diff --git a/vp10/common/reconintra.c b/vp10/common/reconintra.c
index c58da59..9350d67 100644
--- a/vp10/common/reconintra.c
+++ b/vp10/common/reconintra.c
@@ -173,24 +173,24 @@
       /* slower path if the block needs border extension */
       if (x0 + 2 * bs <= frame_width) {
         if (right_available && bs == 4) {
-          memcpy(above_row, above_ref, 2 * bs * sizeof(uint16_t));
+          memcpy(above_row, above_ref, 2 * bs * sizeof(above_row[0]));
         } else {
-          memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+          memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
           vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         }
       } else if (x0 + bs <= frame_width) {
         const int r = frame_width - x0;
         if (right_available && bs == 4) {
-          memcpy(above_row, above_ref, r * sizeof(uint16_t));
+          memcpy(above_row, above_ref, r * sizeof(above_row[0]));
           vpx_memset16(above_row + r, above_row[r - 1],
                        x0 + 2 * bs - frame_width);
         } else {
-          memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+          memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
           vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         }
       } else if (x0 <= frame_width) {
         const int r = frame_width - x0;
-        memcpy(above_row, above_ref, r * sizeof(uint16_t));
+        memcpy(above_row, above_ref, r * sizeof(above_row[0]));
         vpx_memset16(above_row + r, above_row[r - 1],
                        x0 + 2 * bs - frame_width);
       }
@@ -201,9 +201,9 @@
       if (bs == 4 && right_available && left_available) {
         const_above_row = above_ref;
       } else {
-        memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+        memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
         if (bs == 4 && right_available)
-          memcpy(above_row + bs, above_ref + bs, bs * sizeof(uint16_t));
+          memcpy(above_row + bs, above_ref + bs, bs * sizeof(above_row[0]));
         else
           vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         // TODO(Peter): this value should probably change for high bitdepth
diff --git a/vp10/encoder/encodeframe.c b/vp10/encoder/encodeframe.c
index 56e9c0f..262bd34 100644
--- a/vp10/encoder/encodeframe.c
+++ b/vp10/encoder/encodeframe.c
@@ -1046,7 +1046,7 @@
 
   x->skip = ctx->skip;
   memcpy(x->zcoeff_blk[mbmi->tx_size], ctx->zcoeff_blk,
-         sizeof(uint8_t) * ctx->num_4x4_blk);
+         sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
 
   if (!output_enabled)
     return;
@@ -4155,8 +4155,6 @@
   if (x->skip_encode)
     return;
 
-  set_ref_ptrs(cm, xd, mbmi->ref_frame[0], mbmi->ref_frame[1]);
-
   if (!is_inter_block(mbmi)) {
     int plane;
     mbmi->skip = 1;
@@ -4168,6 +4166,7 @@
   } else {
     int ref;
     const int is_compound = has_second_ref(mbmi);
+    set_ref_ptrs(cm, xd, mbmi->ref_frame[0], mbmi->ref_frame[1]);
     for (ref = 0; ref < 1 + is_compound; ++ref) {
       YV12_BUFFER_CONFIG *cfg = get_ref_frame_buffer(cpi,
                                                      mbmi->ref_frame[ref]);
diff --git a/vp10/encoder/encoder.c b/vp10/encoder/encoder.c
index 1d69cd0..1eab6a3 100644
--- a/vp10/encoder/encoder.c
+++ b/vp10/encoder/encoder.c
@@ -4547,6 +4547,7 @@
         cpi->svc.spatial_layer_to_encode = 0;
     }
   }
+  vpx_clear_system_state();
   return 0;
 }
 
diff --git a/vp10/encoder/extend.c b/vp10/encoder/extend.c
index 2ffbac0..1cac5ac 100644
--- a/vp10/encoder/extend.c
+++ b/vp10/encoder/extend.c
@@ -74,7 +74,7 @@
 
