From 0b4498895265474efaa6866397f3c34038674110 Mon Sep 17 00:00:00 2001 From: Frank Galligan Date: Sun, 1 Jun 2014 16:20:26 -0700 Subject: [PATCH] Revert "Remove Wextra warnings from vp9_sad.c" This reverts commit 916550428db803c54c993ff9d3c34b9b0bcebb7c Change-Id: I500822b03f09c64ff6ec5396c68edee9ca3b75cb --- test/sad_test.cc | 350 +++++++++++++--------------------- vp9/common/vp9_rtcd_defs.pl | 52 ++--- vp9/encoder/vp9_encodeframe.c | 6 +- vp9/encoder/vp9_mbgraph.c | 11 +- vp9/encoder/vp9_mcomp.c | 64 ++++--- vp9/encoder/vp9_rdopt.c | 3 +- vp9/encoder/vp9_sad.c | 12 +- vp9/encoder/vp9_variance.h | 6 +- 8 files changed, 221 insertions(+), 283 deletions(-) diff --git a/test/sad_test.cc b/test/sad_test.cc index adb191fd0..a692891ad 100644 --- a/test/sad_test.cc +++ b/test/sad_test.cc @@ -29,22 +29,12 @@ #include "third_party/googletest/src/include/gtest/gtest.h" -#if CONFIG_VP8_ENCODER typedef unsigned int (*sad_m_by_n_fn_t)(const unsigned char *source_ptr, int source_stride, const unsigned char *reference_ptr, int reference_stride, unsigned int max_sad); typedef std::tr1::tuple sad_m_by_n_test_param_t; -#endif -#if CONFIG_VP9_ENCODER -typedef unsigned int (*sad_m_by_n_fn_vp9_t)(const unsigned char *source_ptr, - int source_stride, - const unsigned char *reference_ptr, - int reference_stride); -typedef std::tr1::tuple - sad_m_by_n_test_param_vp9_t; -#endif typedef void (*sad_n_by_n_by_4_fn_t)(const uint8_t *src_ptr, int src_stride, @@ -97,7 +87,7 @@ class SADTestBase : public ::testing::Test { // Sum of Absolute Differences. Given two blocks, calculate the absolute // difference between two pixels in the same relative location; accumulate. - unsigned int ReferenceSAD(unsigned int max_sad, int block_idx) { + unsigned int ReferenceSAD(unsigned int max_sad, int block_idx = 0) { unsigned int sad = 0; const uint8_t* const reference = GetReference(block_idx); @@ -138,43 +128,13 @@ class SADTestBase : public ::testing::Test { ACMRandom rnd_; }; -class SADx4Test - : public SADTestBase, - public ::testing::WithParamInterface { - public: - SADx4Test() : SADTestBase(GET_PARAM(0), GET_PARAM(1)) {} - - protected: - void SADs(unsigned int *results) { - const uint8_t* refs[] = {GetReference(0), GetReference(1), - GetReference(2), GetReference(3)}; - - REGISTER_STATE_CHECK(GET_PARAM(2)(source_data_, source_stride_, - refs, reference_stride_, - results)); - } - - void CheckSADs() { - unsigned int reference_sad, exp_sad[4]; - - SADs(exp_sad); - for (int block = 0; block < 4; ++block) { - reference_sad = ReferenceSAD(UINT_MAX, block); - - EXPECT_EQ(reference_sad, exp_sad[block]) << "block " << block; - } - } -}; - -#if CONFIG_VP8_ENCODER -class SADTest - : public SADTestBase, - public ::testing::WithParamInterface { +class SADTest : public SADTestBase, + public ::testing::WithParamInterface { public: SADTest() : SADTestBase(GET_PARAM(0), GET_PARAM(1)) {} protected: - unsigned int SAD(unsigned int max_sad, int block_idx) { + unsigned int SAD(unsigned int max_sad, int block_idx = 0) { unsigned int ret; const uint8_t* const reference = GetReference(block_idx); @@ -184,11 +144,11 @@ class SADTest return ret; } - void CheckSAD(unsigned int max_sad) { + void CheckSad(unsigned int max_sad) { unsigned int reference_sad, exp_sad; - reference_sad = ReferenceSAD(max_sad, 0); - exp_sad = SAD(max_sad, 0); + reference_sad = ReferenceSAD(max_sad); + exp_sad = SAD(max_sad); if (reference_sad <= max_sad) { ASSERT_EQ(exp_sad, reference_sad); @@ -198,133 +158,43 @@ class SADTest } } }; -#endif // CONFIG_VP8_ENCODER -#if CONFIG_VP9_ENCODER -class SADVP9Test - : public SADTestBase, - public ::testing::WithParamInterface { +class SADx4Test : public SADTestBase, + public ::testing::WithParamInterface { public: - SADVP9Test() : SADTestBase(GET_PARAM(0), GET_PARAM(1)) {} + SADx4Test() : SADTestBase(GET_PARAM(0), GET_PARAM(1)) {} protected: - unsigned int SAD(int block_idx) { - unsigned int ret; - const uint8_t* const reference = GetReference(block_idx); + void SADs(unsigned int *results) { + const uint8_t* refs[] = {GetReference(0), GetReference(1), + GetReference(2), GetReference(3)}; - REGISTER_STATE_CHECK(ret = GET_PARAM(2)(source_data_, source_stride_, - reference, reference_stride_)); - return ret; + REGISTER_STATE_CHECK(GET_PARAM(2)(source_data_, source_stride_, + refs, reference_stride_, + results)); } - void CheckSAD() { - unsigned int reference_sad, exp_sad; + void CheckSADs() { + unsigned int reference_sad, exp_sad[4]; - reference_sad = ReferenceSAD(UINT_MAX, 0); - exp_sad = SAD(0); + SADs(exp_sad); + for (int block = 0; block < 4; block++) { + reference_sad = ReferenceSAD(UINT_MAX, block); - ASSERT_EQ(reference_sad, exp_sad); + EXPECT_EQ(exp_sad[block], reference_sad) << "block " << block; + } } }; -#endif // CONFIG_VP9_ENCODER uint8_t* SADTestBase::source_data_ = NULL; uint8_t* SADTestBase::reference_data_ = NULL; -#if CONFIG_VP8_ENCODER TEST_P(SADTest, MaxRef) { FillConstant(source_data_, source_stride_, 0); FillConstant(reference_data_, reference_stride_, 255); - CheckSAD(UINT_MAX); -} - -TEST_P(SADTest, MaxSrc) { - FillConstant(source_data_, source_stride_, 255); - FillConstant(reference_data_, reference_stride_, 0); - CheckSAD(UINT_MAX); -} - -TEST_P(SADTest, ShortRef) { - int tmp_stride = reference_stride_; - reference_stride_ >>= 1; - FillRandom(source_data_, source_stride_); - FillRandom(reference_data_, reference_stride_); - CheckSAD(UINT_MAX); - reference_stride_ = tmp_stride; -} - -TEST_P(SADTest, UnalignedRef) { - // The reference frame, but not the source frame, may be unaligned for - // certain types of searches. - const int tmp_stride = reference_stride_; - reference_stride_ -= 1; - FillRandom(source_data_, source_stride_); - FillRandom(reference_data_, reference_stride_); - CheckSAD(UINT_MAX); - reference_stride_ = tmp_stride; -} - -TEST_P(SADTest, ShortSrc) { - const int tmp_stride = source_stride_; - source_stride_ >>= 1; - FillRandom(source_data_, source_stride_); - FillRandom(reference_data_, reference_stride_); - CheckSAD(UINT_MAX); - source_stride_ = tmp_stride; -} - -TEST_P(SADTest, MaxSAD) { - // Verify that, when max_sad is set, the implementation does not return a - // value lower than the reference. - FillConstant(source_data_, source_stride_, 255); - FillConstant(reference_data_, reference_stride_, 0); - CheckSAD(128); -} -#endif // CONFIG_VP8_ENCODER - -#if CONFIG_VP9_ENCODER -TEST_P(SADVP9Test, MaxRef) { - FillConstant(source_data_, source_stride_, 0); - FillConstant(reference_data_, reference_stride_, 255); - CheckSAD(); -} - -TEST_P(SADVP9Test, MaxSrc) { - FillConstant(source_data_, source_stride_, 255); - FillConstant(reference_data_, reference_stride_, 0); - CheckSAD(); -} - -TEST_P(SADVP9Test, ShortRef) { - const int tmp_stride = reference_stride_; - reference_stride_ >>= 1; - FillRandom(source_data_, source_stride_); - FillRandom(reference_data_, reference_stride_); - CheckSAD(); - reference_stride_ = tmp_stride; -} - -TEST_P(SADVP9Test, UnalignedRef) { - // The reference frame, but not the source frame, may be unaligned for - // certain types of searches. - const int tmp_stride = reference_stride_; - reference_stride_ -= 1; - FillRandom(source_data_, source_stride_); - FillRandom(reference_data_, reference_stride_); - CheckSAD(); - reference_stride_ = tmp_stride; + CheckSad(UINT_MAX); } -TEST_P(SADVP9Test, ShortSrc) { - const int tmp_stride = source_stride_; - source_stride_ >>= 1; - FillRandom(source_data_, source_stride_); - FillRandom(reference_data_, reference_stride_); - CheckSAD(); - source_stride_ = tmp_stride; -} -#endif // CONFIG_VP9_ENCODER - TEST_P(SADx4Test, MaxRef) { FillConstant(source_data_, source_stride_, 0); FillConstant(GetReference(0), reference_stride_, 255); @@ -334,6 +204,12 @@ TEST_P(SADx4Test, MaxRef) { CheckSADs(); } +TEST_P(SADTest, MaxSrc) { + FillConstant(source_data_, source_stride_, 255); + FillConstant(reference_data_, reference_stride_, 0); + CheckSad(UINT_MAX); +} + TEST_P(SADx4Test, MaxSrc) { FillConstant(source_data_, source_stride_, 255); FillConstant(GetReference(0), reference_stride_, 0); @@ -343,6 +219,15 @@ TEST_P(SADx4Test, MaxSrc) { CheckSADs(); } +TEST_P(SADTest, ShortRef) { + int tmp_stride = reference_stride_; + reference_stride_ >>= 1; + FillRandom(source_data_, source_stride_); + FillRandom(reference_data_, reference_stride_); + CheckSad(UINT_MAX); + reference_stride_ = tmp_stride; +} + TEST_P(SADx4Test, ShortRef) { int tmp_stride = reference_stride_; reference_stride_ >>= 1; @@ -355,6 +240,17 @@ TEST_P(SADx4Test, ShortRef) { reference_stride_ = tmp_stride; } +TEST_P(SADTest, UnalignedRef) { + // The reference frame, but not the source frame, may be unaligned for + // certain types of searches. + int tmp_stride = reference_stride_; + reference_stride_ -= 1; + FillRandom(source_data_, source_stride_); + FillRandom(reference_data_, reference_stride_); + CheckSad(UINT_MAX); + reference_stride_ = tmp_stride; +} + TEST_P(SADx4Test, UnalignedRef) { // The reference frame, but not the source frame, may be unaligned for // certain types of searches. @@ -369,6 +265,15 @@ TEST_P(SADx4Test, UnalignedRef) { reference_stride_ = tmp_stride; } +TEST_P(SADTest, ShortSrc) { + int tmp_stride = source_stride_; + source_stride_ >>= 1; + FillRandom(source_data_, source_stride_); + FillRandom(reference_data_, reference_stride_); + CheckSad(UINT_MAX); + source_stride_ = tmp_stride; +} + TEST_P(SADx4Test, ShortSrc) { int tmp_stride = source_stride_; source_stride_ >>= 1; @@ -381,6 +286,14 @@ TEST_P(SADx4Test, ShortSrc) { source_stride_ = tmp_stride; } +TEST_P(SADTest, MaxSAD) { + // Verify that, when max_sad is set, the implementation does not return a + // value lower than the reference. + FillConstant(source_data_, source_stride_, 255); + FillConstant(reference_data_, reference_stride_, 0); + CheckSad(128); +} + using std::tr1::make_tuple; //------------------------------------------------------------------------------ @@ -391,27 +304,27 @@ const sad_m_by_n_fn_t sad_8x16_c = vp8_sad8x16_c; const sad_m_by_n_fn_t sad_16x8_c = vp8_sad16x8_c; const sad_m_by_n_fn_t sad_8x8_c = vp8_sad8x8_c; const sad_m_by_n_fn_t sad_4x4_c = vp8_sad4x4_c; +#endif +#if CONFIG_VP9_ENCODER +const sad_m_by_n_fn_t sad_64x64_c_vp9 = vp9_sad64x64_c; +const sad_m_by_n_fn_t sad_32x32_c_vp9 = vp9_sad32x32_c; +const sad_m_by_n_fn_t sad_16x16_c_vp9 = vp9_sad16x16_c; +const sad_m_by_n_fn_t sad_8x16_c_vp9 = vp9_sad8x16_c; +const sad_m_by_n_fn_t sad_16x8_c_vp9 = vp9_sad16x8_c; +const sad_m_by_n_fn_t sad_8x8_c_vp9 = vp9_sad8x8_c; +const sad_m_by_n_fn_t sad_8x4_c_vp9 = vp9_sad8x4_c; +const sad_m_by_n_fn_t sad_4x8_c_vp9 = vp9_sad4x8_c; +const sad_m_by_n_fn_t sad_4x4_c_vp9 = vp9_sad4x4_c; +#endif const sad_m_by_n_test_param_t c_tests[] = { +#if CONFIG_VP8_ENCODER make_tuple(16, 16, sad_16x16_c), make_tuple(8, 16, sad_8x16_c), make_tuple(16, 8, sad_16x8_c), make_tuple(8, 8, sad_8x8_c), make_tuple(4, 4, sad_4x4_c), -}; -INSTANTIATE_TEST_CASE_P(C, SADTest, ::testing::ValuesIn(c_tests)); -#endif // CONFIG_VP8_ENCODER - +#endif #if CONFIG_VP9_ENCODER -const sad_m_by_n_fn_vp9_t sad_64x64_c_vp9 = vp9_sad64x64_c; -const sad_m_by_n_fn_vp9_t sad_32x32_c_vp9 = vp9_sad32x32_c; -const sad_m_by_n_fn_vp9_t sad_16x16_c_vp9 = vp9_sad16x16_c; -const sad_m_by_n_fn_vp9_t sad_8x16_c_vp9 = vp9_sad8x16_c; -const sad_m_by_n_fn_vp9_t sad_16x8_c_vp9 = vp9_sad16x8_c; -const sad_m_by_n_fn_vp9_t sad_8x8_c_vp9 = vp9_sad8x8_c; -const sad_m_by_n_fn_vp9_t sad_8x4_c_vp9 = vp9_sad8x4_c; -const sad_m_by_n_fn_vp9_t sad_4x8_c_vp9 = vp9_sad4x8_c; -const sad_m_by_n_fn_vp9_t sad_4x4_c_vp9 = vp9_sad4x4_c; -const sad_m_by_n_test_param_vp9_t c_vp9_tests[] = { make_tuple(64, 64, sad_64x64_c_vp9), make_tuple(32, 32, sad_32x32_c_vp9), make_tuple(16, 16, sad_16x16_c_vp9), @@ -421,9 +334,11 @@ const sad_m_by_n_test_param_vp9_t c_vp9_tests[] = { make_tuple(8, 4, sad_8x4_c_vp9), make_tuple(4, 8, sad_4x8_c_vp9), make_tuple(4, 4, sad_4x4_c_vp9), +#endif }; -INSTANTIATE_TEST_CASE_P(C, SADVP9Test, ::testing::ValuesIn(c_vp9_tests)); +INSTANTIATE_TEST_CASE_P(C, SADTest, ::testing::ValuesIn(c_tests)); +#if CONFIG_VP9_ENCODER const sad_n_by_n_by_4_fn_t sad_64x64x4d_c = vp9_sad64x64x4d_c; const sad_n_by_n_by_4_fn_t sad_64x32x4d_c = vp9_sad64x32x4d_c; const sad_n_by_n_by_4_fn_t sad_32x64x4d_c = vp9_sad32x64x4d_c; @@ -460,8 +375,8 @@ INSTANTIATE_TEST_CASE_P(C, SADx4Test, ::testing::Values( const sad_m_by_n_fn_t sad_16x16_armv6 = vp8_sad16x16_armv6; INSTANTIATE_TEST_CASE_P(MEDIA, SADTest, ::testing::Values( make_tuple(16, 16, sad_16x16_armv6))); -#endif // CONFIG_VP8_ENCODER -#endif // HAVE_MEDIA +#endif +#endif #if HAVE_NEON #if CONFIG_VP8_ENCODER @@ -476,8 +391,8 @@ INSTANTIATE_TEST_CASE_P(NEON, SADTest, ::testing::Values( make_tuple(16, 8, sad_16x8_neon), make_tuple(8, 8, sad_8x8_neon), make_tuple(4, 4, sad_4x4_neon))); -#endif // CONFIG_VP8_ENCODER -#endif // HAVE_NEON +#endif +#endif //------------------------------------------------------------------------------ // x86 functions @@ -488,39 +403,40 @@ const sad_m_by_n_fn_t sad_8x16_mmx = vp8_sad8x16_mmx; const sad_m_by_n_fn_t sad_16x8_mmx = vp8_sad16x8_mmx; const sad_m_by_n_fn_t sad_8x8_mmx = vp8_sad8x8_mmx; const sad_m_by_n_fn_t sad_4x4_mmx = vp8_sad4x4_mmx; +#endif +#if CONFIG_VP9_ENCODER +const sad_m_by_n_fn_t sad_16x16_mmx_vp9 = vp9_sad16x16_mmx; +const sad_m_by_n_fn_t sad_8x16_mmx_vp9 = vp9_sad8x16_mmx; +const sad_m_by_n_fn_t sad_16x8_mmx_vp9 = vp9_sad16x8_mmx; +const sad_m_by_n_fn_t sad_8x8_mmx_vp9 = vp9_sad8x8_mmx; +const sad_m_by_n_fn_t sad_4x4_mmx_vp9 = vp9_sad4x4_mmx; +#endif + const sad_m_by_n_test_param_t mmx_tests[] = { +#if CONFIG_VP8_ENCODER make_tuple(16, 16, sad_16x16_mmx), make_tuple(8, 16, sad_8x16_mmx), make_tuple(16, 8, sad_16x8_mmx), make_tuple(8, 8, sad_8x8_mmx), make_tuple(4, 4, sad_4x4_mmx), -}; -INSTANTIATE_TEST_CASE_P(MMX, SADTest, ::testing::ValuesIn(mmx_tests)); -#endif // CONFIG_VP8_ENCODER - +#endif #if CONFIG_VP9_ENCODER -const sad_m_by_n_fn_vp9_t sad_16x16_mmx_vp9 = vp9_sad16x16_mmx; -const sad_m_by_n_fn_vp9_t sad_8x16_mmx_vp9 = vp9_sad8x16_mmx; -const