diff --git a/test/fdct4x4_test.cc b/test/fdct4x4_test.cc
index 0c91aee..59ce895 100644
--- a/test/fdct4x4_test.cc
+++ b/test/fdct4x4_test.cc
@@ -19,8 +19,8 @@
 #include "test/acm_random.h"
 #include "test/clear_system_state.h"
 #include "test/register_state_check.h"
+#include "test/transform_test_base.h"
 #include "test/util.h"
-#include "vp9/common/vp9_entropy.h"
 #include "vpx/vpx_codec.h"
 #include "vpx/vpx_integer.h"
 #include "vpx_ports/mem.h"
@@ -28,16 +28,16 @@
 using libvpx_test::ACMRandom;
 
 namespace {
-const int kNumCoeffs = 16;
 typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
 typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
-typedef void (*FhtFunc)(const int16_t *in, tran_low_t *out, int stride,
-                        int tx_type);
 typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
                         int tx_type);
+using libvpx_test::FhtFunc;
 
-typedef std::tr1::tuple<FdctFunc, IdctFunc, int, vpx_bit_depth_t> Dct4x4Param;
-typedef std::tr1::tuple<FhtFunc, IhtFunc, int, vpx_bit_depth_t> Ht4x4Param;
+typedef std::tr1::tuple<FdctFunc, IdctFunc, int, vpx_bit_depth_t, int>
+Dct4x4Param;
+typedef std::tr1::tuple<FhtFunc, IhtFunc, int, vpx_bit_depth_t, int>
+Ht4x4Param;
 
 void fdct4x4_ref(const int16_t *in, tran_low_t *out, int stride,
                  int /*tx_type*/) {
@@ -89,197 +89,9 @@
 #endif  // HAVE_SSE2
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
-class Trans4x4TestBase {
- public:
-  virtual ~Trans4x4TestBase() {}
-
- protected:
-  virtual void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) = 0;
-
-  virtual void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) = 0;
-
-  void RunAccuracyCheck(int limit) {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    uint32_t max_error = 0;
-    int64_t total_error = 0;
-    const int count_test_block = 10000;
-    for (int i = 0; i < count_test_block; ++i) {
-      DECLARE_ALIGNED(16, int16_t, test_input_block[kNumCoeffs]);
-      DECLARE_ALIGNED(16, tran_low_t, test_temp_block[kNumCoeffs]);
-      DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-      DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
-#if CONFIG_VP9_HIGHBITDEPTH
-      DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-      DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
-#endif
-
-      // Initialize a test block with input range [-255, 255].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        if (bit_depth_ == VPX_BITS_8) {
-          src[j] = rnd.Rand8();
-          dst[j] = rnd.Rand8();
-          test_input_block[j] = src[j] - dst[j];
-#if CONFIG_VP9_HIGHBITDEPTH
-        } else {
-          src16[j] = rnd.Rand16() & mask_;
-          dst16[j] = rnd.Rand16() & mask_;
-          test_input_block[j] = src16[j] - dst16[j];
-#endif
-        }
-      }
-
-      ASM_REGISTER_STATE_CHECK(RunFwdTxfm(test_input_block,
-                                          test_temp_block, pitch_));
-      if (bit_depth_ == VPX_BITS_8) {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block, dst, pitch_));
-#if CONFIG_VP9_HIGHBITDEPTH
-      } else {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block,
-                                            CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif
-      }
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_VP9_HIGHBITDEPTH
-        const uint32_t diff =
-            bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-        ASSERT_EQ(VPX_BITS_8, bit_depth_);
-        const uint32_t diff = dst[j] - src[j];
-#endif
-        const uint32_t error = diff * diff;
-        if (max_error < error)
-          max_error = error;
-        total_error += error;
-      }
-    }
-
-    EXPECT_GE(static_cast<uint32_t>(limit), max_error)
-        << "Error: 4x4 FHT/IHT has an individual round trip error > "
-        << limit;
-
-    EXPECT_GE(count_test_block * limit, total_error)
-        << "Error: 4x4 FHT/IHT has average round trip error > " << limit
-        << " per block";
-  }
-
-  void RunCoeffCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 5000;
-    DECLARE_ALIGNED(16, int16_t, input_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j)
-        input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
-
-      fwd_txfm_ref(input_block, output_ref_block, pitch_, tx_type_);
-      ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_block, output_block, pitch_));
-
-      // The minimum quant value is 4.
-      for (int j = 0; j < kNumCoeffs; ++j)
-        EXPECT_EQ(output_block[j], output_ref_block[j]);
-    }
-  }
-
-  void RunMemCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 5000;
-    DECLARE_ALIGNED(16, int16_t, input_extreme_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        input_extreme_block[j] = rnd.Rand8() % 2 ? mask_ : -mask_;
-      }
-      if (i == 0) {
-        for (int j = 0; j < kNumCoeffs; ++j)
-          input_extreme_block[j] = mask_;
-      } else if (i == 1) {
-        for (int j = 0; j < kNumCoeffs; ++j)
-          input_extreme_block[j] = -mask_;
-      }
-
-      fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, tx_type_);
-      ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_extreme_block,
-                                          output_block, pitch_));
-
-      // The minimum quant value is 4.
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        EXPECT_EQ(output_block[j], output_ref_block[j]);
-        EXPECT_GE(4 * DCT_MAX_VALUE << (bit_depth_ - 8), abs(output_block[j]))
-            << "Error: 4x4 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
-      }
-    }
-  }
-
-  void RunInvAccuracyCheck(int limit) {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 1000;
-    DECLARE_ALIGNED(16, int16_t, in[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
-#if CONFIG_VP9_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
-#endif
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        if (bit_depth_ == VPX_BITS_8) {
-          src[j] = rnd.Rand8();
-          dst[j] = rnd.Rand8();
-          in[j] = src[j] - dst[j];
-#if CONFIG_VP9_HIGHBITDEPTH
-        } else {
-          src16[j] = rnd.Rand16() & mask_;
-          dst16[j] = rnd.Rand16() & mask_;
-          in[j] = src16[j] - dst16[j];
-#endif
-        }
-      }
-
-      fwd_txfm_ref(in, coeff, pitch_, tx_type_);
-
-      if (bit_depth_ == VPX_BITS_8) {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
-#if CONFIG_VP9_HIGHBITDEPTH
-      } else {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16),
-                                            pitch_));
-#endif
-      }
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_VP9_HIGHBITDEPTH
-        const uint32_t diff =
-            bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-        const uint32_t diff = dst[j] - src[j];
-#endif
-        const uint32_t error = diff * diff;
-        EXPECT_GE(static_cast<uint32_t>(limit), error)
-            << "Error: 4x4 IDCT has error " << error
-            << " at index " << j;
-      }
-    }
-  }
-
-  int pitch_;
-  int tx_type_;
-  FhtFunc fwd_txfm_ref;
-  vpx_bit_depth_t bit_depth_;
-  int mask_;
-};
 
