diff --git a/libavcodec/arm/h264cmc_neon.S b/libavcodec/arm/h264cmc_neon.S
index 3aeb5458c3..c7e54605bb 100644
--- a/libavcodec/arm/h264cmc_neon.S
+++ b/libavcodec/arm/h264cmc_neon.S
@@ -58,15 +58,14 @@ T       cmp             r7,  #0
         vdup.8          d1,  r12
         vld1.8          {d4, d5}, [r1], r4
         vdup.8          d2,  r6
+        vld1.8          {d6, d7}, [r5], r4
         vdup.8          d3,  r7
 
         vext.8          d5,  d4,  d5,  #1
-
-1:
-        vld1.64         {d6, d7}, [r5], r4
-        pld             [r5]
-        vmull.u8        q8,  d4,  d0
         vext.8          d7,  d6,  d7,  #1
+
+1:      pld             [r5]
+        vmull.u8        q8,  d4,  d0
         vmlal.u8        q8,  d5,  d1
         vld1.8          {d4, d5}, [r1], r4
         vmlal.u8        q8,  d6,  d2
@@ -77,6 +76,7 @@ T       cmp             r7,  #0
         vmlal.u8        q9,  d7,  d1
         vmlal.u8        q9,  d4,  d2
         vmlal.u8        q9,  d5,  d3
+        vld1.8          {d6, d7}, [r5], r4
         pld             [r1]
   .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
@@ -92,6 +92,7 @@ T       cmp             r7,  #0
         vld1.8          {d21}, [lr,:64], r2
         vrhadd.u8       q8,  q8,  q10
   .endif
+        vext.8          d7,  d6,  d7,  #1
         vst1.8          {d16}, [r0,:64], r2
         vst1.8          {d17}, [r0,:64], r2
         bgt             1b
@@ -107,16 +108,16 @@ T       cmp             r7,  #0
 
         add             r5,  r1,  r2
         lsl             r4,  r2,  #1
-
-3:
         vld1.8          {d4}, [r1], r4
         vld1.8          {d6}, [r5], r4
 
-        pld             [r5]
+3:      pld             [r5]
         vmull.u8        q8,  d4,  d0
         vmlal.u8        q8,  d6,  d1
+        vld1.8          {d4}, [r1], r4
         vmull.u8        q9,  d6,  d0
         vmlal.u8        q9,  d4,  d1
+        vld1.8          {d6}, [r5], r4
   .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
         vrshrn.u16      d17, q9,  #6
@@ -144,13 +145,15 @@ T       cmp             r7,  #0
         vext.8          d5,  d4,  d5,  #1
         vext.8          d7,  d6,  d7,  #1
 
-        pld             [r1]
+5:      pld             [r1]
         subs            r3,  r3,  #2
         vmull.u8        q8,  d4,  d0
         vmlal.u8        q8,  d5,  d1
+        vld1.8          {d4, d5}, [r1], r2
         vmull.u8        q9,  d6,  d0
         vmlal.u8        q9,  d7,  d1
         pld             [r1]
+        vext.8          d5,  d4,  d5,  #1
   .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
         vrshrn.u16      d17, q9,  #6
@@ -165,9 +168,11 @@ T       cmp             r7,  #0
         vld1.8          {d21}, [lr,:64], r2
         vrhadd.u8       q8,  q8,  q10
   .endif
+        vld1.8          {d6, d7}, [r1], r2
+        vext.8          d7,  d6,  d7,  #1
         vst1.8          {d16}, [r0,:64], r2
         vst1.8          {d17}, [r0,:64], r2
-        bgt             4b
+        bgt             5b
 
         pop             {r4-r7, pc}
 endfunc
@@ -177,7 +182,7 @@ endfunc
 .macro  h264_chroma_mc4 type, codec=h264
 function ff_\type\()_\codec\()_chroma_mc4_neon, export=1
         push            {r4-r7, lr}
-        ldrd            r4,  [sp, #20]
+        ldrd            r4,  r5,  [sp, #20]
   .ifc \type,avg
         mov             lr,  r0
   .endif
@@ -211,27 +216,26 @@ T       cmp             r7,  #0
         vdup.8          d1,  r12
         vld1.8          {d4},     [r1], r4
         vdup.8          d2,  r6
+        vld1.8          {d6},     [r5], r4
         vdup.8          d3,  r7
 
