--- decoder.c 2002/03/26 11:16:08 1.5 +++ decoder.c 2002/04/28 22:12:45 1.12 @@ -1,46 +1,53 @@ /************************************************************************** * - * XVID MPEG-4 VIDEO CODEC - * decoder main + * XVID MPEG-4 VIDEO CODEC + * - Decoder main module - * - * This program is an implementation of a part of one or more MPEG-4 - * Video tools as specified in ISO/IEC 14496-2 standard. Those intending - * to use this software module in hardware or software products are - * advised that its use may infringe existing patents or copyrights, and - * any such use would be at such party's own risk. The original - * developer of this software module and his/her company, and subsequent - * editors and their companies, will have no liability for use of this - * software or modifications or derivatives thereof. + * This program is an implementation of a part of one or more MPEG-4 + * Video tools as specified in ISO/IEC 14496-2 standard. Those intending + * to use this software module in hardware or software products are + * advised that its use may infringe existing patents or copyrights, and + * any such use would be at such party's own risk. The original + * developer of this software module and his/her company, and subsequent + * editors and their companies, will have no liability for use of this + * software or modifications or derivatives thereof. * - * This program is xvid_free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the xvid_free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the xvid_free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * 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 * *************************************************************************/ /************************************************************************** * - * History: + * History: * + * 22.04.2002 add some B-frame decode support; chenm001 + * 29.03.2002 interlacing fix - compensated block wasn't being used when + * reconstructing blocks, thus artifacts + * interlacing speedup - used transfers to re-interlace + * interlaced decoding should be as fast as progressive now * 26.03.2002 interlacing support - moved transfers outside decode loop - * 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block - * 22.12.2001 block based interpolation - * 01.12.2001 inital version; (c)2001 peter ross + * 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block + * 22.12.2001 lock based interpolation + * 01.12.2001 inital version; (c)2001 peter ross + * + * $Id: decoder.c,v 1.12 2002/04/28 22:12:45 edgomez Exp $ * *************************************************************************/ #include -#include // memset +#include #include "xvid.h" #include "portab.h" @@ -55,7 +62,6 @@ #include "dct/fdct.h" #include "utils/mem_transfer.h" #include "image/interpolate8x8.h" -#include "utils/mbfunctions.h" #include "bitstream/mbcoding.h" #include "prediction/mbprediction.h" @@ -92,9 +98,18 @@ return XVID_ERR_MEMORY; } - if (image_create(&dec->refn, dec->edged_width, dec->edged_height)) + if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) + { + image_destroy(&dec->cur, dec->edged_width, dec->edged_height); + xvid_free(dec); + return XVID_ERR_MEMORY; + } + // add by chenm001 + // for support B-frame to reference last 2 frame + if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { image_destroy(&dec->cur, dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); xvid_free(dec); return XVID_ERR_MEMORY; } @@ -108,7 +123,6 @@ } init_timer(); - create_vlc_tables(); return XVID_ERR_OK; } @@ -117,12 +131,10 @@ int decoder_destroy(DECODER * dec) { xvid_free(dec->mbs); - image_destroy(&dec->refn, dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); image_destroy(&dec->cur, dec->edged_width, dec->edged_height); xvid_free(dec); - destroy_vlc_tables(); - write_timer(); return XVID_ERR_OK; } @@ -138,28 +150,31 @@ // decode an intra macroblock void decoder_mbintra(DECODER * dec, - MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - const uint32_t acpred_flag, - const uint32_t cbp, - Bitstream * bs, - const uint32_t quant, - const uint32_t intra_dc_threshold) + MACROBLOCK * pMB, + const uint32_t x_pos, + const uint32_t y_pos, + const uint32_t acpred_flag, + const uint32_t cbp, + Bitstream * bs, + const uint32_t quant, + const uint32_t intra_dc_threshold) { - CACHE_ALIGN int16_t block[6][64]; - CACHE_ALIGN int16_t data[6][64]; - const uint32_t stride = dec->edged_width; + DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); + DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); + + uint32_t stride = dec->edged_width; + uint32_t stride2 = stride / 2; + uint32_t next_block = stride * 