sh4: hpeldsp: Move half-pel assembly from dsputil to hpeldsp
authorMichael Niedermayer <michaelni@gmx.at>
Sun, 10 Mar 2013 23:48:31 +0000 (16:48 -0700)
committerMartin Storsjö <martin@martin.st>
Fri, 19 Apr 2013 20:29:21 +0000 (23:29 +0300)
Signed-off-by: Martin Storsjö <martin@martin.st>
libavcodec/hpeldsp.c
libavcodec/hpeldsp.h
libavcodec/sh4/Makefile
libavcodec/sh4/dsputil_align.c
libavcodec/sh4/dsputil_sh4.h
libavcodec/sh4/hpeldsp.c [new file with mode: 0644]

index 29dd9e3..c5e7a04 100644 (file)
@@ -60,6 +60,8 @@ av_cold void ff_hpeldsp_init(HpelDSPContext *c, int flags)
         ff_hpeldsp_init_bfin(c, flags);
     if (ARCH_PPC)
         ff_hpeldsp_init_ppc(c, flags);
         ff_hpeldsp_init_bfin(c, flags);
     if (ARCH_PPC)
         ff_hpeldsp_init_ppc(c, flags);
+    if (ARCH_SH4)
+        ff_hpeldsp_init_sh4(c, flags);
     if (ARCH_X86)
         ff_hpeldsp_init_x86(c, flags);
 }
     if (ARCH_X86)
         ff_hpeldsp_init_x86(c, flags);
 }
index 9f5bc9f..6fdc4f3 100644 (file)
@@ -97,6 +97,7 @@ void ff_hpeldsp_init(HpelDSPContext *c, int flags);
 void ff_hpeldsp_init_arm(HpelDSPContext *c, int flags);
 void ff_hpeldsp_init_bfin(HpelDSPContext *c, int flags);
 void ff_hpeldsp_init_ppc(HpelDSPContext *c, int flags);
 void ff_hpeldsp_init_arm(HpelDSPContext *c, int flags);
 void ff_hpeldsp_init_bfin(HpelDSPContext *c, int flags);
 void ff_hpeldsp_init_ppc(HpelDSPContext *c, int flags);
+void ff_hpeldsp_init_sh4(HpelDSPContext *c, int flags);
 void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags);
 
 #endif /* AVCODEC_HPELDSP_H */
 void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags);
 
 #endif /* AVCODEC_HPELDSP_H */
index f907408..a24b298 100644 (file)
@@ -3,3 +3,5 @@ OBJS += sh4/dsputil_align.o                                             \
         sh4/idct_sh4.o                                                  \
 
 OBJS-$(CONFIG_H264CHROMA)               += sh4/h264chroma_init.o        \
         sh4/idct_sh4.o                                                  \
 
 OBJS-$(CONFIG_H264CHROMA)               += sh4/h264chroma_init.o        \
+
+OBJS-$(CONFIG_HPELDSP)                  += sh4/hpeldsp.o
index 5c3e65b..b5207b7 100644 (file)
@@ -233,67 +233,15 @@ if (sz==16) { \
         } while(--height); \
 }
 
