From: Jingning Han <jingning@google.com>
Date: Thu, 5 Sep 2013 19:44:03 +0000 (-0700)
Subject: Rework 16x16 transform unit test
X-Git-Tag: v1.3.0~467^2
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=8f92a7efdb24db2ef6d607a4711ab57a786411fc;p=libvpx

Rework 16x16 transform unit test

This commit refactors the 16x16 transform unit test. It enables the
test on all implemented versions of forward and inverse 16x16 transform
modules.

Change-Id: I0c7d5f3c5fdd5d789a25f73e287aeeaf463b9d69
---

diff --git a/test/dct16x16_test.cc b/test/dct16x16_test.cc
index 544d18840..3764ade75 100644
--- a/test/dct16x16_test.cc
+++ b/test/dct16x16_test.cc
@@ -13,15 +13,16 @@
 #include <string.h>
 
 #include "third_party/googletest/src/include/gtest/gtest.h"
-#include "vpx_ports/mem.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
 
 extern "C" {
 #include "vp9/common/vp9_entropy.h"
 #include "./vp9_rtcd.h"
 void vp9_short_idct16x16_add_c(int16_t *input, uint8_t *output, int pitch);
 }
-
-#include "test/acm_random.h"
 #include "vpx/vpx_integer.h"
 
 using libvpx_test::ACMRandom;
@@ -37,6 +38,7 @@ static int round(double x) {
 }
 #endif
 
+const int kNumCoeffs = 256;
 const double PI = 3.1415926535898;
 void reference2_16x16_idct_2d(double *input, double *output) {
   double x;
@@ -61,21 +63,21 @@ void reference2_16x16_idct_2d(double *input, double *output) {
 }
 
 
-static const double C1 = 0.995184726672197;
-static const double C2 = 0.98078528040323;
-static const double C3 = 0.956940335732209;
-static const double C4 = 0.923879532511287;
-static const double C5 = 0.881921264348355;
-static const double C6 = 0.831469612302545;
-static const double C7 = 0.773010453362737;
-static const double C8 = 0.707106781186548;
-static const double C9 = 0.634393284163646;
-static const double C10 = 0.555570233019602;
-static const double C11 = 0.471396736825998;
-static const double C12 = 0.38268343236509;
-static const double C13 = 0.290284677254462;
-static const double C14 = 0.195090322016128;
-static const double C15 = 0.098017140329561;
+const double C1 = 0.995184726672197;
+const double C2 = 0.98078528040323;
+const double C3 = 0.956940335732209;
+const double C4 = 0.923879532511287;
+const double C5 = 0.881921264348355;
+const double C6 = 0.831469612302545;
+const double C7 = 0.773010453362737;
+const double C8 = 0.707106781186548;
+const double C9 = 0.634393284163646;
+const double C10 = 0.555570233019602;
+const double C11 = 0.471396736825998;
+const double C12 = 0.38268343236509;
+const double C13 = 0.290284677254462;
+const double C14 = 0.195090322016128;
+const double C15 = 0.098017140329561;
 
 void butterfly_16x16_dct_1d(double input[16], double output[16]) {
   double step[16];
@@ -110,36 +112,36 @@ void butterfly_16x16_dct_1d(double input[16], double output[16]) {
   output[6] = step[1] - step[6];
   output[7] = step[0] - step[7];
 
-  temp1 = step[ 8]*C7;
-  temp2 = step[15]*C9;
+  temp1 = step[ 8] * C7;
+  temp2 = step[15] * C9;
   output[ 8] = temp1 + temp2;
 
-  temp1 = step[ 9]*C11;
-  temp2 = step[14]*C5;
+  temp1 = step[ 9] * C11;
+  temp2 = step[14] * C5;
   output[ 9] = temp1 - temp2;
 
-  temp1 = step[10]*C3;
-  temp2 = step[13]*C13;
+  temp1 = step[10] * C3;
+  temp2 = step[13] * C13;
   output[10] = temp1 + temp2;
 
-  temp1 = step[11]*C15;
-  temp2 = step[12]*C1;
+  temp1 = step[11] * C15;
+  temp2 = step[12] * C1;
   output[11] = temp1 - temp2;
 
-  temp1 = step[11]*C1;
-  temp2 = step[12]*C15;
+  temp1 = step[11] * C1;
+  temp2 = step[12] * C15;
   output[12] = temp2 + temp1;
 
-  temp1 = step[10]*C13;
-  temp2 = step[13]*C3;
+  temp1 = step[10] * C13;
+  temp2 = step[13] * C3;
   output[13] = temp2 - temp1;
 
