libvpx: Pull from upstream

source/libvpx/vp9/common/arm/neon/vp9_convolve_neon.c

Index: source/libvpx/vp9/common/arm/neon/vp9_convolve_neon.c
===================================================================
--- source/libvpx/vp9/common/arm/neon/vp9_convolve_neon.c	(revision 219822)
+++ source/libvpx/vp9/common/arm/neon/vp9_convolve_neon.c	(working copy)
@@ -10,6 +10,7 @@
 
 #include "./vp9_rtcd.h"
 #include "vp9/common/vp9_common.h"
+#include "vpx_ports/mem.h"
 
 void vp9_convolve8_neon(const uint8_t *src, ptrdiff_t src_stride,
                         uint8_t *dst, ptrdiff_t dst_stride,
@@ -19,7 +20,7 @@
   /* Given our constraints: w <= 64, h <= 64, taps == 8 we can reduce the
    * maximum buffer size to 64 * 64 + 7 (+ 1 to make it divisible by 4).
    */
-  uint8_t temp[64 * 72];
+  DECLARE_ALIGNED_ARRAY(8, uint8_t, temp, 64 * 72);
 
   // Account for the vertical phase needing 3 lines prior and 4 lines post
   int intermediate_height = h + 7;
@@ -53,7 +54,7 @@
                             const int16_t *filter_x, int x_step_q4,
                             const int16_t *filter_y, int y_step_q4,
                             int w, int h) {
-  uint8_t temp[64 * 72];
+  DECLARE_ALIGNED_ARRAY(8, uint8_t, temp, 64 * 72);
   int intermediate_height = h + 7;
 
   if (x_step_q4 != 16 || y_step_q4 != 16)