X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Frv40dsp.c;h=da3efb44424efb8b4f82e8d8f031d0d060ff5376;hb=36779a84051eae6744cc936d91b1d428143665ba;hp=f193b6050d6e18a8cfc0fdc97c30e50420076725;hpb=bb8a6e03cc641035b3fd4dc7c8c77f2dc92b41d5;p=ffmpeg diff --git a/libavcodec/rv40dsp.c b/libavcodec/rv40dsp.c index f193b6050d6..da3efb44424 100644 --- a/libavcodec/rv40dsp.c +++ b/libavcodec/rv40dsp.c @@ -24,14 +24,19 @@ * RV40 decoder motion compensation functions */ +#include "libavutil/common.h" +#include "libavutil/intreadwrite.h" #include "avcodec.h" -#include "dsputil.h" +#include "h264qpel.h" +#include "mathops.h" +#include "pixels.h" +#include "rnd_avg.h" #include "rv34dsp.h" #define RV40_LOWPASS(OPNAME, OP) \ -static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ +static void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\ const int h, const int C1, const int C2, const int SHIFT){\ - uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ + const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;\ int i;\ for(i = 0; i < h; i++)\ {\ @@ -48,9 +53,9 @@ static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, }\ }\ \ -static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ +static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\ const int w, const int C1, const int C2, const int SHIFT){\ - uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ + const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;\ int i;\ for(i = 0; i < w; i++)\ {\ @@ -80,7 +85,7 @@ static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstSt }\ }\ \ -static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ +static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\ const int w, const int C1, const int C2, const int SHIFT){\ OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\ OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\ @@ -90,7 +95,7 @@ static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstS OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\ }\ \ -static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\ +static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\ const int h, const int C1, const int C2, const int SHIFT){\ OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\ OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\ @@ -102,72 +107,84 @@ static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstS \ #define RV40_MC(OPNAME, SIZE) \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\ OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\ }\ \ -static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\ +static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\ +{\ uint8_t full[SIZE*(SIZE+5)];\ uint8_t * const full_mid = full + SIZE*2;\ put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\ @@ -189,6 +206,83 @@ RV40_MC(put_, 16) RV40_MC(avg_, 8) RV40_MC(avg_, 16) +#define PIXOP2(OPNAME, OP) \ +static inline void OPNAME ## _pixels8_xy2_8_c(uint8_t *block, \ + const uint8_t *pixels, \ + ptrdiff_t line_size, \ + int h) \ +{ \ + /* FIXME HIGH BIT DEPTH */ \ + int j; \ + \ + for (j = 0; j < 2; j++) { \ + int i; \ + const uint32_t a = AV_RN32(pixels); \ + const uint32_t b = AV_RN32(pixels + 1); \ + uint32_t l0 = (a & 0x03030303UL) + \ + (b & 0x03030303UL) + \ + 0x02020202UL; \ + uint32_t h0 = ((a & 0xFCFCFCFCUL) >> 2) + \ + ((b & 0xFCFCFCFCUL) >> 2); \ + uint32_t l1, h1; \ + \ + pixels += line_size; \ + for (i = 0; i < h; i += 2) { \ + uint32_t a = AV_RN32(pixels); \ + uint32_t b = AV_RN32(pixels + 1); \ + l1 = (a & 0x03030303UL) + \ + (b & 0x03030303UL); \ + h1 = ((a & 0xFCFCFCFCUL) >> 2) + \ + ((b & 