Separate DWT from snow and dsputil

This moves the DWT functions from snow.c and dsputil.c to a file of
their own.  A new struct, DWTContext, holds the function pointers
previously part of DSPContext.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@22522 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

configure

@@ -894,6 +894,7 @@ CONFIG_LIST="
     bzlib
     dct
     doc
+    dwt
     dxva2
     fastdiv
     ffmpeg
@@ -1276,7 +1277,8 @@ rv30_decoder_select="golomb"
 rv40_decoder_select="golomb"
 shorten_decoder_select="golomb"
 sipr_decoder_select="lsp"
-snow_encoder_select="aandct"
+snow_decoder_select="dwt"
+snow_encoder_select="aandct dwt"
 sonic_decoder_select="golomb"
 sonic_encoder_select="golomb"
 sonic_ls_encoder_select="golomb"

libavcodec/Makefile

@@ -28,6 +28,7 @@ OBJS = allcodecs.o                                                      \
 OBJS-$(CONFIG_AANDCT)                  += aandcttab.o
 OBJS-$(CONFIG_ENCODERS)                += faandct.o jfdctfst.o jfdctint.o
 OBJS-$(CONFIG_DCT)                     += dct.o
+OBJS-$(CONFIG_DWT)                     += dwt.o
 OBJS-$(CONFIG_DXVA2)                   += dxva2.o
 FFT-OBJS-$(CONFIG_HARDCODED_TABLES)    += cos_tables.o
 OBJS-$(CONFIG_FFT)                     += avfft.o fft.o $(FFT-OBJS-yes)
@@ -598,7 +599,7 @@ MMX-OBJS-$(CONFIG_CAVS_DECODER)        += x86/cavsdsp_mmx.o
 MMX-OBJS-$(CONFIG_ENCODERS)            += x86/dsputilenc_mmx.o
 MMX-OBJS-$(CONFIG_GPL)                 += x86/idct_mmx.o
 MMX-OBJS-$(CONFIG_LPC)                 += x86/lpc_mmx.o
-MMX-OBJS-$(CONFIG_SNOW_DECODER)        += x86/snowdsp_mmx.o
+MMX-OBJS-$(CONFIG_DWT)                 += x86/snowdsp_mmx.o
 MMX-OBJS-$(CONFIG_VC1_DECODER)         += x86/vc1dsp_mmx.o
 MMX-OBJS-$(CONFIG_VP3_DECODER)         += x86/vp3dsp_mmx.o              \
                                           x86/vp3dsp_sse2.o

libavcodec/dsputil.c

@@ -33,7 +33,6 @@
 #include "faandct.h"
 #include "faanidct.h"
 #include "mathops.h"
-#include "snow.h"
 #include "mpegvideo.h"
 #include "config.h"
 #include "lpc.h"
@@ -329,102 +328,6 @@ static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
     return s;
 }
 