-  temp1 = step[ 9]*C5;
-  temp2 = step[14]*C11;
+  temp1 = step[ 9] * C5;
+  temp2 = step[14] * C11;
   output[14] = temp2 + temp1;
 
-  temp1 = step[ 8]*C9;
-  temp2 = step[15]*C7;
+  temp1 = step[ 8] * C9;
+  temp2 = step[15] * C7;
   output[15] = temp2 - temp1;
 
   // step 3
@@ -148,20 +150,20 @@ void butterfly_16x16_dct_1d(double input[16], double output[16]) {
   step[ 2] = output[1] - output[2];
   step[ 3] = output[0] - output[3];
 
-  temp1 = output[4]*C14;
-  temp2 = output[7]*C2;
+  temp1 = output[4] * C14;
+  temp2 = output[7] * C2;
   step[ 4] = temp1 + temp2;
 
-  temp1 = output[5]*C10;
-  temp2 = output[6]*C6;
+  temp1 = output[5] * C10;
+  temp2 = output[6] * C6;
   step[ 5] = temp1 + temp2;
 
-  temp1 = output[5]*C6;
-  temp2 = output[6]*C10;
+  temp1 = output[5] * C6;
+  temp2 = output[6] * C10;
   step[ 6] = temp2 - temp1;
 
-  temp1 = output[4]*C2;
-  temp2 = output[7]*C14;
+  temp1 = output[4] * C2;
+  temp2 = output[7] * C14;
   step[ 7] = temp2 - temp1;
 
   step[ 8] = output[ 8] + output[11];
@@ -178,18 +180,18 @@ void butterfly_16x16_dct_1d(double input[16], double output[16]) {
   output[ 0] = (step[ 0] + step[ 1]);
   output[ 8] = (step[ 0] - step[ 1]);
 
-  temp1 = step[2]*C12;
-  temp2 = step[3]*C4;
+  temp1 = step[2] * C12;
+  temp2 = step[3] * C4;
   temp1 = temp1 + temp2;
-  output[ 4] = 2*(temp1*C8);
+  output[ 4] = 2*(temp1 * C8);
 
-  temp1 = step[2]*C4;
-  temp2 = step[3]*C12;
+  temp1 = step[2] * C4;
+  temp2 = step[3] * C12;
   temp1 = temp2 - temp1;
-  output[12] = 2*(temp1*C8);
+  output[12] = 2 * (temp1 * C8);
 
-  output[ 2] = 2*((step[4] + step[ 5])*C8);
-  output[14] = 2*((step[7] - step[ 6])*C8);
+  output[ 2] = 2 * ((step[4] + step[ 5]) * C8);
+  output[14] = 2 * ((step[7] - step[ 6]) * C8);
 
   temp1 = step[4] - step[5];
   temp2 = step[6] + step[7];
@@ -199,17 +201,17 @@ void butterfly_16x16_dct_1d(double input[16], double output[16]) {
   intermediate[8] = step[8] + step[14];
   intermediate[9] = step[9] + step[15];
 
-  temp1 = intermediate[8]*C12;
-  temp2 = intermediate[9]*C4;
+  temp1 = intermediate[8] * C12;
+  temp2 = intermediate[9] * C4;
   temp1 = temp1 - temp2;
-  output[3] = 2*(temp1*C8);
+  output[3] = 2 * (temp1 * C8);
 
-  temp1 = intermediate[8]*C4;
-  temp2 = intermediate[9]*C12;
+  temp1 = intermediate[8] * C4;
+  temp2 = intermediate[9] * C12;
   temp1 = temp2 + temp1;
-  output[13] = 2*(temp1*C8);
+  output[13] = 2 * (temp1 * C8);
 
-  output[ 9] = 2*((step[10] + step[11])*C8);
+  output[ 9] = 2 * ((step[10] + step[11]) * C8);
 
   intermediate[11] = step[10] - step[11];
   intermediate[12] = step[12] + step[13];
@@ -220,216 +222,263 @@ void butterfly_16x16_dct_1d(double input[16], double output[16]) {
   output[15] = (intermediate[11] + intermediate[12]);
   output[ 1] = -(intermediate[11] - intermediate[12]);
 
-  output[ 7] = 2*(intermediate[13]*C8);
+  output[ 7] = 2 * (intermediate[13] * C8);
 
