diracdec: Template DSP functions adding 10-bit versions
authorKieran Kunhya <kierank@ob-encoder.com>
Wed, 9 Dec 2015 08:24:36 +0000 (08:24 +0000)
committerRostislav Pehlivanov <atomnuker@gmail.com>
Thu, 10 Dec 2015 18:25:02 +0000 (18:25 +0000)
libavcodec/dirac_dwt.c
libavcodec/dirac_dwt.h
libavcodec/dirac_dwt_template.c [new file with mode: 0644]
libavcodec/diracdec.c
libavcodec/diracdsp.c
libavcodec/diracdsp.h
libavcodec/x86/dirac_dwt.c
libavcodec/x86/diracdsp_mmx.c
libavcodec/x86/diracdsp_yasm.asm

index ee3665e..ae847f7 100644 (file)
 #include "dirac_dwt.h"
 #include "libavcodec/x86/dirac_dwt.h"
 
+#define TEMPLATE_8bit
+#include "dirac_dwt_template.c"
+#undef TEMPLATE_8bit
 
-static void vertical_compose53iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
-                                  int width)
-{
-    int i;
-
-    for (i = 0; i < width; i++)
-        b1[i] -= (b0[i] + b2[i] + 2) >> 2;
-}
-
-
-static av_always_inline
-void interleave(IDWTELEM *dst, IDWTELEM *src0, IDWTELEM *src1, int w2, int add, int shift)
-{
-    int i;
-    for (i = 0; i < w2; i++) {
-        dst[2*i  ] = (src0[i] + add) >> shift;
-        dst[2*i+1] = (src1[i] + add) >> shift;
-    }
-}
-
-static void horizontal_compose_dirac53i(IDWTELEM *b, IDWTELEM *temp, int w)
-{
-    const int w2 = w >> 1;
-    int x;
-
-    temp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
-    for (x = 1; x < w2; x++) {
-        temp[x     ] = COMPOSE_53iL0     (b[x+w2-1], b[x     ], b[x+w2]);
-        temp[x+w2-1] = COMPOSE_DIRAC53iH0(temp[x-1], b[x+w2-1], temp[x]);
-    }
-    temp[w-1] = COMPOSE_DIRAC53iH0(temp[w2-1], b[w-1], temp[w2-1]);
-
-    interleave(b, temp, temp+w2, w2, 1, 1);
-}
-
-static void horizontal_compose_dd97i(IDWTELEM *b, IDWTELEM *tmp, int w)
-{
-    const int w2 = w >> 1;
-    int x;
-
-    tmp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
-    for (x = 1; x < w2; x++)
-        tmp[x] = COMPOSE_53iL0(b[x+w2-1], b[x], b[x+w2]);
-
-    // extend the edges
-    tmp[-1]   = tmp[0];
-    tmp[w2+1] = tmp[w2] = tmp[w2-1];
-
-    for (x = 0; x < w2; x++) {
-        b[2*x  ] = (tmp[x] + 1)>>1;
-        b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
-    }
-}
-
-static void horizontal_compose_dd137i(IDWTELEM *b, IDWTELEM *tmp, int w)
-{
-    const int w2 = w >> 1;
-    int x;
-
-    tmp[0] = COMPOSE_DD137iL0(b[w2], b[w2], b[0], b[w2  ], b[w2+1]);
-    tmp[1] = COMPOSE_DD137iL0(b[w2], b[w2], b[1], b[w2+1], b[w2+2]);
-    for (x = 2; x < w2-1; x++)
-        tmp[x] = COMPOSE_DD137iL0(b[x+w2-2], b[x+w2-1], b[x], b[x+w2], b[x+w2+1]);
-    tmp[w2-1] = COMPOSE_DD137iL0(b[w-3], b[w-2], b[w2-1], b[w-1], b[w-1]);
-
-    // extend the edges
-    tmp[-1]   = tmp[0];
-    tmp[w2+1] = tmp[w2] = tmp[w2-1];
-
-    for (x = 0; x < w2; x++) {
-        b[2*x  ] = (tmp[x] + 1)>>1;
-        b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
-    }
-}
-
-static av_always_inline
-void horizontal_compose_haari(IDWTELEM *b, IDWTELEM *temp, int w, int shift)
-{
-    const int w2 = w >> 1;
-    int x;
-
-    for (x = 0; x < w2; x++) {
-        temp[x   ] = COMPOSE_HAARiL0(b[x   ], b[x+w2]);
-        temp[x+w2] = COMPOSE_HAARiH0(b[x+w2], temp[x]);
-    }
-
-    interleave(b, temp, temp+w2, w2, shift, shift);
-}
-
-static void horizontal_compose_haar0i(IDWTELEM *b, IDWTELEM *temp, int w)
-{
-    horizontal_compose_haari(b, temp, w, 0);
-}
-
-static void horizontal_compose_haar1i(IDWTELEM *b, IDWTELEM *temp, int w)
-{
-    horizontal_compose_haari(b, temp, w, 1);
-}
-
-static void horizontal_compose_fidelityi(IDWTELEM *b, IDWTELEM *tmp, int w)
-{
-    const int w2 = w >> 1;
-    int i, x;
-    IDWTELEM v[8];
-
-    for (x = 0; x < w2; x++) {
-        for (i = 0; i < 8; i++)
-            v[i] = b[av_clip(x-3+i, 0, w2-1)];
-        tmp[x] = COMPOSE_FIDELITYiH0(v[0], v[1], v[2], v[3], b[x+w2], v[4], v[5], v[6], v[7]);
-    }
-
-    for (x = 0; x < w2; x++) {
-        for (i = 0; i < 8; i++)
-            v[i] = tmp[av_clip(x-4+i, 0, w2-1)];
-        tmp[x+w2] = COMPOSE_FIDELITYiL0(v[0], v[1], v[2], v[3], b[x], v[4], v[5], v[6], v[7]);
-    }
-
-    interleave(b, tmp+w2, tmp, w2, 0, 0);
-}
-
-static void horizontal_compose_daub97i(IDWTELEM *b, IDWTELEM *temp, int w)
-{
-    const int w2 = w >> 1;
-    int x, b0, b1, b2;
-
-    temp[0] = COMPOSE_DAUB97iL1(b[w2], b[0], b[w2]);
-    for (x = 1; x < w2; x++) {
-        temp[x     ] = COMPOSE_DAUB97iL1(b[x+w2-1], b[x     ], b[x+w2]);
-        temp[x+w2-1] = COMPOSE_DAUB97iH1(temp[x-1], b[x+w2-1], temp[x]);
-    }
-    temp[w-1] = COMPOSE_DAUB97iH1(temp[w2-1], b[w-1], temp[w2-1]);
-
-    // second stage combined with interleave and shift
-    b0 = b2 = COMPOSE_DAUB97iL0(temp[w2], temp[0], temp[w2]);
-    b[0] = (b0 + 1) >> 1;
-    for (x = 1; x < w2; x++) {
-        b2 = COMPOSE_DAUB97iL0(temp[x+w2-1], temp[x     ], temp[x+w2]);
-        b1 = COMPOSE_DAUB97iH0(          b0, temp[x+w2-1], b2        );
-        b[2*x-1] = (b1 + 1) >> 1;
-        b[2*x  ] = (b2 + 1) >> 1;
-        b0 = b2;
-    }
-    b[w-1] = (COMPOSE_DAUB97iH0(b2, temp[w-1], b2) + 1) >> 1;
-}
-
-static void vertical_compose_dirac53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        b1[i] = COMPOSE_DIRAC53iH0(b0[i], b1[i], b2[i]);
-    }
-}
-
-static void vertical_compose_dd97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
-                                  IDWTELEM *b3, IDWTELEM *b4, int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        b2[i] = COMPOSE_DD97iH0(b0[i], b1[i], b2[i], b3[i], b4[i]);
-    }
-}
-
-static void vertical_compose_dd137iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
-                                      IDWTELEM *b3, IDWTELEM *b4, int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        b2[i] = COMPOSE_DD137iL0(b0[i], b1[i], b2[i], b3[i], b4[i]);
-    }
-}
-
-static void vertical_compose_haar(IDWTELEM *b0, IDWTELEM *b1, int width)
-{
-    int i;
-
-    for (i = 0; i < width; i++) {
-        b0[i] = COMPOSE_HAARiL0(b0[i], b1[i]);
-        b1[i] = COMPOSE_HAARiH0(b1[i], b0[i]);
-    }
-}
-
-static void vertical_compose_fidelityiH0(IDWTELEM *dst, IDWTELEM *b[8], int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        dst[i] = COMPOSE_FIDELITYiH0(b[0][i], b[1][i], b[2][i], b[3][i], dst[i], b[4][i], b[5][i], b[6][i], b[7][i]);
-    }
-}
-
-static void vertical_compose_fidelityiL0(IDWTELEM *dst, IDWTELEM *b[8], int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        dst[i] = COMPOSE_FIDELITYiL0(b[0][i], b[1][i], b[2][i], b[3][i], dst[i], b[4][i], b[5][i], b[6][i], b[7][i]);
-    }
-}
-
-static void vertical_compose_daub97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        b1[i] = COMPOSE_DAUB97iH0(b0[i], b1[i], b2[i]);
-    }
-}
-
-static void vertical_compose_daub97iH1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        b1[i] = COMPOSE_DAUB97iH1(b0[i], b1[i], b2[i]);
-    }
-}
-
-static void vertical_compose_daub97iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        b1[i] = COMPOSE_DAUB97iL0(b0[i], b1[i], b2[i]);
-    }
-}
-
-static void vertical_compose_daub97iL1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
-{
-    int i;
-
-    for(i=0; i<width; i++){
-        b1[i] = COMPOSE_DAUB97iL1(b0[i], b1[i], b2[i]);
-    }
-}
-
-
-static void spatial_compose_dd97i_dy(DWTContext *d, int level, int width, int height, int stride)
-{
-    vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
-    vertical_compose_5tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
-    DWTCompose *cs = d->cs + level;
-
-    int i, y = cs->y;
-    IDWTELEM *b[8];
-    for (i = 0; i < 6; i++)
-        b[i] = cs->b[i];
-    b[6] = d->buffer + av_clip(y+5, 0, height-2)*stride;
-    b[7] = d->buffer + av_clip(y+6, 1, height-1)*stride;
-
-        if(y+5<(unsigned)height) vertical_compose_l0(      b[5], b[6], b[7],       width);
-        if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);
-
-        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
-        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
-
-    for (i = 0; i < 6; i++)
-        cs->b[i] = b[i+2];
-    cs->y += 2;
-}
-
-static void spatial_compose_dirac53i_dy(DWTContext *d, int level, int width, int height, int stride)
-{
-    vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
-    vertical_compose_3tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
-    DWTCompose *cs = d->cs + level;
-
-    int y= cs->y;
-    IDWTELEM *b[4] = { cs->b[0], cs->b[1] };
-    b[2] = d->buffer + avpriv_mirror(y+1, height-1)*stride;
-    b[3] = d->buffer + avpriv_mirror(y+2, height-1)*stride;
-
-        if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
-        if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);
-
-        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
-        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
-
-    cs->b[0] = b[2];
-    cs->b[1] = b[3];
-    cs->y += 2;
-}
-
-
-static void spatial_compose_dd137i_dy(DWTContext *d, int level, int width, int height, int stride)
-{
-    vertical_compose_5tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
-    vertical_compose_5tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
-    DWTCompose *cs = d->cs + level;
-
-    int i, y = cs->y;
-    IDWTELEM *b[10];
-    for (i = 0; i < 8; i++)
-        b[i] = cs->b[i];
-    b[8] = d->buffer + av_clip(y+7, 0, height-2)*stride;
-    b[9] = d->buffer + av_clip(y+8, 1, height-1)*stride;
-
-        if(y+5<(unsigned)height) vertical_compose_l0(b[3], b[5], b[6], b[7], b[9], width);
-        if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);
+#define TEMPLATE_10bit
+#include "dirac_dwt_template.c"
+#undef TEMPLATE_10bit
 