 class Trans4x4DCT
-    : public Trans4x4TestBase,
+    : public libvpx_test::TransformTestBase,
       public ::testing::TestWithParam<Dct4x4Param> {
  public:
   virtual ~Trans4x4DCT() {}
@@ -292,6 +104,7 @@
     fwd_txfm_ref = fdct4x4_ref;
     bit_depth_ = GET_PARAM(3);
     mask_ = (1 << bit_depth_) - 1;
+    num_coeffs_ = GET_PARAM(4);
   }
   virtual void TearDown() { libvpx_test::ClearSystemState(); }
 
@@ -324,7 +137,7 @@
 }
 
 class Trans4x4HT
-    : public Trans4x4TestBase,
+    : public libvpx_test::TransformTestBase,
       public ::testing::TestWithParam<Ht4x4Param> {
  public:
   virtual ~Trans4x4HT() {}
@@ -337,6 +150,7 @@
     fwd_txfm_ref = fht4x4_ref;
     bit_depth_ = GET_PARAM(3);
     mask_ = (1 << bit_depth_) - 1;
+    num_coeffs_ = GET_PARAM(4);
   }
   virtual void TearDown() { libvpx_test::ClearSystemState(); }
 
@@ -370,7 +184,7 @@
 }
 
 class Trans4x4WHT
-    : public Trans4x4TestBase,
+    : public libvpx_test::TransformTestBase,
       public ::testing::TestWithParam<Dct4x4Param> {
  public:
   virtual ~Trans4x4WHT() {}
@@ -383,6 +197,7 @@
     fwd_txfm_ref = fwht4x4_ref;
     bit_depth_ = GET_PARAM(3);
     mask_ = (1 << bit_depth_) - 1;
+    num_coeffs_ = GET_PARAM(4);
   }
   virtual void TearDown() { libvpx_test::ClearSystemState(); }
 