-  temp1 = intermediate[14]*C12;
-  temp2 = intermediate[15]*C4;
+  temp1 = intermediate[14] * C12;
+  temp2 = intermediate[15] * C4;
   temp1 = temp1 - temp2;
-  output[11] = -2*(temp1*C8);
+  output[11] = -2 * (temp1 * C8);
 
-  temp1 = intermediate[14]*C4;
-  temp2 = intermediate[15]*C12;
+  temp1 = intermediate[14] * C4;
+  temp2 = intermediate[15] * C12;
   temp1 = temp2 + temp1;
-  output[ 5] = 2*(temp1*C8);
+  output[ 5] = 2 * (temp1 * C8);
 }
 
-static void reference_16x16_dct_1d(double in[16], double out[16]) {
-  const double kPi = 3.141592653589793238462643383279502884;
-  const double kInvSqrt2 = 0.707106781186547524400844362104;
-  for (int k = 0; k < 16; k++) {
-    out[k] = 0.0;
-    for (int n = 0; n < 16; n++)
-      out[k] += in[n]*cos(kPi*(2*n+1)*k/32.0);
-    if (k == 0)
-      out[k] = out[k]*kInvSqrt2;
-  }
-}
-
-void reference_16x16_dct_2d(int16_t input[16*16], double output[16*16]) {
+void reference_16x16_dct_2d(int16_t input[256], double output[256]) {
   // First transform columns
   for (int i = 0; i < 16; ++i) {
     double temp_in[16], temp_out[16];
     for (int j = 0; j < 16; ++j)
-      temp_in[j] = input[j*16 + i];
+      temp_in[j] = input[j * 16 + i];
     butterfly_16x16_dct_1d(temp_in, temp_out);
     for (int j = 0; j < 16; ++j)
-      output[j*16 + i] = temp_out[j];
+      output[j * 16 + i] = temp_out[j];
   }
   // Then transform rows
   for (int i = 0; i < 16; ++i) {
     double temp_in[16], temp_out[16];
     for (int j = 0; j < 16; ++j)
-      temp_in[j] = output[j + i*16];
+      temp_in[j] = output[j + i * 16];
     butterfly_16x16_dct_1d(temp_in, temp_out);
     // Scale by some magic number
     for (int j = 0; j < 16; ++j)
-      output[j + i*16] = temp_out[j]/2;
+      output[j + i * 16] = temp_out[j]/2;
   }
 }
 
-void fdct16x16(int16_t *in, int16_t *out, uint8_t* /*dst*/,
-               int stride, int /*tx_type*/) {
-  vp9_short_fdct16x16_c(in, out, stride);
-}
-void idct16x16_add(int16_t* /*in*/, int16_t *out, uint8_t *dst,
-                   int stride, int /*tx_type*/) {
-  vp9_short_idct16x16_add_c(out, dst, stride >> 1);
-}
-void fht16x16(int16_t *in, int16_t *out, uint8_t* /*dst*/,
-              int stride, int tx_type) {
-  // FIXME(jingning): need to test both SSE2 and c
-#if HAVE_SSE2
-  vp9_short_fht16x16_sse2(in, out, stride >> 1, tx_type);
-#else
-  vp9_short_fht16x16_c(in, out, stride >> 1, tx_type);
-#endif
-}
-void iht16x16_add(int16_t* /*in*/, int16_t *out, uint8_t *dst,
-                  int stride, int tx_type) {
-  vp9_short_iht16x16_add_c(out, dst, stride >> 1, tx_type);
-}
+typedef void (*fdct_t)(int16_t *in, int16_t *out, int stride);
+typedef void (*idct_t)(int16_t *in, uint8_t *out, int stride);
+typedef void (*fht_t) (int16_t *in, int16_t *out, int stride, int tx_type);
+typedef void (*iht_t) (int16_t *in, uint8_t *dst, int stride, int tx_type);
 
-class Trans16x16Test : public ::testing::TestWithParam<int> {
+class Trans16x16TestBase {
  public:
-  virtual ~Trans16x16Test() {}
+  virtual ~Trans16x16TestBase() {}
 
