Merge commit '32ff6432'

* commit '32ff6432':
  dsputil: remove avg_no_rnd_pixels8.

Conflicts:
	libavcodec/x86/dsputil_mmx.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>

libavcodec/alpha/dsputil_alpha.c

@@ -308,11 +308,6 @@ void ff_dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx)
     c->avg_pixels_tab[1][2] = avg_pixels_y2_axp;
     c->avg_pixels_tab[1][3] = avg_pixels_xy2_axp;
 
-    c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels_axp;
-    c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels_x2_axp;
-    c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels_y2_axp;
-    c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels_xy2_axp;
-
     c->clear_blocks = clear_blocks_axp;
     }
 

libavcodec/arm/dsputil_init_neon.c

@@ -58,9 +58,6 @@ void ff_avg_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int);
 void ff_avg_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
 void ff_avg_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
 void ff_avg_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
-void ff_avg_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
-void ff_avg_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
-void ff_avg_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
 
 void ff_add_pixels_clamped_neon(const int16_t *, uint8_t *, int);
 void ff_put_pixels_clamped_neon(const int16_t *, uint8_t *, int);
@@ -203,10 +200,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
         c->avg_no_rnd_pixels_tab[0][1] = ff_avg_pixels16_x2_no_rnd_neon;
         c->avg_no_rnd_pixels_tab[0][2] = ff_avg_pixels16_y2_no_rnd_neon;
         c->avg_no_rnd_pixels_tab[0][3] = ff_avg_pixels16_xy2_no_rnd_neon;
-        c->avg_no_rnd_pixels_tab[1][0] = ff_avg_pixels8_neon;
-        c->avg_no_rnd_pixels_tab[1][1] = ff_avg_pixels8_x2_no_rnd_neon;
-        c->avg_no_rnd_pixels_tab[1][2] = ff_avg_pixels8_y2_no_rnd_neon;
-        c->avg_no_rnd_pixels_tab[1][3] = ff_avg_pixels8_xy2_no_rnd_neon;
     }
 
     c->add_pixels_clamped = ff_add_pixels_clamped_neon;

libavcodec/arm/dsputil_neon.S

@@ -421,9 +421,9 @@ function ff_avg_h264_qpel8_mc00_neon, export=1
 endfunc
 
         pixfunc         avg_, pixels8,     avg=1
-        pixfunc2        avg_, pixels8_x2,  avg=1
-        pixfunc2        avg_, pixels8_y2,  avg=1
-        pixfunc2        avg_, pixels8_xy2, avg=1
+        pixfunc         avg_, pixels8_x2,  avg=1
+        pixfunc         avg_, pixels8_y2,  avg=1
+        pixfunc         avg_, pixels8_xy2, avg=1
 
 function ff_put_pixels_clamped_neon, export=1
         vld1.16         {d16-d19}, [r0,:128]!

libavcodec/dsputil.c

@@ -2925,7 +2925,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
     dspfunc1(avg       , 2,  4, depth);\
     dspfunc1(avg       , 3,  2, depth);\
     dspfunc1(avg_no_rnd, 0, 16, depth);\
-    dspfunc1(avg_no_rnd, 1,  8, depth);\
 \
     dspfunc2(put_h264_qpel, 0, 16, depth);\
     dspfunc2(put_h264_qpel, 1,  8, depth);\

libavcodec/dsputil.h

@@ -297,15 +297,15 @@ typedef struct DSPContext {
 
     /**
      * Halfpel motion compensation with no rounding (a+b)>>1.
-     * this is an array[2][4] of motion compensation functions for 2
-     * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
-     * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
+     * this is an array[4] of motion compensation functions for 1
+     * horizontal blocksize (16) and the 4 halfpel positions<br>
+     * *pixels_tab[0][ xhalfpel + 2*yhalfpel ]
      * @param block destination into which the result is averaged (a+b)>>1
      * @param pixels source
      * @param line_size number of bytes in a horizontal line of block
      * @param h height
      */
-    op_pixels_func avg_no_rnd_pixels_tab[4][4];
+    op_pixels_func avg_no_rnd_pixels_tab[1][4];
 