sad_m_by_n_fn_vp9_t sad_16x8_mmx_vp9 = vp9_sad16x8_mmx; -const sad_m_by_n_fn_vp9_t sad_8x8_mmx_vp9 = vp9_sad8x8_mmx; -const sad_m_by_n_fn_vp9_t sad_4x4_mmx_vp9 = vp9_sad4x4_mmx; -const sad_m_by_n_test_param_vp9_t mmx_vp9_tests[] = { make_tuple(16, 16, sad_16x16_mmx_vp9), make_tuple(8, 16, sad_8x16_mmx_vp9), make_tuple(16, 8, sad_16x8_mmx_vp9), make_tuple(8, 8, sad_8x8_mmx_vp9), make_tuple(4, 4, sad_4x4_mmx_vp9), +#endif }; -INSTANTIATE_TEST_CASE_P(MMX, SADVP9Test, ::testing::ValuesIn(mmx_vp9_tests)); -#endif // CONFIG_VP9_ENCODER -#endif // HAVE_MMX +INSTANTIATE_TEST_CASE_P(MMX, SADTest, ::testing::ValuesIn(mmx_tests)); +#endif #if HAVE_SSE #if CONFIG_VP9_ENCODER #if CONFIG_USE_X86INC -const sad_m_by_n_fn_vp9_t sad_4x4_sse_vp9 = vp9_sad4x4_sse; -const sad_m_by_n_fn_vp9_t sad_4x8_sse_vp9 = vp9_sad4x8_sse; -INSTANTIATE_TEST_CASE_P(SSE, SADVP9Test, ::testing::Values( +const sad_m_by_n_fn_t sad_4x4_sse_vp9 = vp9_sad4x4_sse; +const sad_m_by_n_fn_t sad_4x8_sse_vp9 = vp9_sad4x8_sse; +INSTANTIATE_TEST_CASE_P(SSE, SADTest, ::testing::Values( make_tuple(4, 4, sad_4x4_sse_vp9), make_tuple(4, 8, sad_4x8_sse_vp9))); @@ -540,30 +456,32 @@ const sad_m_by_n_fn_t sad_8x16_wmt = vp8_sad8x16_wmt; const sad_m_by_n_fn_t sad_16x8_wmt = vp8_sad16x8_wmt; const sad_m_by_n_fn_t sad_8x8_wmt = vp8_sad8x8_wmt; const sad_m_by_n_fn_t sad_4x4_wmt = vp8_sad4x4_wmt; +#endif +#if CONFIG_VP9_ENCODER +#if CONFIG_USE_X86INC +const sad_m_by_n_fn_t sad_64x64_sse2_vp9 = vp9_sad64x64_sse2; +const sad_m_by_n_fn_t sad_64x32_sse2_vp9 = vp9_sad64x32_sse2; +const sad_m_by_n_fn_t sad_32x64_sse2_vp9 = vp9_sad32x64_sse2; +const sad_m_by_n_fn_t sad_32x32_sse2_vp9 = vp9_sad32x32_sse2; +const sad_m_by_n_fn_t sad_32x16_sse2_vp9 = vp9_sad32x16_sse2; +const sad_m_by_n_fn_t sad_16x32_sse2_vp9 = vp9_sad16x32_sse2; +const sad_m_by_n_fn_t sad_16x16_sse2_vp9 = vp9_sad16x16_sse2; +const sad_m_by_n_fn_t sad_16x8_sse2_vp9 = vp9_sad16x8_sse2; +const sad_m_by_n_fn_t sad_8x16_sse2_vp9 = vp9_sad8x16_sse2; +const sad_m_by_n_fn_t sad_8x8_sse2_vp9 = vp9_sad8x8_sse2; +const sad_m_by_n_fn_t sad_8x4_sse2_vp9 = vp9_sad8x4_sse2; +#endif +#endif const sad_m_by_n_test_param_t sse2_tests[] = { +#if CONFIG_VP8_ENCODER make_tuple(16, 16, sad_16x16_wmt), make_tuple(8, 16, sad_8x16_wmt), make_tuple(16, 8, sad_16x8_wmt), make_tuple(8, 8, sad_8x8_wmt), make_tuple(4, 4, sad_4x4_wmt), -}; -INSTANTIATE_TEST_CASE_P(SSE2, SADTest, ::testing::ValuesIn(sse2_tests)); -#endif // CONFIG_VP8_ENCODER - +#endif #if CONFIG_VP9_ENCODER #if CONFIG_USE_X86INC -const sad_m_by_n_fn_vp9_t sad_64x64_sse2_vp9 = vp9_sad64x64_sse2; -const sad_m_by_n_fn_vp9_t sad_64x32_sse2_vp9 = vp9_sad64x32_sse2; -const sad_m_by_n_fn_vp9_t sad_32x64_sse2_vp9 = vp9_sad32x64_sse2; -const sad_m_by_n_fn_vp9_t sad_32x32_sse2_vp9 = vp9_sad32x32_sse2; -const sad_m_by_n_fn_vp9_t sad_32x16_sse2_vp9 = vp9_sad32x16_sse2; -const sad_m_by_n_fn_vp9_t sad_16x32_sse2_vp9 = vp9_sad16x32_sse2; -const sad_m_by_n_fn_vp9_t sad_16x16_sse2_vp9 = vp9_sad16x16_sse2; -const sad_m_by_n_fn_vp9_t sad_16x8_sse2_vp9 = vp9_sad16x8_sse2; -const sad_m_by_n_fn_vp9_t sad_8x16_sse2_vp9 = vp9_sad8x16_sse2; -const sad_m_by_n_fn_vp9_t sad_8x8_sse2_vp9 = vp9_sad8x8_sse2; -const sad_m_by_n_fn_vp9_t sad_8x4_sse2_vp9 = vp9_sad8x4_sse2; -const sad_m_by_n_test_param_vp9_t sse2_vp9_tests[] = { make_tuple(64, 64, sad_64x64_sse2_vp9), make_tuple(64, 32, sad_64x32_sse2_vp9), make_tuple(32, 64, sad_32x64_sse2_vp9), @@ -575,9 +493,13 @@ const sad_m_by_n_test_param_vp9_t sse2_vp9_tests[] = { make_tuple(8, 16, sad_8x16_sse2_vp9), make_tuple(8, 8, sad_8x8_sse2_vp9), make_tuple(8, 4, sad_8x4_sse2_vp9), +#endif +#endif }; -INSTANTIATE_TEST_CASE_P(SSE2, SADVP9Test, ::testing::ValuesIn(sse2_vp9_tests)); +INSTANTIATE_TEST_CASE_P(SSE2, SADTest, ::testing::ValuesIn(sse2_tests)); +#if CONFIG_VP9_ENCODER +#if CONFIG_USE_X86INC const sad_n_by_n_by_4_fn_t sad_64x64x4d_sse2 = vp9_sad64x64x4d_sse2; const sad_n_by_n_by_4_fn_t sad_64x32x4d_sse2 = vp9_sad64x32x4d_sse2; const sad_n_by_n_by_4_fn_t sad_32x64x4d_sse2 = vp9_sad32x64x4d_sse2; @@ -601,9 +523,9 @@ INSTANTIATE_TEST_CASE_P(SSE2, SADx4Test, ::testing::Values( make_tuple(8, 16, sad_8x16x4d_sse2), make_tuple(8, 8, sad_8x8x4d_sse2), make_tuple(8, 4, sad_8x4x4d_sse2))); -#endif // CONFIG_USE_X86INC -#endif // CONFIG_VP9_ENCODER -#endif // HAVE_SSE2 +#endif +#endif +#endif #if HAVE_SSE3 #if CONFIG_VP8_ENCODER @@ -618,8 +540,8 @@ INSTANTIATE_TEST_CASE_P(SSE3, SADx4Test, ::testing::Values( make_tuple(8, 16, sad_8x16x4d_sse3), make_tuple(8, 8, sad_8x8x4d_sse3), make_tuple(4, 4, sad_4x4x4d_sse3))); -#endif // CONFIG_VP8_ENCODER -#endif // HAVE_SSE3 +#endif +#endif #if HAVE_SSSE3 #if CONFIG_USE_X86INC @@ -627,8 +549,8 @@ INSTANTIATE_TEST_CASE_P(SSE3, SADx4Test, ::testing::Values( const sad_m_by_n_fn_t sad_16x16_sse3 = vp8_sad16x16_sse3; INSTANTIATE_TEST_CASE_P(SSE3, SADTest, ::testing::Values( make_tuple(16, 16, sad_16x16_sse3))); -#endif // CONFIG_VP8_ENCODER -#endif // CONFIG_USE_X86INC -#endif // HAVE_SSSE3 +#endif +#endif +#endif } // namespace diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl index 2b2666aa8..182739620 100644 --- a/vp9/common/vp9_rtcd_defs.pl +++ b/vp9/common/vp9_rtcd_defs.pl @@ -520,82 +520,82 @@ specialize qw/vp9_sub_pixel_variance4x4/, "$sse_x86inc", "$ssse3_x86inc"; add_proto qw/unsigned int vp9_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"; specialize qw/vp9_sub_pixel_avg_variance4x4/, "$sse_x86inc", "$ssse3_x86inc"; -add_proto qw/unsigned int vp9_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad64x64/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad32x64/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad64x32/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad32x16/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad16x32/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad32x32/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad16x16 mmx/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad16x8 mmx/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad8x16 mmx/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad8x8 mmx/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad8x4/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad4x8/, "$sse_x86inc"; -add_proto qw/unsigned int vp9_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride"; +add_proto qw/unsigned int vp9_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"; specialize qw/vp9_sad4x4 mmx/, "$sse_x86inc"; -add_proto qw/unsigned int vp9_sad64x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad64x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad64x64_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad32x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad32x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad32x64_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad64x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad64x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad64x32_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad32x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad32x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad32x16_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad16x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad16x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad16x32_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad32x32_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad16x16_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad16x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad16x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad16x8_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad8x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad8x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad8x16_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad8x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad8x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad8x8_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad8x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad8x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad8x4_avg/, "$sse2_x86inc"; -add_proto qw/unsigned int vp9_sad4x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad4x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad4x8_avg/, "$sse_x86inc"; -add_proto qw/unsigned int vp9_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred"; +add_proto qw/unsigned int vp9_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"; specialize qw/vp9_sad4x4_avg/, "$sse_x86inc"; add_proto qw/void vp9_sad64x64x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"; diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c index ae83d7fbc..ef33fcaf1 100644 --- a/vp9/encoder/vp9_encodeframe.c +++ b/vp9/encoder/vp9_encodeframe.c @@ -1303,14 +1303,14 @@ static int is_background(VP9_COMP *cpi, const TileInfo *const tile, if (row8x8_remaining >= MI_BLOCK_SIZE && col8x8_remaining >= MI_BLOCK_SIZE) { this_sad = cpi->fn_ptr[BLOCK_64X64].sdf(src, src_stride, - pre, pre_stride); + pre, pre_stride, 0x7fffffff); threshold = (1 << 12); } else { int r, c; for (r = 0; r < row8x8_remaining; r += 2) for (c = 0; c < col8x8_remaining; c += 2) - this_sad += cpi->fn_ptr[BLOCK_16X16].sdf(src, src_stride, - pre, pre_stride); + this_sad += cpi->fn_ptr[BLOCK_16X16].sdf(src, src_stride, pre, + pre_stride, 0x7fffffff); threshold = (row8x8_remaining * col8x8_remaining) << 6; } diff --git a/vp9/encoder/vp9_mbgraph.c b/vp9/encoder/vp9_mbgraph.c index 842bc5b9d..041e583fd 100644 --- a/vp9/encoder/vp9_mbgraph.c +++ b/vp9/encoder/vp9_mbgraph.c @@ -72,7 +72,8 @@ static unsigned int do_16x16_motion_iteration(VP9_COMP *cpi, x->mv_row_max = tmp_row_max; return vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride, - xd->plane[0].dst.buf, xd->plane[0].dst.stride); + xd->plane[0].dst.buf, xd->plane[0].dst.stride, + INT_MAX); } static int do_16x16_motion_search(VP9_COMP *cpi, const MV *ref_mv, @@ -85,7 +86,8 @@ static int do_16x16_motion_search(VP9_COMP *cpi, const MV *ref_mv, // Try zero MV first // FIXME should really use something like near/nearest MV and/or MV prediction err = vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride, - xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride); + xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride, + INT_MAX); dst_mv->as_int = 0; // Test last reference frame using the previous best mv as the @@ -121,7 +123,8 @@ static int do_16x16_zerozero_search(VP9_COMP *cpi, int_mv *dst_mv) { // Try zero MV first // FIXME should really use something like near/nearest MV and/or MV prediction err = vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride, - xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride); + xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride, + INT_MAX); dst_mv->as_int = 0; @@ -144,7 +147,7 @@ static int find_best_16x16_intra(VP9_COMP *cpi, PREDICTION_MODE *pbest_mode) { xd->plane[0].dst.buf, xd->plane[0].dst.stride, 0, 0, 0); err = vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride, - xd->plane[0].dst.buf, xd->plane[0].dst.stride); + xd->plane[0].dst.buf, xd->plane[0].dst.stride, best_err); // find best if (err < best_err) { diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c index 9d2b2a497..4f7d6f17c 100644 --- a/vp9/encoder/vp9_mcomp.c +++ b/vp9/encoder/vp9_mcomp.c @@ -524,7 +524,9 @@ static int vp9_pattern_search(const MACROBLOCK *x, // Work out the start point for the search bestsad = vfp->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, ref_mv), in_what->stride); + get_buf_from_mv(in_what, ref_mv), in_what->stride, + 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, + sad_per_bit); // Search all possible scales upto the search param around the center point // pick the scale of the point that is best as the starting scale of @@ -540,7 +542,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, bc + candidates[t][i].col}; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } else { @@ -551,7 +553,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, continue; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } @@ -583,7 +585,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, bc + candidates[s][i].col}; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } else { @@ -594,7 +596,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, continue; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } @@ -621,7 +623,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, bc + candidates[s][next_chkpts_indices[i]].col}; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } else { @@ -632,7 +634,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, continue; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } @@ -659,7 +661,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, bc + neighbors[i].col}; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } else { @@ -670,7 +672,7 @@ static int vp9_pattern_search(const MACROBLOCK *x, continue; thissad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv), - in_what->stride); + in_what->stride, bestsad); CHECK_BETTER } } @@ -892,7 +894,8 @@ int vp9_full_range_search_c(const MACROBLOCK *x, *best_mv = *ref_mv; *num00 = 11; best_sad = fn_ptr->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, ref_mv), in_what->stride) + + get_buf_from_mv(in_what, ref_mv), in_what->stride, + 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit); start_row = MAX(-range, x->mv_row_min - ref_mv->row); start_col = MAX(-range, x->mv_col_min - ref_mv->col); @@ -926,7 +929,7 @@ int vp9_full_range_search_c(const MACROBLOCK *x, for (i = 0; i < end_col - c; ++i) { const MV mv = {ref_mv->row + r, ref_mv->col + c + i}; unsigned int sad = fn_ptr->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, &mv), in_what->stride); + get_buf_from_mv(in_what, &mv), in_what->stride, best_sad); if (sad < best_sad) { sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit); if (sad < best_sad) { @@ -972,7 +975,7 @@ int vp9_diamond_search_sad_c(const MACROBLOCK *x, // Check the starting position