         } while(--height); \
 }
 
-#define         DEFFUNC(op,rnd,xy,sz,OP_N,avgfunc) \
-static void op##_##rnd##_pixels##sz##_##xy (uint8_t * dest, const uint8_t * ref, \
-                                const ptrdiff_t stride, int height) \
-{ \
-        switch((int)ref&3) { \
-        case 0:OP_N##0(sz,rnd##_##avgfunc); return; \
-        case 1:OP_N(1,sz,rnd##_##avgfunc); return; \
-        case 2:OP_N(2,sz,rnd##_##avgfunc); return; \
-        case 3:OP_N(3,sz,rnd##_##avgfunc); return; \
-        } \
-}
-
-#define OP put
-
-DEFFUNC(put,   rnd,o,8,OP_C,avg32)
-DEFFUNC(put,   rnd,x,8,OP_X,avg32)
-DEFFUNC(put,no_rnd,x,8,OP_X,avg32)
-DEFFUNC(put,   rnd,y,8,OP_Y,avg32)
-DEFFUNC(put,no_rnd,y,8,OP_Y,avg32)
-DEFFUNC(put,   rnd,xy,8,OP_XY,PACK)
-DEFFUNC(put,no_rnd,xy,8,OP_XY,PACK)
-DEFFUNC(put,   rnd,o,16,OP_C,avg32)
-DEFFUNC(put,   rnd,x,16,OP_X,avg32)
-DEFFUNC(put,no_rnd,x,16,OP_X,avg32)
-DEFFUNC(put,   rnd,y,16,OP_Y,avg32)
-DEFFUNC(put,no_rnd,y,16,OP_Y,avg32)
-DEFFUNC(put,   rnd,xy,16,OP_XY,PACK)
-DEFFUNC(put,no_rnd,xy,16,OP_XY,PACK)
-
-#undef OP
-#define OP avg
-
-DEFFUNC(avg,   rnd,o,8,OP_C,avg32)
-DEFFUNC(avg,   rnd,x,8,OP_X,avg32)
-DEFFUNC(avg,   rnd,y,8,OP_Y,avg32)
-DEFFUNC(avg,   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)
-DEFFUNC(avg,   rnd,y,16,OP_Y,avg32)
-DEFFUNC(avg,no_rnd,y,16,OP_Y,avg32)
-DEFFUNC(avg,   rnd,xy,16,OP_XY,PACK)
-DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
-
-#undef OP
-
-#define         put_no_rnd_pixels8_o     put_rnd_pixels8_o
-#define         put_no_rnd_pixels16_o    put_rnd_pixels16_o
-#define         avg_no_rnd_pixels16_o    avg_rnd_pixels16_o
+#define         put_pixels8_c            ff_put_rnd_pixels8_o
+#define         put_pixels16_c           ff_put_rnd_pixels16_o
+#define         avg_pixels8_c            ff_avg_rnd_pixels8_o
+#define         avg_pixels16_c           ff_avg_rnd_pixels16_o
+#define         put_no_rnd_pixels8_c     ff_put_rnd_pixels8_o
+#define         put_no_rnd_pixels16_c    ff_put_rnd_pixels16_o
+#define         avg_no_rnd_pixels16_c    ff_avg_rnd_pixels16_o
 
 
-#define         put_pixels8_c            put_rnd_pixels8_o
-#define         put_pixels16_c           put_rnd_pixels16_o
-#define         avg_pixels8_c            avg_rnd_pixels8_o
-#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_pixels16_c    avg_rnd_pixels16_o
-
-#define         QPEL
-
-#ifdef QPEL
+#if CONFIG_HPELDSP
 
 #include "qpel.c"
 
 
 #include "qpel.c"
 
@@ -301,43 +249,7 @@ DEFFUNC(avg,no_rnd,xy,16,OP_XY,PACK)
 
 av_cold void ff_dsputil_init_align(DSPContext *c, AVCodecContext *avctx)
 {
 
 av_cold void ff_dsputil_init_align(DSPContext *c, AVCodecContext *avctx)
 {
-        const int high_bit_depth = avctx->bits_per_raw_sample > 8;
-
-        if (!high_bit_depth) {
-        c->put_pixels_tab[0][0] = put_rnd_pixels16_o;
-        c->put_pixels_tab[0][1] = put_rnd_pixels16_x;
-        c->put_pixels_tab[0][2] = put_rnd_pixels16_y;
-        c->put_pixels_tab[0][3] = put_rnd_pixels16_xy;
-        c->put_pixels_tab[1][0] = put_rnd_pixels8_o;
-        c->put_pixels_tab[1][1] = put_rnd_pixels8_x;
-        c->put_pixels_tab[1][2] = put_rnd_pixels8_y;
-        c->put_pixels_tab[1][3] = put_rnd_pixels8_xy;
-
-        c->put_no_rnd_pixels_tab[0][0] = put_no_rnd_pixels16_o;
-        c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x;
-        c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y;
-        c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy;
-        c->put_no_rnd_pixels_tab[1][0] = put_no_rnd_pixels8_o;
-        c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x;
-        c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y;
-        c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy;
-
-        c->avg_pixels_tab[0][0] = avg_rnd_pixels16_o;
-        c->avg_pixels_tab[0][1] = avg_rnd_pixels16_x;
-        c->avg_pixels_tab[0][2] = avg_rnd_pixels16_y;
-        c->avg_pixels_tab[0][3] = avg_rnd_pixels16_xy;
-        c->avg_pixels_tab[1][0] = avg_rnd_pixels8_o;
-        c->avg_pixels_tab[1][1] = avg_rnd_pixels8_x;
-        c->avg_pixels_tab[1][2] = avg_rnd_pixels8_y;
-        c->avg_pixels_tab[1][3] = avg_rnd_pixels8_xy;
-
-        c->avg_no_rnd_pixels_tab[0] = avg_no_rnd_pixels16_o;
-        c->avg_no_rnd_pixels_tab[1] = avg_no_rnd_pixels16_x;
-        c->avg_no_rnd_pixels_tab[2] = avg_no_rnd_pixels16_y;
-        c->avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels16_xy;
-        }
-
-#ifdef QPEL
+#if CONFIG_HPELDSP
 