-        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
-        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
+#define TEMPLATE_12bit
+#include "dirac_dwt_template.c"
+#undef TEMPLATE_12bit
 
-    for (i = 0; i < 8; i++)
-        cs->b[i] = b[i+2];
-    cs->y += 2;
-}
-
-// haar makes the assumption that height is even (always true for dirac)
-static void spatial_compose_haari_dy(DWTContext *d, int level, int width, int height, int stride)
-{
-    vertical_compose_2tap vertical_compose = (void*)d->vertical_compose;
-    int y = d->cs[level].y;
-    IDWTELEM *b0 = d->buffer + (y-1)*stride;
-    IDWTELEM *b1 = d->buffer + (y  )*stride;
-
-    vertical_compose(b0, b1, width);
-    d->horizontal_compose(b0, d->temp, width);
-    d->horizontal_compose(b1, d->temp, width);
-
-    d->cs[level].y += 2;
-}
-
-// Don't do sliced idwt for fidelity; the 9 tap filter makes it a bit annoying
-// Fortunately, this filter isn't used in practice.
-static void spatial_compose_fidelity(DWTContext *d, int level, int width, int height, int stride)
-{
-    vertical_compose_9tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
-    vertical_compose_9tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
-    int i, y;
-    IDWTELEM *b[8];
-
-    for (y = 1; y < height; y += 2) {
-        for (i = 0; i < 8; i++)
-            b[i] = d->buffer + av_clip((y-7 + 2*i), 0, height-2)*stride;
-        vertical_compose_h0(d->buffer + y*stride, b, width);
-    }
-
-    for (y = 0; y < height; y += 2) {
-        for (i = 0; i < 8; i++)
-            b[i] = d->buffer + av_clip((y-7 + 2*i), 1, height-1)*stride;
-        vertical_compose_l0(d->buffer + y*stride, b, width);
-    }
-
-    for (y = 0; y < height; y++)
-        d->horizontal_compose(d->buffer + y*stride, d->temp, width);
-
-    d->cs[level].y = height+1;
-}
-
-static void spatial_compose_daub97i_dy(DWTContext *d, int level, int width, int height, int stride)
-{
-    vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
-    vertical_compose_3tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
-    vertical_compose_3tap vertical_compose_l1 = (void*)d->vertical_compose_l1;
-    vertical_compose_3tap vertical_compose_h1 = (void*)d->vertical_compose_h1;
-    DWTCompose *cs = d->cs + level;
-
-    int i, y = cs->y;
-    IDWTELEM *b[6];
-    for (i = 0; i < 4; i++)
-        b[i] = cs->b[i];
-    b[4] = d->buffer + avpriv_mirror(y+3, height-1)*stride;
-    b[5] = d->buffer + avpriv_mirror(y+4, height-1)*stride;
-
-        if(y+3<(unsigned)height) vertical_compose_l1(b[3], b[4], b[5], width);
-        if(y+2<(unsigned)height) vertical_compose_h1(b[2], b[3], b[4], width);
-        if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
-        if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);
-
-        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
-        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
-
-    for (i = 0; i < 4; i++)
-        cs->b[i] = b[i+2];
-    cs->y += 2;
-}
-
-
-static void spatial_compose97i_init2(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
-{
-    cs->b[0] = buffer + avpriv_mirror(-3-1, height-1)*stride;
-    cs->b[1] = buffer + avpriv_mirror(-3  , height-1)*stride;
-    cs->b[2] = buffer + avpriv_mirror(-3+1, height-1)*stride;
-    cs->b[3] = buffer + avpriv_mirror(-3+2, height-1)*stride;
-    cs->y = -3;
-}
-
-static void spatial_compose53i_init2(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
-{
-    cs->b[0] = buffer + avpriv_mirror(-1-1, height-1)*stride;
-    cs->b[1] = buffer + avpriv_mirror(-1  , height-1)*stride;
-    cs->y = -1;
-}
-
-static void spatial_compose_dd97i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
-{
-    cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
-    cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
-    cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
-    cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
-    cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
-    cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
-    cs->y = -5;
-}
-
-static void spatial_compose_dd137i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
-{
-    cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
-    cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
-    cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
-    cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
-    cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
-    cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
-    cs->b[6] = buffer + av_clip(-5+5, 0, height-2)*stride;
-    cs->b[7] = buffer + av_clip(-5+6, 1, height-1)*stride;
-    cs->y = -5;
-}
-
-int ff_spatial_idwt_init2(DWTContext *d, IDWTELEM *buffer, int width, int height,
+int ff_spatial_idwt_init2(DWTContext *d, uint8_t *buffer, int width, int height,
                           int stride, enum dwt_type type, int decomposition_count,
-                          IDWTELEM *temp)
+                          uint8_t *temp, int bit_depth)
 {
-    int level;
+    int ret = 0;
 
