rv34: move inverse transform functions to DSP context

6 files changed, 121 insertions, 81 deletions

diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index b7b512478c..04a504be74 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -324,9 +324,9 @@ OBJS-$(CONFIG_RV10_DECODER)            += rv10.o
 OBJS-$(CONFIG_RV10_ENCODER)            += rv10enc.o
 OBJS-$(CONFIG_RV20_DECODER)            += rv10.o
 OBJS-$(CONFIG_RV20_ENCODER)            += rv20enc.o
-OBJS-$(CONFIG_RV30_DECODER)            += rv30.o rv34.o rv30dsp.o        \
+OBJS-$(CONFIG_RV30_DECODER)            += rv30.o rv34.o rv30dsp.o rv34dsp.o \
                                           mpegvideo.o error_resilience.o
-OBJS-$(CONFIG_RV40_DECODER)            += rv40.o rv34.o rv40dsp.o        \
+OBJS-$(CONFIG_RV40_DECODER)            += rv40.o rv34.o rv34dsp.o rv40dsp.o \
                                           mpegvideo.o error_resilience.o
 OBJS-$(CONFIG_S302M_DECODER)           += s302m.o
 OBJS-$(CONFIG_SGI_DECODER)             += sgidec.o
diff --git a/libavcodec/rv30dsp.c b/libavcodec/rv30dsp.c
index 6ba1a6bfc1..bcd1a462b1 100644
--- a/libavcodec/rv30dsp.c
+++ b/libavcodec/rv30dsp.c
@@ -253,6 +253,9 @@ RV30_MC(avg_, 8)
 RV30_MC(avg_, 16)
 
 av_cold void ff_rv30dsp_init(RV34DSPContext *c, DSPContext* dsp) {
+
+    ff_rv34dsp_init(c, dsp);
+
     c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];
     c->put_pixels_tab[0][ 1] = put_rv30_tpel16_mc10_c;
     c->put_pixels_tab[0][ 2] = put_rv30_tpel16_mc20_c;
diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c
index abdfce50a5..b771a7f97e 100644
--- a/libavcodec/rv34.c
+++ b/libavcodec/rv34.c
@@ -171,82 +171,6 @@ static av_cold void rv34_init_tables(void)
 
 /** @} */ // vlc group
 
-
-/**
- * @name RV30/40 inverse transform functions
- * @{
- */
-
-static av_always_inline void rv34_row_transform(int temp[16], DCTELEM *block)
-{
-    int i;
-
-    for(i = 0; i < 4; i++){
-        const int z0 = 13*(block[i+8*0] +    block[i+8*2]);
-        const int z1 = 13*(block[i+8*0] -    block[i+8*2]);
-        const int z2 =  7* block[i+8*1] - 17*block[i+8*3];
-        const int z3 = 17* block[i+8*1] +  7*block[i+8*3];
-
-        temp[4*i+0] = z0 + z3;
-        temp[4*i+1] = z1 + z2;
-        temp[4*i+2] = z1 - z2;
-        temp[4*i+3] = z0 - z3;
-    }
-}
-
-/**
- * Real Video 3.0/4.0 inverse transform
- * Code is almost the same as in SVQ3, only scaling is different.
- */
-static void rv34_inv_transform(DCTELEM *block){
-    int temp[16];
-    int i;
-
-    rv34_row_transform(temp, block);
-
-    for(i = 0; i < 4; i++){
-        const int z0 = 13*(temp[4*0+i] +    temp[4*2+i]) + 0x200;
-        const int z1 = 13*(temp[4*0+i] -    temp[4*2+i]) + 0x200;
-        const int z2 =  7* temp[4*1+i] - 17*temp[4*3+i];
-        const int z3 = 17* temp[4*1+i] +  7*temp[4*3+i];
-
-        block[i*8+0] = (z0 + z3) >> 10;
-        block[i*8+1] = (z1 + z2) >> 10;
-        block[i*8+2] = (z1 - z2) >> 10;
-        block[i*8+3] = (z0 - z3) >> 10;
-    }
-
-}
-
-/**
- * RealVideo 3.0/4.0 inverse transform for DC block
- *
- * Code is almost the same as rv34_inv_transform()
- * but final coefficients are multiplied by 1.5 and have no rounding.
- */
-static void rv34_inv_transform_noround(DCTELEM *block){
-    int temp[16];
-    int i;
-
-    rv34_row_transform(temp, block);
-
-    for(i = 0; i < 4; i++){
-        const int z0 = 13*(temp[4*0+i] +    temp[4*2+i]);
-        const int z1 = 13*(temp[4*0+i] -    temp[4*2+i]);
-        const int z2 =  7* temp[4*1+i] - 17*temp[4*3+i];
-        const int z3 = 17* temp[4*1+i] +  7*temp[4*3+i];
-
-        block[i*8+0] = ((z0 + z3) * 3) >> 11;
-        block[i*8+1] = ((z1 + z2) * 3) >> 11;
-        block[i*8+2] = ((z1 - z2) * 3) >> 11;
-        block[i*8+3] = ((z0 - z3) * 3) >> 11;
-    }
-
-}
-
-/** @} */ // transform
-
-
 /**
  * @name RV30/40 4x4 block decoding functions
  * @{
@@ -1226,7 +1150,7 @@ static int rv34_decode_macroblock(RV34DecContext *r, int8_t *intra_types)
         memset(block16, 0, sizeof(block16));
         rv34_decode_block(block16, gb, r->cur_vlcs, 3, 0);
         rv34_dequant4x4_16x16(block16, rv34_qscale_tab[luma_dc_quant],rv34_qscale_tab[s->qscale]);
-        rv34_inv_transform_noround(block16);
+        r->rdsp.rv34_inv_transform_tab[1](block16);
     }
 
