lavc: Edge emulation with dst/src linesize

Allow supporting files for which the image stride is smaller than
the maximum block size + number of subpel mc taps, e.g. a 64x64 VP9
file or a 16x16 VP8 file with -fflags +emu_edge.

libavcodec/cavs.c

@@ -408,7 +408,8 @@ static inline void mc_dir_part(AVSContext *h, AVFrame *pic, int chroma_height,
         full_mx + 16 /* FIXME */ > pic_width + extra_width ||
         full_my + 16 /* FIXME */ > pic_height + extra_height) {
         h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
-                                 src_y - 2 - 2 * h->l_stride, h->l_stride,
+                                 src_y - 2 - 2 * h->l_stride,
+                                 h->l_stride, h->l_stride,
                                  16 + 5, 16 + 5 /* FIXME */,
                                  full_mx - 2, full_my - 2,
                                  pic_width, pic_height);
@@ -421,7 +422,7 @@ static inline void mc_dir_part(AVSContext *h, AVFrame *pic, int chroma_height,
 
     if (emu) {
         h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cb,
-                                 h->c_stride,
+                                 h->c_stride, h->c_stride,
                                  9, 9 /* FIXME */,
                                  mx >> 3, my >> 3,
                                  pic_width >> 1, pic_height >> 1);
@@ -431,7 +432,7 @@ static inline void mc_dir_part(AVSContext *h, AVFrame *pic, int chroma_height,
 
     if (emu) {
         h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cr,
-                                 h->c_stride,
+                                 h->c_stride, h->c_stride,
                                  9, 9 /* FIXME */,
                                  mx >> 3, my >> 3,
                                  pic_width >> 1, pic_height >> 1);

libavcodec/h264.c

@@ -899,7 +899,7 @@ static av_always_inline void mc_dir_part(H264Context *h, Picture *pic,
         full_my + 16 /*FIXME*/ > pic_height + extra_height) {
         h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
                                  src_y - (2 << pixel_shift) - 2 * h->mb_linesize,
-                                 h->mb_linesize,
+                                 h->mb_linesize, h->mb_linesize,
                                  16 + 5, 16 + 5 /*FIXME*/, full_mx - 2,
                                  full_my - 2, pic_width, pic_height);
         src_y = h->edge_emu_buffer + (2 << pixel_shift) + 2 * h->mb_linesize;
@@ -918,7 +918,7 @@ static av_always_inline void mc_dir_part(H264Context *h, Picture *pic,
         if (emu) {
             h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
                                      src_cb - (2 << pixel_shift) - 2 * h->mb_linesize,
-                                     h->mb_linesize,
+                                     h->mb_linesize, h->mb_linesize,
                                      16 + 5, 16 + 5 /*FIXME*/,
                                      full_mx - 2, full_my - 2,
                                      pic_width, pic_height);
@@ -932,7 +932,7 @@ static av_always_inline void mc_dir_part(H264Context *h, Picture *pic,
         if (emu) {
             h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
                                      src_cr - (2 << pixel_shift) - 2 * h->mb_linesize,
-                                     h->mb_linesize,
+                                     h->mb_linesize, h->mb_linesize,
                                      16 + 5, 16 + 5 /*FIXME*/,
                                      full_mx - 2, full_my - 2,
                                      pic_width, pic_height);
@@ -957,7 +957,8 @@ static av_always_inline void mc_dir_part(H264Context *h, Picture *pic,
              (my >> ysh) * h->mb_uvlinesize;
 
     if (emu) {
-        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cb, h->mb_uvlinesize,
+        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cb,
+                                 h->mb_uvlinesize, h->mb_uvlinesize,
                                  9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh),
                                  pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */));
         src_cb = h->edge_emu_buffer;
@@ -967,7 +968,8 @@ static av_always_inline void mc_dir_part(H264Context *h, Picture *pic,
               mx & 7, (my << (chroma_idc == 2 /* yuv422 */)) & 7);
 
     if (emu) {
-        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cr, h->mb_uvlinesize,
+        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cr,
+                                 h->mb_uvlinesize, h->mb_uvlinesize,
                                  9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh),
                                  pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */));
         src_cr = h->edge_emu_buffer;

libavcodec/hevc.c

@@ -1481,7 +1481,8 @@ static void luma_mc(HEVCContext *s, int16_t *dst, ptrdiff_t dststride,
         y_off >= pic_height - block_h - ff_hevc_qpel_extra_after[my]) {
         int offset = extra_top * srcstride + (extra_left << s->sps->pixel_shift);
 
-        s->vdsp.emulated_edge_mc(lc->edge_emu_buffer, src - offset, srcstride,
+        s->vdsp.emulated_edge_mc(lc->edge_emu_buffer, src - offset,
+                                 srcstride, srcstride,
                                  block_w + ff_hevc_qpel_extra[mx],
                                  block_h + ff_hevc_qpel_extra[my],
                                  x_off - extra_left, y_off - extra_top,
@@ -1532,7 +1533,8 @@ static void chroma_mc(HEVCContext *s, int16_t *dst1, int16_t *dst2,
         int offset1 = EPEL_EXTRA_BEFORE * (src1stride + (1 << s->sps->pixel_shift));
         int offset2 = EPEL_EXTRA_BEFORE * (src2stride + (1 << s->sps->pixel_shift));
 