     /**
      * Thirdpel motion compensation with rounding (a+b+1)>>1.

libavcodec/sh4/dsputil_align.c

@@ -294,11 +294,8 @@ DEFFUNC(put,no_rnd,xy,16,OP_XY,PACK)
 
 DEFFUNC(avg,   rnd,o,8,OP_C,avg32)
 DEFFUNC(avg,   rnd,x,8,OP_X,avg32)
-DEFFUNC(avg,no_rnd,x,8,OP_X,avg32)
 DEFFUNC(avg,   rnd,y,8,OP_Y,avg32)
-DEFFUNC(avg,no_rnd,y,8,OP_Y,avg32)
 DEFFUNC(avg,   rnd,xy,8,OP_XY,PACK)
-DEFFUNC(avg,no_rnd,xy,8,OP_XY,PACK)
 DEFFUNC(avg,   rnd,o,16,OP_C,avg32)
 DEFFUNC(avg,   rnd,x,16,OP_X,avg32)
 DEFFUNC(avg,no_rnd,x,16,OP_X,avg32)
@@ -311,7 +308,6 @@ DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
 
 #define         put_no_rnd_pixels8_o     put_rnd_pixels8_o
 #define         put_no_rnd_pixels16_o    put_rnd_pixels16_o
-#define         avg_no_rnd_pixels8_o     avg_rnd_pixels8_o
 #define         avg_no_rnd_pixels16_o    avg_rnd_pixels16_o
 
 #define         put_pixels8_c            put_rnd_pixels8_o
@@ -320,7 +316,6 @@ DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
 #define         avg_pixels16_c           avg_rnd_pixels16_o
 #define         put_no_rnd_pixels8_c     put_rnd_pixels8_o
 #define         put_no_rnd_pixels16_c    put_rnd_pixels16_o
-#define         avg_no_rnd_pixels8_c     avg_rnd_pixels8_o
 #define         avg_no_rnd_pixels16_c    avg_rnd_pixels16_o
 
 #define         QPEL
@@ -367,10 +362,6 @@ void ff_dsputil_init_align(DSPContext* c, AVCodecContext *avctx)
         c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x;
         c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y;
         c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy;
-        c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_o;
-        c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x;
-        c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y;
-        c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy;
         }
 
 #ifdef QPEL

libavcodec/x86/dsputil_mmx.c

@@ -161,6 +161,7 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 };
 
 /***********************************/
 /* MMX no rounding */
+#define NO_RND 1
 #define DEF(x, y) x ## _no_rnd_ ## y ## _mmx
 #define SET_RND  MOVQ_WONE
 #define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX_NO_RND(a, b, c, d, e, f)
@@ -173,6 +174,7 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 };
 #undef SET_RND
 #undef PAVGBP
 #undef PAVGB
+#undef NO_RND
 /***********************************/
 /* MMX rounding */
 
@@ -2058,7 +2060,6 @@ static void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
         SET_HPEL_FUNCS(put,        1,  8, mmx);
         SET_HPEL_FUNCS(put_no_rnd, 1,  8, mmx);
         SET_HPEL_FUNCS(avg,        1,  8, mmx);
-        SET_HPEL_FUNCS(avg_no_rnd, 1,  8, mmx);
     }
 
 #if ARCH_X86_32 || !HAVE_YASM

libavcodec/x86/dsputil_rnd_template.c

@@ -317,6 +317,7 @@ static void av_unused DEF(avg, pixels4)(uint8_t *block, const uint8_t *pixels, i
     while (--h);
 }
 
+#ifndef NO_RND
 // in case more speed is needed - unroling would certainly help
 static void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
 {
@@ -336,6 +337,7 @@ static void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, int line_si
     }
     while (--h);
 }
+#endif // NO_RND
 
 static void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
 {
@@ -360,6 +362,7 @@ static void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, int line_s
     while (--h);
 }
 
+#ifndef NO_RND
 static void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
 {
     MOVQ_BFE(mm6);
@@ -379,6 +382,7 @@ static void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line
         block += line_size;
     } while (--h);
 }
+#endif // NO_RND
 
 static av_unused void DEF(avg, pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
 {