-    d->buffer = buffer;
-    d->width = width;
-    d->height = height;
-    d->stride = stride;
-    d->decomposition_count = decomposition_count;
-    d->temp = temp + 8;
+    if (bit_depth == 8)
+        ret = ff_spatial_idwt_init2_8bit(d, buffer, width, height, stride, type, decomposition_count, temp);
+    else if (bit_depth == 10)
+        ret = ff_spatial_idwt_init2_10bit(d, buffer, width, height, stride, type, decomposition_count, temp);
+    else if (bit_depth == 12)
+        ret = ff_spatial_idwt_init2_12bit(d, buffer, width, height, stride, type, decomposition_count, temp);
+    else
+        av_log(NULL, AV_LOG_WARNING, "Unsupported bit depth = %i\n", bit_depth);
 
-    for(level=decomposition_count-1; level>=0; level--){
-        int hl = height >> level;
-        int stride_l = stride << level;
-
-        switch(type){
-        case DWT_DIRAC_DD9_7:
-            spatial_compose_dd97i_init(d->cs+level, buffer, hl, stride_l);
-            break;
-        case DWT_DIRAC_LEGALL5_3:
-            spatial_compose53i_init2(d->cs+level, buffer, hl, stride_l);
-            break;
-        case DWT_DIRAC_DD13_7:
-            spatial_compose_dd137i_init(d->cs+level, buffer, hl, stride_l);
-            break;
-        case DWT_DIRAC_HAAR0:
-        case DWT_DIRAC_HAAR1:
-            d->cs[level].y = 1;
-            break;
-        case DWT_DIRAC_DAUB9_7:
-            spatial_compose97i_init2(d->cs+level, buffer, hl, stride_l);
-            break;
-        default:
-            d->cs[level].y = 0;
-            break;
-        }
-    }
-
-    switch (type) {
-    case DWT_DIRAC_DD9_7:
-        d->spatial_compose = spatial_compose_dd97i_dy;
-        d->vertical_compose_l0 = (void*)vertical_compose53iL0;
-        d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0;
-        d->horizontal_compose = horizontal_compose_dd97i;
-        d->support = 7;
-        break;
-    case DWT_DIRAC_LEGALL5_3:
-        d->spatial_compose = spatial_compose_dirac53i_dy;
-        d->vertical_compose_l0 = (void*)vertical_compose53iL0;
-        d->vertical_compose_h0 = (void*)vertical_compose_dirac53iH0;
-        d->horizontal_compose = horizontal_compose_dirac53i;
-        d->support = 3;
-        break;
-    case DWT_DIRAC_DD13_7:
-        d->spatial_compose = spatial_compose_dd137i_dy;
-        d->vertical_compose_l0 = (void*)vertical_compose_dd137iL0;
-        d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0;
-        d->horizontal_compose = horizontal_compose_dd137i;
-        d->support = 7;
-        break;
-    case DWT_DIRAC_HAAR0:
-    case DWT_DIRAC_HAAR1:
-        d->spatial_compose = spatial_compose_haari_dy;
-        d->vertical_compose = (void*)vertical_compose_haar;
-        if (type == DWT_DIRAC_HAAR0)
-            d->horizontal_compose = horizontal_compose_haar0i;
-        else
-            d->horizontal_compose = horizontal_compose_haar1i;
-        d->support = 1;
-        break;
-    case DWT_DIRAC_FIDELITY:
-        d->spatial_compose = spatial_compose_fidelity;
-        d->vertical_compose_l0 = (void*)vertical_compose_fidelityiL0;
-        d->vertical_compose_h0 = (void*)vertical_compose_fidelityiH0;
-        d->horizontal_compose = horizontal_compose_fidelityi;
-        d->support = 0; // not really used
-        break;
-    case DWT_DIRAC_DAUB9_7:
-        d->spatial_compose = spatial_compose_daub97i_dy;
-        d->vertical_compose_l0 = (void*)vertical_compose_daub97iL0;
-        d->vertical_compose_h0 = (void*)vertical_compose_daub97iH0;
-        d->vertical_compose_l1 = (void*)vertical_compose_daub97iL1;
-        d->vertical_compose_h1 = (void*)vertical_compose_daub97iH1;
-        d->horizontal_compose = horizontal_compose_daub97i;
-        d->support = 5;
-        break;
-    default:
+    if (ret) {
         av_log(NULL, AV_LOG_ERROR, "Unknown wavelet type %d\n", type);
         return AVERROR_INVALIDDATA;
     }
 
-    if (HAVE_MMX) ff_spatial_idwt_init_mmx(d, type);
-
+    if (bit_depth == 8 && HAVE_MMX)
+        ff_spatial_idwt_init_mmx(d, type);
     return 0;
 }
 
@@ -558,4 +75,3 @@ void ff_spatial_idwt_slice2(DWTContext *d, int y)
             d->spatial_compose(d, level, wl, hl, stride_l);
     }
 }
-
index e5e447b..41842b5 100644 (file)
@@ -30,21 +30,21 @@ typedef short IDWTELEM;
 #define MAX_DECOMPOSITIONS 8
 
 typedef struct DWTCompose {
-    IDWTELEM *b[MAX_DWT_SUPPORT];
+    uint8_t *b[MAX_DWT_SUPPORT];
     int y;
 } DWTCompose;
 
 struct DWTContext;
 
 // Possible prototypes for vertical_compose functions
-typedef void (*vertical_compose_2tap)(IDWTELEM *b0, IDWTELEM *b1, int width);
-typedef void (*vertical_compose_3tap)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width);
-typedef void (*vertical_compose_5tap)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, int width);
-typedef void (*vertical_compose_9tap)(IDWTELEM *dst, IDWTELEM *b[8], int width);
+typedef void (*vertical_compose_2tap)(uint8_t *b0, uint8_t *b1, int width);
+typedef void (*vertical_compose_3tap)(uint8_t *b0, uint8_t *b1, uint8_t *b2, int width);
+typedef void (*vertical_compose_5tap)(uint8_t *b0, uint8_t *b1, uint8_t *b2, uint8_t *b3, uint8_t *b4, int width);
+typedef void (*vertical_compose_9tap)(uint8_t *dst, uint8_t *b[8], int width);
 
 typedef struct DWTContext {
-    IDWTELEM *buffer;
-    IDWTELEM *temp;
+    uint8_t *buffer;
+    uint8_t *temp;
     int width;
     int height;
     int stride;
@@ -57,7 +57,7 @@ typedef struct DWTContext {
     void (*vertical_compose_l1)(void);
     void (*vertical_compose_h1)(void);
     void (*vertical_compose)(void);     ///< one set of lowpass and highpass combined
-    void (*horizontal_compose)(IDWTELEM *b, IDWTELEM *tmp, int width);
+    void (*horizontal_compose)(uint8_t *b, uint8_t *tmp, int width);
 
     DWTCompose cs[MAX_DECOMPOSITIONS];
 } DWTContext;
@@ -76,9 +76,9 @@ enum dwt_type {
 };
 