8; uint32_t i; uint32_t iQuant = pMB->quant; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; - pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); - pU_Cur = dec->cur.u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); - pV_Cur = dec->cur.v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); + pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); + pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); + pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); - memset(block, 0, sizeof(block)); // clear + memset(block, 0, 6*64*sizeof(int16_t)); // clear for (i = 0; i < 6; i++) { @@ -168,7 +183,7 @@ int start_coeff; start_timer(); - predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, block[i], iQuant, iDcScaler, predictors); + predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors); if (!acpred_flag) { pMB->acpred_directions[i] = 0; @@ -188,7 +203,7 @@ BitstreamSkip(bs, 1); // marker } - block[i][0] = dc_dif; + block[i*64 + 0] = dc_dif; start_coeff = 1; } else @@ -199,44 +214,43 @@ start_timer(); if (cbp & (1 << (5-i))) // coded { - get_intra_block(bs, block[i], pMB->acpred_directions[i], start_coeff); + get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); } stop_coding_timer(); start_timer(); - add_acdc(pMB, i, block[i], iDcScaler, predictors); + add_acdc(pMB, i, &block[i*64], iDcScaler, predictors); stop_prediction_timer(); start_timer(); if (dec->quant_type == 0) { - dequant_intra(data[i], block[i], iQuant, iDcScaler); + dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); } else { - dequant4_intra(data[i], block[i], iQuant, iDcScaler); + dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); } stop_iquant_timer(); start_timer(); - idct(data[i]); + idct(&data[i*64]); stop_idct_timer(); } - start_timer(); if (dec->interlacing && pMB->field_dct) { - MBFieldToFrame(data); + next_block = stride; + stride *= 2; } - stop_interlacing_timer(); start_timer(); - transfer_16to8copy(pY_Cur, data[0], stride); - transfer_16to8copy(pY_Cur + 8, data[1], stride); - transfer_16to8copy(pY_Cur + 8 * stride, data[2], stride); - transfer_16to8copy(pY_Cur + 8 + 8 * stride, data[3], stride); - transfer_16to8copy(pU_Cur, data[4], stride / 2); - transfer_16to8copy(pV_Cur, data[5], stride / 2); + transfer_16to8copy(pY_Cur, &data[0*64], stride); + transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); + transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); + transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); + transfer_16to8copy(pU_Cur, &data[4*64], stride2); + transfer_16to8copy(pV_Cur, &data[5*64], stride2); stop_transfer_timer(); } @@ -247,34 +261,36 @@ #define SIGN(X) (((X)>0)?1:-1) #define ABS(X) (((X)>0)?(X):-(X)) static const uint32_t roundtab[16] = - { 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; +{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; // decode an inter macroblock void decoder_mbinter(DECODER * dec, - const MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - const uint32_t acpred_flag, - const uint32_t cbp, - Bitstream * bs, - const uint32_t quant, - const uint32_t rounding) + const MACROBLOCK * pMB, + const uint32_t x_pos, + const uint32_t y_pos, + const uint32_t acpred_flag, + const uint32_t cbp, + Bitstream * bs, + const uint32_t quant, + const uint32_t rounding) { - CACHE_ALIGN int16_t block[6][64]; - CACHE_ALIGN int16_t data[6][64]; - const uint32_t stride = dec->edged_width; - const uint32_t stride2 = dec->edged_width / 2; - uint32_t i; - uint32_t iQuant = pMB->quant; + DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); + DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); + + uint32_t stride = dec->edged_width; + uint32_t stride2 = stride / 2; + uint32_t next_block = stride * 8; + uint32_t i; + uint32_t iQuant = pMB->quant; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; int uv_dx, uv_dy; - pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); - pU_Cur = dec->cur.u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); - pV_Cur = dec->cur.v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); + pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); + pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); + pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { @@ -295,67 +311,67 @@ } start_timer(); - interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); - interpolate8x8_switch(dec->cur.u, dec->refn.u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); - interpolate8x8_switch(dec->cur.v, dec->refn.v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); + interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); + interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); stop_comp_timer(); for (i = 0; i < 6; i++) { if (cbp & (1 << (5-i))) // coded { - memset(block[i], 0, 64 * sizeof(int16_t)); // clear + memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear start_timer(); - get_inter_block(bs, block[i]); + get_inter_block(bs, &block[i*64]); stop_coding_timer(); start_timer(); if (dec->quant_type == 0) { - dequant_inter(data[i], block[i], iQuant); + dequant_inter(&data[i*64], &block[i*64], iQuant); } else { - dequant4_inter(data[i], block[i], iQuant); + dequant4_inter(&data[i*64], &block[i*64], iQuant); } stop_iquant_timer(); start_timer(); - idct(data[i]); + idct(&data[i*64]); stop_idct_timer(); } } - start_timer(); - if (pMB->field_dct) + if (dec->interlacing && pMB->field_dct) { - MBFieldToFrame(data); + next_block = stride; + stride *= 2; } - stop_interlacing_timer(); start_timer(); if (cbp & 32) - transfer_16to8add(pY_Cur, data[0], stride); + transfer_16to8add(pY_Cur, &data[0*64], stride); if (cbp & 16) - transfer_16to8add(pY_Cur + 8, data[1], stride); + transfer_16to8add(pY_Cur + 8, &data[1*64], stride); if (cbp & 8) - transfer_16to8add(pY_Cur + 8 * stride, data[2], stride); + transfer_16to8add(pY_Cur + next_block, &data[2*64], stride); if (cbp & 4) - transfer_16to8add(pY_Cur + 8 + 8 * stride, data[3], stride); + transfer_16to8add(pY_Cur + 8 + next_block, &data[3*64], stride); if (cbp & 2) - transfer_16to8add(pU_Cur, data[4], stride / 2); + transfer_16to8add(pU_Cur, &data[4*64], stride2); if (cbp & 1) - transfer_16to8add(pV_Cur, data[5], stride / 2); + transfer_16to8add(pV_Cur, &data[5*64], stride2); stop_transfer_timer(); } void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) { + uint32_t x, y; for (y = 0; y < dec->mb_height; y++) @@ -408,11 +424,13 @@ decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); } } + } void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) { + int scale_fac = 1 << (fcode - 1); int high = (32 * scale_fac) - 1; int low = ((-32) * scale_fac); @@ -462,12 +480,11 @@ void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) { + uint32_t x, y; - image_swap(&dec->cur, &dec->refn); - start_timer(); - image_setedges(&dec->refn, dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); + image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); stop_edges_timer(); for (y = 0; y < dec->mb_height; y++) @@ -583,28 +600,28 @@ start_timer(); transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), - dec->refn.y + (16*y)*dec->edged_width + (16*x), - dec->edged_width); + dec->refn[0].y + (16*y)*dec->edged_width + (16*x), + dec->edged_width); transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), - dec->refn.y + (16*y)*dec->edged_width + (16*x+8), - dec->edged_width); + dec->refn[0].y + (16*y)*dec->edged_width + (16*x+8), + dec->edged_width); transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), - dec->refn.y + (16*y+8)*dec->edged_width + (16*x), - dec->edged_width); + dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x), + dec->edged_width); transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), - dec->refn.y + (16*y+8)*dec->edged_width + (16*x+8), - dec->edged_width); + dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x+8), + dec->edged_width); transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), - dec->refn.u + (8*y)*dec->edged_width/2 + (8*x), - dec->edged_width/2); + dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), + dec->edged_width/2); transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), - dec->refn.v + (8*y)*dec->edged_width/2 + (8*x), - dec->edged_width/2); + dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), + dec->edged_width/2); stop_transfer_timer(); } @@ -614,17 +631,28 @@ int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) { + Bitstream bs; uint32_t rounding; uint32_t quant; uint32_t fcode; uint32_t intra_dc_threshold; + uint32_t vop_type; start_global_timer(); BitstreamInit(&bs, frame->bitstream, frame->length); - switch (BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold)) + // add by chenm001 + // for support B-frame to reference last 2 frame + vop_type=BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold); + + if (vop_type==I_VOP || vop_type==P_VOP){ + image_swap(&dec->refn[0], &dec->refn[1]); + image_swap(&dec->cur, &dec->refn[0]); + } + + switch (vop_type) { case P_VOP : decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); @@ -649,7 +677,7 @@ start_timer(); image_output(&dec->cur, dec->width, dec->height, dec->edged_width, - frame->image, frame->stride, frame->colorspace); + frame->image, frame->stride, frame->colorspace); stop_conv_timer(); emms(); @@ -657,4 +685,5 @@ stop_global_timer(); return XVID_ERR_OK; + }