Merge "Fix for loop filter selection procedure" into nextgenv2
diff --git a/test/blend_a64_mask_1d_test.cc b/test/blend_a64_mask_1d_test.cc
new file mode 100644
index 0000000..03e9b7d
--- /dev/null
+++ b/test/blend_a64_mask_1d_test.cc
@@ -0,0 +1,374 @@
+/*
+ * 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 <math.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/register_state_check.h"
+
+#include "test/function_equivalence_test.h"
+
+#include "./vpx_config.h"
+#include "./vpx_dsp_rtcd.h"
+#include "vpx/vpx_integer.h"
+
+#include "./vp10_rtcd.h"
+
+#include "test/acm_random.h"
+#include "vp10/common/enums.h"
+
+#include "vpx_dsp/blend.h"
+
+using libvpx_test::ACMRandom;
+using libvpx_test::FunctionEquivalenceTest;
+using std::tr1::make_tuple;
+
+namespace {
+
+template<typename F, typename T>
+class BlendA64Mask1DTest : public FunctionEquivalenceTest<F> {
+ public:
+ static const int kIterations = 10000;
+ static const int kMaxWidth = MAX_SB_SIZE * 5; // * 5 to cover longer strides
+ static const int kMaxHeight = MAX_SB_SIZE;
+ static const int kBufSize = kMaxWidth * kMaxHeight;
+ static const int kMaxMaskWidth = 2 * MAX_SB_SIZE;
+ static const int kMaxMaskSize = kMaxMaskWidth;
+
+ BlendA64Mask1DTest() : rng_(ACMRandom::DeterministicSeed()) {}
+
+ virtual ~BlendA64Mask1DTest() {}
+
+ virtual void Execute(T *p_src0, T *p_src1) = 0;
+
+ void Common() {
+ w_ = 1 << rng_(MAX_SB_SIZE_LOG2 + 1);
+ h_ = 1 << rng_(MAX_SB_SIZE_LOG2 + 1);
+
+ dst_offset_ = rng_(33);
+ dst_stride_ = rng_(kMaxWidth + 1 - w_) + w_;
+
+ src0_offset_ = rng_(33);
+ src0_stride_ = rng_(kMaxWidth + 1 - w_) + w_;
+
+ src1_offset_ = rng_(33);
+ src1_stride_ = rng_(kMaxWidth + 1 - w_) + w_;
+
+ T *p_src0;
+ T *p_src1;
+
+ switch (rng_(3)) {
+ case 0: // Separate sources
+ p_src0 = src0_;
+ p_src1 = src1_;
+ break;
+ case 1: // src0 == dst
+ p_src0 = dst_tst_;
+ src0_stride_ = dst_stride_;
+ src0_offset_ = dst_offset_;
+ p_src1 = src1_;
+ break;
+ case 2: // src1 == dst
+ p_src0 = src0_;
+ p_src1 = dst_tst_;
+ src1_stride_ = dst_stride_;
+ src1_offset_ = dst_offset_;
+ break;
+ default:
+ FAIL();
+ }
+
+ Execute(p_src0, p_src1);
+
+ for (int r = 0 ; r < h_ ; ++r) {
+ for (int c = 0 ; c < w_ ; ++c) {
+ ASSERT_EQ(dst_ref_[dst_offset_ + r * dst_stride_ + c],
+ dst_tst_[dst_offset_ + r * dst_stride_ + c]);
+ }
+ }
+ }
+
+ ACMRandom rng_;
+
+ T dst_ref_[kBufSize];
+ T dst_tst_[kBufSize];
+ size_t dst_stride_;
+ size_t dst_offset_;
+
+ T src0_[kBufSize];
+ size_t src0_stride_;
+ size_t src0_offset_;
+
+ T src1_[kBufSize];
+ size_t src1_stride_;
+ size_t src1_offset_;
+
+ uint8_t mask_[kMaxMaskSize];
+
+ int w_;
+ int h_;
+};
+
+//////////////////////////////////////////////////////////////////////////////
+// 8 bit version
+//////////////////////////////////////////////////////////////////////////////
+
+typedef void (*F8B)(uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w);
+
+class BlendA64Mask1DTest8B : public BlendA64Mask1DTest<F8B, uint8_t> {
+ protected:
+ void Execute(uint8_t *p_src0, uint8_t *p_src1) {
+ ref_func_(dst_ref_ + dst_offset_, dst_stride_,
+ p_src0 + src0_offset_, src0_stride_,
+ p_src1 + src1_offset_, src1_stride_,
+ mask_, h_, w_);
+
+ tst_func_(dst_tst_ + dst_offset_, dst_stride_,
+ p_src0 + src0_offset_, src0_stride_,
+ p_src1 + src1_offset_, src1_stride_,
+ mask_, h_, w_);
+ }
+};
+
+TEST_P(BlendA64Mask1DTest8B, RandomValues) {
+ for (int iter = 0 ; iter < kIterations && !HasFatalFailure(); ++iter) {
+ for (int i = 0 ; i < kBufSize ; ++i) {
+ dst_ref_[i] = rng_.Rand8();
+ dst_tst_[i] = rng_.Rand8();
+
+ src0_[i] = rng_.Rand8();
+ src1_[i] = rng_.Rand8();
+ }
+
+ for (int i = 0 ; i < kMaxMaskSize ; ++i)
+ mask_[i] = rng_(VPX_BLEND_A64_MAX_ALPHA + 1);
+
+ Common();
+ }
+}
+
+TEST_P(BlendA64Mask1DTest8B, ExtremeValues) {
+ for (int iter = 0 ; iter < kIterations && !HasFatalFailure(); ++iter) {
+ for (int i = 0 ; i < kBufSize ; ++i) {
+ dst_ref_[i] = rng_(2) + 254;
+ dst_tst_[i] = rng_(2) + 254;
+ src0_[i] = rng_(2) + 254;
+ src1_[i] = rng_(2) + 254;
+ }
+
+ for (int i = 0 ; i < kMaxMaskSize ; ++i)
+ mask_[i] = rng_(2) + VPX_BLEND_A64_MAX_ALPHA - 1;
+
+ Common();
+ }
+}
+
+static void blend_a64_hmask_ref(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ uint8_t mask2d[BlendA64Mask1DTest8B::kMaxMaskSize]
+ [BlendA64Mask1DTest8B::kMaxMaskSize];
+
+ for (int row = 0 ; row < h ; ++row)
+ for (int col = 0 ; col < w ; ++col)
+ mask2d[row][col] = mask[col];
+
+ vpx_blend_a64_mask_c(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ &mask2d[0][0], BlendA64Mask1DTest8B::kMaxMaskSize,
+ h, w, 0, 0);
+}
+
+static void blend_a64_vmask_ref(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ uint8_t mask2d[BlendA64Mask1DTest8B::kMaxMaskSize]
+ [BlendA64Mask1DTest8B::kMaxMaskSize];
+
+ for (int row = 0 ; row < h ; ++row)
+ for (int col = 0 ; col < w ; ++col)
+ mask2d[row][col] = mask[row];
+
+ vpx_blend_a64_mask_c(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ &mask2d[0][0], BlendA64Mask1DTest8B::kMaxMaskSize,
+ h, w, 0, 0);
+}
+
+INSTANTIATE_TEST_CASE_P(
+ C, BlendA64Mask1DTest8B,
+ ::testing::Values(
+ make_tuple(blend_a64_hmask_ref, vpx_blend_a64_hmask_c),
+ make_tuple(blend_a64_vmask_ref, vpx_blend_a64_vmask_c)));
+
+#if HAVE_SSE4_1
+INSTANTIATE_TEST_CASE_P(
+ SSE4_1, BlendA64Mask1DTest8B,
+ ::testing::Values(
+ make_tuple(blend_a64_hmask_ref, vpx_blend_a64_hmask_sse4_1),
+ make_tuple(blend_a64_vmask_ref, vpx_blend_a64_vmask_sse4_1)));
+#endif // HAVE_SSE4_1
+
+#if CONFIG_VP9_HIGHBITDEPTH
+//////////////////////////////////////////////////////////////////////////////
+// High bit-depth version
+//////////////////////////////////////////////////////////////////////////////
+
+typedef void (*FHBD)(uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w, int bd);
+
+class BlendA64Mask1DTestHBD : public BlendA64Mask1DTest<FHBD, uint16_t> {
+ protected:
+ void Execute(uint16_t *p_src0, uint16_t *p_src1) {
+ ref_func_(CONVERT_TO_BYTEPTR(dst_ref_ + dst_offset_), dst_stride_,
+ CONVERT_TO_BYTEPTR(p_src0 + src0_offset_), src0_stride_,
+ CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_,
+ mask_, h_, w_, bit_depth_);
+
+ ASM_REGISTER_STATE_CHECK(
+ tst_func_(CONVERT_TO_BYTEPTR(dst_tst_ + dst_offset_), dst_stride_,
+ CONVERT_TO_BYTEPTR(p_src0 + src0_offset_), src0_stride_,
+ CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_,
+ mask_, h_, w_, bit_depth_));
+ }
+
+ int bit_depth_;
+};
+
+TEST_P(BlendA64Mask1DTestHBD, RandomValues) {
+ for (int iter = 0 ; iter < kIterations && !HasFatalFailure(); ++iter) {
+ switch (rng_(3)) {
+ case 0:
+ bit_depth_ = 8;
+ break;
+ case 1:
+ bit_depth_ = 10;
+ break;
+ default:
+ bit_depth_ = 12;
+ break;
+ }
+
+ const int hi = 1 << bit_depth_;
+
+ for (int i = 0 ; i < kBufSize ; ++i) {
+ dst_ref_[i] = rng_(hi);
+ dst_tst_[i] = rng_(hi);
+ src0_[i] = rng_(hi);
+ src1_[i] = rng_(hi);
+ }
+
+ for (int i = 0 ; i < kMaxMaskSize ; ++i)
+ mask_[i] = rng_(VPX_BLEND_A64_MAX_ALPHA + 1);
+
+ Common();
+ }
+}
+
+TEST_P(BlendA64Mask1DTestHBD, ExtremeValues) {
+ for (int iter = 0 ; iter < 1000 && !HasFatalFailure(); ++iter) {
+ switch (rng_(3)) {
+ case 0:
+ bit_depth_ = 8;
+ break;
+ case 1:
+ bit_depth_ = 10;
+ break;
+ default:
+ bit_depth_ = 12;
+ break;
+ }
+
+ const int hi = 1 << bit_depth_;
+ const int lo = hi - 2;
+
+ for (int i = 0 ; i < kBufSize ; ++i) {
+ dst_ref_[i] = rng_(hi - lo) + lo;
+ dst_tst_[i] = rng_(hi - lo) + lo;
+ src0_[i] = rng_(hi - lo) + lo;
+ src1_[i] = rng_(hi - lo) + lo;
+ }
+
+ for (int i = 0 ; i < kMaxMaskSize ; ++i)
+ mask_[i] = rng_(2) + VPX_BLEND_A64_MAX_ALPHA - 1;
+
+ Common();
+ }
+}
+
+static void highbd_blend_a64_hmask_ref(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w, int bd) {
+ uint8_t mask2d[BlendA64Mask1DTestHBD::kMaxMaskSize]
+ [BlendA64Mask1DTestHBD::kMaxMaskSize];
+
+ for (int row = 0 ; row < h ; ++row)
+ for (int col = 0 ; col < w ; ++col)
+ mask2d[row][col] = mask[col];
+
+ vpx_highbd_blend_a64_mask_c(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ &mask2d[0][0],
+ BlendA64Mask1DTestHBD::kMaxMaskSize,
+ h, w, 0, 0, bd);
+}
+
+static void highbd_blend_a64_vmask_ref(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w, int bd) {
+ uint8_t mask2d[BlendA64Mask1DTestHBD::kMaxMaskSize]
+ [BlendA64Mask1DTestHBD::kMaxMaskSize];
+
+ for (int row = 0 ; row < h ; ++row)
+ for (int col = 0 ; col < w ; ++col)
+ mask2d[row][col] = mask[row];
+
+ vpx_highbd_blend_a64_mask_c(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ &mask2d[0][0],
+ BlendA64Mask1DTestHBD::kMaxMaskSize,
+ h, w, 0, 0, bd);
+}
+
+INSTANTIATE_TEST_CASE_P(
+ C, BlendA64Mask1DTestHBD,
+ ::testing::Values(
+ make_tuple(highbd_blend_a64_hmask_ref, vpx_highbd_blend_a64_hmask_c),
+ make_tuple(highbd_blend_a64_vmask_ref, vpx_highbd_blend_a64_vmask_c)));
+
+#if HAVE_SSE4_1
+INSTANTIATE_TEST_CASE_P(
+ SSE4_1, BlendA64Mask1DTestHBD,
+ ::testing::Values(
+ make_tuple(highbd_blend_a64_hmask_ref, vpx_highbd_blend_a64_hmask_sse4_1),
+ make_tuple(highbd_blend_a64_vmask_ref, vpx_highbd_blend_a64_vmask_sse4_1)));
+#endif // HAVE_SSE4_1
+
+#endif // CONFIG_VP9_HIGHBITDEPTH
+} // namespace
diff --git a/test/blend_mask6_test.cc b/test/blend_a64_mask_test.cc
similarity index 77%
rename from test/blend_mask6_test.cc
rename to test/blend_a64_mask_test.cc
index 6afaad7..08ee91d 100644
--- a/test/blend_mask6_test.cc
+++ b/test/blend_a64_mask_test.cc
@@ -26,6 +26,8 @@
#include "test/acm_random.h"
#include "vp10/common/enums.h"
+#include "vpx_dsp/blend.h"
+
using libvpx_test::ACMRandom;
using libvpx_test::FunctionEquivalenceTest;
using std::tr1::make_tuple;
@@ -33,7 +35,7 @@
namespace {
template<typename F, typename T>
-class BlendMask6Test : public FunctionEquivalenceTest<F> {
+class BlendA64MaskTest : public FunctionEquivalenceTest<F> {
protected:
static const int kIterations = 10000;
static const int kMaxWidth = MAX_SB_SIZE * 5; // * 5 to cover longer strides
@@ -42,15 +44,15 @@
static const int kMaxMaskWidth = 2 * MAX_SB_SIZE;
static const int kMaxMaskSize = kMaxMaskWidth * kMaxMaskWidth;
- BlendMask6Test() : rng_(ACMRandom::DeterministicSeed()) {}
+ BlendA64MaskTest() : rng_(ACMRandom::DeterministicSeed()) {}
- virtual ~BlendMask6Test() {}
+ virtual ~BlendA64MaskTest() {}
- virtual void Execute(T *p_src0, T *p_src1) = 0;
+ virtual void Execute(const T *p_src0, const T *p_src1) = 0;
void Common() {
- w_ = 1 << (rng_(MAX_SB_SIZE_LOG2 + 1 - 2) + 2);
- h_ = 1 << (rng_(MAX_SB_SIZE_LOG2 + 1 - 2) + 2);
+ w_ = 1 << rng_(MAX_SB_SIZE_LOG2 + 1);
+ h_ = 1 << rng_(MAX_SB_SIZE_LOG2 + 1);
subx_ = rng_(2);
suby_ = rng_(2);
@@ -131,14 +133,14 @@
//////////////////////////////////////////////////////////////////////////////
typedef void (*F8B)(uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
- const uint8_t *mask, uint32_t mask_stride,
- int h, int w, int suby, int subx);
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, uint32_t mask_stride,
+ int h, int w, int suby, int subx);
-class BlendMask6Test8B : public BlendMask6Test<F8B, uint8_t> {
+class BlendA64MaskTest8B : public BlendA64MaskTest<F8B, uint8_t> {
protected:
- void Execute(uint8_t *p_src0, uint8_t *p_src1) {
+ void Execute(const uint8_t *p_src0, const uint8_t *p_src1) {
ref_func_(dst_ref_ + dst_offset_, dst_stride_,
p_src0 + src0_offset_, src0_stride_,
p_src1 + src1_offset_, src1_stride_,
@@ -153,7 +155,7 @@
}
};
-TEST_P(BlendMask6Test8B, RandomValues) {
+TEST_P(BlendA64MaskTest8B, RandomValues) {
for (int iter = 0 ; iter < kIterations && !HasFatalFailure(); ++iter) {
for (int i = 0 ; i < kBufSize ; ++i) {
dst_ref_[i] = rng_.Rand8();
@@ -164,13 +166,13 @@
}
for (int i = 0 ; i < kMaxMaskSize ; ++i)
- mask_[i] = rng_(65);
+ mask_[i] = rng_(VPX_BLEND_A64_MAX_ALPHA + 1);
Common();
}
}
-TEST_P(BlendMask6Test8B, ExtremeValues) {
+TEST_P(BlendA64MaskTest8B, ExtremeValues) {
for (int iter = 0 ; iter < kIterations && !HasFatalFailure(); ++iter) {
for (int i = 0 ; i < kBufSize ; ++i) {
dst_ref_[i] = rng_(2) + 254;
@@ -180,7 +182,7 @@
}
for (int i = 0 ; i < kMaxMaskSize ; ++i)
- mask_[i] = rng_(2) + 63;
+ mask_[i] = rng_(2) + VPX_BLEND_A64_MAX_ALPHA - 1;
Common();
}
@@ -188,8 +190,9 @@
#if HAVE_SSE4_1
INSTANTIATE_TEST_CASE_P(
- SSE4_1_C_COMPARE, BlendMask6Test8B,
- ::testing::Values(make_tuple(&vpx_blend_mask6_c, &vpx_blend_mask6_sse4_1)));
+ SSE4_1_C_COMPARE, BlendA64MaskTest8B,
+ ::testing::Values(make_tuple(vpx_blend_a64_mask_c,
+ vpx_blend_a64_mask_sse4_1)));
#endif // HAVE_SSE4_1
#if CONFIG_VP9_HIGHBITDEPTH
@@ -198,14 +201,14 @@
//////////////////////////////////////////////////////////////////////////////
typedef void (*FHBD)(uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
- const uint8_t *mask, uint32_t mask_stride,
- int h, int w, int suby, int subx, int bd);
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, uint32_t mask_stride,
+ int h, int w, int suby, int subx, int bd);
-class BlendMask6TestHBD : public BlendMask6Test<FHBD, uint16_t> {
+class BlendA64MaskTestHBD : public BlendA64MaskTest<FHBD, uint16_t> {
protected:
- void Execute(uint16_t *p_src0, uint16_t *p_src1) {
+ void Execute(const uint16_t *p_src0, const uint16_t *p_src1) {
ref_func_(CONVERT_TO_BYTEPTR(dst_ref_ + dst_offset_), dst_stride_,
CONVERT_TO_BYTEPTR(p_src0 + src0_offset_), src0_stride_,
CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_,
@@ -223,7 +226,7 @@
int bit_depth_;
};
-TEST_P(BlendMask6TestHBD, RandomValues) {
+TEST_P(BlendA64MaskTestHBD, RandomValues) {
for (int iter = 0 ; iter < kIterations && !HasFatalFailure(); ++iter) {
switch (rng_(3)) {
case 0:
@@ -247,13 +250,13 @@
}
for (int i = 0 ; i < kMaxMaskSize ; ++i)
- mask_[i] = rng_(65);
+ mask_[i] = rng_(VPX_BLEND_A64_MAX_ALPHA + 1);
Common();
}
}
-TEST_P(BlendMask6TestHBD, ExtremeValues) {
+TEST_P(BlendA64MaskTestHBD, ExtremeValues) {
for (int iter = 0 ; iter < 1000 && !HasFatalFailure(); ++iter) {
switch (rng_(3)) {
case 0:
@@ -278,7 +281,7 @@
}
for (int i = 0 ; i < kMaxMaskSize ; ++i)
- mask_[i] = rng_(65);
+ mask_[i] = rng_(2) + VPX_BLEND_A64_MAX_ALPHA - 1;
Common();
}
@@ -286,9 +289,9 @@
#if HAVE_SSE4_1
INSTANTIATE_TEST_CASE_P(
- SSE4_1_C_COMPARE, BlendMask6TestHBD,
- ::testing::Values(make_tuple(&vpx_highbd_blend_mask6_c,
- &vpx_highbd_blend_mask6_sse4_1)));
+ SSE4_1_C_COMPARE, BlendA64MaskTestHBD,
+ ::testing::Values(make_tuple(vpx_highbd_blend_a64_mask_c,
+ vpx_highbd_blend_a64_mask_sse4_1)));
#endif // HAVE_SSE4_1
#endif // CONFIG_VP9_HIGHBITDEPTH
} // namespace
diff --git a/test/encode_test_driver.cc b/test/encode_test_driver.cc
index cd0b136..753a7e4 100644
--- a/test/encode_test_driver.cc
+++ b/test/encode_test_driver.cc
@@ -188,8 +188,8 @@
const unsigned int w_y = img1->d_w;
const unsigned int h_y = img1->d_h;
- const unsigned int w_uv = ROUNDZ_POWER_OF_TWO(w_y, img1->x_chroma_shift);
- const unsigned int h_uv = ROUNDZ_POWER_OF_TWO(h_y, img1->y_chroma_shift);
+ const unsigned int w_uv = ROUND_POWER_OF_TWO(w_y, img1->x_chroma_shift);
+ const unsigned int h_uv = ROUND_POWER_OF_TWO(h_y, img1->y_chroma_shift);
if (img1->fmt != img2->fmt
|| img1->cs != img2->cs
diff --git a/test/obmc_sad_test.cc b/test/obmc_sad_test.cc
index 9d8c2a2..95d56ae 100644
--- a/test/obmc_sad_test.cc
+++ b/test/obmc_sad_test.cc
@@ -29,7 +29,7 @@
static const int kIterations = 1000;
static const int kMaskMax = 64;
-typedef unsigned int (*ObmcSadF)(const uint8_t *ref, int ref_stride,
+typedef unsigned int (*ObmcSadF)(const uint8_t *pre, int pre_stride,
const int32_t *wsrc, const int32_t *mask);
////////////////////////////////////////////////////////////////////////////////
@@ -45,42 +45,42 @@
};
TEST_P(ObmcSadTest, RandomValues) {
- DECLARE_ALIGNED(32, uint8_t, ref[MAX_SB_SQUARE]);
+ DECLARE_ALIGNED(32, uint8_t, pre[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, wsrc[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, mask[MAX_SB_SQUARE]);
for (int iter = 0 ; iter < kIterations && !HasFatalFailure() ; ++iter) {
- const int ref_stride = rng_(MAX_SB_SIZE + 1);
+ const int pre_stride = rng_(MAX_SB_SIZE + 1);
for (int i = 0 ; i < MAX_SB_SQUARE ; ++i) {
- ref[i] = rng_.Rand8();
+ pre[i] = rng_.Rand8();
wsrc[i] = rng_.Rand8() * rng_(kMaskMax * kMaskMax + 1);
mask[i] = rng_(kMaskMax * kMaskMax + 1);
}
- const unsigned int ref_res = ref_func_(ref, ref_stride, wsrc, mask);
- const unsigned int tst_res = tst_func_(ref, ref_stride, wsrc, mask);
+ const unsigned int ref_res = ref_func_(pre, pre_stride, wsrc, mask);
+ const unsigned int tst_res = tst_func_(pre, pre_stride, wsrc, mask);
ASSERT_EQ(ref_res, tst_res);
}
}
TEST_P(ObmcSadTest, ExtremeValues) {
- DECLARE_ALIGNED(32, uint8_t, ref[MAX_SB_SQUARE]);
+ DECLARE_ALIGNED(32, uint8_t, pre[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, wsrc[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, mask[MAX_SB_SQUARE]);
for (int iter = 0 ; iter < MAX_SB_SIZE && !HasFatalFailure() ; ++iter) {
- const int ref_stride = iter;
+ const int pre_stride = iter;
for (int i = 0 ; i < MAX_SB_SQUARE ; ++i) {
- ref[i] = UINT8_MAX;
+ pre[i] = UINT8_MAX;
wsrc[i] = UINT8_MAX * kMaskMax * kMaskMax;
mask[i] = kMaskMax * kMaskMax;
}
- const unsigned int ref_res = ref_func_(ref, ref_stride, wsrc, mask);
- const unsigned int tst_res = tst_func_(ref, ref_stride, wsrc, mask);
+ const unsigned int ref_res = ref_func_(pre, pre_stride, wsrc, mask);
+ const unsigned int tst_res = tst_func_(pre, pre_stride, wsrc, mask);
ASSERT_EQ(ref_res, tst_res);
}
@@ -126,22 +126,22 @@
};
TEST_P(ObmcSadHBDTest, RandomValues) {
- DECLARE_ALIGNED(32, uint16_t, ref[MAX_SB_SQUARE]);
+ DECLARE_ALIGNED(32, uint16_t, pre[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, wsrc[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, mask[MAX_SB_SQUARE]);
for (int iter = 0 ; iter < kIterations && !HasFatalFailure() ; ++iter) {
- const int ref_stride = rng_(MAX_SB_SIZE + 1);
+ const int pre_stride = rng_(MAX_SB_SIZE + 1);
for (int i = 0 ; i < MAX_SB_SQUARE ; ++i) {
- ref[i] = rng_(1<<12);
+ pre[i] = rng_(1<<12);
wsrc[i] = rng_(1<<12) * rng_(kMaskMax * kMaskMax + 1);
mask[i] = rng_(kMaskMax * kMaskMax + 1);
}
- const unsigned int ref_res = ref_func_(CONVERT_TO_BYTEPTR(ref), ref_stride,
+ const unsigned int ref_res = ref_func_(CONVERT_TO_BYTEPTR(pre), pre_stride,
wsrc, mask);
- const unsigned int tst_res = tst_func_(CONVERT_TO_BYTEPTR(ref), ref_stride,
+ const unsigned int tst_res = tst_func_(CONVERT_TO_BYTEPTR(pre), pre_stride,
wsrc, mask);
ASSERT_EQ(ref_res, tst_res);
@@ -149,22 +149,22 @@
}
TEST_P(ObmcSadHBDTest, ExtremeValues) {
- DECLARE_ALIGNED(32, uint16_t, ref[MAX_SB_SQUARE]);
+ DECLARE_ALIGNED(32, uint16_t, pre[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, wsrc[MAX_SB_SQUARE]);
DECLARE_ALIGNED(32, int32_t, mask[MAX_SB_SQUARE]);
for (int iter = 0 ; iter < MAX_SB_SIZE && !HasFatalFailure() ; ++iter) {
- const int ref_stride = iter;
+ const int pre_stride = iter;
for (int i = 0 ; i < MAX_SB_SQUARE ; ++i) {
- ref[i] = (1 << 12) - 1;
+ pre[i] = (1 << 12) - 1;
wsrc[i] = ((1 << 12) - 1) * kMaskMax * kMaskMax;
mask[i] = kMaskMax * kMaskMax;
}
- const unsigned int ref_res = ref_func_(CONVERT_TO_BYTEPTR(ref), ref_stride,
+ const unsigned int ref_res = ref_func_(CONVERT_TO_BYTEPTR(pre), pre_stride,
wsrc, mask);
- const unsigned int tst_res = tst_func_(CONVERT_TO_BYTEPTR(ref), ref_stride,
+ const unsigned int tst_res = tst_func_(CONVERT_TO_BYTEPTR(pre), pre_stride,
wsrc, mask);
ASSERT_EQ(ref_res, tst_res);
diff --git a/test/test.mk b/test/test.mk
index fcd565c..67fe705 100644
--- a/test/test.mk
+++ b/test/test.mk
@@ -178,11 +178,12 @@
LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += sum_squares_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += subtract_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += blend_a64_mask_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += blend_a64_mask_1d_test.cc
ifeq ($(CONFIG_EXT_INTER),yes)
LIBVPX_TEST_SRCS-$(HAVE_SSSE3) += masked_variance_test.cc
LIBVPX_TEST_SRCS-$(HAVE_SSSE3) += masked_sad_test.cc
-LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += blend_mask6_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += vp10_wedge_utils_test.cc
endif
diff --git a/test/vp10_convolve_optimz_test.cc b/test/vp10_convolve_optimz_test.cc
index 7999087..ec77035 100644
--- a/test/vp10_convolve_optimz_test.cc
+++ b/test/vp10_convolve_optimz_test.cc
@@ -24,12 +24,25 @@
typedef void (*conv_filter_t)(const uint8_t*, int, uint8_t*, int,
int, int, const InterpFilterParams,
const int, int, int);
+#if CONFIG_VP9_HIGHBITDEPTH
+typedef void (*hbd_conv_filter_t)(const uint16_t*, int, uint16_t*, int,
+ int, int, const InterpFilterParams,
+ const int, int, int, int);
+#endif
+
// Test parameter list:
// <convolve_horiz_func, convolve_vert_func,
// <width, height>, filter_params, subpel_x_q4, avg>
typedef tuple<int, int> BlockDimension;
typedef tuple<conv_filter_t, conv_filter_t, BlockDimension, INTERP_FILTER,
int, int> ConvParams;
+#if CONFIG_VP9_HIGHBITDEPTH
+// Test parameter list:
+// <convolve_horiz_func, convolve_vert_func,
+// <width, height>, filter_params, subpel_x_q4, avg, bit_dpeth>
+typedef tuple<hbd_conv_filter_t, hbd_conv_filter_t, BlockDimension,
+ INTERP_FILTER, int, int, int> HbdConvParams;
+#endif
// Note:
// src_ and src_ref_ have special boundary requirement
@@ -75,11 +88,8 @@
void RunVertFilterBitExactCheck();
private:
- void PrepFilterBuffer(uint8_t *src, uint8_t *src_ref,
- uint8_t *dst, uint8_t *dst_ref,
- int w, int h);
- void DiffFilterBuffer(const uint8_t *buf, const uint8_t *buf_ref,
- int w, int h, int fgroup, int findex);
+ void PrepFilterBuffer(int w, int h);
+ void DiffFilterBuffer();
conv_filter_t conv_horiz_;
conv_filter_t conv_vert_;
uint8_t *alloc_;
@@ -94,18 +104,16 @@
int avg_;
};
-void VP10ConvolveOptimzTest::PrepFilterBuffer(uint8_t *src, uint8_t *src_ref,
- uint8_t *dst, uint8_t *dst_ref,
- int w, int h) {
+void VP10ConvolveOptimzTest::PrepFilterBuffer(int w, int h) {
int r, c;
ACMRandom rnd(ACMRandom::DeterministicSeed());
memset(alloc_, 0, 4 * maxBlockSize * sizeof(alloc_[0]));
- uint8_t *src_ptr = src;
- uint8_t *dst_ptr = dst;
- uint8_t *src_ref_ptr = src_ref;
- uint8_t *dst_ref_ptr = dst_ref;
+ uint8_t *src_ptr = src_;
+ uint8_t *dst_ptr = dst_;
+ uint8_t *src_ref_ptr = src_ref_;
+ uint8_t *dst_ref_ptr = dst_ref_;
for (r = 0; r < height_; ++r) {
for (c = 0; c < width_; ++c) {
@@ -121,21 +129,17 @@
}
}
-void VP10ConvolveOptimzTest::DiffFilterBuffer(const uint8_t *buf,
- const uint8_t *buf_ref,
- int w, int h,
- int filter_group,
- int filter_index) {
+void VP10ConvolveOptimzTest::DiffFilterBuffer() {
int r, c;
- const uint8_t *dst_ptr = buf;
- const uint8_t *dst_ref_ptr = buf_ref;
- for (r = 0; r < h; ++r) {
- for (c = 0; c < w; ++c) {
+ const uint8_t *dst_ptr = dst_;
+ const uint8_t *dst_ref_ptr = dst_ref_;
+ for (r = 0; r < height_; ++r) {
+ for (c = 0; c < width_; ++c) {
EXPECT_EQ((uint8_t)dst_ref_ptr[c], (uint8_t)dst_ptr[c])
<< "Error at row: " << r << " col: " << c << " "
- << "w = " << w << " " << "h = " << h << " "
- << "filter group index = " << filter_group << " "
- << "filter index = " << filter_index;
+ << "w = " << width_ << " " << "h = " << height_ << " "
+ << "filter group index = " << filter_ << " "
+ << "filter index = " << subpel_;
}
dst_ptr += stride;
dst_ref_ptr += stride;
@@ -143,7 +147,7 @@
}
void VP10ConvolveOptimzTest::RunHorizFilterBitExactCheck() {
- PrepFilterBuffer(src_, src_ref_, dst_, dst_ref_, testMaxBlk, testMaxBlk);
+ PrepFilterBuffer(testMaxBlk, testMaxBlk);
InterpFilterParams filter_params = vp10_get_interp_filter_params(filter_);
@@ -153,14 +157,14 @@
conv_horiz_(src_, stride, dst_, stride, width_, height_,
filter_params, subpel_, x_step_q4, avg_);
- DiffFilterBuffer(dst_, dst_ref_, width_, height_, filter_, subpel_);
+ DiffFilterBuffer();
// Note:
// Here we need calculate a height which is different from the specified one
// and test again.
