diff --git a/test/vp10_fwd_txfm1d_test.cc b/test/vp10_fwd_txfm1d_test.cc
index a39e0ef..bcbc617 100644
--- a/test/vp10_fwd_txfm1d_test.cc
+++ b/test/vp10_fwd_txfm1d_test.cc
@@ -8,8 +8,8 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
-#include "test/vp10_txfm_test.h"
 #include "vp10/common/vp10_fwd_txfm1d.h"
+#include "test/vp10_txfm_test.h"
 
 using libvpx_test::ACMRandom;
 
@@ -17,12 +17,14 @@
 static int txfm_type_num = 2;
 static TYPE_TXFM txfm_type_ls[2] = {TYPE_DCT, TYPE_ADST};
 
-static int txfm_size_num = 4;
-static int txfm_size_ls[4] = {4, 8, 16, 32};
+static int txfm_size_num = 5;
+static int txfm_size_ls[5] = {4, 8, 16, 32, 64};
 
-static TxfmFunc fwd_txfm_func_ls[2][4] = {
-    {vp10_fdct4_new, vp10_fdct8_new, vp10_fdct16_new, vp10_fdct32_new},
-    {vp10_fadst4_new, vp10_fadst8_new, vp10_fadst16_new, vp10_fadst32_new}};
+static TxfmFunc fwd_txfm_func_ls[2][5] = {
+    {vp10_fdct4_new, vp10_fdct8_new, vp10_fdct16_new, vp10_fdct32_new,
+     vp10_fdct64_new},
+    {vp10_fadst4_new, vp10_fadst8_new, vp10_fadst16_new, vp10_fadst32_new,
+     NULL}};
 
 // the maximum stage number of fwd/inv 1d dct/adst txfm is 12
 static int8_t cos_bit[12] = {14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14};
@@ -104,19 +106,21 @@
       int max_error = 7;
 
       const int count_test_block = 5000;
-      for (int ti = 0; ti < count_test_block; ++ti) {
-        for (int ni = 0; ni < txfm_size; ++ni) {
-          input[ni] = rnd.Rand16() % base - rnd.Rand16() % base;
-          ref_input[ni] = static_cast<double>(input[ni]);
-        }
+      if (fwd_txfm_func != NULL) {
+        for (int ti = 0; ti < count_test_block; ++ti) {
+          for (int ni = 0; ni < txfm_size; ++ni) {
+            input[ni] = rnd.Rand16() % base - rnd.Rand16() % base;
+            ref_input[ni] = static_cast<double>(input[ni]);
+          }
 
-        fwd_txfm_func(input, output, cos_bit, range_bit);
-        reference_hybrid_1d(ref_input, ref_output, txfm_size, txfm_type);
+          fwd_txfm_func(input, output, cos_bit, range_bit);
+          reference_hybrid_1d(ref_input, ref_output, txfm_size, txfm_type);
 
-        for (int ni = 0; ni < txfm_size; ++ni) {
-          EXPECT_LE(
-              abs(output[ni] - static_cast<int32_t>(round(ref_output[ni]))),
-              max_error);
+          for (int ni = 0; ni < txfm_size; ++ni) {
+            EXPECT_LE(
+                abs(output[ni] - static_cast<int32_t>(round(ref_output[ni]))),
+                max_error);
+          }
         }
       }
     }
diff --git a/test/vp10_fwd_txfm2d_test.cc b/test/vp10_fwd_txfm2d_test.cc
index e6416cc..d4115c9 100644
--- a/test/vp10_fwd_txfm2d_test.cc
+++ b/test/vp10_fwd_txfm2d_test.cc
@@ -8,9 +8,9 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
+#include <math.h>
 #include <stdio.h>
 #include <stdlib.h>
-#include <math.h>
 
 #include "third_party/googletest/src/include/gtest/gtest.h"
 
@@ -23,21 +23,22 @@
 
 namespace {
 
-const int txfm_size_num = 4;
-const int txfm_size_ls[4] = {4, 8, 16, 32};
-const TXFM_2D_CFG fwd_txfm_cfg_ls[4][4] = {
-    {fwd_txfm_2d_cfg_dct_dct_4, fwd_txfm_2d_cfg_dct_adst_4,
-     fwd_txfm_2d_cfg_adst_adst_4, fwd_txfm_2d_cfg_adst_dct_4},
-    {fwd_txfm_2d_cfg_dct_dct_8, fwd_txfm_2d_cfg_dct_adst_8,
-     fwd_txfm_2d_cfg_adst_adst_8, fwd_txfm_2d_cfg_adst_dct_8},
-    {fwd_txfm_2d_cfg_dct_dct_16, fwd_txfm_2d_cfg_dct_adst_16,
-     fwd_txfm_2d_cfg_adst_adst_16, fwd_txfm_2d_cfg_adst_dct_16},
-    {fwd_txfm_2d_cfg_dct_dct_32, fwd_txfm_2d_cfg_dct_adst_32,
-     fwd_txfm_2d_cfg_adst_adst_32, fwd_txfm_2d_cfg_adst_dct_32}};
+const int txfm_size_num = 5;
+const int txfm_size_ls[5] = {4, 8, 16, 32, 64};
+const TXFM_2D_CFG* fwd_txfm_cfg_ls[5][4] = {
+    {&fwd_txfm_2d_cfg_dct_dct_4, &fwd_txfm_2d_cfg_dct_adst_4,
+     &fwd_txfm_2d_cfg_adst_adst_4, &fwd_txfm_2d_cfg_adst_dct_4},
+    {&fwd_txfm_2d_cfg_dct_dct_8, &fwd_txfm_2d_cfg_dct_adst_8,
+     &fwd_txfm_2d_cfg_adst_adst_8, &fwd_txfm_2d_cfg_adst_dct_8},
+    {&fwd_txfm_2d_cfg_dct_dct_16, &fwd_txfm_2d_cfg_dct_adst_16,
+     &fwd_txfm_2d_cfg_adst_adst_16, &fwd_txfm_2d_cfg_adst_dct_16},
+    {&fwd_txfm_2d_cfg_dct_dct_32, &fwd_txfm_2d_cfg_dct_adst_32,
+     &fwd_txfm_2d_cfg_adst_adst_32, &fwd_txfm_2d_cfg_adst_dct_32},
+    {&fwd_txfm_2d_cfg_dct_dct_64, NULL, NULL, NULL}};
 
-const Fwd_Txfm2d_Func fwd_txfm_func_ls[4] = {
+const Fwd_Txfm2d_Func fwd_txfm_func_ls[5] = {
     vp10_fwd_txfm2d_4x4, vp10_fwd_txfm2d_8x8, vp10_fwd_txfm2d_16x16,
-    vp10_fwd_txfm2d_32x32};
+    vp10_fwd_txfm2d_32x32, vp10_fwd_txfm2d_64x64};
 
 const int txfm_type_num = 4;
 const TYPE_TXFM type_ls_0[4] = {TYPE_DCT, TYPE_DCT, TYPE_ADST, TYPE_ADST};
@@ -54,44 +55,48 @@
 
     for (int txfm_type_idx = 0; txfm_type_idx < txfm_type_num;
          ++txfm_type_idx) {
-      TXFM_2D_CFG fwd_txfm_cfg = fwd_txfm_cfg_ls[txfm_size_idx][txfm_type_idx];
-      Fwd_Txfm2d_Func fwd_txfm_func = fwd_txfm_func_ls[txfm_size_idx];
-      TYPE_TXFM type0 = type_ls_0[txfm_type_idx];
-      TYPE_TXFM type1 = type_ls_1[txfm_type_idx];
-      int amplify_bit =
-          fwd_txfm_cfg.shift[0] + fwd_txfm_cfg.shift[1] + fwd_txfm_cfg.shift[2];
-      double amplify_factor =
-          amplify_bit >= 0 ? (1 << amplify_bit) : (1.0 / (1 << -amplify_bit));
+      const TXFM_2D_CFG* fwd_txfm_cfg =
+          fwd_txfm_cfg_ls[txfm_size_idx][txfm_type_idx];
+      if (fwd_txfm_cfg != NULL) {
+        Fwd_Txfm2d_Func fwd_txfm_func = fwd_txfm_func_ls[txfm_size_idx];
+        TYPE_TXFM type0 = type_ls_0[txfm_type_idx];
+        TYPE_TXFM type1 = type_ls_1[txfm_type_idx];
+        int amplify_bit = fwd_txfm_cfg->shift[0] + fwd_txfm_cfg->shift[1] +
+                          fwd_txfm_cfg->shift[2];
+        double amplify_factor =
+            amplify_bit >= 0 ? (1 << amplify_bit) : (1.0 / (1 << -amplify_bit));
 
-      ACMRandom rnd(ACMRandom::DeterministicSeed());
-      int count = 5000;
-      double avg_abs_error = 0;
-      for (int ci = 0; ci < count; ci++) {
-        for (int ni = 0; ni < sqr_txfm_size; ++ni) {
-          input[ni] = rnd.Rand16() % base;
-          ref_input[ni] = static_cast<double>(input[ni]);
-          output[ni] = 0;
-          ref_output[ni] = 0;
+        ACMRandom rnd(ACMRandom::DeterministicSeed());
+        int count = 500;
+        double avg_abs_error = 0;
+        for (int ci = 0; ci < count; ci++) {
+          for (int ni = 0; ni < sqr_txfm_size; ++ni) {
+            input[ni] = rnd.Rand16() % base;
+            ref_input[ni] = static_cast<double>(input[ni]);
+            output[ni] = 0;
+            ref_output[ni] = 0;
+          }
+
+          fwd_txfm_func(input, output, txfm_size, fwd_txfm_cfg, bd);
+          reference_hybrid_2d(ref_input, ref_output, txfm_size, type0, type1);
+
+          for (int ni = 0; ni < sqr_txfm_size; ++ni) {
+            ref_output[ni] = round(ref_output[ni] * amplify_factor);
+            EXPECT_LE(fabs(output[ni] - ref_output[ni]) / amplify_factor, 60);
+          }
+          avg_abs_error += compute_avg_abs_error<int32_t, double>(
+              output, ref_output, sqr_txfm_size);
         }
 
-        fwd_txfm_func(input, output, txfm_size, &fwd_txfm_cfg, bd);
-        reference_hybrid_2d(ref_input, ref_output, txfm_size, type0, type1);
-
-        for (int ni = 0; ni < sqr_txfm_size; ++ni) {
-          ref_output[ni] = round(ref_output[ni] * amplify_factor);
-          EXPECT_LE(fabs(output[ni] - ref_output[ni]) / amplify_factor, 30);
-        }
-        avg_abs_error += compute_avg_abs_error<int32_t, double>(
-            output, ref_output, sqr_txfm_size);
+        avg_abs_error /= amplify_factor;
+        avg_abs_error /= count;
+        // max_abs_avg_error comes from upper bound of avg_abs_error
+        // printf("type0: %d type1: %d txfm_size: %d accuracy_avg_abs_error:
+        // %f\n",
+        // type0, type1, txfm_size, avg_abs_error);
+        double max_abs_avg_error = 5;
+        EXPECT_LE(avg_abs_error, max_abs_avg_error);
       }
-
-      avg_abs_error /= amplify_factor;
-      avg_abs_error /= count;
-      // max_abs_avg_error comes from upper bound of avg_abs_error
-      // printf("type0: %d type1: %d txfm_size: %d accuracy_avg_abs_error:
-      // %f\n", type0, type1, txfm_size, avg_abs_error);
-      double max_abs_avg_error = 1.5;
-      EXPECT_LE(avg_abs_error, max_abs_avg_error);
     }
 
     delete[] input;
diff --git a/test/vp10_inv_txfm1d_test.cc b/test/vp10_inv_txfm1d_test.cc
index 3b716c8..2e9e58d 100644
--- a/test/vp10_inv_txfm1d_test.cc
+++ b/test/vp10_inv_txfm1d_test.cc
@@ -16,16 +16,20 @@
 
 namespace {
 static int txfm_type_num = 2;
-static int txfm_size_num = 4;
-static int txfm_size_ls[4] = {4, 8, 16, 32};
+static int txfm_size_num = 5;
+static int txfm_size_ls[5] = {4, 8, 16, 32, 64};
 
-static TxfmFunc fwd_txfm_func_ls[2][4] = {
-    {vp10_fdct4_new, vp10_fdct8_new, vp10_fdct16_new, vp10_fdct32_new},
-    {vp10_fadst4_new, vp10_fadst8_new, vp10_fadst16_new, vp10_fadst32_new}};
+static TxfmFunc fwd_txfm_func_ls[2][5] = {
+    {vp10_fdct4_new, vp10_fdct8_new, vp10_fdct16_new, vp10_fdct32_new,
+     vp10_fdct64_new},
+    {vp10_fadst4_new, vp10_fadst8_new, vp10_fadst16_new, vp10_fadst32_new,
+     NULL}};
 
-static TxfmFunc inv_txfm_func_ls[2][4] = {
-    {vp10_idct4_new, vp10_idct8_new, vp10_idct16_new, vp10_idct32_new},
-    {vp10_iadst4_new, vp10_iadst8_new, vp10_iadst16_new, vp10_iadst32_new}};
+static TxfmFunc inv_txfm_func_ls[2][5] = {
+    {vp10_idct4_new, vp10_idct8_new, vp10_idct16_new, vp10_idct32_new,
+     vp10_idct64_new},
+    {vp10_iadst4_new, vp10_iadst8_new, vp10_iadst16_new, vp10_iadst32_new,
+     NULL}};
 
 // the maximum stage number of fwd/inv 1d dct/adst txfm is 12
 static int8_t cos_bit[12] = {14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14};
@@ -44,19 +48,22 @@
       TxfmFunc inv_txfm_func = inv_txfm_func_ls[ti][si];
       int max_error = 2;
 
-      const int count_test_block = 5000;
-      for (int ci = 0; ci < count_test_block; ++ci) {
-        for (int ni = 0; ni < txfm_size; ++ni) {
-          input[ni] = rnd.Rand16() % base - rnd.Rand16() % base;
-        }
+      if (fwd_txfm_func != NULL) {
+        const int count_test_block = 5000;
+        for (int ci = 0; ci < count_test_block; ++ci) {
+          for (int ni = 0; ni < txfm_size; ++ni) {
+            input[ni] = rnd.Rand16() % base - rnd.Rand16() % base;
+          }
 
-        fwd_txfm_func(input, output, cos_bit, range_bit);
-        inv_txfm_func(output, round_trip_output, cos_bit, range_bit);
+          fwd_txfm_func(input, output, cos_bit, range_bit);
+          inv_txfm_func(output, round_trip_output, cos_bit, range_bit);
 