-        s->vdsp.emulated_edge_mc(lc->edge_emu_buffer, src1 - offset1, src1stride,
+        s->vdsp.emulated_edge_mc(lc->edge_emu_buffer, src1 - offset1,
+                                 src1stride, src1stride,
                                  block_w + EPEL_EXTRA, block_h + EPEL_EXTRA,
                                  x_off - EPEL_EXTRA_BEFORE,
                                  y_off - EPEL_EXTRA_BEFORE,
@@ -1542,7 +1544,8 @@ static void chroma_mc(HEVCContext *s, int16_t *dst1, int16_t *dst2,
         s->hevcdsp.put_hevc_epel[!!my][!!mx](dst1, dststride, src1, src1stride,
                                              block_w, block_h, mx, my, lc->mc_buffer);
 
-        s->vdsp.emulated_edge_mc(lc->edge_emu_buffer, src2 - offset2, src2stride,
+        s->vdsp.emulated_edge_mc(lc->edge_emu_buffer, src2 - offset2,
+                                 src2stride, src2stride,
                                  block_w + EPEL_EXTRA, block_h + EPEL_EXTRA,
                                  x_off - EPEL_EXTRA_BEFORE,
                                  y_off - EPEL_EXTRA_BEFORE,

libavcodec/mpegvideo_enc.c

@@ -1699,15 +1699,19 @@ static av_always_inline void encode_mb_internal(MpegEncContext *s,
 
     if (mb_x * 16 + 16 > s->width || mb_y * 16 + 16 > s->height) {
         uint8_t *ebuf = s->edge_emu_buffer + 32;
-        s->vdsp.emulated_edge_mc(ebuf, ptr_y, wrap_y, 16, 16, mb_x * 16,
-                                 mb_y * 16, s->width, s->height);
+        s->vdsp.emulated_edge_mc(ebuf, ptr_y,
+                                 wrap_y, wrap_y,
+                                 16, 16, mb_x * 16, mb_y * 16,
+                                 s->width, s->height);
         ptr_y = ebuf;
-        s->vdsp.emulated_edge_mc(ebuf + 18 * wrap_y, ptr_cb, wrap_c, 8,
-                                 mb_block_height, mb_x * 8, mb_y * 8,
+        s->vdsp.emulated_edge_mc(ebuf + 18 * wrap_y, ptr_cb,
+                                 wrap_c, wrap_c,
+                                 8, mb_block_height, mb_x * 8, mb_y * 8,
                                  s->width >> 1, s->height >> 1);
         ptr_cb = ebuf + 18 * wrap_y;
-        s->vdsp.emulated_edge_mc(ebuf + 18 * wrap_y + 8, ptr_cr, wrap_c, 8,
-                                 mb_block_height, mb_x * 8, mb_y * 8,
+        s->vdsp.emulated_edge_mc(ebuf + 18 * wrap_y + 8, ptr_cr,
+                                 wrap_c, wrap_c,
+                                 8, mb_block_height, mb_x * 8, mb_y * 8,
                                  s->width >> 1, s->height >> 1);
         ptr_cr = ebuf + 18 * wrap_y + 8;
     }