int intermediate_height =
(((height_ - 1) * 16 + subpel_) >> SUBPEL_BITS) + filter_params.taps;
- PrepFilterBuffer(src_, src_ref_, dst_, dst_ref_, testMaxBlk, testMaxBlk);
+ PrepFilterBuffer(testMaxBlk, testMaxBlk);
vp10_convolve_horiz_c(src_ref_, stride, dst_ref_, stride, width_,
intermediate_height, filter_params, subpel_, x_step_q4,
@@ -170,12 +174,11 @@
intermediate_height, filter_params, subpel_, x_step_q4,
avg_);
- DiffFilterBuffer(dst_, dst_ref_, width_, intermediate_height, filter_,
- subpel_);
+ DiffFilterBuffer();
}
void VP10ConvolveOptimzTest::RunVertFilterBitExactCheck() {
- PrepFilterBuffer(src_, src_ref_, dst_, dst_ref_, testMaxBlk, testMaxBlk);
+ PrepFilterBuffer(testMaxBlk, testMaxBlk);
InterpFilterParams filter_params = vp10_get_interp_filter_params(filter_);
@@ -185,7 +188,7 @@
conv_vert_(src_, stride, dst_, stride, width_, height_,
filter_params, subpel_, x_step_q4, avg_);
- DiffFilterBuffer(dst_, dst_ref_, width_, height_, filter_, subpel_);
+ DiffFilterBuffer();
}
TEST_P(VP10ConvolveOptimzTest, HorizBitExactCheck) {
@@ -197,6 +200,7 @@
using std::tr1::make_tuple;
+#if (HAVE_SSSE3 || HAVE_SSE4_1) && CONFIG_EXT_INTERP
const BlockDimension kBlockDim[] = {
make_tuple(2, 2),
make_tuple(2, 4),
@@ -218,14 +222,15 @@
make_tuple(128, 128),
};
-#if HAVE_SSSE3 && CONFIG_EXT_INTERP
// 10/12-tap filters
const INTERP_FILTER kFilter[] = {6, 4, 2};
const int kSubpelQ4[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};
const int kAvg[] = {0, 1};
+#endif
+#if HAVE_SSSE3 && CONFIG_EXT_INTERP
INSTANTIATE_TEST_CASE_P(
SSSE3, VP10ConvolveOptimzTest,
::testing::Combine(
@@ -236,4 +241,167 @@
::testing::ValuesIn(kSubpelQ4),
::testing::ValuesIn(kAvg)));
#endif // HAVE_SSSE3 && CONFIG_EXT_INTERP
+
+#if CONFIG_VP9_HIGHBITDEPTH
+typedef ::testing::TestWithParam<HbdConvParams> TestWithHbdConvParams;
+class VP10HbdConvolveOptimzTest : public TestWithHbdConvParams {
+ public:
+ virtual ~VP10HbdConvolveOptimzTest() {}
+ virtual void SetUp() {
+ conv_horiz_ = GET_PARAM(0);
+ conv_vert_ = GET_PARAM(1);
+ BlockDimension block = GET_PARAM(2);
+ width_ = std::tr1::get<0>(block);
+ height_ = std::tr1::get<1>(block);
+ filter_ = GET_PARAM(3);
+ subpel_ = GET_PARAM(4);
+ avg_ = GET_PARAM(5);
+ bit_depth_ = GET_PARAM(6);
+
+ alloc_ = new uint16_t[maxBlockSize * 4];
+ src_ = alloc_ + (vertiOffset * maxWidth);
+ src_ += horizOffset;
+ src_ref_ = src_ + maxBlockSize;
+
+ dst_ = alloc_ + 2 * maxBlockSize;
+ dst_ref_ = alloc_ + 3 * maxBlockSize;
+ }
+
+ virtual void TearDown() {
+ delete[] alloc_;
+ libvpx_test::ClearSystemState();
+ }
+
+ protected:
+ void RunHorizFilterBitExactCheck();
+ void RunVertFilterBitExactCheck();
+
+ private:
+ void PrepFilterBuffer(int w, int h);
+ void DiffFilterBuffer();
+ hbd_conv_filter_t conv_horiz_;
+ hbd_conv_filter_t conv_vert_;
+ uint16_t *alloc_;
+ uint16_t *src_;
+ uint16_t *dst_;
+ uint16_t *src_ref_;
+ uint16_t *dst_ref_;
+ int width_;
+ int height_;
+ int filter_;
+ int subpel_;
+ int avg_;
+ int bit_depth_;
+};
+
+void VP10HbdConvolveOptimzTest::PrepFilterBuffer(int w, int h) {
+ int r, c;
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+
+ memset(alloc_, 0, 4 * maxBlockSize * sizeof(alloc_[0]));
+
+ uint16_t *src_ptr = src_;
+ uint16_t *dst_ptr = dst_;
+ uint16_t *dst_ref_ptr = dst_ref_;
+ uint16_t hbd_mask = (1 << bit_depth_) - 1;
+
+ for (r = 0; r < height_; ++r) {
+ for (c = 0; c < width_; ++c) {
+ src_ptr[c] = rnd.Rand16() & hbd_mask;
+ dst_ptr[c] = rnd.Rand16() & hbd_mask;
+ dst_ref_ptr[c] = dst_ptr[c];
+ }
+ src_ptr += stride;
+ dst_ptr += stride;
+ dst_ref_ptr += stride;
+ }
+}
+
+void VP10HbdConvolveOptimzTest::DiffFilterBuffer() {
+ int r, c;
+ const uint16_t *dst_ptr = dst_;
+ const uint16_t *dst_ref_ptr = dst_ref_;
+ for (r = 0; r < height_; ++r) {
+ for (c = 0; c < width_; ++c) {
+ EXPECT_EQ((uint16_t)dst_ref_ptr[c], (uint16_t)dst_ptr[c])
+ << "Error at row: " << r << " col: " << c << " "
+ << "w = " << width_ << " " << "h = " << height_ << " "
+ << "filter group index = " << filter_ << " "
+ << "filter index = " << subpel_ << " "
+ << "bit depth = " << bit_depth_;
+ }
+ dst_ptr += stride;
+ dst_ref_ptr += stride;
+ }
+}
+
+void VP10HbdConvolveOptimzTest::RunHorizFilterBitExactCheck() {
+ PrepFilterBuffer(testMaxBlk, testMaxBlk);
+
+ InterpFilterParams filter_params = vp10_get_interp_filter_params(filter_);
+
+ vp10_highbd_convolve_horiz_c(src_, stride, dst_ref_, stride, width_,
+ height_, filter_params, subpel_, x_step_q4,
+ avg_, bit_depth_);
+
+ conv_horiz_(src_, stride, dst_, stride, width_, height_,
+ filter_params, subpel_, x_step_q4, avg_, bit_depth_);
+
+ DiffFilterBuffer();
+
+ // Note:
+ // Here we need calculate a height which is different from the specified one
+ // and test again.
+ int intermediate_height =
+ (((height_ - 1) * 16 + subpel_) >> SUBPEL_BITS) + filter_params.taps;
+ PrepFilterBuffer(testMaxBlk, testMaxBlk);
+
+ vp10_highbd_convolve_horiz_c(src_, stride, dst_ref_, stride, width_,
+ intermediate_height, filter_params, subpel_,
+ x_step_q4, avg_, bit_depth_);
+
+ conv_horiz_(src_, stride, dst_, stride, width_, intermediate_height,
+ filter_params, subpel_, x_step_q4, avg_, bit_depth_);
+
+ DiffFilterBuffer();
+}
+
+void VP10HbdConvolveOptimzTest::RunVertFilterBitExactCheck() {
+ PrepFilterBuffer(testMaxBlk, testMaxBlk);
+
+ InterpFilterParams filter_params = vp10_get_interp_filter_params(filter_);
+
+ vp10_highbd_convolve_vert_c(src_, stride, dst_ref_, stride, width_, height_,
+ filter_params, subpel_, x_step_q4, avg_,
+ bit_depth_);
+
+ conv_vert_(src_, stride, dst_, stride, width_, height_,
+ filter_params, subpel_, x_step_q4, avg_, bit_depth_);
+
+ DiffFilterBuffer();
+}
+
+TEST_P(VP10HbdConvolveOptimzTest, HorizBitExactCheck) {
+ RunHorizFilterBitExactCheck();
+}
+TEST_P(VP10HbdConvolveOptimzTest, VertBitExactCheck) {
+ RunVertFilterBitExactCheck();
+}
+
+#if HAVE_SSE4_1 && CONFIG_EXT_INTERP
+
+const int kBitdepth[] = {10, 12};
+
+INSTANTIATE_TEST_CASE_P(
+ SSE4_1, VP10HbdConvolveOptimzTest,
+ ::testing::Combine(
+ ::testing::Values(vp10_highbd_convolve_horiz_sse4_1),
+ ::testing::Values(vp10_highbd_convolve_vert_sse4_1),
+ ::testing::ValuesIn(kBlockDim),
+ ::testing::ValuesIn(kFilter),
+ ::testing::ValuesIn(kSubpelQ4),
+ ::testing::ValuesIn(kAvg),
+ ::testing::ValuesIn(kBitdepth)));
+#endif // HAVE_SSE4_1 && CONFIG_EXT_INTERP
+#endif // CONFIG_VP9_HIGHBITDEPTH
} // namespace
diff --git a/test/vp10_wedge_utils_test.cc b/test/vp10_wedge_utils_test.cc
index 930a598..7a541b2 100644
--- a/test/vp10_wedge_utils_test.cc
+++ b/test/vp10_wedge_utils_test.cc
@@ -104,7 +104,7 @@
p1[j] = clamp(s[j] + rng_(33) - 16, 0, UINT8_MAX);
}
- vpx_blend_mask6(p, w, p0, w, p1, w, m, w, h, w, 0, 0);
+ vpx_blend_a64_mask(p, w, p0, w, p1, w, m, w, h, w, 0, 0);
vpx_subtract_block(h, w, r0, w, s, w, p0, w);
vpx_subtract_block(h, w, r1, w, s, w, p1, w);
diff --git a/vp10/common/filter.c b/vp10/common/filter.c
index 5dde3ab..8427237 100644
--- a/vp10/common/filter.c
+++ b/vp10/common/filter.c
@@ -342,3 +342,25 @@
(void)index;
return NULL;
}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+HbdSubpelFilterCoeffs vp10_hbd_get_subpel_filter_ver_signal_dir(
+ const InterpFilterParams p, int index) {
+#if CONFIG_EXT_INTERP && HAVE_SSE4_1
+ if (p.filter_ptr == (const int16_t *)sub_pel_filters_12sharp) {
+ return &sub_pel_filters_12sharp_highbd_ver_signal_dir[index][0];
+ }
+ if (p.filter_ptr == (const int16_t *)sub_pel_filters_10sharp) {
+ return &sub_pel_filters_10sharp_highbd_ver_signal_dir[index][0];
+ }
+#endif
+#if USE_TEMPORALFILTER_12TAP && HAVE_SSE4_1
+ if (p.filter_ptr == (const int16_t *)sub_pel_filters_temporalfilter_12) {
+ return &sub_pel_filters_temporalfilter_12_highbd_ver_signal_dir[index][0];
+ }
+#endif
+ (void)p;
+ (void)index;
+ return NULL;
+}
+#endif
diff --git a/vp10/common/filter.h b/vp10/common/filter.h
index 591ac4d..5ebf2a5 100644
--- a/vp10/common/filter.h
+++ b/vp10/common/filter.h
@@ -95,6 +95,10 @@
#if USE_TEMPORALFILTER_12TAP
extern const int8_t sub_pel_filters_temporalfilter_12_signal_dir[15][2][16];
extern const int8_t sub_pel_filters_temporalfilter_12_ver_signal_dir[15][6][16];
+#if CONFIG_VP9_HIGHBITDEPTH
+extern const
+int16_t sub_pel_filters_temporalfilter_12_highbd_ver_signal_dir[15][6][8];
+#endif
#endif
#if CONFIG_EXT_INTERP
@@ -102,15 +106,26 @@
extern const int8_t sub_pel_filters_10sharp_signal_dir[15][2][16];
extern const int8_t sub_pel_filters_12sharp_ver_signal_dir[15][6][16];
extern const int8_t sub_pel_filters_10sharp_ver_signal_dir[15][6][16];
+#if CONFIG_VP9_HIGHBITDEPTH
+extern const int16_t sub_pel_filters_12sharp_highbd_ver_signal_dir[15][6][8];
+extern const int16_t sub_pel_filters_10sharp_highbd_ver_signal_dir[15][6][8];
+#endif
#endif
typedef const int8_t (*SubpelFilterCoeffs)[16];
+#if CONFIG_VP9_HIGHBITDEPTH
+typedef const int16_t (*HbdSubpelFilterCoeffs)[8];
+#endif
SubpelFilterCoeffs vp10_get_subpel_filter_signal_dir(
const InterpFilterParams p, int index);
SubpelFilterCoeffs vp10_get_subpel_filter_ver_signal_dir(
const InterpFilterParams p, int index);
+#if CONFIG_VP9_HIGHBITDEPTH
+HbdSubpelFilterCoeffs vp10_hbd_get_subpel_filter_ver_signal_dir(
+ const InterpFilterParams p, int index);
+#endif
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/common/idct.h b/vp10/common/idct.h
index ffdad0c..5d52314 100644
--- a/vp10/common/idct.h
+++ b/vp10/common/idct.h
@@ -57,6 +57,12 @@
int eob);
void vp10_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob);
+void vp10_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob);
+void vp10_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob);
+void vp10_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob);
void vp10_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest,
int stride, int eob, TX_TYPE tx_type, int lossless);
diff --git a/vp10/common/reconinter.c b/vp10/common/reconinter.c
index 70cf5e7..53fd1a6 100644
--- a/vp10/common/reconinter.c
+++ b/vp10/common/reconinter.c
@@ -15,6 +15,7 @@
#include "./vpx_config.h"
#include "vpx/vpx_integer.h"
+#include "vpx_dsp/blend.h"
#include "vp10/common/blockd.h"
#include "vp10/common/reconinter.h"
@@ -448,8 +449,8 @@
#if CONFIG_SUPERTX
static void build_masked_compound_wedge_extend(
uint8_t *dst, int dst_stride,
- uint8_t *src0, int src0_stride,
- uint8_t *src1, int src1_stride,
+ const uint8_t *src0, int src0_stride,
+ const uint8_t *src1, int src1_stride,
int wedge_index,
int wedge_sign,
BLOCK_SIZE sb_type,
@@ -459,18 +460,18 @@
const int subw = (2 << b_width_log2_lookup[sb_type]) == w;
const uint8_t *mask = vp10_get_soft_mask(
wedge_index, wedge_sign, sb_type, wedge_offset_x, wedge_offset_y);
- vpx_blend_mask6(dst, dst_stride,
- src0, src0_stride,
- src1, src1_stride,
- mask, MASK_MASTER_STRIDE,
- h, w, subh, subw);
+ vpx_blend_a64_mask(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, MASK_MASTER_STRIDE,
+ h, w, subh, subw);
}
#if CONFIG_VP9_HIGHBITDEPTH
static void build_masked_compound_wedge_extend_highbd(
uint8_t *dst_8, int dst_stride,
- uint8_t *src0_8, int src0_stride,
- uint8_t *src1_8, int src1_stride,
+ const uint8_t *src0_8, int src0_stride,
+ const uint8_t *src1_8, int src1_stride,
int wedge_index, int wedge_sign,
BLOCK_SIZE sb_type,
int wedge_offset_x, int wedge_offset_y,
@@ -479,52 +480,54 @@
const int subw = (2 << b_width_log2_lookup[sb_type]) == w;
const uint8_t *mask = vp10_get_soft_mask(
wedge_index, wedge_sign, sb_type, wedge_offset_x, wedge_offset_y);
- vpx_highbd_blend_mask6(dst_8, dst_stride,
- src0_8, src0_stride,
- src1_8, src1_stride,
- mask, MASK_MASTER_STRIDE,
- h, w, subh, subw, bd);
+ vpx_highbd_blend_a64_mask(dst_8, dst_stride,
+ src0_8, src0_stride,
+ src1_8, src1_stride,
+ mask, MASK_MASTER_STRIDE,
+ h, w, subh, subw, bd);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // CONFIG_SUPERTX
-static void build_masked_compound_wedge(uint8_t *dst, int dst_stride,
- uint8_t *src0, int src0_stride,
- uint8_t *src1, int src1_stride,
- int wedge_index, int wedge_sign,
- BLOCK_SIZE sb_type,
- int h, int w) {
+static void build_masked_compound_wedge(
+ uint8_t *dst, int dst_stride,
+ const uint8_t *src0, int src0_stride,
+ const uint8_t *src1, int src1_stride,
+ int wedge_index, int wedge_sign,
+ BLOCK_SIZE sb_type,
+ int h, int w) {
// Derive subsampling from h and w passed in. May be refactored to
// pass in subsampling factors directly.
const int subh = (2 << b_height_log2_lookup[sb_type]) == h;
const int subw = (2 << b_width_log2_lookup[sb_type]) == w;
const uint8_t *mask = vp10_get_contiguous_soft_mask(wedge_index, wedge_sign,
sb_type);
- vpx_blend_mask6(dst, dst_stride,
- src0, src0_stride,
- src1, src1_stride,
- mask, 4 * num_4x4_blocks_wide_lookup[sb_type],
- h, w, subh, subw);
+ vpx_blend_a64_mask(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, 4 * num_4x4_blocks_wide_lookup[sb_type],
+ h, w, subh, subw);
}
#if CONFIG_VP9_HIGHBITDEPTH
-static void build_masked_compound_wedge_highbd(uint8_t *dst_8, int dst_stride,
- uint8_t *src0_8, int src0_stride,
- uint8_t *src1_8, int src1_stride,
- int wedge_index, int wedge_sign,
- BLOCK_SIZE sb_type,
- int h, int w, int bd) {
+static void build_masked_compound_wedge_highbd(
+ uint8_t *dst_8, int dst_stride,
+ const uint8_t *src0_8, int src0_stride,
+ const uint8_t *src1_8, int src1_stride,
+ int wedge_index, int wedge_sign,
+ BLOCK_SIZE sb_type,
+ int h, int w, int bd) {
// Derive subsampling from h and w passed in. May be refactored to
// pass in subsampling factors directly.