 #define dspfunc(PFX, IDX, NUM) \
     c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_sh4; \
 
 #define dspfunc(PFX, IDX, NUM) \
     c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_sh4; \
index f7dc30b..f474f90 100644 (file)
 void ff_idct_sh4(int16_t *block);
 void ff_dsputil_init_align(DSPContext* c, AVCodecContext *avctx);
 
 void ff_idct_sh4(int16_t *block);
 void ff_dsputil_init_align(DSPContext* c, AVCodecContext *avctx);
 
+void ff_put_rnd_pixels8_o(uint8_t *dest, const uint8_t *ref,
+                          const ptrdiff_t stride, int height);
+void ff_put_rnd_pixels16_o(uint8_t *dest, const uint8_t *ref,
+                           const ptrdiff_t stride, int height);
+void ff_avg_rnd_pixels8_o (uint8_t *dest, const uint8_t *ref,
+                           const ptrdiff_t stride, int height);
+void ff_avg_rnd_pixels16_o(uint8_t *dest, const uint8_t *ref,
+                           const ptrdiff_t stride, int height);
+
 #endif /* AVCODEC_SH4_DSPUTIL_SH4_H */
 #endif /* AVCODEC_SH4_DSPUTIL_SH4_H */
diff --git a/libavcodec/sh4/hpeldsp.c b/libavcodec/sh4/hpeldsp.c
new file mode 100644 (file)
index 0000000..530d566
--- /dev/null
@@ -0,0 +1,350 @@
+/*
+ * aligned/packed access motion
+ *
+ * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/attributes.h"
+#include "libavcodec/avcodec.h"
+#include "libavcodec/dsputil.h"
+#include "libavcodec/hpeldsp.h"
+#include "libavcodec/rnd_avg.h"
+#include "dsputil_sh4.h"
+
+
+#define         LP(p)           *(uint32_t*)(p)
+#define         LPC(p)          *(const uint32_t*)(p)
+
+
+#define         UNPACK(ph,pl,tt0,tt1) do { \
+        uint32_t t0,t1; t0=tt0;t1=tt1; \
+        ph = ( (t0 & ~BYTE_VEC32(0x03))>>2) + ( (t1 & ~BYTE_VEC32(0x03))>>2); \
+        pl = (t0 & BYTE_VEC32(0x03)) + (t1 & BYTE_VEC32(0x03)); } while(0)
+
+#define         rnd_PACK(ph,pl,nph,npl) ph + nph + (((pl + npl + BYTE_VEC32(0x02))>>2) & BYTE_VEC32(0x03))
+#define         no_rnd_PACK(ph,pl,nph,npl)      ph + nph + (((pl + npl + BYTE_VEC32(0x01))>>2) & BYTE_VEC32(0x03))
+
+/* little-endian */
+#define         MERGE1(a,b,ofs) (ofs==0)?a:( ((a)>>(8*ofs))|((b)<<(32-8*ofs)) )
+#define         MERGE2(a,b,ofs) (ofs==3)?b:( ((a)>>(8*(ofs+1)))|((b)<<(32-8*(ofs+1))) )
+/* big
+#define         MERGE1(a,b,ofs) (ofs==0)?a:( ((a)<<(8*ofs))|((b)>>(32-8*ofs)) )
+#define         MERGE2(a,b,ofs) (ofs==3)?b:( ((a)<<(8+8*ofs))|((b)>>(32-8-8*ofs)) )
+*/
+
+
+#define         put(d,s)        d = s
+#define         avg(d,s)        d = rnd_avg32(s,d)
+
+#define         OP_C4(ofs) \
+        ref-=ofs; \
+        do { \
+                OP(LP(dest),MERGE1(LPC(ref),LPC(ref+4),ofs)); \
+                ref+=stride; \
+                dest+=stride; \
+        } while(--height)
+
+#define        OP_C40() \
+        do { \
+                OP(LP(dest),LPC(ref)); \
+                ref+=stride; \
+                dest+=stride; \
+        } while(--height)
+
+
+#define         OP      put
+
+static void put_pixels4_c(uint8_t *dest, const uint8_t *ref,
+                          const int stride, int height)
+{
+        switch((int)ref&3){
+        case 0: OP_C40(); return;