best_sad = fn_ptr->sdf(what->buf, what->stride, - best_address, in_what->stride) + + best_address, in_what->stride, 0x7fffffff) + mvsad_err_cost(x, best_mv, &fcenter_mv, sad_per_bit); i = 1; @@ -983,7 +986,8 @@ int vp9_diamond_search_sad_c(const MACROBLOCK *x, best_mv->col + ss[i].mv.col}; if (is_mv_in(x, &mv)) { int sad = fn_ptr->sdf(what->buf, what->stride, - best_address + ss[i].offset, in_what->stride); + best_address + ss[i].offset, in_what->stride, + best_sad); if (sad < best_sad) { sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit); if (sad < best_sad) { @@ -1008,7 +1012,7 @@ int vp9_diamond_search_sad_c(const MACROBLOCK *x, if (is_mv_in(x, &this_mv)) { int sad = fn_ptr->sdf(what->buf, what->stride, best_address + ss[best_site].offset, - in_what->stride); + in_what->stride, best_sad); if (sad < best_sad) { sad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit); if (sad < best_sad) { @@ -1073,7 +1077,7 @@ int vp9_diamond_search_sadx4(const MACROBLOCK *x, best_address = in_what; // Check the starting position - bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride) + bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride, 0x7fffffff) + mvsad_err_cost(x, best_mv, &fcenter_mv, sad_per_bit); i = 1; @@ -1125,7 +1129,7 @@ int vp9_diamond_search_sadx4(const MACROBLOCK *x, if (is_mv_in(x, &this_mv)) { const uint8_t *const check_here = ss[i].offset + best_address; unsigned int thissad = fn_ptr->sdf(what, what_stride, check_here, - in_what_stride); + in_what_stride, bestsad); if (thissad < bestsad) { thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit); @@ -1150,7 +1154,7 @@ int vp9_diamond_search_sadx4(const MACROBLOCK *x, if (is_mv_in(x, &this_mv)) { const uint8_t *const check_here = ss[best_site].offset + best_address; unsigned int thissad = fn_ptr->sdf(what, what_stride, check_here, - in_what_stride); + in_what_stride, bestsad); if (thissad < bestsad) { thissad += mvsad_err_cost(&this_mv, &fcenter_mv, mvjsadcost, mvsadcost, sad_per_bit); @@ -1249,7 +1253,7 @@ int vp9_full_search_sad_c(const MACROBLOCK *x, const MV *ref_mv, const int col_max = MIN(ref_mv->col + distance, x->mv_col_max); const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3}; int best_sad = fn_ptr->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, ref_mv), in_what->stride) + + get_buf_from_mv(in_what, ref_mv), in_what->stride, 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit); *best_mv = *ref_mv; @@ -1257,7 +1261,7 @@ int vp9_full_search_sad_c(const MACROBLOCK *x, const MV *ref_mv, for (c = col_min; c < col_max; ++c) { const MV mv = {r, c}; const int sad = fn_ptr->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, &mv), in_what->stride) + + get_buf_from_mv(in_what, &mv), in_what->stride, best_sad) + mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit); if (sad < best_sad) { best_sad = sad; @@ -1282,7 +1286,7 @@ int vp9_full_search_sadx3(const MACROBLOCK *x, const MV *ref_mv, const int col_max = MIN(ref_mv->col + distance, x->mv_col_max); const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3}; unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, ref_mv), in_what->stride) + + get_buf_from_mv(in_what, ref_mv), in_what->stride, 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit); *best_mv = *ref_mv; @@ -1316,7 +1320,7 @@ int vp9_full_search_sadx3(const MACROBLOCK *x, const MV *ref_mv, while (c < col_max) { unsigned int sad = fn_ptr->sdf(what->buf, what->stride, - check_here, in_what->stride); + check_here, in_what->stride, best_sad); if (sad < best_sad) { const MV mv = {r, c}; sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit); @@ -1347,7 +1351,7 @@ int vp9_full_search_sadx8(const MACROBLOCK *x, const MV *ref_mv, const int col_max = MIN(ref_mv->col + distance, x->mv_col_max); const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3}; unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, ref_mv), in_what->stride) + + get_buf_from_mv(in_what, ref_mv), in_what->stride, 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit); *best_mv = *ref_mv; @@ -1405,7 +1409,7 @@ int vp9_full_search_sadx8(const MACROBLOCK *x, const MV *ref_mv, while (c < col_max) { unsigned int sad = fn_ptr->sdf(what->buf, what->stride, - check_here, in_what->stride); + check_here, in_what->stride, best_sad); if (sad < best_sad) { const MV mv = {r, c}; sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit); @@ -1434,7 +1438,7 @@ int vp9_refining_search_sad_c(const MACROBLOCK *x, const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3}; unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride, get_buf_from_mv(in_what, ref_mv), - in_what->stride) + + in_what->stride, 