diff --git a/vp9/common/arm/neon/vp9_iht8x8_add_neon.asm b/vp9/common/arm/neon/vp9_iht8x8_add_neon.asm
index 93d3af3011c281ae70f7681b4cc4f362e22acd0a..b41f5661b805127a86df7259d5856e6f0e0fc3ae 100644
--- a/vp9/common/arm/neon/vp9_iht8x8_add_neon.asm
+++ b/vp9/common/arm/neon/vp9_iht8x8_add_neon.asm
@@ -576,6 +576,7 @@
     vld1.s16        {q14,q15}, [r0]!
 
     push            {r0-r10}
+    vpush           {d8-d15}
 
     ; transpose the input data
     TRANSPOSE8X8
@@ -636,6 +637,7 @@ iadst_iadst
     IADST8X8_1D
 
 end_vp9_iht8x8_64_add_neon
+    vpop           {d8-d15}
     pop            {r0-r10}
 
     ; ROUND_POWER_OF_TWO(temp_out[j], 5)