diff --git a/libavcodec/aarch64/h264pred_init.c b/libavcodec/aarch64/h264pred_init.c
index 0ae8f70d23..9819f85f7d 100644
--- a/libavcodec/aarch64/h264pred_init.c
+++ b/libavcodec/aarch64/h264pred_init.c
@@ -25,6 +25,19 @@
 #include "libavcodec/avcodec.h"
 #include "libavcodec/h264pred.h"
 
+/* PERFORMANCE WARNING:
+ * These assembly optimizations have been identified as "performance regressions."
+ * Due to advancements in modern CPU micro-architectures and compiler optimization
+ * the C implementations now consistently outperform these handwritten routines.
+ *
+ * Keep them here for historical reference.
+ *
+ * New optimizations are highly welcome! If you can provide an optimized
+ * implementation that demonstrably beats the current C version in rigorous
+ * benchmarks, please submit a patch.
+ */
+#define ENABLE_INEFFICIENT_ASM 0
+
 void ff_pred16x16_vert_neon(uint8_t *src, ptrdiff_t stride);
 void ff_pred16x16_hor_neon(uint8_t *src, ptrdiff_t stride);
 void ff_pred16x16_plane_neon(uint8_t *src, ptrdiff_t stride);
@@ -69,16 +82,22 @@ static av_cold void h264_pred_init_neon(H264PredContext *h, int codec_id,
 {
     if (bit_depth == 8) {
         if (chroma_format_idc <= 1) {
+#if ENABLE_INEFFICIENT_ASM
             h->pred8x8[VERT_PRED8x8     ] = ff_pred8x8_vert_neon;
             h->pred8x8[HOR_PRED8x8      ] = ff_pred8x8_hor_neon;
+#endif
             if (codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8)
                 h->pred8x8[PLANE_PRED8x8] = ff_pred8x8_plane_neon;
+#if ENABLE_INEFFICIENT_ASM
             h->pred8x8[DC_128_PRED8x8   ] = ff_pred8x8_128_dc_neon;
+#endif
             if (codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP7 &&
                 codec_id != AV_CODEC_ID_VP8) {
+#if ENABLE_INEFFICIENT_ASM
                 h->pred8x8[DC_PRED8x8     ] = ff_pred8x8_dc_neon;
                 h->pred8x8[LEFT_DC_PRED8x8] = ff_pred8x8_left_dc_neon;
                 h->pred8x8[TOP_DC_PRED8x8 ] = ff_pred8x8_top_dc_neon;
+#endif
                 h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8] = ff_pred8x8_l0t_dc_neon;
                 h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8] = ff_pred8x8_0lt_dc_neon;
                 h->pred8x8[ALZHEIMER_DC_L00_PRED8x8] = ff_pred8x8_l00_dc_neon;
@@ -86,27 +105,37 @@ static av_cold void h264_pred_init_neon(H264PredContext *h, int codec_id,
             }
         }
 
+#if ENABLE_INEFFICIENT_ASM
         h->pred16x16[DC_PRED8x8     ] = ff_pred16x16_dc_neon;
+#endif
         h->pred16x16[VERT_PRED8x8   ] = ff_pred16x16_vert_neon;
         h->pred16x16[HOR_PRED8x8    ] = ff_pred16x16_hor_neon;
+#if ENABLE_INEFFICIENT_ASM
         h->pred16x16[LEFT_DC_PRED8x8] = ff_pred16x16_left_dc_neon;
         h->pred16x16[TOP_DC_PRED8x8 ] = ff_pred16x16_top_dc_neon;
         h->pred16x16[DC_128_PRED8x8 ] = ff_pred16x16_128_dc_neon;
+#endif
         if (codec_id != AV_CODEC_ID_SVQ3 && codec_id != AV_CODEC_ID_RV40 &&
             codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8)
             h->pred16x16[PLANE_PRED8x8  ] = ff_pred16x16_plane_neon;
     }
     if (bit_depth == 10) {
         if (chroma_format_idc <= 1) {
+#if ENABLE_INEFFICIENT_ASM
             h->pred8x8[VERT_PRED8x8     ] = ff_pred8x8_vert_neon_10;
+#endif
             h->pred8x8[HOR_PRED8x8      ] = ff_pred8x8_hor_neon_10;
             if (codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8)
                 h->pred8x8[PLANE_PRED8x8] = ff_pred8x8_plane_neon_10;
+#if ENABLE_INEFFICIENT_ASM
             h->pred8x8[DC_128_PRED8x8   ] = ff_pred8x8_128_dc_neon_10;
+#endif
             if (codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP7 &&
                 codec_id != AV_CODEC_ID_VP8) {
+#if ENABLE_INEFFICIENT_ASM
                 h->pred8x8[DC_PRED8x8     ] = ff_pred8x8_dc_neon_10;
                 h->pred8x8[LEFT_DC_PRED8x8] = ff_pred8x8_left_dc_neon_10;
+#endif
                 h->pred8x8[TOP_DC_PRED8x8 ] = ff_pred8x8_top_dc_neon_10;
                 h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8] = ff_pred8x8_l0t_dc_neon_10;
                 h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8] = ff_pred8x8_0lt_dc_neon_10;