shithub: libvpx

Download patch

ref: 5ac88162b9e4648ab35bc1df2d77bd44151ee23a
parent: 794a5ad713afd0533017382de70b7cfad388eb9e
author: Johann <johannkoenig@google.com>
date: Thu Jun 22 13:12:27 EDT 2017

partial fdct test

Test the _1 variant of the fdct, which simply sums the block and applies
a modifying shift based on the block size.

BUG=webm:1424

Change-Id: Ic80d6008abba0c596b575fa0484d5b5855321468

--- a/test/dct16x16_test.cc
+++ b/test/dct16x16_test.cc
@@ -744,66 +744,6 @@
   CompareInvReference(ref_txfm_, thresh_);
 }
 
-class PartialTrans16x16Test : public ::testing::TestWithParam<
-                                  std::tr1::tuple<FdctFunc, vpx_bit_depth_t> > {
- public:
-  virtual ~PartialTrans16x16Test() {}
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    bit_depth_ = GET_PARAM(1);
-  }
-
-  virtual void TearDown() { libvpx_test::ClearSystemState(); }
-
- protected:
-  vpx_bit_depth_t bit_depth_;
-  FdctFunc fwd_txfm_;
-};
-
-TEST_P(PartialTrans16x16Test, Extremes) {
-#if CONFIG_VP9_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  const int minval = -maxval;
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = maxval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
-  EXPECT_EQ((maxval * kNumCoeffs) >> 1, output[0]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = minval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
-  EXPECT_EQ((minval * kNumCoeffs) >> 1, output[0]);
-}
-
-TEST_P(PartialTrans16x16Test, Random) {
-#if CONFIG_VP9_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-
-  int sum = 0;
-  for (int i = 0; i < kNumCoeffs; ++i) {
-    const int val = (i & 1) ? -rnd(maxval + 1) : rnd(maxval + 1);
-    input[i] = val;
-    sum += val;
-  }
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
-  EXPECT_EQ(sum >> 1, output[0]);
-}
-
 using std::tr1::make_tuple;
 
 #if CONFIG_VP9_HIGHBITDEPTH
@@ -836,11 +776,6 @@
         make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1, VPX_BITS_8),
         make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2, VPX_BITS_8),
         make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3, VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(
-    C, PartialTrans16x16Test,
-    ::testing::Values(make_tuple(&vpx_highbd_fdct16x16_1_c, VPX_BITS_8),
-                      make_tuple(&vpx_highbd_fdct16x16_1_c, VPX_BITS_10),
-                      make_tuple(&vpx_highbd_fdct16x16_1_c, VPX_BITS_12)));
 #else
 INSTANTIATE_TEST_CASE_P(
     C, Trans16x16HT,
@@ -849,9 +784,6 @@
         make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1, VPX_BITS_8),
         make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2, VPX_BITS_8),
         make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3, VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(C, PartialTrans16x16Test,
-                        ::testing::Values(make_tuple(&vpx_fdct16x16_1_c,
-                                                     VPX_BITS_8)));
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
 #if HAVE_NEON && !CONFIG_EMULATE_HARDWARE
@@ -876,9 +808,6 @@
                                  2, VPX_BITS_8),
                       make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2,
                                  3, VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans16x16Test,
-                        ::testing::Values(make_tuple(&vpx_fdct16x16_1_sse2,
-                                                     VPX_BITS_8)));
 #endif  // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -913,9 +842,6 @@
                                  &idct16x16_10_add_12_sse2, 3167, VPX_BITS_12),
                       make_tuple(&idct16x16_12, &idct16x16_256_add_12_sse2,
                                  3167, VPX_BITS_12)));
-INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans16x16Test,
-                        ::testing::Values(make_tuple(&vpx_fdct16x16_1_sse2,
-                                                     VPX_BITS_8)));
 #endif  // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -931,8 +857,5 @@
         make_tuple(&vp9_fht16x16_msa, &vp9_iht16x16_256_add_msa, 2, VPX_BITS_8),
         make_tuple(&vp9_fht16x16_msa, &vp9_iht16x16_256_add_msa, 3,
                    VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(MSA, PartialTrans16x16Test,
-                        ::testing::Values(make_tuple(&vpx_fdct16x16_1_msa,
-                                                     VPX_BITS_8)));
 #endif  // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 }  // namespace
--- a/test/dct32x32_test.cc
+++ b/test/dct32x32_test.cc
@@ -292,67 +292,6 @@
   }
 }
 