libavcodec/mpegvideo_motion.c

@@ -63,7 +63,7 @@ static void gmc1_motion(MpegEncContext *s,
         if ((unsigned)src_x >= FFMAX(s->h_edge_pos - 17, 0) ||
             (unsigned)src_y >= FFMAX(s->v_edge_pos - 17, 0)) {
             s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
-                                     linesize,
+                                     linesize, linesize,
                                      17, 17,
                                      src_x, src_y,
                                      s->h_edge_pos, s->v_edge_pos);
@@ -109,7 +109,7 @@ static void gmc1_motion(MpegEncContext *s,
         if ((unsigned)src_x >= FFMAX((s->h_edge_pos >> 1) - 9, 0) ||
             (unsigned)src_y >= FFMAX((s->v_edge_pos >> 1) - 9, 0)) {
             s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
-                                     uvlinesize,
+                                     uvlinesize, uvlinesize,
                                      9, 9,
                                      src_x, src_y,
                                      s->h_edge_pos >> 1, s->v_edge_pos >> 1);
@@ -123,7 +123,7 @@ static void gmc1_motion(MpegEncContext *s,
     ptr = ref_picture[2] + offset;
     if (emu) {
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
-                                 uvlinesize,
+                                 uvlinesize, uvlinesize,
                                  9, 9,
                                  src_x, src_y,
                                  s->h_edge_pos >> 1, s->v_edge_pos >> 1);
@@ -216,7 +216,7 @@ static inline int hpel_motion(MpegEncContext *s,
         if ((unsigned)src_x > FFMAX(s->h_edge_pos - (motion_x & 1) - 8, 0) ||
             (unsigned)src_y > FFMAX(s->v_edge_pos - (motion_y & 1) - 8, 0)) {
             s->vdsp.emulated_edge_mc(s->edge_emu_buffer, src,
-                                     s->linesize,
+                                     s->linesize, s->linesize,
                                      9, 9,
                                      src_x, src_y, s->h_edge_pos,
                                      s->v_edge_pos);
@@ -321,7 +321,7 @@ void mpeg_motion_internal(MpegEncContext *s,
             return;
         }
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr_y,
-                                 s->linesize,
+                                 s->linesize, s->linesize,
                                  17, 17 + field_based,
                                  src_x, src_y << field_based,
                                  s->h_edge_pos, s->v_edge_pos);
@@ -329,12 +329,12 @@ void mpeg_motion_internal(MpegEncContext *s,
         if (!CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
             uint8_t *uvbuf = s->edge_emu_buffer + 18 * s->linesize;
             s->vdsp.emulated_edge_mc(uvbuf, ptr_cb,
-                                     s->uvlinesize,
+                                     s->uvlinesize, s->uvlinesize,
                                      9, 9 + field_based,
                                      uvsrc_x, uvsrc_y << field_based,
                                      s->h_edge_pos >> 1, s->v_edge_pos >> 1);
             s->vdsp.emulated_edge_mc(uvbuf + 16, ptr_cr,
-                                     s->uvlinesize,
+                                     s->uvlinesize, s->uvlinesize,
                                      9, 9 + field_based,
                                      uvsrc_x, uvsrc_y << field_based,
                                      s->h_edge_pos >> 1, s->v_edge_pos >> 1);
@@ -542,7 +542,7 @@ static inline void qpel_motion(MpegEncContext *s,
     if ((unsigned)src_x > FFMAX(s->h_edge_pos - (motion_x & 3) - 16, 0) ||
         (unsigned)src_y > FFMAX(v_edge_pos - (motion_y & 3) - h, 0)) {
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr_y,
-                                 s->linesize,
+                                 s->linesize, s->linesize,
                                  17, 17 + field_based,
                                  src_x, src_y << field_based,
                                  s->h_edge_pos, s->v_edge_pos);
@@ -550,12 +550,12 @@ static inline void qpel_motion(MpegEncContext *s,
         if (!CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
             uint8_t *uvbuf = s->edge_emu_buffer + 18 * s->linesize;
             s->vdsp.emulated_edge_mc(uvbuf, ptr_cb,
-                                     s->uvlinesize,
+                                     s->uvlinesize, s->uvlinesize,
                                      9, 9 + field_based,
                                      uvsrc_x, uvsrc_y << field_based,
                                      s->h_edge_pos >> 1, s->v_edge_pos >> 1);
             s->vdsp.emulated_edge_mc(uvbuf + 16, ptr_cr,
-                                     s->uvlinesize,
+                                     s->uvlinesize, s->uvlinesize,
                                      9, 9 + field_based,
                                      uvsrc_x, uvsrc_y << field_based,
                                      s->h_edge_pos >> 1, s->v_edge_pos >> 1);
@@ -625,8 +625,8 @@ static void chroma_4mv_motion(MpegEncContext *s,
     if (s->flags & CODEC_FLAG_EMU_EDGE) {
         if ((unsigned)src_x > FFMAX((s->h_edge_pos >> 1) - (dxy & 1) - 8, 0) ||
             (unsigned)src_y > FFMAX((s->v_edge_pos >> 1) - (dxy >> 1) - 8, 0)) {
-            s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
-                                     ptr, s->uvlinesize,
+            s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
+                                     s->uvlinesize, s->uvlinesize,
                                      9, 9, src_x, src_y,
                                      s->h_edge_pos >> 1, s->v_edge_pos >> 1);
             ptr = s->edge_emu_buffer;
@@ -637,8 +637,8 @@ static void chroma_4mv_motion(MpegEncContext *s,
 
     ptr = ref_picture[2] + offset;
     if (emu) {
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
-                                 ptr, s->uvlinesize,
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
+                                 s->uvlinesize, s->uvlinesize,
                                  9, 9, src_x, src_y,
                                  s->h_edge_pos >> 1, s->v_edge_pos >> 1);
         ptr = s->edge_emu_buffer;
@@ -786,7 +786,7 @@ static inline void apply_8x8(MpegEncContext *s,
                 if ((unsigned)src_x > FFMAX(s->h_edge_pos - (motion_x & 3) - 8, 0) ||
                     (unsigned)src_y > FFMAX(s->v_edge_pos - (motion_y & 3) - 8, 0)) {
                     s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
-                                             s->linesize,
+                                             s->linesize, s->linesize,
                                              9, 9,
                                              src_x, src_y,
                                              s->h_edge_pos,

libavcodec/rv34.c

@@ -723,12 +723,18 @@ static inline void rv34_mc(RV34DecContext *r, const int block_type,
         uint8_t *uvbuf = s->edge_emu_buffer + 22 * s->linesize;
 