+        case 1: OP_C4(1); return;
+        case 2: OP_C4(2); return;
+        case 3: OP_C4(3); return;
+        }
+}
+
+#undef          OP
+#define         OP      avg
+
+static void avg_pixels4_c(uint8_t *dest, const uint8_t *ref,
+                          const int stride, int height)
+{
+        switch((int)ref&3){
+        case 0: OP_C40(); return;
+        case 1: OP_C4(1); return;
+        case 2: OP_C4(2); return;
+        case 3: OP_C4(3); return;
+        }
+}
+
+#undef          OP
+
+#define         OP_C(ofs,sz,avg2) \
+{ \
+        ref-=ofs; \
+        do { \
+                uint32_t        t0,t1; \
+                t0 = LPC(ref+0); \
+                t1 = LPC(ref+4); \
+                OP(LP(dest+0), MERGE1(t0,t1,ofs)); \
+                t0 = LPC(ref+8); \
+                OP(LP(dest+4), MERGE1(t1,t0,ofs)); \
+if (sz==16) { \
+                t1 = LPC(ref+12); \
+                OP(LP(dest+8), MERGE1(t0,t1,ofs)); \
+                t0 = LPC(ref+16); \
+                OP(LP(dest+12), MERGE1(t1,t0,ofs)); \
+} \
+                ref+=stride; \
+                dest+= stride; \
+        } while(--height); \
+}
+
+/* aligned */
+#define         OP_C0(sz,avg2) \
+{ \
+        do { \
+                OP(LP(dest+0), LPC(ref+0)); \
+                OP(LP(dest+4), LPC(ref+4)); \
+if (sz==16) { \
+                OP(LP(dest+8), LPC(ref+8)); \
+                OP(LP(dest+12), LPC(ref+12)); \
+} \
+                ref+=stride; \
+                dest+= stride; \
+        } while(--height); \
+}
+
+#define         OP_X(ofs,sz,avg2) \
+{ \
+        ref-=ofs; \
+        do { \
+                uint32_t        t0,t1; \
+                t0 = LPC(ref+0); \
+                t1 = LPC(ref+4); \
+                OP(LP(dest+0), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
+                t0 = LPC(ref+8); \
+                OP(LP(dest+4), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
+if (sz==16) { \
+                t1 = LPC(ref+12); \
+                OP(LP(dest+8), avg2(MERGE1(t0,t1,ofs),MERGE2(t0,t1,ofs))); \
+                t0 = LPC(ref+16); \
+                OP(LP(dest+12), avg2(MERGE1(t1,t0,ofs),MERGE2(t1,t0,ofs))); \
+} \
+                ref+=stride; \
+                dest+= stride; \
+        } while(--height); \
+}
+
+/* aligned */
+#define         OP_Y0(sz,avg2) \
+{ \
+        uint32_t t0,t1,t2,t3,t; \
+\
+        t0 = LPC(ref+0); \
+        t1 = LPC(ref+4); \
+if (sz==16) { \
+        t2 = LPC(ref+8); \
+        t3 = LPC(ref+12); \
+} \
+        do { \
+                ref += stride; \
+\
+                t = LPC(ref+0); \
+                OP(LP(dest+0), avg2(t0,t)); t0 = t; \
+                t = LPC(ref+4); \
+                OP(LP(dest+4), avg2(t1,t)); t1 = t; \
+if (sz==16) { \
+                t = LPC(ref+8); \
+                OP(LP(dest+8), avg2(t2,t)); t2 = t; \
+                t = LPC(ref+12); \
+                OP(LP(dest+12), avg2(t3,t)); t3 = t; \
+} \
+                dest+= stride; \
+        } while(--height); \
+}
+
+#define         OP_Y(ofs,sz,avg2) \
+{ \
+        uint32_t t0,t1,t2,t3,t,w0,w1; \
+\
+        ref-=ofs; \
