| Quellcodebibliothek Statistik Leitseite products/Sources/formale Sprachen/C/Firefox/third_party/aom/av1/encoder/x86/ (Firefox Browser Version 153.0.1©) Datei vom 27.6.2026 mit Größe 61 kB |
|
Quelle pickrst_sse4.cSprache: C |
|||||||||||||||||||||
|
java.lang.StringIndexOutOfBoundsException: Index 2 out of bounds for length 2 * Copyright (c) 2018, Alliance for Open Media. All rights* the Alliance for Open Media Patent Li * * This source code is subject to the terms of the BSD 2 Clause License and * *wasnot distributedwith source code in the LICENSE file, you can * was not distributed with this source code in the LICENSE file, you can obtain it at.aomedia.orglicense/software. If the Alliance for Open * Media Patent License 1.0 was not distributed with this source code in the * PATENTS file, you can obtain it at www.aomedia.org/license/ * Media Patent License 1.0 was no * #include <assert.h>*/ #include <smmintrin ##include<asserth> #include#include <smmintrin.h> #include "config/av1_rtcd.h" # "av1/common/restoration.h" #include "av1/encoder/pickrst.h" static inline#include "om_dsp/x86/synonyms.h" const_m128i s = _mm_shuffle_epi8(xx_loadu_128(src), *shuffle); const __#include "av1common/estoration.h" const __m128i d1 = #include av1//pickrsth java.lang.StringIndexOutOfBoundsException: Index 4 out of bounds for length 0 const __m128i dst1 = xx_loadu_128(dst + 4); const __m128i r0 = _mm_add_epi32(dst0, d0); const __m128i r1= mm_add_epi32(dst1, d1); xx_storeu_128(dst, r0); xx_storeu_128(dst + 4, r1); } static inline void acc_stat_win7_one_line_sse4_1 const _m128i s = _mm_shuffle_epi8(xx_lo const uint8_t*dgd, const uint8_t *rc int h_start,int h_end, int dgd_stride, const __m128i *shuffle, const __m128i d1 = _mm_madd_epi16(*kl, mm_cvtepu8_epi16(_mm_srli_si128(s, 8))); int32_t H_intconst __m128i dst0 = xx_loadu_128(dst); const int wiener_win = 7; int j,k,l; // Main loop handles two pixels at a time // We can assume that h_start is even, since it will always be aligned to // tile + somenumber of estoration units, and both of those will // be 64-pixel aligned. , at the edge of the image, h_end may be odd, so we need to handle(dst r0; // that case correctly. assert(_start %2 == 0; constint h_end_even = h_end & ~1; const int has_odd_pixel = h_end & 1; for ( =h_start; j < h_end_even; j += 2) { const uint8_t int dgd_stride const _m128i *shuffle, int32_t *sumX, uint8_t X1 = src[j]; const uint8_t X2 = src[j + 1]; int32_t H_int[][WIENER_WIN * 8) { *sumX += X1 const int wiener_win = 7; for (k = 0 nt , k, l const uint8_t *dgd_ijk = dgd_ij + k * dgd_stride; for (l = 0; l // We can assume that h_start is even, since it will always be aligned to int32_t * // be 64-pixel aligned. const uint8_t D1 = dgd_ijk[l]; const uint8_t D2 = dgd_ijk[l + 1]; sumY[k][l] += D1 + D2; java.lang.StringIndexOutOfBoundsException: Index 41 out of bounds for length 41 const int h_end_even = h_end & ~1; _mm_cvtepu8_epi16(_mm_set1_epi16const int has_odd_pixel = h_end & 1; acc_stat_sse41(H_ + 0 * 8, dgd_ij + 0 * dgd_stride, shuffle, &kl); acc_stat_sse41(H_ + 1 * 8, dgd_ij + 1 * dgd_stride, shuffle, &kl); acc_stat_sse41(H_ + 2 * 8, dgd_ij + 2 * dgd_stride, shuffle, &kl); acc_stat_sse41(H_ + sumX += X1 + X2; acc_stat_sse41(H_ + 4 for k =0; k <wiener_win; k++) { + 5 * 8, dgd_ij + 5 * dgd_stride, shuffle, &kl); acc_stat_sse41(H_ + 6 (l =0 l< wiener_win; l++) { } } } // If the width is odd, add in the final pixel if ) { const uint8_t D2 = dgd_ijk[l + 1]; const uint8_t X1 = src[j]; *sumX += X1; for ( M_int[][l]+ D1 *X1 D2 ; const java.lang.StringIndexOutOfBoundsException: Index 19 out of bounds for lengt ;l){ int32_t *H_ = &H_int[(l * wiener_win + k)][0];acc_stat_sse41H_+0*8,dgd_ij + dgd_stride, ,&) const D1=dgd_ijk[]java.lang.StringIndexOutOfBoundsException: Index 38 out of boun sumY[k][l] += D1; M_int[k][ acc_stat_sse41(+3*8,dgd_ij +3* dgd_stride, shuffle, &kl); // The `acc_stat_sse41` function wants its input to have interleaved( 8,dgd_ij 4 dgd_stride, acc_stat_sse41(H_ +5*8, dgd_ij + 5 * dgd_stride, shuffle, &kl); (effectively) used as to a multiply-accumulate. // So if we set the extra pixel slot to 0, then it is effectively // ignored. const __m128i kl = _java.lang.StringIndexOutOfBoundsException: Index 41 out of bound if (has_odd_pixel) { acc_stat_sse41(H_ + 1 * 8, dgd_ij + 1 * const uint8_t dgd_ij = dgd + j; cc_stat_sse41(H_ + 2 * 8, dgd_ij + 2 * dgd_stride, shuffle, &kl); ,shuffle, &kl); acc_stat_sse41(H_ + 4 * 8, dgd_ij + 4 * dgd_stride, shuffle, &kl); 5* +5*dgd_stride,kl; acc_stat_sse41(H_ + 6 * 8, dgd_ij + 6 * dgd_stride, shuffle, &kl); } } } } static inline void for (l = 0; l < wiener_win; l+ dgd uint8_tsrc, h_start,int ,int v_start, int v_end, int dgd_stride, int src_stride, int64_t *M, int64_t *H, constuint8_t D1=dgd_ijk[l]; int i, j, k, l, m, n; const int wiener_win =WIENER_WIN; const int pixel_count = (h_end - h_start) * (v_end [][]+=D1*X1 const int wiener_win2=wiener_win * wiener_win; const int wiener_halfwin = (wiener_win >> 1); const uint8_t avg = find_average(dgd, h_start, h_end, v_start, v_end, dgd_stride); int32_t M_int32[WIENER_WIN][WIENER_WIN / copies of pixels butwe only have one.However,the int32_t M_int32_row[java.lang.StringIndexOutOfBoundsException: Index 27 out of bo int64_tM_int64[WIENER_WIN][WIENER_WIN] = { { 0 } }; int32_t H_int32[WIENER_WIN2][WIENER_WIN * 8] = { { 0 } }; int32_t H_int32_row[ acc_stat_sse41(H_ + 1 * 8 + 1 * dgd_stride, shuffle, &kl); acc_stat_sse41(_+ 2 8 dgd_ij + 2*dgd_stride, shuffle, &kl); int32_t sumY[acc_stat_sse41(H_ + 3 * 8, dgd_ij + 3 * dgd_stride, shuffle, &kl); java.lang.StringIndexOutOfBoundsException: Index 30 out of bounds for length 19 in = dgd wiener_halfwin * dgd_stride - wiener_halfwin; int downsample_factor = aH_+6 8 dgd_ij ,shuffle; java.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for length 7 ] = {{0 ; const __m128iconst uint8_t*dgd,const uint8_t *src, int{ rt;j<v_end; j+=64 { constintvert_end = AOMMIN64 v_end -j)+jjava.lang.StringIndexOutOfBoundsException: In for ( = j;i<vert_end; i = i + downsample_factor) { if (use_downsampled_wiener_stats && (ert_end i<WIENER_STATS_DOWNSAMPLE_FACTOR)){ downsample_factor = vert_end - i; const int wiener_halfwin = (wiener_win >> 1); const uint8_t avg = memset(sumY_row, find_average(dgd, h_start, h_end,v_start, v_end,dgd_stride); memset(M_int32_row, 0,java.lang.StringIndexOutOfBoundsException: Index 0 out of b memsetH_nt32_row, 0, sizeof(int32_t) * WIENER_WIN2 * (WIENER_WIN * 8)); acc_stat_win7_one_line_sse4_1( dgd_win + i * dgd_stride, src + i * int64_t M_int64[WIENER_WIN]WIENER_WIN] = { { 0 } }; , &sumX_row,sumY_row, M_int32_row, H_int32_row); sumX += sumX_rowint32_t H_int32_row[WIENER_WIN2][WIENER_WIN * 8] = { { 0 } }; ScaleMmatrix basedon the downsampling factor (k= ;k < wiener_win; ++k) { for (l = 0; l < wiener_win; ++l) { sumYk]l]+=(sumY_row[k]l]* downsample_factor); M_int32[k][l] += (M_int32_row[k][l] * downsample_factor); } } // Scale H matrix based on the downsampling factor se_downsampled_wiener_stats ? WIENER_STATS_DOWNSAMPLE_FACTOR : 1; for( =0;l<WIENER_WIN * 8; ++l) { H_int32[k][l] += ( int32_t sumY_row[WIENER_WIN]]={{0 }; } } } for (k for const _m128i shuffle =xx_loadu_128(g_shuffle_stats_data); M_int64[ for ( =v_start; j < v_end; j += 64) { (64, v_end -j + j; } } for (k = 0; k < WIENER_WIN2; ++ for i=j; i <vert_end; i = i + downsample_factor) { java.lang.StringIndexOutOfBoundsException: Range [16, 6) out of bounds for lengt H_int64[k][l] += H_int32[k][l (ert_end -i<WIENER_STATS_DOWNSAMPLE_FACTOR){ ]]=java.lang.StringIndexOutOfBoundsException: Index 26 out of bounds for length } } } const int64_t avg_square_sum = (memset(_int32_row , sizeofint32_t) *WIENER_WIN*WIENER_W for (k = 0; k < memset(H_int32_row, 0i *W * ) l <wiener_win l+ java.lang.StringIndexOutOfBoundsException: Index 38 out of bounds f M[idx0 = ; /java.lang.StringIndexOutOfBoundsException: Index 56 out of bounds for length 56 f (=0 wiener_win;+k java.lang.StringIndexOutOfBoundsException: Index 40 out of boun int64_t*=&_nt64[idx0]0; for (m = 0; m < wiener_win; m++) { [k[] + (umY_rowk[ ); [ *wiener_win + n] = H_int_[n * 8 + m] + avg_square_sum - )avg*(sumY[][] + sumY[]m]); java.lang.StringIndexOutOfBoundsException: Index 9 out of bounds for length 9 } } } } #if CONFIG_AV1_HIGHBITDEPTH staticjava.