         srcY -= 2 + 2*s->linesize;
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, (width<<3)+6, (height<<3)+6,
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
+                                 s->linesize, s->linesize,
+                                 (width << 3) + 6, (height << 3) + 6,
                             src_x - 2, src_y - 2, s->h_edge_pos, s->v_edge_pos);
         srcY = s->edge_emu_buffer + 2 + 2*s->linesize;
-        s->vdsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, (width<<2)+1, (height<<2)+1,
+        s->vdsp.emulated_edge_mc(uvbuf, srcU,
+                                 s->uvlinesize,s->uvlinesize,
+                                 (width << 2) + 1, (height << 2) + 1,
                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
-        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, (width<<2)+1, (height<<2)+1,
+        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV,
+                                 s->uvlinesize, s->uvlinesize,
+                                 (width << 2) + 1, (height << 2) + 1,
                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
         srcU = uvbuf;
         srcV = uvbuf + 16;

libavcodec/svq3.c

@@ -312,7 +312,8 @@ static inline void svq3_mc_dir_part(SVQ3Context *s,
     src  = pic->f.data[0] + mx + my * h->linesize;
 
     if (emu) {
-        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src, h->linesize,
+        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src,
+                                 h->linesize, h->linesize,
                                  width + 1, height + 1,
                                  mx, my, s->h_edge_pos, s->v_edge_pos);
         src = h->edge_emu_buffer;
@@ -338,7 +339,8 @@ static inline void svq3_mc_dir_part(SVQ3Context *s,
             src  = pic->f.data[i] + mx + my * h->uvlinesize;
 
             if (emu) {
-                h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src, h->uvlinesize,
+                h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src,
+                                         h->uvlinesize, h->uvlinesize,
                                          width + 1, height + 1,
                                          mx, my, (s->h_edge_pos >> 1),
                                          s->v_edge_pos >> 1);

libavcodec/vc1dec.c

@@ -452,14 +452,19 @@ static void vc1_mc_1mv(VC1Context *v, int dir)
         uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;
 
         srcY -= s->mspel * (1 + s->linesize);
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
+                                 s->linesize, s->linesize,
                                  17 + s->mspel * 2, 17 + s->mspel * 2,
                                  src_x - s->mspel, src_y - s->mspel,
                                  s->h_edge_pos, v_edge_pos);
         srcY = s->edge_emu_buffer;
-        s->vdsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8 + 1, 8 + 1,
+        s->vdsp.emulated_edge_mc(uvbuf, srcU,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
                                  uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
-        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8 + 1, 8 + 1,
+        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
                                  uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
         srcU = uvbuf;
         srcV = uvbuf + 16;
@@ -693,7 +698,8 @@ static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg)
         || (unsigned)(src_y - (s->mspel << fieldmv)) > v_edge_pos - (my & 3) - ((8 + s->mspel * 2) << fieldmv)) {
         srcY -= s->mspel * (1 + (s->linesize << fieldmv));
         /* check emulate edge stride and offset */
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
+                                 s->linesize, s->linesize,
                                  9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv,
                                  src_x - s->mspel, src_y - (s->mspel << fieldmv),
                                  s->h_edge_pos, v_edge_pos);
@@ -908,10 +914,12 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir)
         || s->h_edge_pos < 18 || v_edge_pos < 18
         || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
         || (unsigned)uvsrc_y > (v_edge_pos    >> 1) - 9) {
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer     , srcU, s->uvlinesize,
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU,
+                                 s->uvlinesize, s->uvlinesize,
                                  8 + 1, 8 + 1, uvsrc_x, uvsrc_y,
                                  s->h_edge_pos >> 1, v_edge_pos >> 1);
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV,
+                                 s->uvlinesize, s->uvlinesize,
                                  8 + 1, 8 + 1, uvsrc_x, uvsrc_y,
                                  s->h_edge_pos >> 1, v_edge_pos >> 1);
         srcU = s->edge_emu_buffer;
@@ -1024,10 +1032,12 @@ static void vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg)
             || s->h_edge_pos < 10 || v_edge_pos < (5 << fieldmv)
             || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 5
             || (unsigned)uvsrc_y > v_edge_pos - (5 << fieldmv)) {
-            s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU, s->uvlinesize,
+            s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU,
+                                     s->uvlinesize, s->uvlinesize,
                                      5, (5 << fieldmv), uvsrc_x, uvsrc_y,
                                      s->h_edge_pos >> 1, v_edge_pos);
-            s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,
+            s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV,
+                                     s->uvlinesize, s->uvlinesize,
                                      5, (5 << fieldmv), uvsrc_x, uvsrc_y,
                                      s->h_edge_pos >> 1, v_edge_pos);
             srcU = s->edge_emu_buffer;
@@ -1955,14 +1965,19 @@ static void vc1_interp_mc(VC1Context *v)
         uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;
 
         srcY -= s->mspel * (1 + s->linesize);
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
+                                 s->linesize, s->linesize,
                                  17 + s->mspel * 2, 17 + s->mspel * 2,
                                  src_x - s->mspel, src_y - s->mspel,
                                  s->h_edge_pos, v_edge_pos);
         srcY = s->edge_emu_buffer;
-        s->vdsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8 + 1, 8 + 1,
+        s->vdsp.emulated_edge_mc(uvbuf, srcU,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
                                  uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
-        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8 + 1, 8 + 1,
+        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
                                  uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
         srcU = uvbuf;
         srcV = uvbuf + 16;

