x86: h264qpel: Move stray comment to the right spot and clarify it

1 files changed, 4 insertions, 0 deletions

diff --git a/libavcodec/x86/h264_qpel.c b/libavcodec/x86/h264_qpel.c
index 9157223c84..64af8fe17c 100644
--- a/libavcodec/x86/h264_qpel.c
+++ b/libavcodec/x86/h264_qpel.c
@@ -611,6 +611,10 @@ void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
     }
 
     if (EXTERNAL_AVX(mm_flags)) {
+        /* AVX implies 64 byte cache lines without the need to avoid unaligned
+         * memory accesses that cross the boundary between two cache lines.
+         * TODO: Port X264_CPU_CACHELINE_32/64 detection from x264 to avoid
+         * having to treat SSE2 functions with such properties as AVX. */
         if (bit_depth == 10) {
             H264_QPEL_FUNCS_10(1, 0, sse2);
             H264_QPEL_FUNCS_10(2, 0, sse2);