diff options
author | Yaowu Xu <yaowu@google.com> | 2013-09-17 16:31:46 -0700 |
---|---|---|
committer | Yaowu Xu <yaowu@google.com> | 2013-09-19 08:14:26 -0700 |
commit | 014acfa2af084034daeb01c06ac0c71a0ef39682 (patch) | |
tree | 580bf34ada8b5ddd7750114635bf9af78e4e7611 /vp9/encoder | |
parent | 1600707d350f572d6768a9adb57248ba44c29b82 (diff) | |
download | libvpx-014acfa2af084034daeb01c06ac0c71a0ef39682.tar libvpx-014acfa2af084034daeb01c06ac0c71a0ef39682.tar.gz libvpx-014acfa2af084034daeb01c06ac0c71a0ef39682.tar.bz2 libvpx-014acfa2af084034daeb01c06ac0c71a0ef39682.zip |
fix integer overflow errors
Change-Id: I76f440a917832c02d7a727697b225bac66b99f56
Diffstat (limited to 'vp9/encoder')
-rw-r--r-- | vp9/encoder/vp9_dct.c | 66 | ||||
-rw-r--r-- | vp9/encoder/vp9_encodeframe.c | 4 | ||||
-rw-r--r-- | vp9/encoder/vp9_firstpass.c | 4 | ||||
-rw-r--r-- | vp9/encoder/vp9_mcomp.c | 96 | ||||
-rw-r--r-- | vp9/encoder/vp9_onyx_int.h | 3 | ||||
-rw-r--r-- | vp9/encoder/vp9_rdopt.c | 18 |
6 files changed, 95 insertions, 96 deletions
diff --git a/vp9/encoder/vp9_dct.c b/vp9/encoder/vp9_dct.c index 4f4ad04b3..ca863931e 100644 --- a/vp9/encoder/vp9_dct.c +++ b/vp9/encoder/vp9_dct.c @@ -58,10 +58,10 @@ void vp9_short_fdct4x4_c(int16_t *input, int16_t *output, int pitch) { for (i = 0; i < 4; ++i) { // Load inputs. if (0 == pass) { - input[0] = in[0 * stride] << 4; - input[1] = in[1 * stride] << 4; - input[2] = in[2 * stride] << 4; - input[3] = in[3 * stride] << 4; + input[0] = in[0 * stride] * 16; + input[1] = in[1 * stride] * 16; + input[2] = in[2 * stride] * 16; + input[3] = in[3 * stride] * 16; if (i == 0 && input[0]) { input[0] += 1; } @@ -160,7 +160,7 @@ void vp9_short_fht4x4_c(int16_t *input, int16_t *output, // Columns for (i = 0; i < 4; ++i) { for (j = 0; j < 4; ++j) - temp_in[j] = input[j * pitch + i] << 4; + temp_in[j] = input[j * pitch + i] * 16; if (i == 0 && temp_in[0]) temp_in[0] += 1; ht.cols(temp_in, temp_out); @@ -250,14 +250,14 @@ void vp9_short_fdct8x8_c(int16_t *input, int16_t *final_output, int pitch) { int i; for (i = 0; i < 8; i++) { // stage 1 - s0 = (input[0 * stride] + input[7 * stride]) << 2; - s1 = (input[1 * stride] + input[6 * stride]) << 2; - s2 = (input[2 * stride] + input[5 * stride]) << 2; - s3 = (input[3 * stride] + input[4 * stride]) << 2; - s4 = (input[3 * stride] - input[4 * stride]) << 2; - s5 = (input[2 * stride] - input[5 * stride]) << 2; - s6 = (input[1 * stride] - input[6 * stride]) << 2; - s7 = (input[0 * stride] - input[7 * stride]) << 2; + s0 = (input[0 * stride] + input[7 * stride]) * 4; + s1 = (input[1 * stride] + input[6 * stride]) * 4; + s2 = (input[2 * stride] + input[5 * stride]) * 4; + s3 = (input[3 * stride] + input[4 * stride]) * 4; + s4 = (input[3 * stride] - input[4 * stride]) * 4; + s5 = (input[2 * stride] - input[5 * stride]) * 4; + s6 = (input[1 * stride] - input[6 * stride]) * 4; + s7 = (input[0 * stride] - input[7 * stride]) * 4; // fdct4_1d(step, step); x0 = s0 + s3; @@ -331,23 +331,23 @@ void vp9_short_fdct16x16_c(int16_t *input, int16_t *output, int pitch) { for (i = 0; i < 16; i++) { if (0 == pass) { // Calculate input for the first 8 results. - input[0] = (in[0 * stride] + in[15 * stride]) << 2; - input[1] = (in[1 * stride] + in[14 * stride]) << 2; - input[2] = (in[2 * stride] + in[13 * stride]) << 2; - input[3] = (in[3 * stride] + in[12 * stride]) << 2; - input[4] = (in[4 * stride] + in[11 * stride]) << 2; - input[5] = (in[5 * stride] + in[10 * stride]) << 2; - input[6] = (in[6 * stride] + in[ 9 * stride]) << 2; - input[7] = (in[7 * stride] + in[ 8 * stride]) << 2; + input[0] = (in[0 * stride] + in[15 * stride]) * 4; + input[1] = (in[1 * stride] + in[14 * stride]) * 4; + input[2] = (in[2 * stride] + in[13 * stride]) * 4; + input[3] = (in[3 * stride] + in[12 * stride]) * 4; + input[4] = (in[4 * stride] + in[11 * stride]) * 4; + input[5] = (in[5 * stride] + in[10 * stride]) * 4; + input[6] = (in[6 * stride] + in[ 9 * stride]) * 4; + input[7] = (in[7 * stride] + in[ 8 * stride]) * 4; // Calculate input for the next 8 results. - step1[0] = (in[7 * stride] - in[ 8 * stride]) << 2; - step1[1] = (in[6 * stride] - in[ 9 * stride]) << 2; - step1[2] = (in[5 * stride] - in[10 * stride]) << 2; - step1[3] = (in[4 * stride] - in[11 * stride]) << 2; - step1[4] = (in[3 * stride] - in[12 * stride]) << 2; - step1[5] = (in[2 * stride] - in[13 * stride]) << 2; - step1[6] = (in[1 * stride] - in[14 * stride]) << 2; - step1[7] = (in[0 * stride] - in[15 * stride]) << 2; + step1[0] = (in[7 * stride] - in[ 8 * stride]) * 4; + step1[1] = (in[6 * stride] - in[ 9 * stride]) * 4; + step1[2] = (in[5 * stride] - in[10 * stride]) * 4; + step1[3] = (in[4 * stride] - in[11 * stride]) * 4; + step1[4] = (in[3 * stride] - in[12 * stride]) * 4; + step1[5] = (in[2 * stride] - in[13 * stride]) * 4; + step1[6] = (in[1 * stride] - in[14 * stride]) * 4; + step1[7] = (in[0 * stride] - in[15 * stride]) * 4; } else { // Calculate input for the first 8 results. input[0] = ((in[0 * 16] + 1) >> 2) + ((in[15 * 16] + 1) >> 2); @@ -575,7 +575,7 @@ void vp9_short_fht8x8_c(int16_t *input, int16_t *output, // Columns for (i = 0; i < 8; ++i) { for (j = 0; j < 8; ++j) - temp_in[j] = input[j * pitch + i] << 2; + temp_in[j] = input[j * pitch + i] * 4; ht.cols(temp_in, temp_out); for (j = 0; j < 8; ++j) outptr[j * 8 + i] = temp_out[j]; @@ -975,7 +975,7 @@ void vp9_short_fht16x16_c(int16_t *input, int16_t *output, // Columns for (i = 0; i < 16; ++i) { for (j = 0; j < 16; ++j) - temp_in[j] = input[j * pitch + i] << 2; + temp_in[j] = input[j * pitch + i] * 4; ht.cols(temp_in, temp_out); for (j = 0; j < 16; ++j) outptr[j * 16 + i] = (temp_out[j] + 1 + (temp_out[j] < 0)) >> 2; @@ -1335,7 +1335,7 @@ void vp9_short_fdct32x32_c(int16_t *input, int16_t *out, int pitch) { for (i = 0; i < 32; ++i) { int temp_in[32], temp_out[32]; for (j = 0; j < 32; ++j) - temp_in[j] = input[j * shortpitch + i] << 2; + temp_in[j] = input[j * shortpitch + i] * 4; dct32_1d(temp_in, temp_out, 0); for (j = 0; j < 32; ++j) output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2; @@ -1364,7 +1364,7 @@ void vp9_short_fdct32x32_rd_c(int16_t *input, int16_t *out, int pitch) { for (i = 0; i < 32; ++i) { int temp_in[32], temp_out[32]; for (j = 0; j < 32; ++j) - temp_in[j] = input[j * shortpitch + i] << 2; + temp_in[j] = input[j * shortpitch + i] * 4; dct32_1d(temp_in, temp_out, 0); for (j = 0; j < 32; ++j) // TODO(cd): see quality impact of only doing diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c index 3b92a3905..ee938bda9 100644 --- a/vp9/encoder/vp9_encodeframe.c +++ b/vp9/encoder/vp9_encodeframe.c @@ -435,7 +435,8 @@ static void update_state(VP9_COMP *cpi, PICK_MODE_CONTEXT *ctx, best_second_mv.as_int = ctx->second_best_ref_mv.as_int; if (mbmi->mode == NEWMV) { best_mv.as_int = mbmi->ref_mvs[rf1][0].as_int; - best_second_mv.as_int = mbmi->ref_mvs[rf2][0].as_int; + if (rf2 > 0) + best_second_mv.as_int = mbmi->ref_mvs[rf2][0].as_int; } mbmi->best_mv.as_int = best_mv.as_int; mbmi->best_second_mv.as_int = best_second_mv.as_int; @@ -2627,7 +2628,6 @@ void vp9_encode_frame(VP9_COMP *cpi) { } else { encode_frame_internal(cpi); } - } static void sum_intra_stats(VP9_COMP *cpi, const MODE_INFO *mi) { diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c index 9cf7b8348..ad7e6d821 100644 --- a/vp9/encoder/vp9_firstpass.c +++ b/vp9/encoder/vp9_firstpass.c @@ -660,8 +660,8 @@ void vp9_first_pass(VP9_COMP *cpi) { neutral_count++; } - mv.as_mv.row <<= 3; - mv.as_mv.col <<= 3; + mv.as_mv.row *= 8; + mv.as_mv.col *= 8; this_error = motion_error; vp9_set_mbmode_and_mvs(x, NEWMV, &mv); xd->this_mi->mbmi.tx_size = TX_4X4; diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c index 8f340d2af..ad8c8999a 100644 --- a/vp9/encoder/vp9_mcomp.c +++ b/vp9/encoder/vp9_mcomp.c @@ -339,13 +339,13 @@ int vp9_find_best_sub_pixel_iterative(MACROBLOCK *x, int rr = ref_mv->as_mv.row; int rc = ref_mv->as_mv.col; - int br = bestmv->as_mv.row << 3; - int bc = bestmv->as_mv.col << 3; + int br = bestmv->as_mv.row * 8; + int bc = bestmv->as_mv.col * 8; int hstep = 4; - const int minc = MAX(x->mv_col_min << 3, ref_mv->as_mv.col - MV_MAX); - const int maxc = MIN(x->mv_col_max << 3, ref_mv->as_mv.col + MV_MAX); - const int minr = MAX(x->mv_row_min << 3, ref_mv->as_mv.row - MV_MAX); - const int maxr = MIN(x->mv_row_max << 3, ref_mv->as_mv.row + MV_MAX); + const int minc = MAX(x->mv_col_min * 8, ref_mv->as_mv.col - MV_MAX); + const int maxc = MIN(x->mv_col_max * 8, ref_mv->as_mv.col + MV_MAX); + const int minr = MAX(x->mv_row_min * 8, ref_mv->as_mv.row - MV_MAX); + const int maxr = MIN(x->mv_row_max * 8, ref_mv->as_mv.row + MV_MAX); int tr = br; int tc = bc; @@ -436,20 +436,20 @@ int vp9_find_best_sub_pixel_tree(MACROBLOCK *x, int rr = ref_mv->as_mv.row; int rc = ref_mv->as_mv.col; - int br = bestmv->as_mv.row << 3; - int bc = bestmv->as_mv.col << 3; + int br = bestmv->as_mv.row * 8; + int bc = bestmv->as_mv.col * 8; int hstep = 4; - const int minc = MAX(x->mv_col_min << 3, ref_mv->as_mv.col - MV_MAX); - const int maxc = MIN(x->mv_col_max << 3, ref_mv->as_mv.col + MV_MAX); - const int minr = MAX(x->mv_row_min << 3, ref_mv->as_mv.row - MV_MAX); - const int maxr = MIN(x->mv_row_max << 3, ref_mv->as_mv.row + MV_MAX); + const int minc = MAX(x->mv_col_min * 8, ref_mv->as_mv.col - MV_MAX); + const int maxc = MIN(x->mv_col_max * 8, ref_mv->as_mv.col + MV_MAX); + const int minr = MAX(x->mv_row_min * 8, ref_mv->as_mv.row - MV_MAX); + const int maxr = MIN(x->mv_row_max * 8, ref_mv->as_mv.row + MV_MAX); int tr = br; int tc = bc; // central mv - bestmv->as_mv.row <<= 3; - bestmv->as_mv.col <<= 3; + bestmv->as_mv.row *= 8; + bestmv->as_mv.col *= 8; // calculate central point error besterr = vfp->vf(y, y_stride, z, src_stride, sse1); @@ -532,20 +532,20 @@ int vp9_find_best_sub_pixel_comp_iterative(MACROBLOCK *x, int rr = ref_mv->as_mv.row; int rc = ref_mv->as_mv.col; - int br = bestmv->as_mv.row << 3; - int bc = bestmv->as_mv.col << 3; + int br = bestmv->as_mv.row * 8; + int bc = bestmv->as_mv.col * 8; int hstep = 4; - const int minc = MAX(x->mv_col_min << 3, ref_mv->as_mv.col - MV_MAX); - const int maxc = MIN(x->mv_col_max << 3, ref_mv->as_mv.col + MV_MAX); - const int minr = MAX(x->mv_row_min << 3, ref_mv->as_mv.row - MV_MAX); - const int maxr = MIN(x->mv_row_max << 3, ref_mv->as_mv.row + MV_MAX); + const int minc = MAX(x->mv_col_min * 8, ref_mv->as_mv.col - MV_MAX); + const int maxc = MIN(x->mv_col_max * 8, ref_mv->as_mv.col + MV_MAX); + const int minr = MAX(x->mv_row_min * 8, ref_mv->as_mv.row - MV_MAX); + const int maxr = MIN(x->mv_row_max * 8, ref_mv->as_mv.row + MV_MAX); int tr = br; int tc = bc; // central mv - bestmv->as_mv.row <<= 3; - bestmv->as_mv.col <<= 3; + bestmv->as_mv.row *= 8; + bestmv->as_mv.col *= 8; // calculate central point error // TODO(yunqingwang): central pointer error was already calculated in full- @@ -634,20 +634,20 @@ int vp9_find_best_sub_pixel_comp_tree(MACROBLOCK *x, int rr = ref_mv->as_mv.row; int rc = ref_mv->as_mv.col; - int br = bestmv->as_mv.row << 3; - int bc = bestmv->as_mv.col << 3; + int br = bestmv->as_mv.row * 8; + int bc = bestmv->as_mv.col * 8; int hstep = 4; - const int minc = MAX(x->mv_col_min << 3, ref_mv->as_mv.col - MV_MAX); - const int maxc = MIN(x->mv_col_max << 3, ref_mv->as_mv.col + MV_MAX); - const int minr = MAX(x->mv_row_min << 3, ref_mv->as_mv.row - MV_MAX); - const int maxr = MIN(x->mv_row_max << 3, ref_mv->as_mv.row + MV_MAX); + const int minc = MAX(x->mv_col_min * 8, ref_mv->as_mv.col - MV_MAX); + const int maxc = MIN(x->mv_col_max * 8, ref_mv->as_mv.col + MV_MAX); + const int minr = MAX(x->mv_row_min * 8, ref_mv->as_mv.row - MV_MAX); + const int maxr = MIN(x->mv_row_max * 8, ref_mv->as_mv.row + MV_MAX); int tr = br; int tc = bc; // central mv - bestmv->as_mv.row <<= 3; - bestmv->as_mv.col <<= 3; + bestmv->as_mv.row *= 8; + bestmv->as_mv.col *= 8; // calculate central point error // TODO(yunqingwang): central pointer error was already calculated in full- @@ -980,8 +980,8 @@ static int vp9_pattern_search(MACROBLOCK *x, this_offset = base_offset + (best_mv->as_mv.row * (in_what_stride)) + best_mv->as_mv.col; - this_mv.as_mv.row = best_mv->as_mv.row << 3; - this_mv.as_mv.col = best_mv->as_mv.col << 3; + this_mv.as_mv.row = best_mv->as_mv.row * 8; + this_mv.as_mv.col = best_mv->as_mv.col * 8; if (bestsad == INT_MAX) return INT_MAX; return @@ -1243,8 +1243,8 @@ int vp9_diamond_search_sad_c(MACROBLOCK *x, (*num00)++; } - this_mv.as_mv.row = best_mv->as_mv.row << 3; - this_mv.as_mv.col = best_mv->as_mv.col << 3; + this_mv.as_mv.row = best_mv->as_mv.row * 8; + this_mv.as_mv.col = best_mv->as_mv.col * 8; if (bestsad == INT_MAX) return INT_MAX; @@ -1416,8 +1416,8 @@ int vp9_diamond_search_sadx4(MACROBLOCK *x, (*num00)++; } - this_mv.as_mv.row = best_mv->as_mv.row << 3; - this_mv.as_mv.col = best_mv->as_mv.col << 3; + this_mv.as_mv.row = best_mv->as_mv.row * 8; + this_mv.as_mv.col = best_mv->as_mv.col * 8; if (bestsad == INT_MAX) return INT_MAX; @@ -1567,8 +1567,8 @@ int vp9_full_search_sad_c(MACROBLOCK *x, int_mv *ref_mv, } } - this_mv.as_mv.row = best_mv->as_mv.row << 3; - this_mv.as_mv.col = best_mv->as_mv.col << 3; + this_mv.as_mv.row = best_mv->as_mv.row * 8; + this_mv.as_mv.col = best_mv->as_mv.col * 8; if (bestsad < INT_MAX) return @@ -1688,8 +1688,8 @@ int vp9_full_search_sadx3(MACROBLOCK *x, int_mv *ref_mv, } - this_mv.as_mv.row = best_mv->as_mv.row << 3; - this_mv.as_mv.col = best_mv->as_mv.col << 3; + this_mv.as_mv.row = best_mv->as_mv.row * 8; + this_mv.as_mv.col = best_mv->as_mv.col * 8; if (bestsad < INT_MAX) return @@ -1836,8 +1836,8 @@ int vp9_full_search_sadx8(MACROBLOCK *x, int_mv *ref_mv, } } - this_mv.as_mv.row = best_mv->as_mv.row << 3; - this_mv.as_mv.col = best_mv->as_mv.col << 3; + this_mv.as_mv.row = best_mv->as_mv.row * 8; + this_mv.as_mv.col = best_mv->as_mv.col * 8; if (bestsad < INT_MAX) return @@ -1912,8 +1912,8 @@ int vp9_refining_search_sad_c(MACROBLOCK *x, } } - this_mv.as_mv.row = ref_mv->as_mv.row << 3; - this_mv.as_mv.col = ref_mv->as_mv.col << 3; + this_mv.as_mv.row = ref_mv->as_mv.row * 8; + this_mv.as_mv.col = ref_mv->as_mv.col * 8; if (bestsad < INT_MAX) return @@ -2018,8 +2018,8 @@ int vp9_refining_search_sadx4(MACROBLOCK *x, } } - this_mv.as_mv.row = ref_mv->as_mv.row << 3; - this_mv.as_mv.col = ref_mv->as_mv.col << 3; + this_mv.as_mv.row = ref_mv->as_mv.row * 8; + this_mv.as_mv.col = ref_mv->as_mv.col * 8; if (bestsad < INT_MAX) return @@ -2109,8 +2109,8 @@ int vp9_refining_search_8p_c(MACROBLOCK *x, } } - this_mv.as_mv.row = ref_mv->as_mv.row << 3; - this_mv.as_mv.col = ref_mv->as_mv.col << 3; + this_mv.as_mv.row = ref_mv->as_mv.row * 8; + this_mv.as_mv.col = ref_mv->as_mv.col * 8; if (bestsad < INT_MAX) { // FIXME(rbultje, yunqing): add full-pixel averaging variance functions diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h index 92edf49ab..9b20dafde 100644 --- a/vp9/encoder/vp9_onyx_int.h +++ b/vp9/encoder/vp9_onyx_int.h @@ -392,8 +392,7 @@ typedef struct VP9_COMP { int rd_thresh_freq_fact[BLOCK_SIZES][MAX_MODES]; int64_t rd_comp_pred_diff[NB_PREDICTION_TYPES]; - // FIXME(rbultje) int64_t? - int rd_prediction_type_threshes[4][NB_PREDICTION_TYPES]; + int64_t rd_prediction_type_threshes[4][NB_PREDICTION_TYPES]; unsigned int intra_inter_count[INTRA_INTER_CONTEXTS][2]; unsigned int comp_inter_count[COMP_INTER_CONTEXTS][2]; unsigned int single_ref_count[REF_CONTEXTS][2][2]; diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c index 64185fcfa..470379c47 100644 --- a/vp9/encoder/vp9_rdopt.c +++ b/vp9/encoder/vp9_rdopt.c @@ -269,7 +269,7 @@ void vp9_initialize_rd_consts(VP9_COMP *cpi, int qindex) { cpi->mb.inter_mode_cost[i][m - NEARESTMV] = cost_token(vp9_inter_mode_tree, cpi->common.fc.inter_mode_probs[i], - vp9_inter_mode_encodings - NEARESTMV + m); + vp9_inter_mode_encodings + (m - NEARESTMV)); } } } @@ -1683,17 +1683,17 @@ static void rd_check_segment_txsize(VP9_COMP *cpi, MACROBLOCK *x, i = idy * 2 + idx; frame_mv[ZEROMV][mbmi->ref_frame[0]].as_int = 0; - frame_mv[ZEROMV][mbmi->ref_frame[1]].as_int = 0; vp9_append_sub8x8_mvs_for_idx(&cpi->common, &x->e_mbd, &frame_mv[NEARESTMV][mbmi->ref_frame[0]], &frame_mv[NEARMV][mbmi->ref_frame[0]], i, 0, mi_row, mi_col); - if (has_second_rf) + if (has_second_rf) { + frame_mv[ZEROMV][mbmi->ref_frame[1]].as_int = 0; vp9_append_sub8x8_mvs_for_idx(&cpi->common, &x->e_mbd, - &frame_mv[NEARESTMV][mbmi->ref_frame[1]], - &frame_mv[NEARMV][mbmi->ref_frame[1]], - i, 1, mi_row, mi_col); - + &frame_mv[NEARESTMV][mbmi->ref_frame[1]], + &frame_mv[NEARMV][mbmi->ref_frame[1]], + i, 1, mi_row, mi_col); + } // search for the best motion vector on this segment for (this_mode = NEARESTMV; this_mode <= NEWMV; ++this_mode) { const struct buf_2d orig_src = x->plane[0].src; @@ -3262,8 +3262,8 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x, continue; // Test best rd so far against threshold for trying this mode. - if ((best_rd < ((cpi->rd_threshes[bsize][mode_index] * - cpi->rd_thresh_freq_fact[bsize][mode_index]) >> 5)) || + if ((best_rd < ((int64_t)cpi->rd_threshes[bsize][mode_index] * + cpi->rd_thresh_freq_fact[bsize][mode_index] >> 5)) || cpi->rd_threshes[bsize][mode_index] == INT_MAX) continue; |