     for(i = 0; i < 16; i++, cbp >>= 1){
@@ -1238,7 +1162,7 @@ static int rv34_decode_macroblock(RV34DecContext *r, int8_t *intra_types)
         rv34_dequant4x4(s->block[blknum] + blkoff, rv34_qscale_tab[s->qscale],rv34_qscale_tab[s->qscale]);
         if(r->is16) //FIXME: optimize
             s->block[blknum][blkoff] = block16[(i & 3) | ((i & 0xC) << 1)];
-        rv34_inv_transform(s->block[blknum] + blkoff);
+        r->rdsp.rv34_inv_transform_tab[0](s->block[blknum] + blkoff);
     }
     if(r->block_type == RV34_MB_P_MIX16x16)
         r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 1);
@@ -1248,7 +1172,7 @@ static int rv34_decode_macroblock(RV34DecContext *r, int8_t *intra_types)
         blkoff = ((i & 1) << 2) + ((i & 2) << 4);
         rv34_decode_block(s->block[blknum] + blkoff, gb, r->cur_vlcs, r->chroma_vlc, 1);
         rv34_dequant4x4(s->block[blknum] + blkoff, rv34_qscale_tab[rv34_chroma_quant[1][s->qscale]],rv34_qscale_tab[rv34_chroma_quant[0][s->qscale]]);
-        rv34_inv_transform(s->block[blknum] + blkoff);
+        r->rdsp.rv34_inv_transform_tab[0](s->block[blknum] + blkoff);
     }
     if (IS_INTRA(s->current_picture_ptr->f.mb_type[mb_pos]))
         rv34_output_macroblock(r, intra_types, cbp2, r->is16);
diff --git a/libavcodec/rv34dsp.c b/libavcodec/rv34dsp.c
new file mode 100644
index 0000000000..59038a7a31
--- /dev/null
+++ b/libavcodec/rv34dsp.c
@@ -0,0 +1,106 @@
+/*
+ * RV30/40 decoder common dsp functions
+ * Copyright (c) 2007 Mike Melanson, Konstantin Shishkov
+ * Copyright (c) 2011 Janne Grunau
+ *
+ * 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
+ */
+
+/**
+ * @file
+ * RV30/40 decoder common dsp functions
+ */
+#include "dsputil.h"
+#include "rv34dsp.h"
+
+/**
+ * @name RV30/40 inverse transform functions
+ * @{
+ */
+
+static av_always_inline void rv34_row_transform(int temp[16], DCTELEM *block)
+{
+    int i;
+
+    for(i = 0; i < 4; i++){
+        const int z0 = 13*(block[i+8*0] +    block[i+8*2]);
+        const int z1 = 13*(block[i+8*0] -    block[i+8*2]);
+        const int z2 =  7* block[i+8*1] - 17*block[i+8*3];
+        const int z3 = 17* block[i+8*1] +  7*block[i+8*3];
+
+        temp[4*i+0] = z0 + z3;
+        temp[4*i+1] = z1 + z2;
+        temp[4*i+2] = z1 - z2;
+        temp[4*i+3] = z0 - z3;
+    }
+}
+
+/**
+ * Real Video 3.0/4.0 inverse transform
+ * Code is almost the same as in SVQ3, only scaling is different.
+ */
+static void rv34_inv_transform_c(DCTELEM *block){
+    int temp[16];
+    int i;
+
+    rv34_row_transform(temp, block);
+
+    for(i = 0; i < 4; i++){
+        const int z0 = 13*(temp[4*0+i] +    temp[4*2+i]) + 0x200;
+        const int z1 = 13*(temp[4*0+i] -    temp[4*2+i]) + 0x200;