const int subh = (2 << b_height_log2_lookup[sb_type]) == h;
const int subw = (2 << b_width_log2_lookup[sb_type]) == w;
const uint8_t *mask = vp10_get_contiguous_soft_mask(wedge_index, wedge_sign,
sb_type);
- vpx_highbd_blend_mask6(dst_8, dst_stride,
- src0_8, src0_stride,
- src1_8, src1_stride,
- mask, 4 * num_4x4_blocks_wide_lookup[sb_type],
- h, w, subh, subw, bd);
+ vpx_highbd_blend_a64_mask(dst_8, dst_stride,
+ src0_8, src0_stride,
+ src1_8, src1_stride,
+ mask, 4 * num_4x4_blocks_wide_lookup[sb_type],
+ h, w, subh, subw, bd);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
@@ -1064,177 +1067,123 @@
28, 18, 10, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
-static void generate_1dmask(int length, uint8_t *mask, int plane) {
+static const uint8_t* get_supertx_mask(int length, int plane) {
switch (length) {
case 8:
- memcpy(mask, plane ? mask_8_uv : mask_8, length);
- break;
+ return plane ? mask_8_uv : mask_8;
case 16:
- memcpy(mask, plane ? mask_16_uv : mask_16, length);
- break;
+ return plane ? mask_16_uv : mask_16;
case 32:
- memcpy(mask, plane ? mask_32_uv : mask_32, length);
- break;
+ return plane ? mask_32_uv : mask_32;
default:
assert(0);
}
+ return NULL;
}
void vp10_build_masked_inter_predictor_complex(
MACROBLOCKD *xd,
- uint8_t *dst, int dst_stride, uint8_t *dst2, int dst2_stride,
+ uint8_t *dst, int dst_stride,
+ const uint8_t *pre, int pre_stride,
int mi_row, int mi_col,
int mi_row_ori, int mi_col_ori, BLOCK_SIZE bsize, BLOCK_SIZE top_bsize,
PARTITION_TYPE partition, int plane) {
- int i, j;
const struct macroblockd_plane *pd = &xd->plane[plane];
- uint8_t mask[MAX_TX_SIZE];
- int top_w = 4 << b_width_log2_lookup[top_bsize];
- int top_h = 4 << b_height_log2_lookup[top_bsize];
- int w = 4 << b_width_log2_lookup[bsize];
- int h = 4 << b_height_log2_lookup[bsize];
- int w_offset = (mi_col - mi_col_ori) * MI_SIZE;
- int h_offset = (mi_row - mi_row_ori) * MI_SIZE;
+ const int ssx = pd->subsampling_x;
+ const int ssy = pd->subsampling_y;
+ const int top_w = (4 << b_width_log2_lookup[top_bsize]) >> ssx;
+ const int top_h = (4 << b_height_log2_lookup[top_bsize]) >> ssy;
+ const int w = (4 << b_width_log2_lookup[bsize]) >> ssx;
+ const int h = (4 << b_height_log2_lookup[bsize]) >> ssy;
+ const int w_offset = ((mi_col - mi_col_ori) * MI_SIZE) >> ssx;
+ const int h_offset = ((mi_row - mi_row_ori) * MI_SIZE) >> ssy;
+
+ int w_remain, h_remain;
#if CONFIG_VP9_HIGHBITDEPTH
- uint16_t *dst16= CONVERT_TO_SHORTPTR(dst);
- uint16_t *dst216 = CONVERT_TO_SHORTPTR(dst2);
- int b_hdb = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? 1 : 0;
+ const int is_hdb = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? 1 : 0;
#endif // CONFIG_VP9_HIGHBITDEPTH
assert(bsize <= BLOCK_32X32);
-
- top_w >>= pd->subsampling_x;
- top_h >>= pd->subsampling_y;
- w >>= pd->subsampling_x;
- h >>= pd->subsampling_y;
- w_offset >>= pd->subsampling_x;
- h_offset >>= pd->subsampling_y;
+ assert(IMPLIES(plane == 0, ssx == 0));
+ assert(IMPLIES(plane == 0, ssy == 0));
switch (partition) {
- case PARTITION_HORZ:
- {
+ case PARTITION_HORZ: {
+ const uint8_t *const mask = get_supertx_mask(h, ssy);
+
+ w_remain = top_w;
+ h_remain = top_h - h_offset - h;
+ dst += h_offset * dst_stride;
+ pre += h_offset * pre_stride;
+
#if CONFIG_VP9_HIGHBITDEPTH
- if (b_hdb) {
- uint16_t *dst_tmp = dst16 + h_offset * dst_stride;
- uint16_t *dst2_tmp = dst216 + h_offset * dst2_stride;
- generate_1dmask(h, mask + h_offset,
- plane && xd->plane[plane].subsampling_y);
-
- for (i = h_offset; i < h_offset + h; i++) {
- for (j = 0; j < top_w; j++) {
- const int m = mask[i]; assert(m >= 0 && m <= 64);
- if (m == 64)
- continue;
-
- if (m == 0)
- dst_tmp[j] = dst2_tmp[j];
- else
- dst_tmp[j] = ROUND_POWER_OF_TWO(dst_tmp[j] * m +
- dst2_tmp[j] * (64 - m), 6);
- }
- dst_tmp += dst_stride;
- dst2_tmp += dst2_stride;
- }
-
- for (; i < top_h; i ++) {
- memcpy(dst_tmp, dst2_tmp, top_w * sizeof(uint16_t));
- dst_tmp += dst_stride;
- dst2_tmp += dst2_stride;
- }
- } else {
+ if (is_hdb)
+ vpx_highbd_blend_a64_vmask(dst, dst_stride,
+ dst, dst_stride,
+ pre, pre_stride,
+ mask, h, top_w, xd->bd);
+ else
#endif // CONFIG_VP9_HIGHBITDEPTH
- uint8_t *dst_tmp = dst + h_offset * dst_stride;
- uint8_t *dst2_tmp = dst2 + h_offset * dst2_stride;
- generate_1dmask(h, mask + h_offset,
- plane && xd->plane[plane].subsampling_y);
+ vpx_blend_a64_vmask(dst, dst_stride,
+ dst, dst_stride,
+ pre, pre_stride,
+ mask, h, top_w);
- for (i = h_offset; i < h_offset + h; i++) {
- for (j = 0; j < top_w; j++) {
- const int m = mask[i]; assert(m >= 0 && m <= 64);
- if (m == 64)
- continue;
-
- if (m == 0)
- dst_tmp[j] = dst2_tmp[j];
- else
- dst_tmp[j] = ROUND_POWER_OF_TWO(dst_tmp[j] * m +
- dst2_tmp[j] * (64 - m), 6);
- }
- dst_tmp += dst_stride;
- dst2_tmp += dst2_stride;
- }
-
- for (; i < top_h; i ++) {
- memcpy(dst_tmp, dst2_tmp, top_w * sizeof(uint8_t));
- dst_tmp += dst_stride;
- dst2_tmp += dst2_stride;
- }
-#if CONFIG_VP9_HIGHBITDEPTH
- }
-#endif // CONFIG_VP9_HIGHBITDEPTH
- }
-
+ dst += h * dst_stride;
+ pre += h * pre_stride;
break;
- case PARTITION_VERT:
- {
-#if CONFIG_VP9_HIGHBITDEPTH
- if (b_hdb) {
- uint16_t *dst_tmp = dst16;
- uint16_t *dst2_tmp = dst216;
- generate_1dmask(w, mask + w_offset,
- plane && xd->plane[plane].subsampling_x);
-
- for (i = 0; i < top_h; i++) {
- for (j = w_offset; j < w_offset + w; j++) {
- const int m = mask[j]; assert(m >= 0 && m <= 64);
- if (m == 64)
- continue;
-
- if (m == 0)
- dst_tmp[j] = dst2_tmp[j];
- else
- dst_tmp[j] = ROUND_POWER_OF_TWO(dst_tmp[j] * m +
- dst2_tmp[j] * (64 - m), 6);
- }
- memcpy(dst_tmp + j, dst2_tmp + j,
- (top_w - w_offset - w) * sizeof(uint16_t));
- dst_tmp += dst_stride;
- dst2_tmp += dst2_stride;
- }
- } else {
-#endif // CONFIG_VP9_HIGHBITDEPTH
- uint8_t *dst_tmp = dst;
- uint8_t *dst2_tmp = dst2;
- generate_1dmask(w, mask + w_offset,
- plane && xd->plane[plane].subsampling_x);
-
- for (i = 0; i < top_h; i++) {
- for (j = w_offset; j < w_offset + w; j++) {
- const int m = mask[j]; assert(m >= 0 && m <= 64);
- if (m == 64)
- continue;
-
- if (m == 0)
- dst_tmp[j] = dst2_tmp[j];
- else
- dst_tmp[j] = ROUND_POWER_OF_TWO(dst_tmp[j] * m +
- dst2_tmp[j] * (64 - m), 6);
- }
- memcpy(dst_tmp + j, dst2_tmp + j,
- (top_w - w_offset - w) * sizeof(uint8_t));
- dst_tmp += dst_stride;
- dst2_tmp += dst2_stride;
- }
-#if CONFIG_VP9_HIGHBITDEPTH
- }
-#endif // CONFIG_VP9_HIGHBITDEPTH
}
+ case PARTITION_VERT: {
+ const uint8_t *const mask = get_supertx_mask(w, ssx);
+
+ w_remain = top_w - w_offset - w;
+ h_remain = top_h;
+ dst += w_offset;
+ pre += w_offset;
+
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (is_hdb)
+ vpx_highbd_blend_a64_hmask(dst, dst_stride,
+ dst, dst_stride,
+ pre, pre_stride,
+ mask, top_h, w, xd->bd);
+ else
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ vpx_blend_a64_hmask(dst, dst_stride,
+ dst, dst_stride,
+ pre, pre_stride,
+ mask, top_h, w);
+
+ dst += w;
+ pre += w;
break;
- default:
+ }
+ default: {
assert(0);
+ return;
+ }
}
- (void) xd;
+
+ if (w_remain == 0 || h_remain == 0) {
+ return;
+ }
+
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (is_hdb) {
+ dst = (uint8_t*)CONVERT_TO_SHORTPTR(dst);
+ pre = (const uint8_t*)CONVERT_TO_SHORTPTR(pre);
+ dst_stride *= 2;
+ pre_stride *= 2;
+ w_remain *= 2;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+ do {
+ memcpy(dst, pre, w_remain * sizeof(uint8_t));
+ dst += dst_stride;
+ pre += pre_stride;
+ } while (--h_remain);
}
void vp10_build_inter_predictors_sb_sub8x8_extend(
@@ -1878,12 +1827,10 @@
BLOCK_SIZE plane_bsize,
uint8_t *comppred,
int compstride,
- uint8_t *interpred,
+ const uint8_t *interpred,
int interstride,
- uint8_t *intrapred,
+ const uint8_t *intrapred,
int intrastride) {
- const int scale_bits = 8;
- const int scale_max = (1 << scale_bits);
const int bw = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
const int bh = 4 * num_4x4_blocks_high_lookup[plane_bsize];
const int size_scale = ii_size_scales[plane_bsize];
@@ -1896,11 +1843,11 @@
bsize);
const int subw = 2 * num_4x4_blocks_wide_lookup[bsize] == bw;
const int subh = 2 * num_4x4_blocks_high_lookup[bsize] == bh;
- vpx_blend_mask6(comppred, compstride,
- intrapred, intrastride,
- interpred, interstride,
- mask, 4 * num_4x4_blocks_wide_lookup[bsize],
- bh, bw, subh, subw);
+ vpx_blend_a64_mask(comppred, compstride,
+ intrapred, intrastride,
+ interpred, interstride,
+ mask, 4 * num_4x4_blocks_wide_lookup[bsize],
+ bh, bw, subh, subw);
}
return;
}
@@ -1911,10 +1858,9 @@
for (j = 0; j < bw; ++j) {
int scale = ii_weights1d[i * size_scale];
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -1924,10 +1870,9 @@
for (j = 0; j < bw; ++j) {
int scale = ii_weights1d[j * size_scale];
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -1939,10 +1884,9 @@
int scale = (ii_weights1d[i * size_scale] * 3 +
ii_weights1d[j * size_scale]) >> 2;
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -1954,10 +1898,9 @@
int scale = (ii_weights1d[j * size_scale] * 3 +
ii_weights1d[i * size_scale]) >> 2;
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -1967,10 +1910,9 @@
for (j = 0; j < bw; ++j) {
int scale = ii_weights1d[(i < j ? i : j) * size_scale];
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -1981,10 +1923,9 @@
int scale = (ii_weights1d[i * size_scale] +
ii_weights1d[j * size_scale]) >> 1;
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -1995,10 +1936,8 @@
for (i = 0; i < bh; ++i) {
for (j = 0; j < bw; ++j) {
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- interpred[i * interstride + j] +
- intrapred[i * intrastride + j],
- 1);
+ VPX_BLEND_AVG(intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -2014,20 +1953,18 @@
BLOCK_SIZE plane_bsize,
uint8_t *comppred8,
int compstride,
- uint8_t *interpred8,
+ const uint8_t *interpred8,
int interstride,
- uint8_t *intrapred8,
+ const uint8_t *intrapred8,
int intrastride, int bd) {
- const int scale_bits = 8;
- const int scale_max = (1 << scale_bits);
const int bw = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
const int bh = 4 * num_4x4_blocks_high_lookup[plane_bsize];
const int size_scale = ii_size_scales[plane_bsize];
int i, j;
uint16_t *comppred = CONVERT_TO_SHORTPTR(comppred8);
- uint16_t *interpred = CONVERT_TO_SHORTPTR(interpred8);
- uint16_t *intrapred = CONVERT_TO_SHORTPTR(intrapred8);
+ const uint16_t *interpred = CONVERT_TO_SHORTPTR(interpred8);
+ const uint16_t *intrapred = CONVERT_TO_SHORTPTR(intrapred8);
if (use_wedge_interintra) {
if (is_interintra_wedge_used(bsize)) {
@@ -2036,11 +1973,11 @@
bsize);
const int subh = 2 * num_4x4_blocks_high_lookup[bsize] == bh;
const int subw = 2 * num_4x4_blocks_wide_lookup[bsize] == bw;
- vpx_highbd_blend_mask6(comppred8, compstride,
- intrapred8, intrastride,
- interpred8, interstride,
- mask, bw,
- bh, bw, subh, subw, bd);
+ vpx_highbd_blend_a64_mask(comppred8, compstride,
+ intrapred8, intrastride,
+ interpred8, interstride,
+ mask, bw,
+ bh, bw, subh, subw, bd);
}
return;
}
@@ -2051,10 +1988,9 @@
for (j = 0; j < bw; ++j) {
int scale = ii_weights1d[i * size_scale];
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -2064,10 +2000,9 @@
for (j = 0; j < bw; ++j) {
int scale = ii_weights1d[j * size_scale];
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -2079,10 +2014,9 @@
int scale = (ii_weights1d[i * size_scale] * 3 +
ii_weights1d[j * size_scale]) >> 2;
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -2094,10 +2028,9 @@
int scale = (ii_weights1d[j * size_scale] * 3 +
ii_weights1d[i * size_scale]) >> 2;
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -2107,10 +2040,9 @@
for (j = 0; j < bw; ++j) {
int scale = ii_weights1d[(i < j ? i : j) * size_scale];
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -2121,10 +2053,9 @@
int scale = (ii_weights1d[i * size_scale] +
ii_weights1d[j * size_scale]) >> 1;
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- (scale_max - scale) * interpred[i * interstride + j] +
- scale * intrapred[i * intrastride + j],
- scale_bits);
+ VPX_BLEND_A256(scale,
+ intrapred[i * intrastride + j],
+ interpred[i * interstride + j]);
}
}
break;
@@ -2135,10 +2066,8 @@
for (i = 0; i < bh; ++i) {
for (j = 0; j < bw; ++j) {
comppred[i * compstride + j] =
- ROUND_POWER_OF_TWO(
- interpred[i * interstride + j] +
- intrapred[i * intrastride + j],
- 1);
+ VPX_BLEND_AVG(interpred[i * interstride + j],
+ intrapred[i * intrastride + j]);
}
}
break;
@@ -2239,8 +2168,8 @@
void vp10_combine_interintra(MACROBLOCKD *xd,
BLOCK_SIZE bsize, int plane,
- uint8_t *inter_pred, int inter_stride,
- uint8_t *intra_pred, int intra_stride) {
+ const uint8_t *inter_pred, int inter_stride,
+ const uint8_t *intra_pred, int intra_stride) {
const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, &xd->plane[plane]);
#if CONFIG_VP9_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
diff --git a/vp10/common/reconinter.h b/vp10/common/reconinter.h
index 4ede3e9..ac4a004 100644
--- a/vp10/common/reconinter.h
+++ b/vp10/common/reconinter.h
@@ -396,7 +396,8 @@
struct macroblockd_plane;
void vp10_build_masked_inter_predictor_complex(
MACROBLOCKD *xd,
- uint8_t *dst, int dst_stride, uint8_t *dst2, int dst2_stride,
+ uint8_t *dst, int dst_stride,
+ const uint8_t *pre, int pre_stride,
int mi_row, int mi_col, int mi_row_ori, int mi_col_ori,
BLOCK_SIZE bsize, BLOCK_SIZE top_bsize,
PARTITION_TYPE partition, int plane);
@@ -631,8 +632,8 @@
void vp10_combine_interintra(
MACROBLOCKD *xd,
BLOCK_SIZE bsize, int plane,
- uint8_t *inter_pred, int inter_stride,
- uint8_t *intra_pred, int intra_stride);
+ const uint8_t *inter_pred, int inter_stride,
+ const uint8_t *intra_pred, int intra_stride);
void vp10_build_interintra_predictors_sbuv(MACROBLOCKD *xd,
uint8_t *upred,
uint8_t *vpred,
diff --git a/vp10/common/vp10_convolve.c b/vp10/common/vp10_convolve.c
index 2026df1..5b1d921 100644
--- a/vp10/common/vp10_convolve.c
+++ b/vp10/common/vp10_convolve.c
@@ -2,6 +2,7 @@
#include <string.h>
#include "./vp10_rtcd.h"
+#include "vp10/common/vp10_convolve.h"
#include "vp10/common/filter.h"
#include "vpx_dsp/vpx_dsp_common.h"
#include "vpx_ports/mem.h"
@@ -182,7 +183,7 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-static void highbd_convolve_horiz(const uint16_t *src, int src_stride,
+void vp10_highbd_convolve_horiz_c(const uint16_t *src, int src_stride,
uint16_t *dst, int dst_stride, int w, int h,
const InterpFilterParams filter_params,
const int subpel_x_q4, int x_step_q4, int avg,
@@ -213,7 +214,7 @@
}
}
-static void highbd_convolve_vert(const uint16_t *src, int src_stride,
+void vp10_highbd_convolve_vert_c(const uint16_t *src, int src_stride,
uint16_t *dst, int dst_stride, int w, int h,
const InterpFilterParams filter_params,
const int subpel_y_q4, int y_step_q4, int avg,
@@ -300,8 +301,9 @@
InterpFilterParams filter_params =
vp10_get_interp_filter_params(interp_filter);
#endif
- highbd_convolve_horiz(src, src_stride, dst, dst_stride, w, h, filter_params,
- subpel_x_q4, x_step_q4, ref_idx, bd);
+ vp10_highbd_convolve_horiz(src, src_stride, dst, dst_stride, w, h,
+ filter_params, subpel_x_q4, x_step_q4, ref_idx,
+ bd);
} else if (ignore_horiz) {
#if CONFIG_DUAL_FILTER
InterpFilterParams filter_params =
@@ -310,8 +312,9 @@
InterpFilterParams filter_params =
vp10_get_interp_filter_params(interp_filter);
#endif
- highbd_convolve_vert(src, src_stride, dst, dst_stride, w, h, filter_params,
- subpel_y_q4, y_step_q4, ref_idx, bd);
+ vp10_highbd_convolve_vert(src, src_stride, dst, dst_stride, w, h,
+ filter_params, subpel_y_q4, y_step_q4, ref_idx,
+ bd);
} else {
// temp's size is set to (maximum possible intermediate_height) *
// MAX_BLOCK_WIDTH
@@ -336,9 +339,10 @@
int intermediate_height =
(((h - 1) * y_step_q4 + subpel_y_q4) >> SUBPEL_BITS) + filter_size;
- highbd_convolve_horiz(src - src_stride * (filter_size / 2 - 1), src_stride,
- temp, temp_stride, w, intermediate_height,
- filter_params, subpel_x_q4, x_step_q4, 0, bd);
+ vp10_highbd_convolve_horiz(src - src_stride * (filter_size / 2 - 1),
+ src_stride, temp, temp_stride, w,
+ intermediate_height, filter_params, subpel_x_q4,
+ x_step_q4, 0, bd);
#if CONFIG_DUAL_FILTER
filter_params = filter_params_y;
@@ -346,9 +350,9 @@
filter_size = filter_params.taps;
assert(filter_params.taps <= MAX_FILTER_TAP);
- highbd_convolve_vert(temp + temp_stride * (filter_size / 2 - 1),
- temp_stride, dst, dst_stride, w, h, filter_params,
- subpel_y_q4, y_step_q4, ref_idx, bd);
+ vp10_highbd_convolve_vert(temp + temp_stride * (filter_size / 2 - 1),
+ temp_stride, dst, dst_stride, w, h, filter_params,
+ subpel_y_q4, y_step_q4, ref_idx, bd);
}
}
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vp10/common/vp10_fwd_txfm.c b/vp10/common/vp10_fwd_txfm.c
index 3211cd0..17935c5 100644
--- a/vp10/common/vp10_fwd_txfm.c
+++ b/vp10/common/vp10_fwd_txfm.c
@@ -8,6 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
+#include "./vp10_rtcd.h"
#include "vp10/common/vp10_fwd_txfm.h"
void vp10_fdct4x4_c(const int16_t *input, tran_low_t *output, int stride) {
diff --git a/vp10/common/vp10_fwd_txfm2d.c b/vp10/common/vp10_fwd_txfm2d.c
index 91a5357..cddd7dc 100644
--- a/vp10/common/vp10_fwd_txfm2d.c
+++ b/vp10/common/vp10_fwd_txfm2d.c
@@ -10,6 +10,7 @@
#include <assert.h>
+#include "./vp10_rtcd.h"
#include "vp10/common/enums.h"
#include "vp10/common/vp10_fwd_txfm1d.h"
#include "vp10/common/vp10_fwd_txfm2d_cfg.h"
diff --git a/vp10/common/vp10_inv_txfm.c b/vp10/common/vp10_inv_txfm.c
index 403b209..0ebac42 100644
--- a/vp10/common/vp10_inv_txfm.c
+++ b/vp10/common/vp10_inv_txfm.c
@@ -12,6 +12,7 @@
#include <math.h>
#include <string.h>
+#include "./vp10_rtcd.h"
#include "vp10/common/vp10_inv_txfm.h"
void vp10_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
diff --git a/vp10/common/vp10_inv_txfm2d.c b/vp10/common/vp10_inv_txfm2d.c
index ccf4614..85a33ba 100644
--- a/vp10/common/vp10_inv_txfm2d.c
+++ b/vp10/common/vp10_inv_txfm2d.c
@@ -8,6 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
+#include "./vp10_rtcd.h"
#include "vp10/common/enums.h"
#include "vp10/common/vp10_txfm.h"
#include "vp10/common/vp10_inv_txfm1d.h"
diff --git a/vp10/common/vp10_rtcd_defs.pl b/vp10/common/vp10_rtcd_defs.pl
index 5a41511..1b501e2 100644
--- a/vp10/common/vp10_rtcd_defs.pl
+++ b/vp10/common/vp10_rtcd_defs.pl
@@ -93,6 +93,13 @@
add_proto qw/void vp10_convolve_vert/, "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int w, int h, const InterpFilterParams fp, const int subpel_x_q4, int x_step_q4, int avg";
specialize qw/vp10_convolve_vert ssse3/;
+if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/void vp10_highbd_convolve_horiz/, "const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams fp, const int subpel_x_q4, int x_step_q4, int avg, int bd";
+ specialize qw/vp10_highbd_convolve_horiz sse4_1/;
+ add_proto qw/void vp10_highbd_convolve_vert/, "const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams fp, const int subpel_x_q4, int x_step_q4, int avg, int bd";
+ specialize qw/vp10_highbd_convolve_vert sse4_1/;
+}
+
#
# dct
#
@@ -260,13 +267,19 @@
specialize qw/vp10_fdct32x32_1/;
} else {
add_proto qw/void vp10_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
- specialize qw/vp10_iht4x4_16_add sse2 neon dspr2 msa/;
+ specialize qw/vp10_iht4x4_16_add sse2 neon dspr2/;
add_proto qw/void vp10_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
- specialize qw/vp10_iht8x8_64_add sse2 neon dspr2 msa/;
+ specialize qw/vp10_iht8x8_64_add sse2 neon dspr2/;
add_proto qw/void vp10_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, int tx_type";
- specialize qw/vp10_iht16x16_256_add sse2 dspr2 msa/;
+ specialize qw/vp10_iht16x16_256_add sse2 dspr2/;
+
+ if (vpx_config("CONFIG_EXT_TX") ne "yes") {
+ specialize qw/vp10_iht4x4_16_add msa/;
+ specialize qw/vp10_iht8x8_64_add msa/;
+ specialize qw/vp10_iht16x16_256_add msa/;
+ }
add_proto qw/void vp10_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/vp10_fdct4x4 sse2/;
@@ -437,13 +450,19 @@
specialize qw/vp10_fwht4x4/, "$mmx_x86inc";
} else {
add_proto qw/void vp10_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/vp10_fht4x4 sse2 msa/;
+ specialize qw/vp10_fht4x4 sse2/;
add_proto qw/void vp10_fht8x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/vp10_fht8x8 sse2 msa/;
+ specialize qw/vp10_fht8x8 sse2/;
add_proto qw/void vp10_fht16x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/vp10_fht16x16 sse2 msa/;
+ specialize qw/vp10_fht16x16 sse2/;
+
+ if (vpx_config("CONFIG_EXT_TX") ne "yes") {
+ specialize qw/vp10_fht4x4 msa/;
+ specialize qw/vp10_fht8x8 msa/;
+ specialize qw/vp10_fht16x16 msa/;
+ }
add_proto qw/void vp10_fht32x32/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
specialize qw/vp10_fht32x32/;
diff --git a/vp10/common/warped_motion.c b/vp10/common/warped_motion.c
index 4990bb3..3b924ea 100644
--- a/vp10/common/warped_motion.c
+++ b/vp10/common/warped_motion.c
@@ -85,19 +85,19 @@
for (i = 0; i < n; ++i) {
const int x = *(points++), y = *(points++);
if (subsampling_x)
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
((x << (WARPEDMODEL_PREC_BITS + 1)) + mat[0]),
WARPEDPIXEL_PREC_BITS + 1);
else
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
((x << WARPEDMODEL_PREC_BITS)) + mat[0],
WARPEDPIXEL_PREC_BITS);
if (subsampling_y)
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
((y << (WARPEDMODEL_PREC_BITS + 1)) + mat[1]),
WARPEDPIXEL_PREC_BITS + 1);
else
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
((y << WARPEDMODEL_PREC_BITS)) + mat[1],
WARPEDPIXEL_PREC_BITS);
points += stride_points - 2;
@@ -115,21 +115,21 @@
for (i = 0; i < n; ++i) {
const int x = *(points++), y = *(points++);
if (subsampling_x)
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
mat[0] * 2 * x + mat[1] * 2 * y + mat[2] +
(mat[0] + mat[1] - (1 << WARPEDMODEL_PREC_BITS)) / 2,
WARPEDDIFF_PREC_BITS + 1);
else
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(mat[0] * x + mat[1] * y + mat[2],
- WARPEDDIFF_PREC_BITS);
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(mat[0] * x + mat[1] * y + mat[2],
+ WARPEDDIFF_PREC_BITS);
if (subsampling_y)
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
-mat[1] * 2 * x + mat[0] * 2 * y + mat[3] +
(-mat[1] + mat[0] - (1 << WARPEDMODEL_PREC_BITS)) / 2,
WARPEDDIFF_PREC_BITS + 1);
else
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(-mat[1] * x + mat[0] * y + mat[3],
- WARPEDDIFF_PREC_BITS);
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(-mat[1] * x + mat[0] * y + mat[3],
+ WARPEDDIFF_PREC_BITS);
points += stride_points - 2;
proj += stride_proj - 2;
}
@@ -145,21 +145,21 @@
for (i = 0; i < n; ++i) {
const int x = *(points++), y = *(points++);
if (subsampling_x)
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
mat[0] * 2 * x + mat[1] * 2 * y + mat[4] +
(mat[0] + mat[1] - (1 << WARPEDMODEL_PREC_BITS)) / 2,
WARPEDDIFF_PREC_BITS + 1);
else
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(mat[0] * x + mat[1] * y + mat[4],
- WARPEDDIFF_PREC_BITS);
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(mat[0] * x + mat[1] * y + mat[4],
+ WARPEDDIFF_PREC_BITS);
if (subsampling_y)
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(
mat[2] * 2 * x + mat[3] * 2 * y + mat[5] +
(mat[2] + mat[3] - (1 << WARPEDMODEL_PREC_BITS)) / 2,
WARPEDDIFF_PREC_BITS + 1);
else
- *(proj++) = ROUNDZ_POWER_OF_TWO_SIGNED(mat[2] * x + mat[3] * y + mat[5],
- WARPEDDIFF_PREC_BITS);
+ *(proj++) = ROUND_POWER_OF_TWO_SIGNED(mat[2] * x + mat[3] * y + mat[5],
+ WARPEDDIFF_PREC_BITS);
points += stride_points - 2;
proj += stride_proj - 2;
}
@@ -357,7 +357,7 @@
const int64_t v2 = x * x * (2 * p[-1] - 5 * p[0] + 4 * p[1] - p[2]);
const int64_t v3 = x * (p[1] - p[-1]);
const int64_t v4 = 2 * p[0];
- return (int32_t)ROUNDZ_POWER_OF_TWO_SIGNED(
+ return (int32_t)ROUND_POWER_OF_TWO_SIGNED(
(v4 << (3 * WARPEDPIXEL_PREC_BITS)) +
(v3 << (2 * WARPEDPIXEL_PREC_BITS)) +
(v2 << WARPEDPIXEL_PREC_BITS) + v1,
diff --git a/vp10/common/x86/idct_intrin_sse2.c b/vp10/common/x86/idct_intrin_sse2.c
index d8b6d95..da60764 100644
--- a/vp10/common/x86/idct_intrin_sse2.c
+++ b/vp10/common/x86/idct_intrin_sse2.c
@@ -8,6 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
+#include "./vp10_rtcd.h"
#include "vpx_dsp/x86/inv_txfm_sse2.h"
#include "vpx_dsp/x86/txfm_common_sse2.h"
#include "vpx_ports/mem.h"
diff --git a/vp10/common/x86/vp10_convolve_filters_ssse3.c b/vp10/common/x86/vp10_convolve_filters_ssse3.c
index 410da89..2f7b3c7 100644
--- a/vp10/common/x86/vp10_convolve_filters_ssse3.c
+++ b/vp10/common/x86/vp10_convolve_filters_ssse3.c
@@ -7,626 +7,936 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
-
#include "./vpx_config.h"
#include "vp10/common/filter.h"
-// Note:
-// Filter coefficients are from "filter.c". We use,
-// sub_pel_filters_temporalfilter_12[],
-// sub_pel_filters_12sharp[],
-// sub_pel_filters_10sharp[].