0xFCFCFCFCUL) >> 2); \ + OP(*((uint32_t *) block), \ + h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL)); \ + pixels += line_size; \ + block += line_size; \ + a = AV_RN32(pixels); \ + b = AV_RN32(pixels + 1); \ + l0 = (a & 0x03030303UL) + \ + (b & 0x03030303UL) + \ + 0x02020202UL; \ + h0 = ((a & 0xFCFCFCFCUL) >> 2) + \ + ((b & 0xFCFCFCFCUL) >> 2); \ + OP(*((uint32_t *) block), \ + h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL)); \ + pixels += line_size; \ + block += line_size; \ + } \ + pixels += 4 - line_size * (h + 1); \ + block += 4 - line_size * h; \ + } \ +} \ + \ +CALL_2X_PIXELS(OPNAME ## _pixels16_xy2_8_c, \ + OPNAME ## _pixels8_xy2_8_c, \ + 8) \ + +#define op_avg(a, b) a = rnd_avg32(a, b) +#define op_put(a, b) a = b +PIXOP2(avg, op_avg) +PIXOP2(put, op_put) +#undef op_avg +#undef op_put + +static void put_rv40_qpel16_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) +{ + put_pixels16_xy2_8_c(dst, src, stride, 16); +} +static void avg_rv40_qpel16_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) +{ + avg_pixels16_xy2_8_c(dst, src, stride, 16); +} +static void put_rv40_qpel8_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) +{ + put_pixels8_xy2_8_c(dst, src, stride, 8); +} +static void avg_rv40_qpel8_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) +{ + avg_pixels8_xy2_8_c(dst, src, stride, 8); +} + static const int rv40_bias[4][4] = { { 0, 16, 32, 16 }, { 32, 28, 32, 28 }, @@ -278,7 +372,7 @@ RV40_CHROMA_MC(put_, op_put) RV40_CHROMA_MC(avg_, op_avg) #define RV40_WEIGHT_FUNC(size) \ -static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, int stride)\ +static void rv40_weight_func_rnd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\ {\ int i, j;\ \ @@ -289,6 +383,18 @@ static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src src2 += stride;\ dst += stride;\ }\ +}\ +static void rv40_weight_func_nornd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\ +{\ + int i, j;\ +\ + for (j = 0; j < size; j++) {\ + for (i = 0; i < size; i++)\ + dst[i] = (w2 * src1[i] + w1 * src2[i] + 0x10) >> 5;\ + src1 += stride;\ + src2 += stride;\ + dst += stride;\ + }\ } RV40_WEIGHT_FUNC(16) @@ -314,224 +420,289 @@ static const uint8_t rv40_dither_r[16] = { /** * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1 */ -static inline void rv40_weak_loop_filter(uint8_t *src, const int step, - const int filter_p1, const int filter_q1, - const int alpha, const int beta, - const int lim_p0q0, - const int lim_q1, const int lim_p1, - const int diff_p1p0, const int diff_q1q0, - const int diff_p1p2, const int diff_q1q2) +static av_always_inline void rv40_weak_loop_filter(uint8_t *src, + const int step, + const ptrdiff_t stride, + const int filter_p1, + const int filter_q1, + const int alpha, + const int beta, + const int lim_p0q0, + const int lim_q1, + const int lim_p1) { - uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; - int t, u, diff; - - t = src[0*step] - src[-1*step]; - if(!t) - return; - u = (alpha * FFABS(t)) >> 7; - if(u > 3 - (filter_p1 && filter_q1)) - return; - - t <<= 2; - if(filter_p1 && filter_q1) - t += src[-2*step] - src[1*step]; - diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0); - src[-1*step] = cm[src[-1*step] + diff]; - src[ 0*step] = cm[src[ 0*step] - diff]; - if(FFABS(diff_p1p2) <= beta && filter_p1){ - t = (diff_p1p0 + diff_p1p2 - diff) >> 1; - src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)]; + const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; + int i, t, u, diff; + + for (i = 0; i < 4; i++, src += stride) { + int diff_p1p0 = src[-2*step] - src[-1*step]; + int diff_q1q0 = src[ 1*step] - src[ 