diff options
Diffstat (limited to 'media/libstagefright/codecs/avc/enc/src/sad_halfpel.cpp')
| -rw-r--r-- | media/libstagefright/codecs/avc/enc/src/sad_halfpel.cpp | 629 |
1 files changed, 629 insertions, 0 deletions
diff --git a/media/libstagefright/codecs/avc/enc/src/sad_halfpel.cpp b/media/libstagefright/codecs/avc/enc/src/sad_halfpel.cpp new file mode 100644 index 0000000..faf2198 --- /dev/null +++ b/media/libstagefright/codecs/avc/enc/src/sad_halfpel.cpp @@ -0,0 +1,629 @@ +/* ------------------------------------------------------------------ + * Copyright (C) 1998-2009 PacketVideo + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either + * express or implied. + * See the License for the specific language governing permissions + * and limitations under the License. + * ------------------------------------------------------------------- + */ +/* contains +int AVCHalfPel1_SAD_MB(uint8 *ref,uint8 *blk,int dmin,int width,int ih,int jh) +int AVCHalfPel2_SAD_MB(uint8 *ref,uint8 *blk,int dmin,int width) +int AVCHalfPel1_SAD_Blk(uint8 *ref,uint8 *blk,int dmin,int width,int ih,int jh) +int AVCHalfPel2_SAD_Blk(uint8 *ref,uint8 *blk,int dmin,int width) + +int AVCSAD_MB_HalfPel_C(uint8 *ref,uint8 *blk,int dmin,int width,int rx,int xh,int yh,void *extra_info) +int AVCSAD_MB_HP_HTFM_Collect(uint8 *ref,uint8 *blk,int dmin,int width,int rx,int xh,int yh,void *extra_info) +int AVCSAD_MB_HP_HTFM(uint8 *ref,uint8 *blk,int dmin,int width,int rx,int xh,int yh,void *extra_info) +int AVCSAD_Blk_HalfPel_C(uint8 *ref,uint8 *blk,int dmin,int width,int rx,int xh,int yh,void *extra_info) +*/ + +#include "avcenc_lib.h" +#include "sad_halfpel_inline.h" + +#ifdef _SAD_STAT +uint32 num_sad_HP_MB = 0; +uint32 num_sad_HP_Blk = 0; +uint32 num_sad_HP_MB_call = 0; +uint32 num_sad_HP_Blk_call = 0; +#define NUM_SAD_HP_MB_CALL() num_sad_HP_MB_call++ +#define NUM_SAD_HP_MB() num_sad_HP_MB++ +#define NUM_SAD_HP_BLK_CALL() num_sad_HP_Blk_call++ +#define NUM_SAD_HP_BLK() num_sad_HP_Blk++ +#else +#define NUM_SAD_HP_MB_CALL() +#define NUM_SAD_HP_MB() +#define NUM_SAD_HP_BLK_CALL() +#define NUM_SAD_HP_BLK() +#endif + + + +/*=============================================================== + Function: SAD_MB_HalfPel + Date: 09/17/2000 + Purpose: Compute the SAD on the half-pel resolution + Input/Output: hmem is assumed to be a pointer to the starting + point of the search in the 33x33 matrix search region + Changes: + 11/7/00: implemented MMX + ===============================================================*/ +/*================================================================== + Function: AVCSAD_MB_HalfPel_C + Date: 04/30/2001 + Purpose: Compute SAD 16x16 between blk and ref in halfpel + resolution, + Changes: + ==================================================================*/ +/* One component is half-pel */ +int AVCSAD_MB_HalfPel_Cxhyh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + (void)(extra_info); + + int i, j; + int sad = 0; + uint8 *kk, *p1, *p2, *p3, *p4; +// int sumref=0; + int temp; + int rx = dmin_rx & 0xFFFF; + + NUM_SAD_HP_MB_CALL(); + + p1 = ref; + p2 = ref + 1; + p3 = ref + rx; + p4 = ref + rx + 1; + kk = blk; + + for (i = 0; i < 16; i++) + { + for (j = 0; j < 16; j++) + { + + temp = ((p1[j] + p2[j] + p3[j] + p4[j] + 2) >> 2) - *kk++; + sad += AVC_ABS(temp); + } + + NUM_SAD_HP_MB(); + + if (sad > (int)((uint32)dmin_rx >> 16)) + return sad; + + p1 += rx; + p3 += rx; + p2 += rx; + p4 += rx; + } + return sad; +} + +int AVCSAD_MB_HalfPel_Cyh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + (void)(extra_info); + + int i, j; + int sad = 0; + uint8 *kk, *p1, *p2; +// int sumref=0; + int temp; + int rx = dmin_rx & 0xFFFF; + + NUM_SAD_HP_MB_CALL(); + + p1 = ref; + p2 = ref + rx; /* either left/right or top/bottom pixel */ + kk = blk; + + for (i = 0; i < 16; i++) + { + for (j = 0; j < 16; j++) + { + + temp = ((p1[j] + p2[j] + 1) >> 1) - *kk++; + sad += AVC_ABS(temp); + } + + NUM_SAD_HP_MB(); + + if (sad > (int)((uint32)dmin_rx >> 16)) + return sad; + p1 += rx; + p2 += rx; + } + return sad; +} + +int AVCSAD_MB_HalfPel_Cxh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + (void)(extra_info); + + int i, j; + int sad = 0; + uint8 *kk, *p1; + int temp; + int rx = dmin_rx & 0xFFFF; + + NUM_SAD_HP_MB_CALL(); + + p1 = ref; + kk = blk; + + for (i = 0; i < 16; i++) + { + for (j = 0; j < 16; j++) + { + + temp = ((p1[j] + p1[j+1] + 1) >> 1) - *kk++; + sad += AVC_ABS(temp); + } + + NUM_SAD_HP_MB(); + + if (sad > (int)((uint32)dmin_rx >> 16)) + return sad; + p1 += rx; + } + return sad; +} + +#ifdef HTFM /* HTFM with uniform subsampling implementation, 2/28/01 */ + +//Checheck here +int AVCAVCSAD_MB_HP_HTFM_Collectxhyh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + int i, j; + int sad = 0; + uint8 *p1, *p2; + int rx = dmin_rx & 0xFFFF; + int refwx4 = rx << 2; + int saddata[16]; /* used when collecting flag (global) is on */ + int difmad, tmp, tmp2; + int madstar; + HTFM_Stat *htfm_stat = (HTFM_Stat*) extra_info; + int *abs_dif_mad_avg = &(htfm_stat->abs_dif_mad_avg); + UInt *countbreak = &(htfm_stat->countbreak); + int *offsetRef = htfm_stat->offsetRef; + uint32 cur_word; + + madstar = (uint32)dmin_rx >> 20; + + NUM_SAD_HP_MB_CALL(); + + blk -= 4; + + for (i = 0; i < 16; i++) /* 16 stages */ + { + p1 = ref + offsetRef[i]; + p2 = p1 + rx; + + j = 4;/* 4 lines */ + do + { + cur_word = *((uint32*)(blk += 4)); + tmp = p1[12] + p2[12]; + tmp2 = p1[13] + p2[13]; + tmp += tmp2; + tmp2 = (cur_word >> 24) & 0xFF; + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[8] + p2[8]; + tmp2 = p1[9] + p2[9]; + tmp += tmp2; + tmp2 = (cur_word >> 16) & 0xFF; + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[4] + p2[4]; + tmp2 = p1[5] + p2[5]; + tmp += tmp2; + tmp2 = (cur_word >> 8) & 0xFF; + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + tmp2 = p1[1] + p2[1]; + tmp = p1[0] + p2[0]; + p1 += refwx4; + p2 += refwx4; + tmp += tmp2; + tmp2 = (cur_word & 0xFF); + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + } + while (--j); + + NUM_SAD_HP_MB(); + + saddata[i] = sad; + + if (i > 0) + { + if (sad > ((uint32)dmin_rx >> 16)) + { + difmad = saddata[0] - ((saddata[1] + 1) >> 1); + (*abs_dif_mad_avg) += ((difmad > 0) ? difmad : -difmad); + (*countbreak)++; + return sad; + } + } + } + difmad = saddata[0] - ((saddata[1] + 1) >> 1); + (*abs_dif_mad_avg) += ((difmad > 0) ? difmad : -difmad); + (*countbreak)++; + + return sad; +} + +int AVCAVCSAD_MB_HP_HTFM_Collectyh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + int i, j; + int sad = 0; + uint8 *p1, *p2; + int rx = dmin_rx & 0xFFFF; + int refwx4 = rx << 2; + int saddata[16]; /* used when collecting flag (global) is on */ + int difmad, tmp, tmp2; + int madstar; + HTFM_Stat *htfm_stat = (HTFM_Stat*) extra_info; + int *abs_dif_mad_avg = &(htfm_stat->abs_dif_mad_avg); + UInt *countbreak = &(htfm_stat->countbreak); + int *offsetRef = htfm_stat->offsetRef; + uint32 cur_word; + + madstar = (uint32)dmin_rx >> 20; + + NUM_SAD_HP_MB_CALL(); + + blk -= 4; + + for (i = 0; i < 16; i++) /* 16 stages */ + { + p1 = ref + offsetRef[i]; + p2 = p1 + rx; + j = 4; + do + { + cur_word = *((uint32*)(blk += 4)); + tmp = p1[12]; + tmp2 = p2[12]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 24) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[8]; + tmp2 = p2[8]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 16) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[4]; + tmp2 = p2[4]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 8) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[0]; + p1 += refwx4; + tmp2 = p2[0]; + p2 += refwx4; + tmp++; + tmp2 += tmp; + tmp = (cur_word & 0xFF); + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + } + while (--j); + + NUM_SAD_HP_MB(); + + saddata[i] = sad; + + if (i > 0) + { + if (sad > ((uint32)dmin_rx >> 16)) + { + difmad = saddata[0] - ((saddata[1] + 1) >> 1); + (*abs_dif_mad_avg) += ((difmad > 0) ? difmad : -difmad); + (*countbreak)++; + return sad; + } + } + } + difmad = saddata[0] - ((saddata[1] + 1) >> 1); + (*abs_dif_mad_avg) += ((difmad > 0) ? difmad : -difmad); + (*countbreak)++; + + return sad; +} + +int AVCAVCSAD_MB_HP_HTFM_Collectxh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + int i, j; + int sad = 0; + uint8 *p1; + int rx = dmin_rx & 0xFFFF; + int refwx4 = rx << 2; + int saddata[16]; /* used when collecting flag (global) is on */ + int difmad, tmp, tmp2; + int madstar; + HTFM_Stat *htfm_stat = (HTFM_Stat*) extra_info; + int *abs_dif_mad_avg = &(htfm_stat->abs_dif_mad_avg); + UInt *countbreak = &(htfm_stat->countbreak); + int *offsetRef = htfm_stat->offsetRef; + uint32 cur_word; + + madstar = (uint32)dmin_rx >> 20; + + NUM_SAD_HP_MB_CALL(); + + blk -= 4; + + for (i = 0; i < 16; i++) /* 16 stages */ + { + p1 = ref + offsetRef[i]; + + j = 4; /* 4 lines */ + do + { + cur_word = *((uint32*)(blk += 4)); + tmp = p1[12]; + tmp2 = p1[13]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 24) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[8]; + tmp2 = p1[9]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 16) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[4]; + tmp2 = p1[5]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 8) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[0]; + tmp2 = p1[1]; + p1 += refwx4; + tmp++; + tmp2 += tmp; + tmp = (cur_word & 0xFF); + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + } + while (--j); + + NUM_SAD_HP_MB(); + + saddata[i] = sad; + + if (i > 0) + { + if (sad > ((uint32)dmin_rx >> 16)) + { + difmad = saddata[0] - ((saddata[1] + 1) >> 1); + (*abs_dif_mad_avg) += ((difmad > 0) ? difmad : -difmad); + (*countbreak)++; + return sad; + } + } + } + difmad = saddata[0] - ((saddata[1] + 1) >> 1); + (*abs_dif_mad_avg) += ((difmad > 0) ? difmad : -difmad); + (*countbreak)++; + + return sad; +} + +int AVCSAD_MB_HP_HTFMxhyh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + int i, j; + int sad = 0, tmp, tmp2; + uint8 *p1, *p2; + int rx = dmin_rx & 0xFFFF; + int refwx4 = rx << 2; + int sadstar = 0, madstar; + int *nrmlz_th = (int*) extra_info; + int *offsetRef = nrmlz_th + 32; + uint32 cur_word; + + madstar = (uint32)dmin_rx >> 20; + + NUM_SAD_HP_MB_CALL(); + + blk -= 4; + + for (i = 0; i < 16; i++) /* 16 stages */ + { + p1 = ref + offsetRef[i]; + p2 = p1 + rx; + + j = 4; /* 4 lines */ + do + { + cur_word = *((uint32*)(blk += 4)); + tmp = p1[12] + p2[12]; + tmp2 = p1[13] + p2[13]; + tmp += tmp2; + tmp2 = (cur_word >> 24) & 0xFF; + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[8] + p2[8]; + tmp2 = p1[9] + p2[9]; + tmp += tmp2; + tmp2 = (cur_word >> 16) & 0xFF; + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[4] + p2[4]; + tmp2 = p1[5] + p2[5]; + tmp += tmp2; + tmp2 = (cur_word >> 8) & 0xFF; + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + tmp2 = p1[1] + p2[1]; + tmp = p1[0] + p2[0]; + p1 += refwx4; + p2 += refwx4; + tmp += tmp2; + tmp2 = (cur_word & 0xFF); + tmp += 2; + sad = INTERP2_SUB_SAD(sad, tmp, tmp2);; + } + while (--j); + + NUM_SAD_HP_MB(); + + sadstar += madstar; + if (sad > sadstar - nrmlz_th[i] || sad > ((uint32)dmin_rx >> 16)) + { + return 65536; + } + } + + return sad; +} + +int AVCSAD_MB_HP_HTFMyh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + int i, j; + int sad = 0, tmp, tmp2; + uint8 *p1, *p2; + int rx = dmin_rx & 0xFFFF; + int refwx4 = rx << 2; + int sadstar = 0, madstar; + int *nrmlz_th = (int*) extra_info; + int *offsetRef = nrmlz_th + 32; + uint32 cur_word; + + madstar = (uint32)dmin_rx >> 20; + + NUM_SAD_HP_MB_CALL(); + + blk -= 4; + + for (i = 0; i < 16; i++) /* 16 stages */ + { + p1 = ref + offsetRef[i]; + p2 = p1 + rx; + j = 4; + do + { + cur_word = *((uint32*)(blk += 4)); + tmp = p1[12]; + tmp2 = p2[12]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 24) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[8]; + tmp2 = p2[8]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 16) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[4]; + tmp2 = p2[4]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 8) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[0]; + p1 += refwx4; + tmp2 = p2[0]; + p2 += refwx4; + tmp++; + tmp2 += tmp; + tmp = (cur_word & 0xFF); + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + } + while (--j); + + NUM_SAD_HP_MB(); + sadstar += madstar; + if (sad > sadstar - nrmlz_th[i] || sad > ((uint32)dmin_rx >> 16)) + { + return 65536; + } + } + + return sad; +} + +int AVCSAD_MB_HP_HTFMxh(uint8 *ref, uint8 *blk, int dmin_rx, void *extra_info) +{ + int i, j; + int sad = 0, tmp, tmp2; + uint8 *p1; + int rx = dmin_rx & 0xFFFF; + int refwx4 = rx << 2; + int sadstar = 0, madstar; + int *nrmlz_th = (int*) extra_info; + int *offsetRef = nrmlz_th + 32; + uint32 cur_word; + + madstar = (uint32)dmin_rx >> 20; + + NUM_SAD_HP_MB_CALL(); + + blk -= 4; + + for (i = 0; i < 16; i++) /* 16 stages */ + { + p1 = ref + offsetRef[i]; + + j = 4;/* 4 lines */ + do + { + cur_word = *((uint32*)(blk += 4)); + tmp = p1[12]; + tmp2 = p1[13]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 24) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[8]; + tmp2 = p1[9]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 16) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[4]; + tmp2 = p1[5]; + tmp++; + tmp2 += tmp; + tmp = (cur_word >> 8) & 0xFF; + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + tmp = p1[0]; + tmp2 = p1[1]; + p1 += refwx4; + tmp++; + tmp2 += tmp; + tmp = (cur_word & 0xFF); + sad = INTERP1_SUB_SAD(sad, tmp, tmp2);; + } + while (--j); + + NUM_SAD_HP_MB(); + + sadstar += madstar; + if (sad > sadstar - nrmlz_th[i] || sad > ((uint32)dmin_rx >> 16)) + { + return 65536; + } + } + + return sad; +} + +#endif /* HTFM */ + + + + + |