-
-#if CONFIG_SNOW_ENCODER //dwt is in snow.c
-static inline int w_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int w, int h, int type){
-    int s, i, j;
-    const int dec_count= w==8 ? 3 : 4;
-    int tmp[32*32];
-    int level, ori;
-    static const int scale[2][2][4][4]={
-      {
-        {
-            // 9/7 8x8 dec=3
-            {268, 239, 239, 213},
-            {  0, 224, 224, 152},
-            {  0, 135, 135, 110},
-        },{
-            // 9/7 16x16 or 32x32 dec=4
-            {344, 310, 310, 280},
-            {  0, 320, 320, 228},
-            {  0, 175, 175, 136},
-            {  0, 129, 129, 102},
-        }
-      },{
-        {
-            // 5/3 8x8 dec=3
-            {275, 245, 245, 218},
-            {  0, 230, 230, 156},
-            {  0, 138, 138, 113},
-        },{
-            // 5/3 16x16 or 32x32 dec=4
-            {352, 317, 317, 286},
-            {  0, 328, 328, 233},
-            {  0, 180, 180, 140},
-            {  0, 132, 132, 105},
-        }
-      }
-    };
-
-    for (i = 0; i < h; i++) {
-        for (j = 0; j < w; j+=4) {
-            tmp[32*i+j+0] = (pix1[j+0] - pix2[j+0])<<4;
-            tmp[32*i+j+1] = (pix1[j+1] - pix2[j+1])<<4;
-            tmp[32*i+j+2] = (pix1[j+2] - pix2[j+2])<<4;
-            tmp[32*i+j+3] = (pix1[j+3] - pix2[j+3])<<4;
-        }
-        pix1 += line_size;
-        pix2 += line_size;
-    }
-
-    ff_spatial_dwt(tmp, w, h, 32, type, dec_count);
-
-    s=0;
-    assert(w==h);
-    for(level=0; level<dec_count; level++){
-        for(ori= level ? 1 : 0; ori<4; ori++){
-            int size= w>>(dec_count-level);
-            int sx= (ori&1) ? size : 0;
-            int stride= 32<<(dec_count-level);
-            int sy= (ori&2) ? stride>>1 : 0;
-
-            for(i=0; i<size; i++){
-                for(j=0; j<size; j++){
-                    int v= tmp[sx + sy + i*stride + j] * scale[type][dec_count-3][level][ori];
-                    s += FFABS(v);
-                }
-            }
-        }
-    }
-    assert(s>=0);
-    return s>>9;
-}
-
-static int w53_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
-    return w_c(v, pix1, pix2, line_size,  8, h, 1);
-}
-
-static int w97_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
-    return w_c(v, pix1, pix2, line_size,  8, h, 0);
-}
-
-static int w53_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
-    return w_c(v, pix1, pix2, line_size, 16, h, 1);
-}
-
-static int w97_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
-    return w_c(v, pix1, pix2, line_size, 16, h, 0);
-}
-
-int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
-    return w_c(v, pix1, pix2, line_size, 32, h, 1);
-}
-
-int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
-    return w_c(v, pix1, pix2, line_size, 32, h, 0);
-}
-#endif
-
 /* draw the edges of width 'w' of an image of size width, height */
 //FIXME check that this is ok for mpeg4 interlaced
 static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
@@ -3531,7 +3434,7 @@ void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type){
         case FF_CMP_NSSE:
             cmp[i]= c->nsse[i];
             break;
-#if CONFIG_SNOW_ENCODER
+#if CONFIG_DWT
         case FF_CMP_W53:
             cmp[i]= c->w53[i];
             break;
@@ -4816,11 +4719,8 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->vsse[5]= vsse_intra8_c;
     c->nsse[0]= nsse16_c;
     c->nsse[1]= nsse8_c;
-#if CONFIG_SNOW_ENCODER
-    c->w53[0]= w53_16_c;
-    c->w53[1]= w53_8_c;
-    c->w97[0]= w97_16_c;
-    c->w97[1]= w97_8_c;
+#if CONFIG_DWT
+    ff_dsputil_init_dwt(c);
 #endif
 
     c->ssd_int8_vs_int16 = ssd_int8_vs_int16_c;
@@ -4865,12 +4765,6 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->try_8x8basis= try_8x8basis_c;
     c->add_8x8basis= add_8x8basis_c;
 
-#if CONFIG_SNOW_DECODER
-    c->vertical_compose97i = ff_snow_vertical_compose97i;
-    c->horizontal_compose97i = ff_snow_horizontal_compose97i;
-    c->inner_add_yblock = ff_snow_inner_add_yblock;
-#endif
-
 #if CONFIG_VORBIS_DECODER
     c->vorbis_inverse_coupling = vorbis_inverse_coupling;
 #endif

libavcodec/dsputil.h

@@ -37,8 +37,6 @@
 //#define DEBUG
 /* dct code */
 typedef short DCTELEM;
-typedef int DWTELEM;
-typedef short IDWTELEM;
 
 void fdct_ifast (DCTELEM *data);
 void fdct_ifast248 (DCTELEM *data);
@@ -185,10 +183,6 @@ static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
 typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size, int h)/* __attribute__ ((const))*/;
 
-
-// for snow slices
-typedef struct slice_buffer_s slice_buffer;
-
 /**
  * Scantable.
  */
@@ -538,11 +532,6 @@ typedef struct DSPContext {
     void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
     void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
 
-    /* snow wavelet */
-    void (*vertical_compose97i)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
-    void (*horizontal_compose97i)(IDWTELEM *b, int width);
-    void (*inner_add_yblock)(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
-
     void (*prefetch)(void *mem, int stride, int h);
 
     void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
@@ -681,6 +670,7 @@ void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
 