libavcodec/videodsp.h

@@ -36,8 +36,10 @@ typedef struct VideoDSPContext {
      *
      * @param buf destination buffer
      * @param src source buffer
-     * @param linesize number of bytes between 2 vertically adjacent samples
-     *                 in both the source and destination buffers
+     * @param buf_linesize number of bytes between 2 vertically adjacent
+     *                     samples in the destination buffer
+     * @param src_linesize number of bytes between 2 vertically adjacent
+     *                     samples in both the source buffer
      * @param block_w width of block
      * @param block_h height of block
      * @param src_x x coordinate of the top left sample of the block in the
@@ -48,7 +50,9 @@ typedef struct VideoDSPContext {
      * @param h height of the source buffer
      */
     void (*emulated_edge_mc)(uint8_t *buf, const uint8_t *src,
-                             ptrdiff_t linesize, int block_w, int block_h,
+                             ptrdiff_t buf_linesize,
+                             ptrdiff_t src_linesize,
+                             int block_w, int block_h,
                              int src_x, int src_y, int w, int h);
 
     /**

libavcodec/videodsp_template.c

@@ -22,7 +22,8 @@
 #include "bit_depth_template.c"
 
 static void FUNC(ff_emulated_edge_mc)(uint8_t *buf, const uint8_t *src,
-                                      ptrdiff_t linesize,
+                                      ptrdiff_t buf_linesize,
+                                      ptrdiff_t src_linesize,
                                       int block_w, int block_h,
                                       int src_x, int src_y, int w, int h)
 {
@@ -30,10 +31,10 @@ static void FUNC(ff_emulated_edge_mc)(uint8_t *buf, const uint8_t *src,
     int start_y, start_x, end_y, end_x;
 
     if (src_y >= h) {
-        src  += (h - 1 - src_y) * linesize;
+        src  += (h - 1 - src_y) * src_linesize;
         src_y = h - 1;
     } else if (src_y <= -block_h) {
-        src  += (1 - block_h - src_y) * linesize;
+        src  += (1 - block_h - src_y) * src_linesize;
         src_y = 1 - block_h;
     }
     if (src_x >= w) {
@@ -52,30 +53,30 @@ static void FUNC(ff_emulated_edge_mc)(uint8_t *buf, const uint8_t *src,
     assert(start_x < end_x && block_w);
 
     w    = end_x - start_x;
-    src += start_y * linesize + start_x * sizeof(pixel);
+    src += start_y * src_linesize + start_x * sizeof(pixel);
     buf += start_x * sizeof(pixel);
 
     // top
     for (y = 0; y < start_y; y++) {
         memcpy(buf, src, w * sizeof(pixel));
-        buf += linesize;
+        buf += buf_linesize;
     }
 
     // copy existing part
     for (; y < end_y; y++) {
         memcpy(buf, src, w * sizeof(pixel));
-        src += linesize;
-        buf += linesize;
+        src += src_linesize;
+        buf += buf_linesize;
     }
 
     // bottom
-    src -= linesize;
+    src -= src_linesize;
     for (; y < block_h; y++) {
         memcpy(buf, src, w * sizeof(pixel));
-        buf += linesize;
+        buf += buf_linesize;
     }
 
-    buf -= block_h * linesize + start_x * sizeof(pixel);
+    buf -= block_h * buf_linesize + start_x * sizeof(pixel);
     while (block_h--) {
         pixel *bufp = (pixel *) buf;
 
@@ -88,6 +89,6 @@ static void FUNC(ff_emulated_edge_mc)(uint8_t *buf, const uint8_t *src,
         for (x = end_x; x < block_w; x++) {
             bufp[x] = bufp[end_x - 1];
         }
-        buf += linesize;
+        buf += buf_linesize;
     }
 }

libavcodec/vp3.c

@@ -1543,7 +1543,11 @@ static void render_slice(Vp3DecodeContext *s, int slice)
                             uint8_t *temp= s->edge_emu_buffer;
                             if(stride<0) temp -= 8*stride;
 
-                            s->vdsp.emulated_edge_mc(temp, motion_source, stride, 9, 9, src_x, src_y, plane_width, plane_height);
+                            s->vdsp.emulated_edge_mc(temp, motion_source,
+                                                     stride, stride,
+                                                     9, 9, src_x, src_y,
+                                                     plane_width,
+                                                     plane_height);
                             motion_source= temp;
                         }
                     }

libavcodec/vp56.c

@@ -341,7 +341,8 @@ static void vp56_mc(VP56Context *s, int b, int plane, uint8_t *src,
         y<0 || y+12>=s->plane_height[plane]) {
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
                             src + s->block_offset[b] + (dy-2)*stride + (dx-2),
-                            stride, 12, 12, x, y,
+                            stride, stride,
+                            12, 12, x, y,
                             s->plane_width[plane],
                             s->plane_height[plane]);
         src_block = s->edge_emu_buffer;