         vext.8          d5,  d4,  d5,  #1
-        vtrn.32         d0,  d1
-        vtrn.32         d2,  d3
-        vtrn.32         d4,  d5
-
-1:
-        vld1.8          {d6},     [r5], r4
-        pld             [r5]
         vext.8          d7,  d6,  d7,  #1
-        vmull.u8        q8,  d4,  d0
+        vtrn.32         d4,  d5
         vtrn.32         d6,  d7
 
-        vld1.8          {d4},     [r1], r4
-        vmlal.u8        q8,  d6,  d2
-        vext.8          d5,  d4,  d5,  #1
-        vmull.u8        q9,  d6,  d0
-        vtrn.32         d4,  d5
-        vmlal.u8        q9,  d4,  d2
+        vtrn.32         d0,  d1
+        vtrn.32         d2,  d3
 
+1:      pld             [r5]
+        vmull.u8        q8,  d4,  d0
+        vmlal.u8        q8,  d6,  d2
+        vld1.8          {d4},     [r1], r4
+        vext.8          d5,  d4,  d5,  #1
+        vtrn.32         d4,  d5
+        vmull.u8        q9,  d6,  d0
+        vmlal.u8        q9,  d4,  d2
+        vld1.8          {d6},     [r5], r4
         vadd.i16        d16, d16, d17
         vadd.i16        d17, d18, d19
   .ifc \codec,h264
@@ -247,6 +251,8 @@ T       cmp             r7,  #0
         vld1.32         {d20[1]}, [lr,:32], r2
         vrhadd.u8       d16, d16, d20
   .endif
+        vext.8          d7,  d6,  d7,  #1
+        vtrn.32         d6,  d7
         vst1.32         {d16[0]}, [r0,:32], r2
         vst1.32         {d16[1]}, [r0,:32], r2
         bgt             1b
@@ -265,14 +271,13 @@ T       cmp             r7,  #0
         add             r5,  r1,  r2
         lsl             r4,  r2,  #1
         vld1.32         {d4[0]},  [r1], r4
-3:
         vld1.32         {d4[1]},  [r5], r4
 
-        pld             [r5]
+3:      pld             [r5]
         vmull.u8        q8,  d4,  d0
         vld1.32         {d4[0]},  [r1], r4
         vmull.u8        q9,  d4,  d1
-
+        vld1.32         {d4[1]},  [r5], r4
         vadd.i16        d16, d16, d17
         vadd.i16        d17, d18, d19
   .ifc \codec,h264
@@ -301,10 +306,12 @@ T       cmp             r7,  #0
         vtrn.32         d4,  d5
         vtrn.32         d6,  d7
 
-        vmull.u8        q8,  d4,  d0
+5:      vmull.u8        q8,  d4,  d0
         vmull.u8        q9,  d6,  d0
         subs            r3,  r3,  #2
-
+        vld1.8          {d4},     [r1], r2
+        vext.8          d5,  d4,  d5,  #1
+        vtrn.32         d4,  d5
         vadd.i16        d16, d16, d17
         vadd.i16        d17, d18, d19
         pld             [r1]
@@ -319,10 +326,13 @@ T       cmp             r7,  #0
         vld1.32         {d20[1]}, [lr,:32], r2
         vrhadd.u8       d16, d16, d20
   .endif
+        vld1.8          {d6},     [r1], r2
+        vext.8          d7,  d6,  d7,  #1
+        vtrn.32         d6,  d7
         pld             [r1]
         vst1.32         {d16[0]}, [r0,:32], r2
         vst1.32         {d16[1]}, [r0,:32], r2
-        bgt             4b
+        bgt             5b
 
         pop             {r4-r7, pc}
 endfunc