lagarith: add YUY2 decoding support

Unlike other variants, for YUY2 we need to use different prediction:
* on line 0 for luma we should left predict starting from the second pixel
* on line 1 we should left predict first 4 pixels for luma and 2 for chroma
* median prediction employed here is taken directly from HuffYUV

libavcodec/lagarith.c

@@ -269,6 +269,40 @@ static void lag_pred_line(LagarithContext *l, uint8_t *buf,
     }
 }
 
+static void lag_pred_line_yuy2(LagarithContext *l, uint8_t *buf,
+                               int width, int stride, int line,
+                               int is_luma)
+{
+    int L, TL;
+
+    if (!line) {
+        if (is_luma) {
+            buf++;
+            width--;
+        }
+        l->dsp.add_hfyu_left_prediction(buf + 1, buf + 1, width - 1, buf[0]);
+        return;
+    }
+    if (line == 1) {
+        const int HEAD = is_luma ? 4 : 2;
+        int i;
+
+        L  = buf[width - stride - 1];
+        TL = buf[HEAD  - stride - 1];
+        for (i = 0; i < HEAD; i++) {
+            L += buf[i];
+            buf[i] = L;
+        }
+        buf   += HEAD;
+        width -= HEAD;
+    } else {
+        TL = buf[width - (2 * stride) - 1];
+        L  = buf[width - stride - 1];
+    }
+    l->dsp.add_hfyu_median_prediction(buf, buf - stride, buf, width,
+                                      &L, &TL);
+}
+
 static int lag_decode_line(LagarithContext *l, lag_rac *rac,
                            uint8_t *dst, int width, int stride,
                            int esc_count)
@@ -432,9 +466,17 @@ static int lag_decode_arith_plane(LagarithContext *l, uint8_t *dst,
         return -1;
     }
 
-    for (i = 0; i < height; i++) {
-        lag_pred_line(l, dst, width, stride, i);
-        dst += stride;
+    if (l->avctx->pix_fmt != PIX_FMT_YUV422P) {
+        for (i = 0; i < height; i++) {
+            lag_pred_line(l, dst, width, stride, i);
+            dst += stride;
+        }
+    } else {
+        for (i = 0; i < height; i++) {
+            lag_pred_line_yuy2(l, dst, width, stride, i,
+                               width == l->avctx->width);
+            dst += stride;
+        }
     }
 
     return 0;
@@ -557,6 +599,32 @@ static int lag_decode_frame(AVCodecContext *avctx,
                 srcs[i] += l->rgb_stride;
         }
         break;
+    case FRAME_ARITH_YUY2:
+        avctx->pix_fmt = PIX_FMT_YUV422P;
+
+        if (avctx->get_buffer(avctx, p) < 0) {
+            av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+            return -1;
+        }
+
+        if (offset_ry >= buf_size ||
+            offset_gu >= buf_size ||
+            offset_bv >= buf_size) {
+            av_log(avctx, AV_LOG_ERROR,
+                   "Invalid frame offsets\n");
+            return AVERROR_INVALIDDATA;
+        }
+
+        lag_decode_arith_plane(l, p->data[0], avctx->width, avctx->height,
+                               p->linesize[0], buf + offset_ry,
+                               buf_size - offset_ry);
+        lag_decode_arith_plane(l, p->data[2], avctx->width / 2,
+                               avctx->height, p->linesize[2],
+                               buf + offset_gu, buf_size - offset_gu);
+        lag_decode_arith_plane(l, p->data[1], avctx->width / 2,
+                               avctx->height, p->linesize[1],
+                               buf + offset_bv, buf_size - offset_bv);
+        break;
     case FRAME_ARITH_YV12:
         avctx->pix_fmt = PIX_FMT_YUV420P;