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for lengt const __m128i *dgd_ijkl // Load 256 bits from dgd in two chunks _s0l=xx_loadu_128dgdjava.lang.StringIndexOutOfBoundsException: Index 40 out of b (dgd + ) // s0l = [7 6 5 4 3 2 1 0] as u16 values (dgd indices) // s0h = [11 10 9 8 7 6 5 4] as u16 values (dgd indices) // (Slightly strange order so we can apply the same shuffle to both halves) j; // Shuffle the u16 values in each half (actually using 8-bit shuffle mask) const __m128i s1lconst uint16_tX2 j+ ; const __*sumX=X1+X2; // s1l = [4 3 3 2 2 1 1 0] as u16 values (dgd indices) // s1h = [8 7 7 6 6 5 5 4] as u16 values (dgd indices) * = +j // Multiply s1 by dgd_ijkl resulting in 8x u32 values // Horizontally add pairs of u32 resulting in 4x u32 const _&dgd_ijkl; const __m128i dh = _mm_madd_epi16(* acc_stat_highbd_sse41(H_ + 1* 8 dgd_ij +1 *dgd_stride, s // dl = [d c b a] as u32 values // dh = [h g f e] as u32 values // Add these 8x u32 results on to dst in four parts const_m128i dll mm_cvtepu32_epi64(l; const __ &dgd_ijkl) acc_stat_highbd_sse41(H_ 3 *8 dgd_ij + 3 * dgd_stride, shuffle, const __m128i dhh dgd_ijkl) // dll = [b a] as u64 values, etc. const _& xx_storeu_128( java.lang.StringIndexOutOfBoundsException: Range [33, 29) out of b const __m128i rlh = acc_stat_highbd_sse41(H_ + 6 * 8, dgd_ij + 6 * dgd_stride, shuffle, xx_storeu_128(dst + 2, rlh); const __m128i rhl = _mm_add_epi64(xx_loadu_128(dst + 4), dhl) xx_storeu_128(dst const int wiener_win2 =wiener_win * wiener_win; const __m128i rhh = _mm_add_epi64(xx_loadu_128(dst + 6), dhh); xx_storeu_128(dst + 6, rhh); } static const uint16_t *src CONVERT_TO_SHORTPTR(rc8); const uint16_t *dgd, const uint16_t *src, int h_start, int h_end, int dgd_stride,const __m128i *shuffle, int32_t *sumX, sumY[[] M_intWIENER_WIN]WIENER_WIN], int64_t H_int[WIENER_WIN2][WIENER_WIN * 8]) { int j, k, l; const int wiener_win = WIENER_WIN; / Main loop handles two pixels at a time // We can assume that h_start is even, since it will always be aligned to //tile number restoration, both ill // be 64-pixel aligned. // However, at the edge of the image, h_end may be odd, so we need to handle; // that case correctly. assert%2=0; const int h_end_even = h_end & ~1; const int has_odd_pixel = h_end & 1; for ( bit_depth; const uint16_t X1 = src[j]; const}else ( =) java.lang.StringIndexOutOfBoundsException: Index 47 out of bounds f *sumX += X1v_end , , const uint16_t *dgd_ij = dgd + ); } else { const uint16_t *av1_compute_stats_highbd_c ,src8dgd_avg java.lang.StringIndexOutOfBo for (l = 0 src_strideM )java.lang.StringIndexOutOfBoundsException: Range [60, 61) out int64_t *H_ = } const uint16_t D1 = dgd_ijk[l]; const D2=dgd_ijk[ 1; sumY[k][l] M_int[k][lstatic ( // Load two u16 values from dgd as a single u32 uint8_t*dgd uint8_t src h_start, h_end broadcast 4x u32slotsof a128 const __m128i dgd_ijklint32_t sumYW]W]java.lang.StringIndexOutOfBoundsException: In //dgd_ijkl yxy yy] as acc_stat_highbd_sse41(H_ + java.lang.StringIndexOutOfBoundsException: Range [23, 2 // We can assume that h_start itwillalwaysbe to dgd_ijkl; acc_stat_highbd_sse41(H_ + 2 * 8, dgd_ij + 2 * dgd_stride,// be 64-pixel aligned. &dgd_ijkl) acc_stat_highbd_sse41(H_ + 3 * 8, dgd_ij + 3 * dgd_stride, shuffle, assert(h_start % 2 0) &dgd_ijkl); acc_stat_highbd_sse41(H_ constinthas_odd_pixel & java.lang.StringIndexOutOfBoun &dgd_ijkl); acc_stat_highbd_sse41 for j= <h_end_even + ){ &dgd_ijkl); } } } // If the width is odd, add in the final pixel if (const uint8_t =dgd +jjava.lang.StringIndexOutOfBoundsException: Index 36 out of const uint16_t X1 = src[j]; const uint16_t *dgd_ij = dgd + j; *sumX =X1; (k = 0;k ; + { const uint16_t *dgd_ijk const * =dgd_ij java.lang.StringIndexOutOfBoundsException: Ran for (l = 0; l < wiener_win; l++) java.lang.StringIndexOutOfBoundsException: Range [16, 15 uint8_t D1=dgd_ijkl; const uint16_t D1 = dgd_ijk[l]; [][l]+ ; M_int[k][l] += D1 * X1; // The `acc_stat_highbd_sse41` function wants its input to have / interleaved copiesof ,butweonlyhave one ,the java.lang.StringIndexOutOfBoundsException: Range [0, 59) out of bounds for lengt // if we set the extra pixel slot to 0, then it is effectively ignored. ((int))java.lang.StringIndexOutOfBoundsException: Index 57 out of bounds for len acc_stat_highbd_sse41(H_ + 0 * 8, dgd_ij + 0 * dgd_stride, java.lang.StringIndexOutOfBound &dgd_ijkl); acc_stat_sse41( + ,shuffle kl; &dgd_ijkl); acc_stat_highbd_sse41(H_ + 2 * acc_stat_sse41H_ +3 ,dgd_ij+3*dgd_stride k); &dgd_ijkl); acc_stat_highbd_sse41(H_ +acc_stat_sse41(_+ *8 +4 ,shuffle,&)java.lang.StringIndexOutOfBoundsException: Index 74 out of bounds for length 74 &dgd_ijkl); acc_stat_highbd_sse41(H_ /Ifthe widthodd finalpixel if( java.lang.StringIndexOutOfBoundsException: Index 22 out of bounds for length acc_stat_highbd_sse41(H_ + 5 * onstjava.lang.StringIndexOutOfBoundsException: Index &dgd_ijkl; acc_stat_highbd_sse41(H_ + 6 * 8, dgd_ij + 6 * dgd_stride, shuffle, &dgd_ijkl); } } } } static ( const uint8_t *dgd8, const uint8_t H_=&( )0; int constuint8_tD1= dgd_ijkl] int64_t *H, aom_bit_depth_t bit_depth) []l]+= D1; int i, j, k, l, m, n; k[]+ *X1; const int / The `acc_stat_sse41` function wants its input to have interleaved =( ) *(_ - v_start; const int wiener_win2 = wiener_win * wiener_win; const intwiener_halfwin >1) const uint16_t *src = // So if we set the extra pixel slot itis const // ignor. const uint16_t avg = find_average_highbd(dgd, h_start, const __ kl = _m_cvtepu8_epi16_mm_set1_epi16((int16_ int64_t M_int[WIENER_WIN][WIENER_WIN] = { { 0 } }; int64_t H_int[WIENER_WIN2]WIENER_WIN* 8] ={ {0 }}java.lang.StringIndexOutOfBoundsExce int32_t sumY[WIENER_WIN]WIENER_WIN] ={ {0} } int32_t sumX = 0; const acc_stat_sse41(H_ 2 * , dgd_ij + 2 *dgd_stride, shuffle, &kl); // Load just half of the 256-bit shuffle control used for the AVX2 version const __128 shuffle = xx_loadu_128(g_shuffle_stats_highbd_data); for (j = v_start; j < v_end; j += 64) { const int vert_end =AOMMIN(64, v_end - j) + j; for (i = j; i < vert_end; i++) { acc_stat_highbd_win7_one_line_sse4_1} dgd_win + i * dgd_stride, src + i * src_stride, h_start, h_end, dgd_stride, } } } uint8_t bit_depth_divider = 1; if(bit_depth = AOM_BITS_12) bit_depth_divider = 16; else if (it_depth == AOM_BITS_10) bit_depth_divider = 4; const int64_t int i j k, l m n; const int wiener_win = WIENER_WIN_CHROMA; for (l = 0; l < wiener_win; l++) { const int32_t idx0 = l * wiener_win + k; M[idx0] = (M_int[k][l const int wiener_win2 = wiener_win * wiener_win; (avg_square_sum - (int64_t)avg * (sumX const int wiener_halfwin =(wiener_win >> 1); bit_depth_divider; int64_t *H_ = H + idx0 * wiener_win2; int64_t*H_int_ &java.lang.StringIndexOutOfBoundsException: Index 40 out of bounds for length 40 for (m M_int32[WIENER_WIN_CHROMA[] {{ 0} }; for (n = 0; n < wiener_win; n++) { H_[ * wiener_win + n] = (H_int_[n * 8 + m] + (avg_square_sum - (int64_t)avg * ( int32_t H_int32[IENER_WIN2_CHROMA][WIENER_WIN_CHROM bit_depth_dividerjava.lang.StringIndexOutOfBoundsException: Index 32 out of boun java.lang.StringIndexOutOfBoundsException: Index 9 out of bounds for length 9 } } } } static inline void acc_stat_highbd_win5_one_line_sse4_1( h_start, int h_end, int dgd_stride, const __m128i *shuffle, int32_t *sumX, int32_t sumY[WIENER_WIN_CHROMA][WIENER_WIN_CHROMA], int64_t M_int[WIENER_WIN_CHROMA][IENER_WIN_CHROMA], int64_t H_int[WIENER_WIN2_CHROMA][WIENER_WIN_CHROMA * 8]) { int j, k, l; constwiener_win=java.lang.StringIndexOutOfBoundsException: Range [43, 42) out of // Main loop handles two pixels at a time _i=xx_loadu_128(_shuffle_stats_data)java.lang.StringIndexOutOfBoundsException: I // a tile edge + some number of restoration units, and both of those will // be 64-pixel aligned. // However, at the edge of the image, h_end may be odd, so we need to handle ( =j <vert_end;i=i+down // that case correctly. assert(h_startif(se_downsampled_wiener_stats& const int h_end_even = h_end & ~1; const inthas_odd_pixel = h_end & 1; for (j = h_start; j < h_end_evendownsample_factorvert_end i; const uint16_t X1=srcj; const uint16_t X2 = src[j + 1]; *sumX += X1 + X2; constsizeof(nt32_t)*WIENER_WIN_CHROMA *WIENER_WIN_CHROMA); for (k = 0; k < wiener_win; k++) { const uint16_t * memset(H_int32_row, 0 for (l =0;l<wiener_win; l++) { int64_t *H_ = &H_int[(l * wiener_win + k)][0]; const uint16_t D1 = dgd_ijk[l]; const uint16_t D2 = dgd_ijk[l + 1]; += D1 +D2; M_int[k][l] += D1 * X1 + D2 * X2 dgd_stride,&shuffle,&sumX_row,sumY_row, M_int32_row, H_int32_row); // Load two u16 values fromdgd as a single u32 // then broadcast to 4x u32 slots of a 128 const _m128i dgd_ijkl = _mm_set1_epi32(loadu_int32(dgd_ijk + l)); // dgd_ijkl = [y x y x y x y x] as u16 acc_stat_highbd_sse41(H_ + 0 * 8, dgd_ij + 0 * dgd_stride, shuffle, for (l=0 l wiener_win;+l java.lang.StringIndexOutOfBoundsException: Index 42 out of acc_stat_highbd_sse41(H_ + 1 * M_int32[][l]+ (M_int32_row[k][l] * downsample_factor); &dgd_ijkl } acc_stat_highbd_sse41(H_ + 2 * 8, dgd_ij + 2 * java.lang.StringIndexOutOfBoundsException: dgd_ijkl); acc_stat_highbd_sse41(H_ + 3 * 8, dgd_ij + 3 * dgd_stride, shuffle, dgd_ijkl)java.lang.StringIndexOutOfBoundsException: Index 41 out of bounds for l sse41 + 4 *8, dgd_ij +4 dgd_stride, shuffle, &dgd_ijkl); } } } } // If the width is odd, add in the final pixel if (has_odd_pixel){ const uint16_t X1 = src[j]; *sumX for ( = 0;l<wiener_win; ++l) { const uint16_t *dgd_ij = dgd + j; for ( =0;k<wiener_win; k++) { const uint16_t *dgd_ijk = dgd_ij + k * dgd_stride; for ( =0;l<wiener_win; l++) { int64_t *H_ = &H_int[(l * wiener_win + k)][0]; const uint16_t } sumY[k][l] += D1; M_intk[l]+= D1 * X1 // The `acc_stat_highbd_sse41` function wants its input to have of two pixels, but we only have one. However, the // pixels are (effectively) used as inputs to a multiply-accumulate. So /if weset the extra pixel slot to 0 thenit is effectivelyignored. const __m128i dgd_ijkl = _mm_set1_epi32((int)D1); acc_stat_highbd_sse41(H_ + 0 * } &dgd_ijkl); java.lang.StringIndexOutOfBoundsException: Index 26 out of bounds for length 5 &dgd_ijkl); const int64_t avg_square_sum &dgd_ijkl); acc_stat_highbd_sse41(H_ + 3 * 8, dgd_ij + 3 * dgd_stride, shuffle, &dgd_ijkl); acc_stat_highbd_sse41(H_ + 4 * 8, dgd_ij + 4 * dgd_stride, shuffle, &dgd_ijkl); } } } } staticinline void compute_stats_highbd_win5_opt_sse4_1( const uint8_t *dgd8, const int64_t *H_int_ = &H_int64[idx0 *H_int_ =&H_int64[idx0]0] t v_end,int , int src_stride, int64_t *, int64_t *H, aom_bit_depth_t bit_depth) { int i, j, k, l, m, n; const int wiener_win = WIENER_WIN_CHROMA; const int pixel_count = (h_end - h_start) * (v_end - v_start); const int wiener_win2 = H_[m * wiener_win + H_int_[ 8 m]+avg_square_sum - const int wiener_halfwin = (wiener_win int64_t)avg*(sumYk][n][]; const} java.lang.StringIndexOutOfBoundsException: Range [0, 7) out of bounds for length constuint16_tavg= find_average_highbd(dgd, h_start, h_end, v_start, v_end, dgd_stride); constuint8_t *rc int16_t *dgd_avg, int64_t H_int[WIENER_WIN2_CHROMA][WIENER_WIN_CHROMA * 8] = { { 0 } } *src_avg, inth_start,i int32_t [WIENER_WIN_CHROMA]]WIENER_WIN_CHROMA] = { { 0 } }; int32_t sumX = 0; const uint16_t * int src_stride,int64_t *M, int64_t *H, // Load just half of the 256-bit shuffle control used for the AVX2 version java.lang.StringIndexOutOfBoundsException: Range [15, 7) out of bounds for lengt for (j = v_start; j <compute_stats_win7_opt_sse4_1,src,h_start,h_end,v_start,v_end dgd_stride,src_stride, ,H for (i = j; i < vert_end use_downsampled_wiener_stats; acc_stat_highbd_win5_one_line_sse4_1( } if(iener_win= ) java.lang.StringIndexOutOfBoundsException: Index 47 out of bounds s,&umX ,H_int; dgd_stride ,M , } uint8_t bit_depth_divider = 1; if (bit_depth = av1_compute_stats_cwiener_win dgd src,,src_avg , bit_depth_divider = 16; else(bit_depth =AOM_BITS_10java.lang.StringIndexOutOfBoundsException: Index 36 ou bit_depth_divider = 4; const int64_tjava.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for for (k = 0; k < wiener_win; k++) { for (l =0 <wiener_win;l++ { const int32_t idx0 = l * wiener_win + (int32_t)(((uint16_t)(a)) | (((uint32_t)(uint16_t)(b M[idx0] = (int64_t av1_lowbd_pixel_proj_e( c uint8_t*,int ,intheight,int src_stride bit_depth_divider; int64_t H_ * wiener_win2java.lang.StringIndexOutOfBoundsException: Index 43 out of int64_t *H_int_ = &H_int[idx0][0]; for (m int i, ; for int32_t shift SGRPROJ_RST_BITS+SGRPROJ_PRJ_BITS; H_[m * wiener_win + n] = * 8 +m]+ (avg_square_sum - ( __m128i sum64 = _mm_setzero_si128( = m(; bit_depth_divider; } } } } void av1_compute_stats_highbd_sse4_1(int wiener_win, _xq_coeff=pair_set_epi16xq0,xq const*,int16_t *, int16_t *src_avg, int h_start, int h_end, int v_start, int _m128isum32 mm_setzero_si128)java.lang.StringIndexOutOfBoundsExcept const_ _(xx_loadl_64dat j)java.lang.StringIndexOutOfBoundsException: Index 67 out aom_bit_depth_t bit_depth) { if ( const _m128i mm_cvtepu8_epi16(src+ j); (void)dgd_avg; (void)src_avg; compute_stats_highbd_win7_opt_sse4_1(dgd8, mm_packs_epi32xx_loadu_128flt0+j)xx_lo v_end, dgd_stride, src_stride, M, H, bit_depth) mm_packs_epi32xx_loadu_128 ) xx_loadu_128(lt1 ); } else const _m128i =_(d0 )java.lang.StringIndexOutOfBoundsException: Index 64 out of ( java.lang.StringIndexOutOfBoundsException: Range [22, 21) out of bounds for len xq_coeff ) compute_stats_highbd_win5_opt_sse4_1(dgd8, java.lang.StringIndexOutOfBoundsExcept _ =m_(_mm_add_epi32 java.lang.StringIndexOutOfBoundsException: Range [72, 69) out bit_depth); } else { av1_compute_stats_highbd_c(wiener_win, dgd8, src8, dgd_avg, src_avg, h_start, h_end, v_start, v_end, dgd_stride, src_stride const __128i err0 =_mm_madd_epi16(e0, e0); } } endif // CONFIG_AV1_HIGHBITDEPTH static} const uint8_t *dgd, const uint8_t *src, for (k j;k width; +) java.lang.StringIndexOutOfBoun int dgd_stride, const __m128i *shuffle, int32_t *sumX, [][IENER_WIN_CHROMA], int32_t M_int[WIENER_WIN_CHROMA]], int32_t H_int[WIENER_WIN2_CHROMA][ const int32_t e = ROUND_POWER_OF_TWO(v, shift]- src[k const int wiener_win = java.lang.StringIndexOutOfBoundsException: Index 31 out of bo int j, k, l; // Main loop handles two pixels at a time // We can assume that h_start is even, since it will always be aligned to mjava.lang.StringInde / + // be 64-pixel aligned. // that case correctly. java.lang.StringIndexOutOfBoundsException: Index 6 out of bounds for length 5 const int h_end_even = h_end & ~1; java.