 // -1 if an error occurred, e.g. the dwt_type isn't recognized
-int ff_spatial_idwt_init2(DWTContext *d, IDWTELEM *buffer, int width, int height,
+int ff_spatial_idwt_init2(DWTContext *d, uint8_t *buffer, int width, int height,
                           int stride, enum dwt_type type, int decomposition_count,
-                          IDWTELEM *temp);
+                          uint8_t *temp, int bit_depth);
 
 void ff_spatial_idwt_slice2(DWTContext *d, int y);
 
diff --git a/libavcodec/dirac_dwt_template.c b/libavcodec/dirac_dwt_template.c
new file mode 100644 (file)
index 0000000..dfcf1de
--- /dev/null
@@ -0,0 +1,615 @@
+/*
+ * Copyright (C) 2004-2010 Michael Niedermayer <michaelni@gmx.at>
+ * Copyright (C) 2008 David Conrad
+ * Copyright (C) 2015 Open Broadcast Systems Ltd.
+ * Author    (C) 2015 Rostislav Pehlivanov <atomnuker@gmail.com>
+ *
+ * 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
+ */
+
+#if defined(TEMPLATE_8bit)
+
+#    define RENAME(N)   N ## _8bit
+#    define TYPE        int16_t
+#    undef  TEMPLATE_8bit
+
+#elif defined(TEMPLATE_10bit)
+
+#    define RENAME(N)   N ## _10bit
+#    define TYPE        int32_t
+#    undef  TEMPLATE_10bit
+
+#elif defined(TEMPLATE_12bit)
+
+#    define RENAME(N)   N ## _12bit
+#    define TYPE        int32_t
+#    undef  TEMPLATE_12bit
+
+#endif
+
+static void RENAME(vertical_compose53iL0)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2,
+                                          int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+    for (i = 0; i < width; i++)
+        b1[i] -= (b0[i] + b2[i] + 2) >> 2;
+}
+
+static av_always_inline void RENAME(interleave)(TYPE *dst, TYPE *src0, TYPE *src1, int w2,
+                                                int add, int shift)
+{
+    int i;
+    for (i = 0; i < w2; i++) {
+        dst[2*i  ] = (src0[i] + add) >> shift;
+        dst[2*i+1] = (src1[i] + add) >> shift;
+    }
+}
+
+static void RENAME(horizontal_compose_dirac53i)(uint8_t *_b, uint8_t *_temp, int w)
+{
+    int x;
+    const int w2 = w >> 1;
+    TYPE *b     = (TYPE *)_b;
+    TYPE *temp  = (TYPE *)_temp;
+
+    temp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
+    for (x = 1; x < w2; x++) {
+        temp[x     ] = COMPOSE_53iL0     (b[x+w2-1], b[x     ], b[x+w2]);
+        temp[x+w2-1] = COMPOSE_DIRAC53iH0(temp[x-1], b[x+w2-1], temp[x]);
+    }
+    temp[w-1] = COMPOSE_DIRAC53iH0(temp[w2-1], b[w-1], temp[w2-1]);
+
+    RENAME(interleave)(b, temp, temp+w2, w2, 1, 1);
+}
+
+static void RENAME(horizontal_compose_dd97i)(uint8_t *_b, uint8_t *_tmp, int w)
+{
+    int x;
+    const int w2 = w >> 1;
+    TYPE *b   = (TYPE *)_b;
+    TYPE *tmp = (TYPE *)_tmp;
+
+    tmp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
+    for (x = 1; x < w2; x++)
+        tmp[x] = COMPOSE_53iL0(b[x+w2-1], b[x], b[x+w2]);
+
+    // extend the edges
+    tmp[-1]   = tmp[0];
+    tmp[w2+1] = tmp[w2] = tmp[w2-1];
+
+    for (x = 0; x < w2; x++) {
+        b[2*x  ] = (tmp[x] + 1)>>1;
+        b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
+    }
+}
+
+static void RENAME(horizontal_compose_dd137i)(uint8_t *_b, uint8_t *_tmp, int w)
+{
+    const int w2 = w >> 1;
+    int x;
+    TYPE *b   = (TYPE *)_b;
+    TYPE *tmp = (TYPE *)_tmp;
+
+    tmp[0] = COMPOSE_DD137iL0(b[w2], b[w2], b[0], b[w2  ], b[w2+1]);
+    tmp[1] = COMPOSE_DD137iL0(b[w2], b[w2], b[1], b[w2+1], b[w2+2]);
+    for (x = 2; x < w2-1; x++)
+        tmp[x] = COMPOSE_DD137iL0(b[x+w2-2], b[x+w2-1], b[x], b[x+w2], b[x+w2+1]);
+    tmp[w2-1] = COMPOSE_DD137iL0(b[w-3], b[w-2], b[w2-1], b[w-1], b[w-1]);
+
+    // extend the edges
+    tmp[-1]   = tmp[0];
+    tmp[w2+1] = tmp[w2] = tmp[w2-1];
+
+    for (x = 0; x < w2; x++) {
+        b[2*x  ] = (tmp[x] + 1)>>1;
+        b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
+    }
+}
+
+static av_always_inline void RENAME(horizontal_compose_haari)(TYPE *b, TYPE *temp,
+                                                              int w, int shift)
+{
+    const int w2 = w >> 1;
+    int x;
+
+    for (x = 0; x < w2; x++) {
+        temp[x   ] = COMPOSE_HAARiL0(b[x   ], b[x+w2]);
+        temp[x+w2] = COMPOSE_HAARiH0(b[x+w2], temp[x]);
+    }
+
+    RENAME(interleave)(b, temp, temp+w2, w2, shift, shift);
+}
+
+static void RENAME(horizontal_compose_haar0i)(uint8_t *_b, uint8_t *_temp, int w)
+{
+    TYPE *b    = (TYPE *)_b;
+    TYPE *temp = (TYPE *)_temp;
+    RENAME(horizontal_compose_haari)(b, temp, w, 0);
+}
+
+static void RENAME(horizontal_compose_haar1i)(uint8_t *_b, uint8_t *_temp, int w)
+{
+    TYPE *b    = (TYPE *)_b;
+    TYPE *temp = (TYPE *)_temp;
+    RENAME(horizontal_compose_haari)(b, temp, w, 1);
+}
+
+static void RENAME(horizontal_compose_fidelityi)(uint8_t *_b, uint8_t *_tmp, int w)
+{
+    const int w2 = w >> 1;
+    int i, x;
+    TYPE v[8];
+    TYPE *b   = (TYPE *)_b;
+    TYPE *tmp = (TYPE *)_tmp;
+
+    for (x = 0; x < w2; x++) {
+        for (i = 0; i < 8; i++)
+            v[i] = b[av_clip(x-3+i, 0, w2-1)];
+        tmp[x] = COMPOSE_FIDELITYiH0(v[0], v[1], v[2], v[3], b[x+w2], v[4], v[5], v[6], v[7]);
+    }
+
+    for (x = 0; x < w2; x++) {
+        for (i = 0; i < 8; i++)
+            v[i] = tmp[av_clip(x-4+i, 0, w2-1)];
+        tmp[x+w2] = COMPOSE_FIDELITYiL0(v[0], v[1], v[2], v[3], b[x], v[4], v[5], v[6], v[7]);
+    }
+
+    RENAME(interleave)(b, tmp+w2, tmp, w2, 0, 0);
+}
+
+static void RENAME(horizontal_compose_daub97i)(uint8_t *_b, uint8_t *_temp, int w)
+{
+    const int w2 = w >> 1;
+    int x, b0, b1, b2;
+    TYPE *b    = (TYPE *)_b;
+    TYPE *temp = (TYPE *)_temp;
+
+    temp[0] = COMPOSE_DAUB97iL1(b[w2], b[0], b[w2]);
+    for (x = 1; x < w2; x++) {
+        temp[x     ] = COMPOSE_DAUB97iL1(b[x+w2-1], b[x     ], b[x+w2]);
+        temp[x+w2-1] = COMPOSE_DAUB97iH1(temp[x-1], b[x+w2-1], temp[x]);
+    }
+    temp[w-1] = COMPOSE_DAUB97iH1(temp[w2-1], b[w-1], temp[w2-1]);
+
+    // second stage combined with interleave and shift
+    b0 = b2 = COMPOSE_DAUB97iL0(temp[w2], temp[0], temp[w2]);
+    b[0] = (b0 + 1) >> 1;
+    for (x = 1; x < w2; x++) {
+        b2 = COMPOSE_DAUB97iL0(temp[x+w2-1], temp[x     ], temp[x+w2]);
+        b1 = COMPOSE_DAUB97iH0(          b0, temp[x+w2-1], b2        );
+        b[2*x-1] = (b1 + 1) >> 1;
+        b[2*x  ] = (b2 + 1) >> 1;
+        b0 = b2;
+    }
+    b[w-1] = (COMPOSE_DAUB97iH0(b2, temp[w-1], b2) + 1) >> 1;
+}
+
+static void RENAME(vertical_compose_dirac53iH0)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2,
+                                                int width)
+{
+    int i;
+    TYPE *b0  = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+    for(i=0; i<width; i++){
+        b1[i] = COMPOSE_DIRAC53iH0(b0[i], b1[i], b2[i]);
+    }
+}
+
+static void RENAME(vertical_compose_dd97iH0)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2,
+                                             uint8_t *_b3, uint8_t *_b4, int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+    TYPE *b3 = (TYPE *)_b3;
+    TYPE *b4 = (TYPE *)_b4;
+    for(i=0; i<width; i++){
+        b2[i] = COMPOSE_DD97iH0(b0[i], b1[i], b2[i], b3[i], b4[i]);
+    }
+}
+
+static void RENAME(vertical_compose_dd137iL0)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2,