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit); int i, j; @@ -1446,7 +1450,7 @@ int vp9_refining_search_sad_c(const MACROBLOCK *x, ref_mv->col + neighbors[j].col}; if (is_mv_in(x, &mv)) { unsigned int sad = fn_ptr->sdf(what->buf, what->stride, - get_buf_from_mv(in_what, &mv), in_what->stride); + get_buf_from_mv(in_what, &mv), in_what->stride, best_sad); if (sad < best_sad) { sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit); if (sad < best_sad) { @@ -1479,7 +1483,7 @@ int vp9_refining_search_sadx4(const MACROBLOCK *x, const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3}; const uint8_t *best_address = get_buf_from_mv(in_what, ref_mv); unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride, best_address, - in_what->stride) + + in_what->stride, 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit); int i, j; @@ -1520,7 +1524,7 @@ int vp9_refining_search_sadx4(const MACROBLOCK *x, if (is_mv_in(x, &mv)) { unsigned int sad = fn_ptr->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &mv), - in_what->stride); + in_what->stride, best_sad); if (sad < best_sad) { sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit); if (sad < best_sad) { @@ -1559,7 +1563,8 @@ int vp9_refining_search_8p_c(const MACROBLOCK *x, const struct buf_2d *const in_what = &xd->plane[0].pre[0]; const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3}; unsigned int best_sad = fn_ptr->sdaf(what->buf, what->stride, - get_buf_from_mv(in_what, ref_mv), in_what->stride, second_pred) + + get_buf_from_mv(in_what, ref_mv), in_what->stride, + second_pred, 0x7fffffff) + mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit); int i, j; @@ -1572,7 +1577,8 @@ int vp9_refining_search_8p_c(const MACROBLOCK *x, if (is_mv_in(x, &mv)) { unsigned int sad = fn_ptr->sdaf(what->buf, what->stride, - get_buf_from_mv(in_what, &mv), in_what->stride, second_pred); + get_buf_from_mv(in_what, &mv), in_what->stride, + second_pred, best_sad); if (sad < best_sad) { sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit); if (sad < best_sad) { diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c index 14a593b1d..601e64d39 100644 --- a/vp9/encoder/vp9_rdopt.c +++ b/vp9/encoder/vp9_rdopt.c @@ -2131,7 +2131,8 @@ static void mv_pred(VP9_COMP *cpi, MACROBLOCK *x, // Find sad for current vector. this_sad = cpi->fn_ptr[block_size].sdf(src_y_ptr, x->plane[0].src.stride, - ref_y_ptr, ref_y_stride); + ref_y_ptr, ref_y_stride, + 0x7fffffff); // Note if it is the best so far. if (this_sad < best_sad) { diff --git a/vp9/encoder/vp9_sad.c b/vp9/encoder/vp9_sad.c index d06263676..892e90551 100644 --- a/vp9/encoder/vp9_sad.c +++ b/vp9/encoder/vp9_sad.c @@ -35,12 +35,14 @@ static INLINE unsigned int sad(const uint8_t *a, int a_stride, #define sadMxN(m, n) \ unsigned int vp9_sad##m##x##n##_c(const uint8_t *src, int src_stride, \ - const uint8_t *ref, int ref_stride) { \ + const uint8_t *ref, int ref_stride, \ + unsigned int max_sad) { \ return sad(src, src_stride, ref, ref_stride, m, n); \ } \ unsigned int vp9_sad##m##x##n##_avg_c(const uint8_t *src, int src_stride, \ const uint8_t *ref, int ref_stride, \ - const uint8_t *second_pred) { \ + const uint8_t *second_pred, \ + unsigned int max_sad) { \ uint8_t comp_pred[m * n]; \ vp9_comp_avg_pred(comp_pred, second_pred, m, n, ref, ref_stride); \ return sad(src, src_stride, comp_pred, m, m, n); \ @@ -52,7 +54,8 @@ void vp9_sad##m##x##n##x##k##_c(const uint8_t *src, int src_stride, \ unsigned int *sads) { \ int i; \ for (i = 0; i < k; ++i) \ - sads[i] = vp9_sad##m##x##n##_c(src, src_stride, &ref[i], ref_stride); \ + sads[i] = vp9_sad##m##x##n##_c(src, src_stride, &ref[i], ref_stride, \ + 0x7fffffff); \ } #define sadMxNx4D(m, n) \ @@ -61,7 +64,8 @@ void vp9_sad##m##x##n##x4d_c(const uint8_t *src, int src_stride, \ unsigned int *sads) { \ int i; \ for (i = 0; i < 4; ++i) \ - sads[i] = vp9_sad##m##x##n##_c(src, src_stride, refs[i], ref_stride); \ + sads[i] = vp9_sad##m##x##n##_c(src, src_stride, refs[i], ref_stride, \ + 0x7fffffff); \ } // 64x64 diff --git a/vp9/encoder/vp9_variance.h b/vp9/encoder/vp9_variance.h index 4a194b72c..c47fe1335 100644 --- a/vp9/encoder/vp9_variance.h +++ b/vp9/encoder/vp9_variance.h @@ -25,13 +25,15 @@ void variance(const uint8_t *a, int a_stride, typedef unsigned int(*vp9_sad_fn_t)(const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, - int ref_stride); + int ref_stride, + unsigned int max_sad); typedef unsigned int(*vp9_sad_avg_fn_t)(const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, - const uint8_t *second_pred); + const uint8_t *second_pred, + unsigned int max_sad); typedef void (*vp9_sad_multi_fn_t)(const uint8_t *src_ptr, int source_stride, -- 2.40.0