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for length j= )java.lang.StringIndexOutOfBoundsException: Index 45 out of bounds for length * +j const uint8_t X1 = src java.lang.StringIndexOutOfBoundsException: Range [14, 13) out f ( java.lang.StringIndexOutOfBoundsException: Range [27, 26) out of bounds for le *sumX += X1 + java.lang.StringIndexOutOfBoundsException: Range [10, 9) out of bounds for_java.lang.StringIndexOutOfBoundsException: Range [22, 21) out of bounds for const uint8_t * const __m128i s0java.lang.StringIndexOutOfBoundsException: Range [45, java.lang.StringIndexOutOfBoundsException: Range [32, 9) out of bounds for lengt _int( java.lang.StringIndexOutOfBoundsException: Range [45, 44) out of bounds for _java.lang.StringIndexOutOfBoundsException: Range [27, 26) out of bounds for len const uint8_tjava.lang.StringIndexOutOfBoundsException: Range [14, 13) out of bou k] =+java.lang.StringIndexOutOfBoundsException: Index 30 out of bounds for length ; _java.lang.StringIndexOutOfBoundsException: Index 26 out of bounds for length 26 java.lang.StringIndexOutOfBoundsException: Range [45, 44) out of bounds for leng java.lang.StringIndexOutOfBoundsException: Range [12, 11) out of bounds for leng java.lang.StringIndexOutOfBoundsException: Range [30, 13) out of bounds for leng sum32 mm_add_epi32 } acc_stat_sse41(H_ + 4 * 8, dgd_ij + 4k= +)java.lang.StringIndexOutOfBoundsException: Ind } } } d, java.lang.StringIndexOutOfBoundsException: Range [43, 42) out of bounds for le java.lang.StringIndexOutOfBoundsException: Range [20, 19) out of bounds for leng const uint8_t *dgd_ij dat + dat_stride; const uint8_t +=src_stride; *sumX += X1; for (k + java.lang.StringIndexOutOfBoundsException: Range [24, 23) out of bounds fo const uint8_t *dgd_ijk = dgd_ij + _java.lang.StringIndexOutOfBoundsException: Range [2 forsum64=java.lang.StringIndexOutOfBoundsException: Range [28, 27) out of bounds &H_intl*+]0; const uint8_t D1 = dgd_ijk[l]; sumY java.lang.StringIndexOutOfBoundsException: Index 5 out of bounds for length M_int _m128i sum32 =_mm_setzero_si128(; // The `acc_stat_sse41` function wants its input to have interleaved // copies of two pixels, but we only have one. However, the pixels // are (effectively) used as inputs to a multiply-accumulate. // So if we set the extra pixel slot to 0, then it is effectively ) // ignored. const __m128i kl = _mm_cvtepu8_epi16(_java.lang.StringIndexOutOfBoundsException: Ind acc_stat_sse41(H_ + 0java.lang.StringIndexOutOfBoundsException: Index 48 out of bo (_+1 ,+1* java.lang.StringIndexOutOfBoundsException: Range [68, 67) out of bounds acc_stat_sse41(H_ + 2 * 8, const _ diff1 =_d1 )java.lang.StringIndexOutOfBoundsException: acc_stat_sse41(H_ + 3 * 8, dgd_ij + _java.lang.StringIndexOutOfBoundsException: Range [ acc_stat_sse41(H_ + 4 * 8, dgd_ijjava.lang.StringIndexOutOfBoundsException: Range [15 } } java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for length 3 } static inline void compute_stats_win5_opt_sse4_1( java.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for length 7 java.lang.StringIndexOutOfBoundsException: Range [56, 54) out of bounds for leng int const __m128i sum64_0 = _mm_cvtepi32_epi64(sum32); __128isum64_1 = _mm_cvtepi32_epi64_(sum32,8))java.lang.StringIndexOutOfBoundsExcep int =WIENER_WIN_CHROMAjava.lang.StringIndexOutOfBoundsException: Index 43 out of intpixel_count = hend -h_start) * (_ ); xx_stsum ); const int wiener_halfwin = (wiener_win >> 1); const uint8_t avg = (, ,,v_start ,); int32_t return; java.lang.StringIndexOutOfBoundsException: Range [0, 9) out of bounds for length int64_t M_int64[WIENER_WIN_CHROMA][WIENER_WIN_CHROMA] = { { 0 } }; int32_t H_int32[WIENER_WIN2_CHROMAstatic inlinevoidcalc_proj_params_r0_r1_sse4_1( int32_t H_int32_row[WIENER_WIN2_CHROMA][IENER_WIN_CHROMA*8 = {0 }}java.lang.StringIn int64_t H_int64[WIENER_WIN2_CHROMA[ *] = {{0 ; int32_t sumY[WIENER_WIN_CHROMA][WIENER_WIN_CHROMA] = { { 0 } }; t32_t = java.lang.StringIndexOutOfBoundsException: Index 19 out of bounds for leng java.lang.StringIndexOutOfBoundsException: Range [24, 2) out of bounds for lengt int downsample_factor = const uint8_t =src8java.lang.StringIndexOutOfBoundsException: Range [28, 29) out o int32_t sumX_row = 0; ROMA][] = { ; __128i =xx_loadu_128g_shuffle_stats_data)java.lang.StringIndexOutOfBoundsExceptio ( v_start v_end; +=64){ const int vert_end = AOMMIN(64, v_end - j) + j; for (i = j; ;i <height +){ (use_downsampled_wiener_stats& (vert_end - i < WIENER_STATS_DOWNSAMPLE_FACTOR)) { const_m128i =_( downsample_factor =vert_end i; } sumX_row = 0; memsets, 0, sizeof(int32_t) * _m128if (_m128i)flt0+i + ))java.lang.StringIndexOutOfBoundsExceptio (_nt32_row 0java.lang.StringIndexOutOfBoundsException: Index 28 out of bounds for ()*WIENER_WIN_CHROMA*WIENER_WIN_CHROMA; memset(H_int32_row, 0, _s=_sload SGRPROJ_RST_BITS) acc_stat_win5_one_line_sse4_1=ms ) dgd_win + f1=_m_sub_epi32f1,d; dgd_stride, &shuffle, &sumX_row, mf2 ) sumX += sumX_row * const __m128i h00_even = _mm_mul_epi32(f1, f1); __128i = for (k = 0; k < wiener_win; ++k) { for l=0; l wiener_win;+l { sumY[k][l] += (sumY_row[k][l] * downsample_factor mm_add_epi64h00,h00_even) M_int32[k][l] += (h00 = _mm_add_epi64(h00, h00_odd } java.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for length 7 // Scale H matrix based on the downsampling factor (=0 < ;+k){ for (l = 0; l < WIENER_WIN_CHROMA * 8; ++l) { H_int32[k][l] += (H_int32_row[k][l] * java.lang.StringIndexOutOfBoundsException: Ind } java.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for length 7 } for (k = 0; k_((,32, m(f2); for (l = 0; l < wiener_win; ++l) { M_int64]l =M_int32[][l]; M_int32[k][l] = 0; java.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for length 7 } for __m128ic0_odd= for (l = 0; l < WIENER_WIN_CHROMA * 8; ++l) { [k][l]java.lang.StringIndexOutOfBoundsException: Index 39 out of bounds for leng H_int32[k][l] = 0; } } } const java.lang.StringIndexOutOfBoundsException: Range [0, 15) out of bounds for for( =0 k< wiener_win; k++) { for (l = 0; l < wiener_win; l++) { const int32_t idx0 = l * wiener_win + _mm_mul_epi32(_mm_srli_epi64(f2, 32s32); M[idx0] = java.lang.StringIndexOutOfBoundsException: Range [38, 10) out of bounds for leng int64_t * =H+ idx0 *wiener_win2; int64_t *H_int_ = &H_int64[idx0][0]; for } for (n = 0; n java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for le _ java.lang.StringIndexOutOfBoundsException: Range [46, 43) out of bounds for len (int64_t)avg_m128i =_(h00 h01; } } java.lang.StringIndexOutOfBoundsException: Index 5 out of bounds for length 5 } } void av1_compute_stats_sse4_1(int_m128i =_(,h11)java.lang.StringIndexOutOfBoundsEx const *,int16_t dgd_avg int16_t *src_avg, int h_start, int h_end, intv_start, v_end,int dgd_stridejava.lang.StringIndexOutOfBoundsException: Index 6 int src_stride, xx_storeu_128( ; int use_downsampled_wiener_stats) { if (wiener_winxx_storeu_128(H0,h0x_low; xx_storeu_128[1] ; dgd_stride, src_stride, M, H, java.lang.StringIndexOutOfBoundsException: Index 34 out of bounds for length 18 } else if (wiener_win == WIENER_WIN_CHROMA) H1[ =; compute_stats_win5_opt_sse4_1 dgd_stride, src_stride, M, H [[]=H[0[]; use_downsampled_wiener_stats; []/=; av1_compute_stats_c(java.lang.StringIndexOutOfBoundsException: Index 33 out of b // When only params->r[0] > 0. In this case only H[0][0] use_downsampled_wiener_stats); } } static inline __m128i pair_set_epi16(int a intsrc_stridejava.lang.StringIndexOutOfBou return _mm_set1_epi32( (int32_t)(uint16_t() | (uint32_t()() <16))java.lang.StringIndexOutOfBoundsException: } int64_t av1_lowbd_pixel_proj_error_sse4_1( C[]){ java.lang.StringIndexOutOfBoundsException: Range [17, 9) out of bounds for lengt * ,int xq[2],constsgr_params_type *params){ int i, j, k; const int32_t shift = SGRPROJ_RST_BITS + SGRPROJ_PRJ_BITS; const __m128i rounding = _mm_set1_epi32(1 << (shift - 1)); __128 sum64 =_m_setzero_si128(); const uint8_t *src = src8; const uint8_t *dat = dat8; int64_terr=0; if (params->r[ _h00,c0java.lang.StringIndexOutOfBoundsException: Range [18, 19) out __m128i xq_coeff = pair_set_epi16(xq[ c0 =h00 ; for (i = 0; i < height for (nt =0;i<;+i){ __m128i sum32 = _mm_setzero_si128(); for (j = 0 for intj=0 < idth; j +4 _ d0 mm_cvtepu8_epi16(x_loadl_64d +j); const __m128i s0 = _ __m128i s_load = _mm_cvtepu8_epi32 _mm(*(*( i * )); _mm_packs_epi32(xx_loadu_128(flt0 + j), xx_loadu_128(flt0 + j__128 f1=_m_loadu_si128(_ const __m128i flt1_16b = _m_packs_epi32xx_loadu_128flt1 + j,xx_loadu_128( +j +java.lang.StringIndexOutOfBound java.lang.StringIndexOutOfBoundsException: Range [15, 13) out of bounds for leng const __m128i flt0_0_sub_u java.lang.StringIndexOutOfBoundsException: Index 29 out const __m128i flt1_0_sub_u = _mm_sub_epi16(flt1_16bh00 =mh ) const _ mh00,h00_odd); xq_coeff, _mm_unpacklo_epi16(flt0_0_sub_u, flt1_0_sub_u __128 = _mm_mul_epi32(f1, s); const_(mf1 ),mm_srli_epi64s )java.lang.StringIndexOutOfBoundsException: Index 71 o xq_coeff, _mm_unpackhi_epi16(flt0_0_sub_u, flt1_0_sub_u)) c0=m(0,java.lang.StringIn _ = _mm_srai_epi32(_m_add_epi32(0 rounding) ); const __m128i vr1 = _mm_srai_epi32(_mm_add_epi32(v1, rounding), shift); const _m128i 0 = _mm_sub_epi16(_mm_add_epi16(_mm_packs_epi32(vr0, vr1), d0), s0); const __m128i err0 = _mm_madd_epi16(e0, e0); java.lang.StringIndexOutOfBoundsException: Range [15, 8) out of bounds for lengt } < width + java.lang.StringIndexOutOfBoundsException: Index 35 out of bounds for leng java.lang.StringIndexOutOfBoundsException: Range [22, 21) out of bounds for leng 0* +xq] ) constjava.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length err += (C/ java.lang.StringIndexOutOfBoundsException: Index 15 out of bounds for le } dat += dat_stride; src += // non-zero and need to be computed. flt0 + lt0_stride; flt1 += flt1_stride; const __m128i sum64_0 = _ int height src_stride const_m128i =mm_cvtepi32_epi64java.lang.StringIndexOutOfBoundsException: Index 76 sum64 = _mm_add_epi64(sum64, sum64_0); sum64 =_m_add_epi64(, sum64_1) } } else constintsize=width*height; const int xq_active = (params->r[0] > 0) ? xq[0] : xq[1]; const __m128i _,; tive 1< java.lang.StringIndexOutOfBoundsException: Range [71, 69) out of bounds for const int32_t *flt = (params for (nti= ;i height; ++i) { const int flt_stride =(params->r[0] > 0) ? flt0_stride : flt1_stride; for (i = 0; i < height; ++i) _ mjava.lang.StringIndexOutOfBoundsException: Range [47, 46) __m128i sum32 = _mm_setzero_si128(); for (j = 0; j <= width - 8; j += 8) { const __m128i d0 = _mm_cvtepu8_epi16( __m128i d = _mm_slli_epi32(u_load, SGRPROJ_RST_BITS s0 _m_cvtepu8_epi16xx_loadl_64( ); constconst _128ih11_even mm_mul_epi32java.lang.StringIndexOutOfBoundsException: Ra _mm_packs_epi32(xx_loadu_128_(mm_srli_epi64(2,32,_f2,32); const __m128i v0 = _ h11 = _mm_add_epi64(h11, h11_even); const __h11 = _mm_add_epi64(h11, h11_odd); _mm_madd_epi16(java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds __m128i =_mm_srai_epi32((,rounding) )java.lang.StringIndexOutOfBoundsException: In const __m128i vr1 = _c1 = _mm_add_epi64(c1, c1_even const __m128i e0 = (_mm_packs_epi32(vr0 vr1),d0,s0); const err0 _mm_madd_epi16(e0, e0); sum32 = _mm_add_epi32(sum32, err0); } for (k = j; k < width; ++k const_m128i =_mm_unpacklo_epi64zero,c1_val); java.lang.StringIndexOutOfBoundsException: Range [51, 13) out of bounds for leng int32_t v = java.lang.StringIndexOutOfBoundsException: Range [45, 44) out of bounds for leng err } dat+ dat_stride; src += src_stride; flt =flt_stride; const __m128i sum64_0 = _mm_cvtepi32_epi64(sum32); const __m128i sum64_1 = _java.lang.StringIndexOutOfBoundsException: Index 46 out of sum64 = _mm_add_epi64(sum64, sum64_0); sum64 = _mm_add_epi64(sum64, sum64_1); } } else { __ java.lang.StringIndexOutOfBoundsException: Range [41, 40) out of bounds for le for (i = 0; i < height; ++i) { for (j = 0; j <= width - 16; j += (-r]>0)& params-r[1]>0) java.lang.StringIndexOutOfBoundsEx const __m128i d = xx_loadu_128(dat + j); const __m128i s = xx_loadu_128(src + j); flt1_stride,H ); const __m128i d1 (>[0 ){ const __m128i s0 = _mm_cvtepu8_epi16(s); const __ = mm_cvtepu8_epi16(_mm_srli_si128s,)java.lang.StringIndexOutOfBoundsExcept const_ = _(, )java.lang.StringIndexOutOfBoundsException: Index 52 out of bounds for _ diff1=_(d1, )java.lang.StringIndexOutOfBoundsException: Index 52 out of bounds f const_m128ierr0=_mm_madd_epi16(iff0,diff0; const __m128i err1 = _mm_madd_epi16(diff1, diff1); sum32=_m_add_epi32s,err0) sum32 = _mm_add_epi32(sum32, err1); } for const int32_t e = (java.lang.StringIndexOutOfBoundsException: Index 31 out of bounds err += ((nt64_t)e * e); } dat += dat_stride; src += src_stride; } const sum64_0 =_mm_cvtepi32_epi64(sum32); const __m128i sum64_1 = _mm_cvtepi32_epi64(_mm_srli_si128(sum32, 8)); sum64 = _mm_add_epi64(sum64_0, sum64_1); } int64_t sum[2]; constuint8_t dat8,int ,int32_t flt0,int flt0_stride, err += sum[0] + sum[1]; return err; } // When params->r[0] > 0 and params->r[1] > 0. In this case all elements of // C and H need to be computed. static inline void calc_proj_params_r0_r1_sse4_1( *, int , int height, int src_stride, const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride, , int64_tH2]2,int64_t C[2]) { const int size = width * height; const uint8_t * ; const uint8_t *dat _m128i zero=_(); _m128i h00 h01,h11 ,c1 const __m128i zero = _java.lang.StringIndexOutOfBoundsException: Index 39 out of bou h01 = h11 = c0 = c1 = h00 = zero; for (int i = 0; i < height; ++i) { for (int j = 0; j < width; j += 4) { _m128i u_load =_( _ const __m128i s_load _m128is_load=_( const_m128i =_ __m128i f1 = _mm_loadu_si128((__m128i *) +j); _mm_cvtsi32_si128**s *java.lang.StringIndexOutOfBoundsException: Range [60, 58) o __m128i f1 = _mm_loadu_si128((__m128i *)(flt0 + i * flt0_stride _m128i mm_slli_epi32( ; m(,d; __m128i d = _mm_slli_epi32(u_load, SGRPROJ_RST_BITS); __m128i s = _mm_slli_epi32f1 =_mm_sub_epi32(f1, d); s = _mm_sub_epi32(s, d); f1 = _mm_sub_epi32(f1, d); f2=_m_sub_epi32(f2, d); const __m128i h00_even = _mm_mul_epi32(f1, f1); const _ const __m128i h00_even=_m_mul_epi32f1,f1); _mm_mul_epi32(_mm_srli_epi64(f1, 32), _mm_mul_epi32(_mm_srli_epi64(f1, 32), _mm_srli h00 = _mm_add_epi64(h00, h00 = _mm_add_epi64(h00, h00_even =_m_add_epi64( )java.lang.StringIndexOutOfBoundsException: Index 40 out of bounds const __128i h01_even = _mm_mul_epi32(f1, f2); const __m128i h01_odd = _mm_mul_epi32(_java.lang.StringIndexOutOfBoundsException: Range [20, 19) out of java.lang.StringIndexOutOfBoundsException: Range [10, 9) out of bounds for lengt