yading@10: /* yading@10: * RV40 decoder motion compensation functions yading@10: * Copyright (c) 2008 Konstantin Shishkov yading@10: * yading@10: * This file is part of FFmpeg. yading@10: * yading@10: * FFmpeg is free software; you can redistribute it and/or yading@10: * modify it under the terms of the GNU Lesser General Public yading@10: * License as published by the Free Software Foundation; either yading@10: * version 2.1 of the License, or (at your option) any later version. yading@10: * yading@10: * FFmpeg is distributed in the hope that it will be useful, yading@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of yading@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU yading@10: * Lesser General Public License for more details. yading@10: * yading@10: * You should have received a copy of the GNU Lesser General Public yading@10: * License along with FFmpeg; if not, write to the Free Software yading@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA yading@10: */ yading@10: yading@10: /** yading@10: * @file yading@10: * RV40 decoder motion compensation functions yading@10: */ yading@10: yading@10: #include "avcodec.h" yading@10: #include "h264qpel.h" yading@10: #include "rv34dsp.h" yading@10: #include "libavutil/avassert.h" yading@10: #include "libavutil/common.h" yading@10: yading@10: #define RV40_LOWPASS(OPNAME, OP) \ yading@10: static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ yading@10: const int h, const int C1, const int C2, const int SHIFT){\ yading@10: const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ yading@10: int i;\ yading@10: for(i = 0; i < h; i++)\ yading@10: {\ yading@10: OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: dst += dstStride;\ yading@10: src += srcStride;\ yading@10: }\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ yading@10: const int w, const int C1, const int C2, const int SHIFT){\ yading@10: const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ yading@10: int i;\ yading@10: for(i = 0; i < w; i++)\ yading@10: {\ yading@10: const int srcB = src[-2*srcStride];\ yading@10: const int srcA = src[-1*srcStride];\ yading@10: const int src0 = src[0 *srcStride];\ yading@10: const int src1 = src[1 *srcStride];\ yading@10: const int src2 = src[2 *srcStride];\ yading@10: const int src3 = src[3 *srcStride];\ yading@10: const int src4 = src[4 *srcStride];\ yading@10: const int src5 = src[5 *srcStride];\ yading@10: const int src6 = src[6 *srcStride];\ yading@10: const int src7 = src[7 *srcStride];\ yading@10: const int src8 = src[8 *srcStride];\ yading@10: const int src9 = src[9 *srcStride];\ yading@10: const int src10 = src[10*srcStride];\ yading@10: OP(dst[0*dstStride], (srcB + src3 - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[1*dstStride], (srcA + src4 - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[2*dstStride], (src0 + src5 - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[3*dstStride], (src1 + src6 - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[4*dstStride], (src2 + src7 - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[5*dstStride], (src3 + src8 - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[6*dstStride], (src4 + src9 - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\ yading@10: dst++;\ yading@10: src++;\ yading@10: }\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ yading@10: const int w, const int C1, const int C2, const int SHIFT){\ yading@10: OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\ yading@10: OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\ yading@10: src += 8*srcStride;\ yading@10: dst += 8*dstStride;\ yading@10: OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, w-8, C1, C2, SHIFT);\ yading@10: OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ yading@10: const int h, const int C1, const int C2, const int SHIFT){\ yading@10: OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\ yading@10: OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\ yading@10: src += 8*srcStride;\ yading@10: dst += 8*dstStride;\ yading@10: OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, h-8, C1, C2, SHIFT);\ yading@10: OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\ yading@10: }\ yading@10: \ yading@10: yading@10: #define RV40_MC(OPNAME, SIZE) \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\ yading@10: {\ yading@10: uint8_t full[SIZE*(SIZE+5)];\ yading@10: uint8_t * const full_mid = full + SIZE*2;\ yading@10: put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\ yading@10: OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\ yading@10: }\ yading@10: \ yading@10: yading@10: #define op_avg(a, b) a = (((a)+cm[b]+1)>>1) yading@10: #define op_put(a, b) a = cm[b] yading@10: yading@10: RV40_LOWPASS(put_ , op_put) yading@10: RV40_LOWPASS(avg_ , op_avg) yading@10: yading@10: #undef op_avg yading@10: #undef op_put yading@10: yading@10: RV40_MC(put_, 8) yading@10: RV40_MC(put_, 16) yading@10: RV40_MC(avg_, 8) yading@10: RV40_MC(avg_, 16) yading@10: yading@10: static const int rv40_bias[4][4] = { yading@10: { 0, 16, 32, 16 }, yading@10: { 32, 28, 32, 28 }, yading@10: { 0, 32, 16, 32 }, yading@10: { 32, 28, 32, 28 } yading@10: }; yading@10: yading@10: #define RV40_CHROMA_MC(OPNAME, OP)\ yading@10: static void OPNAME ## rv40_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ yading@10: const int A = (8-x) * (8-y);\ yading@10: const int B = ( x) * (8-y);\ yading@10: const int C = (8-x) * ( y);\ yading@10: const int D = ( x) * ( y);\ yading@10: int i;\ yading@10: int bias = rv40_bias[y>>1][x>>1];\ yading@10: \ yading@10: av_assert2(x<8 && y<8 && x>=0 && y>=0);\ yading@10: \ yading@10: if(D){\ yading@10: for(i = 0; i < h; i++){\ yading@10: OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\ yading@10: OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\ yading@10: OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\ yading@10: OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\ yading@10: dst += stride;\ yading@10: src += stride;\ yading@10: }\ yading@10: }else{\ yading@10: const int E = B + C;\ yading@10: const int step = C ? stride : 1;\ yading@10: for(i = 0; i < h; i++){\ yading@10: OP(dst[0], (A*src[0] + E*src[step+0] + bias));\ yading@10: OP(dst[1], (A*src[1] + E*src[step+1] + bias));\ yading@10: OP(dst[2], (A*src[2] + E*src[step+2] + bias));\ yading@10: OP(dst[3], (A*src[3] + E*src[step+3] + bias));\ yading@10: dst += stride;\ yading@10: src += stride;\ yading@10: }\ yading@10: }\ yading@10: }\ yading@10: \ yading@10: static void OPNAME ## rv40_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ yading@10: const int A = (8-x) * (8-y);\ yading@10: const int B = ( x) * (8-y);\ yading@10: const int C = (8-x) * ( y);\ yading@10: const int D = ( x) * ( y);\ yading@10: int i;\ yading@10: int bias = rv40_bias[y>>1][x>>1];\ yading@10: \ yading@10: av_assert2(x<8 && y<8 && x>=0 && y>=0);\ yading@10: \ yading@10: if(D){\ yading@10: for(i = 0; i < h; i++){\ yading@10: OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\ yading@10: OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\ yading@10: OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\ yading@10: OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\ yading@10: OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\ yading@10: OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\ yading@10: OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\ yading@10: OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\ yading@10: dst += stride;\ yading@10: src += stride;\ yading@10: }\ yading@10: }else{\ yading@10: const int E = B + C;\ yading@10: const int step = C ? stride : 1;\ yading@10: for(i = 0; i < h; i++){\ yading@10: OP(dst[0], (A*src[0] + E*src[step+0] + bias));\ yading@10: OP(dst[1], (A*src[1] + E*src[step+1] + bias));\ yading@10: OP(dst[2], (A*src[2] + E*src[step+2] + bias));\ yading@10: OP(dst[3], (A*src[3] + E*src[step+3] + bias));\ yading@10: OP(dst[4], (A*src[4] + E*src[step+4] + bias));\ yading@10: OP(dst[5], (A*src[5] + E*src[step+5] + bias));\ yading@10: OP(dst[6], (A*src[6] + E*src[step+6] + bias));\ yading@10: OP(dst[7], (A*src[7] + E*src[step+7] + bias));\ yading@10: dst += stride;\ yading@10: src += stride;\ yading@10: }\ yading@10: }\ yading@10: } yading@10: yading@10: #define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1) yading@10: #define op_put(a, b) a = ((b)>>6) yading@10: yading@10: RV40_CHROMA_MC(put_, op_put) yading@10: RV40_CHROMA_MC(avg_, op_avg) yading@10: yading@10: #define RV40_WEIGHT_FUNC(size) \ yading@10: static void rv40_weight_func_rnd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\ yading@10: {\ yading@10: int i, j;\ yading@10: \ yading@10: for (j = 0; j < size; j++) {\ yading@10: for (i = 0; i < size; i++)\ yading@10: dst[i] = (((w2 * src1[i]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\ yading@10: src1 += stride;\ yading@10: src2 += stride;\ yading@10: dst += stride;\ yading@10: }\ yading@10: }\ yading@10: static void rv40_weight_func_nornd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\ yading@10: {\ yading@10: int i, j;\ yading@10: \ yading@10: for (j = 0; j < size; j++) {\ yading@10: for (i = 0; i < size; i++)\ yading@10: dst[i] = (w2 * src1[i] + w1 * src2[i] + 0x10) >> 5;\ yading@10: src1 += stride;\ yading@10: src2 += stride;\ yading@10: dst += stride;\ yading@10: }\ yading@10: } yading@10: yading@10: RV40_WEIGHT_FUNC(16) yading@10: RV40_WEIGHT_FUNC(8) yading@10: yading@10: /** yading@10: * dither values for deblocking filter - left/top values yading@10: */ yading@10: static const uint8_t rv40_dither_l[16] = { yading@10: 0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30, yading@10: 0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40 yading@10: }; yading@10: yading@10: /** yading@10: * dither values for deblocking filter - right/bottom values yading@10: */ yading@10: static const uint8_t rv40_dither_r[16] = { yading@10: 0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40, yading@10: 0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40 yading@10: }; yading@10: yading@10: #define CLIP_SYMM(a, b) av_clip(a, -(b), b) yading@10: /** yading@10: * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1 yading@10: */ yading@10: static av_always_inline void rv40_weak_loop_filter(uint8_t *src, yading@10: const int step, yading@10: const ptrdiff_t stride, yading@10: const int filter_p1, yading@10: const int filter_q1, yading@10: const int alpha, yading@10: const int beta, yading@10: const int lim_p0q0, yading@10: const int lim_q1, yading@10: const int lim_p1) yading@10: { yading@10: const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; yading@10: int i, t, u, diff; yading@10: yading@10: for (i = 0; i < 4; i++, src += stride) { yading@10: int diff_p1p0 = src[-2*step] - src[-1*step]; yading@10: int diff_q1q0 = src[ 1*step] - src[ 0*step]; yading@10: int diff_p1p2 = src[-2*step] - src[-3*step]; yading@10: int diff_q1q2 = src[ 1*step] - src[ 2*step]; yading@10: yading@10: t = src[0*step] - src[-1*step]; yading@10: if (!t) yading@10: continue; yading@10: yading@10: u = (alpha * FFABS(t)) >> 7; yading@10: if (u > 3 - (filter_p1 && filter_q1)) yading@10: continue; yading@10: yading@10: t <<= 2; yading@10: if (filter_p1 && filter_q1) yading@10: t += src[-2*step] - src[1*step]; yading@10: yading@10: diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0); yading@10: src[-1*step] = cm[src[-1*step] + diff]; yading@10: src[ 0*step] = cm[src[ 0*step] - diff]; yading@10: yading@10: if (filter_p1 && FFABS(diff_p1p2) <= beta) { yading@10: t = (diff_p1p0 + diff_p1p2 - diff) >> 1; yading@10: src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)]; yading@10: } yading@10: yading@10: if (filter_q1 && FFABS(diff_q1q2) <= beta) { yading@10: t = (diff_q1q0 + diff_q1q2 + diff) >> 1; yading@10: src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)]; yading@10: } yading@10: } yading@10: } yading@10: yading@10: static void rv40_h_weak_loop_filter(uint8_t *src, const ptrdiff_t stride, yading@10: const int filter_p1, const int filter_q1, yading@10: const int alpha, const int beta, yading@10: const int lim_p0q0, const int lim_q1, yading@10: const int lim_p1) yading@10: { yading@10: rv40_weak_loop_filter(src, stride, 1, filter_p1, filter_q1, yading@10: alpha, beta, lim_p0q0, lim_q1, lim_p1); yading@10: } yading@10: yading@10: static void rv40_v_weak_loop_filter(uint8_t *src, const ptrdiff_t stride, yading@10: const int filter_p1, const int filter_q1, yading@10: const int alpha, const int beta, yading@10: const int lim_p0q0, const int lim_q1, yading@10: const int lim_p1) yading@10: { yading@10: rv40_weak_loop_filter(src, 1, stride, filter_p1, filter_q1, yading@10: alpha, beta, lim_p0q0, lim_q1, lim_p1); yading@10: } yading@10: yading@10: static av_always_inline void rv40_strong_loop_filter(uint8_t *src, yading@10: const int step, yading@10: const ptrdiff_t stride, yading@10: const int alpha, yading@10: const int lims, yading@10: const int dmode, yading@10: const int chroma) yading@10: { yading@10: int i; yading@10: yading@10: for(i = 0; i < 4; i++, src += stride){ yading@10: int sflag, p0, q0, p1, q1; yading@10: int t = src[0*step] - src[-1*step]; yading@10: yading@10: if (!t) yading@10: continue; yading@10: yading@10: sflag = (alpha * FFABS(t)) >> 7; yading@10: if (sflag > 1) yading@10: continue; yading@10: yading@10: p0 = (25*src[-3*step] + 26*src[-2*step] + 26*src[-1*step] + yading@10: 26*src[ 0*step] + 25*src[ 1*step] + yading@10: rv40_dither_l[dmode + i]) >> 7; yading@10: yading@10: q0 = (25*src[-2*step] + 26*src[-1*step] + 26*src[ 0*step] + yading@10: 26*src[ 1*step] + 25*src[ 2*step] + yading@10: rv40_dither_r[dmode + i]) >> 7; yading@10: yading@10: if (sflag) { yading@10: p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims); yading@10: q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims); yading@10: } yading@10: yading@10: p1 = (25*src[-4*step] + 26*src[-3*step] + 26*src[-2*step] + 26*p0 + yading@10: 25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7; yading@10: q1 = (25*src[-1*step] + 26*q0 + 26*src[ 1*step] + 26*src[ 2*step] + yading@10: 25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7; yading@10: yading@10: if (sflag) { yading@10: p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims); yading@10: q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims); yading@10: } yading@10: yading@10: src[-2*step] = p1; yading@10: src[-1*step] = p0; yading@10: src[ 0*step] = q0; yading@10: src[ 1*step] = q1; yading@10: yading@10: if(!chroma){ yading@10: src[-3*step] = (25*src[-1*step] + 26*src[-2*step] + yading@10: 51*src[-3*step] + 26*src[-4*step] + 64) >> 7; yading@10: src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] + yading@10: 51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7; yading@10: } yading@10: } yading@10: } yading@10: yading@10: static void rv40_h_strong_loop_filter(uint8_t *src, const ptrdiff_t stride, yading@10: const int alpha, const int lims, yading@10: const int dmode, const int chroma) yading@10: { yading@10: rv40_strong_loop_filter(src, stride, 1, alpha, lims, dmode, chroma); yading@10: } yading@10: yading@10: static void rv40_v_strong_loop_filter(uint8_t *src, const ptrdiff_t stride, yading@10: const int alpha, const int lims, yading@10: const int dmode, const int chroma) yading@10: { yading@10: rv40_strong_loop_filter(src, 1, stride, alpha, lims, dmode, chroma); yading@10: } yading@10: yading@10: static av_always_inline int rv40_loop_filter_strength(uint8_t *src, yading@10: int step, ptrdiff_t stride, yading@10: int beta, int beta2, yading@10: int edge, yading@10: int *p1, int *q1) yading@10: { yading@10: int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0; yading@10: int strong0 = 0, strong1 = 0; yading@10: uint8_t *ptr; yading@10: int i; yading@10: yading@10: for (i = 0, ptr = src; i < 4; i++, ptr += stride) { yading@10: sum_p1p0 += ptr[-2*step] - ptr[-1*step]; yading@10: sum_q1q0 += ptr[ 1*step] - ptr[ 0*step]; yading@10: } yading@10: yading@10: *p1 = FFABS(sum_p1p0) < (beta << 2); yading@10: *q1 = FFABS(sum_q1q0) < (beta << 2); yading@10: yading@10: if(!*p1 && !*q1) yading@10: return 0; yading@10: yading@10: if (!edge) yading@10: return 0; yading@10: yading@10: for (i = 0, ptr = src; i < 4; i++, ptr += stride) { yading@10: sum_p1p2 += ptr[-2*step] - ptr[-3*step]; yading@10: sum_q1q2 += ptr[ 1*step] - ptr[ 2*step]; yading@10: } yading@10: yading@10: strong0 = *p1 && (FFABS(sum_p1p2) < beta2); yading@10: strong1 = *q1 && (FFABS(sum_q1q2) < beta2); yading@10: yading@10: return strong0 && strong1; yading@10: } yading@10: yading@10: static int rv40_h_loop_filter_strength(uint8_t *src, ptrdiff_t stride, yading@10: int beta, int beta2, int edge, yading@10: int *p1, int *q1) yading@10: { yading@10: return rv40_loop_filter_strength(src, stride, 1, beta, beta2, edge, p1, q1); yading@10: } yading@10: yading@10: static int rv40_v_loop_filter_strength(uint8_t *src, ptrdiff_t stride, yading@10: int beta, int beta2, int edge, yading@10: int *p1, int *q1) yading@10: { yading@10: return rv40_loop_filter_strength(src, 1, stride, beta, beta2, edge, p1, q1); yading@10: } yading@10: yading@10: av_cold void ff_rv40dsp_init(RV34DSPContext *c) yading@10: { yading@10: H264QpelContext qpel; yading@10: yading@10: ff_rv34dsp_init(c); yading@10: ff_h264qpel_init(&qpel, 8); yading@10: yading@10: c->put_pixels_tab[0][ 0] = qpel.put_h264_qpel_pixels_tab[0][0]; yading@10: c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c; yading@10: c->put_pixels_tab[0][ 2] = qpel.put_h264_qpel_pixels_tab[0][2]; yading@10: c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c; yading@10: c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c; yading@10: c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c; yading@10: c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c; yading@10: c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c; yading@10: c->put_pixels_tab[0][ 8] = qpel.put_h264_qpel_pixels_tab[0][8]; yading@10: c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c; yading@10: c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c; yading@10: c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c; yading@10: c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c; yading@10: c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c; yading@10: c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c; yading@10: c->put_pixels_tab[0][15] = ff_put_rv40_qpel16_mc33_c; yading@10: c->avg_pixels_tab[0][ 0] = qpel.avg_h264_qpel_pixels_tab[0][0]; yading@10: c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c; yading@10: c->avg_pixels_tab[0][ 2] = qpel.avg_h264_qpel_pixels_tab[0][2]; yading@10: c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c; yading@10: c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c; yading@10: c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c; yading@10: c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c; yading@10: c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c; yading@10: c->avg_pixels_tab[0][ 8] = qpel.avg_h264_qpel_pixels_tab[0][8]; yading@10: c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c; yading@10: c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c; yading@10: c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c; yading@10: c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c; yading@10: c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c; yading@10: c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c; yading@10: c->avg_pixels_tab[0][15] = ff_avg_rv40_qpel16_mc33_c; yading@10: c->put_pixels_tab[1][ 0] = qpel.put_h264_qpel_pixels_tab[1][0]; yading@10: c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c; yading@10: c->put_pixels_tab[1][ 2] = qpel.put_h264_qpel_pixels_tab[1][2]; yading@10: c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c; yading@10: c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c; yading@10: c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c; yading@10: c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c; yading@10: c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c; yading@10: c->put_pixels_tab[1][ 8] = qpel.put_h264_qpel_pixels_tab[1][8]; yading@10: c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c; yading@10: c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c; yading@10: c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c; yading@10: c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c; yading@10: c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c; yading@10: c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c; yading@10: c->put_pixels_tab[1][15] = ff_put_rv40_qpel8_mc33_c; yading@10: c->avg_pixels_tab[1][ 0] = qpel.avg_h264_qpel_pixels_tab[1][0]; yading@10: c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c; yading@10: c->avg_pixels_tab[1][ 2] = qpel.avg_h264_qpel_pixels_tab[1][2]; yading@10: c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c; yading@10: c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c; yading@10: c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c; yading@10: c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c; yading@10: c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c; yading@10: c->avg_pixels_tab[1][ 8] = qpel.avg_h264_qpel_pixels_tab[1][8]; yading@10: c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c; yading@10: c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c; yading@10: c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c; yading@10: c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c; yading@10: c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c; yading@10: c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c; yading@10: c->avg_pixels_tab[1][15] = ff_avg_rv40_qpel8_mc33_c; yading@10: yading@10: c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c; yading@10: c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c; yading@10: c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c; yading@10: c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c; yading@10: yading@10: c->rv40_weight_pixels_tab[0][0] = rv40_weight_func_rnd_16; yading@10: c->rv40_weight_pixels_tab[0][1] = rv40_weight_func_rnd_8; yading@10: c->rv40_weight_pixels_tab[1][0] = rv40_weight_func_nornd_16; yading@10: c->rv40_weight_pixels_tab[1][1] = rv40_weight_func_nornd_8; yading@10: yading@10: c->rv40_weak_loop_filter[0] = rv40_h_weak_loop_filter; yading@10: c->rv40_weak_loop_filter[1] = rv40_v_weak_loop_filter; yading@10: c->rv40_strong_loop_filter[0] = rv40_h_strong_loop_filter; yading@10: c->rv40_strong_loop_filter[1] = rv40_v_strong_loop_filter; yading@10: c->rv40_loop_filter_strength[0] = rv40_h_loop_filter_strength; yading@10: c->rv40_loop_filter_strength[1] = rv40_v_loop_filter_strength; yading@10: yading@10: if (ARCH_X86) yading@10: ff_rv40dsp_init_x86(c); yading@10: if (ARCH_ARM) yading@10: ff_rv40dsp_init_arm(c); yading@10: }