+        w0 = LPC(ref+0); \
+        w1 = LPC(ref+4); \
+        t0 = MERGE1(w0,w1,ofs); \
+        w0 = LPC(ref+8); \
+        t1 = MERGE1(w1,w0,ofs); \
+if (sz==16) { \
+        w1 = LPC(ref+12); \
+        t2 = MERGE1(w0,w1,ofs); \
+        w0 = LPC(ref+16); \
+        t3 = MERGE1(w1,w0,ofs); \
+} \
+        do { \
+                ref += stride; \
+\
+                w0 = LPC(ref+0); \
+                w1 = LPC(ref+4); \
+                t = MERGE1(w0,w1,ofs); \
+                OP(LP(dest+0), avg2(t0,t)); t0 = t; \
+                w0 = LPC(ref+8); \
+                t = MERGE1(w1,w0,ofs); \
+                OP(LP(dest+4), avg2(t1,t)); t1 = t; \
+if (sz==16) { \
+                w1 = LPC(ref+12); \
+                t = MERGE1(w0,w1,ofs); \
+                OP(LP(dest+8), avg2(t2,t)); t2 = t; \
+                w0 = LPC(ref+16); \
+                t = MERGE1(w1,w0,ofs); \
+                OP(LP(dest+12), avg2(t3,t)); t3 = t; \
+} \
+                dest+=stride; \
+        } while(--height); \
+}
+
+#define OP_X0(sz,avg2) OP_X(0,sz,avg2)
+#define OP_XY0(sz,PACK) OP_XY(0,sz,PACK)
+#define         OP_XY(ofs,sz,PACK) \
+{ \
+        uint32_t        t2,t3,w0,w1; \
+        uint32_t        a0,a1,a2,a3,a4,a5,a6,a7; \
+\
+        ref -= ofs; \
+        w0 = LPC(ref+0); \
+        w1 = LPC(ref+4); \
+        UNPACK(a0,a1,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
+        w0 = LPC(ref+8); \
+        UNPACK(a2,a3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
+if (sz==16) { \
+        w1 = LPC(ref+12); \
+        UNPACK(a4,a5,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
+        w0 = LPC(ref+16); \
+        UNPACK(a6,a7,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
+} \
+        do { \
+                ref+=stride; \
+                w0 = LPC(ref+0); \
+                w1 = LPC(ref+4); \
+                UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
+                OP(LP(dest+0),PACK(a0,a1,t2,t3)); \
+                a0 = t2; a1 = t3; \
+                w0 = LPC(ref+8); \
+                UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
+                OP(LP(dest+4),PACK(a2,a3,t2,t3)); \
+                a2 = t2; a3 = t3; \
+if (sz==16) { \
+                w1 = LPC(ref+12); \
+                UNPACK(t2,t3,MERGE1(w0,w1,ofs),MERGE2(w0,w1,ofs)); \
+                OP(LP(dest+8),PACK(a4,a5,t2,t3)); \
+                a4 = t2; a5 = t3; \
+                w0 = LPC(ref+16); \
+                UNPACK(t2,t3,MERGE1(w1,w0,ofs),MERGE2(w1,w0,ofs)); \
+                OP(LP(dest+12),PACK(a6,a7,t2,t3)); \
+                a6 = t2; a7 = t3; \
+} \
+                dest+=stride; \
+        } while(--height); \
+}
+
+#define         DEFFUNC(prefix, op, rnd, xy, sz, OP_N, avgfunc) \
+prefix void op##_##rnd##_pixels##sz##_##xy(uint8_t *dest, const uint8_t *ref, \
+                                           const ptrdiff_t stride, int height) \
+{ \
+        switch((int)ref&3) { \
+        case 0:OP_N##0(sz,rnd##_##avgfunc); return; \
+        case 1:OP_N(1,sz,rnd##_##avgfunc); return; \
+        case 2:OP_N(2,sz,rnd##_##avgfunc); return; \
