CONST_128, TMP4, TMP4); + + vis_alignaddr_g0((void *)off); + + vis_faligndata(TMP0, TMP2, REF_0); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + } else { + vis_src1(TMP2, REF_2); + } + + vis_and(TMP4, MASK_7f, TMP4); + + vis_psub16(TMP6, TMP4, DST_0); + vis_st64(DST_0, dest[0]); + dest += stride; + + vis_xor(REF_0, REF_2, TMP12); + + vis_and(TMP12, MASK_fe, TMP12); + + vis_or(REF_0, REF_2, TMP14); + vis_mul8x16(CONST_128, TMP12, TMP12); + + vis_and(TMP12, MASK_7f, TMP12); + + vis_psub16(TMP14, TMP12, DST_0); + vis_st64(DST_0, dest[0]); + dest += stride; +} + +static void MC_avg_x_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + + vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT); + + vis_ld64(constants3[0], CONST_3); + vis_fzero(ZERO); + vis_ld64(constants256_512[0], CONST_256); + + ref = vis_alignaddr(ref); + do { /* 26 cycles */ + vis_ld64(ref[0], TMP0); + + vis_ld64(ref[8], TMP2); + + vis_alignaddr_g0((void *)off); + + vis_ld64(ref[16], TMP4); + + vis_ld64(dest[0], DST_0); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64(dest[8], DST_2); + vis_faligndata(TMP2, TMP4, REF_4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + vis_faligndata(TMP2, TMP4, REF_6); + } else { + vis_src1(TMP2, REF_2); + vis_src1(TMP4, REF_6); + } + + vis_mul8x16au(REF_0, CONST_256, TMP0); + + vis_pmerge(ZERO, REF_2, TMP4); + vis_mul8x16au(REF_0_1, CONST_256, TMP2); + + vis_pmerge(ZERO, REF_2_1, TMP6); + + vis_padd16(TMP0, TMP4, TMP0); + + vis_mul8x16al(DST_0, CONST_512, TMP4); + vis_padd16(TMP2, TMP6, TMP2); + + vis_mul8x16al(DST_1, CONST_512, TMP6); + + vis_mul8x16au(REF_6, CONST_256, TMP12); + + vis_padd16(TMP0, TMP4, TMP0); + vis_mul8x16au(REF_6_1, CONST_256, TMP14); + + vis_padd16(TMP2, TMP6, TMP2); + vis_mul8x16au(REF_4, CONST_256, TMP16); + + vis_padd16(TMP0, CONST_3, TMP8); + vis_mul8x16au(REF_4_1, CONST_256, TMP18); + + vis_padd16(TMP2, CONST_3, TMP10); + vis_pack16(TMP8, DST_0); + + vis_pack16(TMP10, DST_1); + vis_padd16(TMP16, TMP12, TMP0); + + vis_st64(DST_0, dest[0]); + vis_mul8x16al(DST_2, CONST_512, TMP4); + vis_padd16(TMP18, TMP14, TMP2); + + vis_mul8x16al(DST_3, CONST_512, TMP6); + vis_padd16(TMP0, CONST_3, TMP0); + + vis_padd16(TMP2, CONST_3, TMP2); + + vis_padd16(TMP0, TMP4, TMP0); + + vis_padd16(TMP2, TMP6, TMP2); + vis_pack16(TMP0, DST_2); + + vis_pack16(TMP2, DST_3); + vis_st64(DST_2, dest[8]); + + ref += stride; + dest += stride; + } while (--height); +} + +static void MC_avg_x_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + int stride_times_2 = stride << 1; + + vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT); + + vis_ld64(constants3[0], CONST_3); + vis_fzero(ZERO); + vis_ld64(constants256_512[0], CONST_256); + + ref = vis_alignaddr(ref); + height >>= 2; + do { /* 47 cycles */ + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, 8, TMP2); + ref += stride; + + vis_alignaddr_g0((void *)off); + + vis_ld64(ref[0], TMP4); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, 8, TMP6); + ref += stride; + + vis_ld64(ref[0], TMP8); + + vis_ld64_2(ref, 8, TMP10); + ref += stride; + vis_faligndata(TMP4, TMP6, REF_4); + + vis_ld64(ref[0], TMP12); + + vis_ld64_2(ref, 8, TMP14); + ref += stride; + vis_faligndata(TMP8, TMP10, REF_S0); + + vis_faligndata(TMP12, TMP14, REF_S4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + + vis_ld64(dest[0], DST_0); + vis_faligndata(TMP0, TMP2, REF_2); + + vis_ld64_2(dest, stride, DST_2); + vis_faligndata(TMP4, TMP6, REF_6); + + vis_faligndata(TMP8, TMP10, REF_S2); + + vis_faligndata(TMP12, TMP14, REF_S6); + } else { + vis_ld64(dest[0], DST_0); + vis_src1(TMP2, REF_2); + + vis_ld64_2(dest, stride, DST_2); + vis_src1(TMP6, REF_6); + + vis_src1(TMP10, REF_S2); + + vis_src1(TMP14, REF_S6); + } + + vis_pmerge(ZERO, REF_0, TMP0); + vis_mul8x16au(REF_0_1, CONST_256, TMP2); + + vis_pmerge(ZERO, REF_2, TMP4); + vis_mul8x16au(REF_2_1, CONST_256, TMP6); + + vis_padd16(TMP0, CONST_3, TMP0); + vis_mul8x16al(DST_0, CONST_512, TMP16); + + vis_padd16(TMP2, CONST_3, TMP2); + vis_mul8x16al(DST_1, CONST_512, TMP18); + + vis_padd16(TMP0, TMP4, TMP0); + vis_mul8x16au(REF_4, CONST_256, TMP8); + + vis_padd16(TMP2, TMP6, TMP2); + vis_mul8x16au(REF_4_1, CONST_256, TMP10); + + vis_padd16(TMP0, TMP16, TMP0); + vis_mul8x16au(REF_6, CONST_256, TMP12); + + vis_padd16(TMP2, TMP18, TMP2); + vis_mul8x16au(REF_6_1, CONST_256, TMP14); + + vis_padd16(TMP8, CONST_3, TMP8); + vis_mul8x16al(DST_2, CONST_512, TMP16); + + vis_padd16(TMP8, TMP12, TMP8); + vis_mul8x16al(DST_3, CONST_512, TMP18); + + vis_padd16(TMP10, TMP14, TMP10); + vis_pack16(TMP0, DST_0); + + vis_pack16(TMP2, DST_1); + vis_st64(DST_0, dest[0]); + dest += stride; + vis_padd16(TMP10, CONST_3, TMP10); + + vis_ld64_2(dest, stride, DST_0); + vis_padd16(TMP8, TMP16, TMP8); + + vis_ld64_2(dest, stride_times_2, TMP4/*DST_2*/); + vis_padd16(TMP10, TMP18, TMP10); + vis_pack16(TMP8, DST_2); + + vis_pack16(TMP10, DST_3); + vis_st64(DST_2, dest[0]); + dest += stride; + + vis_mul8x16au(REF_S0_1, CONST_256, TMP2); + vis_pmerge(ZERO, REF_S0, TMP0); + + vis_pmerge(ZERO, REF_S2, TMP24); + vis_mul8x16au(REF_S2_1, CONST_256, TMP6); + + vis_padd16(TMP0, CONST_3, TMP0); + vis_mul8x16au(REF_S4, CONST_256, TMP8); + + vis_padd16(TMP2, CONST_3, TMP2); + vis_mul8x16au(REF_S4_1, CONST_256, TMP10); + + vis_padd16(TMP0, TMP24, TMP0); + vis_mul8x16au(REF_S6, CONST_256, TMP12); + + vis_padd16(TMP2, TMP6, TMP2); + vis_mul8x16au(REF_S6_1, CONST_256, TMP14); + + vis_padd16(TMP8, CONST_3, TMP8); + vis_mul8x16al(DST_0, CONST_512, TMP16); + + vis_padd16(TMP10, CONST_3, TMP10); + vis_mul8x16al(DST_1, CONST_512, TMP18); + + vis_padd16(TMP8, TMP12, TMP8); + vis_mul8x16al(TMP4/*DST_2*/, CONST_512, TMP20); + + vis_mul8x16al(TMP5/*DST_3*/, CONST_512, TMP22); + vis_padd16(TMP0, TMP16, TMP0); + + vis_padd16(TMP2, TMP18, TMP2); + vis_pack16(TMP0, DST_0); + + vis_padd16(TMP10, TMP14, TMP10); + vis_pack16(TMP2, DST_1); + vis_st64(DST_0, dest[0]); + dest += stride; + + vis_padd16(TMP8, TMP20, TMP8); + + vis_padd16(TMP10, TMP22, TMP10); + vis_pack16(TMP8, DST_2); + + vis_pack16(TMP10, DST_3); + vis_st64(DST_2, dest[0]); + dest += stride; + } while (--height); +} + +static void MC_put_y_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int offset; + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 16 : 0; + + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, 8, TMP2); + + vis_ld64_2(ref, offset, TMP4); + ref += stride; + + vis_ld64(ref[0], TMP6); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, 8, TMP8); + vis_faligndata(TMP2, TMP4, REF_4); + + vis_ld64_2(ref, offset, TMP10); + ref += stride; + + vis_ld64(constants_fe[0], MASK_fe); + vis_faligndata(TMP6, TMP8, REF_2); + + vis_ld64(constants_7f[0], MASK_7f); + vis_faligndata(TMP8, TMP10, REF_6); + + vis_ld64(constants128[0], CONST_128); + height = (height >> 1) - 1; + do { /* 24 cycles */ + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP12); + + vis_ld64_2(ref, 8, TMP2); + vis_xor(REF_4, REF_6, TMP16); + + vis_ld64_2(ref, offset, TMP4); + ref += stride; + vis_or(REF_0, REF_2, TMP14); + + vis_ld64(ref[0], TMP6); + vis_or(REF_4, REF_6, TMP18); + + vis_ld64_2(ref, 8, TMP8); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, offset, TMP10); + ref += stride; + vis_faligndata(TMP2, TMP4, REF_4); + + vis_and(TMP12, MASK_fe, TMP12); + + vis_and(TMP16, MASK_fe, TMP16); + vis_mul8x16(CONST_128, TMP12, TMP12); + + vis_mul8x16(CONST_128, TMP16, TMP16); + vis_xor(REF_0, REF_2, TMP0); + + vis_xor(REF_4, REF_6, TMP2); + + vis_or(REF_0, REF_2, TMP20); + + vis_and(TMP12, MASK_7f, TMP12); + + vis_and(TMP16, MASK_7f, TMP16); + + vis_psub16(TMP14, TMP12, TMP12); + vis_st64(TMP12, dest[0]); + + vis_psub16(TMP18, TMP16, TMP16); + vis_st64_2(TMP16, dest, 8); + dest += stride; + + vis_or(REF_4, REF_6, TMP18); + + vis_and(TMP0, MASK_fe, TMP0); + + vis_and(TMP2, MASK_fe, TMP2); + vis_mul8x16(CONST_128, TMP0, TMP0); + + vis_faligndata(TMP6, TMP8, REF_2); + vis_mul8x16(CONST_128, TMP2, TMP2); + + vis_faligndata(TMP8, TMP10, REF_6); + + vis_and(TMP0, MASK_7f, TMP0); + + vis_and(TMP2, MASK_7f, TMP2); + + vis_psub16(TMP20, TMP0, TMP0); + vis_st64(TMP0, dest[0]); + + vis_psub16(TMP18, TMP2, TMP2); + vis_st64_2(TMP2, dest, 8); + dest += stride; + } while (--height); + + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP12); + + vis_ld64_2(ref, 8, TMP2); + vis_xor(REF_4, REF_6, TMP16); + + vis_ld64_2(ref, offset, TMP4); + vis_or(REF_0, REF_2, TMP14); + + vis_or(REF_4, REF_6, TMP18); + + vis_faligndata(TMP0, TMP2, REF_0); + + vis_faligndata(TMP2, TMP4, REF_4); + + vis_and(TMP12, MASK_fe, TMP12); + + vis_and(TMP16, MASK_fe, TMP16); + vis_mul8x16(CONST_128, TMP12, TMP12); + + vis_mul8x16(CONST_128, TMP16, TMP16); + vis_xor(REF_0, REF_2, TMP0); + + vis_xor(REF_4, REF_6, TMP2); + + vis_or(REF_0, REF_2, TMP20); + + vis_and(TMP12, MASK_7f, TMP12); + + vis_and(TMP16, MASK_7f, TMP16); + + vis_psub16(TMP14, TMP12, TMP12); + vis_st64(TMP12, dest[0]); + + vis_psub16(TMP18, TMP16, TMP16); + vis_st64_2(TMP16, dest, 8); + dest += stride; + + vis_or(REF_4, REF_6, TMP18); + + vis_and(TMP0, MASK_fe, TMP0); + + vis_and(TMP2, MASK_fe, TMP2); + vis_mul8x16(CONST_128, TMP0, TMP0); + + vis_mul8x16(CONST_128, TMP2, TMP2); + + vis_and(TMP0, MASK_7f, TMP0); + + vis_and(TMP2, MASK_7f, TMP2); + + vis_psub16(TMP20, TMP0, TMP0); + vis_st64(TMP0, dest[0]); + + vis_psub16(TMP18, TMP2, TMP2); + vis_st64_2(TMP2, dest, 8); +} + +static void MC_put_y_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int offset; + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 8 : 0; + + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, offset, TMP2); + ref += stride; + + vis_ld64(ref[0], TMP4); + + vis_ld64_2(ref, offset, TMP6); + ref += stride; + + vis_ld64(constants_fe[0], MASK_fe); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64(constants_7f[0], MASK_7f); + vis_faligndata(TMP4, TMP6, REF_2); + + vis_ld64(constants128[0], CONST_128); + height = (height >> 1) - 1; + do { /* 12 cycles */ + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP4); + + vis_ld64_2(ref, offset, TMP2); + ref += stride; + vis_and(TMP4, MASK_fe, TMP4); + + vis_or(REF_0, REF_2, TMP6); + vis_mul8x16(CONST_128, TMP4, TMP4); + + vis_faligndata(TMP0, TMP2, REF_0); + vis_ld64(ref[0], TMP0); + + vis_ld64_2(ref, offset, TMP2); + ref += stride; + vis_xor(REF_0, REF_2, TMP12); + + vis_and(TMP4, MASK_7f, TMP4); + + vis_and(TMP12, MASK_fe, TMP12); + + vis_mul8x16(CONST_128, TMP12, TMP12); + vis_or(REF_0, REF_2, TMP14); + + vis_psub16(TMP6, TMP4, DST_0); + vis_st64(DST_0, dest[0]); + dest += stride; + + vis_faligndata(TMP0, TMP2, REF_2); + + vis_and(TMP12, MASK_7f, TMP12); + + vis_psub16(TMP14, TMP12, DST_0); + vis_st64(DST_0, dest[0]); + dest += stride; + } while (--height); + + vis_ld64(ref[0], TMP0); + vis_xor(REF_0, REF_2, TMP4); + + vis_ld64_2(ref, offset, TMP2); + vis_and(TMP4, MASK_fe, TMP4); + + vis_or(REF_0, REF_2, TMP6); + vis_mul8x16(CONST_128, TMP4, TMP4); + + vis_faligndata(TMP0, TMP2, REF_0); + + vis_xor(REF_0, REF_2, TMP12); + + vis_and(TMP4, MASK_7f, TMP4); + + vis_and(TMP12, MASK_fe, TMP12); + + vis_mul8x16(CONST_128, TMP12, TMP12); + vis_or(REF_0, REF_2, TMP14); + + vis_psub16(TMP6, TMP4, DST_0); + vis_st64(DST_0, dest[0]); + dest += stride; + + vis_and(TMP12, MASK_7f, TMP12); + + vis_psub16(TMP14, TMP12, DST_0); + vis_st64(DST_0, dest[0]); +} + +static void MC_avg_y_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int stride_8 = stride + 8; + int stride_16; + int offset; + + vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT); + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 16 : 0; + + vis_ld64(ref[ 0], TMP0); + vis_fzero(ZERO); + + vis_ld64(ref[ 8], TMP2); + + vis_ld64_2(ref, offset, TMP4); + stride_16 = stride + offset; + + vis_ld64(constants3[0], CONST_3); + vis_faligndata(TMP0, TMP2, REF_2); + + vis_ld64(constants256_512[0], CONST_256); + vis_faligndata(TMP2, TMP4, REF_6); + height >>= 1; + + do { /* 31 cycles */ + vis_ld64_2(ref, stride, TMP0); + vis_pmerge(ZERO, REF_2, TMP12); + vis_mul8x16au(REF_2_1, CONST_256, TMP14); + + vis_ld64_2(ref, stride_8, TMP2); + vis_pmerge(ZERO, REF_6, TMP16); + vis_mul8x16au(REF_6_1, CONST_256, TMP18); + + vis_ld64_2(ref, stride_16, TMP4); + ref += stride; + + vis_ld64(dest[0], DST_0); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(dest, 8, DST_2); + vis_faligndata(TMP2, TMP4, REF_4); + + vis_ld64_2(ref, stride, TMP6); + vis_pmerge(ZERO, REF_0, TMP0); + vis_mul8x16au(REF_0_1, CONST_256, TMP2); + + vis_ld64_2(ref, stride_8, TMP8); + vis_pmerge(ZERO, REF_4, TMP4); + + vis_ld64_2(ref, stride_16, TMP10); + ref += stride; + + vis_ld64_2(dest, stride, REF_S0/*DST_4*/); + vis_faligndata(TMP6, TMP8, REF_2); + vis_mul8x16au(REF_4_1, CONST_256, TMP6); + + vis_ld64_2(dest, stride_8, REF_S2/*DST_6*/); + vis_faligndata(TMP8, TMP10, REF_6); + vis_mul8x16al(DST_0, CONST_512, TMP20); + + vis_padd16(TMP0, CONST_3, TMP0); + vis_mul8x16al(DST_1, CONST_512, TMP22); + + vis_padd16(TMP2, CONST_3, TMP2); + vis_mul8x16al(DST_2, CONST_512, TMP24); + + vis_padd16(TMP4, CONST_3, TMP4); + vis_mul8x16al(DST_3, CONST_512, TMP26); + + vis_padd16(TMP6, CONST_3, TMP6); + + vis_padd16(TMP12, TMP20, TMP12); + vis_mul8x16al(REF_S0, CONST_512, TMP20); + + vis_padd16(TMP14, TMP22, TMP14); + vis_mul8x16al(REF_S0_1, CONST_512, TMP22); + + vis_padd16(TMP16, TMP24, TMP16); + vis_mul8x16al(REF_S2, CONST_512, TMP24); + + vis_padd16(TMP18, TMP26, TMP18); + vis_mul8x16al(REF_S2_1, CONST_512, TMP26); + + vis_padd16(TMP12, TMP0, TMP12); + vis_mul8x16au(REF_2, CONST_256, TMP28); + + vis_padd16(TMP14, TMP2, TMP14); + vis_mul8x16au(REF_2_1, CONST_256, TMP30); + + vis_padd16(TMP16, TMP4, TMP16); + vis_mul8x16au(REF_6, CONST_256, REF_S4); + + vis_padd16(TMP18, TMP6, TMP18); + vis_mul8x16au(REF_6_1, CONST_256, REF_S6); + + vis_pack16(TMP12, DST_0); + vis_padd16(TMP28, TMP0, TMP12); + + vis_pack16(TMP14, DST_1); + vis_st64(DST_0, dest[0]); + vis_padd16(TMP30, TMP2, TMP14); + + vis_pack16(TMP16, DST_2); + vis_padd16(REF_S4, TMP4, TMP16); + + vis_pack16(TMP18, DST_3); + vis_st64_2(DST_2, dest, 8); + dest += stride; + vis_padd16(REF_S6, TMP6, TMP18); + + vis_padd16(TMP12, TMP20, TMP12); + + vis_padd16(TMP14, TMP22, TMP14); + vis_pack16(TMP12, DST_0); + + vis_padd16(TMP16, TMP24, TMP16); + vis_pack16(TMP14, DST_1); + vis_st64(DST_0, dest[0]); + + vis_padd16(TMP18, TMP26, TMP18); + vis_pack16(TMP16, DST_2); + + vis_pack16(TMP18, DST_3); + vis_st64_2(DST_2, dest, 8); + dest += stride; + } while (--height); +} + +static void MC_avg_y_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + int stride_8; + int offset; + + vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT); + + ref = vis_alignaddr(ref); + offset = (ref != _ref) ? 8 : 0; + + vis_ld64(ref[ 0], TMP0); + vis_fzero(ZERO); + + vis_ld64_2(ref, offset, TMP2); + stride_8 = stride + offset; + + vis_ld64(constants3[0], CONST_3); + vis_faligndata(TMP0, TMP2, REF_2); + + vis_ld64(constants256_512[0], CONST_256); + + height >>= 1; + do { /* 20 cycles */ + vis_ld64_2(ref, stride, TMP0); + vis_pmerge(ZERO, REF_2, TMP8); + vis_mul8x16au(REF_2_1, CONST_256, TMP10); + + vis_ld64_2(ref, stride_8, TMP2); + ref += stride; + + vis_ld64(dest[0], DST_0); + + vis_ld64_2(dest, stride, DST_2); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, stride, TMP4); + vis_mul8x16al(DST_0, CONST_512, TMP16); + vis_pmerge(ZERO, REF_0, TMP12); + + vis_ld64_2(ref, stride_8, TMP6); + ref += stride; + vis_mul8x16al(DST_1, CONST_512, TMP18); + vis_pmerge(ZERO, REF_0_1, TMP14); + + vis_padd16(TMP12, CONST_3, TMP12); + vis_mul8x16al(DST_2, CONST_512, TMP24); + + vis_padd16(TMP14, CONST_3, TMP14); + vis_mul8x16al(DST_3, CONST_512, TMP26); + + vis_faligndata(TMP4, TMP6, REF_2); + + vis_padd16(TMP8, TMP12, TMP8); + + vis_padd16(TMP10, TMP14, TMP10); + vis_mul8x16au(REF_2, CONST_256, TMP20); + + vis_padd16(TMP8, TMP16, TMP0); + vis_mul8x16au(REF_2_1, CONST_256, TMP22); + + vis_padd16(TMP10, TMP18, TMP2); + vis_pack16(TMP0, DST_0); + + vis_pack16(TMP2, DST_1); + vis_st64(DST_0, dest[0]); + dest += stride; + vis_padd16(TMP12, TMP20, TMP12); + + vis_padd16(TMP14, TMP22, TMP14); + + vis_padd16(TMP12, TMP24, TMP0); + + vis_padd16(TMP14, TMP26, TMP2); + vis_pack16(TMP0, DST_2); + + vis_pack16(TMP2, DST_3); + vis_st64(DST_2, dest[0]); + dest += stride; + } while (--height); +} + +static void MC_put_xy_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + int stride_8 = stride + 8; + int stride_16 = stride + 16; + + vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT); + + ref = vis_alignaddr(ref); + + vis_ld64(ref[ 0], TMP0); + vis_fzero(ZERO); + + vis_ld64(ref[ 8], TMP2); + + vis_ld64(ref[16], TMP4); + + vis_ld64(constants2[0], CONST_2); + vis_faligndata(TMP0, TMP2, REF_S0); + + vis_ld64(constants256_512[0], CONST_256); + vis_faligndata(TMP2, TMP4, REF_S4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_S2); + vis_faligndata(TMP2, TMP4, REF_S6); + } else { + vis_src1(TMP2, REF_S2); + vis_src1(TMP4, REF_S6); + } + + height >>= 1; + do { + vis_ld64_2(ref, stride, TMP0); + vis_mul8x16au(REF_S0, CONST_256, TMP12); + vis_pmerge(ZERO, REF_S0_1, TMP14); + + vis_alignaddr_g0((void *)off); + + vis_ld64_2(ref, stride_8, TMP2); + vis_mul8x16au(REF_S2, CONST_256, TMP16); + vis_pmerge(ZERO, REF_S2_1, TMP18); + + vis_ld64_2(ref, stride_16, TMP4); + ref += stride; + vis_mul8x16au(REF_S4, CONST_256, TMP20); + vis_pmerge(ZERO, REF_S4_1, TMP22); + + vis_ld64_2(ref, stride, TMP6); + vis_mul8x16au(REF_S6, CONST_256, TMP24); + vis_pmerge(ZERO, REF_S6_1, TMP26); + + vis_ld64_2(ref, stride_8, TMP8); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, stride_16, TMP10); + ref += stride; + vis_faligndata(TMP2, TMP4, REF_4); + + vis_faligndata(TMP6, TMP8, REF_S0); + + vis_faligndata(TMP8, TMP10, REF_S4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + vis_faligndata(TMP2, TMP4, REF_6); + vis_faligndata(TMP6, TMP8, REF_S2); + vis_faligndata(TMP8, TMP10, REF_S6); + } else { + vis_src1(TMP2, REF_2); + vis_src1(TMP4, REF_6); + vis_src1(TMP8, REF_S2); + vis_src1(TMP10, REF_S6); + } + + vis_mul8x16au(REF_0, CONST_256, TMP0); + vis_pmerge(ZERO, REF_0_1, TMP2); + + vis_mul8x16au(REF_2, CONST_256, TMP4); + vis_pmerge(ZERO, REF_2_1, TMP6); + + vis_padd16(TMP0, CONST_2, TMP8); + vis_mul8x16au(REF_4, CONST_256, TMP0); + + vis_padd16(TMP2, CONST_2, TMP10); + vis_mul8x16au(REF_4_1, CONST_256, TMP2); + + vis_padd16(TMP8, TMP4, TMP8); + vis_mul8x16au(REF_6, CONST_256, TMP4); + + vis_padd16(TMP10, TMP6, TMP10); + vis_mul8x16au(REF_6_1, CONST_256, TMP6); + + vis_padd16(TMP12, TMP8, TMP12); + + vis_padd16(TMP14, TMP10, TMP14); + + vis_padd16(TMP12, TMP16, TMP12); + + vis_padd16(TMP14, TMP18, TMP14); + vis_pack16(TMP12, DST_0); + + vis_pack16(TMP14, DST_1); + vis_st64(DST_0, dest[0]); + vis_padd16(TMP0, CONST_2, TMP12); + + vis_mul8x16au(REF_S0, CONST_256, TMP0); + vis_padd16(TMP2, CONST_2, TMP14); + + vis_mul8x16au(REF_S0_1, CONST_256, TMP2); + vis_padd16(TMP12, TMP4, TMP12); + + vis_mul8x16au(REF_S2, CONST_256, TMP4); + vis_padd16(TMP14, TMP6, TMP14); + + vis_mul8x16au(REF_S2_1, CONST_256, TMP6); + vis_padd16(TMP20, TMP12, TMP20); + + vis_padd16(TMP22, TMP14, TMP22); + + vis_padd16(TMP20, TMP24, TMP20); + + vis_padd16(TMP22, TMP26, TMP22); + vis_pack16(TMP20, DST_2); + + vis_pack16(TMP22, DST_3); + vis_st64_2(DST_2, dest, 8); + dest += stride; + vis_padd16(TMP0, TMP4, TMP24); + + vis_mul8x16au(REF_S4, CONST_256, TMP0); + vis_padd16(TMP2, TMP6, TMP26); + + vis_mul8x16au(REF_S4_1, CONST_256, TMP2); + vis_padd16(TMP24, TMP8, TMP24); + + vis_padd16(TMP26, TMP10, TMP26); + vis_pack16(TMP24, DST_0); + + vis_pack16(TMP26, DST_1); + vis_st64(DST_0, dest[0]); + vis_pmerge(ZERO, REF_S6, TMP4); + + vis_pmerge(ZERO, REF_S6_1, TMP6); + + vis_padd16(TMP0, TMP4, TMP0); + + vis_padd16(TMP2, TMP6, TMP2); + + vis_padd16(TMP0, TMP12, TMP0); + + vis_padd16(TMP2, TMP14, TMP2); + vis_pack16(TMP0, DST_2); + + vis_pack16(TMP2, DST_3); + vis_st64_2(DST_2, dest, 8); + dest += stride; + } while (--height); +} + +static void MC_put_xy_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + int stride_8 = stride + 8; + + vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT); + + ref = vis_alignaddr(ref); + + vis_ld64(ref[ 0], TMP0); + vis_fzero(ZERO); + + vis_ld64(ref[ 8], TMP2); + + vis_ld64(constants2[0], CONST_2); + + vis_ld64(constants256_512[0], CONST_256); + vis_faligndata(TMP0, TMP2, REF_S0); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_S2); + } else { + vis_src1(TMP2, REF_S2); + } + + height >>= 1; + do { /* 26 cycles */ + vis_ld64_2(ref, stride, TMP0); + vis_mul8x16au(REF_S0, CONST_256, TMP8); + vis_pmerge(ZERO, REF_S2, TMP12); + + vis_alignaddr_g0((void *)off); + + vis_ld64_2(ref, stride_8, TMP2); + ref += stride; + vis_mul8x16au(REF_S0_1, CONST_256, TMP10); + vis_pmerge(ZERO, REF_S2_1, TMP14); + + vis_ld64_2(ref, stride, TMP4); + + vis_ld64_2(ref, stride_8, TMP6); + ref += stride; + vis_faligndata(TMP0, TMP2, REF_S4); + + vis_pmerge(ZERO, REF_S4, TMP18); + + vis_pmerge(ZERO, REF_S4_1, TMP20); + + vis_faligndata(TMP4, TMP6, REF_S0); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_S6); + vis_faligndata(TMP4, TMP6, REF_S2); + } else { + vis_src1(TMP2, REF_S6); + vis_src1(TMP6, REF_S2); + } + + vis_padd16(TMP18, CONST_2, TMP18); + vis_mul8x16au(REF_S6, CONST_256, TMP22); + + vis_padd16(TMP20, CONST_2, TMP20); + vis_mul8x16au(REF_S6_1, CONST_256, TMP24); + + vis_mul8x16au(REF_S0, CONST_256, TMP26); + vis_pmerge(ZERO, REF_S0_1, TMP28); + + vis_mul8x16au(REF_S2, CONST_256, TMP30); + vis_padd16(TMP18, TMP22, TMP18); + + vis_mul8x16au(REF_S2_1, CONST_256, TMP32); + vis_padd16(TMP20, TMP24, TMP20); + + vis_padd16(TMP8, TMP18, TMP8); + + vis_padd16(TMP10, TMP20, TMP10); + + vis_padd16(TMP8, TMP12, TMP8); + + vis_padd16(TMP10, TMP14, TMP10); + vis_pack16(TMP8, DST_0); + + vis_pack16(TMP10, DST_1); + vis_st64(DST_0, dest[0]); + dest += stride; + vis_padd16(TMP18, TMP26, TMP18); + + vis_padd16(TMP20, TMP28, TMP20); + + vis_padd16(TMP18, TMP30, TMP18); + + vis_padd16(TMP20, TMP32, TMP20); + vis_pack16(TMP18, DST_2); + + vis_pack16(TMP20, DST_3); + vis_st64(DST_2, dest[0]); + dest += stride; + } while (--height); +} + +static void MC_avg_xy_16_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + int stride_8 = stride + 8; + int stride_16 = stride + 16; + + vis_set_gsr(4 << VIS_GSR_SCALEFACT_SHIFT); + + ref = vis_alignaddr(ref); + + vis_ld64(ref[ 0], TMP0); + vis_fzero(ZERO); + + vis_ld64(ref[ 8], TMP2); + + vis_ld64(ref[16], TMP4); + + vis_ld64(constants6[0], CONST_6); + vis_faligndata(TMP0, TMP2, REF_S0); + + vis_ld64(constants256_1024[0], CONST_256); + vis_faligndata(TMP2, TMP4, REF_S4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_S2); + vis_faligndata(TMP2, TMP4, REF_S6); + } else { + vis_src1(TMP2, REF_S2); + vis_src1(TMP4, REF_S6); + } + + height >>= 1; + do { /* 55 cycles */ + vis_ld64_2(ref, stride, TMP0); + vis_mul8x16au(REF_S0, CONST_256, TMP12); + vis_pmerge(ZERO, REF_S0_1, TMP14); + + vis_alignaddr_g0((void *)off); + + vis_ld64_2(ref, stride_8, TMP2); + vis_mul8x16au(REF_S2, CONST_256, TMP16); + vis_pmerge(ZERO, REF_S2_1, TMP18); + + vis_ld64_2(ref, stride_16, TMP4); + ref += stride; + vis_mul8x16au(REF_S4, CONST_256, TMP20); + vis_pmerge(ZERO, REF_S4_1, TMP22); + + vis_ld64_2(ref, stride, TMP6); + vis_mul8x16au(REF_S6, CONST_256, TMP24); + vis_pmerge(ZERO, REF_S6_1, TMP26); + + vis_ld64_2(ref, stride_8, TMP8); + vis_faligndata(TMP0, TMP2, REF_0); + + vis_ld64_2(ref, stride_16, TMP10); + ref += stride; + vis_faligndata(TMP2, TMP4, REF_4); + + vis_ld64(dest[0], DST_0); + vis_faligndata(TMP6, TMP8, REF_S0); + + vis_ld64_2(dest, 8, DST_2); + vis_faligndata(TMP8, TMP10, REF_S4); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_2); + vis_faligndata(TMP2, TMP4, REF_6); + vis_faligndata(TMP6, TMP8, REF_S2); + vis_faligndata(TMP8, TMP10, REF_S6); + } else { + vis_src1(TMP2, REF_2); + vis_src1(TMP4, REF_6); + vis_src1(TMP8, REF_S2); + vis_src1(TMP10, REF_S6); + } + + vis_mul8x16al(DST_0, CONST_1024, TMP30); + vis_pmerge(ZERO, REF_0, TMP0); + + vis_mul8x16al(DST_1, CONST_1024, TMP32); + vis_pmerge(ZERO, REF_0_1, TMP2); + + vis_mul8x16au(REF_2, CONST_256, TMP4); + vis_pmerge(ZERO, REF_2_1, TMP6); + + vis_mul8x16al(DST_2, CONST_1024, REF_0); + vis_padd16(TMP0, CONST_6, TMP0); + + vis_mul8x16al(DST_3, CONST_1024, REF_2); + vis_padd16(TMP2, CONST_6, TMP2); + + vis_padd16(TMP0, TMP4, TMP0); + vis_mul8x16au(REF_4, CONST_256, TMP4); + + vis_padd16(TMP2, TMP6, TMP2); + vis_mul8x16au(REF_4_1, CONST_256, TMP6); + + vis_padd16(TMP12, TMP0, TMP12); + vis_mul8x16au(REF_6, CONST_256, TMP8); + + vis_padd16(TMP14, TMP2, TMP14); + vis_mul8x16au(REF_6_1, CONST_256, TMP10); + + vis_padd16(TMP12, TMP16, TMP12); + vis_mul8x16au(REF_S0, CONST_256, REF_4); + + vis_padd16(TMP14, TMP18, TMP14); + vis_mul8x16au(REF_S0_1, CONST_256, REF_6); + + vis_padd16(TMP12, TMP30, TMP12); + + vis_padd16(TMP14, TMP32, TMP14); + vis_pack16(TMP12, DST_0); + + vis_pack16(TMP14, DST_1); + vis_st64(DST_0, dest[0]); + vis_padd16(TMP4, CONST_6, TMP4); + + vis_ld64_2(dest, stride, DST_0); + vis_padd16(TMP6, CONST_6, TMP6); + vis_mul8x16au(REF_S2, CONST_256, TMP12); + + vis_padd16(TMP4, TMP8, TMP4); + vis_mul8x16au(REF_S2_1, CONST_256, TMP14); + + vis_padd16(TMP6, TMP10, TMP6); + + vis_padd16(TMP20, TMP4, TMP20); + + vis_padd16(TMP22, TMP6, TMP22); + + vis_padd16(TMP20, TMP24, TMP20); + + vis_padd16(TMP22, TMP26, TMP22); + + vis_padd16(TMP20, REF_0, TMP20); + vis_mul8x16au(REF_S4, CONST_256, REF_0); + + vis_padd16(TMP22, REF_2, TMP22); + vis_pack16(TMP20, DST_2); + + vis_pack16(TMP22, DST_3); + vis_st64_2(DST_2, dest, 8); + dest += stride; + + vis_ld64_2(dest, 8, DST_2); + vis_mul8x16al(DST_0, CONST_1024, TMP30); + vis_pmerge(ZERO, REF_S4_1, REF_2); + + vis_mul8x16al(DST_1, CONST_1024, TMP32); + vis_padd16(REF_4, TMP0, TMP8); + + vis_mul8x16au(REF_S6, CONST_256, REF_4); + vis_padd16(REF_6, TMP2, TMP10); + + vis_mul8x16au(REF_S6_1, CONST_256, REF_6); + vis_padd16(TMP8, TMP12, TMP8); + + vis_padd16(TMP10, TMP14, TMP10); + + vis_padd16(TMP8, TMP30, TMP8); + + vis_padd16(TMP10, TMP32, TMP10); + vis_pack16(TMP8, DST_0); + + vis_pack16(TMP10, DST_1); + vis_st64(DST_0, dest[0]); + + vis_padd16(REF_0, TMP4, REF_0); + + vis_mul8x16al(DST_2, CONST_1024, TMP30); + vis_padd16(REF_2, TMP6, REF_2); + + vis_mul8x16al(DST_3, CONST_1024, TMP32); + vis_padd16(REF_0, REF_4, REF_0); + + vis_padd16(REF_2, REF_6, REF_2); + + vis_padd16(REF_0, TMP30, REF_0); + + /* stall */ + + vis_padd16(REF_2, TMP32, REF_2); + vis_pack16(REF_0, DST_2); + + vis_pack16(REF_2, DST_3); + vis_st64_2(DST_2, dest, 8); + dest += stride; + } while (--height); +} + +static void MC_avg_xy_8_vis (uint8_t * dest, uint8_t * _ref, + int stride, int height) +{ + uint8_t *ref = (uint8_t *) _ref; + unsigned long off = (unsigned long) ref & 0x7; + unsigned long off_plus_1 = off + 1; + int stride_8 = stride + 8; + + vis_set_gsr(4 << VIS_GSR_SCALEFACT_SHIFT); + + ref = vis_alignaddr(ref); + + vis_ld64(ref[0], TMP0); + vis_fzero(ZERO); + + vis_ld64_2(ref, 8, TMP2); + + vis_ld64(constants6[0], CONST_6); + + vis_ld64(constants256_1024[0], CONST_256); + vis_faligndata(TMP0, TMP2, REF_S0); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_S2); + } else { + vis_src1(TMP2, REF_S2); + } + + height >>= 1; + do { /* 31 cycles */ + vis_ld64_2(ref, stride, TMP0); + vis_mul8x16au(REF_S0, CONST_256, TMP8); + vis_pmerge(ZERO, REF_S0_1, TMP10); + + vis_ld64_2(ref, stride_8, TMP2); + ref += stride; + vis_mul8x16au(REF_S2, CONST_256, TMP12); + vis_pmerge(ZERO, REF_S2_1, TMP14); + + vis_alignaddr_g0((void *)off); + + vis_ld64_2(ref, stride, TMP4); + vis_faligndata(TMP0, TMP2, REF_S4); + + vis_ld64_2(ref, stride_8, TMP6); + ref += stride; + + vis_ld64(dest[0], DST_0); + vis_faligndata(TMP4, TMP6, REF_S0); + + vis_ld64_2(dest, stride, DST_2); + + if (off != 0x7) { + vis_alignaddr_g0((void *)off_plus_1); + vis_faligndata(TMP0, TMP2, REF_S6); + vis_faligndata(TMP4, TMP6, REF_S2); + } else { + vis_src1(TMP2, REF_S6); + vis_src1(TMP6, REF_S2); + } + + vis_mul8x16al(DST_0, CONST_1024, TMP30); + vis_pmerge(ZERO, REF_S4, TMP22); + + vis_mul8x16al(DST_1, CONST_1024, TMP32); + vis_pmerge(ZERO, REF_S4_1, TMP24); + + vis_mul8x16au(REF_S6, CONST_256, TMP26); + vis_pmerge(ZERO, REF_S6_1, TMP28); + + vis_mul8x16au(REF_S0, CONST_256, REF_S4); + vis_padd16(TMP22, CONST_6, TMP22); + + vis_mul8x16au(REF_S0_1, CONST_256, REF_S6); + vis_padd16(TMP24, CONST_6, TMP24); + + vis_mul8x16al(DST_2, CONST_1024, REF_0); + vis_padd16(TMP22, TMP26, TMP22); + + vis_mul8x16al(DST_3, CONST_1024, REF_2); + vis_padd16(TMP24, TMP28, TMP24); + + vis_mul8x16au(REF_S2, CONST_256, TMP26); + vis_padd16(TMP8, TMP22, TMP8); + + vis_mul8x16au(REF_S2_1, CONST_256, TMP28); + vis_padd16(TMP10, TMP24, TMP10); + + vis_padd16(TMP8, TMP12, TMP8); + + vis_padd16(TMP10, TMP14, TMP10); + + vis_padd16(TMP8, TMP30, TMP8); + + vis_padd16(TMP10, TMP32, TMP10); + vis_pack16(TMP8, DST_0); + + vis_pack16(TMP10, DST_1); + vis_st64(DST_0, dest[0]); + dest += stride; + + vis_padd16(REF_S4, TMP22, TMP12); + + vis_padd16(REF_S6, TMP24, TMP14); + + vis_padd16(TMP12, TMP26, TMP12); + + vis_padd16(TMP14, TMP28, TMP14); + + vis_padd16(TMP12, REF_0, TMP12); + + vis_padd16(TMP14, REF_2, TMP14); + vis_pack16(TMP12, DST_2); + + vis_pack16(TMP14, DST_3); + vis_st64(DST_2, dest[0]); + dest += stride; + } while (--height); +} + +MPEG2_MC_EXTERN(vis); + +#endif /* defined(ARCH_SPARC) && defined(ENABLE_VIS) */ diff --git a/src/video_dec/libmpeg2/mpeg2.h b/src/video_dec/libmpeg2/mpeg2.h new file mode 100644 index 000000000..ae69688f5 --- /dev/null +++ b/src/video_dec/libmpeg2/mpeg2.h @@ -0,0 +1,100 @@ +/* + * mpeg2.h + * 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 + */ + +/* Structure for the mpeg2dec decoder */ + +#ifndef MPEG2_H +#define MPEG2_H + +#include "libmpeg2_accel.h" + +typedef struct mpeg2dec_s { + xine_video_port_t * output; + uint32_t frame_format; + + /* this is where we keep the state of the decoder */ + struct picture_s * picture; + void *picture_base; + + uint32_t shift; + int new_sequence; + int is_sequence_needed; + int is_wait_for_ip_frames; + int frames_to_drop, drop_frame; + int in_slice; + int seek_mode, is_frame_needed; + + /* the maximum chunk size is determined by vbv_buffer_size */ + /* which is 224K for MP@ML streams. */ + /* (we make no pretenses of decoding anything more than that) */ + /* allocated in init - gcc has problems allocating such big structures */ + uint8_t * chunk_buffer; + void *chunk_base; + /* pointer to current position in chunk_buffer */ + uint8_t * chunk_ptr; + /* last start code ? */ + uint8_t code; + uint32_t chunk_size; + + int64_t pts; + uint32_t rff_pattern; + int force_aspect; + int force_pan_scan; + + /* AFD data can be found after a sequence, group or picture start code */ + /* and will be stored in afd_value_seen. Later it will be transfered to */ + /* a stream property and stored into afd_value_reported to detect changes */ + int afd_value_seen; + int afd_value_reported; + + xine_stream_t *stream; + + /* a spu decoder for possible closed captions */ + spu_decoder_t *cc_dec; + mpeg2dec_accel_t accel; + +} mpeg2dec_t ; + + +/* initialize mpegdec with a opaque user pointer */ +void mpeg2_init (mpeg2dec_t * mpeg2dec, + xine_video_port_t * output); + +/* destroy everything which was allocated, shutdown the output */ +void mpeg2_close (mpeg2dec_t * mpeg2dec); + +int mpeg2_decode_data (mpeg2dec_t * mpeg2dec, + uint8_t * data_start, uint8_t * data_end, + uint64_t pts); + +void mpeg2_find_sequence_header (mpeg2dec_t * mpeg2dec, + uint8_t * data_start, uint8_t * data_end); + +void mpeg2_flush (mpeg2dec_t * mpeg2dec); +void mpeg2_reset (mpeg2dec_t * mpeg2dec); +void mpeg2_discontinuity (mpeg2dec_t * mpeg2dec); + +/* Not needed, it is defined as static in decode.c, and no-one else called it + * currently + */ +/* void process_userdata(mpeg2dec_t *mpeg2dec, uint8_t *buffer); */ + +#endif diff --git a/src/video_dec/libmpeg2/mpeg2_internal.h b/src/video_dec/libmpeg2/mpeg2_internal.h new file mode 100644 index 000000000..eeaa16227 --- /dev/null +++ b/src/video_dec/libmpeg2/mpeg2_internal.h @@ -0,0 +1,294 @@ +/* + * mpeg2_internal.h + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#ifndef MPEG2_INTERNAL_H +#define MPEG2_INTERNAL_H + +#include <xine/video_out.h> +#include "accel_xvmc.h" + +#ifdef ENABLE_ALTIVEC +#include <altivec.h> +#endif + +/* macroblock modes */ +#define MACROBLOCK_INTRA XINE_MACROBLOCK_INTRA +#define MACROBLOCK_PATTERN XINE_MACROBLOCK_PATTERN +#define MACROBLOCK_MOTION_BACKWARD XINE_MACROBLOCK_MOTION_BACKWARD +#define MACROBLOCK_MOTION_FORWARD XINE_MACROBLOCK_MOTION_FORWARD +#define MACROBLOCK_QUANT XINE_MACROBLOCK_QUANT +#define DCT_TYPE_INTERLACED XINE_MACROBLOCK_DCT_TYPE_INTERLACED + +/* motion_type */ +#define MOTION_TYPE_MASK (3*64) +#define MOTION_TYPE_BASE 64 +#define MC_FIELD (1*64) +#define MC_FRAME (2*64) +#define MC_16X8 (2*64) +#define MC_DMV (3*64) + +/* picture structure */ +#define TOP_FIELD VO_TOP_FIELD +#define BOTTOM_FIELD VO_BOTTOM_FIELD +#define FRAME_PICTURE VO_BOTH_FIELDS + +/* picture coding type (mpeg2 header) */ +#define I_TYPE 1 +#define P_TYPE 2 +#define B_TYPE 3 +#define D_TYPE 4 + +typedef struct motion_s { + uint8_t * ref[2][3]; + uint8_t ** ref2[2]; + int pmv[2][2]; + int f_code[2]; +} motion_t; + +typedef struct picture_s { + /* first, state that carries information from one macroblock to the */ + /* next inside a slice, and is never used outside of mpeg2_slice() */ + + /* DCT coefficients - should be kept aligned ! */ + int16_t DCTblock[64]; + + /* XvMC DCT block and macroblock data for XvMC acceleration */ + xine_macroblocks_t *mc; + int XvMC_mb_type; + int XvMC_mv_field_sel[2][2]; + int XvMC_x; + int XvMC_y; + int XvMC_motion_type; + int XvMC_dmvector[2]; + int XvMC_cbp; + int XvMC_dct_type; + + /* bit parsing stuff */ + uint32_t bitstream_buf; /* current 32 bit working set of buffer */ + int bitstream_bits; /* used bits in working set */ + uint8_t * bitstream_ptr; /* buffer with stream data */ + + uint8_t * dest[3]; + int pitches[3]; + int offset; + unsigned int limit_x; + unsigned int limit_y_16; + unsigned int limit_y_8; + unsigned int limit_y; + + /* Motion vectors */ + /* The f_ and b_ correspond to the forward and backward motion */ + /* predictors */ + motion_t b_motion; + motion_t f_motion; + + /* predictor for DC coefficients in intra blocks */ + int16_t dc_dct_pred[3]; + + int quantizer_scale; /* remove */ + int current_field; /* remove */ + int dmv_offset; /* remove */ + unsigned int v_offset; /* remove */ + + + /* now non-slice-specific information */ + + /* sequence header stuff */ + uint8_t intra_quantizer_matrix [64]; + uint8_t non_intra_quantizer_matrix [64]; + int load_intra_quantizer_matrix; + int load_non_intra_quantizer_matrix; + + /* The width and height of the picture snapped to macroblock units */ + int coded_picture_width; + int coded_picture_height; + + /* The width and height as it appears on header sequence */ + unsigned int display_width, display_height; + + /* picture header stuff */ + + /* what type of picture this is (I, P, B, D) */ + int picture_coding_type; + + int vbv_delay; + int low_delay; + + /* picture coding extension stuff */ + + /* quantization factor for intra dc coefficients */ + int intra_dc_precision; + /* top/bottom/both fields */ + int picture_structure; + /* bool to indicate all predictions are frame based */ + int frame_pred_frame_dct; + /* bool to indicate whether intra blocks have motion vectors */ + /* (for concealment) */ + int concealment_motion_vectors; + /* bit to indicate which quantization table to use */ + int q_scale_type; + /* bool to use different vlc tables */ + int intra_vlc_format; + /* used for DMV MC */ + int top_field_first; + + /* stuff derived from bitstream */ + + /* pointer to the zigzag scan we're supposed to be using */ + uint8_t * scan; + + struct vo_frame_s * current_frame; + struct vo_frame_s * forward_reference_frame; + struct vo_frame_s * backward_reference_frame; + + int frame_width, frame_height; + + int second_field; + + int mpeg1; + + int skip_non_intra_dct; + + /* these things are not needed by the decoder */ + /* this is a temporary interface, we will build a better one later. */ + int aspect_ratio_information; + int saved_aspect_ratio; + int frame_rate_code; + int progressive_sequence; + int repeat_first_field; + int progressive_frame; + int32_t frame_centre_horizontal_offset; + int32_t frame_centre_vertical_offset; + uint32_t video_format; + uint32_t colour_description; + uint32_t colour_primatives; + uint32_t transfer_characteristics; + uint32_t matrix_coefficients; + uint32_t display_horizontal_size; + uint32_t display_vertical_size; + uint32_t drop_frame_flag; + uint32_t time_code_hours; + uint32_t time_code_minutes; + uint32_t time_code_seconds; + uint32_t time_code_pictures; + uint32_t closed_gop; + uint32_t broken_link; + + int bitrate; + int frame_rate_ext_n; + int frame_rate_ext_d; + +} picture_t; + +typedef struct cpu_state_s { +#ifdef ARCH_PPC + uint8_t regv[12*16]; +#endif + int dummy; +} cpu_state_t; + +/* cpu_state.c */ +extern void (* mpeg2_cpu_state_save) (cpu_state_t * state); +extern void (* mpeg2_cpu_state_restore) (cpu_state_t * state); +void mpeg2_cpu_state_init (uint32_t mm_accel); + +/* header.c */ +extern uint8_t mpeg2_scan_norm[64]; +extern uint8_t mpeg2_scan_alt[64]; +void mpeg2_header_state_init (picture_t * picture); +int mpeg2_header_picture (picture_t * picture, uint8_t * buffer); +int mpeg2_header_sequence (picture_t * picture, uint8_t * buffer); +int mpeg2_header_extension (picture_t * picture, uint8_t * buffer); +int mpeg2_header_group_of_pictures (picture_t * picture, uint8_t * buffer); + +/* idct.c */ +extern void (* mpeg2_idct_copy) (int16_t * block, uint8_t * dest, int stride); +extern void (* mpeg2_idct_add) (int16_t * block, uint8_t * dest, int stride); +extern void (* mpeg2_idct) (int16_t * block); +extern void (* mpeg2_zero_block) (int16_t * block); +void mpeg2_idct_init (uint32_t mm_accel); + +/* idct_mlib.c */ +void mpeg2_idct_add_mlib (int16_t * block, uint8_t * dest, int stride); +void mpeg2_idct_copy_mlib_non_ieee (int16_t * block, uint8_t * dest, + int stride); +void mpeg2_idct_add_mlib_non_ieee (int16_t * block, uint8_t * dest, + int stride); +void mpeg2_idct_mlib (int16_t * block); + +/* idct_mmx.c */ +void mpeg2_idct_copy_mmxext (int16_t * block, uint8_t * dest, int stride); +void mpeg2_idct_add_mmxext (int16_t * block, uint8_t * dest, int stride); +void mpeg2_idct_mmxext (int16_t * block); +void mpeg2_idct_copy_mmx (int16_t * block, uint8_t * dest, int stride); +void mpeg2_idct_add_mmx (int16_t * block, uint8_t * dest, int stride); +void mpeg2_idct_mmx (int16_t * block); +void mpeg2_zero_block_mmx (int16_t * block); +void mpeg2_idct_mmx_init (void); + +/* idct_altivec.c */ +# ifdef ENABLE_ALTIVEC +void mpeg2_idct_copy_altivec (vector signed short * block, unsigned char * dest, + int stride); +void mpeg2_idct_add_altivec (vector signed short * block, unsigned char * dest, + int stride); +# else /* ! ENABLE_ALTIVEC */ +void mpeg2_idct_copy_altivec (signed short * block, unsigned char * dest, + int stride); +void mpeg2_idct_add_altivec (signed short * block, unsigned char * dest, + int stride); +# endif /* ENABLE_ALTIVEC */ +void mpeg2_idct_altivec_init (void); + +/* motion_comp.c */ +void mpeg2_mc_init (uint32_t mm_accel); + +typedef struct mpeg2_mc_s { + void (* put [8]) (uint8_t * dst, uint8_t *, int32_t, int32_t); + void (* avg [8]) (uint8_t * dst, uint8_t *, int32_t, int32_t); +} mpeg2_mc_t; + +#define MPEG2_MC_EXTERN(x) mpeg2_mc_t mpeg2_mc_##x = { \ + {MC_put_o_16_##x, MC_put_x_16_##x, MC_put_y_16_##x, MC_put_xy_16_##x, \ + MC_put_o_8_##x, MC_put_x_8_##x, MC_put_y_8_##x, MC_put_xy_8_##x}, \ + {MC_avg_o_16_##x, MC_avg_x_16_##x, MC_avg_y_16_##x, MC_avg_xy_16_##x, \ + MC_avg_o_8_##x, MC_avg_x_8_##x, MC_avg_y_8_##x, MC_avg_xy_8_##x} \ +}; + +extern mpeg2_mc_t mpeg2_mc; +extern mpeg2_mc_t mpeg2_mc_c; +extern mpeg2_mc_t mpeg2_mc_mmx; +extern mpeg2_mc_t mpeg2_mc_mmxext; +extern mpeg2_mc_t mpeg2_mc_3dnow; +extern mpeg2_mc_t mpeg2_mc_altivec; +extern mpeg2_mc_t mpeg2_mc_mlib; +extern mpeg2_mc_t mpeg2_mc_vis; + +/* slice.c */ +void mpeg2_slice (picture_t * picture, int code, uint8_t * buffer); + +/* stats.c */ +void mpeg2_stats (int code, uint8_t * buffer); + + +#endif diff --git a/src/video_dec/libmpeg2/slice.c b/src/video_dec/libmpeg2/slice.c new file mode 100644 index 000000000..8247a9a24 --- /dev/null +++ b/src/video_dec/libmpeg2/slice.c @@ -0,0 +1,1833 @@ +/* + * slice.