0*step]; + int diff_p1p2 = src[-2*step] - src[-3*step]; + int diff_q1q2 = src[ 1*step] - src[ 2*step]; + + t = src[0*step] - src[-1*step]; + if (!t) + continue; + + u = (alpha * FFABS(t)) >> 7; + if (u > 3 - (filter_p1 && filter_q1)) + continue; + + t <<= 2; + if (filter_p1 && filter_q1) + t += src[-2*step] - src[1*step]; + + diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0); + src[-1*step] = cm[src[-1*step] + diff]; + src[ 0*step] = cm[src[ 0*step] - diff]; + + if (filter_p1 && FFABS(diff_p1p2) <= beta) { + t = (diff_p1p0 + diff_p1p2 - diff) >> 1; + src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)]; + } + + if (filter_q1 && FFABS(diff_q1q2) <= beta) { + t = (diff_q1q0 + diff_q1q2 + diff) >> 1; + src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)]; + } } - if(FFABS(diff_q1q2) <= beta && filter_q1){ - t = (diff_q1q0 + diff_q1q2 + diff) >> 1; - src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)]; +} + +static void rv40_h_weak_loop_filter(uint8_t *src, const ptrdiff_t stride, + const int filter_p1, const int filter_q1, + const int alpha, const int beta, + const int lim_p0q0, const int lim_q1, + const int lim_p1) +{ + rv40_weak_loop_filter(src, stride, 1, filter_p1, filter_q1, + alpha, beta, lim_p0q0, lim_q1, lim_p1); +} + +static void rv40_v_weak_loop_filter(uint8_t *src, const ptrdiff_t stride, + const int filter_p1, const int filter_q1, + const int alpha, const int beta, + const int lim_p0q0, const int lim_q1, + const int lim_p1) +{ + rv40_weak_loop_filter(src, 1, stride, filter_p1, filter_q1, + alpha, beta, lim_p0q0, lim_q1, lim_p1); +} + +static av_always_inline void rv40_strong_loop_filter(uint8_t *src, + const int step, + const ptrdiff_t stride, + const int alpha, + const int lims, + const int dmode, + const int chroma) +{ + int i; + + for(i = 0; i < 4; i++, src += stride){ + int sflag, p0, q0, p1, q1; + int t = src[0*step] - src[-1*step]; + + if (!t) + continue; + + sflag = (alpha * FFABS(t)) >> 7; + if (sflag > 1) + continue; + + p0 = (25*src[-3*step] + 26*src[-2*step] + 26*src[-1*step] + + 26*src[ 0*step] + 25*src[ 1*step] + + rv40_dither_l[dmode + i]) >> 7; + + q0 = (25*src[-2*step] + 26*src[-1*step] + 26*src[ 0*step] + + 26*src[ 1*step] + 25*src[ 2*step] + + rv40_dither_r[dmode + i]) >> 7; + + if (sflag) { + p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims); + q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims); + } + + p1 = (25*src[-4*step] + 26*src[-3*step] + 26*src[-2*step] + 26*p0 + + 25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7; + q1 = (25*src[-1*step] + 26*q0 + 26*src[ 1*step] + 26*src[ 2*step] + + 25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7; + + if (sflag) { + p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims); + q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims); + } + + src[-2*step] = p1; + src[-1*step] = p0; + src[ 0*step] = q0; + src[ 1*step] = q1; + + if(!chroma){ + src[-3*step] = (25*src[-1*step] + 26*src[-2*step] + + 51*src[-3*step] + 26*src[-4*step] + 64) >> 7; + src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] + + 51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7; + } } } -static av_always_inline void rv40_adaptive_loop_filter(uint8_t *src, const int step, - const int stride, const int dmode, - const int lim_q1, const int lim_p1, - const int alpha, - const int beta, const int beta2, - const int chroma, const int edge) +static void rv40_h_strong_loop_filter(uint8_t *src, const ptrdiff_t stride, + const int alpha, const int lims, + const int dmode, const int chroma) +{ + rv40_strong_loop_filter(src, stride, 1, alpha, lims, dmode, chroma); +} + +static