-        for (int ni = 0; ni < txfm_size; ++ni) {
-          EXPECT_LE(abs(input[ni] - round_shift(round_trip_output[ni],
-                                                get_max_bit(txfm_size) - 1)),
-                    max_error);
+          for (int ni = 0; ni < txfm_size; ++ni) {
+            int node_err =
+                abs(input[ni] - round_shift(round_trip_output[ni],
+                                            get_max_bit(txfm_size) - 1));
+            EXPECT_LE(node_err, max_error);
+          }
         }
       }
     }
diff --git a/test/vp10_inv_txfm2d_test.cc b/test/vp10_inv_txfm2d_test.cc
index 603821e..b0e6af5 100644
--- a/test/vp10_inv_txfm2d_test.cc
+++ b/test/vp10_inv_txfm2d_test.cc
@@ -25,34 +25,36 @@
 
 namespace {
 
-const int txfm_size_num = 4;
-const int txfm_size_ls[4] = {4, 8, 16, 32};
-const TXFM_2D_CFG fwd_txfm_cfg_ls[4][4] = {
-    {fwd_txfm_2d_cfg_dct_dct_4, fwd_txfm_2d_cfg_dct_adst_4,
-     fwd_txfm_2d_cfg_adst_adst_4, fwd_txfm_2d_cfg_adst_dct_4},
-    {fwd_txfm_2d_cfg_dct_dct_8, fwd_txfm_2d_cfg_dct_adst_8,
-     fwd_txfm_2d_cfg_adst_adst_8, fwd_txfm_2d_cfg_adst_dct_8},
-    {fwd_txfm_2d_cfg_dct_dct_16, fwd_txfm_2d_cfg_dct_adst_16,
-     fwd_txfm_2d_cfg_adst_adst_16, fwd_txfm_2d_cfg_adst_dct_16},
-    {fwd_txfm_2d_cfg_dct_dct_32, fwd_txfm_2d_cfg_dct_adst_32,
-     fwd_txfm_2d_cfg_adst_adst_32, fwd_txfm_2d_cfg_adst_dct_32}};
+const int txfm_size_num = 5;
+const int txfm_size_ls[5] = {4, 8, 16, 32, 64};
+const TXFM_2D_CFG* fwd_txfm_cfg_ls[5][4] = {
+    {&fwd_txfm_2d_cfg_dct_dct_4, &fwd_txfm_2d_cfg_dct_adst_4,
+     &fwd_txfm_2d_cfg_adst_adst_4, &fwd_txfm_2d_cfg_adst_dct_4},
+    {&fwd_txfm_2d_cfg_dct_dct_8, &fwd_txfm_2d_cfg_dct_adst_8,
+     &fwd_txfm_2d_cfg_adst_adst_8, &fwd_txfm_2d_cfg_adst_dct_8},
+    {&fwd_txfm_2d_cfg_dct_dct_16, &fwd_txfm_2d_cfg_dct_adst_16,
+     &fwd_txfm_2d_cfg_adst_adst_16, &fwd_txfm_2d_cfg_adst_dct_16},
+    {&fwd_txfm_2d_cfg_dct_dct_32, &fwd_txfm_2d_cfg_dct_adst_32,
+     &fwd_txfm_2d_cfg_adst_adst_32, &fwd_txfm_2d_cfg_adst_dct_32},
+    {&fwd_txfm_2d_cfg_dct_dct_64, NULL, NULL, NULL}};
 
-const TXFM_2D_CFG inv_txfm_cfg_ls[4][4] = {
-    {inv_txfm_2d_cfg_dct_dct_4, inv_txfm_2d_cfg_dct_adst_4,
-     inv_txfm_2d_cfg_adst_adst_4, inv_txfm_2d_cfg_adst_dct_4},
-    {inv_txfm_2d_cfg_dct_dct_8, inv_txfm_2d_cfg_dct_adst_8,
-     inv_txfm_2d_cfg_adst_adst_8, inv_txfm_2d_cfg_adst_dct_8},
-    {inv_txfm_2d_cfg_dct_dct_16, inv_txfm_2d_cfg_dct_adst_16,
-     inv_txfm_2d_cfg_adst_adst_16, inv_txfm_2d_cfg_adst_dct_16},
-    {inv_txfm_2d_cfg_dct_dct_32, inv_txfm_2d_cfg_dct_adst_32,
-     inv_txfm_2d_cfg_adst_adst_32, inv_txfm_2d_cfg_adst_dct_32}};
+const TXFM_2D_CFG* inv_txfm_cfg_ls[5][4] = {
+    {&inv_txfm_2d_cfg_dct_dct_4, &inv_txfm_2d_cfg_dct_adst_4,
+     &inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_dct_4},
+    {&inv_txfm_2d_cfg_dct_dct_8, &inv_txfm_2d_cfg_dct_adst_8,
+     &inv_txfm_2d_cfg_adst_adst_8, &inv_txfm_2d_cfg_adst_dct_8},
+    {&inv_txfm_2d_cfg_dct_dct_16, &inv_txfm_2d_cfg_dct_adst_16,
+     &inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_dct_16},
+    {&inv_txfm_2d_cfg_dct_dct_32, &inv_txfm_2d_cfg_dct_adst_32,
+     &inv_txfm_2d_cfg_adst_adst_32, &inv_txfm_2d_cfg_adst_dct_32},
+    {&inv_txfm_2d_cfg_dct_dct_64, NULL, NULL, NULL}};
 
-const Fwd_Txfm2d_Func fwd_txfm_func_ls[4] = {
+const Fwd_Txfm2d_Func fwd_txfm_func_ls[5] = {
     vp10_fwd_txfm2d_4x4, vp10_fwd_txfm2d_8x8, vp10_fwd_txfm2d_16x16,
-    vp10_fwd_txfm2d_32x32};
-const Inv_Txfm2d_Func inv_txfm_func_ls[4] = {
+    vp10_fwd_txfm2d_32x32, vp10_fwd_txfm2d_64x64};
+const Inv_Txfm2d_Func inv_txfm_func_ls[5] = {
     vp10_inv_txfm2d_add_4x4, vp10_inv_txfm2d_add_8x8, vp10_inv_txfm2d_add_16x16,
-    vp10_inv_txfm2d_add_32x32};
+    vp10_inv_txfm2d_add_32x32, vp10_inv_txfm2d_add_64x64};
 
 const int txfm_type_num = 4;
 
@@ -66,44 +68,46 @@
 
     for (int txfm_type_idx = 0; txfm_type_idx < txfm_type_num;
          ++txfm_type_idx) {
-      const TXFM_2D_CFG fwd_txfm_cfg =
+      const TXFM_2D_CFG* fwd_txfm_cfg =
           fwd_txfm_cfg_ls[txfm_size_idx][txfm_type_idx];
-      const TXFM_2D_CFG inv_txfm_cfg =
+      const TXFM_2D_CFG* inv_txfm_cfg =
           inv_txfm_cfg_ls[txfm_size_idx][txfm_type_idx];
-      const Fwd_Txfm2d_Func fwd_txfm_func = fwd_txfm_func_ls[txfm_size_idx];
-      const Inv_Txfm2d_Func inv_txfm_func = inv_txfm_func_ls[txfm_size_idx];
-      const int count = 5000;
-      double avg_abs_error = 0;
-      ACMRandom rnd(ACMRandom::DeterministicSeed());
-      for (int ci = 0; ci < count; ci++) {
-        for (int ni = 0; ni < sqr_txfm_size; ++ni) {
-          if (ci == 0) {
-            int extreme_input = base - 1;
-            input[ni] = extreme_input;  // extreme case
-            ref_input[ni] = 0;
-          } else {
-            input[ni] = rnd.Rand16() % base;
-            ref_input[ni] = 0;
+      if (fwd_txfm_cfg != NULL) {
+        const Fwd_Txfm2d_Func fwd_txfm_func = fwd_txfm_func_ls[txfm_size_idx];
+        const Inv_Txfm2d_Func inv_txfm_func = inv_txfm_func_ls[txfm_size_idx];
+        const int count = 1000;
+        double avg_abs_error = 0;
+        ACMRandom rnd(ACMRandom::DeterministicSeed());
+        for (int ci = 0; ci < count; ci++) {
+          for (int ni = 0; ni < sqr_txfm_size; ++ni) {
+            if (ci == 0) {
+              int extreme_input = base - 1;
+              input[ni] = extreme_input;  // extreme case
+              ref_input[ni] = 0;
+            } else {
+              input[ni] = rnd.Rand16() % base;
+              ref_input[ni] = 0;
+            }
           }
+
+          fwd_txfm_func(input, output, txfm_size, fwd_txfm_cfg, bd);
+          inv_txfm_func(output, ref_input, txfm_size, inv_txfm_cfg, bd);
+
+          for (int ni = 0; ni < sqr_txfm_size; ++ni) {
+            EXPECT_LE(abs(input[ni] - ref_input[ni]), 2);
+          }
+          avg_abs_error += compute_avg_abs_error<int16_t, uint16_t>(
+              input, ref_input, sqr_txfm_size);
         }
 
-        fwd_txfm_func(input, output, txfm_size, &fwd_txfm_cfg, bd);
-        inv_txfm_func(output, ref_input, txfm_size, &inv_txfm_cfg, bd);
-
-        for (int ni = 0; ni < sqr_txfm_size; ++ni) {
-          EXPECT_LE(abs(input[ni] - ref_input[ni]), 2);
-        }
-        avg_abs_error += compute_avg_abs_error<int16_t, uint16_t>(
-            input, ref_input, sqr_txfm_size);
+        avg_abs_error /= count;
+        // max_abs_avg_error comes from upper bound of
+        // printf("txfm_size: %d accuracy_avg_abs_error: %f\n",
+        // txfm_size, avg_abs_error);
+        // TODO(angiebird): this upper bound is from adst_adst_8
+        const double max_abs_avg_error = 0.024;
+        EXPECT_LE(avg_abs_error, max_abs_avg_error);
       }
-
-      avg_abs_error /= count;
-      // max_abs_avg_error comes from upper bound of
-      // printf("txfm_size: %d accuracy_avg_abs_error: %f\n", txfm_size,
-      // avg_abs_error);
-      // TODO(angiebird): this upper bound is from adst_adst_8
-      const double max_abs_avg_error = 0.024;
-      EXPECT_LE(avg_abs_error, max_abs_avg_error);
     }
 
     delete[] input;
diff --git a/vp10/common/vp10_fwd_txfm1d.c b/vp10/common/vp10_fwd_txfm1d.c
index f3da5c9..ef24362 100644
--- a/vp10/common/vp10_fwd_txfm1d.c
+++ b/vp10/common/vp10_fwd_txfm1d.c
@@ -15,8 +15,8 @@
   {                                                                       \
     int i, j;                                                             \
     for (i = 0; i < size; ++i) {                                          \
-      int buf_bit = get_max_bit(abs(buf[i])) + 1;                         \
-      if (buf_bit > bit) {                                                \
+        int buf_bit = get_max_bit(abs(buf[i])) + 1;                       \
+        if (buf_bit > bit) {                                              \
         printf("======== %s overflow ========\n", __func__);              \
         printf("stage: %d node: %d\n", stage, i);                         \
         printf("bit: %d buf_bit: %d buf[i]: %d\n", bit, buf_bit, buf[i]); \
@@ -32,11 +32,11 @@
 #else
 #define range_check(stage, input, buf, size, bit) \
   {                                               \
-    (void) stage;                                 \
-    (void) input;                                 \
-    (void) buf;                                   \
-    (void) size;                                  \
-    (void) bit;                                   \
+    (void)stage;                                  \
+    (void)input;                                  \
+    (void)buf;                                    \
+    (void)size;                                   \
+    (void)bit;                                    \
   }
 #endif
 
@@ -1092,7 +1092,6 @@
   bf1[14] = bf0[9];
   bf1[15] = -bf0[1];
   range_check(stage, input, bf1, size, stage_range[stage]);
-
 }
 