-
-// (2-1) Parallel filtering along the intended signal direction
-
-// 12-tap filter padding:
-// {filter_coefficients, 0, 0, 0, 0},
-// {0, 0, filter_coefficients, 0, 0),
-#if USE_TEMPORALFILTER_12TAP
+#if CONFIG_EXT_INTERP
DECLARE_ALIGNED(16, const int8_t,
- sub_pel_filters_temporalfilter_12_signal_dir[15][2][16]) = {
+ sub_pel_filters_10sharp_signal_dir[15][2][16]) = {
{
- {0, 1, -1, 3, -7, 127, 8, -4, 2, -1, 0, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -1, 3, -7, 127, 8, -4, 2, -1, 0, 0, 0, 0},
+ { 0, 0, -1, 3, -6, 127, 8, -4,
+ 2, -1, 0, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 0, -1, 3, -6, 127,
+ 8, -4, 2, -1, 0, 0, 0, 0, },
},
{
- {0, 1, -3, 5, -12, 124, 18, -8, 4, -2, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -3, 5, -12, 124, 18, -8, 4, -2, 1, 0, 0, 0},
+ { 0, 1, -2, 5, -12, 124, 18, -7,
+ 3, -2, 0, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -2, 5, -12, 124,
+ 18, -7, 3, -2, 0, 0, 0, 0, },
},
{
- {-1, 2, -4, 8, -17, 120, 28, -11, 6, -3, 1, -1, 0, 0, 0, 0},
- {0, 0, -1, 2, -4, 8, -17, 120, 28, -11, 6, -3, 1, -1, 0, 0},
+ { 0, 1, -3, 7, -17, 119, 28, -11,
+ 5, -2, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -3, 7, -17, 119,
+ 28, -11, 5, -2, 1, 0, 0, 0, },
},
{
- {-1, 2, -4, 10, -21, 114, 38, -15, 8, -4, 2, -1, 0, 0, 0, 0},
- {0, 0, -1, 2, -4, 10, -21, 114, 38, -15, 8, -4, 2, -1, 0, 0},
+ { 0, 1, -4, 8, -20, 114, 38, -14,
+ 7, -3, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -4, 8, -20, 114,
+ 38, -14, 7, -3, 1, 0, 0, 0, },
},
{
- {-1, 3, -5, 11, -23, 107, 49, -18, 9, -5, 2, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -5, 11, -23, 107, 49, -18, 9, -5, 2, -1, 0, 0},
+ { 0, 1, -4, 9, -22, 107, 49, -17,
+ 8, -4, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -4, 9, -22, 107,
+ 49, -17, 8, -4, 1, 0, 0, 0, },
},
{
- {-1, 3, -6, 12, -25, 99, 60, -21, 11, -6, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -6, 12, -25, 99, 60, -21, 11, -6, 3, -1, 0, 0},
+ { 0, 2, -5, 10, -24, 99, 59, -20,
+ 9, -4, 2, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 2, -5, 10, -24, 99,
+ 59, -20, 9, -4, 2, 0, 0, 0, },
},
{
- {-1, 3, -6, 12, -25, 90, 70, -23, 12, -6, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -6, 12, -25, 90, 70, -23, 12, -6, 3, -1, 0, 0},
+ { 0, 2, -5, 10, -24, 90, 70, -22,
+ 10, -5, 2, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 2, -5, 10, -24, 90,
+ 70, -22, 10, -5, 2, 0, 0, 0, },
},
{
- {-1, 3, -6, 12, -24, 80, 80, -24, 12, -6, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -6, 12, -24, 80, 80, -24, 12, -6, 3, -1, 0, 0},
+ { 0, 2, -5, 10, -23, 80, 80, -23,
+ 10, -5, 2, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 2, -5, 10, -23, 80,
+ 80, -23, 10, -5, 2, 0, 0, 0, },
},
{
- {-1, 3, -6, 12, -23, 70, 90, -25, 12, -6, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -6, 12, -23, 70, 90, -25, 12, -6, 3, -1, 0, 0},
+ { 0, 2, -5, 10, -22, 70, 90, -24,
+ 10, -5, 2, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 2, -5, 10, -22, 70,
+ 90, -24, 10, -5, 2, 0, 0, 0, },
},
{
- {-1, 3, -6, 11, -21, 60, 99, -25, 12, -6, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -6, 11, -21, 60, 99, -25, 12, -6, 3, -1, 0, 0},
+ { 0, 2, -4, 9, -20, 59, 99, -24,
+ 10, -5, 2, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 2, -4, 9, -20, 59,
+ 99, -24, 10, -5, 2, 0, 0, 0, },
},
{
- {-1, 2, -5, 9, -18, 49, 107, -23, 11, -5, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 2, -5, 9, -18, 49, 107, -23, 11, -5, 3, -1, 0, 0},
+ { 0, 1, -4, 8, -17, 49, 107, -22,
+ 9, -4, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -4, 8, -17, 49,
+ 107, -22, 9, -4, 1, 0, 0, 0, },
},
{
- {-1, 2, -4, 8, -15, 38, 114, -21, 10, -4, 2, -1, 0, 0, 0, 0},
- {0, 0, -1, 2, -4, 8, -15, 38, 114, -21, 10, -4, 2, -1, 0, 0},
+ { 0, 1, -3, 7, -14, 38, 114, -20,
+ 8, -4, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -3, 7, -14, 38,
+ 114, -20, 8, -4, 1, 0, 0, 0, },
},
{
- {-1, 1, -3, 6, -11, 28, 120, -17, 8, -4, 2, -1, 0, 0, 0, 0},
- {0, 0, -1, 1, -3, 6, -11, 28, 120, -17, 8, -4, 2, -1, 0, 0},
+ { 0, 1, -2, 5, -11, 28, 119, -17,
+ 7, -3, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -2, 5, -11, 28,
+ 119, -17, 7, -3, 1, 0, 0, 0, },
},
{
- {0, 1, -2, 4, -8, 18, 124, -12, 5, -3, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -2, 4, -8, 18, 124, -12, 5, -3, 1, 0, 0, 0},
+ { 0, 0, -2, 3, -7, 18, 124, -12,
+ 5, -2, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 0, -2, 3, -7, 18,
+ 124, -12, 5, -2, 1, 0, 0, 0, },
},
{
- {0, 0, -1, 2, -4, 8, 127, -7, 3, -1, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 0, -1, 2, -4, 8, 127, -7, 3, -1, 1, 0, 0, 0},
+ { 0, 0, -1, 2, -4, 8, 127, -6,
+ 3, -1, 0, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 0, -1, 2, -4, 8,
+ 127, -6, 3, -1, 0, 0, 0, 0, },
},
};
-#endif // USE_TEMPORALFILTER_12TAP
-
+#endif
+#if CONFIG_EXT_INTERP
+DECLARE_ALIGNED(16, const int8_t,
+ sub_pel_filters_10sharp_ver_signal_dir[15][6][16]) = {
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 127, -6, 127, -6, 127, -6, 127,
+ -6, 127, -6, 127, -6, 127, -6, 127, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 5, -2, 5, -2, 5, -2, 5,
+ -2, 5, -2, 5, -2, 5, -2, 5, },
+ {-12, 124, -12, 124, -12, 124, -12, 124,
+ -12, 124, -12, 124, -12, 124, -12, 124, },
+ { 18, -7, 18, -7, 18, -7, 18, -7,
+ 18, -7, 18, -7, 18, -7, 18, -7, },
+ { 3, -2, 3, -2, 3, -2, 3, -2,
+ 3, -2, 3, -2, 3, -2, 3, -2, },
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -3, 7, -3, 7, -3, 7, -3, 7,
+ -3, 7, -3, 7, -3, 7, -3, 7, },
+ {-17, 119, -17, 119, -17, 119, -17, 119,
+ -17, 119, -17, 119, -17, 119, -17, 119, },
+ { 28, -11, 28, -11, 28, -11, 28, -11,
+ 28, -11, 28, -11, 28, -11, 28, -11, },
+ { 5, -2, 5, -2, 5, -2, 5, -2,
+ 5, -2, 5, -2, 5, -2, 5, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-20, 114, -20, 114, -20, 114, -20, 114,
+ -20, 114, -20, 114, -20, 114, -20, 114, },
+ { 38, -14, 38, -14, 38, -14, 38, -14,
+ 38, -14, 38, -14, 38, -14, 38, -14, },
+ { 7, -3, 7, -3, 7, -3, 7, -3,
+ 7, -3, 7, -3, 7, -3, 7, -3, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -4, 9, -4, 9, -4, 9, -4, 9,
+ -4, 9, -4, 9, -4, 9, -4, 9, },
+ {-22, 107, -22, 107, -22, 107, -22, 107,
+ -22, 107, -22, 107, -22, 107, -22, 107, },
+ { 49, -17, 49, -17, 49, -17, 49, -17,
+ 49, -17, 49, -17, 49, -17, 49, -17, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2,
+ 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10,
+ -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-24, 99, -24, 99, -24, 99, -24, 99,
+ -24, 99, -24, 99, -24, 99, -24, 99, },
+ { 59, -20, 59, -20, 59, -20, 59, -20,
+ 59, -20, 59, -20, 59, -20, 59, -20, },
+ { 9, -4, 9, -4, 9, -4, 9, -4,
+ 9, -4, 9, -4, 9, -4, 9, -4, },
+ { 2, 0, 2, 0, 2, 0, 2, 0,
+ 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2,
+ 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10,
+ -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-24, 90, -24, 90, -24, 90, -24, 90,
+ -24, 90, -24, 90, -24, 90, -24, 90, },
+ { 70, -22, 70, -22, 70, -22, 70, -22,
+ 70, -22, 70, -22, 70, -22, 70, -22, },
+ { 10, -5, 10, -5, 10, -5, 10, -5,
+ 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0,
+ 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2,
+ 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10,
+ -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-23, 80, -23, 80, -23, 80, -23, 80,
+ -23, 80, -23, 80, -23, 80, -23, 80, },
+ { 80, -23, 80, -23, 80, -23, 80, -23,
+ 80, -23, 80, -23, 80, -23, 80, -23, },
+ { 10, -5, 10, -5, 10, -5, 10, -5,
+ 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0,
+ 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2,
+ 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10,
+ -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-22, 70, -22, 70, -22, 70, -22, 70,
+ -22, 70, -22, 70, -22, 70, -22, 70, },
+ { 90, -24, 90, -24, 90, -24, 90, -24,
+ 90, -24, 90, -24, 90, -24, 90, -24, },
+ { 10, -5, 10, -5, 10, -5, 10, -5,
+ 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0,
+ 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2,
+ 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -4, 9, -4, 9, -4, 9, -4, 9,
+ -4, 9, -4, 9, -4, 9, -4, 9, },
+ {-20, 59, -20, 59, -20, 59, -20, 59,
+ -20, 59, -20, 59, -20, 59, -20, 59, },
+ { 99, -24, 99, -24, 99, -24, 99, -24,
+ 99, -24, 99, -24, 99, -24, 99, -24, },
+ { 10, -5, 10, -5, 10, -5, 10, -5,
+ 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0,
+ 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-17, 49, -17, 49, -17, 49, -17, 49,
+ -17, 49, -17, 49, -17, 49, -17, 49, },
+ {107, -22, 107, -22, 107, -22, 107, -22,
+ 107, -22, 107, -22, 107, -22, 107, -22, },
+ { 9, -4, 9, -4, 9, -4, 9, -4,
+ 9, -4, 9, -4, 9, -4, 9, -4, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -3, 7, -3, 7, -3, 7, -3, 7,
+ -3, 7, -3, 7, -3, 7, -3, 7, },
+ {-14, 38, -14, 38, -14, 38, -14, 38,
+ -14, 38, -14, 38, -14, 38, -14, 38, },
+ {114, -20, 114, -20, 114, -20, 114, -20,
+ 114, -20, 114, -20, 114, -20, 114, -20, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 5, -2, 5, -2, 5, -2, 5,
+ -2, 5, -2, 5, -2, 5, -2, 5, },
+ {-11, 28, -11, 28, -11, 28, -11, 28,
+ -11, 28, -11, 28, -11, 28, -11, 28, },
+ {119, -17, 119, -17, 119, -17, 119, -17,
+ 119, -17, 119, -17, 119, -17, 119, -17, },
+ { 7, -3, 7, -3, 7, -3, 7, -3,
+ 7, -3, 7, -3, 7, -3, 7, -3, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -2, 3, -2, 3, -2, 3, -2, 3,
+ -2, 3, -2, 3, -2, 3, -2, 3, },
+ { -7, 18, -7, 18, -7, 18, -7, 18,
+ -7, 18, -7, 18, -7, 18, -7, 18, },
+ {124, -12, 124, -12, 124, -12, 124, -12,
+ 124, -12, 124, -12, 124, -12, 124, -12, },
+ { 5, -2, 5, -2, 5, -2, 5, -2,
+ 5, -2, 5, -2, 5, -2, 5, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {127, -6, 127, -6, 127, -6, 127, -6,
+ 127, -6, 127, -6, 127, -6, 127, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+};
+#endif
#if CONFIG_EXT_INTERP
DECLARE_ALIGNED(16, const int8_t,
sub_pel_filters_12sharp_signal_dir[15][2][16]) = {
{
- {0, 1, -2, 3, -7, 127, 8, -4, 2, -1, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -2, 3, -7, 127, 8, -4, 2, -1, 1, 0, 0, 0},
+ { 0, 1, -2, 3, -7, 127, 8, -4,
+ 2, -1, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -2, 3, -7, 127,
+ 8, -4, 2, -1, 1, 0, 0, 0, },
},
{
- {-1, 2, -3, 6, -13, 124, 18, -8, 4, -2, 2, -1, 0, 0, 0, 0},
- {0, 0, -1, 2, -3, 6, -13, 124, 18, -8, 4, -2, 2, -1, 0, 0},
+ { -1, 2, -3, 6, -13, 124, 18, -8,
+ 4, -2, 2, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 2, -3, 6, -13, 124,
+ 18, -8, 4, -2, 2, -1, 0, 0, },
},
{
- {-1, 3, -4, 8, -18, 120, 28, -12, 7, -4, 2, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -4, 8, -18, 120, 28, -12, 7, -4, 2, -1, 0, 0},
+ { -1, 3, -4, 8, -18, 120, 28, -12,
+ 7, -4, 2, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -4, 8, -18, 120,
+ 28, -12, 7, -4, 2, -1, 0, 0, },
},
{
- {-1, 3, -6, 10, -21, 115, 38, -15, 8, -5, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -6, 10, -21, 115, 38, -15, 8, -5, 3, -1, 0, 0},
+ { -1, 3, -6, 10, -21, 115, 38, -15,
+ 8, -5, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -6, 10, -21, 115,
+ 38, -15, 8, -5, 3, -1, 0, 0, },
},
{
- {-2, 4, -6, 12, -24, 108, 49, -18, 10, -6, 3, -2, 0, 0, 0, 0},
- {0, 0, -2, 4, -6, 12, -24, 108, 49, -18, 10, -6, 3, -2, 0, 0},
+ { -2, 4, -6, 12, -24, 108, 49, -18,
+ 10, -6, 3, -2, 0, 0, 0, 0, },
+ { 0, 0, -2, 4, -6, 12, -24, 108,
+ 49, -18, 10, -6, 3, -2, 0, 0, },
},
{
- {-2, 4, -7, 13, -25, 100, 60, -21, 11, -7, 4, -2, 0, 0, 0, 0},
- {0, 0, -2, 4, -7, 13, -25, 100, 60, -21, 11, -7, 4, -2, 0, 0},
+ { -2, 4, -7, 13, -25, 100, 60, -21,
+ 11, -7, 4, -2, 0, 0, 0, 0, },
+ { 0, 0, -2, 4, -7, 13, -25, 100,
+ 60, -21, 11, -7, 4, -2, 0, 0, },
},
{
- {-2, 4, -7, 13, -26, 91, 71, -24, 13, -7, 4, -2, 0, 0, 0, 0},
- {0, 0, -2, 4, -7, 13, -26, 91, 71, -24, 13, -7, 4, -2, 0, 0},
+ { -2, 4, -7, 13, -26, 91, 71, -24,
+ 13, -7, 4, -2, 0, 0, 0, 0, },
+ { 0, 0, -2, 4, -7, 13, -26, 91,
+ 71, -24, 13, -7, 4, -2, 0, 0, },
},
{
- {-2, 4, -7, 13, -25, 81, 81, -25, 13, -7, 4, -2, 0, 0, 0, 0},
- {0, 0, -2, 4, -7, 13, -25, 81, 81, -25, 13, -7, 4, -2, 0, 0},
+ { -2, 4, -7, 13, -25, 81, 81, -25,
+ 13, -7, 4, -2, 0, 0, 0, 0, },
+ { 0, 0, -2, 4, -7, 13, -25, 81,
+ 81, -25, 13, -7, 4, -2, 0, 0, },
},
{
- {-2, 4, -7, 13, -24, 71, 91, -26, 13, -7, 4, -2, 0, 0, 0, 0},
- {0, 0, -2, 4, -7, 13, -24, 71, 91, -26, 13, -7, 4, -2, 0, 0},
+ { -2, 4, -7, 13, -24, 71, 91, -26,
+ 13, -7, 4, -2, 0, 0, 0, 0, },
+ { 0, 0, -2, 4, -7, 13, -24, 71,
+ 91, -26, 13, -7, 4, -2, 0, 0, },
},
{
- {-2, 4, -7, 11, -21, 60, 100, -25, 13, -7, 4, -2, 0, 0, 0, 0},
- {0, 0, -2, 4, -7, 11, -21, 60, 100, -25, 13, -7, 4, -2, 0, 0},
+ { -2, 4, -7, 11, -21, 60, 100, -25,
+ 13, -7, 4, -2, 0, 0, 0, 0, },
+ { 0, 0, -2, 4, -7, 11, -21, 60,
+ 100, -25, 13, -7, 4, -2, 0, 0, },
},
{
- {-2, 3, -6, 10, -18, 49, 108, -24, 12, -6, 4, -2, 0, 0, 0, 0},
- {0, 0, -2, 3, -6, 10, -18, 49, 108, -24, 12, -6, 4, -2, 0, 0},
+ { -2, 3, -6, 10, -18, 49, 108, -24,
+ 12, -6, 4, -2, 0, 0, 0, 0, },
+ { 0, 0, -2, 3, -6, 10, -18, 49,
+ 108, -24, 12, -6, 4, -2, 0, 0, },
},
{
- {-1, 3, -5, 8, -15, 38, 115, -21, 10, -6, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 3, -5, 8, -15, 38, 115, -21, 10, -6, 3, -1, 0, 0},
+ { -1, 3, -5, 8, -15, 38, 115, -21,
+ 10, -6, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -5, 8, -15, 38,
+ 115, -21, 10, -6, 3, -1, 0, 0, },
},
{
- {-1, 2, -4, 7, -12, 28, 120, -18, 8, -4, 3, -1, 0, 0, 0, 0},
- {0, 0, -1, 2, -4, 7, -12, 28, 120, -18, 8, -4, 3, -1, 0, 0},
+ { -1, 2, -4, 7, -12, 28, 120, -18,
+ 8, -4, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 2, -4, 7, -12, 28,
+ 120, -18, 8, -4, 3, -1, 0, 0, },
},
{
- {-1, 2, -2, 4, -8, 18, 124, -13, 6, -3, 2, -1, 0, 0, 0, 0},
- {0, 0, -1, 2, -2, 4, -8, 18, 124, -13, 6, -3, 2, -1, 0, 0},
+ { -1, 2, -2, 4, -8, 18, 124, -13,
+ 6, -3, 2, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 2, -2, 4, -8, 18,
+ 124, -13, 6, -3, 2, -1, 0, 0, },
},
{
- {0, 1, -1, 2, -4, 8, 127, -7, 3, -2, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -1, 2, -4, 8, 127, -7, 3, -2, 1, 0, 0, 0},
+ { 0, 1, -1, 2, -4, 8, 127, -7,
+ 3, -2, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -1, 2, -4, 8,
+ 127, -7, 3, -2, 1, 0, 0, 0, },
},
};
-
-// 10-tap filter padding:
-// {0, filter_coefficients, 0, 0, 0, 0, 0},
-// {0, 0, 0, filter_coefficients, 0, 0, 0),
-DECLARE_ALIGNED(16, const int8_t,
- sub_pel_filters_10sharp_signal_dir[15][2][16]) = {
- {
- {0, 0, -1, 3, -6, 127, 8, -4, 2, -1, 0, 0, 0, 0, 0, 0},
- {0, 0, 0, 0, -1, 3, -6, 127, 8, -4, 2, -1, 0, 0, 0, 0},
- },
- {
- {0, 1, -2, 5, -12, 124, 18, -7, 3, -2, 0, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -2, 5, -12, 124, 18, -7, 3, -2, 0, 0, 0, 0},
- },
- {
- {0, 1, -3, 7, -17, 119, 28, -11, 5, -2, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -3, 7, -17, 119, 28, -11, 5, -2, 1, 0, 0, 0},
- },
- {
- {0, 1, -4, 8, -20, 114, 38, -14, 7, -3, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -4, 8, -20, 114, 38, -14, 7, -3, 1, 0, 0, 0},
- },
- {
- {0, 1, -4, 9, -22, 107, 49, -17, 8, -4, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -4, 9, -22, 107, 49, -17, 8, -4, 1, 0, 0, 0},
- },
- {
- {0, 2, -5, 10, -24, 99, 59, -20, 9, -4, 2, 0, 0, 0, 0, 0},
- {0, 0, 0, 2, -5, 10, -24, 99, 59, -20, 9, -4, 2, 0, 0, 0},
- },
- {
- {0, 2, -5, 10, -24, 90, 70, -22, 10, -5, 2, 0, 0, 0, 0, 0},
- {0, 0, 0, 2, -5, 10, -24, 90, 70, -22, 10, -5, 2, 0, 0, 0},
- },
- {
- {0, 2, -5, 10, -23, 80, 80, -23, 10, -5, 2, 0, 0, 0, 0, 0},
- {0, 0, 0, 2, -5, 10, -23, 80, 80, -23, 10, -5, 2, 0, 0, 0},
- },
- {
- {0, 2, -5, 10, -22, 70, 90, -24, 10, -5, 2, 0, 0, 0, 0, 0},
- {0, 0, 0, 2, -5, 10, -22, 70, 90, -24, 10, -5, 2, 0, 0, 0},
- },
- {
- {0, 2, -4, 9, -20, 59, 99, -24, 10, -5, 2, 0, 0, 0, 0, 0},
- {0, 0, 0, 2, -4, 9, -20, 59, 99, -24, 10, -5, 2, 0, 0, 0},
- },
- {
- {0, 1, -4, 8, -17, 49, 107, -22, 9, -4, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -4, 8, -17, 49, 107, -22, 9, -4, 1, 0, 0, 0},
- },
- {
- {0, 1, -3, 7, -14, 38, 114, -20, 8, -4, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -3, 7, -14, 38, 114, -20, 8, -4, 1, 0, 0, 0},
- },
- {
- {0, 1, -2, 5, -11, 28, 119, -17, 7, -3, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 1, -2, 5, -11, 28, 119, -17, 7, -3, 1, 0, 0, 0},
- },
- {
- {0, 0, -2, 3, -7, 18, 124, -12, 5, -2, 1, 0, 0, 0, 0, 0},
- {0, 0, 0, 0, -2, 3, -7, 18, 124, -12, 5, -2, 1, 0, 0, 0},
- },
- {
- {0, 0, -1, 2, -4, 8, 127, -6, 3, -1, 0, 0, 0, 0, 0, 0},
- {0, 0, 0, 0, -1, 2, -4, 8, 127, -6, 3, -1, 0, 0, 0, 0},
- },
-};
-#endif // CONFIG_EXT_INTERP
-
-// (2-2) Parallel filtering vertically to signal direction
-#if USE_TEMPORALFILTER_12TAP
-DECLARE_ALIGNED(16, const int8_t,
- sub_pel_filters_temporalfilter_12_ver_signal_dir[15][6][16]) = {
- {
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-7, 127, -7, 127, -7, 127, -7, 127, -7, 127, -7, 127,
- -7, 127, -7, 127},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
- },
- {
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-3, 5, -3, 5, -3, 5, -3, 5, -3, 5, -3, 5, -3, 5, -3, 5},
- {-12, 124, -12, 124, -12, 124, -12, 124, -12, 124, -12, 124,
- -12, 124, -12, 124},
- {18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
- },
- {
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {-17, 120, -17, 120, -17, 120, -17, 120, -17, 120, -17, 120,
- -17, 120, -17, 120},
- {28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11},
- {6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3},
- {1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1},
- },
- {
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-4, 10, -4, 10, -4, 10, -4, 10, -4, 10, -4, 10, -4, 10, -4, 10},
- {-21, 114, -21, 114, -21, 114, -21, 114, -21, 114, -21, 114,
- -21, 114, -21, 114},
- {38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15,
- 38, -15, 38, -15},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
- },
- {
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-5, 11, -5, 11, -5, 11, -5, 11, -5, 11, -5, 11, -5, 11, -5, 11},
- {-23, 107, -23, 107, -23, 107, -23, 107, -23, 107, -23, 107,
- -23, 107, -23, 107},
- {49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18},
- {9, -5, 9, -5, 9, -5, 9, -5, 9, -5, 9, -5, 9, -5, 9, -5},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
- },
- {
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12},
- {-25, 99, -25, 99, -25, 99, -25, 99, -25, 99, -25, 99, -25, 99, -25, 99},
- {60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21},
- {11, -6, 11, -6, 11, -6, 11, -6, 11, -6, 11, -6, 11, -6, 11, -6},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- },
- {
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12},
- {-25, 90, -25, 90, -25, 90, -25, 90, -25, 90, -25, 90, -25, 90, -25, 90},
- {70, -23, 70, -23, 70, -23, 70, -23, 70, -23, 70, -23, 70, -23, 70, -23},
- {12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- },
- {
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12},
- {-24, 80, -24, 80, -24, 80, -24, 80, -24, 80, -24, 80, -24, 80, -24, 80},
- {80, -24, 80, -24, 80, -24, 80, -24, 80, -24, 80, -24, 80, -24, 80, -24},
- {12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- },
- {
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12},
- {-23, 70, -23, 70, -23, 70, -23, 70, -23, 70, -23, 70, -23, 70, -23, 70},
- {90, -25, 90, -25, 90, -25, 90, -25, 90, -25, 90, -25, 90, -25, 90, -25},
- {12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- },
- {
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-6, 11, -6, 11, -6, 11, -6, 11, -6, 11, -6, 11, -6, 11, -6, 11},
- {-21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60},
- {99, -25, 99, -25, 99, -25, 99, -25, 99, -25, 99, -25, 99, -25, 99, -25},
- {12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- },
- {
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-5, 9, -5, 9, -5, 9, -5, 9, -5, 9, -5, 9, -5, 9, -5, 9},
- {-18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49},
- {107, -23, 107, -23, 107, -23, 107, -23, 107, -23, 107, -23, 107,
- -23, 107, -23},
- {11, -5, 11, -5, 11, -5, 11, -5, 11, -5, 11, -5, 11, -5, 11, -5},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- },
- {
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {-15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38},
- {114, -21, 114, -21, 114, -21, 114, -21, 114, -21, 114, -21,
- 114, -21, 114, -21},
- {10, -4, 10, -4, 10, -4, 10, -4, 10, -4, 10, -4, 10, -4, 10, -4},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
- },
- {
- {-1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 1},
- {-3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6},
- {-11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28},
- {120, -17, 120, -17, 120, -17, 120, -17, 120, -17, 120, -17,
- 120, -17, 120, -17},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
- },
- {
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18},
- {124, -12, 124, -12, 124, -12, 124, -12, 124, -12, 124, -12,
- 124, -12, 124, -12},
- {5, -3, 5, -3, 5, -3, 5, -3, 5, -3, 5, -3, 5, -3, 5, -3},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
- },
- {
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {127, -7, 127, -7, 127, -7, 127, -7, 127, -7, 127, -7,
- 127, -7, 127, -7},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
- },
-};
-#endif // USE_TEMPORALFILTER_12TAP
-
+#endif
#if CONFIG_EXT_INTERP
DECLARE_ALIGNED(16, const int8_t,
sub_pel_filters_12sharp_ver_signal_dir[15][6][16]) = {
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3},
- {-7, 127, -7, 127, -7, 127, -7, 127, -7, 127, -7, 127,
- -7, 127, -7, 127},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 3, -2, 3, -2, 3, -2, 3,
+ -2, 3, -2, 3, -2, 3, -2, 3, },
+ { -7, 127, -7, 127, -7, 127, -7, 127,
+ -7, 127, -7, 127, -7, 127, -7, 127, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
},
{
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6},
- {-13, 124, -13, 124, -13, 124, -13, 124, -13, 124, -13, 124,
- -13, 124, -13, 124},
- {18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -3, 6, -3, 6, -3, 6, -3, 6,
+ -3, 6, -3, 6, -3, 6, -3, 6, },
+ {-13, 124, -13, 124, -13, 124, -13, 124,
+ -13, 124, -13, 124, -13, 124, -13, 124, },
+ { 18, -8, 18, -8, 18, -8, 18, -8,
+ 18, -8, 18, -8, 18, -8, 18, -8, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
},
{
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {-18, 120, -18, 120, -18, 120, -18, 120, -18, 120, -18, 120,
- -18, 120, -18, 120},
- {28, -12, 28, -12, 28, -12, 28, -12, 28, -12, 28, -12, 28, -12, 28, -12},
- {7, -4, 7, -4, 7, -4, 7, -4, 7, -4, 7, -4, 7, -4, 7, -4},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-18, 120, -18, 120, -18, 120, -18, 120,
+ -18, 120, -18, 120, -18, 120, -18, 120, },
+ { 28, -12, 28, -12, 28, -12, 28, -12,
+ 28, -12, 28, -12, 28, -12, 28, -12, },
+ { 7, -4, 7, -4, 7, -4, 7, -4,
+ 7, -4, 7, -4, 7, -4, 7, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
},
{
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10},
- {-21, 115, -21, 115, -21, 115, -21, 115, -21, 115, -21, 115,
- -21, 115, -21, 115},
- {38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15,
- 38, -15, 38, -15},
- {8, -5, 8, -5, 8, -5, 8, -5, 8, -5, 8, -5, 8, -5, 8, -5},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 10, -6, 10, -6, 10, -6, 10,
+ -6, 10, -6, 10, -6, 10, -6, 10, },
+ {-21, 115, -21, 115, -21, 115, -21, 115,
+ -21, 115, -21, 115, -21, 115, -21, 115, },
+ { 38, -15, 38, -15, 38, -15, 38, -15,
+ 38, -15, 38, -15, 38, -15, 38, -15, },
+ { 8, -5, 8, -5, 8, -5, 8, -5,
+ 8, -5, 8, -5, 8, -5, 8, -5, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
},
{
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12},
- {-24, 108, -24, 108, -24, 108, -24, 108, -24, 108, -24, 108,
- -24, 108, -24, 108},
- {49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18},
- {10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6},
- {3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2},
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -6, 12, -6, 12, -6, 12, -6, 12,
+ -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-24, 108, -24, 108, -24, 108, -24, 108,
+ -24, 108, -24, 108, -24, 108, -24, 108, },
+ { 49, -18, 49, -18, 49, -18, 49, -18,