-  virtual void SetUp() {
-    tx_type_ = GetParam();
-    if (tx_type_ == 0) {
-      fwd_txfm_ = fdct16x16;
-      inv_txfm_ = idct16x16_add;
-    } else {
-      fwd_txfm_ = fht16x16;
-      inv_txfm_ = iht16x16_add;
+ protected:
+  virtual void RunFwdTxfm(int16_t *in, int16_t *out,
+                          uint8_t *dst, int stride) = 0;
+
+  virtual void RunInvTxfm(int16_t *in, int16_t *out,
+                          uint8_t *dst, int stride) = 0;
+
+  void RunAccuracyCheck() {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    int max_error = 0;
+    int total_error = 0;
+    const int count_test_block = 10000;
+    for (int i = 0; i < count_test_block; ++i) {
+      DECLARE_ALIGNED_ARRAY(16, int16_t, test_input_block, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, int16_t, test_temp_block, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, uint8_t, src, kNumCoeffs);
+
+      for (int j = 0; j < kNumCoeffs; ++j) {
+        src[j] = rnd.Rand8();
+        dst[j] = rnd.Rand8();
+        // Initialize a test block with input range [-255, 255].
+        test_input_block[j] = src[j] - dst[j];
+      }
+
+      const int pitch = 32;
+      REGISTER_STATE_CHECK(RunFwdTxfm(test_input_block, test_temp_block,
+                                      dst, pitch));
+      REGISTER_STATE_CHECK(RunInvTxfm(test_input_block, test_temp_block,
+                                      dst, pitch));
+
+      for (int j = 0; j < kNumCoeffs; ++j) {
+        const int diff = dst[j] - src[j];
+        const int error = diff * diff;
+        if (max_error < error)
+          max_error = error;
+        total_error += error;
+      }
     }
-  }
 
- protected:
-  void RunFwdTxfm(int16_t *in, int16_t *out, uint8_t *dst,
-                  int stride, int tx_type) {
-    (*fwd_txfm_)(in, out, dst, stride, tx_type);
-  }
-  void RunInvTxfm(int16_t *in, int16_t *out, uint8_t *dst,
-                  int stride, int tx_type) {
-    (*inv_txfm_)(in, out, dst, stride, tx_type);
+    EXPECT_GE(1, max_error)
+        << "Error: 16x16 FHT/IHT has an individual round trip error > 1";
+
+    EXPECT_GE(count_test_block , total_error)
+        << "Error: 16x16 FHT/IHT has average round trip error > 1 per block";
   }
 
-  int tx_type_;
-  void (*fwd_txfm_)(int16_t*, int16_t*, uint8_t*, int, int);
-  void (*inv_txfm_)(int16_t*, int16_t*, uint8_t*, int, int);
-};
+  void RunCoeffSizeCheck() {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    const int count_test_block = 1000;
+    for (int i = 0; i < count_test_block; ++i) {
+      DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, int16_t, output_block, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, int16_t, output_extreme_block, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
 
-TEST_P(Trans16x16Test, AccuracyCheck) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  int max_error = 0;
-  int total_error = 0;
-  const int count_test_block = 10000;
-  for (int i = 0; i < count_test_block; ++i) {
-    DECLARE_ALIGNED_ARRAY(16, int16_t, test_input_block, 256);
-    DECLARE_ALIGNED_ARRAY(16, int16_t, test_temp_block, 256);
-    DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, 256);
-    DECLARE_ALIGNED_ARRAY(16, uint8_t, src, 256);
-
-    for (int j = 0; j < 256; ++j) {
-      src[j] = rnd.Rand8();
-      dst[j] = rnd.Rand8();
       // Initialize a test block with input range [-255, 255].
-      test_input_block[j] = src[j] - dst[j];
+      for (int j = 0; j < kNumCoeffs; ++j) {
+        input_block[j] = rnd.Rand8() - rnd.Rand8();
+        input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
+      }
+      if (i == 0)
+        for (int j = 0; j < kNumCoeffs; ++j)
+          input_extreme_block[j] = 255;
+
+      const int pitch = 32;
+      REGISTER_STATE_CHECK(RunFwdTxfm(input_block, output_block, dst, pitch));
+      REGISTER_STATE_CHECK(RunFwdTxfm(input_extreme_block,
+                                      output_extreme_block, dst, pitch));
+
+      // The minimum quant value is 4.
+      for (int j = 0; j < kNumCoeffs; ++j) {
+        EXPECT_GE(4 * DCT_MAX_VALUE, abs(output_block[j]))
+            << "Error: 16x16 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
+        EXPECT_GE(4 * DCT_MAX_VALUE, abs(output_extreme_block[j]))
+            << "Error: 16x16 FDCT extreme has coefficient larger "
+            << "than 4*DCT_MAX_VALUE";
+      }
     }
+  }
+
+  void RunInvAccuracyCheck() {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    const int count_test_block = 1000;
 