+                                              uint8_t *_b3, uint8_t *_b4, int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+    TYPE *b3 = (TYPE *)_b3;
+    TYPE *b4 = (TYPE *)_b4;
+    for(i=0; i<width; i++){
+        b2[i] = COMPOSE_DD137iL0(b0[i], b1[i], b2[i], b3[i], b4[i]);
+    }
+}
+
+static void RENAME(vertical_compose_haar)(uint8_t *_b0, uint8_t *_b1, int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+
+    for (i = 0; i < width; i++) {
+        b0[i] = COMPOSE_HAARiL0(b0[i], b1[i]);
+        b1[i] = COMPOSE_HAARiH0(b1[i], b0[i]);
+    }
+}
+
+static void RENAME(vertical_compose_fidelityiH0)(uint8_t *_dst, uint8_t *_b[8], int width)
+{
+    int i;
+    TYPE *dst = (TYPE *)_dst;
+    TYPE *b0  = (TYPE *)_b[0];
+    TYPE *b1  = (TYPE *)_b[1];
+    TYPE *b2  = (TYPE *)_b[2];
+    TYPE *b3  = (TYPE *)_b[3];
+    TYPE *b4  = (TYPE *)_b[4];
+    TYPE *b5  = (TYPE *)_b[5];
+    TYPE *b6  = (TYPE *)_b[6];
+    TYPE *b7  = (TYPE *)_b[7];
+    for(i=0; i<width; i++){
+        dst[i] = COMPOSE_FIDELITYiH0(b0[i], b1[i], b2[i], b3[i], dst[i], b4[i], b5[i], b6[i], b7[i]);
+    }
+}
+
+static void RENAME(vertical_compose_fidelityiL0)(uint8_t *_dst, uint8_t *_b[8], int width)
+{
+    int i;
+    TYPE *dst = (TYPE *)_dst;
+    TYPE *b0  = (TYPE *)_b[0];
+    TYPE *b1  = (TYPE *)_b[1];
+    TYPE *b2  = (TYPE *)_b[2];
+    TYPE *b3  = (TYPE *)_b[3];
+    TYPE *b4  = (TYPE *)_b[4];
+    TYPE *b5  = (TYPE *)_b[5];
+    TYPE *b6  = (TYPE *)_b[6];
+    TYPE *b7  = (TYPE *)_b[7];
+
+    for(i=0; i<width; i++){
+        dst[i] = COMPOSE_FIDELITYiL0(b0[i], b1[i], b2[i], b3[i], dst[i], b4[i], b5[i], b6[i], b7[i]);
+    }
+}
+
+static void RENAME(vertical_compose_daub97iH0)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+
+    for(i=0; i<width; i++){
+        b1[i] = COMPOSE_DAUB97iH0(b0[i], b1[i], b2[i]);
+    }
+}
+
+static void RENAME(vertical_compose_daub97iH1)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+
+    for(i=0; i<width; i++){
+        b1[i] = COMPOSE_DAUB97iH1(b0[i], b1[i], b2[i]);
+    }
+}
+
+static void RENAME(vertical_compose_daub97iL0)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+
+    for(i=0; i<width; i++){
+        b1[i] = COMPOSE_DAUB97iL0(b0[i], b1[i], b2[i]);
+    }
+}
+
+static void RENAME(vertical_compose_daub97iL1)(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, int width)
+{
+    int i;
+    TYPE *b0 = (TYPE *)_b0;
+    TYPE *b1 = (TYPE *)_b1;
+    TYPE *b2 = (TYPE *)_b2;
+
+    for(i=0; i<width; i++){
+        b1[i] = COMPOSE_DAUB97iL1(b0[i], b1[i], b2[i]);
+    }
+}
+
+static void RENAME(spatial_compose_dd97i_dy)(DWTContext *d, int level, int width, int height, int stride)
+{
+    vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
+    vertical_compose_5tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
+    DWTCompose *cs = d->cs + level;
+
+    int i, y = cs->y;
+    uint8_t *b[8];
+    for (i = 0; i < 6; i++)
+        b[i] = cs->b[i];
+    b[6] = d->buffer + av_clip(y+5, 0, height-2)*stride;
+    b[7] = d->buffer + av_clip(y+6, 1, height-1)*stride;
+
+    if(y+5<(unsigned)height) vertical_compose_l0(      b[5], b[6], b[7],       width);
+    if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);
+
+    if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
+    if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
+
+    for (i = 0; i < 6; i++)
+        cs->b[i] = b[i+2];
+    cs->y += 2;
+}
+
+static void RENAME(spatial_compose_dirac53i_dy)(DWTContext *d, int level, int width, int height, int stride)
+{
+    vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
+    vertical_compose_3tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
+    DWTCompose *cs = d->cs + level;
+
+    int y= cs->y;
+    uint8_t *b[4] = { cs->b[0], cs->b[1] };
+    b[2] = d->buffer + avpriv_mirror(y+1, height-1)*stride;
+    b[3] = d->buffer + avpriv_mirror(y+2, height-1)*stride;
+
+    if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
+    if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);
+
+    if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
+    if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
+
+    cs->b[0] = b[2];
+    cs->b[1] = b[3];
+    cs->y += 2;
+}
+
+static void RENAME(spatial_compose_dd137i_dy)(DWTContext *d, int level, int width, int height, int stride)
+{
+    vertical_compose_5tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
+    vertical_compose_5tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
+    DWTCompose *cs = d->cs + level;
+
+    int i, y = cs->y;
+    uint8_t *b[10];
+    for (i = 0; i < 8; i++)
+        b[i] = cs->b[i];
+    b[8] = d->buffer + av_clip(y+7, 0, height-2)*stride;
+    b[9] = d->buffer + av_clip(y+8, 1, height-1)*stride;
+
+    if(y+5<(unsigned)height) vertical_compose_l0(b[3], b[5], b[6], b[7], b[9], width);
+    if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);
+
+    if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
+    if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
+
+    for (i = 0; i < 8; i++)
+        cs->b[i] = b[i+2];
+    cs->y += 2;
+}
+
+// haar makes the assumption that height is even (always true for dirac)
+static void RENAME(spatial_compose_haari_dy)(DWTContext *d, int level, int width, int height, int stride)
+{
+    vertical_compose_2tap vertical_compose = (void*)d->vertical_compose;
+    int y = d->cs[level].y;
+    uint8_t *b0 = d->buffer + (y-1)*stride;
+    uint8_t *b1 = d->buffer + (y  )*stride;
+
+    vertical_compose(b0, b1, width);
+    d->horizontal_compose(b0, d->temp, width);
+    d->horizontal_compose(b1, d->temp, width);
+
+    d->cs[level].y += 2;
+}
+
+// Don't do sliced idwt for fidelity; the 9 tap filter makes it a bit annoying
+// Fortunately, this filter isn't used in practice.
+static void RENAME(spatial_compose_fidelity)(DWTContext *d, int level, int width, int height, int stride)
+{
+    vertical_compose_9tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
+    vertical_compose_9tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
+    int i, y;
+    uint8_t *b[8];
+
+    for (y = 1; y < height; y += 2) {
+        for (i = 0; i < 8; i++)
+            b[i] = d->buffer + av_clip((y-7 + 2*i), 0, height-2)*stride;
+        vertical_compose_h0(d->buffer + y*stride, b, width);
+    }
+
+    for (y = 0; y < height; y += 2) {
+        for (i = 0; i < 8; i++)
+            b[i] = d->buffer + av_clip((y-7 + 2*i), 1, height-1)*stride;
+        vertical_compose_l0(d->buffer + y*stride, b, width);
+    }
+
+    for (y = 0; y < height; y++)
+        d->horizontal_compose(d->buffer + y*stride, d->temp, width);
+
+    d->cs[level].y = height+1;
+}
+
+static void RENAME(spatial_compose_daub97i_dy)(DWTContext *d, int level, int width, int height, int stride)
+{
+    vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
+    vertical_compose_3tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
+    vertical_compose_3tap vertical_compose_l1 = (void*)d->vertical_compose_l1;
+    vertical_compose_3tap vertical_compose_h1 = (void*)d->vertical_compose_h1;
+    DWTCompose *cs = d->cs + level;
+
+    int i, y = cs->y;
+    uint8_t *b[6];
+    for (i = 0; i < 4; i++)
+        b[i] = cs->b[i];