+ 49, -18, 49, -18, 49, -18, 49, -18, },
+ { 10, -6, 10, -6, 10, -6, 10, -6,
+ 10, -6, 10, -6, 10, -6, 10, -6, },
+ { 3, -2, 3, -2, 3, -2, 3, -2,
+ 3, -2, 3, -2, 3, -2, 3, -2, },
},
{
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13},
- {-25, 100, -25, 100, -25, 100, -25, 100, -25, 100, -25, 100,
- -25, 100, -25, 100},
- {60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21},
- {11, -7, 11, -7, 11, -7, 11, -7, 11, -7, 11, -7, 11, -7, 11, -7},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13,
+ -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-25, 100, -25, 100, -25, 100, -25, 100,
+ -25, 100, -25, 100, -25, 100, -25, 100, },
+ { 60, -21, 60, -21, 60, -21, 60, -21,
+ 60, -21, 60, -21, 60, -21, 60, -21, },
+ { 11, -7, 11, -7, 11, -7, 11, -7,
+ 11, -7, 11, -7, 11, -7, 11, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
},
{
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13},
- {-26, 91, -26, 91, -26, 91, -26, 91, -26, 91, -26, 91, -26, 91, -26, 91},
- {71, -24, 71, -24, 71, -24, 71, -24, 71, -24, 71, -24, 71, -24, 71, -24},
- {13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13,
+ -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-26, 91, -26, 91, -26, 91, -26, 91,
+ -26, 91, -26, 91, -26, 91, -26, 91, },
+ { 71, -24, 71, -24, 71, -24, 71, -24,
+ 71, -24, 71, -24, 71, -24, 71, -24, },
+ { 13, -7, 13, -7, 13, -7, 13, -7,
+ 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
},
{
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13},
- {-25, 81, -25, 81, -25, 81, -25, 81, -25, 81, -25, 81, -25, 81, -25, 81},
- {81, -25, 81, -25, 81, -25, 81, -25, 81, -25, 81, -25, 81, -25, 81, -25},
- {13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13,
+ -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-25, 81, -25, 81, -25, 81, -25, 81,
+ -25, 81, -25, 81, -25, 81, -25, 81, },
+ { 81, -25, 81, -25, 81, -25, 81, -25,
+ 81, -25, 81, -25, 81, -25, 81, -25, },
+ { 13, -7, 13, -7, 13, -7, 13, -7,
+ 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
},
{
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13},
- {-24, 71, -24, 71, -24, 71, -24, 71, -24, 71, -24, 71, -24, 71, -24, 71},
- {91, -26, 91, -26, 91, -26, 91, -26, 91, -26, 91, -26, 91, -26, 91, -26},
- {13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13,
+ -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-24, 71, -24, 71, -24, 71, -24, 71,
+ -24, 71, -24, 71, -24, 71, -24, 71, },
+ { 91, -26, 91, -26, 91, -26, 91, -26,
+ 91, -26, 91, -26, 91, -26, 91, -26, },
+ { 13, -7, 13, -7, 13, -7, 13, -7,
+ 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
},
{
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-7, 11, -7, 11, -7, 11, -7, 11, -7, 11, -7, 11, -7, 11, -7, 11},
- {-21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60, -21, 60},
- {100, -25, 100, -25, 100, -25, 100, -25, 100, -25, 100, -25,
- 100, -25, 100, -25},
- {13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7, 13, -7},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 11, -7, 11, -7, 11, -7, 11,
+ -7, 11, -7, 11, -7, 11, -7, 11, },
+ {-21, 60, -21, 60, -21, 60, -21, 60,
+ -21, 60, -21, 60, -21, 60, -21, 60, },
+ {100, -25, 100, -25, 100, -25, 100, -25,
+ 100, -25, 100, -25, 100, -25, 100, -25, },
+ { 13, -7, 13, -7, 13, -7, 13, -7,
+ 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
},
{
- {-2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3},
- {-6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10},
- {-18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49, -18, 49},
- {108, -24, 108, -24, 108, -24, 108, -24, 108, -24, 108, -24,
- 108, -24, 108, -24},
- {12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6, 12, -6},
- {4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2},
+ { -2, 3, -2, 3, -2, 3, -2, 3,
+ -2, 3, -2, 3, -2, 3, -2, 3, },
+ { -6, 10, -6, 10, -6, 10, -6, 10,
+ -6, 10, -6, 10, -6, 10, -6, 10, },
+ {-18, 49, -18, 49, -18, 49, -18, 49,
+ -18, 49, -18, 49, -18, 49, -18, 49, },
+ {108, -24, 108, -24, 108, -24, 108, -24,
+ 108, -24, 108, -24, 108, -24, 108, -24, },
+ { 12, -6, 12, -6, 12, -6, 12, -6,
+ 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
},
{
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-5, 8, -5, 8, -5, 8, -5, 8, -5, 8, -5, 8, -5, 8, -5, 8},
- {-15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38, -15, 38},
- {115, -21, 115, -21, 115, -21, 115, -21, 115, -21, 115, -21,
- 115, -21, 115, -21},
- {10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6, 10, -6},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -5, 8, -5, 8, -5, 8, -5, 8,
+ -5, 8, -5, 8, -5, 8, -5, 8, },
+ {-15, 38, -15, 38, -15, 38, -15, 38,
+ -15, 38, -15, 38, -15, 38, -15, 38, },
+ {115, -21, 115, -21, 115, -21, 115, -21,
+ 115, -21, 115, -21, 115, -21, 115, -21, },
+ { 10, -6, 10, -6, 10, -6, 10, -6,
+ 10, -6, 10, -6, 10, -6, 10, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
},
{
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-4, 7, -4, 7, -4, 7, -4, 7, -4, 7, -4, 7, -4, 7, -4, 7},
- {-12, 28, -12, 28, -12, 28, -12, 28, -12, 28, -12, 28, -12, 28, -12, 28},
- {120, -18, 120, -18, 120, -18, 120, -18, 120, -18, 120, -18,
- 120, -18, 120, -18},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 7, -4, 7, -4, 7, -4, 7,
+ -4, 7, -4, 7, -4, 7, -4, 7, },
+ {-12, 28, -12, 28, -12, 28, -12, 28,
+ -12, 28, -12, 28, -12, 28, -12, 28, },
+ {120, -18, 120, -18, 120, -18, 120, -18,
+ 120, -18, 120, -18, 120, -18, 120, -18, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
},
{
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4, -2, 4},
- {-8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18, -8, 18},
- {124, -13, 124, -13, 124, -13, 124, -13, 124, -13, 124, -13,
- 124, -13, 124, -13},
- {6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3, 6, -3},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -8, 18, -8, 18, -8, 18, -8, 18,
+ -8, 18, -8, 18, -8, 18, -8, 18, },
+ {124, -13, 124, -13, 124, -13, 124, -13,
+ 124, -13, 124, -13, 124, -13, 124, -13, },
+ { 6, -3, 6, -3, 6, -3, 6, -3,
+ 6, -3, 6, -3, 6, -3, 6, -3, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {127, -7, 127, -7, 127, -7, 127, -7, 127, -7, 127, -7,
- 127, -7, 127, -7},
- {3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {127, -7, 127, -7, 127, -7, 127, -7,
+ 127, -7, 127, -7, 127, -7, 127, -7, },
+ { 3, -2, 3, -2, 3, -2, 3, -2,
+ 3, -2, 3, -2, 3, -2, 3, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
},
};
-
+#endif
+#if USE_TEMPORALFILTER_12TAP
DECLARE_ALIGNED(16, const int8_t,
- sub_pel_filters_10sharp_ver_signal_dir[15][6][16]) = {
+ sub_pel_filters_temporalfilter_12_signal_dir[15][2][16]) = {
{
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
- {-1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3},
- {-6, 127, -6, 127, -6, 127, -6, 127, -6, 127, -6, 127, -6, 127, -6, 127},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1},
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
+ { 0, 1, -1, 3, -7, 127, 8, -4,
+ 2, -1, 0, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -1, 3, -7, 127,
+ 8, -4, 2, -1, 0, 0, 0, 0, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5},
- {-12, 124, -12, 124, -12, 124, -12, 124, -12, 124, -12, 124,
- -12, 124, -12, 124},
- {18, -7, 18, -7, 18, -7, 18, -7, 18, -7, 18, -7, 18, -7, 18, -7},
- {3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2},
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
+ { 0, 1, -3, 5, -12, 124, 18, -8,
+ 4, -2, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -3, 5, -12, 124,
+ 18, -8, 4, -2, 1, 0, 0, 0, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7},
- {-17, 119, -17, 119, -17, 119, -17, 119, -17, 119, -17, 119,
- -17, 119, -17, 119},
- {28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11},
- {5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { -1, 2, -4, 8, -17, 120, 28, -11,
+ 6, -3, 1, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 2, -4, 8, -17, 120,
+ 28, -11, 6, -3, 1, -1, 0, 0, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {-20, 114, -20, 114, -20, 114, -20, 114, -20, 114, -20, 114,
- -20, 114, -20, 114},
- {38, -14, 38, -14, 38, -14, 38, -14, 38, -14, 38, -14, 38, -14, 38, -14},
- {7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { -1, 2, -4, 10, -21, 114, 38, -15,
+ 8, -4, 2, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 2, -4, 10, -21, 114,
+ 38, -15, 8, -4, 2, -1, 0, 0, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9},
- {-22, 107, -22, 107, -22, 107, -22, 107, -22, 107, -22, 107,
- -22, 107, -22, 107},
- {49, -17, 49, -17, 49, -17, 49, -17, 49, -17, 49, -17, 49, -17, 49, -17},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { -1, 3, -5, 11, -23, 107, 49, -18,
+ 9, -5, 2, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -5, 11, -23, 107,
+ 49, -18, 9, -5, 2, -1, 0, 0, },
},
{
- {0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2},
- {-5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10},
- {-24, 99, -24, 99, -24, 99, -24, 99, -24, 99, -24, 99, -24, 99, -24, 99},
- {59, -20, 59, -20, 59, -20, 59, -20, 59, -20, 59, -20, 59, -20, 59, -20},
- {9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4},
- {2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0},
+ { -1, 3, -6, 12, -25, 99, 60, -21,
+ 11, -6, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -6, 12, -25, 99,
+ 60, -21, 11, -6, 3, -1, 0, 0, },
},
{
- {0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2},
- {-5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10},
- {-24, 90, -24, 90, -24, 90, -24, 90, -24, 90, -24, 90, -24, 90, -24, 90},
- {70, -22, 70, -22, 70, -22, 70, -22, 70, -22, 70, -22, 70, -22, 70, -22},
- {10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5},
- {2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0},
+ { -1, 3, -6, 12, -25, 90, 70, -23,
+ 12, -6, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -6, 12, -25, 90,
+ 70, -23, 12, -6, 3, -1, 0, 0, },
},
{
- {0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2},
- {-5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10},
- {-23, 80, -23, 80, -23, 80, -23, 80, -23, 80, -23, 80, -23, 80, -23, 80},
- {80, -23, 80, -23, 80, -23, 80, -23, 80, -23, 80, -23, 80, -23, 80, -23},
- {10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5},
- {2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0},
+ { -1, 3, -6, 12, -24, 80, 80, -24,
+ 12, -6, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -6, 12, -24, 80,
+ 80, -24, 12, -6, 3, -1, 0, 0, },
},
{
- {0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2},
- {-5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10},
- {-22, 70, -22, 70, -22, 70, -22, 70, -22, 70, -22, 70, -22, 70, -22, 70},
- {90, -24, 90, -24, 90, -24, 90, -24, 90, -24, 90, -24, 90, -24, 90, -24},
- {10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5},
- {2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0},
+ { -1, 3, -6, 12, -23, 70, 90, -25,
+ 12, -6, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -6, 12, -23, 70,
+ 90, -25, 12, -6, 3, -1, 0, 0, },
},
{
- {0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2},
- {-4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9},
- {-20, 59, -20, 59, -20, 59, -20, 59, -20, 59, -20, 59, -20, 59, -20, 59},
- {99, -24, 99, -24, 99, -24, 99, -24, 99, -24, 99, -24, 99, -24, 99, -24},
- {10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5, 10, -5},
- {2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0},
+ { -1, 3, -6, 11, -21, 60, 99, -25,
+ 12, -6, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 3, -6, 11, -21, 60,
+ 99, -25, 12, -6, 3, -1, 0, 0, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {-17, 49, -17, 49, -17, 49, -17, 49, -17, 49, -17, 49, -17, 49, -17, 49},
- {107, -22, 107, -22, 107, -22, 107, -22, 107, -22, 107, -22,
- 107, -22, 107, -22},
- {9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4, 9, -4},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { -1, 2, -5, 9, -18, 49, 107, -23,
+ 11, -5, 3, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 2, -5, 9, -18, 49,
+ 107, -23, 11, -5, 3, -1, 0, 0, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7},
- {-14, 38, -14, 38, -14, 38, -14, 38, -14, 38, -14, 38, -14, 38, -14, 38},
- {114, -20, 114, -20, 114, -20, 114, -20, 114, -20, 114, -20,
- 114, -20, 114, -20},
- {8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { -1, 2, -4, 8, -15, 38, 114, -21,
+ 10, -4, 2, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 2, -4, 8, -15, 38,
+ 114, -21, 10, -4, 2, -1, 0, 0, },
},
{
- {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1},
- {-2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5},
- {-11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28, -11, 28},
- {119, -17, 119, -17, 119, -17, 119, -17, 119, -17, 119, -17,
- 119, -17, 119, -17},
- {7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3, 7, -3},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { -1, 1, -3, 6, -11, 28, 120, -17,
+ 8, -4, 2, -1, 0, 0, 0, 0, },
+ { 0, 0, -1, 1, -3, 6, -11, 28,
+ 120, -17, 8, -4, 2, -1, 0, 0, },
},
{
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
- {-2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3, -2, 3},
- {-7, 18, -7, 18, -7, 18, -7, 18, -7, 18, -7, 18, -7, 18, -7, 18},
- {124, -12, 124, -12, 124, -12, 124, -12, 124, -12, 124, -12,
- 124, -12, 124, -12},
- {5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2, 5, -2},
- {1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0},
+ { 0, 1, -2, 4, -8, 18, 124, -12,
+ 5, -3, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 1, -2, 4, -8, 18,
+ 124, -12, 5, -3, 1, 0, 0, 0, },
},
{
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
- {-1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2, -1, 2},
- {-4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8, -4, 8},
- {127, -6, 127, -6, 127, -6, 127, -6, 127, -6, 127, -6,
- 127, -6, 127, -6},
- {3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1, 3, -1},
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
+ { 0, 0, -1, 2, -4, 8, 127, -7,
+ 3, -1, 1, 0, 0, 0, 0, 0, },
+ { 0, 0, 0, 0, -1, 2, -4, 8,
+ 127, -7, 3, -1, 1, 0, 0, 0, },
},
};
-#endif // CONFIG_EXT_INTERP
+#endif
+#if USE_TEMPORALFILTER_12TAP
+DECLARE_ALIGNED(16, const int8_t,
+ sub_pel_filters_temporalfilter_12_ver_signal_dir[15][6][16]) = {
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -7, 127, -7, 127, -7, 127, -7, 127,
+ -7, 127, -7, 127, -7, 127, -7, 127, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -3, 5, -3, 5, -3, 5, -3, 5,
+ -3, 5, -3, 5, -3, 5, -3, 5, },
+ {-12, 124, -12, 124, -12, 124, -12, 124,
+ -12, 124, -12, 124, -12, 124, -12, 124, },
+ { 18, -8, 18, -8, 18, -8, 18, -8,
+ 18, -8, 18, -8, 18, -8, 18, -8, },
+ { 4, -2, 4, -2, 4, -2, 4, -2,
+ 4, -2, 4, -2, 4, -2, 4, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-17, 120, -17, 120, -17, 120, -17, 120,
+ -17, 120, -17, 120, -17, 120, -17, 120, },
+ { 28, -11, 28, -11, 28, -11, 28, -11,
+ 28, -11, 28, -11, 28, -11, 28, -11, },
+ { 6, -3, 6, -3, 6, -3, 6, -3,
+ 6, -3, 6, -3, 6, -3, 6, -3, },
+ { 1, -1, 1, -1, 1, -1, 1, -1,
+ 1, -1, 1, -1, 1, -1, 1, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 10, -4, 10, -4, 10, -4, 10,
+ -4, 10, -4, 10, -4, 10, -4, 10, },
+ {-21, 114, -21, 114, -21, 114, -21, 114,
+ -21, 114, -21, 114, -21, 114, -21, 114, },
+ { 38, -15, 38, -15, 38, -15, 38, -15,
+ 38, -15, 38, -15, 38, -15, 38, -15, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -5, 11, -5, 11, -5, 11, -5, 11,
+ -5, 11, -5, 11, -5, 11, -5, 11, },
+ {-23, 107, -23, 107, -23, 107, -23, 107,
+ -23, 107, -23, 107, -23, 107, -23, 107, },
+ { 49, -18, 49, -18, 49, -18, 49, -18,
+ 49, -18, 49, -18, 49, -18, 49, -18, },
+ { 9, -5, 9, -5, 9, -5, 9, -5,
+ 9, -5, 9, -5, 9, -5, 9, -5, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12,
+ -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-25, 99, -25, 99, -25, 99, -25, 99,
+ -25, 99, -25, 99, -25, 99, -25, 99, },
+ { 60, -21, 60, -21, 60, -21, 60, -21,
+ 60, -21, 60, -21, 60, -21, 60, -21, },
+ { 11, -6, 11, -6, 11, -6, 11, -6,
+ 11, -6, 11, -6, 11, -6, 11, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12,
+ -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-25, 90, -25, 90, -25, 90, -25, 90,
+ -25, 90, -25, 90, -25, 90, -25, 90, },
+ { 70, -23, 70, -23, 70, -23, 70, -23,
+ 70, -23, 70, -23, 70, -23, 70, -23, },
+ { 12, -6, 12, -6, 12, -6, 12, -6,
+ 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12,
+ -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-24, 80, -24, 80, -24, 80, -24, 80,
+ -24, 80, -24, 80, -24, 80, -24, 80, },
+ { 80, -24, 80, -24, 80, -24, 80, -24,
+ 80, -24, 80, -24, 80, -24, 80, -24, },
+ { 12, -6, 12, -6, 12, -6, 12, -6,
+ 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12,
+ -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-23, 70, -23, 70, -23, 70, -23, 70,
+ -23, 70, -23, 70, -23, 70, -23, 70, },
+ { 90, -25, 90, -25, 90, -25, 90, -25,
+ 90, -25, 90, -25, 90, -25, 90, -25, },
+ { 12, -6, 12, -6, 12, -6, 12, -6,
+ 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3,
+ -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 11, -6, 11, -6, 11, -6, 11,
+ -6, 11, -6, 11, -6, 11, -6, 11, },
+ {-21, 60, -21, 60, -21, 60, -21, 60,
+ -21, 60, -21, 60, -21, 60, -21, 60, },
+ { 99, -25, 99, -25, 99, -25, 99, -25,
+ 99, -25, 99, -25, 99, -25, 99, -25, },
+ { 12, -6, 12, -6, 12, -6, 12, -6,
+ 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -5, 9, -5, 9, -5, 9, -5, 9,
+ -5, 9, -5, 9, -5, 9, -5, 9, },
+ {-18, 49, -18, 49, -18, 49, -18, 49,
+ -18, 49, -18, 49, -18, 49, -18, 49, },
+ {107, -23, 107, -23, 107, -23, 107, -23,
+ 107, -23, 107, -23, 107, -23, 107, -23, },
+ { 11, -5, 11, -5, 11, -5, 11, -5,
+ 11, -5, 11, -5, 11, -5, 11, -5, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-15, 38, -15, 38, -15, 38, -15, 38,
+ -15, 38, -15, 38, -15, 38, -15, 38, },
+ {114, -21, 114, -21, 114, -21, 114, -21,
+ 114, -21, 114, -21, 114, -21, 114, -21, },
+ { 10, -4, 10, -4, 10, -4, 10, -4,
+ 10, -4, 10, -4, 10, -4, 10, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 1, -1, 1, -1, 1, -1, 1,
+ -1, 1, -1, 1, -1, 1, -1, 1, },
+ { -3, 6, -3, 6, -3, 6, -3, 6,
+ -3, 6, -3, 6, -3, 6, -3, 6, },
+ {-11, 28, -11, 28, -11, 28, -11, 28,
+ -11, 28, -11, 28, -11, 28, -11, 28, },
+ {120, -17, 120, -17, 120, -17, 120, -17,
+ 120, -17, 120, -17, 120, -17, 120, -17, },
+ { 8, -4, 8, -4, 8, -4, 8, -4,
+ 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1,
+ 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1,
+ 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 4, -2, 4, -2, 4, -2, 4,
+ -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -8, 18, -8, 18, -8, 18, -8, 18,
+ -8, 18, -8, 18, -8, 18, -8, 18, },
+ {124, -12, 124, -12, 124, -12, 124, -12,
+ 124, -12, 124, -12, 124, -12, 124, -12, },
+ { 5, -3, 5, -3, 5, -3, 5, -3,
+ 5, -3, 5, -3, 5, -3, 5, -3, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -1, 2, -1, 2, -1, 2, -1, 2,
+ -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8,
+ -4, 8, -4, 8, -4, 8, -4, 8, },
+ {127, -7, 127, -7, 127, -7, 127, -7,
+ 127, -7, 127, -7, 127, -7, 127, -7, },
+ { 3, -1, 3, -1, 3, -1, 3, -1,
+ 3, -1, 3, -1, 3, -1, 3, -1, },
+ { 1, 0, 1, 0, 1, 0, 1, 0,
+ 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+};
+#endif
diff --git a/vp10/common/x86/vp10_convolve_ssse3.c b/vp10/common/x86/vp10_convolve_ssse3.c
index 75520c9..07dc11d 100644
--- a/vp10/common/x86/vp10_convolve_ssse3.c
+++ b/vp10/common/x86/vp10_convolve_ssse3.c
@@ -81,8 +81,9 @@
static store_pixel_t store4pixelTab[2] = {
store_4_pixel_only, accumulate_store_4_pixel};
-void horiz_w4_ssse3(const uint8_t *src, const __m128i *f,
- int tapsNum, store_pixel_t store_func, uint8_t *dst) {
+static void horiz_w4_ssse3(const uint8_t *src, const __m128i *f,
+ int tapsNum, store_pixel_t store_func,
+ uint8_t *dst) {
__m128i sumPairRow[4];
__m128i sumPairCol[8];
__m128i pixel;
@@ -122,40 +123,40 @@
store_func(&sumPairRow[1], dst);
}
-void horiz_w8_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
- store_pixel_t store, uint8_t *buf) {
+static void horiz_w8_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
+ store_pixel_t store, uint8_t *buf) {
horiz_w4_ssse3(src, f, tapsNum, store, buf);
src += 4;
buf += 4;
horiz_w4_ssse3(src, f, tapsNum, store, buf);
}
-void horiz_w16_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
- store_pixel_t store, uint8_t *buf) {
+static void horiz_w16_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
+ store_pixel_t store, uint8_t *buf) {
horiz_w8_ssse3(src, f, tapsNum, store, buf);
src += 8;
buf += 8;
horiz_w8_ssse3(src, f, tapsNum, store, buf);
}
-void horiz_w32_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
- store_pixel_t store, uint8_t *buf) {
+static void horiz_w32_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
+ store_pixel_t store, uint8_t *buf) {
horiz_w16_ssse3(src, f, tapsNum, store, buf);
src += 16;
buf += 16;
horiz_w16_ssse3(src, f, tapsNum, store, buf);
}
-void horiz_w64_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
- store_pixel_t store, uint8_t *buf) {
+static void horiz_w64_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
+ store_pixel_t store, uint8_t *buf) {
horiz_w32_ssse3(src, f, tapsNum, store, buf);
src += 32;
buf += 32;
horiz_w32_ssse3(src, f, tapsNum, store, buf);
}
-void horiz_w128_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
- store_pixel_t store, uint8_t *buf) {
+static void horiz_w128_ssse3(const uint8_t *src, const __m128i *f, int tapsNum,
+ store_pixel_t store, uint8_t *buf) {
horiz_w64_ssse3(src, f, tapsNum, store, buf);
src += 64;
buf += 64;
@@ -172,8 +173,8 @@
horiz_w128_ssse3,
};
-void filter_horiz_ssse3(const uint8_t *src, __m128i *f, int tapsNum, int width,
- store_pixel_t store, uint8_t *dst) {
+static void filter_horiz_ssse3(const uint8_t *src, __m128i *f, int tapsNum,
+ int width, store_pixel_t store, uint8_t *dst) {
switch (width) {
// Note:
// For width=2 and 4, store function must be different
@@ -813,9 +814,10 @@
store_func(&sum, dst);
}
-void filter_vert_compute_small(const uint8_t *src, int src_stride, __m128i *f,
- int tapsNum, store_pixel_t store_func, int h,
- uint8_t *dst, int dst_stride) {
+static void filter_vert_compute_small(const uint8_t *src, int src_stride,
+ __m128i *f, int tapsNum,
+ store_pixel_t store_func, int h,
+ uint8_t *dst, int dst_stride) {
int rowIndex = 0;
do {
filter_vert_horiz_parallel_ssse3(src, src_stride, f, tapsNum, store_func,
@@ -826,9 +828,10 @@
} while (rowIndex < h);
}
-void filter_vert_compute_large(const uint8_t *src, int src_stride, __m128i *f,
- int tapsNum, store_pixel_t store_func, int w,
- int h, uint8_t *dst, int dst_stride) {
+static void filter_vert_compute_large(const uint8_t *src, int src_stride,
+ __m128i *f, int tapsNum,
+ store_pixel_t store_func, int w, int h,
+ uint8_t *dst, int dst_stride) {
int col;
int rowIndex = 0;
const uint8_t *src_ptr = src;
diff --git a/vp10/common/x86/vp10_fwd_txfm2d_sse4.c b/vp10/common/x86/vp10_fwd_txfm2d_sse4.c
index 1d70f14..ff04dc8 100644
--- a/vp10/common/x86/vp10_fwd_txfm2d_sse4.c
+++ b/vp10/common/x86/vp10_fwd_txfm2d_sse4.c
@@ -8,6 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
+#include "./vp10_rtcd.h"
#include "vp10/common/enums.h"
#include "vp10/common/vp10_txfm.h"
#include "vp10/common/x86/vp10_txfm1d_sse4.h"
diff --git a/vp10/common/x86/vp10_fwd_txfm_sse2.c b/vp10/common/x86/vp10_fwd_txfm_sse2.c
index 032c3cc..30bce5f 100644
--- a/vp10/common/x86/vp10_fwd_txfm_sse2.c
+++ b/vp10/common/x86/vp10_fwd_txfm_sse2.c
@@ -10,6 +10,7 @@
#include <emmintrin.h> // SSE2
+#include "./vp10_rtcd.h"
#include "./vpx_config.h"
#include "vpx_dsp/vpx_dsp_common.h"
#include "vpx_dsp/x86/fwd_txfm_sse2.h"
diff --git a/vp10/common/x86/vp10_highbd_convolve_filters_sse4.c b/vp10/common/x86/vp10_highbd_convolve_filters_sse4.c
new file mode 100644
index 0000000..0251022
--- /dev/null
+++ b/vp10/common/x86/vp10_highbd_convolve_filters_sse4.c
@@ -0,0 +1,393 @@
+/*
+ * 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 "./vpx_config.h"
+#include "vp10/common/filter.h"
+
+#if CONFIG_VP9_HIGHBITDEPTH
+#if CONFIG_EXT_INTERP
+DECLARE_ALIGNED(16, const int16_t,