@@ -419,54 +234,54 @@
 INSTANTIATE_TEST_CASE_P(
     C, Trans4x4DCT,
     ::testing::Values(
-        make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10),
-        make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12),
-        make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8)));
+        make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10, 16),
+        make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12, 16),
+        make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8, 16)));
 #else
 INSTANTIATE_TEST_CASE_P(
     C, Trans4x4DCT,
     ::testing::Values(
-        make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8)));
+         make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8, 16)));
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
 #if CONFIG_VP9_HIGHBITDEPTH
 INSTANTIATE_TEST_CASE_P(
     C, Trans4x4HT,
     ::testing::Values(
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 0, VPX_BITS_10),
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 1, VPX_BITS_10),
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 2, VPX_BITS_10),
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 3, VPX_BITS_10),
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 0, VPX_BITS_12),
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 1, VPX_BITS_12),
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 2, VPX_BITS_12),
-        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 3, VPX_BITS_12),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8)));
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 0, VPX_BITS_10, 16),
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 1, VPX_BITS_10, 16),
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 2, VPX_BITS_10, 16),
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 3, VPX_BITS_10, 16),
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 0, VPX_BITS_12, 16),
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 1, VPX_BITS_12, 16),
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 2, VPX_BITS_12, 16),
+        make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 3, VPX_BITS_12, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8, 16)));
 #else
 INSTANTIATE_TEST_CASE_P(
     C, Trans4x4HT,
     ::testing::Values(
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8)));
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8, 16)));
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
 #if CONFIG_VP9_HIGHBITDEPTH
 INSTANTIATE_TEST_CASE_P(
     C, Trans4x4WHT,
     ::testing::Values(
-        make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_10, 0, VPX_BITS_10),
-        make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_12, 0, VPX_BITS_12),
-        make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8)));
+        make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_10, 0, VPX_BITS_10, 16),
+        make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_12, 0, VPX_BITS_12, 16),
+        make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8, 16)));
 #else
 INSTANTIATE_TEST_CASE_P(
     C, Trans4x4WHT,
     ::testing::Values(
-        make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8)));
+        make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8, 16)));
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
 #if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -474,17 +289,17 @@
     NEON, Trans4x4DCT,
     ::testing::Values(
         make_tuple(&vpx_fdct4x4_c,
-                   &vpx_idct4x4_16_add_neon, 0, VPX_BITS_8)));
+                   &vpx_idct4x4_16_add_neon, 0, VPX_BITS_8, 16)));
 #endif  // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 INSTANTIATE_TEST_CASE_P(
     NEON, Trans4x4HT,
     ::testing::Values(
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 0, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 1, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 2, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 3, VPX_BITS_8)));
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 0, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 1, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 2, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 3, VPX_BITS_8, 16)));
 #endif  // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if CONFIG_USE_X86INC && HAVE_MMX && !CONFIG_VP9_HIGHBITDEPTH && \
@@ -492,7 +307,8 @@
 INSTANTIATE_TEST_CASE_P(
     MMX, Trans4x4WHT,
     ::testing::Values(
-        make_tuple(&vp9_fwht4x4_mmx, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8)));
+        make_tuple(&vp9_fwht4x4_mmx, &vpx_iwht4x4_16_add_c, 0,
+                   VPX_BITS_8, 16)));
 #endif
 
 #if CONFIG_USE_X86INC && HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && \
@@ -500,7 +316,8 @@
 INSTANTIATE_TEST_CASE_P(
     SSE2, Trans4x4WHT,
     ::testing::Values(
-        make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_sse2, 0, VPX_BITS_8)));
+        make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_sse2, 0,
+                   VPX_BITS_8, 16)));
 #endif
 