void rv40_v_strong_loop_filter(uint8_t *src, const ptrdiff_t stride, + const int alpha, const int lims, + const int dmode, const int chroma) +{ + rv40_strong_loop_filter(src, 1, stride, alpha, lims, dmode, chroma); +} + +static av_always_inline int rv40_loop_filter_strength(uint8_t *src, + int step, ptrdiff_t stride, + int beta, int beta2, + int edge, + int *p1, int *q1) { - int diff_p1p0[4], diff_q1q0[4], diff_p1p2[4], diff_q1q2[4]; int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0; + int strong0 = 0, strong1 = 0; uint8_t *ptr; - int flag_strong0 = 1, flag_strong1 = 1; - int filter_p1, filter_q1; int i; - int lims; - for(i = 0, ptr = src; i < 4; i++, ptr += stride){ - diff_p1p0[i] = ptr[-2*step] - ptr[-1*step]; - diff_q1q0[i] = ptr[ 1*step] - ptr[ 0*step]; - sum_p1p0 += diff_p1p0[i]; - sum_q1q0 += diff_q1q0[i]; - } - filter_p1 = FFABS(sum_p1p0) < (beta<<2); - filter_q1 = FFABS(sum_q1q0) < (beta<<2); - if(!filter_p1 && !filter_q1) - return; - - for(i = 0, ptr = src; i < 4; i++, ptr += stride){ - diff_p1p2[i] = ptr[-2*step] - ptr[-3*step]; - diff_q1q2[i] = ptr[ 1*step] - ptr[ 2*step]; - sum_p1p2 += diff_p1p2[i]; - sum_q1q2 += diff_q1q2[i]; + for (i = 0, ptr = src; i < 4; i++, ptr += stride) { + sum_p1p0 += ptr[-2*step] - ptr[-1*step]; + sum_q1q0 += ptr[ 1*step] - ptr[ 0*step]; } - if(edge){ - flag_strong0 = filter_p1 && (FFABS(sum_p1p2) < beta2); - flag_strong1 = filter_q1 && (FFABS(sum_q1q2) < beta2); - }else{ - flag_strong0 = flag_strong1 = 0; - } + *p1 = FFABS(sum_p1p0) < (beta << 2); + *q1 = FFABS(sum_q1q0) < (beta << 2); - lims = filter_p1 + filter_q1 + ((lim_q1 + lim_p1) >> 1) + 1; - if(flag_strong0 && flag_strong1){ /* strong filtering */ - for(i = 0; i < 4; i++, src += stride){ - int sflag, p0, q0, p1, q1; - int t = src[0*step] - src[-1*step]; - - if(!t) continue; - sflag = (alpha * FFABS(t)) >> 7; - if(sflag > 1) continue; - - p0 = (25*src[-3*step] + 26*src[-2*step] - + 26*src[-1*step] - + 26*src[ 0*step] + 25*src[ 1*step] + rv40_dither_l[dmode + i]) >> 7; - q0 = (25*src[-2*step] + 26*src[-1*step] - + 26*src[ 0*step] - + 26*src[ 1*step] + 25*src[ 2*step] + rv40_dither_r[dmode + i]) >> 7; - if(sflag){ - p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims); - q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims); - } - p1 = (25*src[-4*step] + 26*src[-3*step] - + 26*src[-2*step] - + 26*p0 + 25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7; - q1 = (25*src[-1*step] + 26*q0 - + 26*src[ 1*step] - + 26*src[ 2*step] + 25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7; - if(sflag){ - p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims); - q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims); - } - src[-2*step] = p1; - src[-1*step] = p0; - src[ 0*step] = q0; - src[ 1*step] = q1; - if(!chroma){ - src[-3*step] = (25*src[-1*step] + 26*src[-2*step] + 51*src[-3*step] + 26*src[-4*step] + 64) >> 7; - src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] + 51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7; - } - } - }else if(filter_p1 && filter_q1){ - for(i = 0; i < 4; i++, src += stride) - rv40_weak_loop_filter(src, step, 1, 1, alpha, beta, lims, lim_q1, lim_p1, - diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]); - }else{ - for(i = 0; i < 4; i++, src += stride) - rv40_weak_loop_filter(src, step, filter_p1, filter_q1, - alpha, beta, lims>>1, lim_q1>>1, lim_p1>>1, - diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]); + if(!*p1 && !