+ sub_pel_filters_10sharp_highbd_ver_signal_dir[15][6][8]) = {
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 127, -6, 127, -6, 127, -6, 127, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 5, -2, 5, -2, 5, -2, 5, },
+ {-12, 124, -12, 124, -12, 124, -12, 124, },
+ { 18, -7, 18, -7, 18, -7, 18, -7, },
+ { 3, -2, 3, -2, 3, -2, 3, -2, },
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -3, 7, -3, 7, -3, 7, -3, 7, },
+ {-17, 119, -17, 119, -17, 119, -17, 119, },
+ { 28, -11, 28, -11, 28, -11, 28, -11, },
+ { 5, -2, 5, -2, 5, -2, 5, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-20, 114, -20, 114, -20, 114, -20, 114, },
+ { 38, -14, 38, -14, 38, -14, 38, -14, },
+ { 7, -3, 7, -3, 7, -3, 7, -3, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -4, 9, -4, 9, -4, 9, -4, 9, },
+ {-22, 107, -22, 107, -22, 107, -22, 107, },
+ { 49, -17, 49, -17, 49, -17, 49, -17, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-24, 99, -24, 99, -24, 99, -24, 99, },
+ { 59, -20, 59, -20, 59, -20, 59, -20, },
+ { 9, -4, 9, -4, 9, -4, 9, -4, },
+ { 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-24, 90, -24, 90, -24, 90, -24, 90, },
+ { 70, -22, 70, -22, 70, -22, 70, -22, },
+ { 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-23, 80, -23, 80, -23, 80, -23, 80, },
+ { 80, -23, 80, -23, 80, -23, 80, -23, },
+ { 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -5, 10, -5, 10, -5, 10, -5, 10, },
+ {-22, 70, -22, 70, -22, 70, -22, 70, },
+ { 90, -24, 90, -24, 90, -24, 90, -24, },
+ { 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 2, 0, 2, 0, 2, 0, 2, },
+ { -4, 9, -4, 9, -4, 9, -4, 9, },
+ {-20, 59, -20, 59, -20, 59, -20, 59, },
+ { 99, -24, 99, -24, 99, -24, 99, -24, },
+ { 10, -5, 10, -5, 10, -5, 10, -5, },
+ { 2, 0, 2, 0, 2, 0, 2, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-17, 49, -17, 49, -17, 49, -17, 49, },
+ {107, -22, 107, -22, 107, -22, 107, -22, },
+ { 9, -4, 9, -4, 9, -4, 9, -4, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -3, 7, -3, 7, -3, 7, -3, 7, },
+ {-14, 38, -14, 38, -14, 38, -14, 38, },
+ {114, -20, 114, -20, 114, -20, 114, -20, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 5, -2, 5, -2, 5, -2, 5, },
+ {-11, 28, -11, 28, -11, 28, -11, 28, },
+ {119, -17, 119, -17, 119, -17, 119, -17, },
+ { 7, -3, 7, -3, 7, -3, 7, -3, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -2, 3, -2, 3, -2, 3, -2, 3, },
+ { -7, 18, -7, 18, -7, 18, -7, 18, },
+ {124, -12, 124, -12, 124, -12, 124, -12, },
+ { 5, -2, 5, -2, 5, -2, 5, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {127, -6, 127, -6, 127, -6, 127, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+};
+#endif
+#endif
+#if CONFIG_VP9_HIGHBITDEPTH
+#if CONFIG_EXT_INTERP
+DECLARE_ALIGNED(16, const int16_t,
+ sub_pel_filters_12sharp_highbd_ver_signal_dir[15][6][8]) = {
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 3, -2, 3, -2, 3, -2, 3, },
+ { -7, 127, -7, 127, -7, 127, -7, 127, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -3, 6, -3, 6, -3, 6, -3, 6, },
+ {-13, 124, -13, 124, -13, 124, -13, 124, },
+ { 18, -8, 18, -8, 18, -8, 18, -8, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-18, 120, -18, 120, -18, 120, -18, 120, },
+ { 28, -12, 28, -12, 28, -12, 28, -12, },
+ { 7, -4, 7, -4, 7, -4, 7, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 10, -6, 10, -6, 10, -6, 10, },
+ {-21, 115, -21, 115, -21, 115, -21, 115, },
+ { 38, -15, 38, -15, 38, -15, 38, -15, },
+ { 8, -5, 8, -5, 8, -5, 8, -5, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-24, 108, -24, 108, -24, 108, -24, 108, },
+ { 49, -18, 49, -18, 49, -18, 49, -18, },
+ { 10, -6, 10, -6, 10, -6, 10, -6, },
+ { 3, -2, 3, -2, 3, -2, 3, -2, },
+ },
+ {
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-25, 100, -25, 100, -25, 100, -25, 100, },
+ { 60, -21, 60, -21, 60, -21, 60, -21, },
+ { 11, -7, 11, -7, 11, -7, 11, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ },
+ {
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-26, 91, -26, 91, -26, 91, -26, 91, },
+ { 71, -24, 71, -24, 71, -24, 71, -24, },
+ { 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ },
+ {
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-25, 81, -25, 81, -25, 81, -25, 81, },
+ { 81, -25, 81, -25, 81, -25, 81, -25, },
+ { 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ },
+ {
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 13, -7, 13, -7, 13, -7, 13, },
+ {-24, 71, -24, 71, -24, 71, -24, 71, },
+ { 91, -26, 91, -26, 91, -26, 91, -26, },
+ { 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ },
+ {
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -7, 11, -7, 11, -7, 11, -7, 11, },
+ {-21, 60, -21, 60, -21, 60, -21, 60, },
+ {100, -25, 100, -25, 100, -25, 100, -25, },
+ { 13, -7, 13, -7, 13, -7, 13, -7, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ },
+ {
+ { -2, 3, -2, 3, -2, 3, -2, 3, },
+ { -6, 10, -6, 10, -6, 10, -6, 10, },
+ {-18, 49, -18, 49, -18, 49, -18, 49, },
+ {108, -24, 108, -24, 108, -24, 108, -24, },
+ { 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -5, 8, -5, 8, -5, 8, -5, 8, },
+ {-15, 38, -15, 38, -15, 38, -15, 38, },
+ {115, -21, 115, -21, 115, -21, 115, -21, },
+ { 10, -6, 10, -6, 10, -6, 10, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 7, -4, 7, -4, 7, -4, 7, },
+ {-12, 28, -12, 28, -12, 28, -12, 28, },
+ {120, -18, 120, -18, 120, -18, 120, -18, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -8, 18, -8, 18, -8, 18, -8, 18, },
+ {124, -13, 124, -13, 124, -13, 124, -13, },
+ { 6, -3, 6, -3, 6, -3, 6, -3, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {127, -7, 127, -7, 127, -7, 127, -7, },
+ { 3, -2, 3, -2, 3, -2, 3, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+};
+#endif
+#endif
+#if CONFIG_VP9_HIGHBITDEPTH
+#if USE_TEMPORALFILTER_12TAP
+DECLARE_ALIGNED(16, const int16_t,
+ sub_pel_filters_temporalfilter_12_highbd_ver_signal_dir[15][6][8]) = {
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -7, 127, -7, 127, -7, 127, -7, 127, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -3, 5, -3, 5, -3, 5, -3, 5, },
+ {-12, 124, -12, 124, -12, 124, -12, 124, },
+ { 18, -8, 18, -8, 18, -8, 18, -8, },
+ { 4, -2, 4, -2, 4, -2, 4, -2, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-17, 120, -17, 120, -17, 120, -17, 120, },
+ { 28, -11, 28, -11, 28, -11, 28, -11, },
+ { 6, -3, 6, -3, 6, -3, 6, -3, },
+ { 1, -1, 1, -1, 1, -1, 1, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 10, -4, 10, -4, 10, -4, 10, },
+ {-21, 114, -21, 114, -21, 114, -21, 114, },
+ { 38, -15, 38, -15, 38, -15, 38, -15, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -5, 11, -5, 11, -5, 11, -5, 11, },
+ {-23, 107, -23, 107, -23, 107, -23, 107, },
+ { 49, -18, 49, -18, 49, -18, 49, -18, },
+ { 9, -5, 9, -5, 9, -5, 9, -5, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-25, 99, -25, 99, -25, 99, -25, 99, },
+ { 60, -21, 60, -21, 60, -21, 60, -21, },
+ { 11, -6, 11, -6, 11, -6, 11, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-25, 90, -25, 90, -25, 90, -25, 90, },
+ { 70, -23, 70, -23, 70, -23, 70, -23, },
+ { 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-24, 80, -24, 80, -24, 80, -24, 80, },
+ { 80, -24, 80, -24, 80, -24, 80, -24, },
+ { 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 12, -6, 12, -6, 12, -6, 12, },
+ {-23, 70, -23, 70, -23, 70, -23, 70, },
+ { 90, -25, 90, -25, 90, -25, 90, -25, },
+ { 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 3, -1, 3, -1, 3, -1, 3, },
+ { -6, 11, -6, 11, -6, 11, -6, 11, },
+ {-21, 60, -21, 60, -21, 60, -21, 60, },
+ { 99, -25, 99, -25, 99, -25, 99, -25, },
+ { 12, -6, 12, -6, 12, -6, 12, -6, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -5, 9, -5, 9, -5, 9, -5, 9, },
+ {-18, 49, -18, 49, -18, 49, -18, 49, },
+ {107, -23, 107, -23, 107, -23, 107, -23, },
+ { 11, -5, 11, -5, 11, -5, 11, -5, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ },
+ {
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {-15, 38, -15, 38, -15, 38, -15, 38, },
+ {114, -21, 114, -21, 114, -21, 114, -21, },
+ { 10, -4, 10, -4, 10, -4, 10, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { -1, 1, -1, 1, -1, 1, -1, 1, },
+ { -3, 6, -3, 6, -3, 6, -3, 6, },
+ {-11, 28, -11, 28, -11, 28, -11, 28, },
+ {120, -17, 120, -17, 120, -17, 120, -17, },
+ { 8, -4, 8, -4, 8, -4, 8, -4, },
+ { 2, -1, 2, -1, 2, -1, 2, -1, },
+ },
+ {
+ { 0, 1, 0, 1, 0, 1, 0, 1, },
+ { -2, 4, -2, 4, -2, 4, -2, 4, },
+ { -8, 18, -8, 18, -8, 18, -8, 18, },
+ {124, -12, 124, -12, 124, -12, 124, -12, },
+ { 5, -3, 5, -3, 5, -3, 5, -3, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+ {
+ { 0, 0, 0, 0, 0, 0, 0, 0, },
+ { -1, 2, -1, 2, -1, 2, -1, 2, },
+ { -4, 8, -4, 8, -4, 8, -4, 8, },
+ {127, -7, 127, -7, 127, -7, 127, -7, },
+ { 3, -1, 3, -1, 3, -1, 3, -1, },
+ { 1, 0, 1, 0, 1, 0, 1, 0, },
+ },
+};
+#endif
+#endif
diff --git a/vp10/common/x86/vp10_highbd_convolve_sse4.c b/vp10/common/x86/vp10_highbd_convolve_sse4.c
new file mode 100644
index 0000000..e828178
--- /dev/null
+++ b/vp10/common/x86/vp10_highbd_convolve_sse4.c
@@ -0,0 +1,474 @@
+/*
+ * 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 <assert.h>
+#include <smmintrin.h>
+
+#include "./vp10_rtcd.h"
+#include "vp10/common/filter.h"
+
+typedef void (*TransposeSave)(const int width, int pixelsNum,
+ uint32_t *src, int src_stride,
+ uint16_t *dst, int dst_stride,
+ int bd);
+
+// pixelsNum 0: write all 4 pixels
+// 1/2/3: residual pixels 1/2/3
+static void writePixel(__m128i *u, int width, int pixelsNum,
+ uint16_t *dst, int dst_stride) {
+ if (2 == width) {
+ if (0 == pixelsNum) {
+ *(int *)dst = _mm_cvtsi128_si32(u[0]);
+ *(int *)(dst + dst_stride) = _mm_cvtsi128_si32(u[1]);
+ *(int *)(dst + 2 * dst_stride) = _mm_cvtsi128_si32(u[2]);
+ *(int *)(dst + 3 * dst_stride) = _mm_cvtsi128_si32(u[3]);
+ } else if (1 == pixelsNum) {
+ *(int *)dst = _mm_cvtsi128_si32(u[0]);
+ } else if (2 == pixelsNum) {
+ *(int *)dst = _mm_cvtsi128_si32(u[0]);
+ *(int *)(dst + dst_stride) = _mm_cvtsi128_si32(u[1]);
+ } else if (3 == pixelsNum) {
+ *(int *)dst = _mm_cvtsi128_si32(u[0]);
+ *(int *)(dst + dst_stride) = _mm_cvtsi128_si32(u[1]);
+ *(int *)(dst + 2 * dst_stride) = _mm_cvtsi128_si32(u[2]);
+ }
+ } else {
+ if (0 == pixelsNum) {
+ _mm_storel_epi64((__m128i *)dst, u[0]);
+ _mm_storel_epi64((__m128i *)(dst + dst_stride), u[1]);
+ _mm_storel_epi64((__m128i *)(dst + 2 * dst_stride), u[2]);
+ _mm_storel_epi64((__m128i *)(dst + 3 * dst_stride), u[3]);
+ } else if (1 == pixelsNum) {
+ _mm_storel_epi64((__m128i *)dst, u[0]);
+ } else if (2 == pixelsNum) {
+ _mm_storel_epi64((__m128i *)dst, u[0]);
+ _mm_storel_epi64((__m128i *)(dst + dst_stride), u[1]);
+ } else if (3 == pixelsNum) {
+ _mm_storel_epi64((__m128i *)dst, u[0]);
+ _mm_storel_epi64((__m128i *)(dst + dst_stride), u[1]);
+ _mm_storel_epi64((__m128i *)(dst + 2 * dst_stride), u[2]);
+ }
+ }
+}
+
+// 16-bit pixels clip with bd (10/12)
+static void highbd_clip(__m128i *p, int numVecs, int bd) {
+ const __m128i zero = _mm_setzero_si128();
+ const __m128i one = _mm_set1_epi16(1);
+ const __m128i max = _mm_sub_epi16(_mm_slli_epi16(one, bd), one);
+ __m128i clamped, mask;
+ int i;
+
+ for (i = 0; i < numVecs; i++) {
+ mask = _mm_cmpgt_epi16(p[i], max);
+ clamped = _mm_andnot_si128(mask, p[i]);
+ mask = _mm_and_si128(mask, max);
+ clamped = _mm_or_si128(mask, clamped);
+ mask = _mm_cmpgt_epi16(clamped, zero);
+ p[i] = _mm_and_si128(clamped, mask);
+ }
+}
+
+static void transClipPixel(uint32_t *src, int src_stride, __m128i *u, int bd) {
+ __m128i v0, v1;
+ __m128i rnd = _mm_set1_epi32(1 << (FILTER_BITS - 1));
+
+ u[0] = _mm_loadu_si128((__m128i const *)src);
+ u[1] = _mm_loadu_si128((__m128i const *)(src + src_stride));
+ u[2] = _mm_loadu_si128((__m128i const *)(src + 2 * src_stride));
+ u[3] = _mm_loadu_si128((__m128i const *)(src + 3 * src_stride));
+
+ u[0] = _mm_add_epi32(u[0], rnd);
+ u[1] = _mm_add_epi32(u[1], rnd);
+ u[2] = _mm_add_epi32(u[2], rnd);
+ u[3] = _mm_add_epi32(u[3], rnd);
+
+ u[0] = _mm_srai_epi32(u[0], FILTER_BITS);
+ u[1] = _mm_srai_epi32(u[1], FILTER_BITS);
+ u[2] = _mm_srai_epi32(u[2], FILTER_BITS);
+ u[3] = _mm_srai_epi32(u[3], FILTER_BITS);
+
+ u[0] = _mm_packus_epi32(u[0], u[1]);
+ u[1] = _mm_packus_epi32(u[2], u[3]);
+
+ highbd_clip(u, 2, bd);
+
+ v0 = _mm_unpacklo_epi16(u[0], u[1]);
+ v1 = _mm_unpackhi_epi16(u[0], u[1]);
+
+ u[0] = _mm_unpacklo_epi16(v0, v1);
+ u[2] = _mm_unpackhi_epi16(v0, v1);
+
+ u[1] = _mm_srli_si128(u[0], 8);
+ u[3] = _mm_srli_si128(u[2], 8);
+}
+
+// pixelsNum = 0 : all 4 rows of pixels will be saved.
+// pixelsNum = 1/2/3 : residual 1/2/4 rows of pixels will be saved.
+void trans_save_4x4(const int width, int pixelsNum,
+ uint32_t *src, int src_stride,
+ uint16_t *dst, int dst_stride,
+ int bd) {
+ __m128i u[4];
+ transClipPixel(src, src_stride, u, bd);
+ writePixel(u, width, pixelsNum, dst, dst_stride);
+}
+
+void trans_accum_save_4x4(const int width, int pixelsNum,
+ uint32_t *src, int src_stride,
+ uint16_t *dst, int dst_stride,
+ int bd) {
+ __m128i u[4], v[4];
+ const __m128i ones = _mm_set1_epi16(1);
+
+ transClipPixel(src, src_stride, u, bd);
+
+ v[0] = _mm_loadl_epi64((__m128i const *)dst);
+ v[1] = _mm_loadl_epi64((__m128i const *)(dst + dst_stride));
+ v[2] = _mm_loadl_epi64((__m128i const *)(dst + 2 * dst_stride));
+ v[3] = _mm_loadl_epi64((__m128i const *)(dst + 3 * dst_stride));
+
+ u[0] = _mm_add_epi16(u[0], v[0]);
+ u[1] = _mm_add_epi16(u[1], v[1]);
+ u[2] = _mm_add_epi16(u[2], v[2]);
+ u[3] = _mm_add_epi16(u[3], v[3]);
+
+ u[0] = _mm_add_epi16(u[0], ones);
+ u[1] = _mm_add_epi16(u[1], ones);
+ u[2] = _mm_add_epi16(u[2], ones);
+ u[3] = _mm_add_epi16(u[3], ones);
+
+ u[0] = _mm_srai_epi16(u[0], 1);
+ u[1] = _mm_srai_epi16(u[1], 1);
+ u[2] = _mm_srai_epi16(u[2], 1);
+ u[3] = _mm_srai_epi16(u[3], 1);
+
+ writePixel(u, width, pixelsNum, dst, dst_stride);
+}
+
+static TransposeSave transSaveTab[2] = {
+ trans_save_4x4, trans_accum_save_4x4};
+
+static INLINE void transpose_pair(__m128i *in, __m128i *out) {
+ __m128i x0, x1;
+
+ x0 = _mm_unpacklo_epi32(in[0], in[1]);
+ x1 = _mm_unpacklo_epi32(in[2], in[3]);
+
+ out[0] = _mm_unpacklo_epi64(x0, x1);
+ out[1] = _mm_unpackhi_epi64(x0, x1);
+
+ x0 = _mm_unpackhi_epi32(in[0], in[1]);
+ x1 = _mm_unpackhi_epi32(in[2], in[3]);
+
+ out[2] = _mm_unpacklo_epi64(x0, x1);
+ out[3] = _mm_unpackhi_epi64(x0, x1);
+
+ x0 = _mm_unpacklo_epi32(in[4], in[5]);
+ x1 = _mm_unpacklo_epi32(in[6], in[7]);
+
+ out[4] = _mm_unpacklo_epi64(x0, x1);
+ out[5] = _mm_unpackhi_epi64(x0, x1);
+}
+
+static void highbd_filter_horiz(const uint16_t *src, int src_stride,
+ __m128i *f, int tapsNum, uint32_t *buf) {
+ __m128i u[8], v[6];
+
+ if (tapsNum == 10) {
+ src -= 1;
+ }
+
+ u[0] = _mm_loadu_si128((__m128i const *)src);
+ u[1] = _mm_loadu_si128((__m128i const *)(src + src_stride));
+ u[2] = _mm_loadu_si128((__m128i const *)(src + 2 * src_stride));
+ u[3] = _mm_loadu_si128((__m128i const *)(src + 3 * src_stride));
+
+ u[4] = _mm_loadu_si128((__m128i const *)(src + 8));
+ u[5] = _mm_loadu_si128((__m128i const *)(src + src_stride + 8));
+ u[6] = _mm_loadu_si128((__m128i const *)(src + 2 * src_stride + 8));
+ u[7] = _mm_loadu_si128((__m128i const *)(src + 3 * src_stride + 8));
+
+ transpose_pair(u, v);
+
+ u[0] = _mm_madd_epi16(v[0], f[0]);
+ u[1] = _mm_madd_epi16(v[1], f[1]);
+ u[2] = _mm_madd_epi16(v[2], f[2]);
+ u[3] = _mm_madd_epi16(v[3], f[3]);
+ u[4] = _mm_madd_epi16(v[4], f[4]);
+ u[5] = _mm_madd_epi16(v[5], f[5]);
+
+ u[6] = _mm_min_epi32(u[2], u[3]);
+ u[7] = _mm_max_epi32(u[2], u[3]);
+
+ u[0] = _mm_add_epi32(u[0], u[1]);
+ u[0] = _mm_add_epi32(u[0], u[5]);
+ u[0] = _mm_add_epi32(u[0], u[4]);
+ u[0] = _mm_add_epi32(u[0], u[6]);
+ u[0] = _mm_add_epi32(u[0], u[7]);
+
+ _mm_storeu_si128((__m128i *)buf, u[0]);
+}
+
+void vp10_highbd_convolve_horiz_sse4_1(const uint16_t *src, int src_stride,
+ uint16_t *dst, int dst_stride,
+ int w, int h,
+ const InterpFilterParams filter_params,
+ const int subpel_x_q4, int x_step_q4,
+ int avg, int bd) {
+ DECLARE_ALIGNED(16, uint32_t, temp[4 * 4]);
+ __m128i verf[6];
+ HbdSubpelFilterCoeffs vCoeffs;
+ const uint16_t *srcPtr;
+ const int tapsNum = filter_params.taps;
+ int i, col, count, blkResidu, blkHeight;
+ TransposeSave transSave = transSaveTab[avg];
+ (void)x_step_q4;
+
+ if (0 == subpel_x_q4 || 16 != x_step_q4) {
+ vp10_highbd_convolve_horiz_c(src, src_stride, dst, dst_stride, w, h,
+ filter_params, subpel_x_q4, x_step_q4, avg,
+ bd);
+ return;
+ }
+
+ vCoeffs = vp10_hbd_get_subpel_filter_ver_signal_dir(
+ filter_params, subpel_x_q4 - 1);
+ if (!vCoeffs) {
+ vp10_highbd_convolve_horiz_c(src, src_stride, dst, dst_stride, w, h,
+ filter_params, subpel_x_q4, x_step_q4, avg,
+ bd);
+ return;
+ }
+
+ verf[0] = *((const __m128i *)(vCoeffs));
+ verf[1] = *((const __m128i *)(vCoeffs + 1));
+ verf[2] = *((const __m128i *)(vCoeffs + 2));
+ verf[3] = *((const __m128i *)(vCoeffs + 3));
+ verf[4] = *((const __m128i *)(vCoeffs + 4));
+ verf[5] = *((const __m128i *)(vCoeffs + 5));
+
+ src -= (tapsNum >> 1) - 1;
+ srcPtr = src;
+
+ count = 0;
+ blkHeight = h >> 2;
+ blkResidu = h & 3;
+
+ while (blkHeight != 0) {
+ for (col = 0; col < w; col += 4) {
+ for (i = 0; i < 4; ++i) {
+ highbd_filter_horiz(srcPtr, src_stride, verf, tapsNum, temp + (i * 4));
+ srcPtr += 1;
+ }
+ transSave(w, 0, temp, 4, dst + col, dst_stride, bd);
+ }
+ count++;
+ srcPtr = src + count * src_stride * 4;
+ dst += dst_stride * 4;
+ blkHeight--;
+ }
+
+ for (col = 0; col < w; col += 4) {
+ for (i = 0; i < 4; ++i) {
+ highbd_filter_horiz(srcPtr, src_stride, verf, tapsNum, temp + (i * 4));
+ srcPtr += 1;
+ }
+ transSave(w, blkResidu, temp, 4, dst + col, dst_stride, bd);
+ }
+}
+
+// Vertical convolutional filter
+
+typedef void (*WritePixels)(__m128i *u, int bd, uint16_t *dst);
+
+static void highbdRndingPacks(__m128i *u) {
+ __m128i rnd = _mm_set1_epi32(1 << (FILTER_BITS - 1));
+ u[0] = _mm_add_epi32(u[0], rnd);
+ u[0] = _mm_srai_epi32(u[0], FILTER_BITS);
+ u[0] = _mm_packus_epi32(u[0], u[0]);
+}
+
+static void write2pixelsOnly(__m128i *u, int bd, uint16_t *dst) {
+ highbdRndingPacks(u);
+ highbd_clip(u, 1, bd);
+ *(uint32_t *)dst = _mm_cvtsi128_si32(u[0]);
+}
+
+static void write2pixelsAccum(__m128i *u, int bd, uint16_t *dst) {
+ __m128i v = _mm_loadl_epi64((__m128i const *)dst);
+ const __m128i ones = _mm_set1_epi16(1);
+
+ highbdRndingPacks(u);
+ highbd_clip(u, 1, bd);
+
+ v = _mm_add_epi16(v, u[0]);
+ v = _mm_add_epi16(v, ones);
+ v = _mm_srai_epi16(v, 1);
+ *(uint32_t *)dst = _mm_cvtsi128_si32(v);
+}
+
+WritePixels write2pixelsTab[2] = {write2pixelsOnly, write2pixelsAccum};
+
+static void write4pixelsOnly(__m128i *u, int bd, uint16_t *dst) {
+ highbdRndingPacks(u);
+ highbd_clip(u, 1, bd);
+ _mm_storel_epi64((__m128i *)dst, u[0]);
+}
+
+static void write4pixelsAccum(__m128i *u, int bd, uint16_t *dst) {
+ __m128i v = _mm_loadl_epi64((__m128i const *)dst);
+ const __m128i ones = _mm_set1_epi16(1);
+
+ highbdRndingPacks(u);
+ highbd_clip(u, 1, bd);
+
+ v = _mm_add_epi16(v, u[0]);
+ v = _mm_add_epi16(v, ones);
+ v = _mm_srai_epi16(v, 1);
+ _mm_storel_epi64((__m128i *)dst, v);
+}
+
+WritePixels write4pixelsTab[2] = {write4pixelsOnly, write4pixelsAccum};
+
+static void filter_vert_horiz_parallel(const uint16_t *src, int src_stride,
+ const __m128i *f, int taps,
+ uint16_t *dst, WritePixels saveFunc,
+ int bd) {
+ __m128i s[12];
+ __m128i zero = _mm_setzero_si128();
+ int i = 0;
+ int r = 0;
+
+ // TODO(luoyi) treat s[12] as a circular buffer in width = 2 case
+ if (10 == taps) {
+ i += 1;
+ s[0] = zero;
+ }
+ while (i < 12) {
+ s[i] = _mm_loadu_si128((__m128i const *)(src + r * src_stride));
+ i += 1;
+ r += 1;
+ }
+
+ s[0] = _mm_unpacklo_epi16(s[0], s[1]);
+ s[2] = _mm_unpacklo_epi16(s[2], s[3]);
+ s[4] = _mm_unpacklo_epi16(s[4], s[5]);
+ s[6] = _mm_unpacklo_epi16(s[6], s[7]);
+ s[8] = _mm_unpacklo_epi16(s[8], s[9]);
+ s[10] = _mm_unpacklo_epi16(s[10], s[11]);
+
+ s[0] = _mm_madd_epi16(s[0], f[0]);
+ s[2] = _mm_madd_epi16(s[2], f[1]);
+ s[4] = _mm_madd_epi16(s[4], f[2]);
+ s[6] = _mm_madd_epi16(s[6], f[3]);
+ s[8] = _mm_madd_epi16(s[8], f[4]);
+ s[10] = _mm_madd_epi16(s[10], f[5]);
+
+ s[1] = _mm_min_epi32(s[4], s[6]);
+ s[3] = _mm_max_epi32(s[4], s[6]);
+
+ s[0] = _mm_add_epi32(s[0], s[2]);
+ s[0] = _mm_add_epi32(s[0], s[10]);
+ s[0] = _mm_add_epi32(s[0], s[8]);
+ s[0] = _mm_add_epi32(s[0], s[1]);
+ s[0] = _mm_add_epi32(s[0], s[3]);
+
+ saveFunc(s, bd, dst);
+}
+
+static void highbd_filter_vert_compute_large(const uint16_t *src,
+ int src_stride,
+ const __m128i *f, int taps,
+ int w, int h,
+ uint16_t *dst, int dst_stride,
+ int avg, int bd) {
+ int col;
+ int rowIndex = 0;
+ const uint16_t *src_ptr = src;
+ uint16_t *dst_ptr = dst;
+ const int step = 4;
+ WritePixels write4pixels = write4pixelsTab[avg];
+
+ do {
+ for (col = 0; col < w; col += step) {
+ filter_vert_horiz_parallel(src_ptr, src_stride, f, taps,
+ dst_ptr, write4pixels, bd);
+ src_ptr += step;
+ dst_ptr += step;
+ }
+ rowIndex++;
+ src_ptr = src + rowIndex * src_stride;
+ dst_ptr = dst + rowIndex * dst_stride;
+ } while (rowIndex < h);
+}
+
+static void highbd_filter_vert_compute_small(const uint16_t *src,
+ int src_stride,
+ const __m128i *f, int taps,
+ int w, int h,
+ uint16_t *dst, int dst_stride,
+ int avg, int bd) {
+ int rowIndex = 0;
+ WritePixels write2pixels = write2pixelsTab[avg];
+ (void)w;
+
+ do {
+ filter_vert_horiz_parallel(src, src_stride, f, taps, dst, write2pixels,
+ bd);
+ rowIndex++;
+ src += src_stride;
+ dst += dst_stride;
+ } while (rowIndex < h);
+}
+
+void vp10_highbd_convolve_vert_sse4_1(const uint16_t *src, int src_stride,
+ uint16_t *dst, int dst_stride,
+ int w, int h,
+ const InterpFilterParams filter_params,
+ const int subpel_y_q4, int y_step_q4,
+ int avg, int bd) {
+ __m128i verf[6];
+ HbdSubpelFilterCoeffs vCoeffs;
+ const int tapsNum = filter_params.taps;
+
+ if (0 == subpel_y_q4 || 16 != y_step_q4) {
+ vp10_highbd_convolve_vert_c(src, src_stride, dst, dst_stride, w, h,
+ filter_params, subpel_y_q4, y_step_q4, avg,
+ bd);
+ return;
+ }
+
+ vCoeffs = vp10_hbd_get_subpel_filter_ver_signal_dir(
+ filter_params, subpel_y_q4 - 1);
+ if (!vCoeffs) {
+ vp10_highbd_convolve_vert_c(src, src_stride, dst, dst_stride, w, h,
+ filter_params, subpel_y_q4, y_step_q4, avg,
+ bd);
+ return;
+ }
+
+ verf[0] = *((const __m128i *)(vCoeffs));
+ verf[1] = *((const __m128i *)(vCoeffs + 1));
+ verf[2] = *((const __m128i *)(vCoeffs + 2));
+ verf[3] = *((const __m128i *)(vCoeffs + 3));
+ verf[4] = *((const __m128i *)(vCoeffs + 4));
+ verf[5] = *((const __m128i *)(vCoeffs + 5));
+
+ src -= src_stride * ((tapsNum >> 1) - 1);
+
+ if (w > 2) {
+ highbd_filter_vert_compute_large(src, src_stride, verf, tapsNum, w, h,
+ dst, dst_stride, avg, bd);
+ } else {
+ highbd_filter_vert_compute_small(src, src_stride, verf, tapsNum, w, h,
+ dst, dst_stride, avg, bd);
+ }
+}
diff --git a/vp10/encoder/bitstream.c b/vp10/encoder/bitstream.c
index 389e40b..3999c94 100644
--- a/vp10/encoder/bitstream.c
+++ b/vp10/encoder/bitstream.c
@@ -121,7 +121,7 @@
static struct vp10_token motvar_encodings[MOTION_VARIATIONS];
#endif // CONFIG_OBMC || CONFIG_WARPED_MOTION
-void vp10_encode_token_init() {
+void vp10_encode_token_init(void) {
#if CONFIG_EXT_TX
int s;
for (s = 1; s < EXT_TX_SETS_INTER; ++s) {
diff --git a/vp10/encoder/bitstream.h b/vp10/encoder/bitstream.h
index 0284920..cacdb43 100644
--- a/vp10/encoder/bitstream.h
+++ b/vp10/encoder/bitstream.h
@@ -20,7 +20,7 @@
void vp10_pack_bitstream(VP10_COMP *const cpi, uint8_t *dest, size_t *size);
-void vp10_encode_token_init();
+void vp10_encode_token_init(void);
static INLINE int vp10_preserve_existing_gf(VP10_COMP *cpi) {
return !cpi->multi_arf_allowed && cpi->refresh_golden_frame &&
diff --git a/vp10/encoder/encodemb.c b/vp10/encoder/encodemb.c
index 3307393..3810be5 100644
--- a/vp10/encoder/encodemb.c
+++ b/vp10/encoder/encodemb.c
@@ -160,7 +160,7 @@
next_shortcut = shortcut;
/* Only add a trellis state for non-zero coefficients. */
- if (x) {
+ if (UNLIKELY(x)) {
error0 = tokens[next][0].error;
error1 = tokens[next][1].error;
/* Evaluate the first possibility for this state. */
@@ -204,7 +204,7 @@
rate1 = tokens[next][1].rate;
// The threshold of 3 is empirically obtained.