+void ff_dsputil_init_dwt(DSPContext *c);
 void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx);
 void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx);
 void ff_rv40dsp_init(DSPContext* c, AVCodecContext *avctx);

libavcodec/dwt.h

+/*
+ * Copyright (C) 2004-2010 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_DWT_H
+#define AVCODEC_DWT_H
+
+#include <stdint.h>
+
+typedef int DWTELEM;
+typedef short IDWTELEM;
+
+typedef struct {
+    IDWTELEM *b0;
+    IDWTELEM *b1;
+    IDWTELEM *b2;
+    IDWTELEM *b3;
+    int y;
+} DWTCompose;
+
+/** Used to minimize the amount of memory used in order to optimize cache performance. **/
+typedef struct slice_buffer_s {
+    IDWTELEM * * line; ///< For use by idwt and predict_slices.
+    IDWTELEM * * data_stack; ///< Used for internal purposes.
+    int data_stack_top;
+    int line_count;
+    int line_width;
+    int data_count;
+    IDWTELEM * base_buffer; ///< Buffer that this structure is caching.
+} slice_buffer;
+
+typedef struct DWTContext {
+    void (*vertical_compose97i)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
+    void (*horizontal_compose97i)(IDWTELEM *b, int width);
+    void (*inner_add_yblock)(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
+} DWTContext;
+
+#define MAX_DECOMPOSITIONS 8
+
+#define DWT_97 0
+#define DWT_53 1
+
+#define liftS lift
+#if 1
+#define W_AM 3
+#define W_AO 0
+#define W_AS 1
+
+#undef liftS
+#define W_BM 1
+#define W_BO 8
+#define W_BS 4
+
+#define W_CM 1
+#define W_CO 0
+#define W_CS 0
+
+#define W_DM 3
+#define W_DO 4
+#define W_DS 3
+#elif 0
+#define W_AM 55
+#define W_AO 16
+#define W_AS 5
+
+#define W_BM 3
+#define W_BO 32
+#define W_BS 6
+
+#define W_CM 127
+#define W_CO 64
+#define W_CS 7
+
+#define W_DM 7
+#define W_DO 8
+#define W_DS 4
+#elif 0
+#define W_AM 97
+#define W_AO 32
+#define W_AS 6
+
+#define W_BM 63
+#define W_BO 512
+#define W_BS 10
+
+#define W_CM 13
+#define W_CO 8
+#define W_CS 4
+
+#define W_DM 15
+#define W_DO 16
+#define W_DS 5
+
+#else
+
+#define W_AM 203
+#define W_AO 64
+#define W_AS 7
+
+#define W_BM 217
+#define W_BO 2048
+#define W_BS 12
+
+#define W_CM 113
+#define W_CO 64
+#define W_CS 7
+
+#define W_DM 227
+#define W_DO 128
+#define W_DS 9
+#endif
+
+#define slice_buffer_get_line(slice_buf, line_num) ((slice_buf)->line[line_num] ? (slice_buf)->line[line_num] : slice_buffer_load_line((slice_buf), (line_num)))
+//#define slice_buffer_get_line(slice_buf, line_num) (slice_buffer_load_line((slice_buf), (line_num)))
+
+void slice_buffer_init(slice_buffer * buf, int line_count, int max_allocated_lines, int line_width, IDWTELEM * base_buffer);
+void slice_buffer_release(slice_buffer * buf, int line);
+void slice_buffer_flush(slice_buffer * buf);
+void slice_buffer_destroy(slice_buffer * buf);
+IDWTELEM * slice_buffer_load_line(slice_buffer * buf, int line);
+
+void ff_snow_vertical_compose97i(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
+void ff_snow_horizontal_compose97i(IDWTELEM *b, int width);
+void ff_snow_inner_add_yblock(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
+
+int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h);
+int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h);
+
+void ff_spatial_dwt(int *buffer, int width, int height, int stride, int type, int decomposition_count);
+
+void ff_spatial_idwt_buffered_init(DWTCompose *cs, slice_buffer * sb, int width, int height, int stride_line, int type, int decomposition_count);
+void ff_spatial_idwt_buffered_slice(DWTContext *dsp, DWTCompose *cs, slice_buffer * slice_buf, int width, int height, int stride_line, int type, int decomposition_count, int y);
+void ff_spatial_idwt_init(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count);
+void ff_spatial_idwt_slice(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count, int y);
+void ff_spatial_idwt(IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count);
+
+void ff_dwt_init(DWTContext *c);
+void ff_dwt_init_x86(DWTContext *c);
+
+#endif /* AVCODEC_DWT_H */

libavcodec/ivi_dsp.c

@@ -28,6 +28,7 @@
 
 #include "avcodec.h"
 #include "dsputil.h"
+#include "dwt.h"
 #include "ivi_common.h"
 #include "ivi_dsp.h"
 

libavcodec/snow.h

@@ -23,10 +23,10 @@
 #define AVCODEC_SNOW_H
 
 #include "dsputil.h"
+#include "dwt.h"
 