   for (i = 0; i < h; i++) {
     vpx_memset16(dst_ptr1, src_ptr1[0], extend_left);
-    memcpy(dst_ptr1 + extend_left, src_ptr1, w * sizeof(uint16_t));
+    memcpy(dst_ptr1 + extend_left, src_ptr1, w * sizeof(src_ptr1[0]));
     vpx_memset16(dst_ptr2, src_ptr2[0], extend_right);
     src_ptr1 += src_pitch;
     src_ptr2 += src_pitch;
@@ -91,12 +91,12 @@
   linesize = extend_left + extend_right + w;
 
   for (i = 0; i < extend_top; i++) {
-    memcpy(dst_ptr1, src_ptr1, linesize * sizeof(uint16_t));
+    memcpy(dst_ptr1, src_ptr1, linesize * sizeof(src_ptr1[0]));
     dst_ptr1 += dst_pitch;
   }
 
   for (i = 0; i < extend_bottom; i++) {
-    memcpy(dst_ptr2, src_ptr2, linesize * sizeof(uint16_t));
+    memcpy(dst_ptr2, src_ptr2, linesize * sizeof(src_ptr2[0]));
     dst_ptr2 += dst_pitch;
   }
 }
diff --git a/vp10/encoder/rdopt.c b/vp10/encoder/rdopt.c
index 6773768..a0de455 100644
--- a/vp10/encoder/rdopt.c
+++ b/vp10/encoder/rdopt.c
@@ -3396,7 +3396,7 @@
         if (!x->select_tx_size)
           swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
         memcpy(ctx->zcoeff_blk, x->zcoeff_blk[mbmi->tx_size],
-               sizeof(uint8_t) * ctx->num_4x4_blk);
+               sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
 
         // TODO(debargha): enhance this test with a better distortion prediction
         // based on qp, activity mask and history
@@ -4137,7 +4137,7 @@
         if (!x->select_tx_size)
           swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
         memcpy(ctx->zcoeff_blk, x->zcoeff_blk[TX_4X4],
-               sizeof(uint8_t) * ctx->num_4x4_blk);
+               sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
 
         for (i = 0; i < 4; i++)
           best_bmodes[i] = xd->mi[0]->bmi[i];
diff --git a/vp10/encoder/resize.c b/vp10/encoder/resize.c
index 5e17532..5572c17 100644
--- a/vp10/encoder/resize.c
+++ b/vp10/encoder/resize.c
@@ -448,7 +448,7 @@
                              uint8_t *buf) {
   int steps;
   if (length == olength) {
-    memcpy(output, input, sizeof(uint8_t) * length);
+    memcpy(output, input, sizeof(output[0]) * length);
     return;
   }
   steps = get_down2_steps(length, olength);
@@ -741,7 +741,7 @@
                                     int bd) {
   int steps;
   if (length == olength) {
-    memcpy(output, input, sizeof(uint16_t) * length);
+    memcpy(output, input, sizeof(output[0]) * length);
     return;
   }
   steps = get_down2_steps(length, olength);
diff --git a/vp9/common/vp9_reconintra.c b/vp9/common/vp9_reconintra.c
index 46f08a7..e60eff8 100644
--- a/vp9/common/vp9_reconintra.c
+++ b/vp9/common/vp9_reconintra.c
@@ -186,24 +186,24 @@
       /* slower path if the block needs border extension */
       if (x0 + 2 * bs <= frame_width) {
         if (right_available && bs == 4) {
-          memcpy(above_row, above_ref, 2 * bs * sizeof(uint16_t));
+          memcpy(above_row, above_ref, 2 * bs * sizeof(above_row[0]));
         } else {
-          memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+          memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
           vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         }
       } else if (x0 + bs <= frame_width) {
         const int r = frame_width - x0;
         if (right_available && bs == 4) {
-          memcpy(above_row, above_ref, r * sizeof(uint16_t));
+          memcpy(above_row, above_ref, r * sizeof(above_row[0]));
           vpx_memset16(above_row + r, above_row[r - 1],
                        x0 + 2 * bs - frame_width);
         } else {
-          memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+          memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
           vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         }
       } else if (x0 <= frame_width) {
         const int r = frame_width - x0;
-        memcpy(above_row, above_ref, r * sizeof(uint16_t));
+        memcpy(above_row, above_ref, r * sizeof(above_row[0]));
         vpx_memset16(above_row + r, above_row[r - 1],
                        x0 + 2 * bs - frame_width);
       }
@@ -214,9 +214,9 @@
       if (bs == 4 && right_available && left_available) {
         const_above_row = above_ref;
       } else {
-        memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+        memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
         if (bs == 4 && right_available)
-          memcpy(above_row + bs, above_ref + bs, bs * sizeof(uint16_t));
+          memcpy(above_row + bs, above_ref + bs, bs * sizeof(above_row[0]));
         else
           vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         // TODO(Peter): this value should probably change for high bitdepth
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index 63ab157..5c12dc4 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -1046,7 +1046,7 @@
 