 void vp10_fadst32_new(const int32_t *input, int32_t *output,
@@ -1529,3 +1528,796 @@
   bf1[31] = -bf0[1];
   range_check(stage, input, bf1, size, stage_range[stage]);
 }
+
+void vp10_fdct64_new(const int32_t *input, int32_t *output,
+                     const int8_t *cos_bit, const int8_t *stage_range) {
+  const int32_t size = 64;
+  const int32_t *cospi;
+
+  int32_t stage = 0;
+  int32_t *bf0, *bf1;
+  int32_t step[64];
+
+  // stage 0;
+  range_check(stage, input, input, size, stage_range[stage]);
+
+  // stage 1;
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf1 = output;
+  bf1[0] = input[0] + input[63];
+  bf1[1] = input[1] + input[62];
+  bf1[2] = input[2] + input[61];
+  bf1[3] = input[3] + input[60];
+  bf1[4] = input[4] + input[59];
+  bf1[5] = input[5] + input[58];
+  bf1[6] = input[6] + input[57];
+  bf1[7] = input[7] + input[56];
+  bf1[8] = input[8] + input[55];
+  bf1[9] = input[9] + input[54];
+  bf1[10] = input[10] + input[53];
+  bf1[11] = input[11] + input[52];
+  bf1[12] = input[12] + input[51];
+  bf1[13] = input[13] + input[50];
+  bf1[14] = input[14] + input[49];
+  bf1[15] = input[15] + input[48];
+  bf1[16] = input[16] + input[47];
+  bf1[17] = input[17] + input[46];
+  bf1[18] = input[18] + input[45];
+  bf1[19] = input[19] + input[44];
+  bf1[20] = input[20] + input[43];
+  bf1[21] = input[21] + input[42];
+  bf1[22] = input[22] + input[41];
+  bf1[23] = input[23] + input[40];
+  bf1[24] = input[24] + input[39];
+  bf1[25] = input[25] + input[38];
+  bf1[26] = input[26] + input[37];
+  bf1[27] = input[27] + input[36];
+  bf1[28] = input[28] + input[35];
+  bf1[29] = input[29] + input[34];
+  bf1[30] = input[30] + input[33];
+  bf1[31] = input[31] + input[32];
+  bf1[32] = -input[32] + input[31];
+  bf1[33] = -input[33] + input[30];
+  bf1[34] = -input[34] + input[29];
+  bf1[35] = -input[35] + input[28];
+  bf1[36] = -input[36] + input[27];
+  bf1[37] = -input[37] + input[26];
+  bf1[38] = -input[38] + input[25];
+  bf1[39] = -input[39] + input[24];
+  bf1[40] = -input[40] + input[23];
+  bf1[41] = -input[41] + input[22];
+  bf1[42] = -input[42] + input[21];
+  bf1[43] = -input[43] + input[20];
+  bf1[44] = -input[44] + input[19];
+  bf1[45] = -input[45] + input[18];
+  bf1[46] = -input[46] + input[17];
+  bf1[47] = -input[47] + input[16];
+  bf1[48] = -input[48] + input[15];
+  bf1[49] = -input[49] + input[14];
+  bf1[50] = -input[50] + input[13];
+  bf1[51] = -input[51] + input[12];
+  bf1[52] = -input[52] + input[11];
+  bf1[53] = -input[53] + input[10];
+  bf1[54] = -input[54] + input[9];
+  bf1[55] = -input[55] + input[8];
+  bf1[56] = -input[56] + input[7];
+  bf1[57] = -input[57] + input[6];
+  bf1[58] = -input[58] + input[5];
+  bf1[59] = -input[59] + input[4];
+  bf1[60] = -input[60] + input[3];
+  bf1[61] = -input[61] + input[2];
+  bf1[62] = -input[62] + input[1];
+  bf1[63] = -input[63] + input[0];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 2
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0] + bf0[31];
+  bf1[1] = bf0[1] + bf0[30];
+  bf1[2] = bf0[2] + bf0[29];
+  bf1[3] = bf0[3] + bf0[28];
+  bf1[4] = bf0[4] + bf0[27];
+  bf1[5] = bf0[5] + bf0[26];
+  bf1[6] = bf0[6] + bf0[25];
+  bf1[7] = bf0[7] + bf0[24];
+  bf1[8] = bf0[8] + bf0[23];
+  bf1[9] = bf0[9] + bf0[22];
+  bf1[10] = bf0[10] + bf0[21];
+  bf1[11] = bf0[11] + bf0[20];
+  bf1[12] = bf0[12] + bf0[19];
+  bf1[13] = bf0[13] + bf0[18];
+  bf1[14] = bf0[14] + bf0[17];
+  bf1[15] = bf0[15] + bf0[16];
+  bf1[16] = -bf0[16] + bf0[15];
+  bf1[17] = -bf0[17] + bf0[14];
+  bf1[18] = -bf0[18] + bf0[13];
+  bf1[19] = -bf0[19] + bf0[12];
+  bf1[20] = -bf0[20] + bf0[11];
+  bf1[21] = -bf0[21] + bf0[10];
+  bf1[22] = -bf0[22] + bf0[9];
+  bf1[23] = -bf0[23] + bf0[8];
+  bf1[24] = -bf0[24] + bf0[7];
+  bf1[25] = -bf0[25] + bf0[6];
+  bf1[26] = -bf0[26] + bf0[5];
+  bf1[27] = -bf0[27] + bf0[4];
+  bf1[28] = -bf0[28] + bf0[3];
+  bf1[29] = -bf0[29] + bf0[2];
+  bf1[30] = -bf0[30] + bf0[1];
+  bf1[31] = -bf0[31] + bf0[0];
+  bf1[32] = bf0[32];
+  bf1[33] = bf0[33];
+  bf1[34] = bf0[34];
+  bf1[35] = bf0[35];
+  bf1[36] = bf0[36];
+  bf1[37] = bf0[37];
+  bf1[38] = bf0[38];
+  bf1[39] = bf0[39];
+  bf1[40] = half_btf(-cospi[32], bf0[40], cospi[32], bf0[55], cos_bit[stage]);
+  bf1[41] = half_btf(-cospi[32], bf0[41], cospi[32], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(-cospi[32], bf0[42], cospi[32], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(-cospi[32], bf0[43], cospi[32], bf0[52], cos_bit[stage]);
+  bf1[44] = half_btf(-cospi[32], bf0[44], cospi[32], bf0[51], cos_bit[stage]);
+  bf1[45] = half_btf(-cospi[32], bf0[45], cospi[32], bf0[50], cos_bit[stage]);
+  bf1[46] = half_btf(-cospi[32], bf0[46], cospi[32], bf0[49], cos_bit[stage]);
+  bf1[47] = half_btf(-cospi[32], bf0[47], cospi[32], bf0[48], cos_bit[stage]);
+  bf1[48] = half_btf(cospi[32], bf0[48], cospi[32], bf0[47], cos_bit[stage]);
+  bf1[49] = half_btf(cospi[32], bf0[49], cospi[32], bf0[46], cos_bit[stage]);
+  bf1[50] = half_btf(cospi[32], bf0[50], cospi[32], bf0[45], cos_bit[stage]);
+  bf1[51] = half_btf(cospi[32], bf0[51], cospi[32], bf0[44], cos_bit[stage]);
+  bf1[52] = half_btf(cospi[32], bf0[52], cospi[32], bf0[43], cos_bit[stage]);
+  bf1[53] = half_btf(cospi[32], bf0[53], cospi[32], bf0[42], cos_bit[stage]);
+  bf1[54] = half_btf(cospi[32], bf0[54], cospi[32], bf0[41], cos_bit[stage]);
+  bf1[55] = half_btf(cospi[32], bf0[55], cospi[32], bf0[40], cos_bit[stage]);
+  bf1[56] = bf0[56];
+  bf1[57] = bf0[57];
+  bf1[58] = bf0[58];
+  bf1[59] = bf0[59];
+  bf1[60] = bf0[60];
+  bf1[61] = bf0[61];
+  bf1[62] = bf0[62];
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 3
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0] + bf0[15];
+  bf1[1] = bf0[1] + bf0[14];
+  bf1[2] = bf0[2] + bf0[13];
+  bf1[3] = bf0[3] + bf0[12];
+  bf1[4] = bf0[4] + bf0[11];
+  bf1[5] = bf0[5] + bf0[10];
+  bf1[6] = bf0[6] + bf0[9];
+  bf1[7] = bf0[7] + bf0[8];
+  bf1[8] = -bf0[8] + bf0[7];
+  bf1[9] = -bf0[9] + bf0[6];
+  bf1[10] = -bf0[10] + bf0[5];
+  bf1[11] = -bf0[11] + bf0[4];
+  bf1[12] = -bf0[12] + bf0[3];
+  bf1[13] = -bf0[13] + bf0[2];
+  bf1[14] = -bf0[14] + bf0[1];
+  bf1[15] = -bf0[15] + bf0[0];
+  bf1[16] = bf0[16];
+  bf1[17] = bf0[17];
+  bf1[18] = bf0[18];
+  bf1[19] = bf0[19];
+  bf1[20] = half_btf(-cospi[32], bf0[20], cospi[32], bf0[27], cos_bit[stage]);
+  bf1[21] = half_btf(-cospi[32], bf0[21], cospi[32], bf0[26], cos_bit[stage]);
+  bf1[22] = half_btf(-cospi[32], bf0[22], cospi[32], bf0[25], cos_bit[stage]);
+  bf1[23] = half_btf(-cospi[32], bf0[23], cospi[32], bf0[24], cos_bit[stage]);
+  bf1[24] = half_btf(cospi[32], bf0[24], cospi[32], bf0[23], cos_bit[stage]);
+  bf1[25] = half_btf(cospi[32], bf0[25], cospi[32], bf0[22], cos_bit[stage]);
+  bf1[26] = half_btf(cospi[32], bf0[26], cospi[32], bf0[21], cos_bit[stage]);
+  bf1[27] = half_btf(cospi[32], bf0[27], cospi[32], bf0[20], cos_bit[stage]);
+  bf1[28] = bf0[28];
+  bf1[29] = bf0[29];
+  bf1[30] = bf0[30];
+  bf1[31] = bf0[31];
+  bf1[32] = bf0[32] + bf0[47];
+  bf1[33] = bf0[33] + bf0[46];
+  bf1[34] = bf0[34] + bf0[45];
+  bf1[35] = bf0[35] + bf0[44];
+  bf1[36] = bf0[36] + bf0[43];
+  bf1[37] = bf0[37] + bf0[42];
+  bf1[38] = bf0[38] + bf0[41];
+  bf1[39] = bf0[39] + bf0[40];
+  bf1[40] = -bf0[40] + bf0[39];
+  bf1[41] = -bf0[41] + bf0[38];
+  bf1[42] = -bf0[42] + bf0[37];
+  bf1[43] = -bf0[43] + bf0[36];
+  bf1[44] = -bf0[44] + bf0[35];
+  bf1[45] = -bf0[45] + bf0[34];
+  bf1[46] = -bf0[46] + bf0[33];
+  bf1[47] = -bf0[47] + bf0[32];
+  bf1[48] = -bf0[48] + bf0[63];
+  bf1[49] = -bf0[49] + bf0[62];
+  bf1[50] = -bf0[50] + bf0[61];
+  bf1[51] = -bf0[51] + bf0[60];
+  bf1[52] = -bf0[52] + bf0[59];
+  bf1[53] = -bf0[53] + bf0[58];
+  bf1[54] = -bf0[54] + bf0[57];
+  bf1[55] = -bf0[55] + bf0[56];
+  bf1[56] = bf0[56] + bf0[55];
+  bf1[57] = bf0[57] + bf0[54];
+  bf1[58] = bf0[58] + bf0[53];
+  bf1[59] = bf0[59] + bf0[52];
+  bf1[60] = bf0[60] + bf0[51];
+  bf1[61] = bf0[61] + bf0[50];
+  bf1[62] = bf0[62] + bf0[49];
+  bf1[63] = bf0[63] + bf0[48];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 4
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0] + bf0[7];
+  bf1[1] = bf0[1] + bf0[6];
+  bf1[2] = bf0[2] + bf0[5];
+  bf1[3] = bf0[3] + bf0[4];
+  bf1[4] = -bf0[4] + bf0[3];
+  bf1[5] = -bf0[5] + bf0[2];
+  bf1[6] = -bf0[6] + bf0[1];
+  bf1[7] = -bf0[7] + bf0[0];
+  bf1[8] = bf0[8];
+  bf1[9] = bf0[9];
+  bf1[10] = half_btf(-cospi[32], bf0[10], cospi[32], bf0[13], cos_bit[stage]);
+  bf1[11] = half_btf(-cospi[32], bf0[11], cospi[32], bf0[12], cos_bit[stage]);
+  bf1[12] = half_btf(cospi[32], bf0[12], cospi[32], bf0[11], cos_bit[stage]);
+  bf1[13] = half_btf(cospi[32], bf0[13], cospi[32], bf0[10], cos_bit[stage]);
+  bf1[14] = bf0[14];
+  bf1[15] = bf0[15];
+  bf1[16] = bf0[16] + bf0[23];
+  bf1[17] = bf0[17] + bf0[22];
+  bf1[18] = bf0[18] + bf0[21];
+  bf1[19] = bf0[19] + bf0[20];
+  bf1[20] = -bf0[20] + bf0[19];
+  bf1[21] = -bf0[21] + bf0[18];
+  bf1[22] = -bf0[22] + bf0[17];
+  bf1[23] = -bf0[23] + bf0[16];
+  bf1[24] = -bf0[24] + bf0[31];
+  bf1[25] = -bf0[25] + bf0[30];
+  bf1[26] = -bf0[26] + bf0[29];
+  bf1[27] = -bf0[27] + bf0[28];
+  bf1[28] = bf0[28] + bf0[27];
+  bf1[29] = bf0[29] + bf0[26];
+  bf1[30] = bf0[30] + bf0[25];
+  bf1[31] = bf0[31] + bf0[24];
+  bf1[32] = bf0[32];
+  bf1[33] = bf0[33];
+  bf1[34] = bf0[34];
+  bf1[35] = bf0[35];
+  bf1[36] = half_btf(-cospi[16], bf0[36], cospi[48], bf0[59], cos_bit[stage]);
+  bf1[37] = half_btf(-cospi[16], bf0[37], cospi[48], bf0[58], cos_bit[stage]);
+  bf1[38] = half_btf(-cospi[16], bf0[38], cospi[48], bf0[57], cos_bit[stage]);
+  bf1[39] = half_btf(-cospi[16], bf0[39], cospi[48], bf0[56], cos_bit[stage]);
+  bf1[40] = half_btf(-cospi[48], bf0[40], -cospi[16], bf0[55], cos_bit[stage]);
+  bf1[41] = half_btf(-cospi[48], bf0[41], -cospi[16], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(-cospi[48], bf0[42], -cospi[16], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(-cospi[48], bf0[43], -cospi[16], bf0[52], cos_bit[stage]);
+  bf1[44] = bf0[44];
+  bf1[45] = bf0[45];
+  bf1[46] = bf0[46];
+  bf1[47] = bf0[47];
+  bf1[48] = bf0[48];
+  bf1[49] = bf0[49];
+  bf1[50] = bf0[50];
+  bf1[51] = bf0[51];
+  bf1[52] = half_btf(cospi[48], bf0[52], -cospi[16], bf0[43], cos_bit[stage]);
+  bf1[53] = half_btf(cospi[48], bf0[53], -cospi[16], bf0[42], cos_bit[stage]);
+  bf1[54] = half_btf(cospi[48], bf0[54], -cospi[16], bf0[41], cos_bit[stage]);
+  bf1[55] = half_btf(cospi[48], bf0[55], -cospi[16], bf0[40], cos_bit[stage]);
+  bf1[56] = half_btf(cospi[16], bf0[56], cospi[48], bf0[39], cos_bit[stage]);
+  bf1[57] = half_btf(cospi[16], bf0[57], cospi[48], bf0[38], cos_bit[stage]);
+  bf1[58] = half_btf(cospi[16], bf0[58], cospi[48], bf0[37], cos_bit[stage]);
+  bf1[59] = half_btf(cospi[16], bf0[59], cospi[48], bf0[36], cos_bit[stage]);
+  bf1[60] = bf0[60];
+  bf1[61] = bf0[61];
+  bf1[62] = bf0[62];
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 5
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0] + bf0[3];
+  bf1[1] = bf0[1] + bf0[2];
+  bf1[2] = -bf0[2] + bf0[1];
+  bf1[3] = -bf0[3] + bf0[0];
+  bf1[4] = bf0[4];
+  bf1[5] = half_btf(-cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
+  bf1[6] = half_btf(cospi[32], bf0[6], cospi[32], bf0[5], cos_bit[stage]);
+  bf1[7] = bf0[7];
+  bf1[8] = bf0[8] + bf0[11];
+  bf1[9] = bf0[9] + bf0[10];
+  bf1[10] = -bf0[10] + bf0[9];
+  bf1[11] = -bf0[11] + bf0[8];
+  bf1[12] = -bf0[12] + bf0[15];
+  bf1[13] = -bf0[13] + bf0[14];
+  bf1[14] = bf0[14] + bf0[13];
+  bf1[15] = bf0[15] + bf0[12];
+  bf1[16] = bf0[16];
+  bf1[17] = bf0[17];
+  bf1[18] = half_btf(-cospi[16], bf0[18], cospi[48], bf0[29], cos_bit[stage]);
+  bf1[19] = half_btf(-cospi[16], bf0[19], cospi[48], bf0[28], cos_bit[stage]);
+  bf1[20] = half_btf(-cospi[48], bf0[20], -cospi[16], bf0[27], cos_bit[stage]);
+  bf1[21] = half_btf(-cospi[48], bf0[21], -cospi[16], bf0[26], cos_bit[stage]);
+  bf1[22] = bf0[22];
+  bf1[23] = bf0[23];
+  bf1[24] = bf0[24];
+  bf1[25] = bf0[25];
+  bf1[26] = half_btf(cospi[48], bf0[26], -cospi[16], bf0[21], cos_bit[stage]);
+  bf1[27] = half_btf(cospi[48], bf0[27], -cospi[16], bf0[20], cos_bit[stage]);
+  bf1[28] = half_btf(cospi[16], bf0[28], cospi[48], bf0[19], cos_bit[stage]);
+  bf1[29] = half_btf(cospi[16], bf0[29], cospi[48], bf0[18], cos_bit[stage]);
+  bf1[30] = bf0[30];
+  bf1[31] = bf0[31];
+  bf1[32] = bf0[32] + bf0[39];
+  bf1[33] = bf0[33] + bf0[38];
+  bf1[34] = bf0[34] + bf0[37];
+  bf1[35] = bf0[35] + bf0[36];
+  bf1[36] = -bf0[36] + bf0[35];
+  bf1[37] = -bf0[37] + bf0[34];
+  bf1[38] = -bf0[38] + bf0[33];
+  bf1[39] = -bf0[39] + bf0[32];
+  bf1[40] = -bf0[40] + bf0[47];
+  bf1[41] = -bf0[41] + bf0[46];
+  bf1[42] = -bf0[42] + bf0[45];
+  bf1[43] = -bf0[43] + bf0[44];
+  bf1[44] = bf0[44] + bf0[43];
+  bf1[45] = bf0[45] + bf0[42];
+  bf1[46] = bf0[46] + bf0[41];
+  bf1[47] = bf0[47] + bf0[40];
+  bf1[48] = bf0[48] + bf0[55];
+  bf1[49] = bf0[49] + bf0[54];
+  bf1[50] = bf0[50] + bf0[53];
+  bf1[51] = bf0[51] + bf0[52];
+  bf1[52] = -bf0[52] + bf0[51];
+  bf1[53] = -bf0[53] + bf0[50];
+  bf1[54] = -bf0[54] + bf0[49];
+  bf1[55] = -bf0[55] + bf0[48];
+  bf1[56] = -bf0[56] + bf0[63];
+  bf1[57] = -bf0[57] + bf0[62];
+  bf1[58] = -bf0[58] + bf0[61];
+  bf1[59] = -bf0[59] + bf0[60];
+  bf1[60] = bf0[60] + bf0[59];
+  bf1[61] = bf0[61] + bf0[58];
+  bf1[62] = bf0[62] + bf0[57];
+  bf1[63] = bf0[63] + bf0[56];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 6
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = half_btf(cospi[32], bf0[0], cospi[32], bf0[1], cos_bit[stage]);
+  bf1[1] = half_btf(-cospi[32], bf0[1], cospi[32], bf0[0], cos_bit[stage]);
+  bf1[2] = half_btf(cospi[48], bf0[2], cospi[16], bf0[3], cos_bit[stage]);
+  bf1[3] = half_btf(cospi[48], bf0[3], -cospi[16], bf0[2], cos_bit[stage]);
+  bf1[4] = bf0[4] + bf0[5];
+  bf1[5] = -bf0[5] + bf0[4];
+  bf1[6] = -bf0[6] + bf0[7];
+  bf1[7] = bf0[7] + bf0[6];
+  bf1[8] = bf0[8];
+  bf1[9] = half_btf(-cospi[16], bf0[9], cospi[48], bf0[14], cos_bit[stage]);
+  bf1[10] = half_btf(-cospi[48], bf0[10], -cospi[16], bf0[13], cos_bit[stage]);
+  bf1[11] = bf0[11];
+  bf1[12] = bf0[12];
+  bf1[13] = half_btf(cospi[48], bf0[13], -cospi[16], bf0[10], cos_bit[stage]);
+  bf1[14] = half_btf(cospi[16], bf0[14], cospi[48], bf0[9], cos_bit[stage]);
+  bf1[15] = bf0[15];
+  bf1[16] = bf0[16] + bf0[19];
+  bf1[17] = bf0[17] + bf0[18];
+  bf1[18] = -bf0[18] + bf0[17];
+  bf1[19] = -bf0[19] + bf0[16];
+  bf1[20] = -bf0[20] + bf0[23];
+  bf1[21] = -bf0[21] + bf0[22];
+  bf1[22] = bf0[22] + bf0[21];
+  bf1[23] = bf0[23] + bf0[20];
+  bf1[24] = bf0[24] + bf0[27];
+  bf1[25] = bf0[25] + bf0[26];
+  bf1[26] = -bf0[26] + bf0[25];
+  bf1[27] = -bf0[27] + bf0[24];
+  bf1[28] = -bf0[28] + bf0[31];
+  bf1[29] = -bf0[29] + bf0[30];
+  bf1[30] = bf0[30] + bf0[29];
+  bf1[31] = bf0[31] + bf0[28];
+  bf1[32] = bf0[32];
+  bf1[33] = bf0[33];
+  bf1[34] = half_btf(-cospi[8], bf0[34], cospi[56], bf0[61], cos_bit[stage]);
+  bf1[35] = half_btf(-cospi[8], bf0[35], cospi[56], bf0[60], cos_bit[stage]);
+  bf1[36] = half_btf(-cospi[56], bf0[36], -cospi[8], bf0[59], cos_bit[stage]);
+  bf1[37] = half_btf(-cospi[56], bf0[37], -cospi[8], bf0[58], cos_bit[stage]);
+  bf1[38] = bf0[38];
+  bf1[39] = bf0[39];
+  bf1[40] = bf0[40];
+  bf1[41] = bf0[41];
+  bf1[42] = half_btf(-cospi[40], bf0[42], cospi[24], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(-cospi[40], bf0[43], cospi[24], bf0[52], cos_bit[stage]);
+  bf1[44] = half_btf(-cospi[24], bf0[44], -cospi[40], bf0[51], cos_bit[stage]);
+  bf1[45] = half_btf(-cospi[24], bf0[45], -cospi[40], bf0[50], cos_bit[stage]);
+  bf1[46] = bf0[46];
+  bf1[47] = bf0[47];
+  bf1[48] = bf0[48];
+  bf1[49] = bf0[49];
+  bf1[50] = half_btf(cospi[24], bf0[50], -cospi[40], bf0[45], cos_bit[stage]);
+  bf1[51] = half_btf(cospi[24], bf0[51], -cospi[40], bf0[44], cos_bit[stage]);
+  bf1[52] = half_btf(cospi[40], bf0[52], cospi[24], bf0[43], cos_bit[stage]);
+  bf1[53] = half_btf(cospi[40], bf0[53], cospi[24], bf0[42], cos_bit[stage]);
+  bf1[54] = bf0[54];
+  bf1[55] = bf0[55];
+  bf1[56] = bf0[56];
+  bf1[57] = bf0[57];
+  bf1[58] = half_btf(cospi[56], bf0[58], -cospi[8], bf0[37], cos_bit[stage]);
+  bf1[59] = half_btf(cospi[56], bf0[59], -cospi[8], bf0[36], cos_bit[stage]);
+  bf1[60] = half_btf(cospi[8], bf0[60], cospi[56], bf0[35], cos_bit[stage]);
+  bf1[61] = half_btf(cospi[8], bf0[61], cospi[56], bf0[34], cos_bit[stage]);
+  bf1[62] = bf0[62];
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 7
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = half_btf(cospi[56], bf0[4], cospi[8], bf0[7], cos_bit[stage]);
+  bf1[5] = half_btf(cospi[24], bf0[5], cospi[40], bf0[6], cos_bit[stage]);
+  bf1[6] = half_btf(cospi[24], bf0[6], -cospi[40], bf0[5], cos_bit[stage]);
+  bf1[7] = half_btf(cospi[56], bf0[7], -cospi[8], bf0[4], cos_bit[stage]);
+  bf1[8] = bf0[8] + bf0[9];
+  bf1[9] = -bf0[9] + bf0[8];
+  bf1[10] = -bf0[10] + bf0[11];
+  bf1[11] = bf0[11] + bf0[10];
+  bf1[12] = bf0[12] + bf0[13];
+  bf1[13] = -bf0[13] + bf0[12];
+  bf1[14] = -bf0[14] + bf0[15];
+  bf1[15] = bf0[15] + bf0[14];
+  bf1[16] = bf0[16];
+  bf1[17] = half_btf(-cospi[8], bf0[17], cospi[56], bf0[30], cos_bit[stage]);
+  bf1[18] = half_btf(-cospi[56], bf0[18], -cospi[8], bf0[29], cos_bit[stage]);
+  bf1[19] = bf0[19];
+  bf1[20] = bf0[20];
+  bf1[21] = half_btf(-cospi[40], bf0[21], cospi[24], bf0[26], cos_bit[stage]);
+  bf1[22] = half_btf(-cospi[24], bf0[22], -cospi[40], bf0[25], cos_bit[stage]);
+  bf1[23] = bf0[23];
+  bf1[24] = bf0[24];
+  bf1[25] = half_btf(cospi[24], bf0[25], -cospi[40], bf0[22], cos_bit[stage]);
+  bf1[26] = half_btf(cospi[40], bf0[26], cospi[24], bf0[21], cos_bit[stage]);
+  bf1[27] = bf0[27];
+  bf1[28] = bf0[28];
+  bf1[29] = half_btf(cospi[56], bf0[29], -cospi[8], bf0[18], cos_bit[stage]);
+  bf1[30] = half_btf(cospi[8], bf0[30], cospi[56], bf0[17], cos_bit[stage]);
+  bf1[31] = bf0[31];
+  bf1[32] = bf0[32] + bf0[35];
+  bf1[33] = bf0[33] + bf0[34];
+  bf1[34] = -bf0[34] + bf0[33];
+  bf1[35] = -bf0[35] + bf0[32];
+  bf1[36] = -bf0[36] + bf0[39];
+  bf1[37] = -bf0[37] + bf0[38];
+  bf1[38] = bf0[38] + bf0[37];
+  bf1[39] = bf0[39] + bf0[36];
+  bf1[40] = bf0[40] + bf0[43];
+  bf1[41] = bf0[41] + bf0[42];
+  bf1[42] = -bf0[42] + bf0[41];
+  bf1[43] = -bf0[43] + bf0[40];
+  bf1[44] = -bf0[44] + bf0[47];
+  bf1[45] = -bf0[45] + bf0[46];
+  bf1[46] = bf0[46] + bf0[45];
+  bf1[47] = bf0[47] + bf0[44];
+  bf1[48] = bf0[48] + bf0[51];
+  bf1[49] = bf0[49] + bf0[50];
+  bf1[50] = -bf0[50] + bf0[49];
+  bf1[51] = -bf0[51] + bf0[48];
+  bf1[52] = -bf0[52] + bf0[55];
+  bf1[53] = -bf0[53] + bf0[54];
+  bf1[54] = bf0[54] + bf0[53];
+  bf1[55] = bf0[55] + bf0[52];
+  bf1[56] = bf0[56] + bf0[59];
+  bf1[57] = bf0[57] + bf0[58];
+  bf1[58] = -bf0[58] + bf0[57];
+  bf1[59] = -bf0[59] + bf0[56];
+  bf1[60] = -bf0[60] + bf0[63];
+  bf1[61] = -bf0[61] + bf0[62];
+  bf1[62] = bf0[62] + bf0[61];
+  bf1[63] = bf0[63] + bf0[60];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 8
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = bf0[4];
+  bf1[5] = bf0[5];
+  bf1[6] = bf0[6];
+  bf1[7] = bf0[7];
+  bf1[8] = half_btf(cospi[60], bf0[8], cospi[4], bf0[15], cos_bit[stage]);
+  bf1[9] = half_btf(cospi[28], bf0[9], cospi[36], bf0[14], cos_bit[stage]);
+  bf1[10] = half_btf(cospi[44], bf0[10], cospi[20], bf0[13], cos_bit[stage]);
+  bf1[11] = half_btf(cospi[12], bf0[11], cospi[52], bf0[12], cos_bit[stage]);
+  bf1[12] = half_btf(cospi[12], bf0[12], -cospi[52], bf0[11], cos_bit[stage]);
+  bf1[13] = half_btf(cospi[44], bf0[13], -cospi[20], bf0[10], cos_bit[stage]);
+  bf1[14] = half_btf(cospi[28], bf0[14], -cospi[36], bf0[9], cos_bit[stage]);
+  bf1[15] = half_btf(cospi[60], bf0[15], -cospi[4], bf0[8], cos_bit[stage]);
+  bf1[16] = bf0[16] + bf0[17];
+  bf1[17] = -bf0[17] + bf0[16];
+  bf1[18] = -bf0[18] + bf0[19];
+  bf1[19] = bf0[19] + bf0[18];
+  bf1[20] = bf0[20] + bf0[21];
+  bf1[21] = -bf0[21] + bf0[20];
+  bf1[22] = -bf0[22] + bf0[23];
+  bf1[23] = bf0[23] + bf0[22];
+  bf1[24] = bf0[24] + bf0[25];
+  bf1[25] = -bf0[25] + bf0[24];
+  bf1[26] = -bf0[26] + bf0[27];
+  bf1[27] = bf0[27] + bf0[26];
+  bf1[28] = bf0[28] + bf0[29];
+  bf1[29] = -bf0[29] + bf0[28];
+  bf1[30] = -bf0[30] + bf0[31];
+  bf1[31] = bf0[31] + bf0[30];
+  bf1[32] = bf0[32];
+  bf1[33] = half_btf(-cospi[4], bf0[33], cospi[60], bf0[62], cos_bit[stage]);
+  bf1[34] = half_btf(-cospi[60], bf0[34], -cospi[4], bf0[61], cos_bit[stage]);
+  bf1[35] = bf0[35];
+  bf1[36] = bf0[36];
+  bf1[37] = half_btf(-cospi[36], bf0[37], cospi[28], bf0[58], cos_bit[stage]);
+  bf1[38] = half_btf(-cospi[28], bf0[38], -cospi[36], bf0[57], cos_bit[stage]);
+  bf1[39] = bf0[39];
+  bf1[40] = bf0[40];
+  bf1[41] = half_btf(-cospi[20], bf0[41], cospi[44], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(-cospi[44], bf0[42], -cospi[20], bf0[53], cos_bit[stage]);
+  bf1[43] = bf0[43];
+  bf1[44] = bf0[44];
+  bf1[45] = half_btf(-cospi[52], bf0[45], cospi[12], bf0[50], cos_bit[stage]);
+  bf1[46] = half_btf(-cospi[12], bf0[46], -cospi[52], bf0[49], cos_bit[stage]);
+  bf1[47] = bf0[47];
+  bf1[48] = bf0[48];
+  bf1[49] = half_btf(cospi[12], bf0[49], -cospi[52], bf0[46], cos_bit[stage]);
+  bf1[50] = half_btf(cospi[52], bf0[50], cospi[12], bf0[45], cos_bit[stage]);
+  bf1[51] = bf0[51];
+  bf1[52] = bf0[52];
+  bf1[53] = half_btf(cospi[44], bf0[53], -cospi[20], bf0[42], cos_bit[stage]);
+  bf1[54] = half_btf(cospi[20], bf0[54], cospi[44], bf0[41], cos_bit[stage]);
+  bf1[55] = bf0[55];
+  bf1[56] = bf0[56];
+  bf1[57] = half_btf(cospi[28], bf0[57], -cospi[36], bf0[38], cos_bit[stage]);
+  bf1[58] = half_btf(cospi[36], bf0[58], cospi[28], bf0[37], cos_bit[stage]);
+  bf1[59] = bf0[59];
+  bf1[60] = bf0[60];
+  bf1[61] = half_btf(cospi[60], bf0[61], -cospi[4], bf0[34], cos_bit[stage]);
+  bf1[62] = half_btf(cospi[4], bf0[62], cospi[60], bf0[33], cos_bit[stage]);
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 9
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = bf0[4];
+  bf1[5] = bf0[5];
+  bf1[6] = bf0[6];
+  bf1[7] = bf0[7];
+  bf1[8] = bf0[8];
+  bf1[9] = bf0[9];
+  bf1[10] = bf0[10];
+  bf1[11] = bf0[11];
+  bf1[12] = bf0[12];
+  bf1[13] = bf0[13];
+  bf1[14] = bf0[14];
+  bf1[15] = bf0[15];
+  bf1[16] = half_btf(cospi[62], bf0[16], cospi[2], bf0[31], cos_bit[stage]);
+  bf1[17] = half_btf(cospi[30], bf0[17], cospi[34], bf0[30], cos_bit[stage]);
+  bf1[18] = half_btf(cospi[46], bf0[18], cospi[18], bf0[29], cos_bit[stage]);
+  bf1[19] = half_btf(cospi[14], bf0[19], cospi[50], bf0[28], cos_bit[stage]);
+  bf1[20] = half_btf(cospi[54], bf0[20], cospi[10], bf0[27], cos_bit[stage]);
+  bf1[21] = half_btf(cospi[22], bf0[21], cospi[42], bf0[26], cos_bit[stage]);
+  bf1[22] = half_btf(cospi[38], bf0[22], cospi[26], bf0[25], cos_bit[stage]);
+  bf1[23] = half_btf(cospi[6], bf0[23], cospi[58], bf0[24], cos_bit[stage]);
+  bf1[24] = half_btf(cospi[6], bf0[24], -cospi[58], bf0[23], cos_bit[stage]);
+  bf1[25] = half_btf(cospi[38], bf0[25], -cospi[26], bf0[22], cos_bit[stage]);
+  bf1[26] = half_btf(cospi[22], bf0[26], -cospi[42], bf0[21], cos_bit[stage]);
+  bf1[27] = half_btf(cospi[54], bf0[27], -cospi[10], bf0[20], cos_bit[stage]);
+  bf1[28] = half_btf(cospi[14], bf0[28], -cospi[50], bf0[19], cos_bit[stage]);
+  bf1[29] = half_btf(cospi[46], bf0[29], -cospi[18], bf0[18], cos_bit[stage]);
+  bf1[30] = half_btf(cospi[30], bf0[30], -cospi[34], bf0[17], cos_bit[stage]);
+  bf1[31] = half_btf(cospi[62], bf0[31], -cospi[2], bf0[16], cos_bit[stage]);
+  bf1[32] = bf0[32] + bf0[33];
+  bf1[33] = -bf0[33] + bf0[32];
+  bf1[34] = -bf0[34] + bf0[35];
+  bf1[35] = bf0[35] + bf0[34];
+  bf1[36] = bf0[36] + bf0[37];
+  bf1[37] = -bf0[37] + bf0[36];
+  bf1[38] = -bf0[38] + bf0[39];
+  bf1[39] = bf0[39] + bf0[38];
+  bf1[40] = bf0[40] + bf0[41];
+  bf1[41] = -bf0[41] + bf0[40];
+  bf1[42] = -bf0[42] + bf0[43];
+  bf1[43] = bf0[43] + bf0[42];
+  bf1[44] = bf0[44] + bf0[45];
+  bf1[45] = -bf0[45] + bf0[44];
+  bf1[46] = -bf0[46] + bf0[47];
+  bf1[47] = bf0[47] + bf0[46];
+  bf1[48] = bf0[48] + bf0[49];
+  bf1[49] = -bf0[49] + bf0[48];
+  bf1[50] = -bf0[50] + bf0[51];
+  bf1[51] = bf0[51] + bf0[50];
+  bf1[52] = bf0[52] + bf0[53];
+  bf1[53] = -bf0[53] + bf0[52];
+  bf1[54] = -bf0[54] + bf0[55];
+  bf1[55] = bf0[55] + bf0[54];
+  bf1[56] = bf0[56] + bf0[57];
+  bf1[57] = -bf0[57] + bf0[56];
+  bf1[58] = -bf0[58] + bf0[59];
+  bf1[59] = bf0[59] + bf0[58];
+  bf1[60] = bf0[60] + bf0[61];
+  bf1[61] = -bf0[61] + bf0[60];
+  bf1[62] = -bf0[62] + bf0[63];
+  bf1[63] = bf0[63] + bf0[62];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 10
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = bf0[4];
+  bf1[5] = bf0[5];
+  bf1[6] = bf0[6];
+  bf1[7] = bf0[7];
+  bf1[8] = bf0[8];
+  bf1[9] = bf0[9];
+  bf1[10] = bf0[10];
+  bf1[11] = bf0[11];
+  bf1[12] = bf0[12];
+  bf1[13] = bf0[13];
+  bf1[14] = bf0[14];
+  bf1[15] = bf0[15];
+  bf1[16] = bf0[16];
+  bf1[17] = bf0[17];
+  bf1[18] = bf0[18];
+  bf1[19] = bf0[19];
+  bf1[20] = bf0[20];
+  bf1[21] = bf0[21];
+  bf1[22] = bf0[22];
+  bf1[23] = bf0[23];
+  bf1[24] = bf0[24];
+  bf1[25] = bf0[25];
+  bf1[26] = bf0[26];
+  bf1[27] = bf0[27];
+  bf1[28] = bf0[28];
+  bf1[29] = bf0[29];
+  bf1[30] = bf0[30];
+  bf1[31] = bf0[31];
+  bf1[32] = half_btf(cospi[63], bf0[32], cospi[1], bf0[63], cos_bit[stage]);
+  bf1[33] = half_btf(cospi[31], bf0[33], cospi[33], bf0[62], cos_bit[stage]);
+  bf1[34] = half_btf(cospi[47], bf0[34], cospi[17], bf0[61], cos_bit[stage]);
+  bf1[35] = half_btf(cospi[15], bf0[35], cospi[49], bf0[60], cos_bit[stage]);
+  bf1[36] = half_btf(cospi[55], bf0[36], cospi[9], bf0[59], cos_bit[stage]);
+  bf1[37] = half_btf(cospi[23], bf0[37], cospi[41], bf0[58], cos_bit[stage]);
+  bf1[38] = half_btf(cospi[39], bf0[38], cospi[25], bf0[57], cos_bit[stage]);
+  bf1[39] = half_btf(cospi[7], bf0[39], cospi[57], bf0[56], cos_bit[stage]);
+  bf1[40] = half_btf(cospi[59], bf0[40], cospi[5], bf0[55], cos_bit[stage]);
+  bf1[41] = half_btf(cospi[27], bf0[41], cospi[37], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(cospi[43], bf0[42], cospi[21], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(cospi[11], bf0[43], cospi[53], bf0[52], cos_bit[stage]);
+  bf1[44] = half_btf(cospi[51], bf0[44], cospi[13], bf0[51], cos_bit[stage]);
+  bf1[45] = half_btf(cospi[19], bf0[45], cospi[45], bf0[50], cos_bit[stage]);
+  bf1[46] = half_btf(cospi[35], bf0[46], cospi[29], bf0[49], cos_bit[stage]);
+  bf1[47] = half_btf(cospi[3], bf0[47], cospi[61], bf0[48], cos_bit[stage]);
+  bf1[48] = half_btf(cospi[3], bf0[48], -cospi[61], bf0[47], cos_bit[stage]);
+  bf1[49] = half_btf(cospi[35], bf0[49], -cospi[29], bf0[46], cos_bit[stage]);
+  bf1[50] = half_btf(cospi[19], bf0[50], -cospi[45], bf0[45], cos_bit[stage]);
+  bf1[51] = half_btf(cospi[51], bf0[51], -cospi[13], bf0[44], cos_bit[stage]);
+  bf1[52] = half_btf(cospi[11], bf0[52], -cospi[53], bf0[43], cos_bit[stage]);
+  bf1[53] = half_btf(cospi[43], bf0[53], -cospi[21], bf0[42], cos_bit[stage]);
+  bf1[54] = half_btf(cospi[27], bf0[54], -cospi[37], bf0[41], cos_bit[stage]);
+  bf1[55] = half_btf(cospi[59], bf0[55], -cospi[5], bf0[40], cos_bit[stage]);
+  bf1[56] = half_btf(cospi[7], bf0[56], -cospi[57], bf0[39], cos_bit[stage]);
+  bf1[57] = half_btf(cospi[39], bf0[57], -cospi[25], bf0[38], cos_bit[stage]);
+  bf1[58] = half_btf(cospi[23], bf0[58], -cospi[41], bf0[37], cos_bit[stage]);
+  bf1[59] = half_btf(cospi[55], bf0[59], -cospi[9], bf0[36], cos_bit[stage]);
+  bf1[60] = half_btf(cospi[15], bf0[60], -cospi[49], bf0[35], cos_bit[stage]);
+  bf1[61] = half_btf(cospi[47], bf0[61], -cospi[17], bf0[34], cos_bit[stage]);
+  bf1[62] = half_btf(cospi[31], bf0[62], -cospi[33], bf0[33], cos_bit[stage]);
+  bf1[63] = half_btf(cospi[63], bf0[63], -cospi[1], bf0[32], cos_bit[stage]);
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 11
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[32];
+  bf1[2] = bf0[16];
+  bf1[3] = bf0[48];
+  bf1[4] = bf0[8];
+  bf1[5] = bf0[40];
+  bf1[6] = bf0[24];
+  bf1[7] = bf0[56];
+  bf1[8] = bf0[4];
+  bf1[9] = bf0[36];
+  bf1[10] = bf0[20];
+  bf1[11] = bf0[52];
+  bf1[12] = bf0[12];
+  bf1[13] = bf0[44];
+  bf1[14] = bf0[28];
+  bf1[15] = bf0[60];
+  bf1[16] = bf0[2];
+  bf1[17] = bf0[34];
+  bf1[18] = bf0[18];
+  bf1[19] = bf0[50];
+  bf1[20] = bf0[10];
+  bf1[21] = bf0[42];
+  bf1[22] = bf0[26];
+  bf1[23] = bf0[58];
+  bf1[24] = bf0[6];
+  bf1[25] = bf0[38];
+  bf1[26] = bf0[22];
+  bf1[27] = bf0[54];
+  bf1[28] = bf0[14];
+  bf1[29] = bf0[46];
+  bf1[30] = bf0[30];
+  bf1[31] = bf0[62];
+  bf1[32] = bf0[1];
+  bf1[33] = bf0[33];
+  bf1[34] = bf0[17];
+  bf1[35] = bf0[49];
+  bf1[36] = bf0[9];
+  bf1[37] = bf0[41];
+  bf1[38] = bf0[25];
+  bf1[39] = bf0[57];
+  bf1[40] = bf0[5];
+  bf1[41] = bf0[37];
+  bf1[42] = bf0[21];
+  bf1[43] = bf0[53];
+  bf1[44] = bf0[13];
+  bf1[45] = bf0[45];
+  bf1[46] = bf0[29];
+  bf1[47] = bf0[61];
+  bf1[48] = bf0[3];
+  bf1[49] = bf0[35];
+  bf1[50] = bf0[19];
+  bf1[51] = bf0[51];
+  bf1[52] = bf0[11];
+  bf1[53] = bf0[43];
+  bf1[54] = bf0[27];
+  bf1[55] = bf0[59];
+  bf1[56] = bf0[7];
+  bf1[57] = bf0[39];
+  bf1[58] = bf0[23];
+  bf1[59] = bf0[55];
+  bf1[60] = bf0[15];
+  bf1[61] = bf0[47];
+  bf1[62] = bf0[31];
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+}
diff --git a/vp10/common/vp10_fwd_txfm1d.h b/vp10/common/vp10_fwd_txfm1d.h
index d5b9f40..d06e305 100644
--- a/vp10/common/vp10_fwd_txfm1d.h
+++ b/vp10/common/vp10_fwd_txfm1d.h
@@ -25,6 +25,8 @@
                      const int8_t *cos_bit, const int8_t *stage_range);
 void vp10_fdct32_new(const int32_t *input, int32_t *output,
                      const int8_t *cos_bit, const int8_t *stage_range);
+void vp10_fdct64_new(const int32_t *input, int32_t *output,
+                     const int8_t *cos_bit, const int8_t *stage_range);
 