-    const int pitch = 32;
-    RunFwdTxfm(test_input_block, test_temp_block, dst, pitch, tx_type_);
-    RunInvTxfm(test_input_block, test_temp_block, dst, pitch, tx_type_);
+    for (int i = 0; i < count_test_block; ++i) {
+      DECLARE_ALIGNED_ARRAY(16, int16_t, in, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, int16_t, coeff, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
+      DECLARE_ALIGNED_ARRAY(16, uint8_t, src, kNumCoeffs);
+      double out_r[kNumCoeffs];
 
-    for (int j = 0; j < 256; ++j) {
-      const int diff = dst[j] - src[j];
-      const int error = diff * diff;
-      if (max_error < error)
-        max_error = error;
-      total_error += error;
+      // Initialize a test block with input range [-255, 255].
+      for (int j = 0; j < kNumCoeffs; ++j) {
+        src[j] = rnd.Rand8();
+        dst[j] = rnd.Rand8();
+        in[j] = src[j] - dst[j];
+      }
+
+      reference_16x16_dct_2d(in, out_r);
+      for (int j = 0; j < kNumCoeffs; ++j)
+        coeff[j] = round(out_r[j]);
+
+      const int pitch = 32;
+      REGISTER_STATE_CHECK(RunInvTxfm(coeff, coeff, dst, pitch));
+
+      for (int j = 0; j < kNumCoeffs; ++j) {
+        const int diff = dst[j] - src[j];
+        const int error = diff * diff;
+        EXPECT_GE(1, error)
+            << "Error: 16x16 IDCT has error " << error
+            << " at index " << j;
+      }
     }
   }
+};
 
-  EXPECT_GE(1, max_error)
-      << "Error: 16x16 FHT/IHT has an individual round trip error > 1";
+class Trans16x16DCT : public Trans16x16TestBase,
+                      public PARAMS(fdct_t, idct_t, int) {
+ public:
+  virtual ~Trans16x16DCT() {}
 
-  EXPECT_GE(count_test_block , total_error)
-      << "Error: 16x16 FHT/IHT has average round trip error > 1 per block";
-}
+  virtual void SetUp() {
+    fwd_txfm_ = GET_PARAM(0);
+    inv_txfm_ = GET_PARAM(1);
+    tx_type_  = GET_PARAM(2);
+  }
+  virtual void TearDown() { libvpx_test::ClearSystemState(); }
 
-TEST_P(Trans16x16Test, CoeffSizeCheck) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  const int count_test_block = 1000;
-  for (int i = 0; i < count_test_block; ++i) {
-    DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, 256);
-    DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, 256);
-    DECLARE_ALIGNED_ARRAY(16, int16_t, output_block, 256);
-    DECLARE_ALIGNED_ARRAY(16, int16_t, output_extreme_block, 256);
-    DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, 256);
-
-    // Initialize a test block with input range [-255, 255].
-    for (int j = 0; j < 256; ++j) {
-      input_block[j] = rnd.Rand8() - rnd.Rand8();
-      input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
-    }
-    if (i == 0)
-      for (int j = 0; j < 256; ++j)
-        input_extreme_block[j] = 255;
-
-    const int pitch = 32;
-    RunFwdTxfm(input_block, output_block, dst, pitch, tx_type_);
-    RunFwdTxfm(input_extreme_block, output_extreme_block, dst, pitch, tx_type_);
-
-    // The minimum quant value is 4.
-    for (int j = 0; j < 256; ++j) {
-      EXPECT_GE(4 * DCT_MAX_VALUE, abs(output_block[j]))
-          << "Error: 16x16 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
-      EXPECT_GE(4 * DCT_MAX_VALUE, abs(output_extreme_block[j]))
-          << "Error: 16x16 FDCT extreme has coefficient larger "
-          << "than 4*DCT_MAX_VALUE";
-    }
+ protected:
+  void RunFwdTxfm(int16_t *in, int16_t *out, uint8_t *dst, int stride) {
+    fwd_txfm_(in, out, stride);
+  }
+  void RunInvTxfm(int16_t *in, int16_t *out, uint8_t *dst, int stride) {
+    inv_txfm_(out, dst, stride >> 1);
   }
+
+  int tx_type_;
+  fdct_t fwd_txfm_;
+  idct_t inv_txfm_;
+};
+
+TEST_P(Trans16x16DCT, AccuracyCheck) {
+  RunAccuracyCheck();
 }
 