 #if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -508,47 +325,60 @@
     SSE2, Trans4x4DCT,
     ::testing::Values(
         make_tuple(&vpx_fdct4x4_sse2,
-                   &vpx_idct4x4_16_add_sse2, 0, VPX_BITS_8)));
+                   &vpx_idct4x4_16_add_sse2, 0, VPX_BITS_8, 16)));
 INSTANTIATE_TEST_CASE_P(
     SSE2, Trans4x4HT,
     ::testing::Values(
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 0, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 1, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 2, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 3, VPX_BITS_8)));
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 0,
+                   VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 1,
+                   VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 2,
+                   VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 3,
+                   VPX_BITS_8, 16)));
 #endif  // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 INSTANTIATE_TEST_CASE_P(
     SSE2, Trans4x4DCT,
     ::testing::Values(
-        make_tuple(&vpx_highbd_fdct4x4_c,    &idct4x4_10_sse2, 0, VPX_BITS_10),
-        make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_10_sse2, 0, VPX_BITS_10),
-        make_tuple(&vpx_highbd_fdct4x4_c,    &idct4x4_12_sse2, 0, VPX_BITS_12),
-        make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_12_sse2, 0, VPX_BITS_12),
+        make_tuple(&vpx_highbd_fdct4x4_c,    &idct4x4_10_sse2, 0,
+                   VPX_BITS_10, 16),
+        make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_10_sse2, 0,
+                   VPX_BITS_10, 16),
+        make_tuple(&vpx_highbd_fdct4x4_c,    &idct4x4_12_sse2, 0,
+                   VPX_BITS_12, 16),
+        make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_12_sse2, 0,
+                   VPX_BITS_12, 16),
         make_tuple(&vpx_fdct4x4_sse2,      &vpx_idct4x4_16_add_c, 0,
-                   VPX_BITS_8)));
+                   VPX_BITS_8, 16)));
 
 INSTANTIATE_TEST_CASE_P(
     SSE2, Trans4x4HT,
     ::testing::Values(
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8)));
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8, 16)));
 #endif  // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 INSTANTIATE_TEST_CASE_P(
     MSA, Trans4x4DCT,
     ::testing::Values(
-        make_tuple(&vpx_fdct4x4_msa, &vpx_idct4x4_16_add_msa, 0, VPX_BITS_8)));
+        make_tuple(&vpx_fdct4x4_msa, &vpx_idct4x4_16_add_msa, 0,
+                   VPX_BITS_8, 16)));
 INSTANTIATE_TEST_CASE_P(
     MSA, Trans4x4HT,
     ::testing::Values(
-        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 0, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 1, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 2, VPX_BITS_8),
-        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 3, VPX_BITS_8)));
+        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 0,
+                   VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 1,
+                   VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 2,
+                   VPX_BITS_8, 16),
+        make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 3,
+                   VPX_BITS_8, 16)));
 #endif  // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 }  // namespace
diff --git a/test/test.mk b/test/test.mk
index a73ebd9..c6ff793 100644
--- a/test/test.mk
+++ b/test/test.mk
@@ -9,6 +9,7 @@
 LIBVPX_TEST_SRCS-yes += test_vectors.h
 LIBVPX_TEST_SRCS-yes += util.h
 LIBVPX_TEST_SRCS-yes += video_source.h
+LIBVPX_TEST_SRCS-yes += transform_test_base.h
 
 ##
 ## BLACK BOX TESTS
@@ -165,6 +166,7 @@
 ifeq ($(CONFIG_VP10),yes)
 LIBVPX_TEST_SRCS-yes                    += vp10_inv_txfm_test.cc
 LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += vp10_dct_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += vp10_fht4x4_test.cc
 LIBVPX_TEST_SRCS-$(CONFIG_ANS)          += vp10_ans_test.cc
 
 LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += sum_squares_test.cc
diff --git a/test/transform_test_base.h b/test/transform_test_base.h
new file mode 100644
index 0000000..cf2facd
--- /dev/null
+++ b/test/transform_test_base.h
@@ -0,0 +1,291 @@
+/*
+ *  Copyright (c) 2016 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+#ifndef TEST_TRANSFORM_TEST_BASE_H_
+#define TEST_TRANSFORM_TEST_BASE_H_
+
+#include "./vpx_config.h"
+#include "vpx_mem/vpx_mem.h"
+#include "vpx/vpx_codec.h"
+
+
+namespace libvpx_test {
+
+//  Note:
+//   Same constant are defined in vp9/common/vp9_entropy.h and
+//   vp10/common/entropy.h.  Goal is to make this base class
+//   to use for future codec transform testing.  But including
+//   either of them would lead to compiling error when we do
+//   unit test for another codec. Suggest to move the definition
+//   to a vpx header file.
+const int kDctMaxValue = 16384;
+
+typedef void (*FhtFunc)(const int16_t *in, tran_low_t *out, int stride,
+                        int tx_type);
+
+class TransformTestBase {
+ public:
+  virtual ~TransformTestBase() {}
+
+ protected:
+  virtual void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) = 0;
+
+  virtual void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) = 0;
+
+  void RunAccuracyCheck(int limit) {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    uint32_t max_error = 0;
+    int64_t total_error = 0;
+    const int count_test_block = 10000;
+
+    int16_t *test_input_block = reinterpret_cast<int16_t *>
+        (vpx_memalign(16, sizeof(int16_t) * num_coeffs_));
+    tran_low_t *test_temp_block = reinterpret_cast<tran_low_t *>
+        (vpx_memalign(16, sizeof(tran_low_t) * num_coeffs_));
+    uint8_t *dst = reinterpret_cast<uint8_t *>
+        (vpx_memalign(16, sizeof(uint8_t) * num_coeffs_));
+    uint8_t *src = reinterpret_cast<uint8_t *>
+        (vpx_memalign(16, sizeof(uint8_t) * num_coeffs_));
+#if CONFIG_VP9_HIGHBITDEPTH
+    uint16_t *dst16 = reinterpret_cast<uint16_t *>
+        (vpx_memalign(16, sizeof(uint16_t) * num_coeffs_));
+    uint16_t *src16 = reinterpret_cast<uint16_t *>
+        (vpx_memalign(16, sizeof(uint16_t) * num_coeffs_));
+#endif
+
+    for (int i = 0; i < count_test_block; ++i) {
+      // Initialize a test block with input range [-255, 255].
+      for (int j = 0; j < num_coeffs_; ++j) {
+        if (bit_depth_ == VPX_BITS_8) {
+          src[j] = rnd.Rand8();
+          dst[j] = rnd.Rand8();
+          test_input_block[j] = src[j] - dst[j];
+#if CONFIG_VP9_HIGHBITDEPTH
+        } else {
+          src16[j] = rnd.Rand16() & mask_;
+          dst16[j] = rnd.Rand16() & mask_;
+          test_input_block[j] = src16[j] - dst16[j];
+#endif
+        }
+      }
+
+      ASM_REGISTER_STATE_CHECK(RunFwdTxfm(test_input_block,
+                                          test_temp_block, pitch_));
+      if (bit_depth_ == VPX_BITS_8) {
+        ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block, dst, pitch_));
+#if CONFIG_VP9_HIGHBITDEPTH
+      } else {
+        ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block,
+                                            CONVERT_TO_BYTEPTR(dst16), pitch_));
+#endif
+      }
+
+      for (int j = 0; j < num_coeffs_; ++j) {
+#if CONFIG_VP9_HIGHBITDEPTH
+        const uint32_t diff =
+            bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
+#else
+        ASSERT_EQ(VPX_BITS_8, bit_depth_);
+        const uint32_t diff = dst[j] - src[j];
+#endif
+        const uint32_t error = diff * diff;
+        if (max_error < error)
+          max_error = error;
+        total_error += error;
+      }
+    }
+
+    EXPECT_GE(static_cast<uint32_t>(limit), max_error)
+        << "Error: 4x4 FHT/IHT has an individual round trip error > "
+        << limit;
+
+    EXPECT_GE(count_test_block * limit, total_error)
+        << "Error: 4x4 FHT/IHT has average round trip error > " << limit
+        << " per block";
+
+    vpx_free(test_input_block);
+    vpx_free(test_temp_block);
+    vpx_free(dst);
+    vpx_free(src);
+#if CONFIG_VP9_HIGHBITDEPTH
+    vpx_free(dst16);
+    vpx_free(src16);
+#endif
+  }
+
+  void RunCoeffCheck() {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    const int count_test_block = 5000;
+
+    int16_t *input_block = reinterpret_cast<int16_t *>
+        (vpx_memalign(16, sizeof(int16_t) * num_coeffs_));
