rv40: rearrange loop filter functions

This splits the loop filter functions into smaller, more SIMD-friendly
functions.

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

1 files changed, 176 insertions, 120 deletions

diff --git a/libavcodec/rv40dsp.c b/libavcodec/rv40dsp.c
index 06bdf18c42..909ccd1732 100644
--- a/libavcodec/rv40dsp.c
+++ b/libavcodec/rv40dsp.c
@@ -314,142 +314,194 @@ 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 int 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)];
+    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 int 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 int 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 int 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 int 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 int 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, int 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, int 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, int 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) {
@@ -529,8 +581,12 @@ 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;
+    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 (HAVE_MMX)
         ff_rv40dsp_init_x86(c, dsp);