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#include <inttypes.h> + +#include <xine/xine_internal.h> +#include <xine/video_out.h> +#include "mpeg2_internal.h" +#include <xine/attributes.h> + +#include "vlc.h" + +static const int non_linear_quantizer_scale [] = { + 0, 1, 2, 3, 4, 5, 6, 7, + 8, 10, 12, 14, 16, 18, 20, 22, + 24, 28, 32, 36, 40, 44, 48, 52, + 56, 64, 72, 80, 88, 96, 104, 112 +}; + +static inline int get_macroblock_modes (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int macroblock_modes; + const MBtab * tab; + + switch (picture->picture_coding_type) { + case I_TYPE: + + tab = MB_I + UBITS (bit_buf, 1); + DUMPBITS (bit_buf, bits, tab->len); + macroblock_modes = tab->modes; + + if ((! (picture->frame_pred_frame_dct)) && + (picture->picture_structure == FRAME_PICTURE)) { + macroblock_modes |= UBITS (bit_buf, 1) * DCT_TYPE_INTERLACED; + DUMPBITS (bit_buf, bits, 1); + } + + return macroblock_modes; + + case P_TYPE: + + tab = MB_P + UBITS (bit_buf, 5); + DUMPBITS (bit_buf, bits, tab->len); + macroblock_modes = tab->modes; + + if (picture->picture_structure != FRAME_PICTURE) { + if (macroblock_modes & MACROBLOCK_MOTION_FORWARD) { + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + } + return macroblock_modes; + } else if (picture->frame_pred_frame_dct) { + if (macroblock_modes & MACROBLOCK_MOTION_FORWARD) + macroblock_modes |= MC_FRAME; + return macroblock_modes; + } else { + if (macroblock_modes & MACROBLOCK_MOTION_FORWARD) { + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + } + if (macroblock_modes & (MACROBLOCK_INTRA | MACROBLOCK_PATTERN)) { + macroblock_modes |= UBITS (bit_buf, 1) * DCT_TYPE_INTERLACED; + DUMPBITS (bit_buf, bits, 1); + } + return macroblock_modes; + } + + case B_TYPE: + + tab = MB_B + UBITS (bit_buf, 6); + DUMPBITS (bit_buf, bits, tab->len); + macroblock_modes = tab->modes; + + if (picture->picture_structure != FRAME_PICTURE) { + if (! (macroblock_modes & MACROBLOCK_INTRA)) { + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + } + return macroblock_modes; + } else if (picture->frame_pred_frame_dct) { + /* if (! (macroblock_modes & MACROBLOCK_INTRA)) */ + macroblock_modes |= MC_FRAME; + return macroblock_modes; + } else { + if (macroblock_modes & MACROBLOCK_INTRA) + goto intra; + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + if (macroblock_modes & (MACROBLOCK_INTRA | MACROBLOCK_PATTERN)) { + intra: + macroblock_modes |= UBITS (bit_buf, 1) * DCT_TYPE_INTERLACED; + DUMPBITS (bit_buf, bits, 1); + } + return macroblock_modes; + } + + case D_TYPE: + + DUMPBITS (bit_buf, bits, 1); + return MACROBLOCK_INTRA; + + default: + return 0; + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_quantizer_scale (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + int quantizer_scale_code; + + quantizer_scale_code = UBITS (bit_buf, 5); + DUMPBITS (bit_buf, bits, 5); + + if (picture->q_scale_type) + return non_linear_quantizer_scale [quantizer_scale_code]; + else + return quantizer_scale_code << 1; +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_motion_delta (picture_t * picture, int f_code) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + int delta; + int sign; + const MVtab * tab; + + if (bit_buf & 0x80000000) { + DUMPBITS (bit_buf, bits, 1); + return 0; + } else if (bit_buf >= 0x0c000000) { + + tab = MV_4 + UBITS (bit_buf, 4); + delta = (tab->delta << f_code) + 1; + bits += tab->len + f_code + 1; + bit_buf <<= tab->len; + + sign = SBITS (bit_buf, 1); + bit_buf <<= 1; + + if (f_code) + delta += UBITS (bit_buf, f_code); + bit_buf <<= f_code; + + return (delta ^ sign) - sign; + + } else { + + tab = MV_10 + UBITS (bit_buf, 10); + delta = (tab->delta << f_code) + 1; + bits += tab->len + 1; + bit_buf <<= tab->len; + + sign = SBITS (bit_buf, 1); + bit_buf <<= 1; + + if (f_code) { + NEEDBITS (bit_buf, bits, bit_ptr); + delta += UBITS (bit_buf, f_code); + DUMPBITS (bit_buf, bits, f_code); + } + + return (delta ^ sign) - sign; + + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int bound_motion_vector (int vec, int f_code) +{ +#if 1 + unsigned int limit; + int sign; + + limit = 16 << f_code; + + if ((unsigned int)(vec + limit) < 2 * limit) + return vec; + else { + sign = ((int32_t)vec) >> 31; + return vec - ((2 * limit) ^ sign) + sign; + } +#else + return ((int32_t)vector << (27 - f_code)) >> (27 - f_code); +#endif +} + +static inline int get_dmv (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + const DMVtab * tab; + + tab = DMV_2 + UBITS (bit_buf, 2); + DUMPBITS (bit_buf, bits, tab->len); + return tab->dmv; +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_coded_block_pattern (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + const CBPtab * tab; + + NEEDBITS (bit_buf, bits, bit_ptr); + + if (bit_buf >= 0x20000000) { + + tab = CBP_7 + (UBITS (bit_buf, 7) - 16); + DUMPBITS (bit_buf, bits, tab->len); + return tab->cbp; + + } else { + + tab = CBP_9 + UBITS (bit_buf, 9); + DUMPBITS (bit_buf, bits, tab->len); + return tab->cbp; + } + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_luma_dc_dct_diff (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + const DCtab * tab; + int size; + int dc_diff; + + if (bit_buf < 0xf8000000) { + tab = DC_lum_5 + UBITS (bit_buf, 5); + size = tab->size; + if (size) { + bits += tab->len + size; + bit_buf <<= tab->len; + dc_diff = + UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + bit_buf <<= size; + return dc_diff; + } else { + DUMPBITS (bit_buf, bits, 3); + return 0; + } + } else { + tab = DC_long + (UBITS (bit_buf, 9) - 0x1e0); + size = tab->size; + DUMPBITS (bit_buf, bits, tab->len); + NEEDBITS (bit_buf, bits, bit_ptr); + dc_diff = UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + DUMPBITS (bit_buf, bits, size); + return dc_diff; + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_chroma_dc_dct_diff (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + const DCtab * tab; + int size; + int dc_diff; + + if (bit_buf < 0xf8000000) { + tab = DC_chrom_5 + UBITS (bit_buf, 5); + size = tab->size; + if (size) { + bits += tab->len + size; + bit_buf <<= tab->len; + dc_diff = + UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + bit_buf <<= size; + return dc_diff; + } else { + DUMPBITS (bit_buf, bits, 2); + return 0; + } + } else { + tab = DC_long + (UBITS (bit_buf, 10) - 0x3e0); + size = tab->size; + DUMPBITS (bit_buf, bits, tab->len + 1); + NEEDBITS (bit_buf, bits, bit_ptr); + dc_diff = UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + DUMPBITS (bit_buf, bits, size); + return dc_diff; + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +#define SATURATE(val) \ +do { \ + if ((uint32_t)(val + 2048) > 4095) \ + val = (val > 0) ? 2047 : -2048; \ +} while (0) + +static void get_intra_block_B14 (picture_t * picture) +{ + int i; + int j; + int val; + uint8_t * scan = picture->scan; + uint8_t * quant_matrix = picture->intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + int mismatch; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + dest = picture->DCTblock; + i = 0; + mismatch = ~dest[0]; + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + j = scan[i]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = (tab->level * quantizer_scale * quant_matrix[j]) >> 4; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + j = scan[i]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = (SBITS (bit_buf, 12) * + quantizer_scale * quant_matrix[j]) / 16; + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + dest[63] ^= mismatch & 1; + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_intra_block_B15 (picture_t * picture) +{ + int i; + int j; + int val; + uint8_t * scan = picture->scan; + uint8_t * quant_matrix = picture->intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + int mismatch; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + dest = picture->DCTblock; + i = 0; + mismatch = ~dest[0]; + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + + while (1) { + if (bit_buf >= 0x04000000) { + + tab = DCT_B15_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) { + + normal_code: + j = scan[i]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = (tab->level * quantizer_scale * quant_matrix[j]) >> 4; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else { + + /* end of block. I commented out this code because if we */ + /* dont exit here we will still exit at the later test :) */ + + /* if (i >= 128) break; */ /* end of block */ + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check against buffer overflow */ + + j = scan[i]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = (SBITS (bit_buf, 12) * + quantizer_scale * quant_matrix[j]) / 16; + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } + } else if (bit_buf >= 0x02000000) { + tab = DCT_B15_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + dest[63] ^= mismatch & 1; + DUMPBITS (bit_buf, bits, 4); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_non_intra_block (picture_t * picture) +{ + int i; + int j; + int val; + uint8_t * scan = picture->scan; + uint8_t * quant_matrix = picture->non_intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + int mismatch; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + i = -1; + mismatch = 1; + dest = picture->DCTblock; + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + if (bit_buf >= 0x28000000) { + tab = DCT_B14DC_5 + (UBITS (bit_buf, 5) - 5); + goto entry_1; + } else + goto entry_2; + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + entry_1: + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + j = scan[i]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = ((2*tab->level+1) * quantizer_scale * quant_matrix[j]) >> 5; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } + + entry_2: + if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + j = scan[i]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = 2 * (SBITS (bit_buf, 12) + SBITS (bit_buf, 1)) + 1; + val = (val * quantizer_scale * quant_matrix[j]) / 32; + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + dest[63] ^= mismatch & 1; + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_mpeg1_intra_block (picture_t * picture) +{ + int i; + int j; + int val; + uint8_t * scan = picture->scan; + uint8_t * quant_matrix = picture->intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + i = 0; + dest = picture->DCTblock; + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + j = scan[i]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = (tab->level * quantizer_scale * quant_matrix[j]) >> 4; + + /* oddification */ + val = (val - 1) | 1; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + j = scan[i]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = SBITS (bit_buf, 8); + if (! (val & 0x7f)) { + DUMPBITS (bit_buf, bits, 8); + val = UBITS (bit_buf, 8) + 2 * val; + } + val = (val * quantizer_scale * quant_matrix[j]) / 16; + + /* oddification */ + val = (val + ~SBITS (val, 1)) | 1; + + SATURATE (val); + dest[j] = val; + + DUMPBITS (bit_buf, bits, 8); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_mpeg1_non_intra_block (picture_t * picture) +{ + int i; + int j; + int val; + uint8_t * scan = picture->scan; + uint8_t * quant_matrix = picture->non_intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + i = -1; + dest = picture->DCTblock; + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + if (bit_buf >= 0x28000000) { + tab = DCT_B14DC_5 + (UBITS (bit_buf, 5) - 5); + goto entry_1; + } else + goto entry_2; + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + entry_1: + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + j = scan[i]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = ((2*tab->level+1) * quantizer_scale * quant_matrix[j]) >> 5; + + /* oddification */ + val = (val - 1) | 1; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } + + entry_2: + if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + j = scan[i]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = SBITS (bit_buf, 8); + if (! (val & 0x7f)) { + DUMPBITS (bit_buf, bits, 8); + val = UBITS (bit_buf, 8) + 2 * val; + } + val = 2 * (val + SBITS (val, 1)) + 1; + val = (val * quantizer_scale * quant_matrix[j]) / 32; + + /* oddification */ + val = (val + ~SBITS (val, 1)) | 1; + + SATURATE (val); + dest[j] = val; + + DUMPBITS (bit_buf, bits, 8); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static inline void slice_intra_DCT (picture_t * picture, int cc, + uint8_t * dest, int stride) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + NEEDBITS (bit_buf, bits, bit_ptr); + /* Get the intra DC coefficient and inverse quantize it */ + if (cc == 0) + picture->dc_dct_pred[0] += get_luma_dc_dct_diff (picture); + else + picture->dc_dct_pred[cc] += get_chroma_dc_dct_diff (picture); + picture->DCTblock[0] = + picture->dc_dct_pred[cc] << (3 - picture->intra_dc_precision); + + if (picture->mpeg1) { + if (picture->picture_coding_type != D_TYPE) + get_mpeg1_intra_block (picture); + } else if (picture->intra_vlc_format) + get_intra_block_B15 (picture); + else + get_intra_block_B14 (picture); + mpeg2_idct_copy (picture->DCTblock, dest, stride); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline void slice_non_intra_DCT (picture_t * picture, uint8_t * dest, + int stride) +{ + if (picture->mpeg1) + get_mpeg1_non_intra_block (picture); + else + get_non_intra_block (picture); + mpeg2_idct_add (picture->DCTblock, dest, stride); +} + +#define MOTION(table,ref,motion_x,motion_y,size,y) \ + pos_x = 2 * picture->offset + motion_x; \ + pos_y = 2 * picture->v_offset + motion_y + 2 * y; \ + if (pos_x > picture->limit_x) { \ + pos_x = ((int)pos_x < 0) ? 0 : picture->limit_x; \ + motion_x = pos_x - 2 * picture->offset; \ + } \ + if (pos_y > picture->limit_y_ ## size){ \ + pos_y = ((int)pos_y < 0) ? 0 : picture->limit_y_ ## size; \ + motion_y = pos_y - 2 * picture->v_offset - 2 * y; \ + } \ + xy_half = ((pos_y & 1) << 1) | (pos_x & 1); \ + table[xy_half] (picture->dest[0] + y * picture->pitches[0] + \ + picture->offset, ref[0] + (pos_x >> 1) + \ + (pos_y >> 1) * picture->pitches[0], picture->pitches[0], \ + size); \ + motion_x /= 2; motion_y /= 2; \ + xy_half = ((motion_y & 1) << 1) | (motion_x & 1); \ + table[4+xy_half] (picture->dest[1] + y/2 * picture->pitches[1] + \ + (picture->offset >> 1), ref[1] + \ + (((picture->offset + motion_x) >> 1) + \ + ((((picture->v_offset + motion_y) >> 1) + y/2) * \ + picture->pitches[1])), picture->pitches[1], size/2); \ + table[4+xy_half] (picture->dest[2] + y/2 * picture->pitches[2] + \ + (picture->offset >> 1), ref[2] + \ + (((picture->offset + motion_x) >> 1) + \ + ((((picture->v_offset + motion_y) >> 1) + y/2) * \ + picture->pitches[2])), picture->pitches[2], size/2) \ + +#define MOTION_FIELD(table,ref,motion_x,motion_y,dest_field,op,src_field) \ + pos_x = 2 * picture->offset + motion_x; \ + pos_y = picture->v_offset + motion_y; \ + if (pos_x > picture->limit_x) { \ + pos_x = ((int)pos_x < 0) ? 0 : picture->limit_x; \ + motion_x = pos_x - 2 * picture->offset; \ + } \ + if (pos_y > picture->limit_y){ \ + pos_y = ((int)pos_y < 0) ? 0 : picture->limit_y; \ + motion_y = pos_y - picture->v_offset; \ + } \ + xy_half = ((pos_y & 1) << 1) | (pos_x & 1); \ + table[xy_half] (picture->dest[0] + dest_field * picture->pitches[0] + \ + picture->offset, \ + (ref[0] + (pos_x >> 1) + \ + ((pos_y op) + src_field) * picture->pitches[0]), \ + 2 * picture->pitches[0], 8); \ + motion_x /= 2; motion_y /= 2; \ + xy_half = ((motion_y & 1) << 1) | (motion_x & 1); \ + table[4+xy_half] (picture->dest[1] + dest_field * picture->pitches[1] + \ + (picture->offset >> 1), ref[1] + \ + (((picture->offset + motion_x) >> 1) + \ + (((picture->v_offset >> 1) + \ + (motion_y op) + src_field) * picture->pitches[1])), \ + 2 * picture->pitches[1], 4); \ + table[4+xy_half] (picture->dest[2] + dest_field * picture->pitches[2] + \ + (picture->offset >> 1), ref[2] + \ + (((picture->offset + motion_x) >> 1) + \ + (((picture->v_offset >> 1) + \ + (motion_y op) + src_field) * picture->pitches[2])), \ + 2 * picture->pitches[2], 4) + +static void motion_mp1 (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + unsigned int pos_x, pos_y, xy_half; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = (motion->pmv[0][0] + + (get_motion_delta (picture, + motion->f_code[0]) << motion->f_code[1])); + motion_x = bound_motion_vector (motion_x, + motion->f_code[0] + motion->f_code[1]); + motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = (motion->pmv[0][1] + + (get_motion_delta (picture, + motion->f_code[0]) << motion->f_code[1])); + motion_y = bound_motion_vector (motion_y, + motion->f_code[0] + motion->f_code[1]); + motion->pmv[0][1] = motion_y; + + MOTION (table, motion->ref[0], motion_x, motion_y, 16, 0); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fr_frame (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + unsigned int pos_x, pos_y, xy_half; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = motion->pmv[0][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[0][1] + get_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion->pmv[0][1] = motion_y; + + MOTION (table, motion->ref[0], motion_x, motion_y, 16, 0); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fr_field (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y, field; + unsigned int pos_x, pos_y, xy_half; + + NEEDBITS (bit_buf, bits, bit_ptr); + field = UBITS (bit_buf, 1); + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[0][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = (motion->pmv[0][1] >> 1) + get_motion_delta (picture, + motion->f_code[1]); + /* motion_y = bound_motion_vector (motion_y, motion->f_code[1]); */ + motion->pmv[0][1] = motion_y << 1; + + MOTION_FIELD (table, motion->ref[0], motion_x, motion_y, 0, & ~1, field); + + NEEDBITS (bit_buf, bits, bit_ptr); + field = UBITS (bit_buf, 1); + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[1][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = (motion->pmv[1][1] >> 1) + get_motion_delta (picture, + motion->f_code[1]); + /* motion_y = bound_motion_vector (motion_y, motion->f_code[1]); */ + motion->pmv[1][1] = motion_y << 1; + + MOTION_FIELD (table, motion->ref[0], motion_x, motion_y, 1, & ~1, field); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fr_dmv (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y, dmv_x, dmv_y, m, other_x, other_y; + unsigned int pos_x, pos_y, xy_half, offset; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = motion->pmv[0][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + NEEDBITS (bit_buf, bits, bit_ptr); + dmv_x = get_dmv (picture); + + motion_y = (motion->pmv[0][1] >> 1) + get_motion_delta (picture, + motion->f_code[1]); + /* motion_y = bound_motion_vector (motion_y, motion->f_code[1]); */ + motion->pmv[1][1] = motion->pmv[0][1] = motion_y << 1; + dmv_y = get_dmv (picture); + + m = picture->top_field_first ? 1 : 3; + other_x = ((motion_x * m + (motion_x > 0)) >> 1) + dmv_x; + other_y = ((motion_y * m + (motion_y > 0)) >> 1) + dmv_y - 1; + MOTION_FIELD (mpeg2_mc.put, motion->ref[0], other_x, other_y, 0, | 1, 0); + + m = picture->top_field_first ? 3 : 1; + other_x = ((motion_x * m + (motion_x > 0)) >> 1) + dmv_x; + other_y = ((motion_y * m + (motion_y > 0)) >> 1) + dmv_y + 1; + MOTION_FIELD (mpeg2_mc.put, motion->ref[0], other_x, other_y, 1, & ~1, 0); + + pos_x = 2 * picture->offset + motion_x; + pos_y = picture->v_offset + motion_y; + if(pos_x > picture->limit_x){ + pos_x = ((int)pos_x < 0) ? 0 : picture->limit_x; + motion_x = pos_x - 2 * picture->offset; + } + if(pos_y > picture->limit_y){ + pos_y = ((int)pos_y < 0) ? 