+    b[4] = d->buffer + avpriv_mirror(y+3, height-1)*stride;
+    b[5] = d->buffer + avpriv_mirror(y+4, height-1)*stride;
+
+    if(y+3<(unsigned)height) vertical_compose_l1(b[3], b[4], b[5], width);
+    if(y+2<(unsigned)height) vertical_compose_h1(b[2], b[3], b[4], width);
+    if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
+    if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);
+
+    if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
+    if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
+
+    for (i = 0; i < 4; i++)
+        cs->b[i] = b[i+2];
+    cs->y += 2;
+}
+
+static void RENAME(spatial_compose97i_init2)(DWTCompose *cs, uint8_t *buffer, int height, int stride)
+{
+    cs->b[0] = buffer + avpriv_mirror(-3-1, height-1)*stride;
+    cs->b[1] = buffer + avpriv_mirror(-3  , height-1)*stride;
+    cs->b[2] = buffer + avpriv_mirror(-3+1, height-1)*stride;
+    cs->b[3] = buffer + avpriv_mirror(-3+2, height-1)*stride;
+    cs->y = -3;
+}
+
+static void RENAME(spatial_compose53i_init2)(DWTCompose *cs, uint8_t *buffer, int height, int stride)
+{
+    cs->b[0] = buffer + avpriv_mirror(-1-1, height-1)*stride;
+    cs->b[1] = buffer + avpriv_mirror(-1  , height-1)*stride;
+    cs->y = -1;
+}
+
+static void RENAME(spatial_compose_dd97i_init)(DWTCompose *cs, uint8_t *buffer, int height, int stride)
+{
+    cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
+    cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
+    cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
+    cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
+    cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
+    cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
+    cs->y = -5;
+}
+
+static void RENAME(spatial_compose_dd137i_init)(DWTCompose *cs, uint8_t *buffer, int height, int stride)
+{
+    cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
+    cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
+    cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
+    cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
+    cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
+    cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
+    cs->b[6] = buffer + av_clip(-5+5, 0, height-2)*stride;
+    cs->b[7] = buffer + av_clip(-5+6, 1, height-1)*stride;
+    cs->y = -5;
+}
+
+static int RENAME(ff_spatial_idwt_init2)(DWTContext *d, uint8_t *buffer, int width, int height,
+                                         int stride, enum dwt_type type, int decomposition_count,
+                                         uint8_t *temp)
+{
+    int level;
+
+    d->buffer = buffer;
+    d->width  = width;
+    d->height = height;
+    d->stride = stride;
+    d->decomposition_count = decomposition_count;
+    d->temp = (uint8_t *)(((TYPE *)temp) + 8);
+
+    for(level=decomposition_count-1; level>=0; level--){
+        int hl = height >> level;
+        int stride_l = stride << level;
+
+        switch(type){
+            case DWT_DIRAC_DD9_7:
+                RENAME(spatial_compose_dd97i_init)(d->cs+level, buffer, hl, stride_l);
+                break;
+            case DWT_DIRAC_LEGALL5_3:
+                RENAME(spatial_compose53i_init2)(d->cs+level, buffer, hl, stride_l);
+                break;
+            case DWT_DIRAC_DD13_7:
+                RENAME(spatial_compose_dd137i_init)(d->cs+level, buffer, hl, stride_l);
+                break;
+            case DWT_DIRAC_HAAR0:
+            case DWT_DIRAC_HAAR1:
+                d->cs[level].y = 1;
+                break;
+            case DWT_DIRAC_DAUB9_7:
+                RENAME(spatial_compose97i_init2)(d->cs+level, buffer, hl, stride_l);
+                break;
+            default:
+                d->cs[level].y = 0;
+                break;
+        }
+    }
+
+    switch (type) {
+        case DWT_DIRAC_DD9_7:
+            d->spatial_compose = RENAME(spatial_compose_dd97i_dy);
+            d->vertical_compose_l0 = (void*)RENAME(vertical_compose53iL0);
+            d->vertical_compose_h0 = (void*)RENAME(vertical_compose_dd97iH0);
+            d->horizontal_compose = RENAME(horizontal_compose_dd97i);
+            d->support = 7;
+            break;
+        case DWT_DIRAC_LEGALL5_3:
+            d->spatial_compose = RENAME(spatial_compose_dirac53i_dy);
+            d->vertical_compose_l0 = (void*)RENAME(vertical_compose53iL0);
+            d->vertical_compose_h0 = (void*)RENAME(vertical_compose_dirac53iH0);
+            d->horizontal_compose = RENAME(horizontal_compose_dirac53i);
+            d->support = 3;
+            break;
+        case DWT_DIRAC_DD13_7:
+            d->spatial_compose = RENAME(spatial_compose_dd137i_dy);
+            d->vertical_compose_l0 = (void*)RENAME(vertical_compose_dd137iL0);
+            d->vertical_compose_h0 = (void*)RENAME(vertical_compose_dd97iH0);
+            d->horizontal_compose = RENAME(horizontal_compose_dd137i);
+            d->support = 7;
+            break;
+        case DWT_DIRAC_HAAR0:
+        case DWT_DIRAC_HAAR1:
+            d->spatial_compose = RENAME(spatial_compose_haari_dy);
+            d->vertical_compose = (void*)RENAME(vertical_compose_haar);
+            if (type == DWT_DIRAC_HAAR0)
+                d->horizontal_compose = RENAME(horizontal_compose_haar0i);
+            else
+                d->horizontal_compose = RENAME(horizontal_compose_haar1i);
+            d->support = 1;
+            break;
+        case DWT_DIRAC_FIDELITY:
+            d->spatial_compose = RENAME(spatial_compose_fidelity);
+            d->vertical_compose_l0 = (void*)RENAME(vertical_compose_fidelityiL0);
+            d->vertical_compose_h0 = (void*)RENAME(vertical_compose_fidelityiH0);
+            d->horizontal_compose = RENAME(horizontal_compose_fidelityi);
+            d->support = 0; // not really used
+            break;
+        case DWT_DIRAC_DAUB9_7:
+            d->spatial_compose = RENAME(spatial_compose_daub97i_dy);
+            d->vertical_compose_l0 = (void*)RENAME(vertical_compose_daub97iL0);
+            d->vertical_compose_h0 = (void*)RENAME(vertical_compose_daub97iH0);
+            d->vertical_compose_l1 = (void*)RENAME(vertical_compose_daub97iL1);
+            d->vertical_compose_h1 = (void*)RENAME(vertical_compose_daub97iH1);
+            d->horizontal_compose = RENAME(horizontal_compose_daub97i);
+            d->support = 5;
+            break;
+        default:
+            return AVERROR_INVALIDDATA;
+    }
+
+    return 0;
+}
+
+#undef RENAME
+#undef TYPE
index 4bb9ca3..1d4601e 100644 (file)
@@ -1687,16 +1687,16 @@ static int dirac_decode_frame_internal(DiracContext *s)
             memset(p->idwt_buf, 0, p->idwt_stride * p->idwt_height);
             decode_component(s, comp); /* [DIRAC_STD] 13.4.1 core_transform_data() */
         }
-        ret = ff_spatial_idwt_init2(&d, (int16_t*)p->idwt_buf, p->idwt_width, p->idwt_height, p->idwt_stride >> 1,
-                                    s->wavelet_idx+2, s->wavelet_depth, (int16_t*)p->idwt_tmp);
+        ret = ff_spatial_idwt_init2(&d, p->idwt_buf, p->idwt_width, p->idwt_height, p->idwt_stride,
+                                    s->wavelet_idx+2, s->wavelet_depth, p->idwt_tmp, s->bit_depth);
         if (ret < 0)
             return ret;
 