 void vp10_fadst4_new(const int32_t *input, int32_t *output,
                      const int8_t *cos_bit, const int8_t *stage_range);
diff --git a/vp10/common/vp10_fwd_txfm2d.c b/vp10/common/vp10_fwd_txfm2d.c
index 67449ec..045ca2b 100644
--- a/vp10/common/vp10_fwd_txfm2d.c
+++ b/vp10/common/vp10_fwd_txfm2d.c
@@ -82,3 +82,11 @@
   (void)bd;
   fwd_txfm2d_c(input, output, stride, cfg, txfm_buf);
 }
+
+void vp10_fwd_txfm2d_64x64(const int16_t *input, int32_t *output,
+                           const int stride, const TXFM_2D_CFG *cfg,
+                           const int bd) {
+  int txfm_buf[64 * 64 + 64 + 64];
+  (void)bd;
+  fwd_txfm2d_c(input, output, stride, cfg, txfm_buf);
+}
diff --git a/vp10/common/vp10_fwd_txfm2d.h b/vp10/common/vp10_fwd_txfm2d.h
index 64e6f56..3829609 100644
--- a/vp10/common/vp10_fwd_txfm2d.h
+++ b/vp10/common/vp10_fwd_txfm2d.h
@@ -27,6 +27,9 @@
 void vp10_fwd_txfm2d_32x32(const int16_t *input, int32_t *output,
                            const int stride, const TXFM_2D_CFG *cfg,
                            const int bd);
+void vp10_fwd_txfm2d_64x64(const int16_t *input, int32_t *output,
+                           const int stride, const TXFM_2D_CFG *cfg,
+                           const int bd);
 #ifdef __cplusplus
 }
 #endif
diff --git a/vp10/common/vp10_fwd_txfm2d_cfg.h b/vp10/common/vp10_fwd_txfm2d_cfg.h
index 5c2b4ca..bbabafe 100644
--- a/vp10/common/vp10_fwd_txfm2d_cfg.h
+++ b/vp10/common/vp10_fwd_txfm2d_cfg.h
@@ -95,6 +95,29 @@
     vp10_fdct32_new,                 // .txfm_func_col
     vp10_fdct32_new};                // .txfm_func_row;
 