libavcodec/vp8.c

@@ -1196,7 +1196,9 @@ void vp8_mc_luma(VP8Context *s, VP8ThreadData *td, uint8_t *dst,
         src += y_off * linesize + x_off;
         if (x_off < mx_idx || x_off >= width  - block_w - subpel_idx[2][mx] ||
             y_off < my_idx || y_off >= height - block_h - subpel_idx[2][my]) {
-            s->vdsp.emulated_edge_mc(td->edge_emu_buffer, src - my_idx * linesize - mx_idx, linesize,
+            s->vdsp.emulated_edge_mc(td->edge_emu_buffer,
+                                     src - my_idx * linesize - mx_idx,
+                                     linesize, linesize,
                                      block_w + subpel_idx[1][mx], block_h + subpel_idx[1][my],
                                      x_off - mx_idx, y_off - my_idx, width, height);
             src = td->edge_emu_buffer + mx_idx + linesize * my_idx;
@@ -1246,13 +1248,17 @@ void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1, uint8_t *dst
         ff_thread_await_progress(ref, (3 + y_off + block_h + subpel_idx[2][my]) >> 3, 0);
         if (x_off < mx_idx || x_off >= width  - block_w - subpel_idx[2][mx] ||
             y_off < my_idx || y_off >= height - block_h - subpel_idx[2][my]) {
-            s->vdsp.emulated_edge_mc(td->edge_emu_buffer, src1 - my_idx * linesize - mx_idx, linesize,
+            s->vdsp.emulated_edge_mc(td->edge_emu_buffer,
+                                     src1 - my_idx * linesize - mx_idx,
+                                     linesize, linesize,
                                      block_w + subpel_idx[1][mx], block_h + subpel_idx[1][my],
                                      x_off - mx_idx, y_off - my_idx, width, height);
             src1 = td->edge_emu_buffer + mx_idx + linesize * my_idx;
             mc_func[my_idx][mx_idx](dst1, linesize, src1, linesize, block_h, mx, my);
 
-            s->vdsp.emulated_edge_mc(td->edge_emu_buffer, src2 - my_idx * linesize - mx_idx, linesize,
+            s->vdsp.emulated_edge_mc(td->edge_emu_buffer,
+                                     src2 - my_idx * linesize - mx_idx,
+                                     linesize, linesize,
                                      block_w + subpel_idx[1][mx], block_h + subpel_idx[1][my],
                                      x_off - mx_idx, y_off - my_idx, width, height);
             src2 = td->edge_emu_buffer + mx_idx + linesize * my_idx;

libavcodec/wmv2.c

@@ -120,8 +120,12 @@ void ff_mspel_motion(MpegEncContext *s,
     if(s->flags&CODEC_FLAG_EMU_EDGE){
         if(src_x<1 || src_y<1 || src_x + 17  >= s->h_edge_pos
                               || src_y + h+1 >= v_edge_pos){
-            s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr - 1 - s->linesize, s->linesize, 19, 19,
-                             src_x-1, src_y-1, s->h_edge_pos, s->v_edge_pos);
+            s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
+                                     ptr - 1 - s->linesize,
+                                     s->linesize, s->linesize,
+                                     19, 19,
+                                     src_x - 1, src_y - 1,
+                                     s->h_edge_pos, s->v_edge_pos);
             ptr= s->edge_emu_buffer + 1 + s->linesize;
             emu=1;
         }
@@ -161,16 +165,22 @@ void ff_mspel_motion(MpegEncContext *s,
     offset = (src_y * uvlinesize) + src_x;
     ptr = ref_picture[1] + offset;
     if(emu){
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9,
-                         src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1);
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
+                                 s->uvlinesize, s->uvlinesize,
+                                 9, 9,
+                                 src_x, src_y,
+                                 s->h_edge_pos >> 1, s->v_edge_pos >> 1);
         ptr= s->edge_emu_buffer;
     }
     pix_op[1][dxy](dest_cb, ptr, uvlinesize, h >> 1);
 
     ptr = ref_picture[2] + offset;
     if(emu){
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9,
-                         src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1);
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
+                                 s->uvlinesize, s->uvlinesize,
+                                 9, 9,
+                                 src_x, src_y,
+                                 s->h_edge_pos >> 1, s->v_edge_pos >> 1);
         ptr= s->edge_emu_buffer;
     }
     pix_op[1][dxy](dest_cr, ptr, uvlinesize, h >> 1);

libavcodec/x86/videodsp_init.c

@@ -28,29 +28,127 @@
 #include "libavcodec/videodsp.h"
 