0 : picture->limit_y; + motion_y = pos_y - picture->v_offset; + } + xy_half = ((pos_y & 1) << 1) | (pos_x & 1); + offset = (pos_x >> 1) + (pos_y & ~1) * picture->pitches[0]; + mpeg2_mc.avg[xy_half] + (picture->dest[0] + picture->offset, + motion->ref[0][0] + offset, 2 * picture->pitches[0], 8); + mpeg2_mc.avg[xy_half] + (picture->dest[0] + picture->pitches[0] + picture->offset, + motion->ref[0][0] + picture->pitches[0] + offset, + 2 * picture->pitches[0], 8); + motion_x /= 2; motion_y /= 2; + xy_half = ((motion_y & 1) << 1) | (motion_x & 1); + offset = (((picture->offset + motion_x) >> 1) + + (((picture->v_offset >> 1) + (motion_y & ~1)) * + picture->pitches[1])); + mpeg2_mc.avg[4+xy_half] + (picture->dest[1] + (picture->offset >> 1), + motion->ref[0][1] + offset, 2 * picture->pitches[1], 4); + mpeg2_mc.avg[4+xy_half] + (picture->dest[1] + picture->pitches[1] + (picture->offset >> 1), + motion->ref[0][1] + picture->pitches[1] + offset, + 2 * picture->pitches[1], 4); + offset = (((picture->offset + motion_x) >> 1) + + (((picture->v_offset >> 1) + (motion_y & ~1)) * + picture->pitches[2])); + mpeg2_mc.avg[4+xy_half] + (picture->dest[2] + (picture->offset >> 1), + motion->ref[0][2] + offset, 2 * picture->pitches[2], 4); + mpeg2_mc.avg[4+xy_half] + (picture->dest[2] + picture->pitches[2] + (picture->offset >> 1), + motion->ref[0][2] + picture->pitches[2] + offset, + 2 * picture->pitches[2], 4); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_reuse (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ + int motion_x, motion_y; + unsigned int pos_x, pos_y, xy_half; + + motion_x = motion->pmv[0][0]; + motion_y = motion->pmv[0][1]; + + MOTION (table, motion->ref[0], motion_x, motion_y, 16, 0); +} + +static void motion_zero (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ + table[0] (picture->dest[0] + picture->offset, + (motion->ref[0][0] + picture->offset + + picture->v_offset * picture->pitches[0]), + picture->pitches[0], 16); + + table[4] (picture->dest[1] + (picture->offset >> 1), + motion->ref[0][1] + (picture->offset >> 1) + + (picture->v_offset >> 1) * picture->pitches[1], + picture->pitches[1], 8); + table[4] (picture->dest[2] + (picture->offset >> 1), + motion->ref[0][2] + (picture->offset >> 1) + + (picture->v_offset >> 1) * picture->pitches[2], + picture->pitches[2], 8); +} + +/* like motion_frame, but parsing without actual motion compensation */ +static void motion_fr_conceal (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + tmp = (picture->f_motion.pmv[0][0] + + get_motion_delta (picture, picture->f_motion.f_code[0])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[0]); + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[0][0] = tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + tmp = (picture->f_motion.pmv[0][1] + + get_motion_delta (picture, picture->f_motion.f_code[1])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[1]); + picture->f_motion.pmv[1][1] = picture->f_motion.pmv[0][1] = tmp; + + DUMPBITS (bit_buf, bits, 1); /* remove marker_bit */ +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fi_field (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + uint8_t ** ref_field; + unsigned int pos_x, pos_y, xy_half; + + NEEDBITS (bit_buf, bits, bit_ptr); + ref_field = motion->ref2[UBITS (bit_buf, 1)]; + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[0][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[0][1] + get_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion->pmv[0][1] = motion_y; + + MOTION (table, ref_field, motion_x, motion_y, 16, 0); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fi_16x8 (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + uint8_t ** ref_field; + unsigned int pos_x, pos_y, xy_half; + + NEEDBITS (bit_buf, bits, bit_ptr); + ref_field = motion->ref2[UBITS (bit_buf, 1)]; + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[0][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[0][1] + get_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[0][1] = motion_y; + + MOTION (table, ref_field, motion_x, motion_y, 8, 0); + + NEEDBITS (bit_buf, bits, bit_ptr); + ref_field = motion->ref2[UBITS (bit_buf, 1)]; + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[1][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[1][1] + get_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion_y; + + MOTION (table, ref_field, motion_x, motion_y, 8, 8); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fi_dmv (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y, other_x, other_y; + unsigned int pos_x, pos_y, xy_half; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = motion->pmv[0][0] + get_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + NEEDBITS (bit_buf, bits, bit_ptr); + other_x = ((motion_x + (motion_x > 0)) >> 1) + get_dmv (picture); + + motion_y = motion->pmv[0][1] + get_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion->pmv[0][1] = motion_y; + other_y = (((motion_y + (motion_y > 0)) >> 1) + get_dmv (picture) + + picture->dmv_offset); + + MOTION (mpeg2_mc.put, motion->ref[0], motion_x, motion_y, 16, 0); + MOTION (mpeg2_mc.avg, motion->ref[1], other_x, other_y, 16, 0); +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fi_conceal (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + DUMPBITS (bit_buf, bits, 1); /* remove field_select */ + + tmp = (picture->f_motion.pmv[0][0] + + get_motion_delta (picture, picture->f_motion.f_code[0])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[0]); + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[0][0] = tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + tmp = (picture->f_motion.pmv[0][1] + + get_motion_delta (picture, picture->f_motion.f_code[1])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[1]); + picture->f_motion.pmv[1][1] = picture->f_motion.pmv[0][1] = tmp; + + DUMPBITS (bit_buf, bits, 1); /* remove marker_bit */ +#undef bit_buf +#undef bits +#undef bit_ptr +} + +#define MOTION_CALL(routine,direction) \ +do { \ + if ((direction) & MACROBLOCK_MOTION_FORWARD) \ + routine (picture, &(picture->f_motion), mpeg2_mc.put); \ + if ((direction) & MACROBLOCK_MOTION_BACKWARD) \ + routine (picture, &(picture->b_motion), \ + ((direction) & MACROBLOCK_MOTION_FORWARD ? \ + mpeg2_mc.avg : mpeg2_mc.put)); \ +} while (0) + +#define NEXT_MACROBLOCK \ +do { \ + picture->offset += 16; \ + if (picture->offset == picture->coded_picture_width) { \ + do { /* just so we can use the break statement */ \ + if (picture->current_frame->proc_slice) { \ + picture->current_frame->proc_slice (picture->current_frame, \ + picture->dest); \ + } \ + picture->dest[0] += 16 * picture->pitches[0]; \ + picture->dest[1] += 8 * picture->pitches[1]; \ + picture->dest[2] += 8 * picture->pitches[2]; \ + } while (0); \ + picture->v_offset += 16; \ + if (picture->v_offset > picture->limit_y) { \ + if (mpeg2_cpu_state_restore) \ + mpeg2_cpu_state_restore (&cpu_state); \ + return; \ + } \ + picture->offset = 0; \ + } \ +} while (0) + +static inline int slice_init (picture_t * picture, int code) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int offset, height; + struct vo_frame_s * forward_reference_frame; + struct vo_frame_s * backward_reference_frame; + const MBAtab * mba; + + offset = picture->picture_structure == BOTTOM_FIELD; + picture->pitches[0] = picture->current_frame->pitches[0]; + picture->pitches[1] = picture->current_frame->pitches[1]; + picture->pitches[2] = picture->current_frame->pitches[2]; + + if( picture->forward_reference_frame ) { + forward_reference_frame = picture->forward_reference_frame; + } + else { + /* return 1; */ + forward_reference_frame = picture->current_frame; + } + + if( picture->backward_reference_frame ) { + backward_reference_frame = picture->backward_reference_frame; + } + else { + /* return 1; */ + backward_reference_frame = picture->current_frame; + } + + picture->f_motion.ref[0][0] = + forward_reference_frame->base[0] + (offset ? picture->pitches[0] : 0); + picture->f_motion.ref[0][1] = + forward_reference_frame->base[1] + (offset ? picture->pitches[1] : 0); + picture->f_motion.ref[0][2] = + forward_reference_frame->base[2] + (offset ? picture->pitches[2] : 0); + + picture->b_motion.ref[0][0] = + backward_reference_frame->base[0] + (offset ? picture->pitches[0] : 0); + picture->b_motion.ref[0][1] = + backward_reference_frame->base[1] + (offset ? picture->pitches[1] : 0); + picture->b_motion.ref[0][2] = + backward_reference_frame->base[2] + (offset ? picture->pitches[2] : 0); + + if (picture->picture_structure != FRAME_PICTURE) { + uint8_t ** forward_ref; + int bottom_field; + + bottom_field = (picture->picture_structure == BOTTOM_FIELD); + picture->dmv_offset = bottom_field ? 1 : -1; + picture->f_motion.ref2[0] = picture->f_motion.ref[bottom_field]; + picture->f_motion.ref2[1] = picture->f_motion.ref[!bottom_field]; + picture->b_motion.ref2[0] = picture->b_motion.ref[bottom_field]; + picture->b_motion.ref2[1] = picture->b_motion.ref[!bottom_field]; + + forward_ref = forward_reference_frame->base; + if (picture->second_field && (picture->picture_coding_type != B_TYPE)) + forward_ref = picture->current_frame->base; + + picture->f_motion.ref[1][0] = forward_ref[0] + (bottom_field ? 0 : picture->pitches[0]); + picture->f_motion.ref[1][1] = forward_ref[1] + (bottom_field ? 0 : picture->pitches[1]); + picture->f_motion.ref[1][2] = forward_ref[2] + (bottom_field ? 0 : picture->pitches[2]); + + picture->b_motion.ref[1][0] = + backward_reference_frame->base[0] + (bottom_field ? 0 : picture->pitches[0]); + picture->b_motion.ref[1][1] = + backward_reference_frame->base[1] + (bottom_field ? 0 : picture->pitches[1]); + picture->b_motion.ref[1][2] = + backward_reference_frame->base[2] + (bottom_field ? 0 : picture->pitches[2]); + } + + picture->f_motion.pmv[0][0] = picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[1][1] = 0; + picture->b_motion.pmv[0][0] = picture->b_motion.pmv[0][1] = 0; + picture->b_motion.pmv[1][0] = picture->b_motion.pmv[1][1] = 0; + + picture->v_offset = (code - 1) * 16; + offset = (code - 1); + if (picture->picture_structure != FRAME_PICTURE) + offset = 2 * offset; + + picture->dest[0] = picture->current_frame->base[0] + picture->pitches[0] * offset * 16; + picture->dest[1] = picture->current_frame->base[1] + picture->pitches[1] * offset * 8; + picture->dest[2] = picture->current_frame->base[2] + picture->pitches[2] * offset * 8; + + height = picture->coded_picture_height; + switch (picture->picture_structure) { + case BOTTOM_FIELD: + picture->dest[0] += picture->pitches[0]; + picture->dest[1] += picture->pitches[1]; + picture->dest[2] += picture->pitches[2]; + /* follow thru */ + case TOP_FIELD: + picture->pitches[0] <<= 1; + picture->pitches[1] <<= 1; + picture->pitches[2] <<= 1; + height >>= 1; + } + picture->limit_x = 2 * picture->coded_picture_width - 32; + picture->limit_y_16 = 2 * height - 32; + picture->limit_y_8 = 2 * height - 16; + picture->limit_y = height - 16; + + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 1 << (picture->intra_dc_precision + 7); + + picture->quantizer_scale = get_quantizer_scale (picture); + + /* ignore intra_slice and all the extra data */ + while (bit_buf & 0x80000000) { + DUMPBITS (bit_buf, bits, 9); + NEEDBITS (bit_buf, bits, bit_ptr); + } + + /* decode initial macroblock address increment */ + offset = 0; + while (1) { + if (bit_buf >= 0x08000000) { + mba = MBA_5 + (UBITS (bit_buf, 6) - 2); + break; + } else if (bit_buf >= 0x01800000) { + mba = MBA_11 + (UBITS (bit_buf, 12) - 24); + break; + } else switch (UBITS (bit_buf, 12)) { + case 8: /* macroblock_escape */ + offset += 33; + DUMPBITS (bit_buf, bits, 11); + NEEDBITS (bit_buf, bits, bit_ptr); + continue; + case 15: /* macroblock_stuffing (MPEG1 only) */ + bit_buf &= 0xfffff; + DUMPBITS (bit_buf, bits, 11); + NEEDBITS (bit_buf, bits, bit_ptr); + continue; + default: /* error */ + return 1; + } + } + DUMPBITS (bit_buf, bits, mba->len + 1); + picture->offset = (offset + mba->mba) << 4; + + while (picture->offset - picture->coded_picture_width >= 0) { + picture->offset -= picture->coded_picture_width; + if ((picture->current_frame->proc_slice == NULL) || + (picture->picture_coding_type != B_TYPE)) { + picture->dest[0] += 16 * picture->pitches[0]; + picture->dest[1] += 8 * picture->pitches[1]; + picture->dest[2] += 8 * picture->pitches[2]; + } + picture->v_offset += 16; + } + if (picture->v_offset > picture->limit_y) + return 1; + + return 0; +#undef bit_buf +#undef bits +#undef bit_ptr +} + +void mpeg2_slice (picture_t * picture, int code, uint8_t * buffer) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + cpu_state_t cpu_state; + + bitstream_init (picture, buffer); + + if (slice_init (picture, code)) + return; + + if (mpeg2_cpu_state_save) + mpeg2_cpu_state_save (&cpu_state); + + while (1) { + int macroblock_modes; + int mba_inc; + const MBAtab * mba; + + NEEDBITS (bit_buf, bits, bit_ptr); + + macroblock_modes = get_macroblock_modes (picture); + + /* maybe integrate MACROBLOCK_QUANT test into get_macroblock_modes ? */ + if (macroblock_modes & MACROBLOCK_QUANT) + picture->quantizer_scale = get_quantizer_scale (picture); + + if (macroblock_modes & MACROBLOCK_INTRA) { + + int DCT_offset, DCT_stride; + int offset; + uint8_t * dest_y; + + if (picture->concealment_motion_vectors) { + if (picture->picture_structure == FRAME_PICTURE) + motion_fr_conceal (picture); + else + motion_fi_conceal (picture); + } else { + picture->f_motion.pmv[0][0] = picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[1][1] = 0; + picture->b_motion.pmv[0][0] = picture->b_motion.pmv[0][1] = 0; + picture->b_motion.pmv[1][0] = picture->b_motion.pmv[1][1] = 0; + } + + if (macroblock_modes & DCT_TYPE_INTERLACED) { + DCT_offset = picture->pitches[0]; + DCT_stride = picture->pitches[0] * 2; + } else { + DCT_offset = picture->pitches[0] * 8; + DCT_stride = picture->pitches[0]; + } + + offset = picture->offset; + dest_y = picture->dest[0] + offset; + slice_intra_DCT (picture, 0, dest_y, DCT_stride); + slice_intra_DCT (picture, 0, dest_y + 8, DCT_stride); + slice_intra_DCT (picture, 0, dest_y + DCT_offset, DCT_stride); + slice_intra_DCT (picture, 0, dest_y + DCT_offset + 8, DCT_stride); + slice_intra_DCT (picture, 1, picture->dest[1] + (offset >> 1), + picture->pitches[1]); + slice_intra_DCT (picture, 2, picture->dest[2] + (offset >> 1), + picture->pitches[2]); + + if (picture->picture_coding_type == D_TYPE) { + NEEDBITS (bit_buf, bits, bit_ptr); + DUMPBITS (bit_buf, bits, 1); + } + } else { + + if (picture->picture_structure == FRAME_PICTURE) + switch (macroblock_modes & MOTION_TYPE_MASK) { + case MC_FRAME: + if (picture->mpeg1) + MOTION_CALL (motion_mp1, macroblock_modes); + else + MOTION_CALL (motion_fr_frame, macroblock_modes); + break; + + case MC_FIELD: + MOTION_CALL (motion_fr_field, macroblock_modes); + break; + + case MC_DMV: + MOTION_CALL (motion_fr_dmv, MACROBLOCK_MOTION_FORWARD); + break; + + case 0: + /* non-intra mb without forward mv in a P picture */ + picture->f_motion.pmv[0][0] = 0; + picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = 0; + picture->f_motion.pmv[1][1] = 0; + MOTION_CALL (motion_zero, MACROBLOCK_MOTION_FORWARD); + break; + } + else + switch (macroblock_modes & MOTION_TYPE_MASK) { + case MC_FIELD: + MOTION_CALL (motion_fi_field, macroblock_modes); + break; + + case MC_16X8: + MOTION_CALL (motion_fi_16x8, macroblock_modes); + break; + + case MC_DMV: + MOTION_CALL (motion_fi_dmv, MACROBLOCK_MOTION_FORWARD); + break; + + case 0: + /* non-intra mb without forward mv in a P picture */ + picture->f_motion.pmv[0][0] = 0; + picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = 0; + picture->f_motion.pmv[1][1] = 0; + MOTION_CALL (motion_zero, MACROBLOCK_MOTION_FORWARD); + break; + } + + if (macroblock_modes & MACROBLOCK_PATTERN) { + int coded_block_pattern; + int DCT_offset, DCT_stride; + int offset; + uint8_t * dest_y; + + if (macroblock_modes & DCT_TYPE_INTERLACED) { + DCT_offset = picture->pitches[0]; + DCT_stride = picture->pitches[0] * 2; + } else { + DCT_offset = picture->pitches[0] * 8; + DCT_stride = picture->pitches[0]; + } + + coded_block_pattern = get_coded_block_pattern (picture); + + offset = picture->offset; + dest_y = picture->dest[0] + offset; + if (coded_block_pattern & 0x20) + slice_non_intra_DCT (picture, dest_y, DCT_stride); + if (coded_block_pattern & 0x10) + slice_non_intra_DCT (picture, dest_y + 8, DCT_stride); + if (coded_block_pattern & 0x08) + slice_non_intra_DCT (picture, dest_y + DCT_offset, + DCT_stride); + if (coded_block_pattern & 0x04) + slice_non_intra_DCT (picture, dest_y + DCT_offset + 8, + DCT_stride); + if (coded_block_pattern & 0x2) + slice_non_intra_DCT (picture, + picture->dest[1] + (offset >> 1), + picture->pitches[1]); + if (coded_block_pattern & 0x1) + slice_non_intra_DCT (picture, + picture->dest[2] + (offset >> 1), + picture->pitches[2]); + } + + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 128 << picture->intra_dc_precision; + } + + NEXT_MACROBLOCK; + + NEEDBITS (bit_buf, bits, bit_ptr); + mba_inc = 0; + while (1) { + if (bit_buf >= 0x10000000) { + mba = MBA_5 + (UBITS (bit_buf, 5) - 2); + break; + } else if (bit_buf >= 0x03000000) { + mba = MBA_11 + (UBITS (bit_buf, 11) - 24); + break; + } else switch (UBITS (bit_buf, 11)) { + case 8: /* macroblock_escape */ + mba_inc += 33; + /* pass through */ + case 15: /* macroblock_stuffing (MPEG1 only) */ + DUMPBITS (bit_buf, bits, 11); + NEEDBITS (bit_buf, bits, bit_ptr); + continue; + default: /* end of slice, or error */ + if (mpeg2_cpu_state_restore) + mpeg2_cpu_state_restore (&cpu_state); + return; + } + } + DUMPBITS (bit_buf, bits, mba->len); + mba_inc += mba->mba; + + if (mba_inc) { + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 128 << picture->intra_dc_precision; + + if (picture->picture_coding_type == P_TYPE) { + picture->f_motion.pmv[0][0] = picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[1][1] = 0; + + do { + MOTION_CALL (motion_zero, MACROBLOCK_MOTION_FORWARD); + NEXT_MACROBLOCK; + } while (--mba_inc); + } else { + do { + MOTION_CALL (motion_reuse, macroblock_modes); + NEXT_MACROBLOCK; + } while (--mba_inc); + } + } + } +#undef bit_buf +#undef bits +#undef bit_ptr +} diff --git a/src/video_dec/libmpeg2/slice_xvmc.c b/src/video_dec/libmpeg2/slice_xvmc.c new file mode 100644 index 000000000..014ae7924 --- /dev/null +++ b/src/video_dec/libmpeg2/slice_xvmc.c @@ -0,0 +1,1988 @@ +/* + * slice_xvmc.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#include <stdio.h> +#include <string.h> /* memcpy/memset, try to remove */ +#include <stdlib.h> +#include <inttypes.h> + +#include <xine/xine_internal.h> +#include <xine/video_out.h> +#include "mpeg2_internal.h" +#include <xine/xineutils.h> + +#include <xine/attributes.h> +#include "accel_xvmc.h" +#include "xvmc.h" + + +#define MOTION_ACCEL XINE_VO_MOTION_ACCEL +#define IDCT_ACCEL XINE_VO_IDCT_ACCEL +#define SIGNED_INTRA XINE_VO_SIGNED_INTRA +#define ACCEL (MOTION_ACCEL | IDCT_ACCEL) + +#include "vlc.h" +/* original (non-patched) scan tables */ + +static const uint8_t mpeg2_scan_norm_orig[64] ATTR_ALIGN(16) = +{ + /* Zig-Zag scan pattern */ + 0, 1, 8,16, 9, 2, 3,10, + 17,24,32,25,18,11, 4, 5, + 12,19,26,33,40,48,41,34, + 27,20,13, 6, 7,14,21,28, + 35,42,49,56,57,50,43,36, + 29,22,15,23,30,37,44,51, + 58,59,52,45,38,31,39,46, + 53,60,61,54,47,55,62,63 +}; + +static const uint8_t mpeg2_scan_alt_orig[64] ATTR_ALIGN(16) = +{ + /* Alternate scan pattern */ + 0,8,16,24,1,9,2,10,17,25,32,40,48,56,57,49, + 41,33,26,18,3,11,4,12,19,27,34,42,50,58,35,43, + 51,59,20,28,5,13,6,14,21,29,36,44,52,60,37,45, + 53,61,22,30,7,15,23,31,38,46,54,62,39,47,55,63 +}; + +static uint8_t mpeg2_scan_alt_ptable[64] ATTR_ALIGN(16); +static uint8_t mpeg2_scan_norm_ptable[64] ATTR_ALIGN(16); +static uint8_t mpeg2_scan_orig_ptable[64] ATTR_ALIGN(16); + +void xvmc_setup_scan_ptable( void ) +{ + int i; + for (i=0; i<64; ++i) { + mpeg2_scan_norm_ptable[mpeg2_scan_norm_orig[i]] = mpeg2_scan_norm[i]; + mpeg2_scan_alt_ptable[mpeg2_scan_alt_orig[i]] = mpeg2_scan_alt[i]; + mpeg2_scan_orig_ptable[i] = i; + } +} + + +static const int non_linear_quantizer_scale [] = { + 0, 1, 2, 3, 4, 5, 6, 7, + 8, 10, 12, 14, 16, 18, 20, 22, + 24, 28, 32, 36, 40, 44, 48, 52, + 56, 64, 72, 80, 88, 96, 104, 112 +}; + +static inline int get_xvmc_macroblock_modes (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int macroblock_modes; + const MBtab * tab; + + switch (picture->picture_coding_type) { + case I_TYPE: + + tab = MB_I + UBITS (bit_buf, 1); + DUMPBITS (bit_buf, bits, tab->len); + macroblock_modes = tab->modes; + + if ((! (picture->frame_pred_frame_dct)) && + (picture->picture_structure == FRAME_PICTURE)) { + macroblock_modes |= UBITS (bit_buf, 1) * DCT_TYPE_INTERLACED; + DUMPBITS (bit_buf, bits, 1); + } + + return macroblock_modes; + + case P_TYPE: + + tab = MB_P + UBITS (bit_buf, 5); + DUMPBITS (bit_buf, bits, tab->len); + macroblock_modes = tab->modes; + + if (picture->picture_structure != FRAME_PICTURE) { + if (macroblock_modes & MACROBLOCK_MOTION_FORWARD) { + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + } + return macroblock_modes; + } else if (picture->frame_pred_frame_dct) { + if (macroblock_modes & MACROBLOCK_MOTION_FORWARD) + macroblock_modes |= MC_FRAME; + return macroblock_modes; + } else { + if (macroblock_modes & MACROBLOCK_MOTION_FORWARD) { + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + } + if (macroblock_modes & (MACROBLOCK_INTRA | MACROBLOCK_PATTERN)) { + macroblock_modes |= UBITS (bit_buf, 1) * DCT_TYPE_INTERLACED; + DUMPBITS (bit_buf, bits, 1); + } + return macroblock_modes; + } + + case B_TYPE: + + tab = MB_B + UBITS (bit_buf, 6); + DUMPBITS (bit_buf, bits, tab->len); + macroblock_modes = tab->modes; + + if (picture->picture_structure != FRAME_PICTURE) { + if (! (macroblock_modes & MACROBLOCK_INTRA)) { + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + } + return macroblock_modes; + } else if (picture->frame_pred_frame_dct) { + /* if (! (macroblock_modes & MACROBLOCK_INTRA)) */ + macroblock_modes |= MC_FRAME; + return macroblock_modes; + } else { + if (macroblock_modes & MACROBLOCK_INTRA) + goto intra; + macroblock_modes |= UBITS (bit_buf, 2) * MOTION_TYPE_BASE; + DUMPBITS (bit_buf, bits, 2); + if (macroblock_modes & (MACROBLOCK_INTRA | MACROBLOCK_PATTERN)) { + intra: + macroblock_modes |= UBITS (bit_buf, 1) * DCT_TYPE_INTERLACED; + DUMPBITS (bit_buf, bits, 1); + } + return macroblock_modes; + } + + case D_TYPE: + + DUMPBITS (bit_buf, bits, 