-class PartialTrans32x32Test
-    : public ::testing::TestWithParam<
-          std::tr1::tuple<FwdTxfmFunc, vpx_bit_depth_t> > {
- public:
-  virtual ~PartialTrans32x32Test() {}
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    bit_depth_ = GET_PARAM(1);
-  }
-
-  virtual void TearDown() { libvpx_test::ClearSystemState(); }
-
- protected:
-  vpx_bit_depth_t bit_depth_;
-  FwdTxfmFunc fwd_txfm_;
-};
-
-TEST_P(PartialTrans32x32Test, Extremes) {
-#if CONFIG_VP9_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  const int minval = -maxval;
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = maxval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
-  EXPECT_EQ((maxval * kNumCoeffs) >> 3, output[0]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = minval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
-  EXPECT_EQ((minval * kNumCoeffs) >> 3, output[0]);
-}
-
-TEST_P(PartialTrans32x32Test, Random) {
-#if CONFIG_VP9_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-
-  int sum = 0;
-  for (int i = 0; i < kNumCoeffs; ++i) {
-    const int val = (i & 1) ? -rnd(maxval + 1) : rnd(maxval + 1);
-    input[i] = val;
-    sum += val;
-  }
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
-  EXPECT_EQ(sum >> 3, output[0]);
-}
-
 using std::tr1::make_tuple;
 