*q1) + return 0; + + if (!edge) + return 0; + + for (i = 0, ptr = src; i < 4; i++, ptr += stride) { + sum_p1p2 += ptr[-2*step] - ptr[-3*step]; + sum_q1q2 += ptr[ 1*step] - ptr[ 2*step]; } + + strong0 = *p1 && (FFABS(sum_p1p2) < beta2); + strong1 = *q1 && (FFABS(sum_q1q2) < beta2); + + return strong0 && strong1; } -static void rv40_v_loop_filter(uint8_t *src, int stride, int dmode, - int lim_q1, int lim_p1, - int alpha, int beta, int beta2, int chroma, int edge){ - rv40_adaptive_loop_filter(src, 1, stride, dmode, lim_q1, lim_p1, - alpha, beta, beta2, chroma, edge); +static int rv40_h_loop_filter_strength(uint8_t *src, ptrdiff_t stride, + int beta, int beta2, int edge, + int *p1, int *q1) +{ + return rv40_loop_filter_strength(src, stride, 1, beta, beta2, edge, p1, q1); } -static void rv40_h_loop_filter(uint8_t *src, int stride, int dmode, - int lim_q1, int lim_p1, - int alpha, int beta, int beta2, int chroma, int edge){ - rv40_adaptive_loop_filter(src, stride, 1, dmode, lim_q1, lim_p1, - alpha, beta, beta2, chroma, edge); + +static int rv40_v_loop_filter_strength(uint8_t *src, ptrdiff_t stride, + int beta, int beta2, int edge, + int *p1, int *q1) +{ + return rv40_loop_filter_strength(src, 1, stride, beta, beta2, edge, p1, q1); } -av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) { +av_cold void ff_rv40dsp_init(RV34DSPContext *c) +{ + H264QpelContext qpel; - ff_rv34dsp_init(c, dsp); + ff_rv34dsp_init(c); + ff_h264qpel_init(&qpel, 8); - c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0]; + c->put_pixels_tab[0][ 0] = qpel.put_h264_qpel_pixels_tab[0][0]; c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c; - c->put_pixels_tab[0][ 2] = dsp->put_h264_qpel_pixels_tab[0][2]; + c->put_pixels_tab[0][ 2] = qpel.put_h264_qpel_pixels_tab[0][2]; c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c; c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c; c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c; c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c; c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c; - c->put_pixels_tab[0][ 8] = dsp->put_h264_qpel_pixels_tab[0][8]; + c->put_pixels_tab[0][ 8] = qpel.put_h264_qpel_pixels_tab[0][8]; c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c; c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c; c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c; c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c; c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c; c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c; - c->put_pixels_tab[0][15] = ff_put_rv40_qpel16_mc33_c; - c->avg_pixels_tab[0][ 0] = dsp->avg_h264_qpel_pixels_tab[0][0]; + c->put_pixels_tab[0][15] = put_rv40_qpel16_mc33_c; + c->avg_pixels_tab[0][ 0] = qpel.avg_h264_qpel_pixels_tab[0][0]; c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c; - c->avg_pixels_tab[0][ 2] = dsp->avg_h264_qpel_pixels_tab[0][2]; + c->avg_pixels_tab[0][ 2] = qpel.avg_h264_qpel_pixels_tab[0][2]; c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c; c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c; c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c; c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c; c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c; - c->avg_pixels_tab[0][ 8] = dsp->avg_h264_qpel_pixels_tab[0][8]; + c->avg_pixels_tab[0][ 8] = qpel.avg_h264_qpel_pixels_tab[0][8]; c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c; c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c; c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c; c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c; c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c; c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c; - c->avg_pixels_tab[0][15] = ff_avg_rv40_qpel16_mc33_c; - c->put_pixels_tab[1][ 