+//  ---------------- config fwd_dct_dct_64 ----------------
+static int8_t fwd_shift_dct_dct_64[3] = {2, -2, -2};
+static int8_t fwd_stage_range_col_dct_dct_64[12] = {13, 14, 15, 16, 17, 18,
+                                                    19, 19, 19, 19, 19, 19};
+static int8_t fwd_stage_range_row_dct_dct_64[12] = {17, 18, 19, 20, 21, 22,
+                                                    22, 22, 22, 22, 22, 22};
+static int8_t fwd_cos_bit_col_dct_dct_64[12] = {15, 15, 15, 15, 15, 14,
+                                                13, 13, 13, 13, 13, 13};
+static int8_t fwd_cos_bit_row_dct_dct_64[12] = {15, 14, 13, 12, 11, 10,
+                                                10, 10, 10, 10, 10, 10};
+
+static const TXFM_2D_CFG fwd_txfm_2d_cfg_dct_dct_64 = {
+    64,                              // .txfm_size
+    12,                              // .stage_num_col
+    12,                              // .stage_num_row
+    fwd_shift_dct_dct_64,            // .shift
+    fwd_stage_range_col_dct_dct_64,  // .stage_range_col
+    fwd_stage_range_row_dct_dct_64,  // .stage_range_row
+    fwd_cos_bit_col_dct_dct_64,      // .cos_bit_col
+    fwd_cos_bit_row_dct_dct_64,      // .cos_bit_row
+    vp10_fdct64_new,                 // .txfm_func_col
+    vp10_fdct64_new};                // .txfm_func_row;
+
 //  ---------------- config fwd_dct_adst_4 ----------------
 static const int8_t fwd_shift_dct_adst_4[3] = {5, -2, -1};
 static const int8_t fwd_stage_range_col_dct_adst_4[4] = {16, 17, 18, 18};
diff --git a/vp10/common/vp10_inv_txfm1d.c b/vp10/common/vp10_inv_txfm1d.c
index 606ca55..494000f 100644
--- a/vp10/common/vp10_inv_txfm1d.c
+++ b/vp10/common/vp10_inv_txfm1d.c
@@ -32,11 +32,11 @@
 #else
 #define range_check(stage, input, buf, size, bit) \
   {                                               \
-    (void) stage;                                 \
-    (void) input;                                 \
-    (void) buf;                                   \
-    (void) size;                                  \
-    (void) bit;                                   \
+    (void)stage;                                  \
+    (void)input;                                  \
+    (void)buf;                                    \
+    (void)size;                                   \
+    (void)bit;                                    \
   }
 #endif
 