+    tran_low_t *output_ref_block = reinterpret_cast<tran_low_t *>
+        (vpx_memalign(16, sizeof(tran_low_t) * num_coeffs_));
+    tran_low_t *output_block = reinterpret_cast<tran_low_t *>
+        (vpx_memalign(16, sizeof(tran_low_t) * num_coeffs_));
+
+    for (int i = 0; i < count_test_block; ++i) {
+      // Initialize a test block with input range [-mask_, mask_].
+      for (int j = 0; j < num_coeffs_; ++j)
+        input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
+
+      fwd_txfm_ref(input_block, output_ref_block, pitch_, tx_type_);
+      ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_block, output_block, pitch_));
+
+      // The minimum quant value is 4.
+      for (int j = 0; j < num_coeffs_; ++j) {
+        EXPECT_EQ(output_block[j], output_ref_block[j])
+            << "Error: not bit-exact result at index: " << j
+            << " at test block: " << i;
+      }
+    }
+    vpx_free(input_block);
+    vpx_free(output_ref_block);
+    vpx_free(output_block);
+  }
+
+  void RunMemCheck() {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    const int count_test_block = 5000;
+
+    int16_t *input_extreme_block = reinterpret_cast<int16_t *>
+        (vpx_memalign(16, sizeof(int16_t) * num_coeffs_));
+    tran_low_t *output_ref_block = reinterpret_cast<tran_low_t *>
+        (vpx_memalign(16, sizeof(tran_low_t) * num_coeffs_));
+    tran_low_t *output_block = reinterpret_cast<tran_low_t *>
+        (vpx_memalign(16, sizeof(tran_low_t) * num_coeffs_));
+
+    for (int i = 0; i < count_test_block; ++i) {
+      // Initialize a test block with input range [-mask_, mask_].
+      for (int j = 0; j < num_coeffs_; ++j) {
+        input_extreme_block[j] = rnd.Rand8() % 2 ? mask_ : -mask_;
+      }
+      if (i == 0) {
+        for (int j = 0; j < num_coeffs_; ++j)
+          input_extreme_block[j] = mask_;
+      } else if (i == 1) {
+        for (int j = 0; j < num_coeffs_; ++j)
+          input_extreme_block[j] = -mask_;
+      }
+
+      fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, tx_type_);
+      ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_extreme_block,
+                                          output_block, pitch_));
+
+      int row_length = FindRowLength();
+      // The minimum quant value is 4.
+      for (int j = 0; j < num_coeffs_; ++j) {
+        EXPECT_EQ(output_block[j], output_ref_block[j]);
+        EXPECT_GE(row_length * kDctMaxValue << (bit_depth_ - 8),
+                  abs(output_block[j]))
+            << "Error: NxN FDCT has coefficient larger than N*DCT_MAX_VALUE";
+      }
+    }
+    vpx_free(input_extreme_block);
+    vpx_free(output_ref_block);
+    vpx_free(output_block);
+  }
+
+  void RunInvAccuracyCheck(int limit) {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    const int count_test_block = 1000;
+
+    int16_t *in = reinterpret_cast<int16_t *>
+        (vpx_memalign(16, sizeof(int16_t) * num_coeffs_));
+    tran_low_t *coeff = reinterpret_cast<tran_low_t *>
+        (vpx_memalign(16, sizeof(tran_low_t) * num_coeffs_));
+    uint8_t *dst = reinterpret_cast<uint8_t *>
+        (vpx_memalign(16, sizeof(uint8_t) * num_coeffs_));
+    uint8_t *src = reinterpret_cast<uint8_t *>
+        (vpx_memalign(16, sizeof(uint8_t) * num_coeffs_));
+
+#if CONFIG_VP9_HIGHBITDEPTH
+    uint16_t *dst16 = reinterpret_cast<uint16_t *>
+        (vpx_memalign(16, sizeof(uint16_t) * num_coeffs_));
+    uint16_t *src16 = reinterpret_cast<uint16_t *>
+        (vpx_memalign(16, sizeof(uint16_t) * num_coeffs_));
+#endif
+
+    for (int i = 0; i < count_test_block; ++i) {
+      // Initialize a test block with input range [-mask_, mask_].
+      for (int j = 0; j < num_coeffs_; ++j) {
+        if (bit_depth_ == VPX_BITS_8) {
+          src[j] = rnd.Rand8();
+          dst[j] = rnd.Rand8();
+          in[j] = src[j] - dst[j];
+#if CONFIG_VP9_HIGHBITDEPTH
+        } else {
+          src16[j] = rnd.Rand16() & mask_;
+          dst16[j] = rnd.Rand16() & mask_;
+          in[j] = src16[j] - dst16[j];
+#endif
+        }
+      }
+
+      fwd_txfm_ref(in, coeff, pitch_, tx_type_);
+
+      if (bit_depth_ == VPX_BITS_8) {
+        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
+#if CONFIG_VP9_HIGHBITDEPTH
+      } else {
+        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16),
+                                            pitch_));