+        case 3:OP_N(3,sz,rnd##_##avgfunc); return; \
+        } \
+}
+
+#define OP put
+
+DEFFUNC(      ,ff_put,rnd,o,8,OP_C,avg32)
+DEFFUNC(static,put,   rnd,x,8,OP_X,avg32)
+DEFFUNC(static,put,no_rnd,x,8,OP_X,avg32)
+DEFFUNC(static,put,   rnd,y,8,OP_Y,avg32)
+DEFFUNC(static,put,no_rnd,y,8,OP_Y,avg32)
+DEFFUNC(static,put,   rnd,xy,8,OP_XY,PACK)
+DEFFUNC(static,put,no_rnd,xy,8,OP_XY,PACK)
+DEFFUNC(      ,ff_put,rnd,o,16,OP_C,avg32)
+DEFFUNC(static,put,   rnd,x,16,OP_X,avg32)
+DEFFUNC(static,put,no_rnd,x,16,OP_X,avg32)
+DEFFUNC(static,put,   rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,put,no_rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,put,   rnd,xy,16,OP_XY,PACK)
+DEFFUNC(static,put,no_rnd,xy,16,OP_XY,PACK)
+
+#undef OP
+#define OP avg
+
+DEFFUNC(      ,ff_avg,rnd,o,8,OP_C,avg32)
+DEFFUNC(static,avg,   rnd,x,8,OP_X,avg32)
+DEFFUNC(static,avg,   rnd,y,8,OP_Y,avg32)
+DEFFUNC(static,avg,   rnd,xy,8,OP_XY,PACK)
+DEFFUNC(      ,ff_avg,rnd,o,16,OP_C,avg32)
+DEFFUNC(static,avg,   rnd,x,16,OP_X,avg32)
+DEFFUNC(static,avg,no_rnd,x,16,OP_X,avg32)
+DEFFUNC(static,avg,   rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,avg,no_rnd,y,16,OP_Y,avg32)
+DEFFUNC(static,avg,   rnd,xy,16,OP_XY,PACK)
+DEFFUNC(static,avg,no_rnd,xy,16,OP_XY,PACK)
+
+#undef OP
+
+#define         ff_put_no_rnd_pixels8_o     ff_put_rnd_pixels8_o
+#define         ff_put_no_rnd_pixels16_o    ff_put_rnd_pixels16_o
+#define         ff_avg_no_rnd_pixels16_o    ff_avg_rnd_pixels16_o
+
+av_cold void ff_hpeldsp_init_sh4(HpelDSPContext *c, int flags)
+{
+    c->put_pixels_tab[0][0] = ff_put_rnd_pixels16_o;
+    c->put_pixels_tab[0][1] = put_rnd_pixels16_x;
+    c->put_pixels_tab[0][2] = put_rnd_pixels16_y;
+    c->put_pixels_tab[0][3] = put_rnd_pixels16_xy;
+    c->put_pixels_tab[1][0] = ff_put_rnd_pixels8_o;
+    c->put_pixels_tab[1][1] = put_rnd_pixels8_x;
+    c->put_pixels_tab[1][2] = put_rnd_pixels8_y;
+    c->put_pixels_tab[1][3] = put_rnd_pixels8_xy;
+
+    c->put_no_rnd_pixels_tab[0][0] = ff_put_no_rnd_pixels16_o;
+    c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x;
+    c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y;
+    c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy;
+    c->put_no_rnd_pixels_tab[1][0] = ff_put_no_rnd_pixels8_o;
+    c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x;
+    c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y;
+    c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy;
+
+    c->avg_pixels_tab[0][0] = ff_avg_rnd_pixels16_o;
+    c->avg_pixels_tab[0][1] = avg_rnd_pixels16_x;
+    c->avg_pixels_tab[0][2] = avg_rnd_pixels16_y;
+    c->avg_pixels_tab[0][3] = avg_rnd_pixels16_xy;
+    c->avg_pixels_tab[1][0] = ff_avg_rnd_pixels8_o;
+    c->avg_pixels_tab[1][1] = avg_rnd_pixels8_x;
+    c->avg_pixels_tab[1][2] = avg_rnd_pixels8_y;
+    c->avg_pixels_tab[1][3] = avg_rnd_pixels8_xy;
+
+    c->avg_no_rnd_pixels_tab[0] = ff_avg_no_rnd_pixels16_o;
+    c->avg_no_rnd_pixels_tab[1] = avg_no_rnd_pixels16_x;
+    c->avg_no_rnd_pixels_tab[2] = avg_no_rnd_pixels16_y;
+    c->avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels16_xy;
+}