Merge commit '5ae72f54532960cb9eae82a1c9e8d505106c022b'

* commit '5ae72f54532960cb9eae82a1c9e8d505106c022b':
  flashsv: check for keyframe before using differential coding
  h264: enable low delay only if no delayed frames were seen
  x86: fix build without inline asm

Conflicts:
	libavcodec/h264.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

3 files changed, 36 insertions, 19 deletions

diff --git a/libavcodec/flashsv.c b/libavcodec/flashsv.c
index 46a103b22c..1b65ecf6d1 100644
--- a/libavcodec/flashsv.c
+++ b/libavcodec/flashsv.c
@@ -379,6 +379,11 @@ static int flashsv_decode_frame(AVCodecContext *avctx, void *data,
                 }
 
                 if (has_diff) {
+                    if (!s->keyframe) {
+                        av_log(avctx, AV_LOG_ERROR,
+                               "inter frame without keyframe\n");
+                        return AVERROR_INVALIDDATA;
+                    }
                     s->diff_start  = get_bits(&gb, 8);
                     s->diff_height = get_bits(&gb, 8);
                     av_log(avctx, AV_LOG_DEBUG,
@@ -405,10 +410,6 @@ static int flashsv_decode_frame(AVCodecContext *avctx, void *data,
                 int k;
                 int off = (s->image_height - y_pos - 1) * s->frame.linesize[0];
 
-                if (!s->keyframe) {
-                    av_log(avctx, AV_LOG_ERROR, "no keyframe yet\n");
-                    return AVERROR_INVALIDDATA;
-                }
                 for (k = 0; k < cur_blk_height; k++)
                     memcpy(s->frame.data[0] + off - k*s->frame.linesize[0] + x_pos*3,
                            s->keyframe + off - k*s->frame.linesize[0] + x_pos*3,
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index d9469e82b6..f1c5104e80 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -4007,8 +4007,15 @@ again:
 
                 if (s->flags & CODEC_FLAG_LOW_DELAY ||
                     (h->sps.bitstream_restriction_flag &&
-                     !h->sps.num_reorder_frames))
-                    s->low_delay = 1;
+                     !h->sps.num_reorder_frames)) {
+                    if (s->avctx->has_b_frames > 1 || h->delayed_pic[0])
+                        av_log(avctx, AV_LOG_WARNING, "Delayed frames seen "
+                               "reenabling low delay requires a codec "
+                               "flush.\n");
+                        else
+                            s->low_delay = 1;
+                }
+
                 if (avctx->has_b_frames < 2)
                     avctx->has_b_frames = !s->low_delay;
                 break;
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index d09ce5c5a7..3fb9d211b5 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -2431,6 +2431,18 @@ static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
 #if HAVE_INLINE_ASM
     c->prefetch = prefetch_mmxext;
 
+    SET_QPEL_FUNCS(avg_qpel,        0, 16, mmxext, );
+    SET_QPEL_FUNCS(avg_qpel,        1,  8, mmxext, );
+    SET_QPEL_FUNCS(avg_2tap_qpel,   0, 16, mmxext, );
+    SET_QPEL_FUNCS(avg_2tap_qpel,   1,  8, mmxext, );
+
+    SET_QPEL_FUNCS(put_qpel,        0, 16, mmxext, );
+    SET_QPEL_FUNCS(put_qpel,        1,  8, mmxext, );
+    SET_QPEL_FUNCS(put_2tap_qpel,   0, 16, mmxext, );
+    SET_QPEL_FUNCS(put_2tap_qpel,   1,  8, mmxext, );
+    SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmxext, );
+    SET_QPEL_FUNCS(put_no_rnd_qpel, 1,  8, mmxext, );
+
     if (!high_bit_depth) {
         c->put_pixels_tab[0][1] = put_pixels16_x2_mmxext;
         c->put_pixels_tab[0][2] = put_pixels16_y2_mmxext;
@@ -2468,13 +2480,6 @@ static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
 
 #if HAVE_MMXEXT_EXTERNAL
     if (CONFIG_H264QPEL) {
-        SET_QPEL_FUNCS(put_qpel,        0, 16, mmxext, );
-        SET_QPEL_FUNCS(put_qpel,        1,  8, mmxext, );
-        SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmxext, );
-        SET_QPEL_FUNCS(put_no_rnd_qpel, 1,  8, mmxext, );
-        SET_QPEL_FUNCS(avg_qpel,        0, 16, mmxext, );
-        SET_QPEL_FUNCS(avg_qpel,        1,  8, mmxext, );
-
         if (!high_bit_depth) {
             SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmxext, );
             SET_QPEL_FUNCS(put_h264_qpel, 1,  8, mmxext, );
@@ -2492,11 +2497,6 @@ static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
             SET_QPEL_FUNCS(put_h264_qpel, 2, 4,  10_mmxext, ff_);
             SET_QPEL_FUNCS(avg_h264_qpel, 2, 4,  10_mmxext, ff_);
         }
-
-        SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmxext, );
-        SET_QPEL_FUNCS(put_2tap_qpel, 1,  8, mmxext, );
-        SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmxext, );
-        SET_QPEL_FUNCS(avg_2tap_qpel, 1,  8, mmxext, );
     }
 
     if (!high_bit_depth && CONFIG_H264CHROMA) {
@@ -2626,10 +2626,19 @@ static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags)
 static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
                               int mm_flags)
 {
-#if HAVE_SSE2_EXTERNAL
     const int bit_depth      = avctx->bits_per_raw_sample;
     const int high_bit_depth = bit_depth > 8;
 
+#if HAVE_SSE2_INLINE
+    if (!high_bit_depth && avctx->idct_algo == FF_IDCT_XVIDMMX) {
+        c->idct_put              = ff_idct_xvid_sse2_put;
+        c->idct_add              = ff_idct_xvid_sse2_add;
+        c->idct                  = ff_idct_xvid_sse2;
+        c->idct_permutation_type = FF_SSE2_IDCT_PERM;
+    }
+#endif /* HAVE_SSE2_INLINE */
+
+#if HAVE_SSE2_EXTERNAL
     if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
         // these functions are slower than mmx on AMD, but faster on Intel
         if (!high_bit_depth) {