+#endif
+      }
+
+      for (int j = 0; j < num_coeffs_; ++j) {
+#if CONFIG_VP9_HIGHBITDEPTH
+        const uint32_t diff =
+            bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
+#else
+        const uint32_t diff = dst[j] - src[j];
+#endif
+        const uint32_t error = diff * diff;
+        EXPECT_GE(static_cast<uint32_t>(limit), error)
+            << "Error: 4x4 IDCT has error " << error
+            << " at index " << j;
+      }
+    }
+    vpx_free(in);
+    vpx_free(coeff);
+    vpx_free(dst);
+    vpx_free(src);
+#if CONFIG_VP9_HIGHBITDEPTH
+    vpx_free(src16);
+    vpx_free(dst16);
+#endif
+  }
+
+  int pitch_;
+  int tx_type_;
+  FhtFunc fwd_txfm_ref;
+  vpx_bit_depth_t bit_depth_;
+  int mask_;
+  int num_coeffs_;
+
+ private:
+  //  Assume transform size is 4x4, 8x8, 16x16,...
+  int FindRowLength() const {
+    int row = 4;
+    if (16 == num_coeffs_) {
+      row = 4;
+    } else if (64 == num_coeffs_) {
+      row = 8;
+    } else if (256 == num_coeffs_) {
+      row = 16;
+    } else if (1024 == num_coeffs_) {
+      row = 32;
+    }
+    return row;
+  }
+};
+
+}  // namespace libvpx_test
+
+#endif  // TEST_TRANSFORM_TEST_BASE_H_
diff --git a/test/vp10_fht4x4_test.cc b/test/vp10_fht4x4_test.cc
new file mode 100644
index 0000000..d2598f9
--- /dev/null
+++ b/test/vp10_fht4x4_test.cc
@@ -0,0 +1,123 @@
+/*
+ *  Copyright (c) 2016 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+#include "./vp10_rtcd.h"
+#include "./vpx_dsp_rtcd.h"
+
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/transform_test_base.h"
+#include "test/util.h"
+#include "vpx_ports/mem.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {
+typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
+                        int tx_type);
+
+using libvpx_test::FhtFunc;
+typedef std::tr1::tuple<FhtFunc, IhtFunc, int, vpx_bit_depth_t, int> Ht4x4Param;
+
+void fht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
+                int tx_type) {
+  vp10_fht4x4_c(in, out, stride, tx_type);
+}
+
+class VP10Trans4x4HT
+    : public libvpx_test::TransformTestBase,
+      public ::testing::TestWithParam<Ht4x4Param> {
+ public:
+  virtual ~VP10Trans4x4HT() {}
+
+  virtual void SetUp() {
+    fwd_txfm_ = GET_PARAM(0);
+    inv_txfm_ = GET_PARAM(1);
+    tx_type_  = GET_PARAM(2);
+    pitch_    = 4;
+    fwd_txfm_ref = fht4x4_ref;
+    bit_depth_ = GET_PARAM(3);
+    mask_ = (1 << bit_depth_) - 1;
+    num_coeffs_ = GET_PARAM(4);
+  }
+  virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
+ protected:
+  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
+    fwd_txfm_(in, out, stride, tx_type_);
+  }
+
+  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
+    inv_txfm_(out, dst, stride, tx_type_);
+  }
+
+  FhtFunc fwd_txfm_;
+  IhtFunc inv_txfm_;
+};
+
+TEST_P(VP10Trans4x4HT, CoeffCheck) {
+  RunCoeffCheck();
+}
+
+using std::tr1::make_tuple;
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(
+    SSE2, VP10Trans4x4HT,
+    ::testing::Values(
+#if !CONFIG_EXT_TX
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 0,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 1,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 2,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 3,
+                 VPX_BITS_8, 16)));
+#else
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 0,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 1,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 2,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 3,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 4,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 5,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 6,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 7,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 8,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 9,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 10,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 11,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 12,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 13,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 14,
+                 VPX_BITS_8, 16),
+      make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 15,
+                 VPX_BITS_8, 16)));
+#endif  // !CONFIG_EXT_TX
+#endif  // HAVE_SSE2
+
+}  // namespace