         if (!s->num_refs) { /* intra */
             for (y = 0; y < p->height; y += 16) {
                 ff_spatial_idwt_slice2(&d, y+16); /* decode */
-                s->diracdsp.put_signed_rect_clamped(frame + y*p->stride, p->stride,
-                                                   (int16_t*)(p->idwt_buf) + y*(p->idwt_stride >> 1), (p->idwt_stride >> 1), p->width, 16);
+                s->diracdsp.put_signed_rect_clamped[s->pshift](frame + y*p->stride, p->stride,
+                                                               p->idwt_buf + y*p->idwt_stride, p->idwt_stride, p->width, 16);
             }
         } else { /* inter */
             int rowheight = p->ybsep*p->stride;
index 6b02779..e82b587 100644 (file)
@@ -135,9 +135,10 @@ ADD_OBMC(8)
 ADD_OBMC(16)
 ADD_OBMC(32)
 
-static void put_signed_rect_clamped_c(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height)
+static void put_signed_rect_clamped_8bit_c(uint8_t *dst, int dst_stride, const uint8_t *_src, int src_stride, int width, int height)
 {
     int x, y;
+    int16_t *src = (int16_t *)_src;
     for (y = 0; y < height; y++) {
         for (x = 0; x < width; x+=4) {
             dst[x  ] = av_clip_uint8(src[x  ] + 128);
@@ -146,7 +147,24 @@ static void put_signed_rect_clamped_c(uint8_t *dst, int dst_stride, const int16_
             dst[x+3] = av_clip_uint8(src[x+3] + 128);
         }
         dst += dst_stride;
-        src += src_stride;
+        src += src_stride >> 1;
+    }
+}
+
+static void put_signed_rect_clamped_10bit_c(uint8_t *_dst, int dst_stride, const uint8_t *_src, int src_stride, int width, int height)
+{
+    int x, y;
+    uint16_t *dst = (uint16_t *)_dst;
+    int32_t *src = (int32_t *)_src;
+    for (y = 0; y < height; y++) {
+        for (x = 0; x < width; x+=4) {
+            dst[x  ] = av_clip(src[x  ] + 512, 0, (1 << 10) - 1);
+            dst[x+1] = av_clip(src[x+1] + 512, 0, (1 << 10) - 1);
+            dst[x+2] = av_clip(src[x+2] + 512, 0, (1 << 10) - 1);
+            dst[x+3] = av_clip(src[x+3] + 512, 0, (1 << 10) - 1);
+        }
+        dst += dst_stride >> 1;
+        src += src_stride >> 2;
     }
 }
 
@@ -177,7 +195,8 @@ av_cold void ff_diracdsp_init(DiracDSPContext *c)
 {
     c->dirac_hpel_filter = dirac_hpel_filter;
     c->add_rect_clamped = add_rect_clamped_c;
-    c->put_signed_rect_clamped = put_signed_rect_clamped_c;
+    c->put_signed_rect_clamped[0] = put_signed_rect_clamped_8bit_c;
+    c->put_signed_rect_clamped[1] = put_signed_rect_clamped_10bit_c;
 
     c->add_dirac_obmc[0] = add_obmc8_c;
     c->add_dirac_obmc[1] = add_obmc16_c;
index 613ca5b..439bda2 100644 (file)
@@ -41,8 +41,8 @@ typedef struct {
     void (*put_dirac_pixels_tab[3][4])(uint8_t *dst, const uint8_t *src[5], int stride, int h);
     void (*avg_dirac_pixels_tab[3][4])(uint8_t *dst, const uint8_t *src[5], int stride, int h);
 
-    void (*put_signed_rect_clamped)(uint8_t *dst/*align 16*/, int dst_stride, const int16_t *src/*align 16*/, int src_stride, int width, int height/*mod 2*/);
-    void (*put_rect_clamped)(uint8_t *dst/*align 16*/, int dst_stride, const int16_t *src/*align 16*/, int src_stride, int width, int height/*mod 2*/);
+    void (*put_signed_rect_clamped[3])(uint8_t *dst/*align 16*/, int dst_stride, const uint8_t *src/*align 16*/, int src_stride, int width, int height/*mod 2*/);
+    void (*put_rect_clamped)(uint8_t *dst/*align 16*/, int dst_stride, const uint8_t *src/*align 16*/, int src_stride, int width, int height/*mod 2*/);
     void (*add_rect_clamped)(uint8_t *dst/*align 16*/, const uint16_t *src/*align 16*/, int stride, const int16_t *idwt/*align 16*/, int idwt_stride, int width, int height/*mod 2*/);
     void (*add_dirac_obmc[3])(uint16_t *dst, const uint8_t *src, int stride, const uint8_t *obmc_weight, int yblen);
 
index 3c51ea6..c9c7093 100644 (file)
 #include "dirac_dwt.h"
 
