diff --git a/av1/encoder/arm/neon/av1_fwd_txfm2d_neon.c b/av1/encoder/arm/neon/av1_fwd_txfm2d_neon.c
index 249b3a6..3640cf1 100644
--- a/av1/encoder/arm/neon/av1_fwd_txfm2d_neon.c
+++ b/av1/encoder/arm/neon/av1_fwd_txfm2d_neon.c
@@ -2317,10 +2317,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   if (ud_flip) {
     load_buffer_16bit_to_16bit_w4_flip(input, stride, buf0, height);
   } else {
@@ -2360,10 +2359,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   if (ud_flip) {
     load_buffer_16bit_to_16bit_w4_flip(input, stride, buf0, height);
   } else {
@@ -2401,10 +2399,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   if (ud_flip) {
     load_buffer_16bit_to_16bit_w4_flip(input, stride, buf0, height);
   } else {
@@ -2446,10 +2443,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   if (ud_flip)
     load_buffer_16bit_to_16bit_flip(input, stride, buf0, height);
   else
@@ -2486,10 +2482,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   if (ud_flip)
     load_buffer_16bit_to_16bit_flip(input, stride, buf0, height);
   else
@@ -2526,10 +2521,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   if (ud_flip) {
     load_buffer_16bit_to_16bit_flip(input, stride, buf0, height);
   } else {
@@ -2571,10 +2565,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   if (ud_flip) {
     load_buffer_16bit_to_16bit_flip(input, stride, buf0, height);
   } else {
@@ -2619,10 +2612,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   for (int i = 0; i < 2; i++) {
     if (ud_flip) {
       load_buffer_16bit_to_16bit_flip(input + 8 * i, stride, buf0, height);
@@ -2663,10 +2655,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
   for (int i = 0; i < 2; i++) {
     if (ud_flip) {
       load_buffer_16bit_to_16bit_flip(input + 8 * i, stride, buf0, height);
@@ -2706,10 +2697,9 @@
   int ud_flip, lr_flip;
 
   get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-  const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-  const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-  const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-  const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+  const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+  const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+  const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
 
   for (int i = 0; i < 2; i++) {
     if (ud_flip) {
@@ -2755,10 +2745,9 @@
   if (col_txfm != NULL && row_txfm != NULL) {
     int ud_flip, lr_flip;
     get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-    const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-    const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-    const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-    const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+    const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+    const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+    const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
 
     for (int i = 0; i < 2; i++) {
       if (ud_flip) {
@@ -2809,10 +2798,9 @@
   if (col_txfm != NULL && row_txfm != NULL) {
     int ud_flip, lr_flip;
     get_flip_cfg(tx_type, &ud_flip, &lr_flip);
-    const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-    const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-    const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-    const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+    const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+    const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+    const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
 
     for (int i = 0; i < 4; i++) {
       if (ud_flip) {
@@ -2858,10 +2846,9 @@
   const transform_1d_lbd_neon row_txfm = row_txfm8x32_arr[tx_type];
 
   if (col_txfm != NULL && row_txfm != NULL) {
-    const int16x4_t v_shifts = vget_low_s16(vmovl_s8(vld1_s8(&shift[0])));
-    const int16x8_t v_shift0 = vdupq_lane_s16(v_shifts, 0);
-    const int16x8_t v_shift1 = vdupq_lane_s16(v_shifts, 1);
-    const int16x8_t v_shift2 = vdupq_lane_s16(v_shifts, 2);
+    const int16x8_t v_shift0 = vdupq_n_s16(shift[0]);
+    const int16x8_t v_shift1 = vdupq_n_s16(shift[1]);
+    const int16x8_t v_shift2 = vdupq_n_s16(shift[2]);
     int ud_flip, lr_flip;
     get_flip_cfg(tx_type, &ud_flip, &lr_flip);