rv40: move loop filter to rv34dsp context

Signed-off-by: Mans Rullgard <mans@mansr.com>

libavcodec/rv34dsp.h

@@ -36,6 +36,11 @@ typedef void (*rv40_weight_func)(uint8_t *dst/*align width (8 or 16)*/,
 
 typedef void (*rv34_inv_transform_func)(DCTELEM *block);
 
+typedef void (*rv40_loop_filter_func)(uint8_t *src, int stride, int dmode,
+                                      int lim_q1, int lim_p1, int alpha,
+                                      int beta, int beta2, int chroma,
+                                      int edge);
+
 typedef struct RV34DSPContext {
     qpel_mc_func put_pixels_tab[4][16];
     qpel_mc_func avg_pixels_tab[4][16];
@@ -43,6 +48,8 @@ typedef struct RV34DSPContext {
     h264_chroma_mc_func avg_chroma_pixels_tab[3];
     rv40_weight_func rv40_weight_pixels_tab[2];
     rv34_inv_transform_func rv34_inv_transform_tab[2];
+    rv40_loop_filter_func rv40_h_loop_filter;
+    rv40_loop_filter_func rv40_v_loop_filter;
 } RV34DSPContext;
 
 void ff_rv30dsp_init(RV34DSPContext *c, DSPContext* dsp);

libavcodec/rv40data.h

@@ -68,20 +68,6 @@ static const uint8_t rv40_luma_dc_quant[2][32] = {
  * @name Coefficients used by the RV40 loop filter
  * @{
  */
-/**
- * dither values for deblocking filter - left/top values
- */
-static const uint8_t rv40_dither_l[16] = {
-    0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,
-    0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40
-};
-/**
- * dither values for deblocking filter - right/bottom values
- */
-static const uint8_t rv40_dither_r[16] = {
-    0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,
-    0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40
-};
 
 /** alpha parameter for RV40 loop filter - almost the same as in JVT-A003r1 */
 static const uint8_t rv40_alpha_tab[32] = {

libavcodec/rv40dsp.c

@@ -294,6 +294,164 @@ static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src
 RV40_WEIGHT_FUNC(16)
 RV40_WEIGHT_FUNC(8)
 
+/**
+ * dither values for deblocking filter - left/top values
+ */
+static const uint8_t rv40_dither_l[16] = {
+    0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,
+    0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40
+};
+
+/**
+ * dither values for deblocking filter - right/bottom values
+ */
+static const uint8_t rv40_dither_r[16] = {
+    0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,
+    0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40
+};
+
+#define CLIP_SYMM(a, b) av_clip(a, -(b), b)
+/**
+ * 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)
+{
+    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)];
+    }
+    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 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)
+{
+    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;
+    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];
+    }
+
+    if(edge){
+        flag_strong0 = filter_p1 && (FFABS(sum_p1p2) < beta2);
+        flag_strong1 = filter_q1 && (FFABS(sum_q1q2) < beta2);
+    }else{
+        flag_strong0 = flag_strong1 = 0;
+    }
+
+    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]);
+    }
+}
+
+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 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);
+}
+
 av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
 
     ff_rv34dsp_init(c, dsp);
@@ -371,6 +529,9 @@ av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
     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);
 }