-TEST_P(Trans16x16Test, InvAccuracyCheck) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  const int count_test_block = 1000;
-  // TODO(jingning): is this unit test necessary? if so, need to add
-  // check sets for inverse hybrid transforms too.
-  if (tx_type_ != DCT_DCT)
-    return;
-
-  for (int i = 0; i < count_test_block; ++i) {
-    DECLARE_ALIGNED_ARRAY(16, int16_t, in, 256);
-    DECLARE_ALIGNED_ARRAY(16, int16_t, coeff, 256);
-    DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, 256);
-    DECLARE_ALIGNED_ARRAY(16, uint8_t, src, 256);
-    double out_r[256];
-
-    for (int j = 0; j < 256; ++j) {
-      src[j] = rnd.Rand8();
-      dst[j] = rnd.Rand8();
-    }
-    // Initialize a test block with input range [-255, 255].
-    for (int j = 0; j < 256; ++j)
-      in[j] = src[j] - dst[j];
-
-    reference_16x16_dct_2d(in, out_r);
-    for (int j = 0; j < 256; j++)
-      coeff[j] = round(out_r[j]);
-
-    const int pitch = 32;
-    RunInvTxfm(coeff, coeff, dst, pitch, tx_type_);
-
-    for (int j = 0; j < 256; ++j) {
-      const int diff = dst[j] - src[j];
-      const int error = diff * diff;
-      EXPECT_GE(1, error)
-          << "Error: 16x16 IDCT has error " << error
-          << " at index " << j;
-    }
+TEST_P(Trans16x16DCT, CoeffSizeCheck) {
+  RunCoeffSizeCheck();
+}
+
+TEST_P(Trans16x16DCT, InvAccuracyCheck) {
+  RunInvAccuracyCheck();
+}
+
+class Trans16x16HT : public Trans16x16TestBase,
+                     public PARAMS(fht_t, iht_t, int) {
+ public:
+  virtual ~Trans16x16HT() {}
+
+  virtual void SetUp() {
+    fwd_txfm_ = GET_PARAM(0);
+    inv_txfm_ = GET_PARAM(1);
+    tx_type_  = GET_PARAM(2);
   }
+  virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
+ protected:
+  void RunFwdTxfm(int16_t *in, int16_t *out, uint8_t *dst, int stride) {
+    fwd_txfm_(in, out, stride >> 1, tx_type_);
+  }
+  void RunInvTxfm(int16_t *in, int16_t *out, uint8_t *dst, int stride) {
+    inv_txfm_(out, dst, stride >> 1, tx_type_);
+  }
+
+  int tx_type_;
+  fht_t fwd_txfm_;
+  iht_t inv_txfm_;
+};
+
+TEST_P(Trans16x16HT, AccuracyCheck) {
+  RunAccuracyCheck();
+}
+
+TEST_P(Trans16x16HT, CoeffSizeCheck) {
+  RunCoeffSizeCheck();
 }
 
-INSTANTIATE_TEST_CASE_P(VP9, Trans16x16Test, ::testing::Range(0, 4));
+using std::tr1::make_tuple;
+
+INSTANTIATE_TEST_CASE_P(
+    C, Trans16x16DCT,
+    ::testing::Values(
+        make_tuple(&vp9_short_fdct16x16_c, &vp9_short_idct16x16_add_c, 0)));
+INSTANTIATE_TEST_CASE_P(
+    C, Trans16x16HT,
+    ::testing::Values(
+        make_tuple(&vp9_short_fht16x16_c, &vp9_short_iht16x16_add_c, 0),
+        make_tuple(&vp9_short_fht16x16_c, &vp9_short_iht16x16_add_c, 1),
+        make_tuple(&vp9_short_fht16x16_c, &vp9_short_iht16x16_add_c, 2),
+        make_tuple(&vp9_short_fht16x16_c, &vp9_short_iht16x16_add_c, 3)));
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(
+    SSE2, Trans16x16DCT,
+    ::testing::Values(
+        make_tuple(&vp9_short_fdct16x16_sse2, &vp9_short_idct16x16_add_c, 0)));
+INSTANTIATE_TEST_CASE_P(
+    SSE2, Trans16x16HT,
+    ::testing::Values(
+        make_tuple(&vp9_short_fht16x16_sse2, &vp9_short_iht16x16_add_sse2, 0),
+        make_tuple(&vp9_short_fht16x16_sse2, &vp9_short_iht16x16_add_sse2, 1),
+        make_tuple(&vp9_short_fht16x16_sse2, &vp9_short_iht16x16_add_sse2, 2),
+        make_tuple(&vp9_short_fht16x16_sse2, &vp9_short_iht16x16_add_sse2, 3)));
+#endif
 }  // namespace