 #define MID_STATE 128
 
-#define MAX_DECOMPOSITIONS 8
 #define MAX_PLANES 4
 #define QSHIFT 5
 #define QROOT (1<<QSHIFT)
@@ -37,101 +37,6 @@
 #define LOG2_OBMC_MAX 8
 #define OBMC_MAX (1<<(LOG2_OBMC_MAX))
 
-#define DWT_97 0
-#define DWT_53 1
-
-/** Used to minimize the amount of memory used in order to optimize cache performance. **/
-struct slice_buffer_s {
-    IDWTELEM * * line; ///< For use by idwt and predict_slices.
-    IDWTELEM * * data_stack; ///< Used for internal purposes.
-    int data_stack_top;
-    int line_count;
-    int line_width;
-    int data_count;
-    IDWTELEM * base_buffer; ///< Buffer that this structure is caching.
-};
-
-#define liftS lift
-#if 1
-#define W_AM 3
-#define W_AO 0
-#define W_AS 1
-
-#undef liftS
-#define W_BM 1
-#define W_BO 8
-#define W_BS 4
-
-#define W_CM 1
-#define W_CO 0
-#define W_CS 0
-
-#define W_DM 3
-#define W_DO 4
-#define W_DS 3
-#elif 0
-#define W_AM 55
-#define W_AO 16
-#define W_AS 5
-
-#define W_BM 3
-#define W_BO 32
-#define W_BS 6
-
-#define W_CM 127
-#define W_CO 64
-#define W_CS 7
-
-#define W_DM 7
-#define W_DO 8
-#define W_DS 4
-#elif 0
-#define W_AM 97
-#define W_AO 32
-#define W_AS 6
-
-#define W_BM 63
-#define W_BO 512
-#define W_BS 10
-
-#define W_CM 13
-#define W_CO 8
-#define W_CS 4
-
-#define W_DM 15
-#define W_DO 16
-#define W_DS 5
-
-#else
-
-#define W_AM 203
-#define W_AO 64
-#define W_AS 7
-
-#define W_BM 217
-#define W_BO 2048
-#define W_BS 12
-
-#define W_CM 113
-#define W_CO 64
-#define W_CS 7
-
-#define W_DM 227
-#define W_DO 128
-#define W_DS 9
-#endif
-
-void ff_snow_vertical_compose97i(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
-void ff_snow_horizontal_compose97i(IDWTELEM *b, int width);
-void ff_snow_inner_add_yblock(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
-
-#if CONFIG_SNOW_ENCODER
-int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h);
-int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h);
-#endif
-
-void ff_spatial_dwt(int *buffer, int width, int height, int stride, int type, int decomposition_count);
-
 /* C bits used by mmx/sse2/altivec */
 
 static av_always_inline void snow_interleave_line_header(int * i, int width, IDWTELEM * low, IDWTELEM * high){

libavcodec/x86/dsputil_mmx.c

@@ -2894,25 +2894,6 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         }
 #endif
 
-#if CONFIG_SNOW_DECODER
-        if(mm_flags & FF_MM_SSE2 & 0){
-            c->horizontal_compose97i = ff_snow_horizontal_compose97i_sse2;
-#if HAVE_7REGS
-            c->vertical_compose97i = ff_snow_vertical_compose97i_sse2;
-#endif
-            c->inner_add_yblock = ff_snow_inner_add_yblock_sse2;
-        }
-        else{
-            if(mm_flags & FF_MM_MMX2){
-            c->horizontal_compose97i = ff_snow_horizontal_compose97i_mmx;
-#if HAVE_7REGS
-            c->vertical_compose97i = ff_snow_vertical_compose97i_mmx;
-#endif
-            }
-            c->inner_add_yblock = ff_snow_inner_add_yblock_mmx;
-        }
-#endif
-
         if(mm_flags & FF_MM_3DNOW){
             c->vorbis_inverse_coupling = vorbis_inverse_coupling_3dnow;
             c->vector_fmul = vector_fmul_3dnow;

libavcodec/x86/dsputil_mmx.h

@@ -167,15 +167,6 @@ void ff_vc1dsp_init_mmx(DSPContext* dsp, AVCodecContext *avctx);
 void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd);
 void ff_avg_vc1_mspel_mc00_mmx2(uint8_t *dst, const uint8_t *src, int stride, int rnd);
 