1); + return MACROBLOCK_INTRA; + + default: + return 0; + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_xvmc_quantizer_scale (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + int quantizer_scale_code; + + quantizer_scale_code = UBITS (bit_buf, 5); + DUMPBITS (bit_buf, bits, 5); + + if (picture->q_scale_type) + return non_linear_quantizer_scale [quantizer_scale_code]; + else + return quantizer_scale_code << 1; +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_xvmc_motion_delta (picture_t * picture, int f_code) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + int delta; + int sign; + const MVtab * tab; + + if (bit_buf & 0x80000000) { + DUMPBITS (bit_buf, bits, 1); + return 0; + } else if (bit_buf >= 0x0c000000) { + + tab = MV_4 + UBITS (bit_buf, 4); + delta = (tab->delta << f_code) + 1; + bits += tab->len + f_code + 1; + bit_buf <<= tab->len; + + sign = SBITS (bit_buf, 1); + bit_buf <<= 1; + + if (f_code) + delta += UBITS (bit_buf, f_code); + bit_buf <<= f_code; + + return (delta ^ sign) - sign; + + } else { + + tab = MV_10 + UBITS (bit_buf, 10); + delta = (tab->delta << f_code) + 1; + bits += tab->len + 1; + bit_buf <<= tab->len; + + sign = SBITS (bit_buf, 1); + bit_buf <<= 1; + + if (f_code) { + NEEDBITS (bit_buf, bits, bit_ptr); + delta += UBITS (bit_buf, f_code); + DUMPBITS (bit_buf, bits, f_code); + } + + return (delta ^ sign) - sign; + + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int bound_motion_vector (int vec, int f_code) +{ +#if 1 + unsigned int limit; + int sign; + + limit = 16 << f_code; + + if ((unsigned int)(vec + limit) < 2 * limit) + return vec; + else { + sign = ((int32_t)vec) >> 31; + return vec - ((2 * limit) ^ sign) + sign; + } +#else + return ((int32_t)vec << (27 - f_code)) >> (27 - f_code); +#endif +} + +static inline int get_xvmc_dmv (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + const DMVtab * tab; + + tab = DMV_2 + UBITS (bit_buf, 2); + DUMPBITS (bit_buf, bits, tab->len); + return tab->dmv; +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_xvmc_coded_block_pattern (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + + const CBPtab * tab; + + NEEDBITS (bit_buf, bits, bit_ptr); + + if (bit_buf >= 0x20000000) { + + tab = CBP_7 + (UBITS (bit_buf, 7) - 16); + DUMPBITS (bit_buf, bits, tab->len); + return tab->cbp; + + } else { + + tab = CBP_9 + UBITS (bit_buf, 9); + DUMPBITS (bit_buf, bits, tab->len); + return tab->cbp; + } + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_xvmc_luma_dc_dct_diff (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + const DCtab * tab; + int size; + int dc_diff; + + if (bit_buf < 0xf8000000) { + tab = DC_lum_5 + UBITS (bit_buf, 5); + size = tab->size; + if (size) { + bits += tab->len + size; + bit_buf <<= tab->len; + dc_diff = + UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + bit_buf <<= size; + return dc_diff; + } else { + DUMPBITS (bit_buf, bits, 3); + return 0; + } + } else { + tab = DC_long + (UBITS (bit_buf, 9) - 0x1e0); + size = tab->size; + DUMPBITS (bit_buf, bits, tab->len); + NEEDBITS (bit_buf, bits, bit_ptr); + dc_diff = UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + DUMPBITS (bit_buf, bits, size); + return dc_diff; + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline int get_xvmc_chroma_dc_dct_diff (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + const DCtab * tab; + int size; + int dc_diff; + + if (bit_buf < 0xf8000000) { + tab = DC_chrom_5 + UBITS (bit_buf, 5); + size = tab->size; + if (size) { + bits += tab->len + size; + bit_buf <<= tab->len; + dc_diff = + UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + bit_buf <<= size; + return dc_diff; + } else { + DUMPBITS (bit_buf, bits, 2); + return 0; + } + } else { + tab = DC_long + (UBITS (bit_buf, 10) - 0x3e0); + size = tab->size; + DUMPBITS (bit_buf, bits, tab->len + 1); + NEEDBITS (bit_buf, bits, bit_ptr); + dc_diff = UBITS (bit_buf, size) - UBITS (SBITS (~bit_buf, 1), size); + DUMPBITS (bit_buf, bits, size); + return dc_diff; + } +#undef bit_buf +#undef bits +#undef bit_ptr +} + +#define SATURATE(val) \ +do { \ + if ((uint32_t)(val + 2048) > 4095) \ + val = (val > 0) ? 2047 : -2048; \ +} while (0) + +static void get_xvmc_intra_block_B14 (picture_t * picture) +{ + int i; + int j; + int l; + int val; + const uint8_t * scan = picture->scan; + uint8_t * scan_ptable = mpeg2_scan_orig_ptable; + uint8_t * quant_matrix = picture->intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + int mismatch; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + dest = picture->mc->blockptr; + + if( picture->mc->xvmc_accel & IDCT_ACCEL ) { + if ( scan == mpeg2_scan_norm ) { + scan = mpeg2_scan_norm_orig; + scan_ptable = mpeg2_scan_norm_ptable; + } else { + scan = mpeg2_scan_alt_orig; + scan_ptable = mpeg2_scan_alt_ptable; + } + } + + i = 0; + mismatch = ~dest[0]; + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + l = scan_ptable[j = scan[i]]; + + bit_buf <<= tab->len; + bits += tab->len + 1; + val = (tab->level * quantizer_scale * quant_matrix[l]) >> 4; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + l = scan_ptable[j = scan[i]]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = (SBITS (bit_buf, 12) * + quantizer_scale * quant_matrix[l]) / 16; + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + + dest[63] ^= mismatch & 1; + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_xvmc_intra_block_B15 (picture_t * picture) +{ + int i; + int j; + int l; + int val; + const uint8_t * scan = picture->scan; + uint8_t * scan_ptable = mpeg2_scan_orig_ptable; + uint8_t * quant_matrix = picture->intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + int mismatch; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + dest = picture->mc->blockptr; + + if( picture->mc->xvmc_accel & IDCT_ACCEL ) { + if ( scan == mpeg2_scan_norm ) { + scan = mpeg2_scan_norm_orig; + scan_ptable = mpeg2_scan_norm_ptable; + } else { + scan = mpeg2_scan_alt_orig; + scan_ptable = mpeg2_scan_alt_ptable; + } + } + + i = 0; + mismatch = ~dest[0]; + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + + while (1) { + if (bit_buf >= 0x04000000) { + + tab = DCT_B15_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) { + + normal_code: + l = scan_ptable[j = scan[i]]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = (tab->level * quantizer_scale * quant_matrix[l]) >> 4; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else { + + /* end of block. I commented out this code because if we */ + /* dont exit here we will still exit at the later test :) */ + + /* if (i >= 128) break; */ /* end of block */ + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check against buffer overflow */ + + l = scan_ptable[j = scan[i]]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = (SBITS (bit_buf, 12) * + quantizer_scale * quant_matrix[l]) / 16; + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } + } else if (bit_buf >= 0x02000000) { + tab = DCT_B15_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + + dest[63] ^= mismatch & 1; + DUMPBITS (bit_buf, bits, 4); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_xvmc_non_intra_block (picture_t * picture) +{ + int i; + int j; + int l; + int val; + const uint8_t * scan = picture->scan; + uint8_t * scan_ptable = mpeg2_scan_orig_ptable; + uint8_t * quant_matrix = picture->non_intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + int mismatch; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + i = -1; + mismatch = 1; + + dest = picture->mc->blockptr; + + if( picture->mc->xvmc_accel & IDCT_ACCEL ) { + if ( scan == mpeg2_scan_norm ) { + scan = mpeg2_scan_norm_orig; + scan_ptable = mpeg2_scan_norm_ptable; + } else { + scan = mpeg2_scan_alt_orig; + scan_ptable = mpeg2_scan_alt_ptable; + } + } + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + if (bit_buf >= 0x28000000) { + tab = DCT_B14DC_5 + (UBITS (bit_buf, 5) - 5); + goto entry_1; + } else + goto entry_2; + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + entry_1: + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + l = scan_ptable[j = scan[i]]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = ((2*tab->level+1) * quantizer_scale * quant_matrix[l]) >> 5; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } + + entry_2: + if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + l = scan_ptable[j = scan[i]]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = 2 * (SBITS (bit_buf, 12) + SBITS (bit_buf, 1)) + 1; + val = (val * quantizer_scale * quant_matrix[l]) / 32; + + SATURATE (val); + dest[j] = val; + mismatch ^= val; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + dest[63] ^= mismatch & 1; + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_xvmc_mpeg1_intra_block (picture_t * picture) +{ + int i; + int j; + int l; + int val; + const uint8_t * scan = picture->scan; + uint8_t * scan_ptable = mpeg2_scan_orig_ptable; + uint8_t * quant_matrix = picture->intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + i = 0; + + dest = picture->mc->blockptr; + + if( picture->mc->xvmc_accel & IDCT_ACCEL ) { + if ( scan == mpeg2_scan_norm ) { + scan = mpeg2_scan_norm_orig; + scan_ptable = mpeg2_scan_norm_ptable; + } else { + scan = mpeg2_scan_alt_orig; + scan_ptable = mpeg2_scan_alt_ptable; + } + } + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + l = scan_ptable[j = scan[i]]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = (tab->level * quantizer_scale * quant_matrix[l]) >> 4; + + /* oddification */ + val = (val - 1) | 1; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + l = scan_ptable[j = scan[i]]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = SBITS (bit_buf, 8); + if (! (val & 0x7f)) { + DUMPBITS (bit_buf, bits, 8); + val = UBITS (bit_buf, 8) + 2 * val; + } + val = (val * quantizer_scale * quant_matrix[l]) / 16; + + /* oddification */ + val = (val + ~SBITS (val, 1)) | 1; + + SATURATE (val); + dest[j] = val; + + DUMPBITS (bit_buf, bits, 8); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static void get_xvmc_mpeg1_non_intra_block (picture_t * picture) +{ + int i; + int j; + int l; + int val; + const uint8_t * scan = picture->scan; + uint8_t * scan_ptable = mpeg2_scan_orig_ptable; + uint8_t * quant_matrix = picture->non_intra_quantizer_matrix; + int quantizer_scale = picture->quantizer_scale; + const DCTtab * tab; + uint32_t bit_buf; + int bits; + uint8_t * bit_ptr; + int16_t * dest; + + i = -1; + + dest = picture->mc->blockptr; + + if( picture->mc->xvmc_accel & IDCT_ACCEL ) { + if ( scan == mpeg2_scan_norm ) { + scan = mpeg2_scan_norm_orig; + scan_ptable = mpeg2_scan_norm_ptable; + } else { + scan = mpeg2_scan_alt_orig; + scan_ptable = mpeg2_scan_alt_ptable; + } + } + + bit_buf = picture->bitstream_buf; + bits = picture->bitstream_bits; + bit_ptr = picture->bitstream_ptr; + + NEEDBITS (bit_buf, bits, bit_ptr); + if (bit_buf >= 0x28000000) { + tab = DCT_B14DC_5 + (UBITS (bit_buf, 5) - 5); + goto entry_1; + } else + goto entry_2; + + while (1) { + if (bit_buf >= 0x28000000) { + + tab = DCT_B14AC_5 + (UBITS (bit_buf, 5) - 5); + + entry_1: + i += tab->run; + if (i >= 64) + break; /* end of block */ + + normal_code: + l = scan_ptable[j = scan[i]]; + bit_buf <<= tab->len; + bits += tab->len + 1; + val = ((2*tab->level+1) * quantizer_scale * quant_matrix[l]) >> 5; + + /* oddification */ + val = (val - 1) | 1; + + /* if (bitstream_get (1)) val = -val; */ + val = (val ^ SBITS (bit_buf, 1)) - SBITS (bit_buf, 1); + + SATURATE (val); + dest[j] = val; + + bit_buf <<= 1; + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } + + entry_2: + if (bit_buf >= 0x04000000) { + + tab = DCT_B14_8 + (UBITS (bit_buf, 8) - 4); + + i += tab->run; + if (i < 64) + goto normal_code; + + /* escape code */ + + i += UBITS (bit_buf << 6, 6) - 64; + if (i >= 64) + break; /* illegal, check needed to avoid buffer overflow */ + + l = scan_ptable[j = scan[i]]; + + DUMPBITS (bit_buf, bits, 12); + NEEDBITS (bit_buf, bits, bit_ptr); + val = SBITS (bit_buf, 8); + if (! (val & 0x7f)) { + DUMPBITS (bit_buf, bits, 8); + val = UBITS (bit_buf, 8) + 2 * val; + } + val = 2 * (val + SBITS (val, 1)) + 1; + val = (val * quantizer_scale * quant_matrix[l]) / 32; + + /* oddification */ + val = (val + ~SBITS (val, 1)) | 1; + + SATURATE (val); + dest[j] = val; + + DUMPBITS (bit_buf, bits, 8); + NEEDBITS (bit_buf, bits, bit_ptr); + + continue; + + } else if (bit_buf >= 0x02000000) { + tab = DCT_B14_10 + (UBITS (bit_buf, 10) - 8); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00800000) { + tab = DCT_13 + (UBITS (bit_buf, 13) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else if (bit_buf >= 0x00200000) { + tab = DCT_15 + (UBITS (bit_buf, 15) - 16); + i += tab->run; + if (i < 64) + goto normal_code; + } else { + tab = DCT_16 + UBITS (bit_buf, 16); + bit_buf <<= 16; + GETWORD (bit_buf, bits + 16, bit_ptr); + i += tab->run; + if (i < 64) + goto normal_code; + } + break; /* illegal, check needed to avoid buffer overflow */ + } + DUMPBITS (bit_buf, bits, 2); /* dump end of block code */ + picture->bitstream_buf = bit_buf; + picture->bitstream_bits = bits; + picture->bitstream_ptr = bit_ptr; +} + +static inline void slice_xvmc_intra_DCT (picture_t * picture, int cc, + uint8_t * dest, int stride) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + NEEDBITS (bit_buf, bits, bit_ptr); + /* Get the intra DC coefficient and inverse quantize it */ + + // printf("slice: slice_xvmc_intra_DCT cc=%d pred[0]=%d\n",cc,picture->dc_dct_pred[0]); + if (cc == 0) + picture->dc_dct_pred[0] += get_xvmc_luma_dc_dct_diff (picture); + else + picture->dc_dct_pred[cc] += get_xvmc_chroma_dc_dct_diff (picture); + //TODO conversion to signed format + // printf("slice: pred[0]=%d presision=%d\n",picture->dc_dct_pred[0], + // picture->intra_dc_precision); + + mpeg2_zero_block(picture->mc->blockptr); + + picture->mc->blockptr[0] = picture->dc_dct_pred[cc] << (3 - picture->intra_dc_precision); + + if (picture->mpeg1) { + if (picture->picture_coding_type != D_TYPE) + get_xvmc_mpeg1_intra_block (picture); + } else if (picture->intra_vlc_format) + get_xvmc_intra_block_B15 (picture); + else + get_xvmc_intra_block_B14 (picture); + + if((picture->mc->xvmc_accel & ACCEL) == MOTION_ACCEL) { + //motion_comp only no idct acceleration so do it in software + mpeg2_idct (picture->mc->blockptr); + } + picture->mc->blockptr += 64; +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static inline void slice_xvmc_non_intra_DCT (picture_t * picture, uint8_t * dest, + int stride) +{ + mpeg2_zero_block(picture->mc->blockptr); + + if (picture->mpeg1) + get_xvmc_mpeg1_non_intra_block (picture); + else + get_xvmc_non_intra_block (picture); + + if((picture->mc->xvmc_accel & ACCEL) == MOTION_ACCEL) { + // motion comp only no idct acceleration so do it in sw + mpeg2_idct (picture->mc->blockptr); + } + picture->mc->blockptr += 64; +} + +static void motion_mp1 (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = (motion->pmv[0][0] + + (get_xvmc_motion_delta (picture, + motion->f_code[0]) << motion->f_code[1])); + motion_x = bound_motion_vector (motion_x, + motion->f_code[0] + motion->f_code[1]); + motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = (motion->pmv[0][1] + + (get_xvmc_motion_delta (picture, + motion->f_code[0]) << motion->f_code[1])); + motion_y = bound_motion_vector (motion_y, + motion->f_code[0] + motion->f_code[1]); + motion->pmv[0][1] = motion_y; + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fr_frame (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = motion->pmv[0][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[0][1] + get_xvmc_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion->pmv[0][1] = motion_y; + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fr_field (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int), + int dir) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y, field; + // unsigned int pos_x, pos_y, xy_half; + + NEEDBITS (bit_buf, bits, bit_ptr); + field = UBITS (bit_buf, 1); + picture->XvMC_mv_field_sel[0][dir] = field; + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[0][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = (motion->pmv[0][1] >> 1) + get_xvmc_motion_delta (picture, + motion->f_code[1]); + /* motion_y = bound_motion_vector (motion_y, motion->f_code[1]); */ + motion->pmv[0][1] = motion_y << 1; + + NEEDBITS (bit_buf, bits, bit_ptr); + field = UBITS (bit_buf, 1); + //TODO look at field select need bob (weave ok) + picture->XvMC_mv_field_sel[1][dir] = field; + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[1][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = (motion->pmv[1][1] >> 1) + get_xvmc_motion_delta (picture, + motion->f_code[1]); + /* motion_y = bound_motion_vector (motion_y, motion->f_code[1]); */ + motion->pmv[1][1] = motion_y << 1; + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fr_dmv (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + + // TODO field select ?? possible need to be 0 + picture->XvMC_mv_field_sel[0][0] = picture->XvMC_mv_field_sel[1][0] = 0; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = motion->pmv[0][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + NEEDBITS (bit_buf, bits, bit_ptr); + + motion_y = (motion->pmv[0][1] >> 1) + get_xvmc_motion_delta (picture, + motion->f_code[1]); + /* motion_y = bound_motion_vector (motion_y, motion->f_code[1]); */ + motion->pmv[1][1] = motion->pmv[0][1] = motion_y << 1; + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_reuse (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ + int motion_x, motion_y; + + motion_x = motion->pmv[0][0]; + motion_y = motion->pmv[0][1]; + +} + +/* like motion_frame, but parsing without actual motion compensation */ +static void motion_fr_conceal (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + tmp = (picture->f_motion.pmv[0][0] + + get_xvmc_motion_delta (picture, picture->f_motion.f_code[0])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[0]); + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[0][0] = tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + tmp = (picture->f_motion.pmv[0][1] + + get_xvmc_motion_delta (picture, picture->f_motion.f_code[1])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[1]); + picture->f_motion.pmv[1][1] = picture->f_motion.pmv[0][1] = tmp; + + DUMPBITS (bit_buf, bits, 1); /* remove marker_bit */ +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fi_field (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + uint8_t ** ref_field; + + NEEDBITS (bit_buf, bits, bit_ptr); + ref_field = motion->ref2[UBITS (bit_buf, 1)]; + + // TODO field select may need to do something here for bob (weave ok) + picture->XvMC_mv_field_sel[0][0] = picture->XvMC_mv_field_sel[1][0] = 0; + + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[0][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[0][1] + get_xvmc_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion->pmv[0][1] = motion_y; + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fi_16x8 (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + uint8_t ** ref_field; + + NEEDBITS (bit_buf, bits, bit_ptr); + ref_field = motion->ref2[UBITS (bit_buf, 1)]; + + // TODO field select may need to do something here bob (weave ok) + picture->XvMC_mv_field_sel[0][0] = picture->XvMC_mv_field_sel[1][0] = 0; + + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[0][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[0][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[0][1] + get_xvmc_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[0][1] = motion_y; + + + NEEDBITS (bit_buf, bits, bit_ptr); + ref_field = motion->ref2[UBITS (bit_buf, 1)]; + + // TODO field select may need to do something here for bob (weave ok) + picture->XvMC_mv_field_sel[0][0] = picture->XvMC_mv_field_sel[1][0] = 0; + + DUMPBITS (bit_buf, bits, 1); + + motion_x = motion->pmv[1][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion_x; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_y = motion->pmv[1][1] + get_xvmc_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion_y; + +#undef bit_buf +#undef bits +#undef bit_ptr +} + +static void motion_fi_dmv (picture_t * picture, motion_t * motion, + void (** table) (uint8_t *, uint8_t *, int, int)) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int motion_x, motion_y; + + NEEDBITS (bit_buf, bits, bit_ptr); + motion_x = motion->pmv[0][0] + get_xvmc_motion_delta (picture, + motion->f_code[0]); + motion_x = bound_motion_vector (motion_x, motion->f_code[0]); + motion->pmv[1][0] = motion->pmv[0][0] = motion_x; + NEEDBITS (bit_buf, bits, bit_ptr); + + motion_y = motion->pmv[0][1] + get_xvmc_motion_delta (picture, + motion->f_code[1]); + motion_y = bound_motion_vector (motion_y, motion->f_code[1]); + motion->pmv[1][1] = motion->pmv[0][1] = motion_y; + + // TODO field select may need to do something here for bob (weave ok) + picture->XvMC_mv_field_sel[0][0] = picture->XvMC_mv_field_sel[1][0] = 0; + +#undef bit_buf +#undef bits +#undef bit_ptr +} + + +static void motion_fi_conceal (picture_t * picture) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + DUMPBITS (bit_buf, bits, 1); /* remove field_select */ + + tmp = (picture->f_motion.pmv[0][0] + + get_xvmc_motion_delta (picture, picture->f_motion.f_code[0])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[0]); + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[0][0] = tmp; + + NEEDBITS (bit_buf, bits, bit_ptr); + tmp = (picture->f_motion.pmv[0][1] + + get_xvmc_motion_delta (picture, picture->f_motion.f_code[1])); + tmp = bound_motion_vector (tmp, picture->f_motion.f_code[1]); + picture->f_motion.pmv[1][1] = picture->f_motion.pmv[0][1] = tmp; + + DUMPBITS (bit_buf, bits, 1); /* remove marker_bit */ +#undef bit_buf +#undef bits +#undef bit_ptr +} + +#define MOTION_CALL(routine,direction) \ +do { \ + if ((direction) & MACROBLOCK_MOTION_FORWARD) \ + routine (picture, &(picture->f_motion), mpeg2_mc.put); \ + if ((direction) & MACROBLOCK_MOTION_BACKWARD) \ + routine (picture, &(picture->b_motion), \ + ((direction) & MACROBLOCK_MOTION_FORWARD ? \ + mpeg2_mc.avg : mpeg2_mc.put)); \ +} while (0) + +#define NEXT_MACROBLOCK \ +do { \ + picture->offset += 16; \ + if (picture->offset == picture->coded_picture_width) { \ + do { /* just so we can use the break statement */ \ + if (picture->current_frame->proc_slice) { \ + picture->current_frame->proc_slice (picture->current_frame, \ + picture->dest); \ + if (picture->picture_coding_type == B_TYPE) \ + break; \ + } \ + picture->dest[0] += 16 * picture->pitches[0]; \ + picture->dest[1] += 8 * picture->pitches[1]; \ + picture->dest[2] += 8 * picture->pitches[2]; \ + } while (0); \ + picture->v_offset += 16; \ + if (picture->v_offset > picture->limit_y) { \ + if (mpeg2_cpu_state_restore) \ + mpeg2_cpu_state_restore (&cpu_state); \ + return; \ + } \ + picture->offset = 0; \ + } \ +} while (0) + +static inline int slice_xvmc_init (picture_t * picture, int code) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + int offset, height; + struct vo_frame_s * forward_reference_frame; + struct vo_frame_s * backward_reference_frame; + const MBAtab * mba; + + offset = picture->picture_structure == BOTTOM_FIELD; + picture->pitches[0] = picture->current_frame->pitches[0]; + picture->pitches[1] = picture->current_frame->pitches[1]; + picture->pitches[2] = picture->current_frame->pitches[2]; + + if( picture->forward_reference_frame ) { + forward_reference_frame = picture->forward_reference_frame; + } + else { + /* return 1; */ + forward_reference_frame = picture->current_frame; + } + + if( picture->backward_reference_frame ) { + backward_reference_frame = picture->backward_reference_frame; + } + else { + /* return 1; */ + backward_reference_frame = picture->current_frame; + } + + picture->f_motion.ref[0][0] = + forward_reference_frame->base[0] + (offset ? picture->pitches[0] : 0); + picture->f_motion.ref[0][1] = + forward_reference_frame->base[1] + (offset ? picture->pitches[1] : 0); + picture->f_motion.ref[0][2] = + forward_reference_frame->base[2] + (offset ? picture->pitches[2] : 0); + + picture->b_motion.ref[0][0] = + backward_reference_frame->base[0] + (offset ? picture->pitches[0] : 0); + picture->b_motion.ref[0][1] = + backward_reference_frame->base[1] + (offset ? picture->pitches[1] : 0); + picture->b_motion.ref[0][2] = + backward_reference_frame->base[2] + (offset ? picture->pitches[2] : 0); + + if (picture->picture_structure != FRAME_PICTURE) { + uint8_t ** forward_ref; + int bottom_field; + + bottom_field = (picture->picture_structure == BOTTOM_FIELD); + picture->dmv_offset = bottom_field ? 1 : -1; + picture->f_motion.ref2[0] = picture->f_motion.ref[bottom_field]; + picture->f_motion.ref2[1] = picture->f_motion.ref[!bottom_field]; + picture->b_motion.ref2[0] = picture->b_motion.ref[bottom_field]; + picture->b_motion.ref2[1] = picture->b_motion.ref[!bottom_field]; + + forward_ref = forward_reference_frame->base; + if (picture->second_field && (picture->picture_coding_type != B_TYPE)) + forward_ref = picture->current_frame->base; + + picture->f_motion.ref[1][0] = forward_ref[0] + (bottom_field ? 0 : picture->pitches[0]); + picture->f_motion.ref[1][1] = forward_ref[1] + (bottom_field ? 0 : picture->pitches[1]); + picture->f_motion.ref[1][2] = forward_ref[2] + (bottom_field ? 0 : picture->pitches[2]); + + picture->b_motion.ref[1][0] = + backward_reference_frame->base[0] + (bottom_field ? 0 : picture->pitches[0]); + picture->b_motion.ref[1][1] = + backward_reference_frame->base[1] + (bottom_field ? 0 : picture->pitches[1]); + picture->b_motion.ref[1][2] = + backward_reference_frame->base[2] + (bottom_field ? 0 : picture->pitches[2]); + } + + picture->f_motion.pmv[0][0] = picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[1][1] = 0; + picture->b_motion.pmv[0][0] = picture->b_motion.pmv[0][1] = 0; + picture->b_motion.pmv[1][0] = picture->b_motion.pmv[1][1] = 0; + + picture->v_offset = (code - 1) * 16; + offset = (code - 1); + if (picture->current_frame->proc_slice && picture->picture_coding_type == B_TYPE) + offset = 0; + else if (picture->picture_structure != FRAME_PICTURE) + offset = 2 * offset; + + picture->dest[0] = picture->current_frame->base[0] + picture->pitches[0] * offset * 16; + picture->dest[1] = picture->current_frame->base[1] + picture->pitches[1] * offset * 8; + picture->dest[2] = picture->current_frame->base[2] + picture->pitches[2] * offset * 8; + + height = picture->coded_picture_height; + switch (picture->picture_structure) { + case BOTTOM_FIELD: + picture->dest[0] += picture->pitches[0]; + picture->dest[1] += picture->pitches[1]; + picture->dest[2] += picture->pitches[2]; + /* follow thru */ + case TOP_FIELD: + picture->pitches[0] <<= 1; + picture->pitches[1] <<= 1; + picture->pitches[2] <<= 1; + height >>= 1; + } + picture->limit_x = 2 * picture->coded_picture_width - 32; + picture->limit_y_16 = 2 * height - 32; + picture->limit_y_8 = 2 * height - 16; + picture->limit_y = height - 16; + + //TODO conversion to signed format signed format + if((picture->mc->xvmc_accel & ACCEL) == MOTION_ACCEL && + !(picture->mc->xvmc_accel & SIGNED_INTRA)) { + //Motion Comp only unsigned intra + // original: + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 1 << (picture->intra_dc_precision + 7); + } else { + //Motion Comp only signed intra MOTION_ACCEL+SIGNED_INTRA + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 0; + } + + picture->quantizer_scale = get_xvmc_quantizer_scale (picture); + + /* ignore intra_slice and all the extra data */ + while (bit_buf & 0x80000000) { + DUMPBITS (bit_buf, bits, 9); + NEEDBITS (bit_buf, bits, bit_ptr); + } + + /* decode initial macroblock address increment */ + offset = 0; + while (1) { + if (bit_buf >= 0x08000000) { + mba = MBA_5 + (UBITS (bit_buf, 6) - 2); + break; + } else if (bit_buf >= 0x01800000) { + mba = MBA_11 + (UBITS (bit_buf, 12) - 24); + break; + } else switch (UBITS (bit_buf, 12)) { + case 8: /* macroblock_escape */ + offset += 33; + DUMPBITS (bit_buf, bits, 11); + NEEDBITS (bit_buf, bits, bit_ptr); + continue; + case 15: /* macroblock_stuffing (MPEG1 only) */ + bit_buf &= 0xfffff; + DUMPBITS (bit_buf, bits, 11); + NEEDBITS (bit_buf, bits, bit_ptr); + continue; + default: /* error */ + return 1; + } + } + DUMPBITS (bit_buf, bits, mba->len + 1); + picture->offset = (offset + mba->mba) << 4; + + while (picture->offset - picture->coded_picture_width >= 0) { + picture->offset -= picture->coded_picture_width; + if ((picture->current_frame->proc_slice == NULL) || + (picture->picture_coding_type != B_TYPE)) { + picture->dest[0] += 16 * picture->pitches[0]; + picture->dest[1] += 8 * picture->pitches[1]; + picture->dest[2] += 8 * picture->pitches[2]; + } + picture->v_offset += 16; + } + if (picture->v_offset > picture->limit_y) + return 1; + + return 0; +#undef bit_buf +#undef bits +#undef bit_ptr +} + +void mpeg2_xvmc_slice (mpeg2dec_accel_t *accel, picture_t * picture, int code, uint8_t * buffer) +{ +#define bit_buf (picture->bitstream_buf) +#define bits (picture->bitstream_bits) +#define bit_ptr (picture->bitstream_ptr) + cpu_state_t cpu_state; + xine_xvmc_t *xvmc = (xine_xvmc_t *) picture->current_frame->accel_data; + + if (1 == code) { + accel->xvmc_last_slice_code = 0; + } + if ((code != accel->xvmc_last_slice_code + 1) && + (code != accel->xvmc_last_slice_code)) + return; + + bitstream_init (picture, buffer); + + if (slice_xvmc_init (picture, code)) + return; + + if (mpeg2_cpu_state_save) + mpeg2_cpu_state_save (&cpu_state); + + while (1) { + int macroblock_modes; + int mba_inc; + const MBAtab * mba; + + NEEDBITS (bit_buf, bits, bit_ptr); + + macroblock_modes = get_xvmc_macroblock_modes (picture); //macroblock_modes() + picture->XvMC_mb_type = macroblock_modes & 0x1F; + picture->XvMC_dct_type = (macroblock_modes & DCT_TYPE_INTERLACED)>>5; + picture->XvMC_motion_type = (macroblock_modes & MOTION_TYPE_MASK)>>6; + + picture->XvMC_x = picture->offset/16; + picture->XvMC_y = picture->v_offset/16; + + if((picture->XvMC_x == 0) && (picture->XvMC_y == 0)) { + picture->XvMC_mv_field_sel[0][0] = + picture->XvMC_mv_field_sel[1][0] = + picture->XvMC_mv_field_sel[0][1] = + picture->XvMC_mv_field_sel[1][1] = 0; + } + + picture->XvMC_cbp = 0x3f; //TODO set for intra 4:2:0 6 blocks yyyyuv all enabled + + /* maybe integrate MACROBLOCK_QUANT test into get_xvmc_macroblock_modes ? */ + if (macroblock_modes & MACROBLOCK_QUANT) + picture->quantizer_scale = get_xvmc_quantizer_scale (picture); + if (macroblock_modes & MACROBLOCK_INTRA) { + + int DCT_offset, DCT_stride; + int offset; + uint8_t * dest_y; + + if (picture->concealment_motion_vectors) { + if (picture->picture_structure == FRAME_PICTURE) + motion_fr_conceal (picture); + else + motion_fi_conceal (picture); + } else { + picture->f_motion.pmv[0][0] = picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[1][1] = 0; + picture->b_motion.pmv[0][0] = picture->b_motion.pmv[0][1] = 0; + picture->b_motion.pmv[1][0] = picture->b_motion.pmv[1][1] = 0; + } + + if (macroblock_modes & DCT_TYPE_INTERLACED) { + DCT_offset = picture->pitches[0]; + DCT_stride = picture->pitches[0] * 2; + } else { + DCT_offset = picture->pitches[0] * 8; + DCT_stride = picture->pitches[0]; + } + offset = picture->offset; + dest_y = picture->dest[0] + offset; + // unravaled loop of 6 block(i) calls in macroblock() + slice_xvmc_intra_DCT (picture, 0, dest_y, DCT_stride); + slice_xvmc_intra_DCT (picture, 0, dest_y + 8, DCT_stride); + slice_xvmc_intra_DCT (picture, 0, dest_y + DCT_offset, DCT_stride); + slice_xvmc_intra_DCT (picture, 0, dest_y + DCT_offset + 8, DCT_stride); + slice_xvmc_intra_DCT (picture, 1, picture->dest[1] + (offset >> 1), + picture->pitches[1]); + slice_xvmc_intra_DCT (picture, 2, picture->dest[2] + (offset >> 1), + picture->pitches[2]); + + if (picture->picture_coding_type == D_TYPE) { + NEEDBITS (bit_buf, bits, bit_ptr); + DUMPBITS (bit_buf, bits, 1); + } + } else { + picture->XvMC_cbp = 0; + + if (picture->picture_structure == FRAME_PICTURE) + switch (macroblock_modes & MOTION_TYPE_MASK) { + case MC_FRAME: + if (picture->mpeg1) { + MOTION_CALL (motion_mp1, macroblock_modes); + } else { + MOTION_CALL (motion_fr_frame, macroblock_modes); + } + break; + + case MC_FIELD: + //MOTION_CALL (motion_fr_field, macroblock_modes); + + if ((macroblock_modes) & MACROBLOCK_MOTION_FORWARD) + motion_fr_field(picture, &(picture->f_motion), + mpeg2_mc.put,0); + if ((macroblock_modes) & MACROBLOCK_MOTION_BACKWARD) + motion_fr_field(picture, &(picture->b_motion), + ((macroblock_modes) & MACROBLOCK_MOTION_FORWARD ? + mpeg2_mc.avg : mpeg2_mc.put),1); + + break; + + case MC_DMV: + MOTION_CALL (motion_fr_dmv, MACROBLOCK_MOTION_FORWARD); + break; + + case 0: + /* non-intra mb without forward mv in a P picture */ + picture->f_motion.pmv[0][0] = 0; + picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = 0; + picture->f_motion.pmv[1][1] = 0; + // MOTION_CALL (motion_zero, MACROBLOCK_MOTION_FORWARD); + break; + } + else + switch (macroblock_modes & MOTION_TYPE_MASK) { + case MC_FIELD: + MOTION_CALL (motion_fi_field, macroblock_modes); + break; + + case MC_16X8: + MOTION_CALL (motion_fi_16x8, macroblock_modes); + break; + + case MC_DMV: + MOTION_CALL (motion_fi_dmv, MACROBLOCK_MOTION_FORWARD); + break; + + case 0: + /* non-intra mb without forward mv in a P picture */ + picture->f_motion.pmv[0][0] = 0; + picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = 0; + picture->f_motion.pmv[1][1] = 0; + // MOTION_CALL (motion_zero, MACROBLOCK_MOTION_FORWARD); + break; + } + + if (macroblock_modes & MACROBLOCK_PATTERN) { + int coded_block_pattern; + int DCT_offset, DCT_stride; + int offset; + uint8_t * dest_y; + + if (macroblock_modes & DCT_TYPE_INTERLACED) { + DCT_offset = picture->pitches[0]; + DCT_stride = picture->pitches[0] * 2; + } else { + DCT_offset = picture->pitches[0] * 8; + DCT_stride = picture->pitches[0]; + } + + picture->XvMC_cbp = coded_block_pattern = get_xvmc_coded_block_pattern (picture); + offset = picture->offset; + dest_y = picture->dest[0] + offset; + // TODO optimize not fully used for idct accel only mc. + if (coded_block_pattern & 0x20) + slice_xvmc_non_intra_DCT (picture, dest_y, DCT_stride); // cc0 luma 0 + if (coded_block_pattern & 0x10) + slice_xvmc_non_intra_DCT (picture, dest_y + 8, DCT_stride); // cc0 luma 1 + if (coded_block_pattern & 0x08) + slice_xvmc_non_intra_DCT (picture, dest_y + DCT_offset, + DCT_stride); // cc0 luma 2 + if (coded_block_pattern & 0x04) + slice_xvmc_non_intra_DCT (picture, dest_y + DCT_offset + 8, + DCT_stride); // cc0 luma 3 + if (coded_block_pattern & 0x2) + slice_xvmc_non_intra_DCT (picture, + picture->dest[1] + (offset >> 1), + picture->pitches[1]); // cc1 croma + if (coded_block_pattern & 0x1) + slice_xvmc_non_intra_DCT (picture, + picture->dest[2] + (offset >> 1), + picture->pitches[2]); // cc2 croma + } + + if((picture->mc->xvmc_accel & ACCEL) == MOTION_ACCEL && + !(picture->mc->xvmc_accel & SIGNED_INTRA)) { + // original: + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 128 << picture->intra_dc_precision; + + } else { // MOTION_ACCEL+SIGNED_INTRA + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 0; + } + + } + xvmc->proc_macro_block(picture->XvMC_x, picture->XvMC_y, + picture->XvMC_mb_type, + picture->XvMC_motion_type, + picture->XvMC_mv_field_sel, + picture->XvMC_dmvector, + picture->XvMC_cbp, + picture->XvMC_dct_type, + picture->current_frame, + picture->forward_reference_frame, + picture->backward_reference_frame, + picture->picture_structure, + picture->second_field, + picture->f_motion.pmv, + picture->b_motion.pmv); + + + NEXT_MACROBLOCK; + + NEEDBITS (bit_buf, bits, bit_ptr); + mba_inc = 0; + while (1) { + if (bit_buf >= 0x10000000) { + mba = MBA_5 + (UBITS (bit_buf, 5) - 2); + break; + } else if (bit_buf >= 0x03000000) { + mba = MBA_11 + (UBITS (bit_buf, 11) - 24); + break; + } else switch (UBITS (bit_buf, 11)) { + case 8: /* macroblock_escape */ + mba_inc += 33; + /* pass through */ + case 15: /* macroblock_stuffing (MPEG1 only) */ + DUMPBITS (bit_buf, bits, 11); + NEEDBITS (bit_buf, bits, bit_ptr); + continue; + default: /* end of slice, or error */ + if (mpeg2_cpu_state_restore) + mpeg2_cpu_state_restore (&cpu_state); + accel->xvmc_last_slice_code = code; + return; + } + } + DUMPBITS (bit_buf, bits, mba->len); + mba_inc += mba->mba; + if (mba_inc) { + //TODO conversion to signed format signed format + if((picture->mc->xvmc_accel & ACCEL) == MOTION_ACCEL && + !(picture->mc->xvmc_accel & SIGNED_INTRA)) { + // original: + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 128 << picture->intra_dc_precision; + } else { // MOTION_ACCEL+SIGNED_INTRA + picture->dc_dct_pred[0] = picture->dc_dct_pred[1] = + picture->dc_dct_pred[2] = 0; + } + + picture->XvMC_cbp = 0; + if (picture->picture_coding_type == P_TYPE) { + picture->f_motion.pmv[0][0] = picture->f_motion.pmv[0][1] = 0; + picture->f_motion.pmv[1][0] = picture->f_motion.pmv[1][1] = 0; + + do { + if(picture->mc->xvmc_accel) { + + /* derive motion_type */ + if(picture->picture_structure == FRAME_PICTURE) { + picture->XvMC_motion_type = XINE_MC_FRAME; + } else { + picture->XvMC_motion_type = XINE_MC_FIELD; + /* predict from field of same parity */ + picture->XvMC_mv_field_sel[0][0] = + picture->XvMC_mv_field_sel[0][1] = + (picture->picture_structure==BOTTOM_FIELD); + } + picture->XvMC_mb_type = macroblock_modes & 0x1E; + picture->XvMC_x = picture->offset/16; + picture->XvMC_y = picture->v_offset/16; + + xvmc->proc_macro_block(picture->XvMC_x,picture->XvMC_y, + picture->XvMC_mb_type, + picture->XvMC_motion_type, + picture->XvMC_mv_field_sel, + picture->XvMC_dmvector, + picture->XvMC_cbp, + picture->XvMC_dct_type, + picture->current_frame, + picture->forward_reference_frame, + picture->backward_reference_frame, + picture->picture_structure, + picture->second_field, + picture->f_motion.pmv, + picture->b_motion.pmv); + } else { + // MOTION_CALL (motion_zero, MACROBLOCK_MOTION_FORWARD); + } + NEXT_MACROBLOCK; + } while (--mba_inc); + } else { + do { + if(picture->mc->xvmc_accel) { + + /* derive motion_type */ + if(picture->picture_structure == FRAME_PICTURE) { + picture->XvMC_motion_type = XINE_MC_FRAME; + } else { + picture->XvMC_motion_type = XINE_MC_FIELD; + /* predict from field of same parity */ + picture->XvMC_mv_field_sel[0][0] = + picture->XvMC_mv_field_sel[0][1] = + (picture->picture_structure==BOTTOM_FIELD); + } + + picture->XvMC_mb_type = macroblock_modes & 0x1E; + picture->XvMC_x = picture->offset/16; + picture->XvMC_y = picture->v_offset/16; + + xvmc->proc_macro_block(picture->XvMC_x,picture->XvMC_y, + picture->XvMC_mb_type, + picture->XvMC_motion_type, + picture->XvMC_mv_field_sel, + picture->XvMC_dmvector, + picture->XvMC_cbp, + picture->XvMC_dct_type, + picture->current_frame, + picture->forward_reference_frame, + picture->backward_reference_frame, + picture->picture_structure, + picture->second_field, + picture->f_motion.pmv, + picture->b_motion.pmv); + } else { + MOTION_CALL (motion_reuse, macroblock_modes); + } + NEXT_MACROBLOCK; + } while (--mba_inc); + } + } + } + accel->xvmc_last_slice_code = code; +#undef bit_buf +#undef bits +#undef bit_ptr +} + diff --git a/src/video_dec/libmpeg2/slice_xvmc_vld.c b/src/video_dec/libmpeg2/slice_xvmc_vld.c new file mode 100644 index 000000000..3606cf66b --- /dev/null +++ b/src/video_dec/libmpeg2/slice_xvmc_vld.c @@ -0,0 +1,225 @@ +/* + * Copyright (c) 2004 The Unichrome project. All rights reserved. + * + * 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, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTIES OR REPRESENTATIONS; 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. + * + * + */ + +#include <xine/xine_internal.h> +#include <xine/video_out.h> +#include "mpeg2.h" +#include "mpeg2_internal.h" +#include "xvmc_vld.h" + +static const uint8_t zig_zag_scan[64] ATTR_ALIGN(16) = +{ + /* Zig-Zag scan pattern */ + 0, 1, 8,16, 9, 2, 3,10, + 17,24,32,25,18,11, 4, 5, + 12,19,26,33,40,48,41,34, + 27,20,13, 6, 7,14,21,28, + 35,42,49,56,57,50,43,36, + 29,22,15,23,30,37,44,51, + 58,59,52,45,38,31,39,46, + 53,60,61,54,47,55,62,63 +}; + +static const uint8_t alternate_scan [64] ATTR_ALIGN(16) = +{ + /* Alternate scan pattern */ + 0,8,16,24,1,9,2,10,17,25,32,40,48,56,57,49, + 41,33,26,18,3,11,4,12,19,27,34,42,50,58,35,43, + 51,59,20,28,5,13,6,14,21,29,36,44,52,60,37,45, + 53,61,22,30,7,15,23,31,38,46,54,62,39,47,55,63 +}; + +void mpeg2_xxmc_slice( mpeg2dec_accel_t *accel, picture_t *picture, + int code, uint8_t *buffer, uint32_t chunk_size, + uint8_t *chunk_buffer) + +{ + vo_frame_t + *frame = picture->current_frame; + xine_xxmc_t + *xxmc = (xine_xxmc_t *) frame->accel_data; + xine_vld_frame_t + *vft = &xxmc->vld_frame; + unsigned + mb_frame_height; + int + i; + const uint8_t * + scan_pattern; + float + ms_per_slice; + + if (1 == code && accel->xvmc_last_slice_code != 1) { + frame->bad_frame = 1; + accel->slices_per_row = 1; + accel->row_slice_count = 1; + + /* + * Check that first field went through OK. Otherwise, + * indicate bad frame. + */ + + if (picture->second_field) { + accel->xvmc_last_slice_code = (xxmc->decoded) ? 0 : -1; + xxmc->decoded = 0; + } else { + accel->xvmc_last_slice_code = 0; + } + + mb_frame_height = + (!