   x->skip = ctx->skip;
   memcpy(x->zcoeff_blk[mbmi->tx_size], ctx->zcoeff_blk,
-         sizeof(uint8_t) * ctx->num_4x4_blk);
+         sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
 
   if (!output_enabled)
     return;
@@ -4157,8 +4157,6 @@
   if (x->skip_encode)
     return;
 
-  set_ref_ptrs(cm, xd, mbmi->ref_frame[0], mbmi->ref_frame[1]);
-
   if (!is_inter_block(mbmi)) {
     int plane;
     mbmi->skip = 1;
@@ -4170,6 +4168,7 @@
   } else {
     int ref;
     const int is_compound = has_second_ref(mbmi);
+    set_ref_ptrs(cm, xd, mbmi->ref_frame[0], mbmi->ref_frame[1]);
     for (ref = 0; ref < 1 + is_compound; ++ref) {
       YV12_BUFFER_CONFIG *cfg = get_ref_frame_buffer(cpi,
                                                      mbmi->ref_frame[ref]);
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index 0680550..5261feb 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -4547,6 +4547,7 @@
         cpi->svc.spatial_layer_to_encode = 0;
     }
   }
+  vpx_clear_system_state();
   return 0;
 }
 
diff --git a/vp9/encoder/vp9_extend.c b/vp9/encoder/vp9_extend.c
index 6e1ed36..0c304dc 100644
--- a/vp9/encoder/vp9_extend.c
+++ b/vp9/encoder/vp9_extend.c
@@ -74,7 +74,7 @@
 
   for (i = 0; i < h; i++) {
     vpx_memset16(dst_ptr1, src_ptr1[0], extend_left);
-    memcpy(dst_ptr1 + extend_left, src_ptr1, w * sizeof(uint16_t));
+    memcpy(dst_ptr1 + extend_left, src_ptr1, w * sizeof(src_ptr1[0]));
     vpx_memset16(dst_ptr2, src_ptr2[0], extend_right);
     src_ptr1 += src_pitch;
     src_ptr2 += src_pitch;
@@ -91,12 +91,12 @@
   linesize = extend_left + extend_right + w;
 
   for (i = 0; i < extend_top; i++) {
-    memcpy(dst_ptr1, src_ptr1, linesize * sizeof(uint16_t));
+    memcpy(dst_ptr1, src_ptr1, linesize * sizeof(src_ptr1[0]));
     dst_ptr1 += dst_pitch;
   }
 
   for (i = 0; i < extend_bottom; i++) {
-    memcpy(dst_ptr2, src_ptr2, linesize * sizeof(uint16_t));
+    memcpy(dst_ptr2, src_ptr2, linesize * sizeof(src_ptr2[0]));
     dst_ptr2 += dst_pitch;
   }
 }
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 9f38736..e690a7c 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -3394,7 +3394,7 @@
         if (!x->select_tx_size)
           swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
         memcpy(ctx->zcoeff_blk, x->zcoeff_blk[mbmi->tx_size],
-               sizeof(uint8_t) * ctx->num_4x4_blk);
+               sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
 