+        const int z2 =  7* temp[4*1+i] - 17*temp[4*3+i];
+        const int z3 = 17* temp[4*1+i] +  7*temp[4*3+i];
+
+        block[i*8+0] = (z0 + z3) >> 10;
+        block[i*8+1] = (z1 + z2) >> 10;
+        block[i*8+2] = (z1 - z2) >> 10;
+        block[i*8+3] = (z0 - z3) >> 10;
+    }
+}
+
+/**
+ * RealVideo 3.0/4.0 inverse transform for DC block
+ *
+ * Code is almost the same as rv34_inv_transform()
+ * but final coefficients are multiplied by 1.5 and have no rounding.
+ */
+static void rv34_inv_transform_noround_c(DCTELEM *block){
+    int temp[16];
+    int i;
+
+    rv34_row_transform(temp, block);
+
+    for(i = 0; i < 4; i++){
+        const int z0 = 13*(temp[4*0+i] +    temp[4*2+i]);
+        const int z1 = 13*(temp[4*0+i] -    temp[4*2+i]);
+        const int z2 =  7* temp[4*1+i] - 17*temp[4*3+i];
+        const int z3 = 17* temp[4*1+i] +  7*temp[4*3+i];
+
+        block[i*8+0] = ((z0 + z3) * 3) >> 11;
+        block[i*8+1] = ((z1 + z2) * 3) >> 11;
+        block[i*8+2] = ((z1 - z2) * 3) >> 11;
+        block[i*8+3] = ((z0 - z3) * 3) >> 11;
+    }
+}
+
+/** @} */ // transform
+
+
+av_cold void ff_rv34dsp_init(RV34DSPContext *c, DSPContext* dsp) {
+    c->rv34_inv_transform_tab[0] = rv34_inv_transform_c;
+    c->rv34_inv_transform_tab[1] = rv34_inv_transform_noround_c;
+}
diff --git a/libavcodec/rv34dsp.h b/libavcodec/rv34dsp.h
index e1def7dc26..a2ab5f232b 100644
--- a/libavcodec/rv34dsp.h
+++ b/libavcodec/rv34dsp.h
@@ -34,15 +34,19 @@ typedef void (*rv40_weight_func)(uint8_t *dst/*align width (8 or 16)*/,
                                  uint8_t *src2/*align width (8 or 16)*/,
                                  int w1, int w2, int stride);
 
+typedef void (*rv34_inv_transform_func)(DCTELEM *block);
+
 typedef struct RV34DSPContext {
     qpel_mc_func put_pixels_tab[4][16];
     qpel_mc_func avg_pixels_tab[4][16];
     h264_chroma_mc_func put_chroma_pixels_tab[3];
     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];
 } RV34DSPContext;
 
 void ff_rv30dsp_init(RV34DSPContext *c, DSPContext* dsp);
+void ff_rv34dsp_init(RV34DSPContext *c, DSPContext* dsp);
 void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp);
 
 void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp);
diff --git a/libavcodec/rv40dsp.c b/libavcodec/rv40dsp.c
index c54f96556f..efc049f1cb 100644
--- a/libavcodec/rv40dsp.c
+++ b/libavcodec/rv40dsp.c
@@ -295,6 +295,9 @@ RV40_WEIGHT_FUNC(16)
 RV40_WEIGHT_FUNC(8)
 
 av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
+
+    ff_rv34dsp_init(c, dsp);
+
     c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];
     c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;
     c->put_pixels_tab[0][ 2] = dsp->put_h264_qpel_pixels_tab[0][2];