 #if HAVE_YASM
-typedef void emu_edge_core_func(uint8_t *buf, const uint8_t *src,
-                                x86_reg linesize, x86_reg start_y,
-                                x86_reg end_y, x86_reg block_h,
-                                x86_reg start_x, x86_reg end_x,
-                                x86_reg block_w);
-extern emu_edge_core_func ff_emu_edge_core_mmx;
-extern emu_edge_core_func ff_emu_edge_core_sse;
-
-static av_always_inline void emulated_edge_mc(uint8_t *buf, const uint8_t *src,
-                                              ptrdiff_t linesize,
-                                              int block_w, int block_h,
-                                              int src_x, int src_y,
-                                              int w, int h,
-                                              emu_edge_core_func *core_fn)
+typedef void emu_edge_vfix_func(uint8_t *dst, const uint8_t *src,
+                                x86_reg dst_stride, x86_reg src_stride,
+                                x86_reg start_y, x86_reg end_y, x86_reg bh);
+typedef void emu_edge_vvar_func(uint8_t *dst, const uint8_t *src,
+                                x86_reg dst_stride, x86_reg src_stride,
+                                x86_reg start_y, x86_reg end_y, x86_reg bh,
+                                x86_reg w);
+
+extern emu_edge_vfix_func ff_emu_edge_vfix1_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix2_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix3_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix4_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix5_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix6_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix7_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix8_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix9_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix10_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix11_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix12_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix13_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix14_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix15_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix16_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix17_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix18_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix19_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix20_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix21_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix22_mmx;
+#if ARCH_X86_32
+static emu_edge_vfix_func *vfixtbl_mmx[22] = {
+    &ff_emu_edge_vfix1_mmx,  &ff_emu_edge_vfix2_mmx,  &ff_emu_edge_vfix3_mmx,
+    &ff_emu_edge_vfix4_mmx,  &ff_emu_edge_vfix5_mmx,  &ff_emu_edge_vfix6_mmx,
+    &ff_emu_edge_vfix7_mmx,  &ff_emu_edge_vfix8_mmx,  &ff_emu_edge_vfix9_mmx,
+    &ff_emu_edge_vfix10_mmx, &ff_emu_edge_vfix11_mmx, &ff_emu_edge_vfix12_mmx,
+    &ff_emu_edge_vfix13_mmx, &ff_emu_edge_vfix14_mmx, &ff_emu_edge_vfix15_mmx,
+    &ff_emu_edge_vfix16_mmx, &ff_emu_edge_vfix17_mmx, &ff_emu_edge_vfix18_mmx,
+    &ff_emu_edge_vfix19_mmx, &ff_emu_edge_vfix20_mmx, &ff_emu_edge_vfix21_mmx,
+    &ff_emu_edge_vfix22_mmx
+};
+#endif
+extern emu_edge_vvar_func ff_emu_edge_vvar_mmx;
+extern emu_edge_vfix_func ff_emu_edge_vfix16_sse;
+extern emu_edge_vfix_func ff_emu_edge_vfix17_sse;
+extern emu_edge_vfix_func ff_emu_edge_vfix18_sse;
+extern emu_edge_vfix_func ff_emu_edge_vfix19_sse;
+extern emu_edge_vfix_func ff_emu_edge_vfix20_sse;
+extern emu_edge_vfix_func ff_emu_edge_vfix21_sse;
+extern emu_edge_vfix_func ff_emu_edge_vfix22_sse;
+static emu_edge_vfix_func *vfixtbl_sse[22] = {
+    ff_emu_edge_vfix1_mmx,  ff_emu_edge_vfix2_mmx,  ff_emu_edge_vfix3_mmx,
+    ff_emu_edge_vfix4_mmx,  ff_emu_edge_vfix5_mmx,  ff_emu_edge_vfix6_mmx,
+    ff_emu_edge_vfix7_mmx,  ff_emu_edge_vfix8_mmx,  ff_emu_edge_vfix9_mmx,
+    ff_emu_edge_vfix10_mmx, ff_emu_edge_vfix11_mmx, ff_emu_edge_vfix12_mmx,
+    ff_emu_edge_vfix13_mmx, ff_emu_edge_vfix14_mmx, ff_emu_edge_vfix15_mmx,
+    ff_emu_edge_vfix16_sse, ff_emu_edge_vfix17_sse, ff_emu_edge_vfix18_sse,
+    ff_emu_edge_vfix19_sse, ff_emu_edge_vfix20_sse, ff_emu_edge_vfix21_sse,
+    ff_emu_edge_vfix22_sse
+};
+extern emu_edge_vvar_func ff_emu_edge_vvar_sse;
+
+typedef void emu_edge_hfix_func(uint8_t *dst, x86_reg dst_stride,
+                                x86_reg start_x, x86_reg bh);
+typedef void emu_edge_hvar_func(uint8_t *dst, x86_reg dst_stride,
+                                x86_reg start_x, x86_reg n_words, x86_reg bh);
+
+extern emu_edge_hfix_func ff_emu_edge_hfix2_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix4_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix6_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix8_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix10_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix12_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix14_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix16_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix18_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix20_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix22_mmx;
+#if ARCH_X86_32
+static emu_edge_hfix_func *hfixtbl_mmx[11] = {
+    ff_emu_edge_hfix2_mmx,  ff_emu_edge_hfix4_mmx,  ff_emu_edge_hfix6_mmx,
+    ff_emu_edge_hfix8_mmx,  ff_emu_edge_hfix10_mmx, ff_emu_edge_hfix12_mmx,
+    ff_emu_edge_hfix14_mmx, ff_emu_edge_hfix16_mmx, ff_emu_edge_hfix18_mmx,
+    ff_emu_edge_hfix20_mmx, ff_emu_edge_hfix22_mmx
+};
+#endif
+extern emu_edge_hvar_func ff_emu_edge_hvar_mmx;
+extern emu_edge_hfix_func ff_emu_edge_hfix16_sse;
+extern emu_edge_hfix_func ff_emu_edge_hfix18_sse;
+extern emu_edge_hfix_func ff_emu_edge_hfix20_sse;
+extern emu_edge_hfix_func ff_emu_edge_hfix22_sse;
+static emu_edge_hfix_func *hfixtbl_sse[11] = {
+    ff_emu_edge_hfix2_mmx,  ff_emu_edge_hfix4_mmx,  ff_emu_edge_hfix6_mmx,
+    ff_emu_edge_hfix8_mmx,  ff_emu_edge_hfix10_mmx, ff_emu_edge_hfix12_mmx,
+    ff_emu_edge_hfix14_mmx, ff_emu_edge_hfix16_sse, ff_emu_edge_hfix18_sse,
+    ff_emu_edge_hfix20_sse, ff_emu_edge_hfix22_sse
+};
+extern emu_edge_hvar_func ff_emu_edge_hvar_sse;
+
+static av_always_inline void emulated_edge_mc(uint8_t *dst, const uint8_t *src,
+                                              ptrdiff_t dst_stride,
+                                              ptrdiff_t src_stride,
+                                              x86_reg block_w, x86_reg block_h,
+                                              x86_reg src_x, x86_reg src_y,
+                                              x86_reg w, x86_reg h,
+                                              emu_edge_vfix_func **vfix_tbl,
+                                              emu_edge_vvar_func *v_extend_var,
+                                              emu_edge_hfix_func **hfix_tbl,
+                                              emu_edge_hvar_func *h_extend_var)
 {
-    int start_y, start_x, end_y, end_x, src_y_add = 0;
+    x86_reg start_y, start_x, end_y, end_x, src_y_add = 0, p;
+
+    if (!w || !h)
+         return;
 