(picture->mpeg1) && (picture->progressive_sequence)) ? + 2*((picture->coded_picture_height+31) >> 5) : + (picture->coded_picture_height+15) >> 4; + accel->xxmc_mb_pic_height = (picture->picture_structure == FRAME_PICTURE ) ? + mb_frame_height : mb_frame_height >> 1; + + ms_per_slice = 1000. / (90000. * mb_frame_height) * frame->duration; + xxmc->sleep = 1. / (ms_per_slice * 0.45); + if (xxmc->sleep < 1.) xxmc->sleep = 1.; + + if (picture->mpeg1) { + vft->mv_ranges[0][0] = picture->b_motion.f_code[0]; + vft->mv_ranges[0][1] = picture->b_motion.f_code[0]; + vft->mv_ranges[1][0] = picture->f_motion.f_code[0]; + vft->mv_ranges[1][1] = picture->f_motion.f_code[0]; + } else { + vft->mv_ranges[0][0] = picture->b_motion.f_code[0]; + vft->mv_ranges[0][1] = picture->b_motion.f_code[1]; + vft->mv_ranges[1][0] = picture->f_motion.f_code[0]; + vft->mv_ranges[1][1] = picture->f_motion.f_code[1]; + } + + vft->picture_structure = picture->picture_structure; + vft->picture_coding_type = picture->picture_coding_type; + vft->mpeg_coding = (picture->mpeg1) ? 0 : 1; + vft->progressive_sequence = picture->progressive_sequence; + vft->scan = (picture->scan == mpeg2_scan_alt); + vft->pred_dct_frame = picture->frame_pred_frame_dct; + vft->concealment_motion_vectors = + picture->concealment_motion_vectors; + vft->q_scale_type = picture->q_scale_type; + vft->intra_vlc_format = picture->intra_vlc_format; + vft->intra_dc_precision = picture->intra_dc_precision; + vft->second_field = picture->second_field; + + /* + * Translation of libmpeg2's Q-matrix layout to VLD XvMC's. + * Errors here will give + * blocky artifacts and sometimes wrong colors. + */ + + scan_pattern = (vft->scan) ? alternate_scan : zig_zag_scan; + + if ((vft->load_intra_quantizer_matrix = picture->load_intra_quantizer_matrix)) { + for (i=0; i<64; ++i) { + vft->intra_quantizer_matrix[scan_pattern[i]] = + picture->intra_quantizer_matrix[picture->scan[i]]; + } + } + + if ((vft->load_non_intra_quantizer_matrix = picture->load_non_intra_quantizer_matrix)) { + for (i=0; i<64; ++i) { + vft->non_intra_quantizer_matrix[scan_pattern[i]] = + picture->non_intra_quantizer_matrix[picture->scan[i]]; + } + } + + picture->load_intra_quantizer_matrix = 0; + picture->load_non_intra_quantizer_matrix = 0; + vft->forward_reference_frame = picture->forward_reference_frame; + vft->backward_reference_frame = picture->backward_reference_frame; + xxmc->proc_xxmc_begin( frame ); + if (xxmc->result != 0) { + accel->xvmc_last_slice_code=-1; + } + } + + if (((code == accel->xvmc_last_slice_code + 1) || + (code == accel->xvmc_last_slice_code))) { + + /* + * Send this slice to the output plugin. May stall for a long + * time in proc_slice; + */ + + frame->bad_frame = 1; + xxmc->slice_data_size = chunk_size; + xxmc->slice_data = chunk_buffer; + xxmc->slice_code = code; + + xxmc->proc_xxmc_slice( frame ); + + if (xxmc->result != 0) { + accel->xvmc_last_slice_code=-1; + return; + } + /* + * Keep track of slices. + */ + + accel->row_slice_count = (accel->xvmc_last_slice_code == code) ? + accel->row_slice_count + 1 : 1; + accel->slices_per_row = (accel->row_slice_count > accel->slices_per_row) ? + accel->row_slice_count:accel->slices_per_row; + accel->xvmc_last_slice_code = code; + + } else { + + /* + * An error has occured. + */ + + lprintf("libmpeg2: VLD XvMC: Slice error.\n"); + accel->xvmc_last_slice_code = -1; + return; + } +} + +void mpeg2_xxmc_vld_frame_complete(mpeg2dec_accel_t *accel, picture_t *picture, int code) +{ + vo_frame_t + *frame = picture->current_frame; + xine_xxmc_t + *xxmc = (xine_xxmc_t *) frame->accel_data; + + if (xxmc->decoded) return; + if (accel->xvmc_last_slice_code == -1) { + xxmc->proc_xxmc_flush( frame ); + return; + } + + if ((code != 0xff) || ((accel->xvmc_last_slice_code == + accel->xxmc_mb_pic_height) && + accel->slices_per_row == accel->row_slice_count)) { + + xxmc->proc_xxmc_flush( frame ); + + if (xxmc->result) { + accel->xvmc_last_slice_code=-1; + frame->bad_frame = 1; + return; + } + xxmc->decoded = 1; + accel->xvmc_last_slice_code = 0; + if (picture->picture_structure == 3 || picture->second_field) { + if (xxmc->result == 0) + frame->bad_frame = 0; + } + } +} diff --git a/src/video_dec/libmpeg2/stats.c b/src/video_dec/libmpeg2/stats.c new file mode 100644 index 000000000..63c701179 --- /dev/null +++ b/src/video_dec/libmpeg2/stats.c @@ -0,0 +1,317 @@ +/* + * stats.c + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#include "config.h" + +#include <stdio.h> +#include <stdlib.h> +#include <inttypes.h> + +#include "mpeg2_internal.h" + +static int debug_level = -1; + +/* Determine is debug output is required. */ +/* We could potentially have multiple levels of debug info */ +static int debug_is_on (void) +{ + char * env_var; + + if (debug_level < 0) { + env_var = getenv ("MPEG2_DEBUG"); + + if (env_var) + debug_level = 1; + else + debug_level = 0; + } + + return debug_level; +} + +static void stats_picture (uint8_t * buffer) +{ + static const char *const picture_coding_type_str [8] = { + "Invalid picture type", + "I-type", + "P-type", + "B-type", + "D (very bad)", + "Invalid","Invalid","Invalid" + }; + + int picture_coding_type; + int temporal_reference; + int vbv_delay; + + temporal_reference = (buffer[0] << 2) | (buffer[1] >> 6); + picture_coding_type = (buffer [1] >> 3) & 7; + vbv_delay = ((buffer[1] << 13) | (buffer[2] << 5) | + (buffer[3] >> 3)) & 0xffff; + + fprintf (stderr, " (picture) %s temporal_reference %d, vbv_delay %d\n", + picture_coding_type_str [picture_coding_type], + temporal_reference, vbv_delay); +} + +static void stats_user_data (uint8_t * buffer) +{ + fprintf (stderr, " (user_data)\n"); +} + +static void stats_sequence (uint8_t * buffer) +{ + static const char *const aspect_ratio_information_str[8] = { + "Invalid Aspect Ratio", + "1:1", + "4:3", + "16:9", + "2.21:1", + "Invalid Aspect Ratio", + "Invalid Aspect Ratio", + "Invalid Aspect Ratio" + }; + static const char *const frame_rate_str[16] = { + "Invalid frame_rate_code", + "23.976", "24", "25" , "29.97", + "30" , "50", "59.94", "60" , + "Invalid frame_rate_code", "Invalid frame_rate_code", + "Invalid frame_rate_code", "Invalid frame_rate_code", + "Invalid frame_rate_code", "Invalid frame_rate_code", + "Invalid frame_rate_code" + }; + + int horizontal_size; + int vertical_size; + int aspect_ratio_information; + int frame_rate_code; + int bit_rate_value; + int vbv_buffer_size_value; + int constrained_parameters_flag; + int load_intra_quantizer_matrix; + int load_non_intra_quantizer_matrix; + + vertical_size = (buffer[0] << 16) | (buffer[1] << 8) | buffer[2]; + horizontal_size = vertical_size >> 12; + vertical_size &= 0xfff; + aspect_ratio_information = buffer[3] >> 4; + frame_rate_code = buffer[3] & 15; + bit_rate_value = (buffer[4] << 10) | (buffer[5] << 2) | (buffer[6] >> 6); + vbv_buffer_size_value = ((buffer[6] << 5) | (buffer[7] >> 3)) & 0x3ff; + constrained_parameters_flag = buffer[7] & 4; + load_intra_quantizer_matrix = buffer[7] & 2; + if (load_intra_quantizer_matrix) + buffer += 64; + load_non_intra_quantizer_matrix = buffer[7] & 1; + + fprintf (stderr, " (seq) %dx%d %s, %s fps, %5.0f kbps, VBV %d kB%s%s%s\n", + horizontal_size, vertical_size, + aspect_ratio_information_str [aspect_ratio_information], + frame_rate_str [frame_rate_code], + bit_rate_value * 400.0 / 1000.0, + 2 * vbv_buffer_size_value, + constrained_parameters_flag ? " , CP":"", + load_intra_quantizer_matrix ? " , Custom Intra Matrix":"", + load_non_intra_quantizer_matrix ? " , Custom Non-Intra Matrix":""); +} + +static void stats_sequence_error (uint8_t * buffer) +{ + fprintf (stderr, " (sequence_error)\n"); +} + +static void stats_sequence_end (uint8_t * buffer) +{ + fprintf (stderr, " (sequence_end)\n"); +} + +static void stats_group (uint8_t * buffer) +{ + fprintf (stderr, " (group)%s%s\n", + (buffer[4] & 0x40) ? " closed_gop" : "", + (buffer[4] & 0x20) ? " broken_link" : ""); +} + +static void stats_slice (int code, uint8_t * buffer) +{ + /* fprintf (stderr, " (slice %d)\n", code); */ +} + +static void stats_sequence_extension (uint8_t * buffer) +{ + static const char *const chroma_format_str[4] = { + "Invalid Chroma Format", + "4:2:0 Chroma", + "4:2:2 Chroma", + "4:4:4 Chroma" + }; + + int progressive_sequence; + int chroma_format; + + progressive_sequence = (buffer[1] >> 3) & 1; + chroma_format = (buffer[1] >> 1) & 3; + + fprintf (stderr, " (seq_ext) progressive_sequence %d, %s\n", + progressive_sequence, chroma_format_str [chroma_format]); +} + +static void stats_sequence_display_extension (uint8_t * buffer) +{ + fprintf (stderr, " (sequence_display_extension)\n"); +} + +static void stats_quant_matrix_extension (uint8_t * buffer) +{ + fprintf (stderr, " (quant_matrix_extension)\n"); +} + +static void stats_copyright_extension (uint8_t * buffer) +{ + fprintf (stderr, " (copyright_extension)\n"); +} + + +static void stats_sequence_scalable_extension (uint8_t * buffer) +{ + fprintf (stderr, " (sequence_scalable_extension)\n"); +} + +static void stats_picture_display_extension (uint8_t * buffer) +{ + fprintf (stderr, " (picture_display_extension)\n"); +} + +static void stats_picture_coding_extension (uint8_t * buffer) +{ + static const char *const picture_structure_str[4] = { + "Invalid Picture Structure", + "Top field", + "Bottom field", + "Frame Picture" + }; + + int f_code[2][2]; + int intra_dc_precision; + int picture_structure; + int top_field_first; + int frame_pred_frame_dct; + int concealment_motion_vectors; + int q_scale_type; + int intra_vlc_format; + int alternate_scan; + int repeat_first_field; + int progressive_frame; + + f_code[0][0] = buffer[0] & 15; + f_code[0][1] = buffer[1] >> 4; + f_code[1][0] = buffer[1] & 15; + f_code[1][1] = buffer[2] >> 4; + intra_dc_precision = (buffer[2] >> 2) & 3; + picture_structure = buffer[2] & 3; + top_field_first = buffer[3] >> 7; + frame_pred_frame_dct = (buffer[3] >> 6) & 1; + concealment_motion_vectors = (buffer[3] >> 5) & 1; + q_scale_type = (buffer[3] >> 4) & 1; + intra_vlc_format = (buffer[3] >> 3) & 1; + alternate_scan = (buffer[3] >> 2) & 1; + repeat_first_field = (buffer[3] >> 1) & 1; + progressive_frame = buffer[4] >> 7; + + fprintf (stderr, + " (pic_ext) %s\n", picture_structure_str [picture_structure]); + fprintf (stderr, + " (pic_ext) forward horizontal f_code % d, forward vertical f_code % d\n", + f_code[0][0], f_code[0][1]); + fprintf (stderr, + " (pic_ext) backward horizontal f_code % d, backward vertical f_code % d\n", + f_code[1][0], f_code[1][1]); + fprintf (stderr, + " (pic_ext) intra_dc_precision %d, top_field_first %d, frame_pred_frame_dct %d\n", + intra_dc_precision, top_field_first, frame_pred_frame_dct); + fprintf (stderr, + " (pic_ext) concealment_motion_vectors %d, q_scale_type %d, intra_vlc_format %d\n", + concealment_motion_vectors, q_scale_type, intra_vlc_format); + fprintf (stderr, + " (pic_ext) alternate_scan %d, repeat_first_field %d, progressive_frame %d\n", + alternate_scan, repeat_first_field, progressive_frame); +} + +void mpeg2_stats (int code, uint8_t * buffer) +{ + if (! (debug_is_on ())) + return; + + switch (code) { + case 0x00: + stats_picture (buffer); + break; + case 0xb2: + stats_user_data (buffer); + break; + case 0xb3: + stats_sequence (buffer); + break; + case 0xb4: + stats_sequence_error (buffer); + break; + case 0xb5: + switch (buffer[0] >> 4) { + case 1: + stats_sequence_extension (buffer); + break; + case 2: + stats_sequence_display_extension (buffer); + break; + case 3: + stats_quant_matrix_extension (buffer); + break; + case 4: + stats_copyright_extension (buffer); + break; + case 5: + stats_sequence_scalable_extension (buffer); + break; + case 7: + stats_picture_display_extension (buffer); + break; + case 8: + stats_picture_coding_extension (buffer); + break; + default: + fprintf (stderr, " (unknown extension %#x)\n", buffer[0] >> 4); + } + break; + case 0xb7: + stats_sequence_end (buffer); + break; + case 0xb8: + stats_group (buffer); + break; + default: + if (code < 0xb0) + stats_slice (code, buffer); + else + fprintf (stderr, " (unknown start code %#02x)\n", code); + } +} diff --git a/src/video_dec/libmpeg2/vis.h b/src/video_dec/libmpeg2/vis.h new file mode 100644 index 000000000..69dd49075 --- /dev/null +++ b/src/video_dec/libmpeg2/vis.h @@ -0,0 +1,328 @@ +/* + * vis.h + * Copyright (C) 2003 David S. Miller <davem@redhat.com> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +/* You may be asking why I hard-code the instruction opcodes and don't + * use the normal VIS assembler mnenomics for the VIS instructions. + * + * The reason is that Sun, in their infinite wisdom, decided that a binary + * using a VIS instruction will cause it to be marked (in the ELF headers) + * as doing so, and this prevents the OS from loading such binaries if the + * current cpu doesn't have VIS. There is no way to easily override this + * behavior of the assembler that I am aware of. + * + * This totally defeats what libmpeg2 is trying to do which is allow a + * single binary to be created, and then detect the availability of VIS + * at runtime. + * + * I'm not saying that tainting the binary by default is bad, rather I'm + * saying that not providing a way to override this easily unnecessarily + * ties people's hands. + * + * Thus, we do the opcode encoding by hand and output 32-bit words in + * the assembler to keep the binary from becoming tainted. + */ + +#define vis_opc_base ((0x1 << 31) | (0x36 << 19)) +#define vis_opf(X) ((X) << 5) +#define vis_sreg(X) (X) +#define vis_dreg(X) (((X)&0x1f)|((X)>>5)) +#define vis_rs1_s(X) (vis_sreg(X) << 14) +#define vis_rs1_d(X) (vis_dreg(X) << 14) +#define vis_rs2_s(X) (vis_sreg(X) << 0) +#define vis_rs2_d(X) (vis_dreg(X) << 0) +#define vis_rd_s(X) (vis_sreg(X) << 25) +#define vis_rd_d(X) (vis_dreg(X) << 25) + +#define vis_ss2s(opf,rs1,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs1_s(rs1) | \ + vis_rs2_s(rs2) | \ + vis_rd_s(rd))) + +#define vis_dd2d(opf,rs1,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs1_d(rs1) | \ + vis_rs2_d(rs2) | \ + vis_rd_d(rd))) + +#define vis_ss2d(opf,rs1,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs1_s(rs1) | \ + vis_rs2_s(rs2) | \ + vis_rd_d(rd))) + +#define vis_sd2d(opf,rs1,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs1_s(rs1) | \ + vis_rs2_d(rs2) | \ + vis_rd_d(rd))) + +#define vis_d2s(opf,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs2_d(rs2) | \ + vis_rd_s(rd))) + +#define vis_s2d(opf,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs2_s(rs2) | \ + vis_rd_d(rd))) + +#define vis_d12d(opf,rs1,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs1_d(rs1) | \ + vis_rd_d(rd))) + +#define vis_d22d(opf,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs2_d(rs2) | \ + vis_rd_d(rd))) + +#define vis_s12s(opf,rs1,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs1_s(rs1) | \ + vis_rd_s(rd))) + +#define vis_s22s(opf,rs2,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rs2_s(rs2) | \ + vis_rd_s(rd))) + +#define vis_s(opf,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rd_s(rd))) + +#define vis_d(opf,rd) \ + __asm__ __volatile__ (".word %0" \ + : : "i" (vis_opc_base | vis_opf(opf) | \ + vis_rd_d(rd))) + +#define vis_r2m(op,rd,mem) \ + __asm__ __volatile__ (#op "\t%%f" #rd ", [%0]" : : "r" (&(mem)) ) + +#define vis_r2m_2(op,rd,mem1,mem2) \ + __asm__ __volatile__ (#op "\t%%f" #rd ", [%0 + %1]" : : "r" (mem1), "r" (mem2) ) + +#define vis_m2r(op,mem,rd) \ + __asm__ __volatile__ (#op "\t[%0], %%f" #rd : : "r" (&(mem)) ) + +#define vis_m2r_2(op,mem1,mem2,rd) \ + __asm__ __volatile__ (#op "\t[%0 + %1], %%f" #rd : : "r" (mem1), "r" (mem2) ) + +static inline void vis_set_gsr(unsigned int _val) +{ + register unsigned int val asm("g1"); + + val = _val; + __asm__ __volatile__(".word 0xa7804000" + : : "r" (val)); +} + +#define VIS_GSR_ALIGNADDR_MASK 0x0000007 +#define VIS_GSR_ALIGNADDR_SHIFT 0 +#define VIS_GSR_SCALEFACT_MASK 0x0000078 +#define VIS_GSR_SCALEFACT_SHIFT 3 + +#define vis_ld32(mem,rs1) vis_m2r(ld, mem, rs1) +#define vis_ld32_2(mem1,mem2,rs1) vis_m2r_2(ld, mem1, mem2, rs1) +#define vis_st32(rs1,mem) vis_r2m(st, rs1, mem) +#define vis_st32_2(rs1,mem1,mem2) vis_r2m_2(st, rs1, mem1, mem2) +#define vis_ld64(mem,rs1) vis_m2r(ldd, mem, rs1) +#define vis_ld64_2(mem1,mem2,rs1) vis_m2r_2(ldd, mem1, mem2, rs1) +#define vis_st64(rs1,mem) vis_r2m(std, rs1, mem) +#define vis_st64_2(rs1,mem1,mem2) vis_r2m_2(std, rs1, mem1, mem2) + +#define vis_ldblk(mem, rd) \ +do { register void *__mem asm("g1"); \ + __mem = &(mem); \ + __asm__ __volatile__(".word 0xc1985e00 | %1" \ + : \ + : "r" (__mem), \ + "i" (vis_rd_d(rd)) \ + : "memory"); \ +} while (0) + +#define vis_stblk(rd, mem) \ +do { register void *__mem asm("g1"); \ + __mem = &(mem); \ + __asm__ __volatile__(".word 0xc1b85e00 | %1" \ + : \ + : "r" (__mem), \ + "i" (vis_rd_d(rd)) \ + : "memory"); \ +} while (0) + +#define vis_membar_storestore() \ + __asm__ __volatile__(".word 0x8143e008" : : : "memory") + +#define vis_membar_sync() \ + __asm__ __volatile__(".word 0x8143e040" : : : "memory") + +/* 16 and 32 bit partitioned addition and subtraction. The normal + * versions perform 4 16-bit or 2 32-bit additions or subtractions. + * The 's' versions perform 2 16-bit or 2 32-bit additions or + * subtractions. + */ + +#define vis_padd16(rs1,rs2,rd) vis_dd2d(0x50, rs1, rs2, rd) +#define vis_padd16s(rs1,rs2,rd) vis_ss2s(0x51, rs1, rs2, rd) +#define vis_padd32(rs1,rs2,rd) vis_dd2d(0x52, rs1, rs2, rd) +#define vis_padd32s(rs1,rs2,rd) vis_ss2s(0x53, rs1, rs2, rd) +#define vis_psub16(rs1,rs2,rd) vis_dd2d(0x54, rs1, rs2, rd) +#define vis_psub16s(rs1,rs2,rd) vis_ss2s(0x55, rs1, rs2, rd) +#define vis_psub32(rs1,rs2,rd) vis_dd2d(0x56, rs1, rs2, rd) +#define vis_psub32s(rs1,rs2,rd) vis_ss2s(0x57, rs1, rs2, rd) + +/* Pixel formatting instructions. */ + +#define vis_pack16(rs2,rd) vis_d2s( 0x3b, rs2, rd) +#define vis_pack32(rs1,rs2,rd) vis_dd2d(0x3a, rs1, rs2, rd) +#define vis_packfix(rs2,rd) vis_d2s( 0x3d, rs2, rd) +#define vis_expand(rs2,rd) vis_s2d( 0x4d, rs2, rd) +#define vis_pmerge(rs1,rs2,rd) vis_ss2d(0x4b, rs1, rs2, rd) + +/* Partitioned multiply instructions. */ + +#define vis_mul8x16(rs1,rs2,rd) vis_sd2d(0x31, rs1, rs2, rd) +#define vis_mul8x16au(rs1,rs2,rd) vis_ss2d(0x33, rs1, rs2, rd) +#define vis_mul8x16al(rs1,rs2,rd) vis_ss2d(0x35, rs1, rs2, rd) +#define vis_mul8sux16(rs1,rs2,rd) vis_dd2d(0x36, rs1, rs2, rd) +#define vis_mul8ulx16(rs1,rs2,rd) vis_dd2d(0x37, rs1, rs2, rd) +#define vis_muld8sux16(rs1,rs2,rd) vis_ss2d(0x38, rs1, rs2, rd) +#define vis_muld8ulx16(rs1,rs2,rd) vis_ss2d(0x39, rs1, rs2, rd) + +/* Alignment instructions. */ + +static inline void *vis_alignaddr(void *_ptr) +{ + register void *ptr asm("g1"); + + ptr = _ptr; + + __asm__ __volatile__(".word %2" + : "=&r" (ptr) + : "0" (ptr), + "i" (vis_opc_base | vis_opf(0x18) | + vis_rs1_s(1) | + vis_rs2_s(0) | + vis_rd_s(1))); + + return ptr; +} + +static inline void vis_alignaddr_g0(void *_ptr) +{ + register void *ptr asm("g1"); + + ptr = _ptr; + + __asm__ __volatile__(".word %2" + : "=&r" (ptr) + : "0" (ptr), + "i" (vis_opc_base | vis_opf(0x18) | + vis_rs1_s(1) | + vis_rs2_s(0) | + vis_rd_s(0))); +} + +static inline void *vis_alignaddrl(void *_ptr) +{ + register void *ptr asm("g1"); + + ptr = _ptr; + + __asm__ __volatile__(".word %2" + : "=&r" (ptr) + : "0" (ptr), + "i" (vis_opc_base | vis_opf(0x19) | + vis_rs1_s(1) | + vis_rs2_s(0) | + vis_rd_s(1))); + + return ptr; +} + +static inline void vis_alignaddrl_g0(void *_ptr) +{ + register void *ptr asm("g1"); + + ptr = _ptr; + + __asm__ __volatile__(".word %2" + : "=&r" (ptr) + : "0" (ptr), + "i" (vis_opc_base | vis_opf(0x19) | + vis_rs1_s(1) | + vis_rs2_s(0) | + vis_rd_s(0))); +} + +#define vis_faligndata(rs1,rs2,rd) vis_dd2d(0x48, rs1, rs2, rd) + +/* Logical operate instructions. */ + +#define vis_fzero(rd) vis_d( 0x60, rd) +#define vis_fzeros(rd) vis_s( 0x61, rd) +#define vis_fone(rd) vis_d( 0x7e, rd) +#define vis_fones(rd) vis_s( 0x7f, rd) +#define vis_src1(rs1,rd) vis_d12d(0x74, rs1, rd) +#define vis_src1s(rs1,rd) vis_s12s(0x75, rs1, rd) +#define vis_src2(rs2,rd) vis_d22d(0x78, rs2, rd) +#define vis_src2s(rs2,rd) vis_s22s(0x79, rs2, rd) +#define vis_not1(rs1,rd) vis_d12d(0x6a, rs1, rd) +#define vis_not1s(rs1,rd) vis_s12s(0x6b, rs1, rd) +#define vis_not2(rs2,rd) vis_d22d(0x66, rs2, rd) +#define vis_not2s(rs2,rd) vis_s22s(0x67, rs2, rd) +#define vis_or(rs1,rs2,rd) vis_dd2d(0x7c, rs1, rs2, rd) +#define vis_ors(rs1,rs2,rd) vis_ss2s(0x7d, rs1, rs2, rd) +#define vis_nor(rs1,rs2,rd) vis_dd2d(0x62, rs1, rs2, rd) +#define vis_nors(rs1,rs2,rd) vis_ss2s(0x63, rs1, rs2, rd) +#define vis_and(rs1,rs2,rd) vis_dd2d(0x70, rs1, rs2, rd) +#define vis_ands(rs1,rs2,rd) vis_ss2s(0x71, rs1, rs2, rd) +#define