@@ -1535,3 +1535,796 @@
   bf1[31] = bf0[0];
   range_check(stage, input, bf1, size, stage_range[stage]);
 }
+
+void vp10_idct64_new(const int32_t *input, int32_t *output,
+                     const int8_t *cos_bit, const int8_t *stage_range) {
+  const int32_t size = 64;
+  const int32_t *cospi;
+
+  int32_t stage = 0;
+  int32_t *bf0, *bf1;
+  int32_t step[64];
+
+  // stage 0;
+  range_check(stage, input, input, size, stage_range[stage]);
+
+  // stage 1;
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf1 = output;
+  bf1[0] = input[0];
+  bf1[1] = input[32];
+  bf1[2] = input[16];
+  bf1[3] = input[48];
+  bf1[4] = input[8];
+  bf1[5] = input[40];
+  bf1[6] = input[24];
+  bf1[7] = input[56];
+  bf1[8] = input[4];
+  bf1[9] = input[36];
+  bf1[10] = input[20];
+  bf1[11] = input[52];
+  bf1[12] = input[12];
+  bf1[13] = input[44];
+  bf1[14] = input[28];
+  bf1[15] = input[60];
+  bf1[16] = input[2];
+  bf1[17] = input[34];
+  bf1[18] = input[18];
+  bf1[19] = input[50];
+  bf1[20] = input[10];
+  bf1[21] = input[42];
+  bf1[22] = input[26];
+  bf1[23] = input[58];
+  bf1[24] = input[6];
+  bf1[25] = input[38];
+  bf1[26] = input[22];
+  bf1[27] = input[54];
+  bf1[28] = input[14];
+  bf1[29] = input[46];
+  bf1[30] = input[30];
+  bf1[31] = input[62];
+  bf1[32] = input[1];
+  bf1[33] = input[33];
+  bf1[34] = input[17];
+  bf1[35] = input[49];
+  bf1[36] = input[9];
+  bf1[37] = input[41];
+  bf1[38] = input[25];
+  bf1[39] = input[57];
+  bf1[40] = input[5];
+  bf1[41] = input[37];
+  bf1[42] = input[21];
+  bf1[43] = input[53];
+  bf1[44] = input[13];
+  bf1[45] = input[45];
+  bf1[46] = input[29];
+  bf1[47] = input[61];
+  bf1[48] = input[3];
+  bf1[49] = input[35];
+  bf1[50] = input[19];
+  bf1[51] = input[51];
+  bf1[52] = input[11];
+  bf1[53] = input[43];
+  bf1[54] = input[27];
+  bf1[55] = input[59];
+  bf1[56] = input[7];
+  bf1[57] = input[39];
+  bf1[58] = input[23];
+  bf1[59] = input[55];
+  bf1[60] = input[15];
+  bf1[61] = input[47];
+  bf1[62] = input[31];
+  bf1[63] = input[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 2
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = bf0[4];
+  bf1[5] = bf0[5];
+  bf1[6] = bf0[6];
+  bf1[7] = bf0[7];
+  bf1[8] = bf0[8];
+  bf1[9] = bf0[9];
+  bf1[10] = bf0[10];
+  bf1[11] = bf0[11];
+  bf1[12] = bf0[12];
+  bf1[13] = bf0[13];
+  bf1[14] = bf0[14];
+  bf1[15] = bf0[15];
+  bf1[16] = bf0[16];
+  bf1[17] = bf0[17];
+  bf1[18] = bf0[18];
+  bf1[19] = bf0[19];
+  bf1[20] = bf0[20];
+  bf1[21] = bf0[21];
+  bf1[22] = bf0[22];
+  bf1[23] = bf0[23];
+  bf1[24] = bf0[24];
+  bf1[25] = bf0[25];
+  bf1[26] = bf0[26];
+  bf1[27] = bf0[27];
+  bf1[28] = bf0[28];
+  bf1[29] = bf0[29];
+  bf1[30] = bf0[30];
+  bf1[31] = bf0[31];
+  bf1[32] = half_btf(cospi[63], bf0[32], -cospi[1], bf0[63], cos_bit[stage]);
+  bf1[33] = half_btf(cospi[31], bf0[33], -cospi[33], bf0[62], cos_bit[stage]);
+  bf1[34] = half_btf(cospi[47], bf0[34], -cospi[17], bf0[61], cos_bit[stage]);
+  bf1[35] = half_btf(cospi[15], bf0[35], -cospi[49], bf0[60], cos_bit[stage]);
+  bf1[36] = half_btf(cospi[55], bf0[36], -cospi[9], bf0[59], cos_bit[stage]);
+  bf1[37] = half_btf(cospi[23], bf0[37], -cospi[41], bf0[58], cos_bit[stage]);
+  bf1[38] = half_btf(cospi[39], bf0[38], -cospi[25], bf0[57], cos_bit[stage]);
+  bf1[39] = half_btf(cospi[7], bf0[39], -cospi[57], bf0[56], cos_bit[stage]);
+  bf1[40] = half_btf(cospi[59], bf0[40], -cospi[5], bf0[55], cos_bit[stage]);
+  bf1[41] = half_btf(cospi[27], bf0[41], -cospi[37], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(cospi[43], bf0[42], -cospi[21], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(cospi[11], bf0[43], -cospi[53], bf0[52], cos_bit[stage]);
+  bf1[44] = half_btf(cospi[51], bf0[44], -cospi[13], bf0[51], cos_bit[stage]);
+  bf1[45] = half_btf(cospi[19], bf0[45], -cospi[45], bf0[50], cos_bit[stage]);
+  bf1[46] = half_btf(cospi[35], bf0[46], -cospi[29], bf0[49], cos_bit[stage]);
+  bf1[47] = half_btf(cospi[3], bf0[47], -cospi[61], bf0[48], cos_bit[stage]);
+  bf1[48] = half_btf(cospi[61], bf0[47], cospi[3], bf0[48], cos_bit[stage]);
+  bf1[49] = half_btf(cospi[29], bf0[46], cospi[35], bf0[49], cos_bit[stage]);
+  bf1[50] = half_btf(cospi[45], bf0[45], cospi[19], bf0[50], cos_bit[stage]);
+  bf1[51] = half_btf(cospi[13], bf0[44], cospi[51], bf0[51], cos_bit[stage]);
+  bf1[52] = half_btf(cospi[53], bf0[43], cospi[11], bf0[52], cos_bit[stage]);
+  bf1[53] = half_btf(cospi[21], bf0[42], cospi[43], bf0[53], cos_bit[stage]);
+  bf1[54] = half_btf(cospi[37], bf0[41], cospi[27], bf0[54], cos_bit[stage]);
+  bf1[55] = half_btf(cospi[5], bf0[40], cospi[59], bf0[55], cos_bit[stage]);
+  bf1[56] = half_btf(cospi[57], bf0[39], cospi[7], bf0[56], cos_bit[stage]);
+  bf1[57] = half_btf(cospi[25], bf0[38], cospi[39], bf0[57], cos_bit[stage]);
+  bf1[58] = half_btf(cospi[41], bf0[37], cospi[23], bf0[58], cos_bit[stage]);
+  bf1[59] = half_btf(cospi[9], bf0[36], cospi[55], bf0[59], cos_bit[stage]);
+  bf1[60] = half_btf(cospi[49], bf0[35], cospi[15], bf0[60], cos_bit[stage]);
+  bf1[61] = half_btf(cospi[17], bf0[34], cospi[47], bf0[61], cos_bit[stage]);
+  bf1[62] = half_btf(cospi[33], bf0[33], cospi[31], bf0[62], cos_bit[stage]);
+  bf1[63] = half_btf(cospi[1], bf0[32], cospi[63], bf0[63], cos_bit[stage]);
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 3
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = bf0[4];
+  bf1[5] = bf0[5];
+  bf1[6] = bf0[6];
+  bf1[7] = bf0[7];
+  bf1[8] = bf0[8];
+  bf1[9] = bf0[9];
+  bf1[10] = bf0[10];
+  bf1[11] = bf0[11];
+  bf1[12] = bf0[12];
+  bf1[13] = bf0[13];
+  bf1[14] = bf0[14];
+  bf1[15] = bf0[15];
+  bf1[16] = half_btf(cospi[62], bf0[16], -cospi[2], bf0[31], cos_bit[stage]);
+  bf1[17] = half_btf(cospi[30], bf0[17], -cospi[34], bf0[30], cos_bit[stage]);
+  bf1[18] = half_btf(cospi[46], bf0[18], -cospi[18], bf0[29], cos_bit[stage]);
+  bf1[19] = half_btf(cospi[14], bf0[19], -cospi[50], bf0[28], cos_bit[stage]);
+  bf1[20] = half_btf(cospi[54], bf0[20], -cospi[10], bf0[27], cos_bit[stage]);
+  bf1[21] = half_btf(cospi[22], bf0[21], -cospi[42], bf0[26], cos_bit[stage]);
+  bf1[22] = half_btf(cospi[38], bf0[22], -cospi[26], bf0[25], cos_bit[stage]);
+  bf1[23] = half_btf(cospi[6], bf0[23], -cospi[58], bf0[24], cos_bit[stage]);
+  bf1[24] = half_btf(cospi[58], bf0[23], cospi[6], bf0[24], cos_bit[stage]);
+  bf1[25] = half_btf(cospi[26], bf0[22], cospi[38], bf0[25], cos_bit[stage]);
+  bf1[26] = half_btf(cospi[42], bf0[21], cospi[22], bf0[26], cos_bit[stage]);
+  bf1[27] = half_btf(cospi[10], bf0[20], cospi[54], bf0[27], cos_bit[stage]);
+  bf1[28] = half_btf(cospi[50], bf0[19], cospi[14], bf0[28], cos_bit[stage]);
+  bf1[29] = half_btf(cospi[18], bf0[18], cospi[46], bf0[29], cos_bit[stage]);
+  bf1[30] = half_btf(cospi[34], bf0[17], cospi[30], bf0[30], cos_bit[stage]);
+  bf1[31] = half_btf(cospi[2], bf0[16], cospi[62], bf0[31], cos_bit[stage]);
+  bf1[32] = bf0[32] + bf0[33];
+  bf1[33] = bf0[32] - bf0[33];
+  bf1[34] = -bf0[34] + bf0[35];
+  bf1[35] = bf0[34] + bf0[35];
+  bf1[36] = bf0[36] + bf0[37];
+  bf1[37] = bf0[36] - bf0[37];
+  bf1[38] = -bf0[38] + bf0[39];
+  bf1[39] = bf0[38] + bf0[39];
+  bf1[40] = bf0[40] + bf0[41];
+  bf1[41] = bf0[40] - bf0[41];
+  bf1[42] = -bf0[42] + bf0[43];
+  bf1[43] = bf0[42] + bf0[43];
+  bf1[44] = bf0[44] + bf0[45];
+  bf1[45] = bf0[44] - bf0[45];
+  bf1[46] = -bf0[46] + bf0[47];
+  bf1[47] = bf0[46] + bf0[47];
+  bf1[48] = bf0[48] + bf0[49];
+  bf1[49] = bf0[48] - bf0[49];
+  bf1[50] = -bf0[50] + bf0[51];
+  bf1[51] = bf0[50] + bf0[51];
+  bf1[52] = bf0[52] + bf0[53];
+  bf1[53] = bf0[52] - bf0[53];
+  bf1[54] = -bf0[54] + bf0[55];
+  bf1[55] = bf0[54] + bf0[55];
+  bf1[56] = bf0[56] + bf0[57];
+  bf1[57] = bf0[56] - bf0[57];
+  bf1[58] = -bf0[58] + bf0[59];
+  bf1[59] = bf0[58] + bf0[59];
+  bf1[60] = bf0[60] + bf0[61];
+  bf1[61] = bf0[60] - bf0[61];
+  bf1[62] = -bf0[62] + bf0[63];
+  bf1[63] = bf0[62] + bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 4
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = bf0[4];
+  bf1[5] = bf0[5];
+  bf1[6] = bf0[6];
+  bf1[7] = bf0[7];
+  bf1[8] = half_btf(cospi[60], bf0[8], -cospi[4], bf0[15], cos_bit[stage]);
+  bf1[9] = half_btf(cospi[28], bf0[9], -cospi[36], bf0[14], cos_bit[stage]);
+  bf1[10] = half_btf(cospi[44], bf0[10], -cospi[20], bf0[13], cos_bit[stage]);
+  bf1[11] = half_btf(cospi[12], bf0[11], -cospi[52], bf0[12], cos_bit[stage]);
+  bf1[12] = half_btf(cospi[52], bf0[11], cospi[12], bf0[12], cos_bit[stage]);
+  bf1[13] = half_btf(cospi[20], bf0[10], cospi[44], bf0[13], cos_bit[stage]);
+  bf1[14] = half_btf(cospi[36], bf0[9], cospi[28], bf0[14], cos_bit[stage]);
+  bf1[15] = half_btf(cospi[4], bf0[8], cospi[60], bf0[15], cos_bit[stage]);
+  bf1[16] = bf0[16] + bf0[17];
+  bf1[17] = bf0[16] - bf0[17];
+  bf1[18] = -bf0[18] + bf0[19];
+  bf1[19] = bf0[18] + bf0[19];
+  bf1[20] = bf0[20] + bf0[21];
+  bf1[21] = bf0[20] - bf0[21];
+  bf1[22] = -bf0[22] + bf0[23];
+  bf1[23] = bf0[22] + bf0[23];
+  bf1[24] = bf0[24] + bf0[25];
+  bf1[25] = bf0[24] - bf0[25];
+  bf1[26] = -bf0[26] + bf0[27];
+  bf1[27] = bf0[26] + bf0[27];
+  bf1[28] = bf0[28] + bf0[29];
+  bf1[29] = bf0[28] - bf0[29];
+  bf1[30] = -bf0[30] + bf0[31];
+  bf1[31] = bf0[30] + bf0[31];
+  bf1[32] = bf0[32];
+  bf1[33] = half_btf(-cospi[4], bf0[33], cospi[60], bf0[62], cos_bit[stage]);
+  bf1[34] = half_btf(-cospi[60], bf0[34], -cospi[4], bf0[61], cos_bit[stage]);
+  bf1[35] = bf0[35];
+  bf1[36] = bf0[36];
+  bf1[37] = half_btf(-cospi[36], bf0[37], cospi[28], bf0[58], cos_bit[stage]);
+  bf1[38] = half_btf(-cospi[28], bf0[38], -cospi[36], bf0[57], cos_bit[stage]);
+  bf1[39] = bf0[39];
+  bf1[40] = bf0[40];
+  bf1[41] = half_btf(-cospi[20], bf0[41], cospi[44], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(-cospi[44], bf0[42], -cospi[20], bf0[53], cos_bit[stage]);
+  bf1[43] = bf0[43];
+  bf1[44] = bf0[44];
+  bf1[45] = half_btf(-cospi[52], bf0[45], cospi[12], bf0[50], cos_bit[stage]);
+  bf1[46] = half_btf(-cospi[12], bf0[46], -cospi[52], bf0[49], cos_bit[stage]);
+  bf1[47] = bf0[47];
+  bf1[48] = bf0[48];
+  bf1[49] = half_btf(-cospi[52], bf0[46], cospi[12], bf0[49], cos_bit[stage]);
+  bf1[50] = half_btf(cospi[12], bf0[45], cospi[52], bf0[50], cos_bit[stage]);
+  bf1[51] = bf0[51];
+  bf1[52] = bf0[52];
+  bf1[53] = half_btf(-cospi[20], bf0[42], cospi[44], bf0[53], cos_bit[stage]);
+  bf1[54] = half_btf(cospi[44], bf0[41], cospi[20], bf0[54], cos_bit[stage]);
+  bf1[55] = bf0[55];
+  bf1[56] = bf0[56];
+  bf1[57] = half_btf(-cospi[36], bf0[38], cospi[28], bf0[57], cos_bit[stage]);
+  bf1[58] = half_btf(cospi[28], bf0[37], cospi[36], bf0[58], cos_bit[stage]);
+  bf1[59] = bf0[59];
+  bf1[60] = bf0[60];
+  bf1[61] = half_btf(-cospi[4], bf0[34], cospi[60], bf0[61], cos_bit[stage]);
+  bf1[62] = half_btf(cospi[60], bf0[33], cospi[4], bf0[62], cos_bit[stage]);
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 5
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0];
+  bf1[1] = bf0[1];
+  bf1[2] = bf0[2];
+  bf1[3] = bf0[3];
+  bf1[4] = half_btf(cospi[56], bf0[4], -cospi[8], bf0[7], cos_bit[stage]);
+  bf1[5] = half_btf(cospi[24], bf0[5], -cospi[40], bf0[6], cos_bit[stage]);
+  bf1[6] = half_btf(cospi[40], bf0[5], cospi[24], bf0[6], cos_bit[stage]);
+  bf1[7] = half_btf(cospi[8], bf0[4], cospi[56], bf0[7], cos_bit[stage]);
+  bf1[8] = bf0[8] + bf0[9];
+  bf1[9] = bf0[8] - bf0[9];
+  bf1[10] = -bf0[10] + bf0[11];
+  bf1[11] = bf0[10] + bf0[11];
+  bf1[12] = bf0[12] + bf0[13];
+  bf1[13] = bf0[12] - bf0[13];
+  bf1[14] = -bf0[14] + bf0[15];
+  bf1[15] = bf0[14] + bf0[15];
+  bf1[16] = bf0[16];
+  bf1[17] = half_btf(-cospi[8], bf0[17], cospi[56], bf0[30], cos_bit[stage]);
+  bf1[18] = half_btf(-cospi[56], bf0[18], -cospi[8], bf0[29], cos_bit[stage]);
+  bf1[19] = bf0[19];
+  bf1[20] = bf0[20];
+  bf1[21] = half_btf(-cospi[40], bf0[21], cospi[24], bf0[26], cos_bit[stage]);
+  bf1[22] = half_btf(-cospi[24], bf0[22], -cospi[40], bf0[25], cos_bit[stage]);
+  bf1[23] = bf0[23];
+  bf1[24] = bf0[24];
+  bf1[25] = half_btf(-cospi[40], bf0[22], cospi[24], bf0[25], cos_bit[stage]);
+  bf1[26] = half_btf(cospi[24], bf0[21], cospi[40], bf0[26], cos_bit[stage]);
+  bf1[27] = bf0[27];
+  bf1[28] = bf0[28];
+  bf1[29] = half_btf(-cospi[8], bf0[18], cospi[56], bf0[29], cos_bit[stage]);
+  bf1[30] = half_btf(cospi[56], bf0[17], cospi[8], bf0[30], cos_bit[stage]);
+  bf1[31] = bf0[31];
+  bf1[32] = bf0[32] + bf0[35];
+  bf1[33] = bf0[33] + bf0[34];
+  bf1[34] = bf0[33] - bf0[34];
+  bf1[35] = bf0[32] - bf0[35];
+  bf1[36] = -bf0[36] + bf0[39];
+  bf1[37] = -bf0[37] + bf0[38];
+  bf1[38] = bf0[37] + bf0[38];
+  bf1[39] = bf0[36] + bf0[39];
+  bf1[40] = bf0[40] + bf0[43];
+  bf1[41] = bf0[41] + bf0[42];
+  bf1[42] = bf0[41] - bf0[42];
+  bf1[43] = bf0[40] - bf0[43];
+  bf1[44] = -bf0[44] + bf0[47];
+  bf1[45] = -bf0[45] + bf0[46];
+  bf1[46] = bf0[45] + bf0[46];
+  bf1[47] = bf0[44] + bf0[47];
+  bf1[48] = bf0[48] + bf0[51];
+  bf1[49] = bf0[49] + bf0[50];
+  bf1[50] = bf0[49] - bf0[50];
+  bf1[51] = bf0[48] - bf0[51];
+  bf1[52] = -bf0[52] + bf0[55];
+  bf1[53] = -bf0[53] + bf0[54];
+  bf1[54] = bf0[53] + bf0[54];
+  bf1[55] = bf0[52] + bf0[55];
+  bf1[56] = bf0[56] + bf0[59];
+  bf1[57] = bf0[57] + bf0[58];
+  bf1[58] = bf0[57] - bf0[58];
+  bf1[59] = bf0[56] - bf0[59];
+  bf1[60] = -bf0[60] + bf0[63];
+  bf1[61] = -bf0[61] + bf0[62];
+  bf1[62] = bf0[61] + bf0[62];
+  bf1[63] = bf0[60] + bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 6
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = half_btf(cospi[32], bf0[0], cospi[32], bf0[1], cos_bit[stage]);
+  bf1[1] = half_btf(cospi[32], bf0[0], -cospi[32], bf0[1], cos_bit[stage]);
+  bf1[2] = half_btf(cospi[48], bf0[2], -cospi[16], bf0[3], cos_bit[stage]);
+  bf1[3] = half_btf(cospi[16], bf0[2], cospi[48], bf0[3], cos_bit[stage]);
+  bf1[4] = bf0[4] + bf0[5];
+  bf1[5] = bf0[4] - bf0[5];
+  bf1[6] = -bf0[6] + bf0[7];
+  bf1[7] = bf0[6] + bf0[7];
+  bf1[8] = bf0[8];
+  bf1[9] = half_btf(-cospi[16], bf0[9], cospi[48], bf0[14], cos_bit[stage]);
+  bf1[10] = half_btf(-cospi[48], bf0[10], -cospi[16], bf0[13], cos_bit[stage]);
+  bf1[11] = bf0[11];
+  bf1[12] = bf0[12];
+  bf1[13] = half_btf(-cospi[16], bf0[10], cospi[48], bf0[13], cos_bit[stage]);
+  bf1[14] = half_btf(cospi[48], bf0[9], cospi[16], bf0[14], cos_bit[stage]);
+  bf1[15] = bf0[15];
+  bf1[16] = bf0[16] + bf0[19];
+  bf1[17] = bf0[17] + bf0[18];
+  bf1[18] = bf0[17] - bf0[18];
+  bf1[19] = bf0[16] - bf0[19];
+  bf1[20] = -bf0[20] + bf0[23];
+  bf1[21] = -bf0[21] + bf0[22];
+  bf1[22] = bf0[21] + bf0[22];
+  bf1[23] = bf0[20] + bf0[23];
+  bf1[24] = bf0[24] + bf0[27];
+  bf1[25] = bf0[25] + bf0[26];
+  bf1[26] = bf0[25] - bf0[26];
+  bf1[27] = bf0[24] - bf0[27];
+  bf1[28] = -bf0[28] + bf0[31];
+  bf1[29] = -bf0[29] + bf0[30];
+  bf1[30] = bf0[29] + bf0[30];
+  bf1[31] = bf0[28] + bf0[31];
+  bf1[32] = bf0[32];
+  bf1[33] = bf0[33];
+  bf1[34] = half_btf(-cospi[8], bf0[34], cospi[56], bf0[61], cos_bit[stage]);
+  bf1[35] = half_btf(-cospi[8], bf0[35], cospi[56], bf0[60], cos_bit[stage]);
+  bf1[36] = half_btf(-cospi[56], bf0[36], -cospi[8], bf0[59], cos_bit[stage]);
+  bf1[37] = half_btf(-cospi[56], bf0[37], -cospi[8], bf0[58], cos_bit[stage]);
+  bf1[38] = bf0[38];
+  bf1[39] = bf0[39];
+  bf1[40] = bf0[40];
+  bf1[41] = bf0[41];
+  bf1[42] = half_btf(-cospi[40], bf0[42], cospi[24], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(-cospi[40], bf0[43], cospi[24], bf0[52], cos_bit[stage]);
+  bf1[44] = half_btf(-cospi[24], bf0[44], -cospi[40], bf0[51], cos_bit[stage]);
+  bf1[45] = half_btf(-cospi[24], bf0[45], -cospi[40], bf0[50], cos_bit[stage]);
+  bf1[46] = bf0[46];
+  bf1[47] = bf0[47];
+  bf1[48] = bf0[48];
+  bf1[49] = bf0[49];
+  bf1[50] = half_btf(-cospi[40], bf0[45], cospi[24], bf0[50], cos_bit[stage]);
+  bf1[51] = half_btf(-cospi[40], bf0[44], cospi[24], bf0[51], cos_bit[stage]);
+  bf1[52] = half_btf(cospi[24], bf0[43], cospi[40], bf0[52], cos_bit[stage]);
+  bf1[53] = half_btf(cospi[24], bf0[42], cospi[40], bf0[53], cos_bit[stage]);
+  bf1[54] = bf0[54];
+  bf1[55] = bf0[55];
+  bf1[56] = bf0[56];
+  bf1[57] = bf0[57];
+  bf1[58] = half_btf(-cospi[8], bf0[37], cospi[56], bf0[58], cos_bit[stage]);
+  bf1[59] = half_btf(-cospi[8], bf0[36], cospi[56], bf0[59], cos_bit[stage]);
+  bf1[60] = half_btf(cospi[56], bf0[35], cospi[8], bf0[60], cos_bit[stage]);
+  bf1[61] = half_btf(cospi[56], bf0[34], cospi[8], bf0[61], cos_bit[stage]);
+  bf1[62] = bf0[62];
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 7
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0] + bf0[3];
+  bf1[1] = bf0[1] + bf0[2];
+  bf1[2] = bf0[1] - bf0[2];
+  bf1[3] = bf0[0] - bf0[3];
+  bf1[4] = bf0[4];
+  bf1[5] = half_btf(-cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
+  bf1[6] = half_btf(cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
+  bf1[7] = bf0[7];
+  bf1[8] = bf0[8] + bf0[11];
+  bf1[9] = bf0[9] + bf0[10];
+  bf1[10] = bf0[9] - bf0[10];
+  bf1[11] = bf0[8] - bf0[11];
+  bf1[12] = -bf0[12] + bf0[15];
+  bf1[13] = -bf0[13] + bf0[14];
+  bf1[14] = bf0[13] + bf0[14];
+  bf1[15] = bf0[12] + bf0[15];
+  bf1[16] = bf0[16];
+  bf1[17] = bf0[17];
+  bf1[18] = half_btf(-cospi[16], bf0[18], cospi[48], bf0[29], cos_bit[stage]);
+  bf1[19] = half_btf(-cospi[16], bf0[19], cospi[48], bf0[28], cos_bit[stage]);
+  bf1[20] = half_btf(-cospi[48], bf0[20], -cospi[16], bf0[27], cos_bit[stage]);
+  bf1[21] = half_btf(-cospi[48], bf0[21], -cospi[16], bf0[26], cos_bit[stage]);
+  bf1[22] = bf0[22];
+  bf1[23] = bf0[23];
+  bf1[24] = bf0[24];
+  bf1[25] = bf0[25];
+  bf1[26] = half_btf(-cospi[16], bf0[21], cospi[48], bf0[26], cos_bit[stage]);
+  bf1[27] = half_btf(-cospi[16], bf0[20], cospi[48], bf0[27], cos_bit[stage]);
+  bf1[28] = half_btf(cospi[48], bf0[19], cospi[16], bf0[28], cos_bit[stage]);
+  bf1[29] = half_btf(cospi[48], bf0[18], cospi[16], bf0[29], cos_bit[stage]);
+  bf1[30] = bf0[30];
+  bf1[31] = bf0[31];
+  bf1[32] = bf0[32] + bf0[39];
+  bf1[33] = bf0[33] + bf0[38];
+  bf1[34] = bf0[34] + bf0[37];
+  bf1[35] = bf0[35] + bf0[36];
+  bf1[36] = bf0[35] - bf0[36];
+  bf1[37] = bf0[34] - bf0[37];
+  bf1[38] = bf0[33] - bf0[38];
+  bf1[39] = bf0[32] - bf0[39];
+  bf1[40] = -bf0[40] + bf0[47];
+  bf1[41] = -bf0[41] + bf0[46];
+  bf1[42] = -bf0[42] + bf0[45];
+  bf1[43] = -bf0[43] + bf0[44];
+  bf1[44] = bf0[43] + bf0[44];
+  bf1[45] = bf0[42] + bf0[45];
+  bf1[46] = bf0[41] + bf0[46];
+  bf1[47] = bf0[40] + bf0[47];
+  bf1[48] = bf0[48] + bf0[55];
+  bf1[49] = bf0[49] + bf0[54];
+  bf1[50] = bf0[50] + bf0[53];
+  bf1[51] = bf0[51] + bf0[52];
+  bf1[52] = bf0[51] - bf0[52];
+  bf1[53] = bf0[50] - bf0[53];
+  bf1[54] = bf0[49] - bf0[54];
+  bf1[55] = bf0[48] - bf0[55];
+  bf1[56] = -bf0[56] + bf0[63];
+  bf1[57] = -bf0[57] + bf0[62];
+  bf1[58] = -bf0[58] + bf0[61];
+  bf1[59] = -bf0[59] + bf0[60];
+  bf1[60] = bf0[59] + bf0[60];
+  bf1[61] = bf0[58] + bf0[61];
+  bf1[62] = bf0[57] + bf0[62];
+  bf1[63] = bf0[56] + bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 8
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0] + bf0[7];
+  bf1[1] = bf0[1] + bf0[6];
+  bf1[2] = bf0[2] + bf0[5];
+  bf1[3] = bf0[3] + bf0[4];
+  bf1[4] = bf0[3] - bf0[4];
+  bf1[5] = bf0[2] - bf0[5];
+  bf1[6] = bf0[1] - bf0[6];
+  bf1[7] = bf0[0] - bf0[7];
+  bf1[8] = bf0[8];
+  bf1[9] = bf0[9];
+  bf1[10] = half_btf(-cospi[32], bf0[10], cospi[32], bf0[13], cos_bit[stage]);
+  bf1[11] = half_btf(-cospi[32], bf0[11], cospi[32], bf0[12], cos_bit[stage]);
+  bf1[12] = half_btf(cospi[32], bf0[11], cospi[32], bf0[12], cos_bit[stage]);
+  bf1[13] = half_btf(cospi[32], bf0[10], cospi[32], bf0[13], cos_bit[stage]);
+  bf1[14] = bf0[14];
+  bf1[15] = bf0[15];
+  bf1[16] = bf0[16] + bf0[23];
+  bf1[17] = bf0[17] + bf0[22];
+  bf1[18] = bf0[18] + bf0[21];
+  bf1[19] = bf0[19] + bf0[20];
+  bf1[20] = bf0[19] - bf0[20];
+  bf1[21] = bf0[18] - bf0[21];
+  bf1[22] = bf0[17] - bf0[22];
+  bf1[23] = bf0[16] - bf0[23];
+  bf1[24] = -bf0[24] + bf0[31];
+  bf1[25] = -bf0[25] + bf0[30];
+  bf1[26] = -bf0[26] + bf0[29];
+  bf1[27] = -bf0[27] + bf0[28];
+  bf1[28] = bf0[27] + bf0[28];
+  bf1[29] = bf0[26] + bf0[29];
+  bf1[30] = bf0[25] + bf0[30];
+  bf1[31] = bf0[24] + bf0[31];
+  bf1[32] = bf0[32];
+  bf1[33] = bf0[33];
+  bf1[34] = bf0[34];
+  bf1[35] = bf0[35];
+  bf1[36] = half_btf(-cospi[16], bf0[36], cospi[48], bf0[59], cos_bit[stage]);
+  bf1[37] = half_btf(-cospi[16], bf0[37], cospi[48], bf0[58], cos_bit[stage]);
+  bf1[38] = half_btf(-cospi[16], bf0[38], cospi[48], bf0[57], cos_bit[stage]);
+  bf1[39] = half_btf(-cospi[16], bf0[39], cospi[48], bf0[56], cos_bit[stage]);
+  bf1[40] = half_btf(-cospi[48], bf0[40], -cospi[16], bf0[55], cos_bit[stage]);
+  bf1[41] = half_btf(-cospi[48], bf0[41], -cospi[16], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(-cospi[48], bf0[42], -cospi[16], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(-cospi[48], bf0[43], -cospi[16], bf0[52], cos_bit[stage]);
+  bf1[44] = bf0[44];
+  bf1[45] = bf0[45];
+  bf1[46] = bf0[46];
+  bf1[47] = bf0[47];
+  bf1[48] = bf0[48];
+  bf1[49] = bf0[49];
+  bf1[50] = bf0[50];
+  bf1[51] = bf0[51];
+  bf1[52] = half_btf(-cospi[16], bf0[43], cospi[48], bf0[52], cos_bit[stage]);
+  bf1[53] = half_btf(-cospi[16], bf0[42], cospi[48], bf0[53], cos_bit[stage]);
+  bf1[54] = half_btf(-cospi[16], bf0[41], cospi[48], bf0[54], cos_bit[stage]);
+  bf1[55] = half_btf(-cospi[16], bf0[40], cospi[48], bf0[55], cos_bit[stage]);
+  bf1[56] = half_btf(cospi[48], bf0[39], cospi[16], bf0[56], cos_bit[stage]);
+  bf1[57] = half_btf(cospi[48], bf0[38], cospi[16], bf0[57], cos_bit[stage]);
+  bf1[58] = half_btf(cospi[48], bf0[37], cospi[16], bf0[58], cos_bit[stage]);
+  bf1[59] = half_btf(cospi[48], bf0[36], cospi[16], bf0[59], cos_bit[stage]);
+  bf1[60] = bf0[60];
+  bf1[61] = bf0[61];
+  bf1[62] = bf0[62];
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 9
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0] + bf0[15];
+  bf1[1] = bf0[1] + bf0[14];
+  bf1[2] = bf0[2] + bf0[13];
+  bf1[3] = bf0[3] + bf0[12];
+  bf1[4] = bf0[4] + bf0[11];
+  bf1[5] = bf0[5] + bf0[10];
+  bf1[6] = bf0[6] + bf0[9];
+  bf1[7] = bf0[7] + bf0[8];
+  bf1[8] = bf0[7] - bf0[8];
+  bf1[9] = bf0[6] - bf0[9];
+  bf1[10] = bf0[5] - bf0[10];
+  bf1[11] = bf0[4] - bf0[11];
+  bf1[12] = bf0[3] - bf0[12];
+  bf1[13] = bf0[2] - bf0[13];
+  bf1[14] = bf0[1] - bf0[14];
+  bf1[15] = bf0[0] - bf0[15];
+  bf1[16] = bf0[16];
+  bf1[17] = bf0[17];
+  bf1[18] = bf0[18];
+  bf1[19] = bf0[19];
+  bf1[20] = half_btf(-cospi[32], bf0[20], cospi[32], bf0[27], cos_bit[stage]);
+  bf1[21] = half_btf(-cospi[32], bf0[21], cospi[32], bf0[26], cos_bit[stage]);
+  bf1[22] = half_btf(-cospi[32], bf0[22], cospi[32], bf0[25], cos_bit[stage]);
+  bf1[23] = half_btf(-cospi[32], bf0[23], cospi[32], bf0[24], cos_bit[stage]);
+  bf1[24] = half_btf(cospi[32], bf0[23], cospi[32], bf0[24], cos_bit[stage]);
+  bf1[25] = half_btf(cospi[32], bf0[22], cospi[32], bf0[25], cos_bit[stage]);
+  bf1[26] = half_btf(cospi[32], bf0[21], cospi[32], bf0[26], cos_bit[stage]);
+  bf1[27] = half_btf(cospi[32], bf0[20], cospi[32], bf0[27], cos_bit[stage]);
+  bf1[28] = bf0[28];
+  bf1[29] = bf0[29];
+  bf1[30] = bf0[30];
+  bf1[31] = bf0[31];
+  bf1[32] = bf0[32] + bf0[47];
+  bf1[33] = bf0[33] + bf0[46];
+  bf1[34] = bf0[34] + bf0[45];
+  bf1[35] = bf0[35] + bf0[44];
+  bf1[36] = bf0[36] + bf0[43];
+  bf1[37] = bf0[37] + bf0[42];
+  bf1[38] = bf0[38] + bf0[41];
+  bf1[39] = bf0[39] + bf0[40];
+  bf1[40] = bf0[39] - bf0[40];
+  bf1[41] = bf0[38] - bf0[41];
+  bf1[42] = bf0[37] - bf0[42];
+  bf1[43] = bf0[36] - bf0[43];
+  bf1[44] = bf0[35] - bf0[44];
+  bf1[45] = bf0[34] - bf0[45];
+  bf1[46] = bf0[33] - bf0[46];
+  bf1[47] = bf0[32] - bf0[47];
+  bf1[48] = -bf0[48] + bf0[63];
+  bf1[49] = -bf0[49] + bf0[62];
+  bf1[50] = -bf0[50] + bf0[61];
+  bf1[51] = -bf0[51] + bf0[60];
+  bf1[52] = -bf0[52] + bf0[59];
+  bf1[53] = -bf0[53] + bf0[58];
+  bf1[54] = -bf0[54] + bf0[57];
+  bf1[55] = -bf0[55] + bf0[56];
+  bf1[56] = bf0[55] + bf0[56];
+  bf1[57] = bf0[54] + bf0[57];
+  bf1[58] = bf0[53] + bf0[58];
+  bf1[59] = bf0[52] + bf0[59];
+  bf1[60] = bf0[51] + bf0[60];
+  bf1[61] = bf0[50] + bf0[61];
+  bf1[62] = bf0[49] + bf0[62];
+  bf1[63] = bf0[48] + bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 10
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = output;
+  bf1 = step;
+  bf1[0] = bf0[0] + bf0[31];
+  bf1[1] = bf0[1] + bf0[30];
+  bf1[2] = bf0[2] + bf0[29];
+  bf1[3] = bf0[3] + bf0[28];
+  bf1[4] = bf0[4] + bf0[27];
+  bf1[5] = bf0[5] + bf0[26];
+  bf1[6] = bf0[6] + bf0[25];
+  bf1[7] = bf0[7] + bf0[24];
+  bf1[8] = bf0[8] + bf0[23];
+  bf1[9] = bf0[9] + bf0[22];
+  bf1[10] = bf0[10] + bf0[21];
+  bf1[11] = bf0[11] + bf0[20];
+  bf1[12] = bf0[12] + bf0[19];
+  bf1[13] = bf0[13] + bf0[18];
+  bf1[14] = bf0[14] + bf0[17];
+  bf1[15] = bf0[15] + bf0[16];
+  bf1[16] = bf0[15] - bf0[16];
+  bf1[17] = bf0[14] - bf0[17];
+  bf1[18] = bf0[13] - bf0[18];
+  bf1[19] = bf0[12] - bf0[19];
+  bf1[20] = bf0[11] - bf0[20];
+  bf1[21] = bf0[10] - bf0[21];
+  bf1[22] = bf0[9] - bf0[22];
+  bf1[23] = bf0[8] - bf0[23];
+  bf1[24] = bf0[7] - bf0[24];
+  bf1[25] = bf0[6] - bf0[25];
+  bf1[26] = bf0[5] - bf0[26];
+  bf1[27] = bf0[4] - bf0[27];
+  bf1[28] = bf0[3] - bf0[28];
+  bf1[29] = bf0[2] - bf0[29];
+  bf1[30] = bf0[1] - bf0[30];
+  bf1[31] = bf0[0] - bf0[31];
+  bf1[32] = bf0[32];
+  bf1[33] = bf0[33];
+  bf1[34] = bf0[34];
+  bf1[35] = bf0[35];
+  bf1[36] = bf0[36];
+  bf1[37] = bf0[37];
+  bf1[38] = bf0[38];
+  bf1[39] = bf0[39];
+  bf1[40] = half_btf(-cospi[32], bf0[40], cospi[32], bf0[55], cos_bit[stage]);
+  bf1[41] = half_btf(-cospi[32], bf0[41], cospi[32], bf0[54], cos_bit[stage]);
+  bf1[42] = half_btf(-cospi[32], bf0[42], cospi[32], bf0[53], cos_bit[stage]);
+  bf1[43] = half_btf(-cospi[32], bf0[43], cospi[32], bf0[52], cos_bit[stage]);
+  bf1[44] = half_btf(-cospi[32], bf0[44], cospi[32], bf0[51], cos_bit[stage]);
+  bf1[45] = half_btf(-cospi[32], bf0[45], cospi[32], bf0[50], cos_bit[stage]);
+  bf1[46] = half_btf(-cospi[32], bf0[46], cospi[32], bf0[49], cos_bit[stage]);
+  bf1[47] = half_btf(-cospi[32], bf0[47], cospi[32], bf0[48], cos_bit[stage]);
+  bf1[48] = half_btf(cospi[32], bf0[47], cospi[32], bf0[48], cos_bit[stage]);
+  bf1[49] = half_btf(cospi[32], bf0[46], cospi[32], bf0[49], cos_bit[stage]);
+  bf1[50] = half_btf(cospi[32], bf0[45], cospi[32], bf0[50], cos_bit[stage]);
+  bf1[51] = half_btf(cospi[32], bf0[44], cospi[32], bf0[51], cos_bit[stage]);
+  bf1[52] = half_btf(cospi[32], bf0[43], cospi[32], bf0[52], cos_bit[stage]);
+  bf1[53] = half_btf(cospi[32], bf0[42], cospi[32], bf0[53], cos_bit[stage]);
+  bf1[54] = half_btf(cospi[32], bf0[41], cospi[32], bf0[54], cos_bit[stage]);
+  bf1[55] = half_btf(cospi[32], bf0[40], cospi[32], bf0[55], cos_bit[stage]);
+  bf1[56] = bf0[56];
+  bf1[57] = bf0[57];
+  bf1[58] = bf0[58];
+  bf1[59] = bf0[59];
+  bf1[60] = bf0[60];
+  bf1[61] = bf0[61];
+  bf1[62] = bf0[62];
+  bf1[63] = bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+
+  // stage 11
+  stage++;
+  cospi = cospi_arr[cos_bit[stage] - cos_bit_min];
+  bf0 = step;
+  bf1 = output;
+  bf1[0] = bf0[0] + bf0[63];
+  bf1[1] = bf0[1] + bf0[62];
+  bf1[2] = bf0[2] + bf0[61];
+  bf1[3] = bf0[3] + bf0[60];
+  bf1[4] = bf0[4] + bf0[59];
+  bf1[5] = bf0[5] + bf0[58];
+  bf1[6] = bf0[6] + bf0[57];
+  bf1[7] = bf0[7] + bf0[56];
+  bf1[8] = bf0[8] + bf0[55];
+  bf1[9] = bf0[9] + bf0[54];
+  bf1[10] = bf0[10] + bf0[53];
+  bf1[11] = bf0[11] + bf0[52];
+  bf1[12] = bf0[12] + bf0[51];
+  bf1[13] = bf0[13] + bf0[50];
+  bf1[14] = bf0[14] + bf0[49];
+  bf1[15] = bf0[15] + bf0[48];
+  bf1[16] = bf0[16] + bf0[47];
+  bf1[17] = bf0[17] + bf0[46];
+  bf1[18] = bf0[18] + bf0[45];
+  bf1[19] = bf0[19] + bf0[44];
+  bf1[20] = bf0[20] + bf0[43];
+  bf1[21] = bf0[21] + bf0[42];
+  bf1[22] = bf0[22] + bf0[41];
+  bf1[23] = bf0[23] + bf0[40];
+  bf1[24] = bf0[24] + bf0[39];
+  bf1[25] = bf0[25] + bf0[38];
+  bf1[26] = bf0[26] + bf0[37];
+  bf1[27] = bf0[27] + bf0[36];
+  bf1[28] = bf0[28] + bf0[35];
+  bf1[29] = bf0[29] + bf0[34];
+  bf1[30] = bf0[30] + bf0[33];
+  bf1[31] = bf0[31] + bf0[32];
+  bf1[32] = bf0[31] - bf0[32];
+  bf1[33] = bf0[30] - bf0[33];
+  bf1[34] = bf0[29] - bf0[34];
+  bf1[35] = bf0[28] - bf0[35];
+  bf1[36] = bf0[27] - bf0[36];
+  bf1[37] = bf0[26] - bf0[37];
+  bf1[38] = bf0[25] - bf0[38];
+  bf1[39] = bf0[24] - bf0[39];
+  bf1[40] = bf0[23] - bf0[40];
+  bf1[41] = bf0[22] - bf0[41];
+  bf1[42] = bf0[21] - bf0[42];
+  bf1[43] = bf0[20] - bf0[43];
+  bf1[44] = bf0[19] - bf0[44];
+  bf1[45] = bf0[18] - bf0[45];
+  bf1[46] = bf0[17] - bf0[46];
+  bf1[47] = bf0[16] - bf0[47];
+  bf1[48] = bf0[15] - bf0[48];
+  bf1[49] = bf0[14] - bf0[49];
+  bf1[50] = bf0[13] - bf0[50];
+  bf1[51] = bf0[12] - bf0[51];
+  bf1[52] = bf0[11] - bf0[52];
+  bf1[53] = bf0[10] - bf0[53];
+  bf1[54] = bf0[9] - bf0[54];
+  bf1[55] = bf0[8] - bf0[55];
+  bf1[56] = bf0[7] - bf0[56];
+  bf1[57] = bf0[6] - bf0[57];
+  bf1[58] = bf0[5] - bf0[58];
+  bf1[59] = bf0[4] - bf0[59];
+  bf1[60] = bf0[3] - bf0[60];
+  bf1[61] = bf0[2] - bf0[61];
+  bf1[62] = bf0[1] - bf0[62];
+  bf1[63] = bf0[0] - bf0[63];
+  range_check(stage, input, bf1, size, stage_range[stage]);
+}
diff --git a/vp10/common/vp10_inv_txfm1d.h b/vp10/common/vp10_inv_txfm1d.h
index 0609b65..fd547a6 100644
--- a/vp10/common/vp10_inv_txfm1d.h
+++ b/vp10/common/vp10_inv_txfm1d.h
@@ -25,6 +25,8 @@
                      const int8_t *cos_bit, const int8_t *stage_range);
 void vp10_idct32_new(const int32_t *input, int32_t *output,
                      const int8_t *cos_bit, const int8_t *stage_range);
+void vp10_idct64_new(const int32_t *input, int32_t *output,
+                     const int8_t *cos_bit, const int8_t *stage_range);
 