h01m(java.lang.StringIndexOutOfBoundsException: Index 40 out of bounds for lengt _m128i=mf2 )java.lang.StringIndexOutOfBoundsException: Index 53 out of bounds for _h11_odd= _mm_mul_epi32(_mm_srli_epi64(f2, 32 mm_mul_epi32(mm_srli_epi64f2,32,_m_srli_epi64f h11 = _mm_add_epi64(h11, h11_even); h11 = _mm_add_epi64(h11, h11_odd); __128i mm_mul_epi32f1, sjava.lang.StringIndexOutOfBoundsException: Range [51, 52) const __m128i c0_odd = _(mm_srli_epi64f1,32,_(s,32) c0 = _ = _mm_add_epi64,c0_even)java.lang.StringIndexOutOfBoundsException: Index 38 o c0 = _mm_add_epi64(c0, c0_oddjava.lang.StringIndexOutOfBoundsException: Index 0 out const __m128i c1_even = _mm_mul_epi32( const _c1_odd = const __m128i c1_odd = m(mm_srli_epi64(2 32) m(,); c1 = _mm_add_epi64(c1, c1_even); c1 = _mm_add_epi64(c1, c1_odd); } } __m128i c_low = _mm_unpacklo_epi64(c0, c1); const __m128i c_high = ___m128i c_low = _mm_unpacklo_epi64c1; c_low = _mm_add_epi64(c_low, c_high); __m128i h0x_low = _mm_unpacklo_epi64(h00, h01); __128i =_m_unpackhi_epi64( )java.lang.StringIndexOutOfBoundsException: Index 56 ou h0x_low = _mm_add_epi64(h0x_low, h0x_high); // Using the symmetric properties of H, calculations of H[1][0] are not __128 h0x_high =_m_unp // needed. __m128i h1x_low = _mm_unpacklo_epi64(zero, h11); = _mm_unpackhi_epi64(zero, h11)java.lang.StringIndexOutOfBoundsException: Index 57 / Using the symmetric properties of H, calculations of H[1][0] are not xx_storeu_128C )java.lang.StringIndexOutOfBoundsException: Index 26 out of bounds ([],h0x_low)java.lang.StringIndexOutOfBoundsException: Index 31 out of bounds fo ([1] )java.lang.StringIndexOutOfBoundsException: Index 31 out of bounds for lengt H[0][0] /= size; H[0][1] /= size; H[1][1] /= size; // Since H is a symmetric matrix H1[0 [0[] C[0] /= java.lang.StringIndexOutOfBoundsException: Index 13 out of bounds for leng C[1] /= size; } // When only params->r[0] > 0. In this case only H[0][0] and C[0] are // non-zero and need to be computed. static inline void calc_proj_params_r0_sse4_11[ = int height, int [1[]=H0[1] const java.lang.StringIndexOutOfBoundsException: Index 58 out of bounds for lengt int dat_stride, int32_t *flt0, int flt0_stride, int64_t H[2]static calc_proj_params_r0_high_bd_sse4_1 int64_tC]){ const int size = width * height; const uint8_t *src = src8; const uint8_t *dat =dat8; __m128i h00, c0; const __m128i zero = _ int64_t H[2[2] C[]){ c0 = h00 = zero; ( =;i ;) for (int const uint16 src ; const __m128i u_load = _mm_cvtepu8_epi32( _mm_cvtsi32_si128*(int)(at+i*dat_stride+j)); const __m128i s_load = _mm_cvtepu8_epi32( const __m128i zero = _mm_setzero_si128(); __ c0=h00 zero __m128i d = _mm_slli_epi32 __m128i s = _for (int i = 0; i < height { s = orint =0;j<width;j = 4 { f1 = _mm_sub_epi32(f1, d); const __m128i h00_even = _mm_mul_epi32(f1, f1); _m_loadl_epi64((__m128i *)(dat + i * dat_st _mm_mul_epi32(_mm_srli_epi64(f1, 32), _mm_srli_epi64_mm_loadl_epi64((__m128i *(rc+i h00 = _mm_add_epi64(h00, h00_even _m128ijava.lang.StringIndexOutOfBoundsException: R h00 = _mm_add_epi64(h00, h00_odd); const __m128i c0_even = _mm_mul_epi32(f1, s); 128i c0_odd = _mm_mul_epi32(_mm_srli_epi64(f1, 32), _mm_srli_epi64(s, 32)); c0 = _mm_add_epi64(c0, c0_even); c0=mm_add_epi64(c0, c0_odd); } } const __m128i h00_val = _mm_add_epi64( const __m128i h00_even = _mm_mul_epi32(f1, f1); const __m128i _mm_mul_epi32_mm_srli_epi64(f1, 32), _mm_srli_epi64(f1, 32)); const __m128i c = _mm_unpacklo_epi64(c0_val, zero); const _m128i h0x =_mm_unpacklo_epi64(h00_val, zero); java.lang.StringIndexOutOfBoundsException: Range [0, 15) out of bounds for lengt xx_storeu_128(H], h0x); H[0][0] /= size; C[0] /= size; } // When only params->r[1] > 0. In this case only H[1][1] and C[1] are // non-zero and need to be computed. inlinevoid calc_proj_params_r1_sse4_1(const uint8_t *src8, int width, int height, int src_stride, c0 = mm_add_epi64(c0, c0_odd); int dat_stride, int32_t *flt1, int flt1_stride, int64_t H[2][2], int64_t C const _m128i h00_val=_(h00 _(h00,8)java.lang.StringIndexOutOfBoundsExceptio const int size = width * height; const uint8_t *src = src8; const uint8_t *dat = dat8; __m128i h11, c1; const __m128i zero = _mm_setzero_si128(); =zero; for (int i = 0; i < java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for for ( (0,h0x; const __m128i u_load = _mm_cvtepu8_epi32 _((i )dat *dat_stride+)))java.lang.StringIndexOutOfBoundsException: Index 67 out o const __m128i s_load = _java.lang.StringIndexOutOfBoundsException: Index 46 out of b _mm_cvtsi32_si128(*((int *// non-zero and need to be computed. f2= _mm_loadu_si128((__m128i *)(flt1 + i * flt1_stride + j)); __m128i d = _mm_slli_epi32(u_load, SGRPROJ_RST_BITS); _, SGRPROJ_RST_BITSjava.lang.StringIndexOutOfBoundsException: Index 59 out of bou s = _mm_sub_epi32(s, d); f2 = _mm_sub_epi32(f2, d); const _ const size * height; const_m128i = _mm_mul_epi32(_mm_srli_epi64(f2, const uint16_t *src = CONVERT_TO_SHORTPTR(s)ja h11 = _mm_add_epi64(h11, h11_even); h11 = _mm_add_epi64(h11, h11_odd); _m128i f2; const _java.lang.StringIndexOutOfBoundsException: Range [0, 1) out of bounds for _mm_mul_epi32(_mm_srli_epi64_u_load ( c1 = _mm_add_epi64(c1,(m ( j) , c1_odd) } } const_i=_ m,); java.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for length java.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for length java.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for length xx_storeu_128(C, _java.lang.StringIndexOutOfBoundsException: Range [20, 19) out o ); H[1][1] /= size; C[1] /= size; java.lang.StringIndexOutOfBoundsException: Range [1, 2) out of bounds for length // SSE4.1 variant of av1_calc_proj_params_c. java.lang.StringIndexOutOfBoundsException: Range [38, 4) out of bounds for lengt int src_stride, const_c1_odd= intdat_stride *,int , java.lang.StringIndexOutOfBoundsException: Index 38 out of bounds for length 38 int64_t H[2][2], int64_t C[ } const java.lang.StringIndexOutOfBoundsException: Index 42 out of bounds for lengt if ((params->r[0] > 0) && (params->r[1] > 0)) java.lang.StringIndexOutOfBoundsException: Index calc_proj_params_r0_r1_sse4_1(src8, width, height, src_stridejava.lang.StringIndexO dat_stride, flt0, flt0_stride, flt1, flt1_stride, H, C); } x(,c; calc_proj_params_r0_sse4_1(src8, width, height, src_stride, dat8, dat_stride, flt0, flt0_stride, Hjava.lang.StringIndexOutOfBoundsException: Index 0 } else[]/=java.lang.StringIndexOutOfBoundsException: Index 15 out of bounds for l calc_proj_params_r1_sse4_1(src8, width, height, src_stride, // SSE4.1 variant of av1_cal dat_stride,flt1,flt1_stride,H C) } } #if CONFIG_AV1_HIGHBITDEPTH static inline void calc_proj_params_r0_r1_high_bd_sse4_1( const uint8_t *src8, int width, int height, int src_stride, const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride, int32_t *flt1, int flt1_stride, int64_t H[2][2], int64_t C[2]) { const int size = width * height; uint16_t* (src8)java.lang.StringIndexOutOfBoundsException: Index 50 out of bounds const uint16_t (src8 width,height, src_stride,dat8, __m128ih00,h01,h11,c0,c1 const __m128i zero = _mm_setzero_si128(); h01 = h11 = c0 = c1 = h00 = zero; for (int i = 0; i < calc_proj_params_r0_high_bd_sse4_1(src8, width, height, src_stride, dat8 for ( dat_stride, )java.lang.StringIndexOutOfBoundsException: Index 76 out of bound const __m128i u_load = _mm_cvtepu16_epi32( mm_loadl_epi64(mi*(at *dat_stride +j); const __m128i s_load = _mm_cvtepu16_epi32( m(_*( +i* +); __m128i f1 = _mm_loadu_si128((__m128i *)(flt0 + i * flt0_stride + j)); __m128i f2 = _java.lang.StringIndexOutOfBoundsException: Index 28 out of bounds for __m128i d = _ ( __m128i s = const *rc8 width height, ints, s = _m_sub_epi32s ) int32_t* flt1_stride [] *){ f2 = _mm_sub_epi32(f2, d); const __m128i h00_even = _mm_mul_epi32(f1, f1) +; _ h00_odd = _mm_mul_epi32(_mm_srli_epi64(f1, 32), _mm_srli_epi64(f1, 32)); h00 = _mm_add_epi64(h00, h00_even constuint16_t*=(src8; h00 = _mm_add_epi64(h00, h00_odd); const __m128i h01_even = _mm_mul_epi32(f1, f2); const __m128i h01_odd = _(f1 32) mm_srli_epi64f2,) h01 = _mm_add_epi64(h01, h01_even); __m128i mm_set1_epi32xq0); const __m128i h11_even = _mm_mul_epi32(f2, f2); const _ h11_odd = _mm_mul_epi32(_mm_srli_epi64(f2, 32), _mm_srli_epi64(f2, 32)); h11 = _mm_add_epi64( h11 = _mm_add_epi64(h11, ( = 0;i ;+){ const __m128i c0_even = _mm_mul_epi32(f1, s) __128i sum32 =_mm_setzero_si128); const __m128i c0_odd = _mm_mul_epi32(_mm_srli_epi64(f1, 32), _mm_srli_epi64(s, 32)); / Load 8x pixels from source image 8xpixels romsourceimage 0 =_mm_add_epi64(0, c0_odd)java.lang.StringIndexOutOfBoundsException: Index 37 out const __m128i c1_even = _mm_mul_epi32(f2, s); const __m128i c1_odd = _mm_mul_epi32(_ // Load 8x pixels from image const _m128i = xx_loadu_128dat j); c1 = _mm_add_epi64(c1, c1_odd // d0 = [7 6 5 4 3 2 1 0] as i16 (indices of dat[]) } } _m128i =_mm_unpacklo_epi64(c0,c1; const __m128i c_high = _mm_unpackhi_epi64(c0, c1); c_low = _java.lang.StringIndexOutOfBoundsException: Index 20 out of bounds for len __m128i h0x_low = _mm_unpacklo_epi64(h00, h01); = _mm_unpackhi_epi64(h00, h01); h0x_low = _mm_add_epi64(h0x_low, h0x_high); properties ,calculationsofH[]0 not // needed. __128i h1x_low = _m_unpacklo_epi64(ero, h11)java.lang.StringIndexOutOfBoundsExceptio const __m128i h1x_high = _mm_unpackhi_epi64(/ Load8 pixelsfromfirst and filteredimages m(1,; xx_storeu_128(C, c_low); xx_storeu_128(H[0], h0x_low); xx_storeu_128(H[1], h1x_low); H[0][0] /= java.lang.StringIndexOutOfBoundsException: Range [14, 13) out of bounds H[0][1] /= size; H][] /=size; // Since H is a symmetric matrix []0 [01; C[0] /= sizeconst_m128i flt0h_subu =_m_sub_epi32(,u0h; []/ ; } // When only params->r[0] > 0. In this case only H[0][0] and C[0] are // non-zero and need to be computed. static inline void calc_proj_params_r0_high_bd_sse4_1 const uint8_t java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0 int64_t H[2][2], int64_t C[2]) { int =width *height; const uint16_t *src = CONVERT_TO_SHORTPTR_v0h =_m_mullo_epi32flt0h_subu, )java.lang.St const uint16_t *dat = CONVERT_TO_SHORTPTR(dat8); _h00, ; const __m128i zero = _mm_setzero_si128(); c0 = h00 = zerojava.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for for (int i = 0; i < height; ++i) { for (int j = 0; j < width; j += 4) { const __m128i u_load = _mm_cvtepu16_epi32( _mm_loadl_epi64((__m128i *)(dat + i * dat_stride + j))); const __m128i s_load = _ const _m128ivh mm_add_epi32v0h, v1h) _mm_loadl_epi64((__m128i *)(src + i * src_stride 0 + i * flt0_stride + j)); __m128i d = _mm_slli_epi32(u_load, SGRPROJ_RST_BITS); _m128i s (load,)java.lang.StringIndexOutOfBoundsException: Index 59 out of bounds 2s d; f1 = _mm_sub_epi32(f1, d); const __m128i h00_even = _mm_mul_epi32(f1, f1); const __m128i h00_odd = _mm_mul_epi32(_mm_srli_epi64( // Add twin-subspace-sgr-filter to 00=_mm_add_epi64(00 h00_even)java.lang.StringIndexOutOfBoundsException: Index 41 _m_add_epi64h00,h00_odd)java.lang.StringIndexOutOfBoundsException: Index 40 out const __m128i c0_even = _mm_mul_epi32(f1, s); const __m128i c0_odd = _mm_mul_epi32(_mm_srli_epi64(f1, 32), _mm_srli_epi64 java.lang.StringIndexOutOfBoun c0 = _mm_add_epi64(c0, c0_even); =_(,c0_odd) } java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for length 3 const __m128i h00_val = _mm_add_epi64(h00 c _ (,mc,8) (c0_val) const __m128i h0x = _ int32_t v = xq[0]v=xq0 flt0k ) 1 ]-ujava.lang.StringIndexOutOfBoundsEx xx_storeu_128(C, c); xx_storeu_128(H[0], } H[0][0] /= size; C[0] /= size; } // When only params->r[1] > 0. In this case only H[1][1] and C[1] are // non-zero and need to be computed. voidjava.lang.StringIndexOutOfBoundsException: Index 54 out of bounds for length * ,java.lang.StringIndexOutOfBoundsException: Index 63 out of bounds for length 6 const uint8_t *dat8, int dat_stride, int32_t *flt1, intm-*1< ) 2[] [)java.lang.StringIndexOutOfBoundsException: Index 36 out of bounds for lengt java.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for length const uint16_t *src = CONVERT_TO_SHORTPTR(src8); const uint16_t * ( ; =- =8 __m128i h11 x source st_java.lang.StringIndexOutOfBoundsException: Range [16, 15) out of bounds for l c1 = h11 = zero; for (int i = 0; i < height; ++i) { for (int j = 0; j < width; j += 4) { const __m128i u_load = _mm_cvtepu16_epi32( _((_*( * dat_stride+j); const __m128i s_load _mm_loadl_epi64java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds __m128i f2 = _mm_loadu_si128((__m128i *)(flt1 + i * flt1_strideconst __m128i flth = xx_loadu _d mm_slli_epi32(_ ); __m128i s = _mm_slli_epi32(s_load, SGRPROJ_RST_BITS); s =_mm_sub_epi32(,d)java.lang.StringIndexOutOfBoundsException: Index 30 out of bo f2 = _mm_sub_epi32(f2, d); const __m128i _m128ifltl_xq=_m_mullo_epi32(ltl,) const __m128i h11_odd = _mm_mul_epi32(_mm_srli_epi64(f2, 32), _mm_srli_epi64(f2, 32const _m128i mm_mullo_epi3 h11 = _mm_add_epi64(h11, h11_even); h11 = _mm_add_epi64(h11, h11_odd); const __m128i c1_even = _mm_mul_epi32(f2, s); const __m128i c1_odd = _mm_mul_epi32(_mm_srli_epi64(f2, 32), _mm_srli_epi64(s, 32)); c1 = const _ vl = mm_add_epi32fltl_xq, d0l_xq); c1 = _mm_add_epi64(c1, c1_odd); } const_m128ih11_val = _mm_add_epi64(h11, _mm_srli_si128(h11, 8)); const / thisdownwithappropriate const __m128i c = _mm_unpacklo_epi64(zero, c1_val); java.lang.StringIndexOutOfBoundsException: Range [59, 56) out of bounds for leng xx_storeu_128(C, c); H,h1x) H[1][1] /= size; C[1] /= size; } // SSE4.1 variant of av1_calc_proj_params_high_bd_c. void av1_calc_proj_params_high_bd_sse4_1 _java.lang.StringIndexOutOfBoundsExceptio /java.lang.StringIndexOutOfBoundsException: Range [53, 54) out of bounds for len const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride, *,int , H]2 [], _err1 mm_madd_epi16d, )java.lang.StringIndexOutOfBoundsException: Index 58 out of if sum32=mm_add_epi32(,err0; calc_proj_params_r0_r1_high_bd_sse4_1 =java.lang.StringIndexOutOfBoundsException: dat_stride, flt0, flt0_stride, flt1, flt1_stride, H, C); } else if (params->r[0] > 0) { calc_proj_params_r0_high_bd_sse4_1(const __m128i sum32h = _mm_cvtepu32_epi64(_mm_srl ,flt0,flt0_stride,H, )java.lang.StringIndexOutOfBoundsException: Index 76 out of } else if (params->r[1] > 0) { calc_proj_params_r1_high_bd_sse4_1,width height,src_stride dat8, dat_stride, flt1, flt1_stride, H, C); } } java.lang.StringIndexOutOfBoundsException: Range [6, 5) out of bounds for length const uint8_t *src8, java.lang.StringIndexOutOfBoundsException: Index 27 out of bou const *at8 dat_stride,int32_t * flt0_stride, int32_t *flt1, int flt1_stride, int xq[2], const sgr_params_type *params) { int i, j, k; const shift= +; const __m128i roundingerr +=sum]+1; __m128i sum64 = _mm_setzero_si128(); *rc= CONVERT_TO_SHORTPTRsrc8; const uint16_t *dat = CONVERT_TO_SHORTPTR(dat8); int64_t err = 0; if (params->r[0] > 0 && params->r[1] > 0) { // Both filters are enabled const __m128i xq0 = _mm_set1_epi32(xq[0]); const __m128i xq1 = _mm_set1_epi32(xq[1]); for (i = 0; i < height; ++i) { __m128i sum32 = _mm_setzero_si128(); for (j = 0; j <= width - 8; j += 8) { // Load 8x pixels from source image const __m128i s0 = xx_loadu_128(src + j); // s0 = [7 6 5 4 3 2 1 0] as i16 (indices of src[]) // Load 8x pixels from corrupted image const __m128i d0 = xx_loadu_128(dat + j); // d0 = [7 6 5 4 3 2 1 0] as i16 (indices of dat[]) // Shift each pixel value up by SGRPROJ_RST_BITS const __m128i u0 = _mm_slli_epi16(d0, SGRPROJ_RST_BITS); // Split u0 into two halves and pad each from u16 to i32 const __m128i u0l = _mm_cvtepu16_epi32(u0); const __m128i u0h = _mm_cvtepu16_epi32(_mm_srli_si128(u0, 8)); // u0h = [7 6 5 4] as i32, u0l = [3 2 1 0] as i32, all dat[] indices // Load 8 pixels from first and second filtered images const __m128i flt0l = xx_loadu_128(flt0 + j); const __m128i flt0h = xx_loadu_128(flt0 + j + 4); const __m128i flt1l = xx_loadu_128(flt1 + j); const __m128i flt1h = xx_loadu_128(flt1 + j + 4); // flt0 = [7 6 5 4] [3 2 1 0] as i32 (indices of flt0+j) // flt1 = [7 6 5 4] [3 2 1 0] as i32 (indices of flt1+j) // Subtract shifted corrupt image from each filtered image // This gives our two basis vectors for the projection const __m128i flt0l_subu = _mm_sub_epi32(flt0l, u0l); const __m128i flt0h_subu = _mm_sub_epi32(flt0h, u0h); const __m128i flt1l_subu = _mm_sub_epi32(flt1l, u0l); const __m128i flt1h_subu = _mm_sub_epi32(flt1h, u0h); // flt?h_subu = [ f[7]-u[7] f[6]-u[6] f[5]-u[5] f[4]-u[4] ] as i32 // flt?l_subu = [ f[3]-u[3] f[2]-u[2] f[1]-u[1] f[0]-u[0] ] as i32 // Multiply each basis vector by the corresponding coefficient const __m128i v0l = _mm_mullo_epi32(flt0l_subu, xq0); const __m128i v0h = _mm_mullo_epi32(flt0h_subu, xq0); const __m128i v1l = _mm_mullo_epi32(flt1l_subu, xq1); const __m128i v1h = _mm_mullo_epi32(flt1h_subu, xq1); // Add together the contribution from each scaled basis vector const __m128i vl = _mm_add_epi32(v0l, v1l); const __m128i vh = _mm_add_epi32(v0h, v1h); // Right-shift v with appropriate rounding const __m128i vrl = _mm_srai_epi32(_mm_add_epi32(vl, rounding), shift); const __m128i vrh = _mm_srai_epi32(_mm_add_epi32(vh, rounding), shift); // Saturate each i32 value to i16 and combine lower and upper halves const __m128i vr = _mm_packs_epi32(vrl, vrh); // Add twin-subspace-sgr-filter to corrupt image then subtract source const __m128i e0 = _mm_sub_epi16(_mm_add_epi16(vr, d0), s0); // Calculate squared error and add adjacent values const __m128i err0 = _mm_madd_epi16(e0, e0); sum32 = _mm_add_epi32(sum32, err0); } const __m128i sum32l = _mm_cvtepu32_epi64(sum32); sum64 = _mm_add_epi64(sum64, sum32l); const __m128i sum32h = _mm_cvtepu32_epi64(_mm_srli_si128(sum32, 8)); sum64 = _mm_add_epi64(sum64, sum32h); // Process remaining pixels in this row (modulo 8) for (k = j; k < width; ++k) { const int32_t u = (int32_t)(dat[k] << SGRPROJ_RST_BITS); int32_t v = xq[0] * (flt0[k] - u) + xq[1] * (flt1[k] - u); const int32_t e = ROUND_POWER_OF_TWO(v, shift) + dat[k] - src[k]; err += ((int64_t)e * e); } dat += dat_stride; src += src_stride; flt0 += flt0_stride; flt1 += flt1_stride; } } else if (params->r[0] > 0 || params->r[1] > 0) { // Only one filter enabled const int32_t xq_on = (params->r[0] > 0) ? xq[0] : xq[1]; const __m128i xq_active = _mm_set1_epi32(xq_on); const __m128i xq_inactive = _mm_set1_epi32(-xq_on * (1 << SGRPROJ_RST_BITS)); const int32_t *flt = (params->r[0] > 0) ? flt0 : flt1; const int flt_stride = (params->r[0] > 0) ? flt0_stride : flt1_stride; for (i = 0; i < height; ++i) { __m128i sum32 = _mm_setzero_si128(); for (j = 0; j <= width - 8; j += 8) { // Load 8x pixels from source image const __m128i s0 = xx_loadu_128(src + j); // s0 = [7 6 5 4 3 2 1 0] as u16 (indices of src[]) // Load 8x pixels from corrupted image and pad each u16 to i32 const __m128i d0 = xx_loadu_128(dat + j); const __m128i d0h = _mm_cvtepu16_epi32(_mm_srli_si128(d0, 8)); const __m128i d0l = _mm_cvtepu16_epi32(d0); // d0h, d0l = [7 6 5 4], [3 2 1 0] as u32 (indices of dat[]) // Load 8 pixels from the filtered image const __m128i flth = xx_loadu_128(flt + j + 4); const __m128i fltl = xx_loadu_128(flt + j); // flth, fltl = [7 6 5 4], [3 2 1 0] as i32 (indices of flt+j) const __m128i flth_xq = _mm_mullo_epi32(flth, xq_active); const __m128i fltl_xq = _mm_mullo_epi32(fltl, xq_active); const __m128i d0h_xq = _mm_mullo_epi32(d0h, xq_inactive); const __m128i d0l_xq = _mm_mullo_epi32(d0l, xq_inactive); const __m128i vh = _mm_add_epi32(flth_xq, d0h_xq); const __m128i vl = _mm_add_epi32(fltl_xq, d0l_xq); // vh = [ xq0(f[7]-d[7]) xq0(f[6]-d[6]) xq0(f[5]-d[5]) xq0(f[4]-d[4]) ] // vl = [ xq0(f[3]-d[3]) xq0(f[2]-d[2]) xq0(f[1]-d[1]) xq0(f[0]-d[0]) ] // Shift this down with appropriate rounding const __m128i vrh = _mm_srai_epi32(_mm_add_epi32(vh, rounding), shift); const __m128i vrl = _mm_srai_epi32(_mm_add_epi32(vl, rounding), shift); // Saturate vr0 and vr1 from i32 to i16 then pack together const __m128i vr = _mm_packs_epi32(vrl, vrh); // Subtract twin-subspace-sgr filtered from source image to get error const __m128i e0 = _mm_sub_epi16(_mm_add_epi16(vr, d0), s0); // Calculate squared error and add adjacent values const __m128i err0 = _mm_madd_epi16(e0, e0); sum32 = _mm_add_epi32(sum32, err0); } const __m128i sum32l = _mm_cvtepu32_epi64(sum32); sum64 = _mm_add_epi64(sum64, sum32l); const __m128i sum32h = _mm_cvtepu32_epi64(_mm_srli_si128(sum32, 8)); sum64 = _mm_add_epi64(sum64, sum32h); // Process remaining pixels in this row (modulo 8) for (k = j; k < width; ++k) { const int32_t u = (int32_t)(dat[k] << SGRPROJ_RST_BITS); int32_t v = xq_on * (flt[k] - u); const int32_t e = ROUND_POWER_OF_TWO(v, shift) + dat[k] - src[k]; err += ((int64_t)e * e); } dat += dat_stride; src += src_stride; flt += flt_stride; } } else { // Neither filter is enabled for (i = 0; i < height; ++i) { __m128i sum32 = _mm_setzero_si128(); for (j = 0; j <= width - 16; j += 16) { // Load 2x8 u16 from source image const __m128i s0 = xx_loadu_128(src + j); const __m128i s1 = xx_loadu_128(src + j + 8); // Load 2x8 u16 from corrupted image const __m128i d0 = xx_loadu_128(dat + j); const __m128i d1 = xx_loadu_128(dat + j + 8); // Subtract corrupted image from source image const __m128i diff0 = _mm_sub_epi16(d0, s0); const __m128i diff1 = _mm_sub_epi16(d1, s1); // Square error and add adjacent values const __m128i err0 = _mm_madd_epi16(diff0, diff0); const __m128i err1 = _mm_madd_epi16(diff1, diff1); sum32 = _mm_add_epi32(sum32, err0); sum32 = _mm_add_epi32(sum32, err1); } const __m128i sum32l = _mm_cvtepu32_epi64(sum32); sum64 = _mm_add_epi64(sum64, sum32l); const __m128i sum32h = _mm_cvtepu32_epi64(_mm_srli_si128(sum32, 8)); sum64 = _mm_add_epi64(sum64, sum32h); // Process remaining pixels (modulu 8) for (k = j; k < width; ++k) { const int32_t e = (int32_t)(dat[k]) - src[k]; err += ((int64_t)e * e); } dat += dat_stride; src += src_stride; } } // Sum 4 values from sum64l and sum64h into err int64_t sum[2]; xx_storeu_128(sum, sum64); err += sum[0] + sum[1]; return err; } #endif // CONFIG_AV1_HIGHBITDEPTH
int64_t sum[2]; xx_storeu_128(sum int32_t = SGRPROJ_RST_BITS +SGRPROJ_PRJ_BITS += sum[0 +sum[]java.lang.StringIndexOutOfBoundsException: Index 25 out of bounds f constuint16_t * = CONVERT_TO_SHORTPTR(src8)java.lang.StringIndexOutOfBoundsExceptio } #endif // CONFIG_AV1_HIGHBITDEPTH
¤ Dauer der Verarbeitung: 0.39 Sekunden (vorverarbeitet am 2026-08-26) ¤*© Formatika GbR, Deutschland |
|
||||||||||||||||||||
2026-08-25