     if (src_y >= h) {
-        src_y_add = h - 1 - src_y;
-        src_y     = h - 1;
+        src  -= src_y * src_stride;
+        src_y = src_y_add = h - 1;
     } else if (src_y <= -block_h) {
-        src_y_add = 1 - block_h - src_y;
-        src_y     = 1 - block_h;
+        src  -= src_y*src_stride;
+        src_y = src_y_add = 1 - block_h;
     }
     if (src_x >= w) {
         src   += w - 1 - src_x;
@@ -68,30 +166,62 @@ static av_always_inline void emulated_edge_mc(uint8_t *buf, const uint8_t *src,
     assert(start_y < end_y && block_h > 0);
 
     // fill in the to-be-copied part plus all above/below
-    src += (src_y_add + start_y) * linesize + start_x;
-    buf += start_x;
-    core_fn(buf, src, linesize, start_y, end_y,
-            block_h, start_x, end_x, block_w);
+    src += (src_y_add + start_y) * src_stride + start_x;
+    w = end_x - start_x;
+    if (w <= 22) {
+        vfix_tbl[w - 1](dst + start_x, src,
+                        dst_stride, src_stride,
+                        start_y, end_y, block_h);
+    } else {
+        v_extend_var(dst + start_x, src, dst_stride, src_stride,
+                     start_y, end_y, block_h, w);
+    }
+
+    // fill left
+    if (start_x) {
+        if (start_x <= 22) {
+            hfix_tbl[(start_x - 1) >> 1](dst, dst_stride, start_x, block_h);
+        } else {
+            h_extend_var(dst, dst_stride,
+                         start_x, (start_x + 1) >> 1, block_h);
+        }
+    }
+
+    // fill right
+    p = block_w - end_x;
+    if (p) {
+        if (p <= 22) {
+            hfix_tbl[(p - 1) >> 1](dst + end_x - (p & 1), dst_stride,
+                                   -!(p & 1), block_h);
+        } else {
+            h_extend_var(dst + end_x - (p & 1), dst_stride,
+                         -!(p & 1), (p + 1) >> 1, block_h);
+        }
+    }
 }
 
 #if ARCH_X86_32
 static av_noinline void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src,
-                                             ptrdiff_t linesize,
+                                             ptrdiff_t buf_stride,
+                                             ptrdiff_t src_stride,
                                              int block_w, int block_h,
                                              int src_x, int src_y, int w, int h)
 {
-    emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y,
-                     w, h, &ff_emu_edge_core_mmx);
+    emulated_edge_mc(buf, src, buf_stride, src_stride, block_w, block_h,
+                     src_x, src_y, w, h, vfixtbl_mmx, &ff_emu_edge_vvar_mmx,
+                     hfixtbl_mmx, &ff_emu_edge_hvar_mmx);
 }
 #endif
 
-static av_noinline void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src,
-                                             ptrdiff_t linesize,
+static av_noinline void emulated_edge_mc_sse(uint8_t * buf,const uint8_t *src,
+                                             ptrdiff_t buf_stride,
+                                             ptrdiff_t src_stride,
                                              int block_w, int block_h,
                                              int src_x, int src_y, int w, int h)
 {
-    emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y,
-                     w, h, &ff_emu_edge_core_sse);
+    emulated_edge_mc(buf, src, buf_stride, src_stride, block_w, block_h,
+                     src_x, src_y, w, h, vfixtbl_sse, &ff_emu_edge_vvar_sse,
+                     hfixtbl_sse, &ff_emu_edge_hvar_sse);
 }
 #endif /* HAVE_YASM */