         // TODO(debargha): enhance this test with a better distortion prediction
         // based on qp, activity mask and history
@@ -4135,7 +4135,7 @@
         if (!x->select_tx_size)
           swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
         memcpy(ctx->zcoeff_blk, x->zcoeff_blk[TX_4X4],
-               sizeof(uint8_t) * ctx->num_4x4_blk);
+               sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
 
         for (i = 0; i < 4; i++)
           best_bmodes[i] = xd->mi[0]->bmi[i];
diff --git a/vp9/encoder/vp9_resize.c b/vp9/encoder/vp9_resize.c
index e5ae959..59c7478 100644
--- a/vp9/encoder/vp9_resize.c
+++ b/vp9/encoder/vp9_resize.c
@@ -448,7 +448,7 @@
                              uint8_t *buf) {
   int steps;
   if (length == olength) {
-    memcpy(output, input, sizeof(uint8_t) * length);
+    memcpy(output, input, sizeof(output[0]) * length);
     return;
   }
   steps = get_down2_steps(length, olength);
@@ -741,7 +741,7 @@
                                     int bd) {
   int steps;
   if (length == olength) {
-    memcpy(output, input, sizeof(uint16_t) * length);
+    memcpy(output, input, sizeof(output[0]) * length);
     return;
   }
   steps = get_down2_steps(length, olength);
diff --git a/vpx_dsp/x86/convolve.h b/vpx_dsp/x86/convolve.h
index d883463..b6fbfcf 100644
--- a/vpx_dsp/x86/convolve.h
+++ b/vpx_dsp/x86/convolve.h
@@ -32,75 +32,74 @@
                                     const int16_t *filter_y, int y_step_q4, \
                                     int w, int h) { \
   assert(filter[3] != 128); \
-  if (step_q4 == 16 && filter[3] != 128) { \
-    if (filter[0] || filter[1] || filter[2]) { \
-      while (w >= 16) { \
-        vpx_filter_block1d16_##dir##8_##avg##opt(src_start, \
-                                                 src_stride, \
-                                                 dst, \
-                                                 dst_stride, \
-                                                 h, \
-                                                 filter); \
-        src += 16; \
-        dst += 16; \
-        w -= 16; \
-      } \
-      while (w >= 8) { \
-        vpx_filter_block1d8_##dir##8_##avg##opt(src_start, \
-                                                src_stride, \
-                                                dst, \
-                                                dst_stride, \
-                                                h, \
-                                                filter); \
-        src += 8; \
-        dst += 8; \
-        w -= 8; \
-      } \
-      while (w >= 4) { \
-        vpx_filter_block1d4_##dir##8_##avg##opt(src_start, \
-                                                src_stride, \
-                                                dst, \
-                                                dst_stride, \
-                                                h, \
-                                                filter); \
-        src += 4; \
-        dst += 4; \
-        w -= 4; \
-      } \
-    } else { \
-      while (w >= 16) { \
-        vpx_filter_block1d16_##dir##2_##avg##opt(src, \
-                                                 src_stride, \
-                                                 dst, \
-                                                 dst_stride, \
-                                                 h, \
-                                                 filter); \
-        src += 16; \
-        dst += 16; \
-        w -= 16; \
-      } \
-      while (w >= 8) { \
-        vpx_filter_block1d8_##dir##2_##avg##opt(src, \
-                                                src_stride, \
-                                                dst, \
-                                                dst_stride, \
-                                                h, \
-                                                filter); \
-        src += 8; \
-        dst += 8; \
-        w -= 8; \
-      } \
-      while (w >= 4) { \
-        vpx_filter_block1d4_##dir##2_##avg##opt(src, \
-                                                src_stride, \
-                                                dst, \
-                                                dst_stride, \
-                                                h, \
-                                                filter); \
-        src += 4; \
-        dst += 4; \
-        w -= 4; \
-      } \
+  assert(step_q4 == 16); \
+  if (filter[0] || filter[1] || filter[2]) { \
+    while (w >= 16) { \
+      vpx_filter_block1d16_##dir##8_##avg##opt(src_start, \
+                                               src_stride, \
+                                               dst, \
