hevc: simplify SAO computation, delay from one row its computation

(cherry picked from commit f2c5f647cec786df26f442a85e6d685a131a50c9)
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libavcodec/hevc.c

@@ -2092,7 +2092,7 @@ static int hls_decode_entry(AVCodecContext *avctxt, void *isFilterThread)
 
     if (x_ctb + ctb_size >= s->sps->width &&
         y_ctb + ctb_size >= s->sps->height)
-        ff_hevc_hls_filter(s, x_ctb, y_ctb);
+        ff_hevc_hls_filter(s, x_ctb, y_ctb, ctb_size);
 
     return ctb_addr_ts;
 }
@@ -2167,7 +2167,7 @@ static int hls_decode_entry_wpp(AVCodecContext *avctxt, void *input_ctb_row, int
         }
 
         if ((x_ctb+ctb_size) >= s->sps->width && (y_ctb+ctb_size) >= s->sps->height ) {
-            ff_hevc_hls_filter(s, x_ctb, y_ctb);
+            ff_hevc_hls_filter(s, x_ctb, y_ctb, ctb_size);
             ff_thread_report_progress2(s->avctx, ctb_row , thread, SHIFT_CTB_WPP);
             return ctb_addr_ts;
         }

libavcodec/hevc.h

@@ -1001,7 +1001,7 @@ void ff_hevc_deblocking_boundary_strengths(HEVCContext *s, int x0, int y0,
                                            int log2_trafo_size);
 int ff_hevc_cu_qp_delta_sign_flag(HEVCContext *s);
 int ff_hevc_cu_qp_delta_abs(HEVCContext *s);
-void ff_hevc_hls_filter(HEVCContext *s, int x, int y);
+void ff_hevc_hls_filter(HEVCContext *s, int x, int y, int ctb_size);
 void ff_hevc_hls_filters(HEVCContext *s, int x_ctb, int y_ctb, int ctb_size);
 void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0,
                                  int log2_trafo_size, enum ScanType scan_idx,

libavcodec/hevcdsp.c

@@ -207,15 +207,9 @@ void ff_hevc_dsp_init(HEVCDSPContext *hevcdsp, int bit_depth)
     hevcdsp->transform_dc_add[2]    = FUNC(transform_16x16_dc_add, depth);  \
     hevcdsp->transform_dc_add[3]    = FUNC(transform_32x32_dc_add, depth);  \
                                                                             \
-    hevcdsp->sao_band_filter[0] = FUNC(sao_band_filter_0, depth);           \
-    hevcdsp->sao_band_filter[1] = FUNC(sao_band_filter_1, depth);           \
-    hevcdsp->sao_band_filter[2] = FUNC(sao_band_filter_2, depth);           \
-    hevcdsp->sao_band_filter[3] = FUNC(sao_band_filter_3, depth);           \
-                                                                            \
-    hevcdsp->sao_edge_filter[0] = FUNC(sao_edge_filter_0, depth);           \
-    hevcdsp->sao_edge_filter[1] = FUNC(sao_edge_filter_1, depth);           \
-    hevcdsp->sao_edge_filter[2] = FUNC(sao_edge_filter_2, depth);           \
-    hevcdsp->sao_edge_filter[3] = FUNC(sao_edge_filter_3, depth);           \
+    hevcdsp->sao_band_filter    = FUNC(sao_band_filter_0, depth);              \
+    hevcdsp->sao_edge_filter[0] = FUNC(sao_edge_filter_0, depth);              \
+    hevcdsp->sao_edge_filter[1] = FUNC(sao_edge_filter_1, depth);              \
                                                                                \
     QPEL_FUNCS(depth);                                                         \
     QPEL_UNI_FUNCS(depth);                                                     \

libavcodec/hevcdsp.h

@@ -54,13 +54,14 @@ typedef struct HEVCDSPContext {
 
     void (*transform_dc_add[4])(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
 
-    void (*sao_band_filter[4])(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
-                               struct SAOParams *sao, int *borders,
-                               int width, int height, int c_idx);
-    void (*sao_edge_filter[4])(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
-                               struct SAOParams *sao, int *borders, int width,
-                               int height, int c_idx, uint8_t vert_edge,
-                               uint8_t horiz_edge, uint8_t diag_edge);
+    void (*sao_band_filter)(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride,
+                            struct SAOParams *sao, int *borders,
+                            int width, int height, int c_idx);
+
+    void (*sao_edge_filter[2])(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride,
+                               struct SAOParams *sao, int *borders, int _width,
+                               int _height, int c_idx, uint8_t *vert_edge,
+                               uint8_t *horiz_edge, uint8_t *diag_edge);
 
     void (*put_hevc_qpel[10][2][2])(int16_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride,
                                     int height, intptr_t mx, intptr_t my, int width);