Fix convolve_copy SIMD functions This continued to fix convolve_copy SIMD functions following "commit f191d8703412d5ba7b5b2c3d76354eca6c06e24e" to avoid potential out-of-boundary reads. Change-Id: I1e136ac0b0b9db30bec21f3694530fcef065bae3

diff --git a/av1/common/x86/convolve_2d_sse2.c b/av1/common/x86/convolve_2d_sse2.c
index 174b6dd..4cc3341 100644
--- a/av1/common/x86/convolve_2d_sse2.c
+++ b/av1/common/x86/convolve_2d_sse2.c

@@ -255,24 +255,20 @@
 
   if (w == 2) {
     do {
-      __m128i s = _mm_loadl_epi64((__m128i *)src);
-      *(uint16_t *)dst = _mm_cvtsi128_si32(s);
+      memcpy(dst, src, 2 * sizeof(*src));
       src += src_stride;
       dst += dst_stride;
-      s = _mm_loadl_epi64((__m128i *)src);
-      *(uint16_t *)dst = _mm_cvtsi128_si32(s);
+      memcpy(dst, src, 2 * sizeof(*src));
       src += src_stride;
       dst += dst_stride;
       h -= 2;
     } while (h);
   } else if (w == 4) {
     do {
-      __m128i s = _mm_loadl_epi64((__m128i *)src);
-      *(uint32_t *)dst = _mm_cvtsi128_si32(s);
+      memcpy(dst, src, 4 * sizeof(*src));
       src += src_stride;
       dst += dst_stride;
-      s = _mm_loadl_epi64((__m128i *)src);
-      *(uint32_t *)dst = _mm_cvtsi128_si32(s);
+      memcpy(dst, src, 4 * sizeof(*src));
       src += src_stride;
       dst += dst_stride;
       h -= 2;

diff --git a/av1/common/x86/highbd_convolve_2d_avx2.c b/av1/common/x86/highbd_convolve_2d_avx2.c
index 63c1cf1..75dadf8 100644
--- a/av1/common/x86/highbd_convolve_2d_avx2.c
+++ b/av1/common/x86/highbd_convolve_2d_avx2.c

@@ -238,12 +238,10 @@
 
   if (w == 2) {
     do {
-      __m128i s = _mm_loadl_epi64((__m128i *)src);
-      *(uint32_t *)dst = _mm_cvtsi128_si32(s);
+      memcpy(dst, src, 2 * sizeof(*src));
       src += src_stride;
       dst += dst_stride;
-      s = _mm_loadl_epi64((__m128i *)src);
-      *(uint32_t *)dst = _mm_cvtsi128_si32(s);
+      memcpy(dst, src, 2 * sizeof(*src));
       src += src_stride;
       dst += dst_stride;
       h -= 2;