vis_nand(rs1,rs2,rd) vis_dd2d(0x6e, rs1, rs2, rd) +#define vis_nands(rs1,rs2,rd) vis_ss2s(0x6f, rs1, rs2, rd) +#define vis_xor(rs1,rs2,rd) vis_dd2d(0x6c, rs1, rs2, rd) +#define vis_xors(rs1,rs2,rd) vis_ss2s(0x6d, rs1, rs2, rd) +#define vis_xnor(rs1,rs2,rd) vis_dd2d(0x72, rs1, rs2, rd) +#define vis_xnors(rs1,rs2,rd) vis_ss2s(0x73, rs1, rs2, rd) +#define vis_ornot1(rs1,rs2,rd) vis_dd2d(0x7a, rs1, rs2, rd) +#define vis_ornot1s(rs1,rs2,rd) vis_ss2s(0x7b, rs1, rs2, rd) +#define vis_ornot2(rs1,rs2,rd) vis_dd2d(0x76, rs1, rs2, rd) +#define vis_ornot2s(rs1,rs2,rd) vis_ss2s(0x77, rs1, rs2, rd) +#define vis_andnot1(rs1,rs2,rd) vis_dd2d(0x68, rs1, rs2, rd) +#define vis_andnot1s(rs1,rs2,rd) vis_ss2s(0x69, rs1, rs2, rd) +#define vis_andnot2(rs1,rs2,rd) vis_dd2d(0x64, rs1, rs2, rd) +#define vis_andnot2s(rs1,rs2,rd) vis_ss2s(0x65, rs1, rs2, rd) + +/* Pixel component distance. */ + +#define vis_pdist(rs1,rs2,rd) vis_dd2d(0x3e, rs1, rs2, rd) diff --git a/src/video_dec/libmpeg2/vlc.h b/src/video_dec/libmpeg2/vlc.h new file mode 100644 index 000000000..65de9a840 --- /dev/null +++ b/src/video_dec/libmpeg2/vlc.h @@ -0,0 +1,428 @@ +/* + * vlc.h + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#define GETWORD(bit_buf,shift,bit_ptr) \ +do { \ + bit_buf |= ((bit_ptr[0] << 8) | bit_ptr[1]) << (shift); \ + bit_ptr += 2; \ +} while (0) + +static inline void bitstream_init (picture_t * picture, uint8_t * start) +{ + picture->bitstream_buf = + (start[0] << 24) | (start[1] << 16) | (start[2] << 8) | start[3]; + picture->bitstream_ptr = start + 4; + picture->bitstream_bits = -16; +} + +/* make sure that there are at least 16 valid bits in bit_buf */ +#define NEEDBITS(bit_buf,bits,bit_ptr) \ +do { \ + if (bits > 0) { \ + GETWORD (bit_buf, bits, bit_ptr); \ + bits -= 16; \ + } \ +} while (0) + +/* remove num valid bits from bit_buf */ +#define DUMPBITS(bit_buf,bits,num) \ +do { \ + bit_buf <<= (num); \ + bits += (num); \ +} while (0) + +/* take num bits from the high part of bit_buf and zero extend them */ +#define UBITS(bit_buf,num) (((uint32_t)(bit_buf)) >> (32 - (num))) + +/* take num bits from the high part of bit_buf and sign extend them */ +#define SBITS(bit_buf,num) (((int32_t)(bit_buf)) >> (32 - (num))) + +typedef struct { + uint8_t modes; + uint8_t len; +} MBtab; + +typedef struct { + uint8_t delta; + uint8_t len; +} MVtab; + +typedef struct { + int8_t dmv; + uint8_t len; +} DMVtab; + +typedef struct { + uint8_t cbp; + uint8_t len; +} CBPtab; + +typedef struct { + uint8_t size; + uint8_t len; +} DCtab; + +typedef struct { + uint8_t run; + uint8_t level; + uint8_t len; +} DCTtab; + +typedef struct { + uint8_t mba; + uint8_t len; +} MBAtab; + + +#define INTRA MACROBLOCK_INTRA +#define QUANT MACROBLOCK_QUANT + +static const MBtab MB_I [] = { + {INTRA|QUANT, 2}, {INTRA, 1} +}; + +#define MC MACROBLOCK_MOTION_FORWARD +#define CODED MACROBLOCK_PATTERN + +static const MBtab MB_P [] = { + {INTRA|QUANT, 6}, {CODED|QUANT, 5}, {MC|CODED|QUANT, 5}, {INTRA, 5}, + {MC, 3}, {MC, 3}, {MC, 3}, {MC, 3}, + {CODED, 2}, {CODED, 2}, {CODED, 2}, {CODED, 2}, + {CODED, 2}, {CODED, 2}, {CODED, 2}, {CODED, 2}, + {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1}, + {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1}, + {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1}, + {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1}, {MC|CODED, 1} +}; + +#define FWD MACROBLOCK_MOTION_FORWARD +#define BWD MACROBLOCK_MOTION_BACKWARD +#define INTER MACROBLOCK_MOTION_FORWARD|MACROBLOCK_MOTION_BACKWARD + +static const MBtab MB_B [] = { + {0, 0}, {INTRA|QUANT, 6}, + {BWD|CODED|QUANT, 6}, {FWD|CODED|QUANT, 6}, + {INTER|CODED|QUANT, 5}, {INTER|CODED|QUANT, 5}, + {INTRA, 5}, {INTRA, 5}, + {FWD, 4}, {FWD, 4}, {FWD, 4}, {FWD, 4}, + {FWD|CODED, 4}, {FWD|CODED, 4}, {FWD|CODED, 4}, {FWD|CODED, 4}, + {BWD, 3}, {BWD, 3}, {BWD, 3}, {BWD, 3}, + {BWD, 3}, {BWD, 3}, {BWD, 3}, {BWD, 3}, + {BWD|CODED, 3}, {BWD|CODED, 3}, {BWD|CODED, 3}, {BWD|CODED, 3}, + {BWD|CODED, 3}, {BWD|CODED, 3}, {BWD|CODED, 3}, {BWD|CODED, 3}, + {INTER, 2}, {INTER, 2}, {INTER, 2}, {INTER, 2}, + {INTER, 2}, {INTER, 2}, {INTER, 2}, {INTER, 2}, + {INTER, 2}, {INTER, 2}, {INTER, 2}, {INTER, 2}, + {INTER, 2}, {INTER, 2}, {INTER, 2}, {INTER, 2}, + {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2}, + {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2}, + {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2}, + {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2}, {INTER|CODED, 2} +}; + +#undef INTRA +#undef QUANT +#undef MC +#undef CODED +#undef FWD +#undef BWD +#undef INTER + + +static const MVtab MV_4 [] = { + { 3, 6}, { 2, 4}, { 1, 3}, { 1, 3}, { 0, 2}, { 0, 2}, { 0, 2}, { 0, 2} +}; + +static const MVtab MV_10 [] = { + { 0,10}, { 0,10}, { 0,10}, { 0,10}, { 0,10}, { 0,10}, { 0,10}, { 0,10}, + { 0,10}, { 0,10}, { 0,10}, { 0,10}, {15,10}, {14,10}, {13,10}, {12,10}, + {11,10}, {10,10}, { 9, 9}, { 9, 9}, { 8, 9}, { 8, 9}, { 7, 9}, { 7, 9}, + { 6, 7}, { 6, 7}, { 6, 7}, { 6, 7}, { 6, 7}, { 6, 7}, { 6, 7}, { 6, 7}, + { 5, 7}, { 5, 7}, { 5, 7}, { 5, 7}, { 5, 7}, { 5, 7}, { 5, 7}, { 5, 7}, + { 4, 7}, { 4, 7}, { 4, 7}, { 4, 7}, { 4, 7}, { 4, 7}, { 4, 7}, { 4, 7} +}; + + +static const DMVtab DMV_2 [] = { + { 0, 1}, { 0, 1}, { 1, 2}, {-1, 2} +}; + + +static const CBPtab CBP_7 [] = { + {0x22, 7}, {0x12, 7}, {0x0a, 7}, {0x06, 7}, + {0x21, 7}, {0x11, 7}, {0x09, 7}, {0x05, 7}, + {0x3f, 6}, {0x3f, 6}, {0x03, 6}, {0x03, 6}, + {0x24, 6}, {0x24, 6}, {0x18, 6}, {0x18, 6}, + {0x3e, 5}, {0x3e, 5}, {0x3e, 5}, {0x3e, 5}, + {0x02, 5}, {0x02, 5}, {0x02, 5}, {0x02, 5}, + {0x3d, 5}, {0x3d, 5}, {0x3d, 5}, {0x3d, 5}, + {0x01, 5}, {0x01, 5}, {0x01, 5}, {0x01, 5}, + {0x38, 5}, {0x38, 5}, {0x38, 5}, {0x38, 5}, + {0x34, 5}, {0x34, 5}, {0x34, 5}, {0x34, 5}, + {0x2c, 5}, {0x2c, 5}, {0x2c, 5}, {0x2c, 5}, + {0x1c, 5}, {0x1c, 5}, {0x1c, 5}, {0x1c, 5}, + {0x28, 5}, {0x28, 5}, {0x28, 5}, {0x28, 5}, + {0x14, 5}, {0x14, 5}, {0x14, 5}, {0x14, 5}, + {0x30, 5}, {0x30, 5}, {0x30, 5}, {0x30, 5}, + {0x0c, 5}, {0x0c, 5}, {0x0c, 5}, {0x0c, 5}, + {0x20, 4}, {0x20, 4}, {0x20, 4}, {0x20, 4}, + {0x20, 4}, {0x20, 4}, {0x20, 4}, {0x20, 4}, + {0x10, 4}, {0x10, 4}, {0x10, 4}, {0x10, 4}, + {0x10, 4}, {0x10, 4}, {0x10, 4}, {0x10, 4}, + {0x08, 4}, {0x08, 4}, {0x08, 4}, {0x08, 4}, + {0x08, 4}, {0x08, 4}, {0x08, 4}, {0x08, 4}, + {0x04, 4}, {0x04, 4}, {0x04, 4}, {0x04, 4}, + {0x04, 4}, {0x04, 4}, {0x04, 4}, {0x04, 4}, + {0x3c, 3}, {0x3c, 3}, {0x3c, 3}, {0x3c, 3}, + {0x3c, 3}, {0x3c, 3}, {0x3c, 3}, {0x3c, 3}, + {0x3c, 3}, {0x3c, 3}, {0x3c, 3}, {0x3c, 3}, + {0x3c, 3}, {0x3c, 3}, {0x3c, 3}, {0x3c, 3} +}; + +static const CBPtab CBP_9 [] = { + {0, 0}, {0x00, 9}, {0x27, 9}, {0x1b, 9}, + {0x3b, 9}, {0x37, 9}, {0x2f, 9}, {0x1f, 9}, + {0x3a, 8}, {0x3a, 8}, {0x36, 8}, {0x36, 8}, + {0x2e, 8}, {0x2e, 8}, {0x1e, 8}, {0x1e, 8}, + {0x39, 8}, {0x39, 8}, {0x35, 8}, {0x35, 8}, + {0x2d, 8}, {0x2d, 8}, {0x1d, 8}, {0x1d, 8}, + {0x26, 8}, {0x26, 8}, {0x1a, 8}, {0x1a, 8}, + {0x25, 8}, {0x25, 8}, {0x19, 8}, {0x19, 8}, + {0x2b, 8}, {0x2b, 8}, {0x17, 8}, {0x17, 8}, + {0x33, 8}, {0x33, 8}, {0x0f, 8}, {0x0f, 8}, + {0x2a, 8}, {0x2a, 8}, {0x16, 8}, {0x16, 8}, + {0x32, 8}, {0x32, 8}, {0x0e, 8}, {0x0e, 8}, + {0x29, 8}, {0x29, 8}, {0x15, 8}, {0x15, 8}, + {0x31, 8}, {0x31, 8}, {0x0d, 8}, {0x0d, 8}, + {0x23, 8}, {0x23, 8}, {0x13, 8}, {0x13, 8}, + {0x0b, 8}, {0x0b, 8}, {0x07, 8}, {0x07, 8} +}; + + +static const DCtab DC_lum_5 [] = { + {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, + {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, + {0, 3}, {0, 3}, {0, 3}, {0, 3}, {3, 3}, {3, 3}, {3, 3}, {3, 3}, + {4, 3}, {4, 3}, {4, 3}, {4, 3}, {5, 4}, {5, 4}, {6, 5} +}; + +static const DCtab DC_chrom_5 [] = { + {0, 2}, {0, 2}, {0, 2}, {0, 2}, {0, 2}, {0, 2}, {0, 2}, {0, 2}, + {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, {1, 2}, + {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, {2, 2}, + {3, 3}, {3, 3}, {3, 3}, {3, 3}, {4, 4}, {4, 4}, {5, 5} +}; + +static const DCtab DC_long [] = { + {6, 5}, {6, 5}, {6, 5}, {6, 5}, {6, 5}, {6, 5}, { 6, 5}, { 6, 5}, + {6, 5}, {6, 5}, {6, 5}, {6, 5}, {6, 5}, {6, 5}, { 6, 5}, { 6, 5}, + {7, 6}, {7, 6}, {7, 6}, {7, 6}, {7, 6}, {7, 6}, { 7, 6}, { 7, 6}, + {8, 7}, {8, 7}, {8, 7}, {8, 7}, {9, 8}, {9, 8}, {10, 9}, {11, 9} +}; + + +static const DCTtab DCT_16 [] = { + {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, + {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, + {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, + {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, {129, 0, 0}, + { 2,18, 0}, { 2,17, 0}, { 2,16, 0}, { 2,15, 0}, + { 7, 3, 0}, { 17, 2, 0}, { 16, 2, 0}, { 15, 2, 0}, + { 14, 2, 0}, { 13, 2, 0}, { 12, 2, 0}, { 32, 1, 0}, + { 31, 1, 0}, { 30, 1, 0}, { 29, 1, 0}, { 28, 1, 0} +}; + +static const DCTtab DCT_15 [] = { + { 1,40,15}, { 1,39,15}, { 1,38,15}, { 1,37,15}, + { 1,36,15}, { 1,35,15}, { 1,34,15}, { 1,33,15}, + { 1,32,15}, { 2,14,15}, { 2,13,15}, { 2,12,15}, + { 2,11,15}, { 2,10,15}, { 2, 9,15}, { 2, 8,15}, + { 1,31,14}, { 1,31,14}, { 1,30,14}, { 1,30,14}, + { 1,29,14}, { 1,29,14}, { 1,28,14}, { 1,28,14}, + { 1,27,14}, { 1,27,14}, { 1,26,14}, { 1,26,14}, + { 1,25,14}, { 1,25,14}, { 1,24,14}, { 1,24,14}, + { 1,23,14}, { 1,23,14}, { 1,22,14}, { 1,22,14}, + { 1,21,14}, { 1,21,14}, { 1,20,14}, { 1,20,14}, + { 1,19,14}, { 1,19,14}, { 1,18,14}, { 1,18,14}, + { 1,17,14}, { 1,17,14}, { 1,16,14}, { 1,16,14} +}; + +static const DCTtab DCT_13 [] = { + { 11, 2,13}, { 10, 2,13}, { 6, 3,13}, { 4, 4,13}, + { 3, 5,13}, { 2, 7,13}, { 2, 6,13}, { 1,15,13}, + { 1,14,13}, { 1,13,13}, { 1,12,13}, { 27, 1,13}, + { 26, 1,13}, { 25, 1,13}, { 24, 1,13}, { 23, 1,13}, + { 1,11,12}, { 1,11,12}, { 9, 2,12}, { 9, 2,12}, + { 5, 3,12}, { 5, 3,12}, { 1,10,12}, { 1,10,12}, + { 3, 4,12}, { 3, 4,12}, { 8, 2,12}, { 8, 2,12}, + { 22, 1,12}, { 22, 1,12}, { 21, 1,12}, { 21, 1,12}, + { 1, 9,12}, { 1, 9,12}, { 20, 1,12}, { 20, 1,12}, + { 19, 1,12}, { 19, 1,12}, { 2, 5,12}, { 2, 5,12}, + { 4, 3,12}, { 4, 3,12}, { 1, 8,12}, { 1, 8,12}, + { 7, 2,12}, { 7, 2,12}, { 18, 1,12}, { 18, 1,12} +}; + +static const DCTtab DCT_B14_10 [] = { + { 17, 1,10}, { 6, 2,10}, { 1, 7,10}, { 3, 3,10}, + { 2, 4,10}, { 16, 1,10}, { 15, 1,10}, { 5, 2,10} +}; + +static const DCTtab DCT_B14_8 [] = { + { 65, 0, 6}, { 65, 0, 6}, { 65, 0, 6}, { 65, 0, 6}, + { 3, 2, 7}, { 3, 2, 7}, { 10, 1, 7}, { 10, 1, 7}, + { 1, 4, 7}, { 1, 4, 7}, { 9, 1, 7}, { 9, 1, 7}, + { 8, 1, 6}, { 8, 1, 6}, { 8, 1, 6}, { 8, 1, 6}, + { 7, 1, 6}, { 7, 1, 6}, { 7, 1, 6}, { 7, 1, 6}, + { 2, 2, 6}, { 2, 2, 6}, { 2, 2, 6}, { 2, 2, 6}, + { 6, 1, 6}, { 6, 1, 6}, { 6, 1, 6}, { 6, 1, 6}, + { 14, 1, 8}, { 1, 6, 8}, { 13, 1, 8}, { 12, 1, 8}, + { 4, 2, 8}, { 2, 3, 8}, { 1, 5, 8}, { 11, 1, 8} +}; + +static const DCTtab DCT_B14AC_5 [] = { + { 1, 3, 5}, { 5, 1, 5}, { 4, 1, 5}, + { 1, 2, 4}, { 1, 2, 4}, { 3, 1, 4}, { 3, 1, 4}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + {129, 0, 2}, {129, 0, 2}, {129, 0, 2}, {129, 0, 2}, + {129, 0, 2}, {129, 0, 2}, {129, 0, 2}, {129, 0, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2} +}; + +static const DCTtab DCT_B14DC_5 [] = { + { 1, 3, 5}, { 5, 1, 5}, { 4, 1, 5}, + { 1, 2, 4}, { 1, 2, 4}, { 3, 1, 4}, { 3, 1, 4}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1}, + { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1}, + { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1}, + { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1}, { 1, 1, 1} +}; + +static const DCTtab DCT_B15_10 [] = { + { 6, 2, 9}, { 6, 2, 9}, { 15, 1, 9}, { 15, 1, 9}, + { 3, 4,10}, { 17, 1,10}, { 16, 1, 9}, { 16, 1, 9} +}; + +static const DCTtab DCT_B15_8 [] = { + { 65, 0, 6}, { 65, 0, 6}, { 65, 0, 6}, { 65, 0, 6}, + { 8, 1, 7}, { 8, 1, 7}, { 9, 1, 7}, { 9, 1, 7}, + { 7, 1, 7}, { 7, 1, 7}, { 3, 2, 7}, { 3, 2, 7}, + { 1, 7, 6}, { 1, 7, 6}, { 1, 7, 6}, { 1, 7, 6}, + { 1, 6, 6}, { 1, 6, 6}, { 1, 6, 6}, { 1, 6, 6}, + { 5, 1, 6}, { 5, 1, 6}, { 5, 1, 6}, { 5, 1, 6}, + { 6, 1, 6}, { 6, 1, 6}, { 6, 1, 6}, { 6, 1, 6}, + { 2, 5, 8}, { 12, 1, 8}, { 1,11, 8}, { 1,10, 8}, + { 14, 1, 8}, { 13, 1, 8}, { 4, 2, 8}, { 2, 4, 8}, + { 3, 1, 5}, { 3, 1, 5}, { 3, 1, 5}, { 3, 1, 5}, + { 3, 1, 5}, { 3, 1, 5}, { 3, 1, 5}, { 3, 1, 5}, + { 2, 2, 5}, { 2, 2, 5}, { 2, 2, 5}, { 2, 2, 5}, + { 2, 2, 5}, { 2, 2, 5}, { 2, 2, 5}, { 2, 2, 5}, + { 4, 1, 5}, { 4, 1, 5}, { 4, 1, 5}, { 4, 1, 5}, + { 4, 1, 5}, { 4, 1, 5}, { 4, 1, 5}, { 4, 1, 5}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, { 2, 1, 3}, + {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, + {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, + {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, + {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, {129, 0, 4}, + { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, + { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, + { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, + { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, { 1, 3, 4}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, { 1, 1, 2}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, { 1, 2, 3}, + { 1, 4, 5}, { 1, 4, 5}, { 1, 4, 5}, { 1, 4, 5}, + { 1, 4, 5}, { 1, 4, 5}, { 1, 4, 5}, { 1, 4, 5}, + { 1, 5, 5}, { 1, 5, 5}, { 1, 5, 5}, { 1, 5, 5}, + { 1, 5, 5}, { 1, 5, 5}, { 1, 5, 5}, { 1, 5, 5}, + { 10, 1, 7}, { 10, 1, 7}, { 2, 3, 7}, { 2, 3, 7}, + { 11, 1, 7}, { 11, 1, 7}, { 1, 8, 7}, { 1, 8, 7}, + { 1, 9, 7}, { 1, 9, 7}, { 1,12, 8}, { 1,13, 8}, + { 3, 3, 8}, { 5, 2, 8}, { 1,14, 8}, { 1,15, 8} +}; + + +static const MBAtab MBA_5 [] = { + {6, 5}, {5, 5}, {4, 4}, {4, 4}, {3, 4}, {3, 4}, + {2, 3}, {2, 3}, {2, 3}, {2, 3}, {1, 3}, {1, 3}, {1, 3}, {1, 3}, + {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, + {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1}, {0, 1} +}; + +static const MBAtab MBA_11 [] = { + {32, 11}, {31, 11}, {30, 11}, {29, 11}, + {28, 11}, {27, 11}, {26, 11}, {25, 11}, + {24, 11}, {23, 11}, {22, 11}, {21, 11}, + {20, 10}, {20, 10}, {19, 10}, {19, 10}, + {18, 10}, {18, 10}, {17, 10}, {17, 10}, + {16, 10}, {16, 10}, {15, 10}, {15, 10}, + {14, 8}, {14, 8}, {14, 8}, {14, 8}, + {14, 8}, {14, 8}, {14, 8}, {14, 8}, + {13, 8}, {13, 8}, {13, 8}, {13, 8}, + {13, 8}, {13, 8}, {13, 8}, {13, 8}, + {12, 8}, {12, 8}, {12, 8}, {12, 8}, + {12, 8}, {12, 8}, {12, 8}, {12, 8}, + {11, 8}, {11, 8}, {11, 8}, {11, 8}, + {11, 8}, {11, 8}, {11, 8}, {11, 8}, + {10, 8}, {10, 8}, {10, 8}, {10, 8}, + {10, 8}, {10, 8}, {10, 8}, {10, 8}, + { 9, 8}, { 9, 8}, { 9, 8}, { 9, 8}, + { 9, 8}, { 9, 8}, { 9, 8}, { 9, 8}, + { 8, 7}, { 8, 7}, { 8, 7}, { 8, 7}, + { 8, 7}, { 8, 7}, { 8, 7}, { 8, 7}, + { 8, 7}, { 8, 7}, { 8, 7}, { 8, 7}, + { 8, 7}, { 8, 7}, { 8, 7}, { 8, 7}, + { 7, 7}, { 7, 7}, { 7, 7}, { 7, 7}, + { 7, 7}, { 7, 7}, { 7, 7}, { 7, 7}, + { 7, 7}, { 7, 7}, { 7, 7}, { 7, 7}, + { 7, 7}, { 7, 7}, { 7, 7}, { 7, 7} +}; diff --git a/src/video_dec/libmpeg2/xine_mpeg2_decoder.c b/src/video_dec/libmpeg2/xine_mpeg2_decoder.c new file mode 100644 index 000000000..0419efa61 --- /dev/null +++ b/src/video_dec/libmpeg2/xine_mpeg2_decoder.c @@ -0,0 +1,169 @@ +/* + * Copyright (C) 2000-2003 the xine project + * + * This file is part of xine, a free 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110, USA + * + * stuff needed to turn libmpeg2 into a xine decoder plugin + */ + + +#include <stdlib.h> +#include <string.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <fcntl.h> +#include <unistd.h> + +#define LOG_MODULE "mpeg2_decoder" +#define LOG_VERBOSE +/* +#define LOG +*/ + +#include <xine/xine_internal.h> +#include <xine/video_out.h> +#include "mpeg2.h" +#include "mpeg2_internal.h" +#include <xine/buffer.h> + +typedef struct { + video_decoder_class_t decoder_class; +} mpeg2_class_t; + + +typedef struct mpeg2dec_decoder_s { + video_decoder_t video_decoder; + mpeg2dec_t mpeg2; + mpeg2_class_t *class; + xine_stream_t *stream; +} mpeg2dec_decoder_t; + +static void mpeg2dec_decode_data (video_decoder_t *this_gen, buf_element_t *buf) { + mpeg2dec_decoder_t *this = (mpeg2dec_decoder_t *) this_gen; + + lprintf ("decode_data, flags=0x%08x ...\n", buf->decoder_flags); + + /* handle aspect hints from xine-dvdnav */ + if (buf->decoder_flags & BUF_FLAG_SPECIAL) { + if (buf->decoder_info[1] == BUF_SPECIAL_ASPECT) { + this->mpeg2.force_aspect = buf->decoder_info[2]; + if (buf->decoder_info[3] == 0x1 && buf->decoder_info[2] == 3) + /* letterboxing is denied, we have to do pan&scan */ + this->mpeg2.force_pan_scan = 1; + else + this->mpeg2.force_pan_scan = 0; + } + return; + } + + if (buf->decoder_flags & BUF_FLAG_PREVIEW) { + mpeg2_find_sequence_header (&this->mpeg2, buf->content, buf->content + buf->size); + } else { + + mpeg2_decode_data (&this->mpeg2, buf->content, buf->content + buf->size, + buf->pts); + } + + lprintf ("decode_data...done\n"); +} + +static void mpeg2dec_flush (video_decoder_t *this_gen) { + mpeg2dec_decoder_t *this = (mpeg2dec_decoder_t *) this_gen; + + lprintf ("flush\n"); + + mpeg2_flush (&this->mpeg2); +} + +static void mpeg2dec_reset (video_decoder_t *this_gen) { + mpeg2dec_decoder_t *this = (mpeg2dec_decoder_t *) this_gen; + + mpeg2_reset (&this->mpeg2); +} + +static void mpeg2dec_discontinuity (video_decoder_t *this_gen) { + mpeg2dec_decoder_t *this = (mpeg2dec_decoder_t *) this_gen; + + mpeg2_discontinuity (&this->mpeg2); +} + +static void mpeg2dec_dispose (video_decoder_t *this_gen) { + + mpeg2dec_decoder_t *this = (mpeg2dec_decoder_t *) this_gen; + + lprintf ("close\n"); + + mpeg2_close (&this->mpeg2); + + this->stream->video_out->close(this->stream->video_out, this->stream); + + free (this); +} + +static video_decoder_t *open_plugin (video_decoder_class_t *class_gen, xine_stream_t *stream) { + mpeg2dec_decoder_t *this ; + + this = (mpeg2dec_decoder_t *) calloc(1, sizeof(mpeg2dec_decoder_t)); + + this->video_decoder.decode_data = mpeg2dec_decode_data; + this->video_decoder.flush = mpeg2dec_flush; + this->video_decoder.reset = mpeg2dec_reset; + this->video_decoder.discontinuity = mpeg2dec_discontinuity; + this->video_decoder.dispose = mpeg2dec_dispose; + this->stream = stream; + this->class = (mpeg2_class_t *) class_gen; + this->mpeg2.stream = stream; + + mpeg2_init (&this->mpeg2, stream->video_out); + (stream->video_out->open) (stream->video_out, stream); + this->mpeg2.force_aspect = this->mpeg2.force_pan_scan = 0; + + return &this->video_decoder; +} + +/* + * mpeg2 plugin class + */ +static void *init_plugin (xine_t *xine, void *data) { + + mpeg2_class_t *this; + + this = (mpeg2_class_t *) calloc(1, sizeof(mpeg2_class_t)); + + this->decoder_class.open_plugin = open_plugin; + this->decoder_class.identifier = "mpeg2dec"; + this->decoder_class.description = N_("mpeg2 based video decoder plugin"); + this->decoder_class.dispose = default_video_decoder_class_dispose; + + return this; +} +/* + * exported plugin catalog entry + */ + +static const uint32_t supported_types[] = { BUF_VIDEO_MPEG, 0 }; + +static const decoder_info_t dec_info_mpeg2 = { + supported_types, /* supported types */ + 7 /* priority */ +}; + +const plugin_info_t xine_plugin_info[] EXPORTED = { + /* type, API, "name", version, special_info, init_function */ + { PLUGIN_VIDEO_DECODER, 19, "mpeg2", XINE_VERSION_CODE, &dec_info_mpeg2, init_plugin }, + { PLUGIN_NONE, 0, "", 0, NULL, NULL } +}; diff --git a/src/video_dec/libmpeg2/xvmc.h b/src/video_dec/libmpeg2/xvmc.h new file mode 100644 index 000000000..5d61bcf83 --- /dev/null +++ b/src/video_dec/libmpeg2/xvmc.h @@ -0,0 +1,32 @@ +/* + * mpeg2_internal.h + * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * 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 + */ + +#ifndef _XVMC_H +#include "libmpeg2_accel.h" + +/* slice_xvmc.c */ + +void mpeg2_xvmc_slice (mpeg2dec_accel_t *accel, picture_t * picture, int code, uint8_t * buffer); +void xvmc_setup_scan_ptable( void ); + +#endif diff --git a/src/video_dec/libmpeg2/xvmc_vld.h b/src/video_dec/libmpeg2/xvmc_vld.h new file mode 100644 index 000000000..561d1789d --- /dev/null +++ b/src/video_dec/libmpeg2/xvmc_vld.h @@ -0,0 +1,32 @@ +/* + * Copyright (c) 2004 The Unichrome project. All rights reserved. + * + * 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, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTIES OR REPRESENTATIONS; 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. + * + * + */ + +#ifndef _XVMC_VLD_H +#define _XVMC_VLD_H + +#include "accel_xvmc.h" +#include "xvmc.h" + +extern void mpeg2_xxmc_slice( mpeg2dec_accel_t *accel, picture_t *picture, + int code, uint8_t *buffer, uint32_t chunk_size, + uint8_t *chunk_buffer); +extern void mpeg2_xxmc_vld_frame_complete(mpeg2dec_accel_t *accel, picture_t *picture, int code); + + +#endif |