 #if CONFIG_VP9_HIGHBITDEPTH
@@ -366,11 +305,6 @@
         make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_c, 0, VPX_BITS_8),
         make_tuple(&vpx_fdct32x32_rd_c, &vpx_idct32x32_1024_add_c, 1,
                    VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(
-    C, PartialTrans32x32Test,
-    ::testing::Values(make_tuple(&vpx_highbd_fdct32x32_1_c, VPX_BITS_8),
-                      make_tuple(&vpx_highbd_fdct32x32_1_c, VPX_BITS_10),
-                      make_tuple(&vpx_highbd_fdct32x32_1_c, VPX_BITS_12)));
 #else
 INSTANTIATE_TEST_CASE_P(
     C, Trans32x32Test,
@@ -378,9 +312,6 @@
                                  VPX_BITS_8),
                       make_tuple(&vpx_fdct32x32_rd_c, &vpx_idct32x32_1024_add_c,
                                  1, VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(C, PartialTrans32x32Test,
-                        ::testing::Values(make_tuple(&vpx_fdct32x32_1_c,
-                                                     VPX_BITS_8)));
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
 #if HAVE_NEON && !CONFIG_EMULATE_HARDWARE
@@ -399,9 +330,6 @@
                                  &vpx_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
                       make_tuple(&vpx_fdct32x32_rd_sse2,
                                  &vpx_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans32x32Test,
-                        ::testing::Values(make_tuple(&vpx_fdct32x32_1_sse2,
-                                                     VPX_BITS_8)));
 #endif  // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -418,9 +346,6 @@
                    VPX_BITS_8),
         make_tuple(&vpx_fdct32x32_rd_sse2, &vpx_idct32x32_1024_add_c, 1,
                    VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans32x32Test,
-                        ::testing::Values(make_tuple(&vpx_fdct32x32_1_sse2,
-                                                     VPX_BITS_8)));
 #endif  // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 
 #if HAVE_AVX2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -439,8 +364,5 @@
                                  &vpx_idct32x32_1024_add_msa, 0, VPX_BITS_8),
                       make_tuple(&vpx_fdct32x32_rd_msa,
                                  &vpx_idct32x32_1024_add_msa, 1, VPX_BITS_8)));
-INSTANTIATE_TEST_CASE_P(MSA, PartialTrans32x32Test,
-                        ::testing::Values(make_tuple(&vpx_fdct32x32_1_msa,
-                                                     VPX_BITS_8)));
 #endif  // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
 }  // namespace
--- /dev/null
+++ b/test/dct_partial_test.cc
@@ -1,0 +1,154 @@
+/*
+ *  Copyright (c) 2017 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 <math.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <limits>
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+#include "./vpx_dsp_rtcd.h"
+#include "test/acm_random.h"
+#include "test/buffer.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+#include "vpx/vpx_codec.h"
+#include "vpx/vpx_integer.h"
+#include "vpx_dsp/vpx_dsp_common.h"
+
+using libvpx_test::ACMRandom;
+using libvpx_test::Buffer;
+using std::tr1::tuple;
+using std::tr1::make_tuple;
+
+namespace {
+typedef void (*PartialFdctFunc)(const int16_t *in, tran_low_t *out, int stride);
+
+typedef tuple<PartialFdctFunc, int /* size */, vpx_bit_depth_t>
+    PartialFdctParam;
+
+tran_low_t partial_fdct_ref(const Buffer<int16_t> &in, int size) {
+  int64_t sum = 0;
+  for (int y = 0; y < size; ++y) {
+    for (int x = 0; x < size; ++x) {
+      sum += in.TopLeftPixel()[y * in.stride() + x];
+    }
+  }
+
+  switch (size) {
+    case 4: sum <<= 1; break;
+    case 8: /*sum = sum;*/ break;
+    case 16: sum >>= 1; break;
+    case 32: sum >>= 3; break;
+  }
+
+  return static_cast<tran_low_t>(sum);
+}
+
+class PartialFdctTest : public ::testing::TestWithParam<PartialFdctParam> {
+ public:
+  PartialFdctTest() {
+    fwd_txfm_ = GET_PARAM(0);
+    size_ = GET_PARAM(1);
+    bit_depth_ = GET_PARAM(2);
+  }
+
+  virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
+ protected:
+  void RunTest() {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    const int16_t maxvalue =
+        clip_pixel_highbd(std::numeric_limits<int16_t>::max(), bit_depth_);
+    const int16_t minvalue = -maxvalue;
+    Buffer<int16_t> input_block =
+        Buffer<int16_t>(size_, size_, 8, size_ == 4 ? 0 : 16);
+    ASSERT_TRUE(input_block.Init());
+    Buffer<tran_low_t> output_block = Buffer<tran_low_t>(size_, size_, 0, 16);