-void ff_snow_horizontal_compose97i_sse2(IDWTELEM *b, int width);
-void ff_snow_horizontal_compose97i_mmx(IDWTELEM *b, int width);
-void ff_snow_vertical_compose97i_sse2(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
-void ff_snow_vertical_compose97i_mmx(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
-void ff_snow_inner_add_yblock_sse2(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
-                                   int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
-void ff_snow_inner_add_yblock_mmx(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
-                                  int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
-
 void ff_lpc_compute_autocorr_sse2(const int32_t *data, int len, int lag,
                                    double *autoc);
 

libavcodec/x86/snowdsp_mmx.c

@@ -22,9 +22,10 @@
 #include "libavutil/x86_cpu.h"
 #include "libavcodec/avcodec.h"
 #include "libavcodec/snow.h"
+#include "libavcodec/dwt.h"
 #include "dsputil_mmx.h"
 
-void ff_snow_horizontal_compose97i_sse2(IDWTELEM *b, int width){
+static void ff_snow_horizontal_compose97i_sse2(IDWTELEM *b, int width){
     const int w2= (width+1)>>1;
     DECLARE_ALIGNED(16, IDWTELEM, temp)[width>>1];
     const int w_l= (width>>1);
@@ -213,7 +214,7 @@ void ff_snow_horizontal_compose97i_sse2(IDWTELEM *b, int width){
     }
 }
 
-void ff_snow_horizontal_compose97i_mmx(IDWTELEM *b, int width){
+static void ff_snow_horizontal_compose97i_mmx(IDWTELEM *b, int width){
     const int w2= (width+1)>>1;
     IDWTELEM temp[width >> 1];
     const int w_l= (width>>1);
@@ -436,7 +437,7 @@ void ff_snow_horizontal_compose97i_mmx(IDWTELEM *b, int width){
         "movdqa %%"s2", %%"t2" \n\t"\
         "movdqa %%"s3", %%"t3" \n\t"
 
-void ff_snow_vertical_compose97i_sse2(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width){
+static void ff_snow_vertical_compose97i_sse2(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width){
     x86_reg i = width;
 
     while(i & 0x1F)
@@ -534,7 +535,7 @@ void ff_snow_vertical_compose97i_sse2(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
         "movq %%"s3", %%"t3" \n\t"
 
 
-void ff_snow_vertical_compose97i_mmx(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width){
+static void ff_snow_vertical_compose97i_mmx(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width){
     x86_reg i = width;
     while(i & 15)
     {
@@ -847,7 +848,7 @@ snow_inner_add_yblock_mmx_mix("16", "8")
 snow_inner_add_yblock_mmx_end("32")
 }
 
-void ff_snow_inner_add_yblock_sse2(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
+static void ff_snow_inner_add_yblock_sse2(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
                            int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8){
 
     if (b_w == 16)
@@ -861,7 +862,7 @@ void ff_snow_inner_add_yblock_sse2(const uint8_t *obmc, const int obmc_stride, u
          ff_snow_inner_add_yblock(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8);
 }
 
-void ff_snow_inner_add_yblock_mmx(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
+static void ff_snow_inner_add_yblock_mmx(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
                           int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8){
     if (b_w == 16)
         inner_add_yblock_bw_16_obmc_32_mmx(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8);
@@ -870,3 +871,27 @@ void ff_snow_inner_add_yblock_mmx(const uint8_t *obmc, const int obmc_stride, ui
     else
         ff_snow_inner_add_yblock(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8);
 }
+
+void ff_dwt_init_x86(DWTContext *c)
+{
+    mm_flags = mm_support();
+
+    if (mm_flags & FF_MM_MMX) {
+        if(mm_flags & FF_MM_SSE2 & 0){
+            c->horizontal_compose97i = ff_snow_horizontal_compose97i_sse2;
+#if HAVE_7REGS
+            c->vertical_compose97i = ff_snow_vertical_compose97i_sse2;
+#endif
+            c->inner_add_yblock = ff_snow_inner_add_yblock_sse2;
+        }
+        else{
+            if(mm_flags & FF_MM_MMX2){
+            c->horizontal_compose97i = ff_snow_horizontal_compose97i_mmx;
+#if HAVE_7REGS
+            c->vertical_compose97i = ff_snow_vertical_compose97i_mmx;
+#endif
+            }
+            c->inner_add_yblock = ff_snow_inner_add_yblock_mmx;
+        }
+    }
+}