- if (abs(x) > 3) {
+ if (UNLIKELY(abs(x) > 3)) {
shortcut = 0;
} else {
#if CONFIG_NEW_QUANT
@@ -233,7 +233,7 @@
best_index[i][1] = best_index[i][0];
next = i;
- if (!(--band_left)) {
+ if (UNLIKELY(!(--band_left))) {
--band_counts;
band_left = *band_counts;
--token_costs;
@@ -255,7 +255,7 @@
}
if (next_shortcut) {
- if (next < default_eob) {
+ if (LIKELY(next < default_eob)) {
if (t0 != EOB_TOKEN) {
token_cache[rc] = vp10_pt_energy_class[t0];
pt = get_coef_context(nb, token_cache, i + 1);
@@ -350,7 +350,7 @@
/* Don't update next, because we didn't add a new node. */
}
- if (!(--band_left)) {
+ if (UNLIKELY(!(--band_left))) {
--band_counts;
band_left = *band_counts;
--token_costs;
diff --git a/vp10/encoder/rd.c b/vp10/encoder/rd.c
index f82e74c..9e0a339 100644
--- a/vp10/encoder/rd.c
+++ b/vp10/encoder/rd.c
@@ -582,10 +582,10 @@
}
}
-void vp10_get_entropy_contexts_plane(BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
- const struct macroblockd_plane *pd,
- ENTROPY_CONTEXT t_above[2 * MAX_MIB_SIZE],
- ENTROPY_CONTEXT t_left[2 * MAX_MIB_SIZE]) {
+static void get_entropy_contexts_plane(
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size, const struct macroblockd_plane *pd,
+ ENTROPY_CONTEXT t_above[2 * MAX_MIB_SIZE],
+ ENTROPY_CONTEXT t_left[2 * MAX_MIB_SIZE]) {
const int num_4x4_w = num_4x4_blocks_wide_lookup[plane_bsize];
const int num_4x4_h = num_4x4_blocks_high_lookup[plane_bsize];
const ENTROPY_CONTEXT *const above = pd->above_context;
@@ -626,7 +626,7 @@
ENTROPY_CONTEXT t_above[2 * MAX_MIB_SIZE],
ENTROPY_CONTEXT t_left[2 * MAX_MIB_SIZE]) {
const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
- vp10_get_entropy_contexts_plane(plane_bsize, tx_size, pd, t_above, t_left);
+ get_entropy_contexts_plane(plane_bsize, tx_size, pd, t_above, t_left);
}
void vp10_mv_pred(VP10_COMP *cpi, MACROBLOCK *x,
diff --git a/vp10/encoder/rdopt.c b/vp10/encoder/rdopt.c
index e8209ba..b96e6e4 100644
--- a/vp10/encoder/rdopt.c
+++ b/vp10/encoder/rdopt.c
@@ -604,11 +604,10 @@
(void) var[15];
}
-int adst_vs_flipadst(const VP10_COMP *cpi,
- BLOCK_SIZE bsize,
- uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- double *hdist, double *vdist) {
+static int adst_vs_flipadst(const VP10_COMP *cpi, BLOCK_SIZE bsize,
+ uint8_t *src, int src_stride,
+ uint8_t *dst, int dst_stride,
+ double *hdist, double *vdist) {
int prune_bitmask = 0;
double svm_proj_h = 0, svm_proj_v = 0;
get_energy_distribution_fine(cpi, bsize, src, src_stride,
@@ -1217,7 +1216,7 @@
sse = vpx_sum_squares_2d_i16(diff, diff_stride, bs);
#if CONFIG_VP9_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
- sse = ROUNDZ_POWER_OF_TWO(sse, (xd->bd - 8) * 2);
+ sse = ROUND_POWER_OF_TWO(sse, (xd->bd - 8) * 2);
#endif // CONFIG_VP9_HIGHBITDEPTH
sse = (int64_t)sse * 16;
@@ -3027,7 +3026,7 @@
#if CONFIG_VP9_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
- tmp = ROUNDZ_POWER_OF_TWO(tmp, (xd->bd - 8) * 2);
+ tmp = ROUND_POWER_OF_TWO(tmp, (xd->bd - 8) * 2);
#endif // CONFIG_VP9_HIGHBITDEPTH
*bsse += tmp * 16;
@@ -6664,7 +6663,7 @@
mask = vp10_get_contiguous_soft_mask(wedge_index, wedge_sign, bsize);
sse = vp10_wedge_sse_from_residuals(r1, d10, mask, N);
- sse = ROUNDZ_POWER_OF_TWO(sse, bd_round);
+ sse = ROUND_POWER_OF_TWO(sse, bd_round);
model_rd_from_sse(cpi, xd, bsize, 0, sse, &rate, &dist);
rd = RDCOST(x->rdmult, x->rddiv, rate, dist);
@@ -6726,7 +6725,7 @@
for (wedge_index = 0; wedge_index < wedge_types; ++wedge_index) {
mask = vp10_get_contiguous_soft_mask(wedge_index, wedge_sign, bsize);
sse = vp10_wedge_sse_from_residuals(r1, d10, mask, N);
- sse = ROUNDZ_POWER_OF_TWO(sse, bd_round);
+ sse = ROUND_POWER_OF_TWO(sse, bd_round);
model_rd_from_sse(cpi, xd, bsize, 0, sse, &rate, &dist);
rd = RDCOST(x->rdmult, x->rddiv, rate, dist);
@@ -7909,6 +7908,10 @@
*rate2 += vp10_cost_bit(vp10_get_skip_prob(cm, xd), 1);
*distortion = skip_sse_sb;
+ *psse = skip_sse_sb;
+ *rate_y = 0;
+ *rate_uv = 0;
+ *skippable = 1;
}
#if CONFIG_OBMC || CONFIG_WARPED_MOTION
diff --git a/vp10/vp10_common.mk b/vp10/vp10_common.mk
index 0253b4c..e68e083 100644
--- a/vp10/vp10_common.mk
+++ b/vp10/vp10_common.mk
@@ -74,6 +74,10 @@
VP10_COMMON_SRCS-yes += common/vp10_inv_txfm2d_cfg.h
VP10_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp10_convolve_ssse3.c
VP10_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp10_convolve_filters_ssse3.c
+ifeq ($(CONFIG_VP9_HIGHBITDEPTH),yes)
+VP10_COMMON_SRCS-$(HAVE_SSE4_1) += common/x86/vp10_highbd_convolve_sse4.c
+VP10_COMMON_SRCS-$(HAVE_SSE4_1) += common/x86/vp10_highbd_convolve_filters_sse4.c
+endif
VP10_COMMON_SRCS-yes += common/vp10_convolve.c
VP10_COMMON_SRCS-yes += common/vp10_convolve.h
VP10_COMMON_SRCS-$(CONFIG_ANS) += common/ans.h
diff --git a/vpx_dsp/blend.h b/vpx_dsp/blend.h
new file mode 100644
index 0000000..109183a
--- /dev/null
+++ b/vpx_dsp/blend.h
@@ -0,0 +1,40 @@
+/*
+* 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 VPX_DSP_BLEND_H_
+#define VPX_DSP_BLEND_H_
+
+#include "vpx_ports/mem.h"
+
+// Various blending functions and macros.
+// See also the vpx_blend_* functions in vpx_dsp_rtcd.h
+
+// Alpha blending with alpha values from the range [0, 64], where 64
+// means use the first input and 0 means use the second input.
+#define VPX_BLEND_A64_ROUND_BITS 6
+#define VPX_BLEND_A64_MAX_ALPHA (1 << VPX_BLEND_A64_ROUND_BITS) // 64
+
+#define VPX_BLEND_A64(a, v0, v1) \
+ ROUND_POWER_OF_TWO((a) * (v0) + (VPX_BLEND_A64_MAX_ALPHA - (a)) * (v1), \
+ VPX_BLEND_A64_ROUND_BITS)
+
+// Alpha blending with alpha values from the range [0, 256], where 256
+// means use the first input and 0 means use the second input.
+#define VPX_BLEND_A256_ROUND_BITS 8
+#define VPX_BLEND_A256_MAX_ALPHA (1 << VPX_BLEND_A256_ROUND_BITS) // 256
+
+#define VPX_BLEND_A256(a, v0, v1) \
+ ROUND_POWER_OF_TWO((a) * (v0) + (VPX_BLEND_A256_MAX_ALPHA - (a)) * (v1), \
+ VPX_BLEND_A256_ROUND_BITS)
+
+// Blending by averaging.
+#define VPX_BLEND_AVG(v0, v1) ROUND_POWER_OF_TWO((v0) + (v1), 1)
+
+#endif // VPX_DSP_BLEND_H_
diff --git a/vpx_dsp/blend_a64_hmask.c b/vpx_dsp/blend_a64_hmask.c
new file mode 100644
index 0000000..90f3415
--- /dev/null
+++ b/vpx_dsp/blend_a64_hmask.c
@@ -0,0 +1,73 @@
+/*
+ * 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 <assert.h>
+
+#include "vpx/vpx_integer.h"
+#include "vpx_ports/mem.h"
+#include "vpx_dsp/vpx_dsp_common.h"
+#include "vpx_dsp/blend.h"
+
+#include "./vpx_dsp_rtcd.h"
+
+void vpx_blend_a64_hmask_c(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ int i, j;
+
+ assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
+ assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ dst[i * dst_stride + j] = VPX_BLEND_A64(mask[j],
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+void vpx_highbd_blend_a64_hmask_c(
+ uint8_t *dst_8, uint32_t dst_stride,
+ const uint8_t *src0_8, uint32_t src0_stride,
+ const uint8_t *src1_8, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w, int bd) {
+ int i, j;
+ uint16_t *dst = CONVERT_TO_SHORTPTR(dst_8);
+ const uint16_t *src0 = CONVERT_TO_SHORTPTR(src0_8);
+ const uint16_t *src1 = CONVERT_TO_SHORTPTR(src1_8);
+
+ assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
+ assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ assert(bd == 8 || bd == 10 || bd == 12);
+
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ dst[i * dst_stride + j] = VPX_BLEND_A64(mask[j],
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/blend_a64_mask.c b/vpx_dsp/blend_a64_mask.c
new file mode 100644
index 0000000..1649798
--- /dev/null
+++ b/vpx_dsp/blend_a64_mask.c
@@ -0,0 +1,151 @@
+/*
+ * 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 <assert.h>
+
+#include "vpx/vpx_integer.h"
+#include "vpx_ports/mem.h"
+#include "vpx_dsp/blend.h"
+#include "vpx_dsp/vpx_dsp_common.h"
+
+#include "./vpx_dsp_rtcd.h"
+
+// Blending with alpha mask. Mask values come from the range [0, 64],
+// as described for VPX_BLEND_A64 in vpx_dsp/blned.h. src0 or src1 can
+// be the same as dst, or dst can be different from both sources.
+
+void vpx_blend_a64_mask_c(uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, uint32_t mask_stride,
+ int h, int w, int subh, int subw) {
+ int i, j;
+
+ assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
+ assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ if (subw == 0 && subh == 0) {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m = mask[i * mask_stride + j];
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ } else if (subw == 1 && subh == 1) {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m =
+ ROUND_POWER_OF_TWO(mask[(2 * i) * mask_stride + (2 * j)] +
+ mask[(2 * i + 1) * mask_stride + (2 * j)] +
+ mask[(2 * i) * mask_stride + (2 * j + 1)] +
+ mask[(2 * i + 1) * mask_stride + (2 * j + 1)],
+ 2);
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ } else if (subw == 1 && subh == 0) {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m = VPX_BLEND_AVG(mask[i * mask_stride + (2 * j)],
+ mask[i * mask_stride + (2 * j + 1)]);
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ } else {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m = VPX_BLEND_AVG(mask[(2 * i) * mask_stride + j],
+ mask[(2 * i + 1) * mask_stride + j]);
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ }
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+void vpx_highbd_blend_a64_mask_c(uint8_t *dst_8, uint32_t dst_stride,
+ const uint8_t *src0_8, uint32_t src0_stride,
+ const uint8_t *src1_8, uint32_t src1_stride,
+ const uint8_t *mask, uint32_t mask_stride,
+ int h, int w, int subh, int subw, int bd) {
+ int i, j;
+ uint16_t *dst = CONVERT_TO_SHORTPTR(dst_8);
+ const uint16_t *src0 = CONVERT_TO_SHORTPTR(src0_8);
+ const uint16_t *src1 = CONVERT_TO_SHORTPTR(src1_8);
+
+ assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
+ assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ assert(bd == 8 || bd == 10 || bd == 12);
+
+ if (subw == 0 && subh == 0) {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m = mask[i * mask_stride + j];
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ } else if (subw == 1 && subh == 1) {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m =
+ ROUND_POWER_OF_TWO(mask[(2 * i) * mask_stride + (2 * j)] +
+ mask[(2 * i + 1) * mask_stride + (2 * j)] +
+ mask[(2 * i) * mask_stride + (2 * j + 1)] +
+ mask[(2 * i + 1) * mask_stride + (2 * j + 1)],
+ 2);
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ } else if (subw == 1 && subh == 0) {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m = VPX_BLEND_AVG(mask[i * mask_stride + (2 * j)],
+ mask[i * mask_stride + (2 * j + 1)]);
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ } else {
+ for (i = 0; i < h; ++i) {
+ for (j = 0; j < w; ++j) {
+ const int m = VPX_BLEND_AVG(mask[(2 * i) * mask_stride + j],
+ mask[(2 * i + 1) * mask_stride + j]);
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/blend_a64_vmask.c b/vpx_dsp/blend_a64_vmask.c
new file mode 100644
index 0000000..5d48a83
--- /dev/null
+++ b/vpx_dsp/blend_a64_vmask.c
@@ -0,0 +1,75 @@
+/*
+ * 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 <assert.h>
+
+#include "vpx/vpx_integer.h"
+#include "vpx_ports/mem.h"
+#include "vpx_dsp/vpx_dsp_common.h"
+#include "vpx_dsp/blend.h"
+
+#include "./vpx_dsp_rtcd.h"
+
+void vpx_blend_a64_vmask_c(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ int i, j;
+
+ assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
+ assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ for (i = 0; i < h; ++i) {
+ const int m = mask[i];
+ for (j = 0; j < w; ++j) {
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+void vpx_highbd_blend_a64_vmask_c(
+ uint8_t *dst_8, uint32_t dst_stride,
+ const uint8_t *src0_8, uint32_t src0_stride,
+ const uint8_t *src1_8, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w, int bd) {
+ int i, j;
+ uint16_t *dst = CONVERT_TO_SHORTPTR(dst_8);
+ const uint16_t *src0 = CONVERT_TO_SHORTPTR(src0_8);
+ const uint16_t *src1 = CONVERT_TO_SHORTPTR(src1_8);
+
+ assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
+ assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ assert(bd == 8 || bd == 10 || bd == 12);
+
+ for (i = 0; i < h; ++i) {
+ const int m = mask[i];
+ for (j = 0; j < w; ++j) {
+ dst[i * dst_stride + j] = VPX_BLEND_A64(m,
+ src0[i * src0_stride + j],
+ src1[i * src1_stride + j]);
+ }
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/blend_mask6.c b/vpx_dsp/blend_mask6.c
deleted file mode 100644
index 584ee6a..0000000
--- a/vpx_dsp/blend_mask6.c
+++ /dev/null
@@ -1,152 +0,0 @@
-/*
- * 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 <assert.h>
-
-#include "vpx/vpx_integer.h"
-#include "vpx_ports/mem.h"
-#include "vpx_dsp/vpx_dsp_common.h"
-
-#include "./vpx_dsp_rtcd.h"
-
-#define MASK_BITS 6
-
-void vpx_blend_mask6_c(uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
- const uint8_t *mask, uint32_t mask_stride,
- int h, int w, int subh, int subw) {
- int i, j;
-
- assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
- assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
- assert(h >= 4);
- assert(w >= 4);
- assert(IS_POWER_OF_TWO(h));
- assert(IS_POWER_OF_TWO(w));
-
- if (subw == 0 && subh == 0) {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 = mask[i * mask_stride + j];
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- } else if (subw == 1 && subh == 1) {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 =
- ROUND_POWER_OF_TWO(mask[(2 * i) * mask_stride + (2 * j)] +
- mask[(2 * i + 1) * mask_stride + (2 * j)] +
- mask[(2 * i) * mask_stride + (2 * j + 1)] +
- mask[(2 * i + 1) * mask_stride + (2 * j + 1)],
- 2);
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- } else if (subw == 1 && subh == 0) {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 =
- ROUND_POWER_OF_TWO(mask[i * mask_stride + (2 * j)] +
- mask[i * mask_stride + (2 * j + 1)], 1);
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- } else {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 =
- ROUND_POWER_OF_TWO(mask[(2 * i) * mask_stride + j] +
- mask[(2 * i + 1) * mask_stride + j], 1);
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- }
-}
-
-#if CONFIG_VP9_HIGHBITDEPTH
-void vpx_highbd_blend_mask6_c(uint8_t *dst_8, uint32_t dst_stride,
- uint8_t *src0_8, uint32_t src0_stride,
- uint8_t *src1_8, uint32_t src1_stride,
- const uint8_t *mask, uint32_t mask_stride,
- int h, int w, int subh, int subw, int bd) {
- int i, j;
- uint16_t *dst = CONVERT_TO_SHORTPTR(dst_8);
- uint16_t *src0 = CONVERT_TO_SHORTPTR(src0_8);
- uint16_t *src1 = CONVERT_TO_SHORTPTR(src1_8);
-
- assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
- assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
- assert(h >= 4);
- assert(w >= 4);
- assert(IS_POWER_OF_TWO(h));
- assert(IS_POWER_OF_TWO(w));
-
- assert(bd == 8 || bd == 10 || bd == 12);
-
- if (subw == 0 && subh == 0) {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 = mask[i * mask_stride + j];
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- } else if (subw == 1 && subh == 1) {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 =
- ROUND_POWER_OF_TWO(mask[(2 * i) * mask_stride + (2 * j)] +
- mask[(2 * i + 1) * mask_stride + (2 * j)] +
- mask[(2 * i) * mask_stride + (2 * j + 1)] +
- mask[(2 * i + 1) * mask_stride + (2 * j + 1)],
- 2);
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- } else if (subw == 1 && subh == 0) {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 =
- ROUND_POWER_OF_TWO(mask[i * mask_stride + (2 * j)] +
- mask[i * mask_stride + (2 * j + 1)], 1);
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- } else {
- for (i = 0; i < h; ++i)
- for (j = 0; j < w; ++j) {
- const int m0 =
- ROUND_POWER_OF_TWO(mask[(2 * i) * mask_stride + j] +
- mask[(2 * i + 1) * mask_stride + j], 1);
- const int m1 = ((1 << MASK_BITS) - m0);
- dst[i * dst_stride + j] =
- ROUND_POWER_OF_TWO(src0[i * src0_stride + j] * m0 +
- src1[i * src1_stride + j] * m1, MASK_BITS);
- }
- }
-}
-#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/sad.c b/vpx_dsp/sad.c
index bb1daf8..e64dae3 100644
--- a/vpx_dsp/sad.c
+++ b/vpx_dsp/sad.c
@@ -452,23 +452,23 @@
#endif // CONFIG_VP10 && CONFIG_EXT_INTER
#if CONFIG_VP10 && CONFIG_OBMC
-// a: pred
-// b: target weighted prediction (has been *4096 to keep precision)
-// m: 2d weights (scaled by 4096)
-static INLINE unsigned int obmc_sad(const uint8_t *a, int a_stride,
- const int32_t *b,
- const int32_t *m,
+// pre: predictor being evaluated
+// wsrc: target weighted prediction (has been *4096 to keep precision)
+// mask: 2d weights (scaled by 4096)
+static INLINE unsigned int obmc_sad(const uint8_t *pre, int pre_stride,
+ const int32_t *wsrc,
+ const int32_t *mask,
int width, int height) {
int y, x;
unsigned int sad = 0;
for (y = 0; y < height; y++) {
for (x = 0; x < width; x++)
- sad += ROUND_POWER_OF_TWO(abs(b[x] - a[x] * m[x]), 12);
+ sad += ROUND_POWER_OF_TWO(abs(wsrc[x] - pre[x] * mask[x]), 12);
- a += a_stride;
- b += width;
- m += width;
+ pre += pre_stride;
+ wsrc += width;
+ mask += width;
}
return sad;
@@ -477,8 +477,8 @@
#define OBMCSADMxN(m, n) \
unsigned int vpx_obmc_sad##m##x##n##_c(const uint8_t *ref, int ref_stride, \
const int32_t *wsrc, \
- const int32_t *msk) { \
- return obmc_sad(ref, ref_stride, wsrc, msk, m, n); \
+ const int32_t *mask) { \
+ return obmc_sad(ref, ref_stride, wsrc, mask, m, n); \
}
#if CONFIG_EXT_PARTITION
@@ -501,21 +501,21 @@
OBMCSADMxN(4, 4)
#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE unsigned int highbd_obmc_sad(const uint8_t *a8, int a_stride,
- const int32_t *b,
- const int32_t *m,
+static INLINE unsigned int highbd_obmc_sad(const uint8_t *pre8, int pre_stride,
+ const int32_t *wsrc,
+ const int32_t *mask,
int width, int height) {
int y, x;
unsigned int sad = 0;
- const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
+ const uint16_t *pre = CONVERT_TO_SHORTPTR(pre8);
for (y = 0; y < height; y++) {
for (x = 0; x < width; x++)
- sad += ROUND_POWER_OF_TWO(abs(b[x] - a[x] * m[x]), 12);
+ sad += ROUND_POWER_OF_TWO(abs(wsrc[x] - pre[x] * mask[x]), 12);
- a += a_stride;
- b += width;
- m += width;
+ pre += pre_stride;
+ wsrc += width;
+ mask += width;
}
return sad;
@@ -525,8 +525,8 @@
unsigned int vpx_highbd_obmc_sad##m##x##n##_c(const uint8_t *ref, \
int ref_stride, \
const int32_t *wsrc, \
- const int32_t *msk) { \
- return highbd_obmc_sad(ref, ref_stride, wsrc, msk, m, n); \
+ const int32_t *mask) { \
+ return highbd_obmc_sad(ref, ref_stride, wsrc, mask, m, n); \
}
#if CONFIG_EXT_PARTITION
diff --git a/vpx_dsp/vpx_dsp.mk b/vpx_dsp/vpx_dsp.mk
index 3c519b6..3eb7a9f 100644
--- a/vpx_dsp/vpx_dsp.mk
+++ b/vpx_dsp/vpx_dsp.mk
@@ -70,10 +70,14 @@
# inter predictions
ifeq ($(CONFIG_VP10),yes)
-ifeq ($(CONFIG_EXT_INTER),yes)
-DSP_SRCS-yes += blend_mask6.c
-DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_mask6_sse4.c
-endif #CONFIG_EXT_INTER
+DSP_SRCS-yes += blend.h
+DSP_SRCS-yes += blend_a64_mask.c
+DSP_SRCS-yes += blend_a64_hmask.c
+DSP_SRCS-yes += blend_a64_vmask.c
+DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_sse4.h
+DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_a64_mask_sse4.c
+DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_a64_hmask_sse4.c
+DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_a64_vmask_sse4.c
endif #CONFIG_VP10
# interpolation filters
diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl
index e630994..d8055e9 100644
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -959,6 +959,27 @@
} # CONFIG_VP9_HIGHBITDEPTH
} # CONFIG_VP9_ENCODER || CONFIG_VP10_ENCODER
+if (vpx_config("CONFIG_VP10") eq "yes") {
+ #
+ # Alpha blending with mask
+ #
+ add_proto qw/void vpx_blend_a64_mask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx";
+ add_proto qw/void vpx_blend_a64_hmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w";
+ add_proto qw/void vpx_blend_a64_vmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w";
+ specialize "vpx_blend_a64_mask", qw/sse4_1/;
+ specialize "vpx_blend_a64_hmask", qw/sse4_1/;
+ specialize "vpx_blend_a64_vmask", qw/sse4_1/;
+
+ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/void vpx_highbd_blend_a64_mask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx, int bd";
+ add_proto qw/void vpx_highbd_blend_a64_hmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w, int bd";
+ add_proto qw/void vpx_highbd_blend_a64_vmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w, int bd";
+ specialize "vpx_highbd_blend_a64_mask", qw/sse4_1/;
+ specialize "vpx_highbd_blend_a64_hmask", qw/sse4_1/;
+ specialize "vpx_highbd_blend_a64_vmask", qw/sse4_1/;
+ }
+} # CONFIG_VP10
+
if (vpx_config("CONFIG_ENCODERS") eq "yes") {
#
# Block subtraction
@@ -1103,14 +1124,14 @@
if (vpx_config("CONFIG_OBMC") eq "yes") {
foreach (@block_sizes) {
($w, $h) = @$_;
- add_proto qw/unsigned int/, "vpx_obmc_sad${w}x${h}", "const uint8_t *ref_ptr, int ref_stride, const int32_t *wsrc_ptr, const int32_t *mask";
+ add_proto qw/unsigned int/, "vpx_obmc_sad${w}x${h}", "const uint8_t *pre, int pre_stride, const int32_t *wsrc, const int32_t *mask";
specialize "vpx_obmc_sad${w}x${h}", qw/sse4_1/;
}
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
foreach (@block_sizes) {
($w, $h) = @$_;
- add_proto qw/unsigned int/, "vpx_highbd_obmc_sad${w}x${h}", "const uint8_t *ref_ptr, int ref_stride, const int32_t *wsrc_ptr, const int32_t *mask";
+ add_proto qw/unsigned int/, "vpx_highbd_obmc_sad${w}x${h}", "const uint8_t *pre, int pre_stride, const int32_t *wsrc, const int32_t *mask";
specialize "vpx_highbd_obmc_sad${w}x${h}", qw/sse4_1/;
}
}
@@ -1384,14 +1405,6 @@
}
}
}
-
- add_proto qw/void vpx_blend_mask6/, "uint8_t *dst, uint32_t dst_stride, uint8_t *src0, uint32_t src0_stride, uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx";
- specialize "vpx_blend_mask6", qw/sse4_1/;
-
- if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
- add_proto qw/void vpx_highbd_blend_mask6/, "uint8_t *dst, uint32_t dst_stride, uint8_t *src0, uint32_t src0_stride, uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx, int bd";
- specialize "vpx_highbd_blend_mask6", qw/sse4_1/;
- }
}
#
diff --git a/vpx_dsp/x86/blend_a64_hmask_sse4.c b/vpx_dsp/x86/blend_a64_hmask_sse4.c
new file mode 100644
index 0000000..a10e077
--- /dev/null
+++ b/vpx_dsp/x86/blend_a64_hmask_sse4.c
@@ -0,0 +1,41 @@
+/*
+ * 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 "vpx/vpx_integer.h"
+
+#include "./vpx_dsp_rtcd.h"
+
+// To start out, just dispatch to the function using the 2D mask and
+// pass mask stride as 0. This can be improved upon if necessary.