 #define COMPOSE_VERTICAL(ext, align) \
-void ff_vertical_compose53iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width); \
-void ff_vertical_compose_dirac53iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width); \
-void ff_vertical_compose_dd137iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, int width); \
-void ff_vertical_compose_dd97iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, int width); \
-void ff_vertical_compose_haar##ext(IDWTELEM *b0, IDWTELEM *b1, int width); \
-void ff_horizontal_compose_haar0i##ext(IDWTELEM *b, IDWTELEM *tmp, int w);\
-void ff_horizontal_compose_haar1i##ext(IDWTELEM *b, IDWTELEM *tmp, int w);\
-\
-static void vertical_compose53iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width) \
+void ff_vertical_compose53iL0##ext(int16_t *b0, int16_t *b1, int16_t *b2, int width); \
+void ff_vertical_compose_dirac53iH0##ext(int16_t *b0, int16_t *b1, int16_t *b2, int width); \
+void ff_vertical_compose_dd137iL0##ext(int16_t *b0, int16_t *b1, int16_t *b2, int16_t *b3, int16_t *b4, int width); \
+void ff_vertical_compose_dd97iH0##ext(int16_t *b0, int16_t *b1, int16_t *b2, int16_t *b3, int16_t *b4, int width); \
+void ff_vertical_compose_haar##ext(int16_t *b0, int16_t *b1, int width); \
+void ff_horizontal_compose_haar0i##ext(int16_t *b, int16_t *tmp, int w);\
+void ff_horizontal_compose_haar1i##ext(int16_t *b, int16_t *tmp, int w);\
+\
+static void vertical_compose53iL0##ext(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, int width) \
 { \
     int i, width_align = width&~(align-1); \
+    int16_t *b0 = (int16_t *)_b0; \
+    int16_t *b1 = (int16_t *)_b1; \
+    int16_t *b2 = (int16_t *)_b2; \
 \
     for(i=width_align; i<width; i++) \
         b1[i] = COMPOSE_53iL0(b0[i], b1[i], b2[i]); \
@@ -43,9 +46,12 @@ static void vertical_compose53iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
     ff_vertical_compose53iL0##ext(b0, b1, b2, width_align); \
 } \
 \
-static void vertical_compose_dirac53iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width) \
+static void vertical_compose_dirac53iH0##ext(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, int width) \
 { \
     int i, width_align = width&~(align-1); \
+    int16_t *b0 = (int16_t *)_b0; \
+    int16_t *b1 = (int16_t *)_b1; \
+    int16_t *b2 = (int16_t *)_b2; \
 \
     for(i=width_align; i<width; i++) \
         b1[i] = COMPOSE_DIRAC53iH0(b0[i], b1[i], b2[i]); \
@@ -53,10 +59,15 @@ static void vertical_compose_dirac53iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELE
     ff_vertical_compose_dirac53iH0##ext(b0, b1, b2, width_align); \
 } \
 \
-static void vertical_compose_dd137iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, \
-                                           IDWTELEM *b3, IDWTELEM *b4, int width) \
+static void vertical_compose_dd137iL0##ext(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, \
+                                           uint8_t *_b3, uint8_t *_b4, int width) \
 { \
     int i, width_align = width&~(align-1); \
+    int16_t *b0 = (int16_t *)_b0; \
+    int16_t *b1 = (int16_t *)_b1; \
+    int16_t *b2 = (int16_t *)_b2; \
+    int16_t *b3 = (int16_t *)_b3; \
+    int16_t *b4 = (int16_t *)_b4; \
 \
     for(i=width_align; i<width; i++) \
         b2[i] = COMPOSE_DD137iL0(b0[i], b1[i], b2[i], b3[i], b4[i]); \
@@ -64,19 +75,26 @@ static void vertical_compose_dd137iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM
     ff_vertical_compose_dd137iL0##ext(b0, b1, b2, b3, b4, width_align); \
 } \
 \
-static void vertical_compose_dd97iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, \
-                                          IDWTELEM *b3, IDWTELEM *b4, int width) \
+static void vertical_compose_dd97iH0##ext(uint8_t *_b0, uint8_t *_b1, uint8_t *_b2, \
+                                          uint8_t *_b3, uint8_t *_b4, int width) \
 { \
     int i, width_align = width&~(align-1); \
+    int16_t *b0 = (int16_t *)_b0; \
+    int16_t *b1 = (int16_t *)_b1; \
+    int16_t *b2 = (int16_t *)_b2; \
+    int16_t *b3 = (int16_t *)_b3; \
+    int16_t *b4 = (int16_t *)_b4; \
 \
     for(i=width_align; i<width; i++) \
         b2[i] = COMPOSE_DD97iH0(b0[i], b1[i], b2[i], b3[i], b4[i]); \
 \
     ff_vertical_compose_dd97iH0##ext(b0, b1, b2, b3, b4, width_align); \
 } \
-static void vertical_compose_haar##ext(IDWTELEM *b0, IDWTELEM *b1, int width) \
+static void vertical_compose_haar##ext(uint8_t *_b0, uint8_t *_b1, int width) \
 { \
     int i, width_align = width&~(align-1); \
+    int16_t *b0 = (int16_t *)_b0; \
+    int16_t *b1 = (int16_t *)_b1; \
 \
     for(i=width_align; i<width; i++) { \
         b0[i] = COMPOSE_HAARiL0(b0[i], b1[i]); \
@@ -85,10 +103,13 @@ static void vertical_compose_haar##ext(IDWTELEM *b0, IDWTELEM *b1, int width) \
 \
     ff_vertical_compose_haar##ext(b0, b1, width_align); \
 } \
-static void horizontal_compose_haar0i##ext(IDWTELEM *b, IDWTELEM *tmp, int w)\
+static void horizontal_compose_haar0i##ext(uint8_t *_b, uint8_t *_tmp, int w)\
 {\
     int w2= w>>1;\
     int x= w2 - (w2&(align-1));\
+    int16_t *b = (int16_t *)_b; \
+    int16_t *tmp = (int16_t *)_tmp; \
+\
     ff_horizontal_compose_haar0i##ext(b, tmp, w);\
 \
     for (; x < w2; x++) {\
@@ -96,10 +117,13 @@ static void horizontal_compose_haar0i##ext(IDWTELEM *b, IDWTELEM *tmp, int w)\
         b[2*x+1] = COMPOSE_HAARiH0(b[x+w2], tmp[x]);\
     }\
 }\
-static void horizontal_compose_haar1i##ext(IDWTELEM *b, IDWTELEM *tmp, int w)\
+static void horizontal_compose_haar1i##ext(uint8_t *_b, uint8_t *_tmp, int w)\
 {\
     int w2= w>>1;\
     int x= w2 - (w2&(align-1));\
+    int16_t *b = (int16_t *)_b; \
+    int16_t *tmp = (int16_t *)_tmp; \
+\
     ff_horizontal_compose_haar1i##ext(b, tmp, w);\
 \
     for (; x < w2; x++) {\
@@ -116,12 +140,15 @@ COMPOSE_VERTICAL(_mmx, 4)
 COMPOSE_VERTICAL(_sse2, 8)
 
 
-void ff_horizontal_compose_dd97i_ssse3(IDWTELEM *b, IDWTELEM *tmp, int w);
+void ff_horizontal_compose_dd97i_ssse3(int16_t *_b, int16_t *_tmp, int w);
 
-static void horizontal_compose_dd97i_ssse3(IDWTELEM *b, IDWTELEM *tmp, int w)
+static void horizontal_compose_dd97i_ssse3(uint8_t *_b, uint8_t *_tmp, int w)
 {
     int w2= w>>1;
     int x= w2 - (w2&7);
+    int16_t *b = (int16_t *)_b;
+    int16_t *tmp = (int16_t *)_tmp;
+
     ff_horizontal_compose_dd97i_ssse3(b, tmp, w);
 
     for (; x < w2; x++) {
index 11df5e3..b2e949c 100644 (file)
@@ -130,7 +130,7 @@ void ff_diracdsp_init_mmx(DiracDSPContext* c)
     c->add_dirac_obmc[2] = ff_add_dirac_obmc32_mmx;
     c->dirac_hpel_filter = dirac_hpel_filter_mmx;
     c->add_rect_clamped = ff_add_rect_clamped_mmx;
-    c->put_signed_rect_clamped = ff_put_signed_rect_clamped_mmx;
+    c->put_signed_rect_clamped[0] = (void *)ff_put_signed_rect_clamped_mmx;
 #endif
     PIXFUNC(put, 0, mmx);
     PIXFUNC(avg, 0, mmx);
@@ -143,7 +143,7 @@ void ff_diracdsp_init_mmx(DiracDSPContext* c)
     if (EXTERNAL_SSE2(mm_flags)) {
         c->dirac_hpel_filter = dirac_hpel_filter_sse2;
         c->add_rect_clamped = ff_add_rect_clamped_sse2;
-        c->put_signed_rect_clamped = ff_put_signed_rect_clamped_sse2;
+        c->put_signed_rect_clamped[0] = (void *)ff_put_signed_rect_clamped_sse2;
 
         c->add_dirac_obmc[1] = ff_add_dirac_obmc16_sse2;
         c->add_dirac_obmc[2] = ff_add_dirac_obmc32_sse2;
index 40fe2c8..a042413 100644 (file)
@@ -150,7 +150,7 @@ cglobal put_signed_rect_clamped_%1, 5,9,3, dst, dst_stride, src, src_stride, w,
 %endif
 
 .loopy:
-    lea     src2q, [srcq+src_strideq*2]
+    lea     src2q, [srcq+src_strideq]
     lea     dst2q, [dstq+dst_strideq]
 .loopx:
     sub      wd, mmsize
@@ -164,7 +164,7 @@ cglobal put_signed_rect_clamped_%1, 5,9,3, dst, dst_stride, src, src_stride, w,
     mova    [dst2q+wq], m2
     jg      .loopx
 
-    lea   srcq, [srcq+src_strideq*4]
+    lea   srcq, [srcq+src_strideq*2]
     lea   dstq, [dstq+dst_strideq*2]
     sub     hd, 2
     mov     wd, wspill