diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
index c9fa31900f..b14ed117b9 100644
--- a/libavcodec/hevc.c
+++ b/libavcodec/hevc.c
@@ -834,18 +834,18 @@ static int hls_transform_unit(HEVCContext *s, int x0, int y0,
         int trafo_size = 1 << log2_trafo_size;
         ff_hevc_set_neighbour_available(s, x0, y0, trafo_size, trafo_size);
 
-        s->hpc.intra_pred(s, x0, y0, log2_trafo_size, 0);
+        s->hpc.intra_pred[log2_trafo_size - 2](s, x0, y0, 0);
         if (log2_trafo_size > 2) {
             trafo_size = trafo_size << (s->sps->hshift[1] - 1);
             ff_hevc_set_neighbour_available(s, x0, y0, trafo_size, trafo_size);
-            s->hpc.intra_pred(s, x0, y0, log2_trafo_size - 1, 1);
-            s->hpc.intra_pred(s, x0, y0, log2_trafo_size - 1, 2);
+            s->hpc.intra_pred[log2_trafo_size - 3](s, x0, y0, 1);
+            s->hpc.intra_pred[log2_trafo_size - 3](s, x0, y0, 2);
         } else if (blk_idx == 3) {
             trafo_size = trafo_size << s->sps->hshift[1];
             ff_hevc_set_neighbour_available(s, xBase, yBase,
                                             trafo_size, trafo_size);
-            s->hpc.intra_pred(s, xBase, yBase, log2_trafo_size, 1);
-            s->hpc.intra_pred(s, xBase, yBase, log2_trafo_size, 2);
+            s->hpc.intra_pred[log2_trafo_size - 2](s, xBase, yBase, 1);
+            s->hpc.intra_pred[log2_trafo_size - 2](s, xBase, yBase, 2);
         }
     }
 
diff --git a/libavcodec/hevcpred.c b/libavcodec/hevcpred.c
index e4987f8eca..4dfbfbfed0 100644
--- a/libavcodec/hevcpred.c
+++ b/libavcodec/hevcpred.c
@@ -42,7 +42,10 @@ void ff_hevc_pred_init(HEVCPredContext *hpc, int bit_depth)
 #define FUNC(a, depth) a ## _ ## depth
 
 #define HEVC_PRED(depth)                                \
-    hpc->intra_pred      = FUNC(intra_pred, depth);     \
+    hpc->intra_pred[0]   = FUNC(intra_pred_2, depth);   \
+    hpc->intra_pred[1]   = FUNC(intra_pred_3, depth);   \
+    hpc->intra_pred[2]   = FUNC(intra_pred_4, depth);   \
+    hpc->intra_pred[3]   = FUNC(intra_pred_5, depth);   \
     hpc->pred_planar[0]  = FUNC(pred_planar_0, depth);  \
     hpc->pred_planar[1]  = FUNC(pred_planar_1, depth);  \
     hpc->pred_planar[2]  = FUNC(pred_planar_2, depth);  \
diff --git a/libavcodec/hevcpred.h b/libavcodec/hevcpred.h
index 4dead2ecdc..7f14a76da1 100644
--- a/libavcodec/hevcpred.h
+++ b/libavcodec/hevcpred.h
@@ -29,13 +29,15 @@
 struct HEVCContext;
 
 typedef struct HEVCPredContext {
-    void (*intra_pred)(struct HEVCContext *s, int x0, int y0, int log2_size, int c_idx);
+    void (*intra_pred[4])(struct HEVCContext *s, int x0, int y0, int c_idx);
 
-    void(*pred_planar[4])(uint8_t *src, const uint8_t *top, const uint8_t *left, ptrdiff_t stride);
-    void(*pred_dc)(uint8_t *src, const uint8_t *top, const uint8_t *left, ptrdiff_t stride,
-                   int log2_size, int c_idx);
-    void(*pred_angular[4])(uint8_t *src, const uint8_t *top, const uint8_t *left, ptrdiff_t stride,
-                         int c_idx, int mode);
+    void (*pred_planar[4])(uint8_t *src, const uint8_t *top,
+                           const uint8_t *left, ptrdiff_t stride);
+    void (*pred_dc)(uint8_t *src, const uint8_t *top, const uint8_t *left,
+                    ptrdiff_t stride, int log2_size, int c_idx);
+    void (*pred_angular[4])(uint8_t *src, const uint8_t *top,
+                            const uint8_t *left, ptrdiff_t stride,
+                            int c_idx, int mode);
 } HEVCPredContext;
 
 void ff_hevc_pred_init(HEVCPredContext *hpc, int bit_depth);
diff --git a/libavcodec/hevcpred_template.c b/libavcodec/hevcpred_template.c
index ece40ddbaf..ad99a8493e 100644
--- a/libavcodec/hevcpred_template.c
+++ b/libavcodec/hevcpred_template.c
@@ -27,7 +27,8 @@
 
 #define POS(x, y) src[(x) + stride * (y)]
 
-static void FUNC(intra_pred)(HEVCContext *s, int x0, int y0, int log2_size, int c_idx)
+static av_always_inline void FUNC(intra_pred)(HEVCContext *s, int x0, int y0,
+                                              int log2_size, int c_idx)
 {
 #define PU(x) \
     ((x) >> s->sps->log2_min_pu_size)
@@ -362,6 +363,19 @@ static void FUNC(intra_pred)(HEVCContext *s, int x0, int y0, int log2_size, int
     }
 }
 
+#define INTRA_PRED(size)                                                            \
+static void FUNC(intra_pred_ ## size)(HEVCContext *s, int x0, int y0, int c_idx)    \
+{                                                                                   \
+    FUNC(intra_pred)(s, x0, y0, size, c_idx);                                       \
+}
+
+INTRA_PRED(2)
+INTRA_PRED(3)
+INTRA_PRED(4)
+INTRA_PRED(5)
+
+#undef INTRA_PRED
+
 static av_always_inline void FUNC(pred_planar)(uint8_t *_src, const uint8_t *_top,
                                   const uint8_t *_left, ptrdiff_t stride,
                                   int trafo_size)