 void vp10_iadst4_new(const int32_t *input, int32_t *output,
                      const int8_t *cos_bit, const int8_t *stage_range);
diff --git a/vp10/common/vp10_inv_txfm2d.c b/vp10/common/vp10_inv_txfm2d.c
index c894a42..cacbd0b 100644
--- a/vp10/common/vp10_inv_txfm2d.c
+++ b/vp10/common/vp10_inv_txfm2d.c
@@ -96,3 +96,15 @@
   inv_txfm2d_add_c(input, (int16_t *)output, stride, cfg, txfm_buf);
   clamp_block((int16_t *)output, 32, stride, 0, (1 << bd) - 1);
 }
+
+void vp10_inv_txfm2d_add_64x64(const int32_t *input, uint16_t *output,
+                               const int stride, const TXFM_2D_CFG *cfg,
+                               const int bd) {
+  int txfm_buf[64 * 64 + 64 + 64];
+  // output contains the prediction signal which is always positive and smaller
+  // than (1 << bd) - 1
+  // since bd < 16-1, therefore we can treat the uint16_t* output buffer as an
+  // int16_t*
+  inv_txfm2d_add_c(input, (int16_t *)output, stride, cfg, txfm_buf);
+  clamp_block((int16_t *)output, 64, stride, 0, (1 << bd) - 1);
+}
diff --git a/vp10/common/vp10_inv_txfm2d.h b/vp10/common/vp10_inv_txfm2d.h
index 1b570ef..fd4b23c 100644
--- a/vp10/common/vp10_inv_txfm2d.h
+++ b/vp10/common/vp10_inv_txfm2d.h
@@ -27,6 +27,9 @@
 void vp10_inv_txfm2d_add_32x32(const int32_t *input, uint16_t *output,
                                const int stride, const TXFM_2D_CFG *cfg,
                                const int bd);
+void vp10_inv_txfm2d_add_64x64(const int32_t *input, uint16_t *output,
+                               const int stride, const TXFM_2D_CFG *cfg,
+                               const int bd);
 #ifdef __cplusplus
 }
 #endif
diff --git a/vp10/common/vp10_inv_txfm2d_cfg.h b/vp10/common/vp10_inv_txfm2d_cfg.h
index fc552fe..a1c4892 100644
--- a/vp10/common/vp10_inv_txfm2d_cfg.h
+++ b/vp10/common/vp10_inv_txfm2d_cfg.h
@@ -96,6 +96,29 @@
     vp10_idct32_new,                 // .txfm_func_col
     vp10_idct32_new};                // .txfm_func_row;
 