0] = dsp->put_h264_qpel_pixels_tab[1][0]; + c->avg_pixels_tab[0][15] = avg_rv40_qpel16_mc33_c; + c->put_pixels_tab[1][ 0] = qpel.put_h264_qpel_pixels_tab[1][0]; c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c; - c->put_pixels_tab[1][ 2] = dsp->put_h264_qpel_pixels_tab[1][2]; + c->put_pixels_tab[1][ 2] = qpel.put_h264_qpel_pixels_tab[1][2]; c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c; c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c; c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c; c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c; c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c; - c->put_pixels_tab[1][ 8] = dsp->put_h264_qpel_pixels_tab[1][8]; + c->put_pixels_tab[1][ 8] = qpel.put_h264_qpel_pixels_tab[1][8]; c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c; c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c; c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c; c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c; c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c; c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c; - c->put_pixels_tab[1][15] = ff_put_rv40_qpel8_mc33_c; - c->avg_pixels_tab[1][ 0] = dsp->avg_h264_qpel_pixels_tab[1][0]; + c->put_pixels_tab[1][15] = put_rv40_qpel8_mc33_c; + c->avg_pixels_tab[1][ 0] = qpel.avg_h264_qpel_pixels_tab[1][0]; c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c; - c->avg_pixels_tab[1][ 2] = dsp->avg_h264_qpel_pixels_tab[1][2]; + c->avg_pixels_tab[1][ 2] = qpel.avg_h264_qpel_pixels_tab[1][2]; c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c; c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c; c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c; c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c; c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c; - c->avg_pixels_tab[1][ 8] = dsp->avg_h264_qpel_pixels_tab[1][8]; + c->avg_pixels_tab[1][ 8] = qpel.avg_h264_qpel_pixels_tab[1][8]; c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c; c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c; c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c; c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c; c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c; c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c; - c->avg_pixels_tab[1][15] = ff_avg_rv40_qpel8_mc33_c; + c->avg_pixels_tab[1][15] = avg_rv40_qpel8_mc33_c; c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c; c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c; c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c; c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c; - c->rv40_weight_pixels_tab[0] = rv40_weight_func_16; - c->rv40_weight_pixels_tab[1] = rv40_weight_func_8; - - c->rv40_h_loop_filter = rv40_h_loop_filter; - c->rv40_v_loop_filter = rv40_v_loop_filter; - - if (HAVE_MMX) - ff_rv40dsp_init_x86(c, dsp); + c->rv40_weight_pixels_tab[0][0] = rv40_weight_func_rnd_16; + c->rv40_weight_pixels_tab[0][1] = rv40_weight_func_rnd_8; + c->rv40_weight_pixels_tab[1][0] = rv40_weight_func_nornd_16; + c->rv40_weight_pixels_tab[1][1] = rv40_weight_func_nornd_8; + + c->rv40_weak_loop_filter[0] = rv40_h_weak_loop_filter; + c->rv40_weak_loop_filter[1] = rv40_v_weak_loop_filter; + c->rv40_strong_loop_filter[0] = rv40_h_strong_loop_filter; + c->rv40_strong_loop_filter[1] = rv40_v_strong_loop_filter; + c->rv40_loop_filter_strength[0] = rv40_h_loop_filter_strength; + c->rv40_loop_filter_strength[1] = rv40_v_loop_filter_strength; + + if (ARCH_AARCH64) + ff_rv40dsp_init_aarch64(c); + if (ARCH_ARM) + ff_rv40dsp_init_arm(c); + if (ARCH_X86) + ff_rv40dsp_init_x86(c); }