+    ASSERT_TRUE(output_block.Init());
+
+    for (int i = 0; i < 100; ++i) {
+      if (i == 0) {
+        input_block.Set(maxvalue);
+      } else if (i == 1) {
+        input_block.Set(minvalue);
+      } else {
+        for (int y = 0; y < size_; ++y) {
+          for (int x = 0; x < size_; ++x) {
+            input_block.TopLeftPixel()[y * input_block.stride() + x] =
+                clamp((rnd.Rand16() - rnd.Rand16()), minvalue, maxvalue);
+          }
+        }
+      }
+
+      ASM_REGISTER_STATE_CHECK(fwd_txfm_(input_block.TopLeftPixel(),
+                                         output_block.TopLeftPixel(),
+                                         input_block.stride()));
+
+      EXPECT_EQ(partial_fdct_ref(input_block, size_),
+                output_block.TopLeftPixel()[0]);
+    }
+  }
+
+  PartialFdctFunc fwd_txfm_;
+  vpx_bit_depth_t bit_depth_;
+  int size_;
+};
+
+TEST_P(PartialFdctTest, PartialFdctTest) { RunTest(); }
+
+#if CONFIG_VP9_HIGHBITDEPTH
+INSTANTIATE_TEST_CASE_P(
+    C, PartialFdctTest,
+    ::testing::Values(make_tuple(&vpx_highbd_fdct32x32_1_c, 32, VPX_BITS_12),
+                      make_tuple(&vpx_highbd_fdct32x32_1_c, 32, VPX_BITS_10),
+                      make_tuple(&vpx_fdct32x32_1_c, 32, VPX_BITS_8),
+                      make_tuple(&vpx_highbd_fdct16x16_1_c, 16, VPX_BITS_12),
+                      make_tuple(&vpx_highbd_fdct16x16_1_c, 16, VPX_BITS_10),
+                      make_tuple(&vpx_fdct16x16_1_c, 16, VPX_BITS_8),
+                      make_tuple(&vpx_highbd_fdct8x8_1_c, 8, VPX_BITS_12),
+                      make_tuple(&vpx_highbd_fdct8x8_1_c, 8, VPX_BITS_10),
+                      make_tuple(&vpx_fdct8x8_1_c, 8, VPX_BITS_8),
+                      make_tuple(&vpx_fdct4x4_1_c, 4, VPX_BITS_8)));
+#else
+INSTANTIATE_TEST_CASE_P(
+    C, PartialFdctTest,
+    ::testing::Values(make_tuple(&vpx_fdct32x32_1_c, 32, VPX_BITS_8),
+                      make_tuple(&vpx_fdct16x16_1_c, 16, VPX_BITS_8),
+                      make_tuple(&vpx_fdct8x8_1_c, 8, VPX_BITS_8),
+                      make_tuple(&vpx_fdct4x4_1_c, 4, VPX_BITS_8)));
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(
+    SSE2, PartialFdctTest,
+    ::testing::Values(make_tuple(&vpx_fdct32x32_1_sse2, 32, VPX_BITS_8),
+                      make_tuple(&vpx_fdct16x16_1_sse2, 16, VPX_BITS_8),
+                      make_tuple(&vpx_fdct8x8_1_sse2, 8, VPX_BITS_8),
+                      make_tuple(&vpx_fdct4x4_1_sse2, 4, VPX_BITS_8)));
+#endif  // HAVE_SSE2
+
+#if HAVE_NEON
+INSTANTIATE_TEST_CASE_P(NEON, PartialFdctTest,
+                        ::testing::Values(make_tuple(&vpx_fdct8x8_1_neon, 8,
+                                                     VPX_BITS_8)));
+#endif  // HAVE_NEON
+
+#if HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH
+INSTANTIATE_TEST_CASE_P(
+    MSA, PartialFdctTest,
+    ::testing::Values(make_tuple(&vpx_fdct32x32_1_msa, 32, VPX_BITS_8),
+                      make_tuple(&vpx_fdct16x16_1_msa, 16, VPX_BITS_8),
+                      make_tuple(&vpx_fdct8x8_1_msa, 8, VPX_BITS_8)));
+#endif  // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH
+}  // namespace
--- a/test/test.mk
+++ b/test/test.mk
@@ -154,6 +154,7 @@
 LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += comp_avg_pred_test.cc
 LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct16x16_test.cc
 LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct32x32_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct_partial_test.cc
 LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct_test.cc
 LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct8x8_test.cc
 LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += hadamard_test.cc
--- a/vpx_dsp/vpx_dsp_common.h
+++ b/vpx_dsp/vpx_dsp_common.h
@@ -55,7 +55,6 @@
   return value < low ? low : (value > high ? high : value);
 }
 
-#if CONFIG_VP9_HIGHBITDEPTH
 static INLINE uint16_t clip_pixel_highbd(int val, int bd) {
   switch (bd) {
     case 8:
@@ -64,7 +63,6 @@
     case 12: return (uint16_t)clamp(val, 0, 4095);
   }
 }
-#endif  // CONFIG_VP9_HIGHBITDEPTH
 
 #ifdef __cplusplus
 }  // extern "C"
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -493,7 +493,7 @@
   specialize qw/vpx_fdct8x8 neon sse2/;
 
   add_proto qw/void vpx_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
-  specialize qw/vpx_fdct8x8_1 neon sse2/;
+  specialize qw/vpx_fdct8x8_1 neon sse2 msa/;
 
   add_proto qw/void vpx_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
   specialize qw/vpx_fdct16x16 neon sse2/;