+//  ---------------- config inv_dct_dct_64 ----------------
+static int8_t inv_shift_dct_dct_64[2] = {-1, -7};
+static int8_t inv_stage_range_col_dct_dct_64[12] = {19, 19, 19, 19, 19, 19,
+                                                    19, 19, 19, 19, 18, 18};
+static int8_t inv_stage_range_row_dct_dct_64[12] = {20, 20, 20, 20, 20, 20,
+                                                    20, 20, 20, 20, 20, 20};
+static int8_t inv_cos_bit_col_dct_dct_64[12] = {13, 13, 13, 13, 13, 13,
+                                                13, 13, 13, 13, 13, 14};
+static int8_t inv_cos_bit_row_dct_dct_64[12] = {12, 12, 12, 12, 12, 12,
+                                                12, 12, 12, 12, 12, 12};
+
+static const TXFM_2D_CFG inv_txfm_2d_cfg_dct_dct_64 = {
+    64,                              // .txfm_size
+    12,                              // .stage_num_col
+    12,                              // .stage_num_row
+    inv_shift_dct_dct_64,            // .shift
+    inv_stage_range_col_dct_dct_64,  // .stage_range_col
+    inv_stage_range_row_dct_dct_64,  // .stage_range_row
+    inv_cos_bit_col_dct_dct_64,      // .cos_bit_col
+    inv_cos_bit_row_dct_dct_64,      // .cos_bit_row
+    vp10_idct64_new,                 // .txfm_func_col
+    vp10_idct64_new};                // .txfm_func_row;
+
 //  ---------------- config inv_dct_adst_4 ----------------
 static const int8_t inv_shift_dct_adst_4[2] = {1, -5};
 static const int8_t inv_stage_range_col_dct_adst_4[4] = {17, 17, 16, 16};