+                                               dst_stride, \
+                                               h, \
+                                               filter); \
+      src += 16; \
+      dst += 16; \
+      w -= 16; \
+    } \
+    while (w >= 8) { \
+      vpx_filter_block1d8_##dir##8_##avg##opt(src_start, \
+                                              src_stride, \
+                                              dst, \
+                                              dst_stride, \
+                                              h, \
+                                              filter); \
+      src += 8; \
+      dst += 8; \
+      w -= 8; \
+    } \
+    while (w >= 4) { \
+      vpx_filter_block1d4_##dir##8_##avg##opt(src_start, \
+                                              src_stride, \
+                                              dst, \
+                                              dst_stride, \
+                                              h, \
+                                              filter); \
+      src += 4; \
+      dst += 4; \
+      w -= 4; \
+    } \
+  } else { \
+    while (w >= 16) { \
+      vpx_filter_block1d16_##dir##2_##avg##opt(src, \
+                                               src_stride, \
+                                               dst, \
+                                               dst_stride, \
+                                               h, \
+                                               filter); \
+      src += 16; \
+      dst += 16; \
+      w -= 16; \
+    } \
+    while (w >= 8) { \
+      vpx_filter_block1d8_##dir##2_##avg##opt(src, \
+                                              src_stride, \
+                                              dst, \
+                                              dst_stride, \
+                                              h, \
+                                              filter); \
+      src += 8; \
+      dst += 8; \
+      w -= 8; \
+    } \
+    while (w >= 4) { \
+      vpx_filter_block1d4_##dir##2_##avg##opt(src, \
+                                              src_stride, \
+                                              dst, \
+                                              dst_stride, \
+                                              h, \
+                                              filter); \
+      src += 4; \
+      dst += 4; \
+      w -= 4; \
     } \
   } \
 }
@@ -115,25 +114,25 @@
   assert(filter_y[3] != 128); \
   assert(w <= 64); \
   assert(h <= 64); \
-  if (x_step_q4 == 16 && y_step_q4 == 16) { \
-    if (filter_x[0] || filter_x[1] || filter_x[2] || filter_x[3] == 128 || \
-        filter_y[0] || filter_y[1] || filter_y[2] || filter_y[3] == 128) { \
-      DECLARE_ALIGNED(16, uint8_t, fdata2[64 * 71]); \
-      vpx_convolve8_horiz_##opt(src - 3 * src_stride, src_stride, fdata2, 64, \
-                                filter_x, x_step_q4, filter_y, y_step_q4, \
-                                w, h + 7); \
-      vpx_convolve8_##avg##vert_##opt(fdata2 + 3 * 64, 64, dst, dst_stride, \
-                                      filter_x, x_step_q4, filter_y, \
-                                      y_step_q4, w, h); \
-    } else { \
-      DECLARE_ALIGNED(16, uint8_t, fdata2[64 * 65]); \
-      vpx_convolve8_horiz_##opt(src, src_stride, fdata2, 64, \
-                                filter_x, x_step_q4, filter_y, y_step_q4, \
-                                w, h + 1); \
-      vpx_convolve8_##avg##vert_##opt(fdata2, 64, dst, dst_stride, \
-                                      filter_x, x_step_q4, filter_y, \
-                                      y_step_q4, w, h); \
-    } \
+  assert(x_step_q4 == 16); \
+  assert(y_step_q4 == 16); \
+  if (filter_x[0] || filter_x[1] || filter_x[2]|| \
+      filter_y[0] || filter_y[1] || filter_y[2]) { \
+    DECLARE_ALIGNED(16, uint8_t, fdata2[64 * 71]); \
+    vpx_convolve8_horiz_##opt(src - 3 * src_stride, src_stride, fdata2, 64, \
+                              filter_x, x_step_q4, filter_y, y_step_q4, \
+                              w, h + 7); \
+    vpx_convolve8_##avg##vert_##opt(fdata2 + 3 * 64, 64, dst, dst_stride, \
+                                    filter_x, x_step_q4, filter_y, \
+                                    y_step_q4, w, h); \
+  } else { \
+    DECLARE_ALIGNED(16, uint8_t, fdata2[64 * 65]); \
+    vpx_convolve8_horiz_##opt(src, src_stride, fdata2, 64, \
+                              filter_x, x_step_q4, filter_y, y_step_q4, \
+                              w, h + 1); \
+    vpx_convolve8_##avg##vert_##opt(fdata2, 64, dst, dst_stride, \
+                                    filter_x, x_step_q4, filter_y, \
+                                    y_step_q4, w, h); \
   } \
 }