+
+void vpx_blend_a64_hmask_sse4_1(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ vpx_blend_a64_mask_sse4_1(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, 0, h, w, 0, 0);
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+void vpx_highbd_blend_a64_hmask_sse4_1(
+ uint8_t *dst_8, uint32_t dst_stride,
+ const uint8_t *src0_8, uint32_t src0_stride,
+ const uint8_t *src1_8, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w,
+ int bd) {
+ vpx_highbd_blend_a64_mask_sse4_1(dst_8, dst_stride,
+ src0_8, src0_stride,
+ src1_8, src1_stride,
+ mask, 0, h, w, 0, 0, bd);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/x86/blend_mask6_sse4.c b/vpx_dsp/x86/blend_a64_mask_sse4.c
similarity index 61%
rename from vpx_dsp/x86/blend_mask6_sse4.c
rename to vpx_dsp/x86/blend_a64_mask_sse4.c
index 28693a4..cdb40c2 100644
--- a/vpx_dsp/x86/blend_mask6_sse4.c
+++ b/vpx_dsp/x86/blend_a64_mask_sse4.c
@@ -15,62 +15,24 @@
#include "vpx/vpx_integer.h"
#include "vpx_ports/mem.h"
#include "vpx_dsp/vpx_dsp_common.h"
+#include "vpx_dsp/blend.h"
#include "vpx_dsp/x86/synonyms.h"
+#include "vpx_dsp/x86/blend_sse4.h"
#include "./vpx_dsp_rtcd.h"
-#define MASK_BITS 6
-
-//////////////////////////////////////////////////////////////////////////////
-// Common kernels
-//////////////////////////////////////////////////////////////////////////////
-
-static INLINE __m128i blend_4(uint8_t*src0, uint8_t *src1,
- const __m128i v_m0_w, const __m128i v_m1_w) {
- const __m128i v_s0_b = xx_loadl_32(src0);
- const __m128i v_s1_b = xx_loadl_32(src1);
- const __m128i v_s0_w = _mm_cvtepu8_epi16(v_s0_b);
- const __m128i v_s1_w = _mm_cvtepu8_epi16(v_s1_b);
-
- const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
- const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
-
- const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
-
- const __m128i v_res_w = xx_roundn_epu16(v_sum_w, MASK_BITS);
-
- return v_res_w;
-}
-
-static INLINE __m128i blend_8(uint8_t*src0, uint8_t *src1,
- const __m128i v_m0_w, const __m128i v_m1_w) {
- const __m128i v_s0_b = xx_loadl_64(src0);
- const __m128i v_s1_b = xx_loadl_64(src1);
- const __m128i v_s0_w = _mm_cvtepu8_epi16(v_s0_b);
- const __m128i v_s1_w = _mm_cvtepu8_epi16(v_s1_b);
-
- const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
- const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
-
- const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
-
- const __m128i v_res_w = xx_roundn_epu16(v_sum_w, MASK_BITS);
-
- return v_res_w;
-}
-
//////////////////////////////////////////////////////////////////////////////
// No sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static void blend_mask6_w4_sse4_1(
+static void blend_a64_mask_w4_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -92,13 +54,13 @@
} while (--h);
}
-static void blend_mask6_w8_sse4_1(
+static void blend_a64_mask_w8_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -120,13 +82,13 @@
} while (--h);
}
-static void blend_mask6_w16n_sse4_1(
+static void blend_a64_mask_w16n_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -158,15 +120,15 @@
// Horizontal sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static void blend_mask6_sx_w4_sse4_1(
+static void blend_a64_mask_sx_w4_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -190,15 +152,15 @@
} while (--h);
}
-static void blend_mask6_sx_w8_sse4_1(
+static void blend_a64_mask_sx_w8_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -222,15 +184,15 @@
} while (--h);
}
-static void blend_mask6_sx_w16n_sse4_1(
+static void blend_a64_mask_sx_w16n_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -265,13 +227,13 @@
// Vertical sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static void blend_mask6_sy_w4_sse4_1(
+static void blend_a64_mask_sy_w4_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -296,13 +258,13 @@
} while (--h);
}
-static void blend_mask6_sy_w8_sse4_1(
+static void blend_a64_mask_sy_w8_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -327,14 +289,14 @@
} while (--h);
}
-static void blend_mask6_sy_w16n_sse4_1(
+static void blend_a64_mask_sy_w16n_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
const __m128i v_zero = _mm_setzero_si128();
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -368,15 +330,15 @@
// Horizontal and Vertical sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static void blend_mask6_sx_sy_w4_sse4_1(
+static void blend_a64_mask_sx_sy_w4_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -405,15 +367,15 @@
} while (--h);
}
-static void blend_mask6_sx_sy_w8_sse4_1(
+static void blend_a64_mask_sx_sy_w8_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
(void)w;
@@ -442,15 +404,15 @@
} while (--h);
}
-static void blend_mask6_sx_sy_w16n_sse4_1(
+static void blend_a64_mask_sx_sy_w16n_sse4_1(
uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -495,146 +457,67 @@
// Dispatch
//////////////////////////////////////////////////////////////////////////////
-void vpx_blend_mask6_sse4_1(uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
- const uint8_t *mask, uint32_t mask_stride,
- int h, int w, int suby, int subx) {
+void vpx_blend_a64_mask_sse4_1(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, uint32_t mask_stride,
+ int h, int w, int suby, int subx) {
typedef void (*blend_fn)(uint8_t *dst, uint32_t dst_stride,
- uint8_t *src0, uint32_t src0_stride,
- uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w);
- static blend_fn blend[3][2][2] = { // width_index X subx X suby
+ // Dimensions are: width_index X subx X suby
+ static const blend_fn blend[3][2][2] = {
{ // w % 16 == 0
- {blend_mask6_w16n_sse4_1, blend_mask6_sy_w16n_sse4_1},
- {blend_mask6_sx_w16n_sse4_1, blend_mask6_sx_sy_w16n_sse4_1}
+ {blend_a64_mask_w16n_sse4_1, blend_a64_mask_sy_w16n_sse4_1},
+ {blend_a64_mask_sx_w16n_sse4_1, blend_a64_mask_sx_sy_w16n_sse4_1}
}, { // w == 4
- {blend_mask6_w4_sse4_1, blend_mask6_sy_w4_sse4_1},
- {blend_mask6_sx_w4_sse4_1, blend_mask6_sx_sy_w4_sse4_1}
+ {blend_a64_mask_w4_sse4_1, blend_a64_mask_sy_w4_sse4_1},
+ {blend_a64_mask_sx_w4_sse4_1, blend_a64_mask_sx_sy_w4_sse4_1}
}, { // w == 8
- {blend_mask6_w8_sse4_1, blend_mask6_sy_w8_sse4_1},
- {blend_mask6_sx_w8_sse4_1, blend_mask6_sx_sy_w8_sse4_1}
+ {blend_a64_mask_w8_sse4_1, blend_a64_mask_sy_w8_sse4_1},
+ {blend_a64_mask_sx_w8_sse4_1, blend_a64_mask_sx_sy_w8_sse4_1}
}
};
assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
- assert(h >= 4);
- assert(w >= 4);
+ assert(h >= 1);
+ assert(w >= 1);
assert(IS_POWER_OF_TWO(h));
assert(IS_POWER_OF_TWO(w));
- blend[(w >> 2) & 3][subx != 0][suby != 0](dst, dst_stride,
- src0, src0_stride,
- src1, src1_stride,
- mask, mask_stride,
- h, w);
+ if (UNLIKELY((h | w) & 3)) { // if (w <= 2 || h <= 2)
+ vpx_blend_a64_mask_c(dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, mask_stride,
+ h, w, suby, subx);
+ } else {
+ blend[(w >> 2) & 3][subx != 0][suby != 0](dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, mask_stride,
+ h, w);
+ }
}
#if CONFIG_VP9_HIGHBITDEPTH
//////////////////////////////////////////////////////////////////////////////
-// Common kernels
-//////////////////////////////////////////////////////////////////////////////
-
-typedef __m128i (*blend_unit_fn)(uint16_t*src0, uint16_t *src1,
- const __m128i v_m0_w, const __m128i v_m1_w);
-
-static INLINE __m128i blend_4_b10(uint16_t*src0, uint16_t *src1,
- const __m128i v_m0_w, const __m128i v_m1_w) {
- const __m128i v_s0_w = xx_loadl_64(src0);
- const __m128i v_s1_w = xx_loadl_64(src1);
-
- const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
- const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
-
- const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
-
- const __m128i v_res_w = xx_roundn_epu16(v_sum_w, MASK_BITS);
-
- return v_res_w;
-}
-
-static INLINE __m128i blend_8_b10(uint16_t*src0, uint16_t *src1,
- const __m128i v_m0_w, const __m128i v_m1_w) {
- const __m128i v_s0_w = xx_loadu_128(src0);
- const __m128i v_s1_w = xx_loadu_128(src1);
-
- const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
- const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
-
- const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
-
- const __m128i v_res_w = xx_roundn_epu16(v_sum_w, MASK_BITS);
-
- return v_res_w;
-}
-
-static INLINE __m128i blend_4_b12(uint16_t*src0, uint16_t *src1,
- const __m128i v_m0_w, const __m128i v_m1_w) {
- const __m128i v_s0_w = xx_loadl_64(src0);
- const __m128i v_s1_w = xx_loadl_64(src1);
-
- // Interleave
- const __m128i v_m01_w = _mm_unpacklo_epi16(v_m0_w, v_m1_w);
- const __m128i v_s01_w = _mm_unpacklo_epi16(v_s0_w, v_s1_w);
-
- // Multiply-Add
- const __m128i v_sum_d = _mm_madd_epi16(v_s01_w, v_m01_w);
-
- // Scale
- const __m128i v_ssum_d = _mm_srli_epi32(v_sum_d, MASK_BITS - 1);
-
- // Pack
- const __m128i v_pssum_d = _mm_packs_epi32(v_ssum_d, v_ssum_d);
-
- // Round
- const __m128i v_res_w = xx_round_epu16(v_pssum_d);
-
- return v_res_w;
-}
-
-static INLINE __m128i blend_8_b12(uint16_t*src0, uint16_t *src1,
- const __m128i v_m0_w, const __m128i v_m1_w) {
- const __m128i v_s0_w = xx_loadu_128(src0);
- const __m128i v_s1_w = xx_loadu_128(src1);
-
- // Interleave
- const __m128i v_m01l_w = _mm_unpacklo_epi16(v_m0_w, v_m1_w);
- const __m128i v_m01h_w = _mm_unpackhi_epi16(v_m0_w, v_m1_w);
- const __m128i v_s01l_w = _mm_unpacklo_epi16(v_s0_w, v_s1_w);
- const __m128i v_s01h_w = _mm_unpackhi_epi16(v_s0_w, v_s1_w);
-
- // Multiply-Add
- const __m128i v_suml_d = _mm_madd_epi16(v_s01l_w, v_m01l_w);
- const __m128i v_sumh_d = _mm_madd_epi16(v_s01h_w, v_m01h_w);
-
- // Scale
- const __m128i v_ssuml_d = _mm_srli_epi32(v_suml_d, MASK_BITS - 1);
- const __m128i v_ssumh_d = _mm_srli_epi32(v_sumh_d, MASK_BITS - 1);
-
- // Pack
- const __m128i v_pssum_d = _mm_packs_epi32(v_ssuml_d, v_ssumh_d);
-
- // Round
- const __m128i v_res_w = xx_round_epu16(v_pssum_d);
-
- return v_res_w;
-}
-
-//////////////////////////////////////////////////////////////////////////////
// No sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static INLINE void blend_mask6_bn_w4_sse4_1(
+static INLINE void blend_a64_mask_bn_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, blend_unit_fn blend) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
const __m128i v_m0_b = xx_loadl_32(mask);
@@ -652,37 +535,37 @@
} while (--h);
}
-static void blend_mask6_b10_w4_sse4_1(
+static void blend_a64_mask_b10_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b10);
+ blend_a64_mask_bn_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b10);
}
-static void blend_mask6_b12_w4_sse4_1(
+static void blend_a64_mask_b12_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b12);
+ blend_a64_mask_bn_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b12);
}
-static inline void blend_mask6_bn_w8n_sse4_1(
+static inline void blend_a64_mask_bn_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w, blend_unit_fn blend) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -702,41 +585,41 @@
} while (--h);
}
-static void blend_mask6_b10_w8n_sse4_1(
+static void blend_a64_mask_b10_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b10);
+ blend_a64_mask_bn_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b10);
}
-static void blend_mask6_b12_w8n_sse4_1(
+static void blend_a64_mask_b12_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b12);
+ blend_a64_mask_bn_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b12);
}
//////////////////////////////////////////////////////////////////////////////
// Horizontal sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static INLINE void blend_mask6_bn_sx_w4_sse4_1(
+static INLINE void blend_a64_mask_bn_sx_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, blend_unit_fn blend) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
const __m128i v_r_b = xx_loadl_64(mask);
@@ -756,39 +639,39 @@
} while (--h);
}
-static void blend_mask6_b10_sx_w4_sse4_1(
+static void blend_a64_mask_b10_sx_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_sx_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b10);
+ blend_a64_mask_bn_sx_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b10);
}
-static void blend_mask6_b12_sx_w4_sse4_1(
+static void blend_a64_mask_b12_sx_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_sx_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b12);
+ blend_a64_mask_bn_sx_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b12);
}
-static INLINE void blend_mask6_bn_sx_w8n_sse4_1(
+static INLINE void blend_a64_mask_bn_sx_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w, blend_unit_fn blend) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -810,39 +693,39 @@
} while (--h);
}
-static void blend_mask6_b10_sx_w8n_sse4_1(
+static void blend_a64_mask_b10_sx_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_sx_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b10);
+ blend_a64_mask_bn_sx_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b10);
}
-static void blend_mask6_b12_sx_w8n_sse4_1(
+static void blend_a64_mask_b12_sx_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_sx_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b12);
+ blend_a64_mask_bn_sx_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b12);
}
//////////////////////////////////////////////////////////////////////////////
// Vertical sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static INLINE void blend_mask6_bn_sy_w4_sse4_1(
+static INLINE void blend_a64_mask_bn_sy_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, blend_unit_fn blend) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
const __m128i v_ra_b = xx_loadl_32(mask);
@@ -863,37 +746,37 @@
} while (--h);
}
-static void blend_mask6_b10_sy_w4_sse4_1(
+static void blend_a64_mask_b10_sy_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b10);
+ blend_a64_mask_bn_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b10);
}
-static void blend_mask6_b12_sy_w4_sse4_1(
+static void blend_a64_mask_b12_sy_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b12);
+ blend_a64_mask_bn_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b12);
}
-static INLINE void blend_mask6_bn_sy_w8n_sse4_1(
+static INLINE void blend_a64_mask_bn_sy_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w, blend_unit_fn blend) {
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -916,41 +799,41 @@
} while (--h);
}
-static void blend_mask6_b10_sy_w8n_sse4_1(
+static void blend_a64_mask_b10_sy_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b10);
+ blend_a64_mask_bn_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b10);
}
-static void blend_mask6_b12_sy_w8n_sse4_1(
+static void blend_a64_mask_b12_sy_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b12);
+ blend_a64_mask_bn_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b12);
}
//////////////////////////////////////////////////////////////////////////////
// Horizontal and Vertical sub-sampling
//////////////////////////////////////////////////////////////////////////////
-static INLINE void blend_mask6_bn_sx_sy_w4_sse4_1(
+static INLINE void blend_a64_mask_bn_sx_sy_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, blend_unit_fn blend) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
const __m128i v_ra_b = xx_loadl_64(mask);
@@ -975,39 +858,39 @@
} while (--h);
}
-static void blend_mask6_b10_sx_sy_w4_sse4_1(
+static void blend_a64_mask_b10_sx_sy_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_sx_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b10);
+ blend_a64_mask_bn_sx_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b10);
}
-static void blend_mask6_b12_sx_sy_w4_sse4_1(
+static void blend_a64_mask_b12_sx_sy_w4_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
(void)w;
- blend_mask6_bn_sx_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h,
- blend_4_b12);
+ blend_a64_mask_bn_sx_sy_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h,
+ blend_4_b12);
}
-static INLINE void blend_mask6_bn_sx_sy_w8n_sse4_1(
+static INLINE void blend_a64_mask_bn_sx_sy_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w, blend_unit_fn blend) {
const __m128i v_zmask_b = _mm_set_epi8(0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff,
0, 0xff, 0, 0xff, 0, 0xff, 0, 0xff);
- const __m128i v_maxval_w = _mm_set1_epi16(1 << MASK_BITS);
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
do {
int c;
@@ -1034,82 +917,91 @@
} while (--h);
}
-static void blend_mask6_b10_sx_sy_w8n_sse4_1(
+static void blend_a64_mask_b10_sx_sy_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_sx_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b10);
+ blend_a64_mask_bn_sx_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b10);
}
-static void blend_mask6_b12_sx_sy_w8n_sse4_1(
+static void blend_a64_mask_b12_sx_sy_w8n_sse4_1(
uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w) {
- blend_mask6_bn_sx_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
- src1_stride, mask, mask_stride, h, w,
- blend_8_b12);
+ blend_a64_mask_bn_sx_sy_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, mask_stride, h, w,
+ blend_8_b12);
}
//////////////////////////////////////////////////////////////////////////////
// Dispatch
//////////////////////////////////////////////////////////////////////////////
-void vpx_highbd_blend_mask6_sse4_1(uint8_t *dst_8, uint32_t dst_stride,
- uint8_t *src0_8, uint32_t src0_stride,
- uint8_t *src1_8, uint32_t src1_stride,
- const uint8_t *mask, uint32_t mask_stride,
- int h, int w, int suby, int subx, int bd) {
- uint16_t *const dst = CONVERT_TO_SHORTPTR(dst_8);
- uint16_t *const src0 = CONVERT_TO_SHORTPTR(src0_8);
- uint16_t *const src1 = CONVERT_TO_SHORTPTR(src1_8);
-
+void vpx_highbd_blend_a64_mask_sse4_1(
+ uint8_t *dst_8, uint32_t dst_stride,
+ const uint8_t *src0_8, uint32_t src0_stride,
+ const uint8_t *src1_8, uint32_t src1_stride,
+ const uint8_t *mask, uint32_t mask_stride,
+ int h, int w, int suby, int subx, int bd) {
typedef void (*blend_fn)(uint16_t *dst, uint32_t dst_stride,
- uint16_t *src0, uint32_t src0_stride,
- uint16_t *src1, uint32_t src1_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
const uint8_t *mask, uint32_t mask_stride,
int h, int w);
- static blend_fn blend[2][2][2][2] = { // bd_index X width_index X subx X suby
+ // Dimensions are: bd_index X width_index X subx X suby
+ static const blend_fn blend[2][2][2][2] = {
{ // bd == 8 or 10
{ // w % 8 == 0
- {blend_mask6_b10_w8n_sse4_1, blend_mask6_b10_sy_w8n_sse4_1},
- {blend_mask6_b10_sx_w8n_sse4_1, blend_mask6_b10_sx_sy_w8n_sse4_1}
+ {blend_a64_mask_b10_w8n_sse4_1, blend_a64_mask_b10_sy_w8n_sse4_1},
+ {blend_a64_mask_b10_sx_w8n_sse4_1, blend_a64_mask_b10_sx_sy_w8n_sse4_1}
}, { // w == 4
- {blend_mask6_b10_w4_sse4_1, blend_mask6_b10_sy_w4_sse4_1},
- {blend_mask6_b10_sx_w4_sse4_1, blend_mask6_b10_sx_sy_w4_sse4_1}
+ {blend_a64_mask_b10_w4_sse4_1, blend_a64_mask_b10_sy_w4_sse4_1},
+ {blend_a64_mask_b10_sx_w4_sse4_1, blend_a64_mask_b10_sx_sy_w4_sse4_1}
}
},
{ // bd == 12
{ // w % 8 == 0
- {blend_mask6_b12_w8n_sse4_1, blend_mask6_b12_sy_w8n_sse4_1},
- {blend_mask6_b12_sx_w8n_sse4_1, blend_mask6_b12_sx_sy_w8n_sse4_1}
+ {blend_a64_mask_b12_w8n_sse4_1, blend_a64_mask_b12_sy_w8n_sse4_1},
+ {blend_a64_mask_b12_sx_w8n_sse4_1, blend_a64_mask_b12_sx_sy_w8n_sse4_1}
}, { // w == 4
- {blend_mask6_b12_w4_sse4_1, blend_mask6_b12_sy_w4_sse4_1},
- {blend_mask6_b12_sx_w4_sse4_1, blend_mask6_b12_sx_sy_w4_sse4_1}
+ {blend_a64_mask_b12_w4_sse4_1, blend_a64_mask_b12_sy_w4_sse4_1},
+ {blend_a64_mask_b12_sx_w4_sse4_1, blend_a64_mask_b12_sx_sy_w4_sse4_1}
}
}
};
- assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
- assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+ assert(IMPLIES(src0_8 == dst_8, src0_stride == dst_stride));
+ assert(IMPLIES(src1_8 == dst_8, src1_stride == dst_stride));
- assert(h >= 4);
- assert(w >= 4);
+ assert(h >= 1);
+ assert(w >= 1);
assert(IS_POWER_OF_TWO(h));
assert(IS_POWER_OF_TWO(w));
assert(bd == 8 || bd == 10 || bd == 12);
+ if (UNLIKELY((h | w) & 3)) { // if (w <= 2 || h <= 2)
+ vpx_highbd_blend_a64_mask_c(dst_8, dst_stride,
+ src0_8, src0_stride,
+ src1_8, src1_stride,
+ mask, mask_stride,
+ h, w, suby, subx, bd);
+ } else {
+ uint16_t *const dst = CONVERT_TO_SHORTPTR(dst_8);
+ const uint16_t *const src0 = CONVERT_TO_SHORTPTR(src0_8);
+ const uint16_t *const src1 = CONVERT_TO_SHORTPTR(src1_8);
- blend[bd == 12][(w >> 2) & 1][subx != 0][suby != 0](dst, dst_stride,
- src0, src0_stride,
- src1, src1_stride,
- mask, mask_stride,
- h, w);
+ blend[bd == 12][(w >> 2) & 1][subx != 0][suby != 0](dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, mask_stride,
+ h, w);
+ }
}
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/x86/blend_a64_vmask_sse4.c b/vpx_dsp/x86/blend_a64_vmask_sse4.c
new file mode 100644
index 0000000..4b0f38d
--- /dev/null
+++ b/vpx_dsp/x86/blend_a64_vmask_sse4.c
@@ -0,0 +1,293 @@
+/*
+ * 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 <smmintrin.h> // SSE4.1
+
+#include <assert.h>
+
+#include "vpx/vpx_integer.h"
+#include "vpx_ports/mem.h"
+#include "vpx_dsp/vpx_dsp_common.h"
+#include "vpx_dsp/blend.h"
+
+#include "vpx_dsp/x86/synonyms.h"
+#include "vpx_dsp/x86/blend_sse4.h"
+
+#include "./vpx_dsp_rtcd.h"
+
+//////////////////////////////////////////////////////////////////////////////
+// Implementation - No sub-sampling
+//////////////////////////////////////////////////////////////////////////////
+
+static void blend_a64_vmask_w4_sse4_1(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
+
+ (void)w;
+
+ do {
+ const __m128i v_m0_w = _mm_set1_epi16(*mask);
+ const __m128i v_m1_w = _mm_sub_epi16(v_maxval_w, v_m0_w);
+
+ const __m128i v_res_w = blend_4(src0, src1, v_m0_w, v_m1_w);
+
+ const __m128i v_res_b = _mm_packus_epi16(v_res_w, v_res_w);
+
+ xx_storel_32(dst, v_res_b);
+
+ dst += dst_stride;
+ src0 += src0_stride;
+ src1 += src1_stride;
+ mask += 1;
+ } while (--h);
+}
+
+static void blend_a64_vmask_w8_sse4_1(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
+
+ (void)w;
+
+ do {
+ const __m128i v_m0_w = _mm_set1_epi16(*mask);
+ const __m128i v_m1_w = _mm_sub_epi16(v_maxval_w, v_m0_w);
+
+ const __m128i v_res_w = blend_8(src0, src1, v_m0_w, v_m1_w);
+
+ const __m128i v_res_b = _mm_packus_epi16(v_res_w, v_res_w);
+
+ xx_storel_64(dst, v_res_b);
+
+ dst += dst_stride;
+ src0 += src0_stride;
+ src1 += src1_stride;
+ mask += 1;
+ } while (--h);
+}
+
+static void blend_a64_vmask_w16n_sse4_1(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
+
+ do {
+ int c;
+ const __m128i v_m0_w = _mm_set1_epi16(*mask);
+ const __m128i v_m1_w = _mm_sub_epi16(v_maxval_w, v_m0_w);
+ for (c = 0; c < w; c += 16) {
+ const __m128i v_resl_w = blend_8(src0 + c, src1 + c,
+ v_m0_w, v_m1_w);
+ const __m128i v_resh_w = blend_8(src0 + c + 8, src1 + c + 8,
+ v_m0_w, v_m1_w);
+
+ const __m128i v_res_b = _mm_packus_epi16(v_resl_w, v_resh_w);
+
+ xx_storeu_128(dst + c, v_res_b);
+ }
+ dst += dst_stride;
+ src0 += src0_stride;
+ src1 += src1_stride;
+ mask += 1;
+ } while (--h);
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// Dispatch
+//////////////////////////////////////////////////////////////////////////////
+
+void vpx_blend_a64_vmask_sse4_1(
+ uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ typedef void (*blend_fn)(uint8_t *dst, uint32_t dst_stride,
+ const uint8_t *src0, uint32_t src0_stride,
+ const uint8_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w);
+
+ // Dimension: width_index
+ static const blend_fn blend[9] = {
+ blend_a64_vmask_w16n_sse4_1, // w % 16 == 0
+ vpx_blend_a64_vmask_c, // w == 1
+ vpx_blend_a64_vmask_c, // w == 2
+ NULL, // INVALID
+ blend_a64_vmask_w4_sse4_1, // w == 4
+ NULL, // INVALID
+ NULL, // INVALID
+ NULL, // INVALID
+ blend_a64_vmask_w8_sse4_1, // w == 8
+ };
+
+ assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
+ assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ blend[w & 0xf](dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, h, w);
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+//////////////////////////////////////////////////////////////////////////////
+// Implementation - No sub-sampling
+//////////////////////////////////////////////////////////////////////////////
+
+static INLINE void blend_a64_vmask_bn_w4_sse4_1(
+ uint16_t *dst, uint32_t dst_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, blend_unit_fn blend) {
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
+
+ do {
+ const __m128i v_m0_w = _mm_set1_epi16(*mask);
+ const __m128i v_m1_w = _mm_sub_epi16(v_maxval_w, v_m0_w);
+
+ const __m128i v_res_w = blend(src0, src1, v_m0_w, v_m1_w);
+
+ xx_storel_64(dst, v_res_w);
+
+ dst += dst_stride;
+ src0 += src0_stride;
+ src1 += src1_stride;
+ mask += 1;
+ } while (--h);
+}
+
+static void blend_a64_vmask_b10_w4_sse4_1(
+ uint16_t *dst, uint32_t dst_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ (void)w;
+ blend_a64_vmask_bn_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, h,
+ blend_4_b10);
+}
+
+static void blend_a64_vmask_b12_w4_sse4_1(
+ uint16_t *dst, uint32_t dst_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ (void)w;
+ blend_a64_vmask_bn_w4_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, h,
+ blend_4_b12);
+}
+
+static inline void blend_a64_vmask_bn_w8n_sse4_1(
+ uint16_t *dst, uint32_t dst_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w, blend_unit_fn blend) {
+ const __m128i v_maxval_w = _mm_set1_epi16(VPX_BLEND_A64_MAX_ALPHA);
+
+ do {
+ int c;
+ const __m128i v_m0_w = _mm_set1_epi16(*mask);
+ const __m128i v_m1_w = _mm_sub_epi16(v_maxval_w, v_m0_w);
+ for (c = 0; c < w; c += 8) {
+ const __m128i v_res_w = blend(src0 + c, src1 + c, v_m0_w, v_m1_w);
+
+ xx_storeu_128(dst + c, v_res_w);
+ }
+ dst += dst_stride;
+ src0 += src0_stride;
+ src1 += src1_stride;
+ mask += 1;
+ } while (--h);
+}
+
+static void blend_a64_vmask_b10_w8n_sse4_1(
+ uint16_t *dst, uint32_t dst_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ blend_a64_vmask_bn_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, h, w,
+ blend_8_b10);
+}
+
+static void blend_a64_vmask_b12_w8n_sse4_1(
+ uint16_t *dst, uint32_t dst_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w) {
+ blend_a64_vmask_bn_w8n_sse4_1(dst, dst_stride, src0, src0_stride, src1,
+ src1_stride, mask, h, w,
+ blend_8_b12);
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// Dispatch
+//////////////////////////////////////////////////////////////////////////////
+
+void vpx_highbd_blend_a64_vmask_sse4_1(
+ uint8_t *dst_8, uint32_t dst_stride,
+ const uint8_t *src0_8, uint32_t src0_stride,
+ const uint8_t *src1_8, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w, int bd) {
+ typedef void (*blend_fn)(uint16_t *dst, uint32_t dst_stride,
+ const uint16_t *src0, uint32_t src0_stride,
+ const uint16_t *src1, uint32_t src1_stride,
+ const uint8_t *mask, int h, int w);
+
+ // Dimensions are: bd_index X width_index
+ static const blend_fn blend[2][2] = {
+ { // bd == 8 or 10
+ blend_a64_vmask_b10_w8n_sse4_1, // w % 8 == 0
+ blend_a64_vmask_b10_w4_sse4_1, // w == 4
+ }, { // bd == 12
+ blend_a64_vmask_b12_w8n_sse4_1, // w % 8 == 0
+ blend_a64_vmask_b12_w4_sse4_1, // w == 4
+ }
+ };
+
+ assert(IMPLIES(src0_8 == dst_8, src0_stride == dst_stride));
+ assert(IMPLIES(src1_8 == dst_8, src1_stride == dst_stride));
+
+ assert(h >= 1);
+ assert(w >= 1);
+ assert(IS_POWER_OF_TWO(h));
+ assert(IS_POWER_OF_TWO(w));
+
+ assert(bd == 8 || bd == 10 || bd == 12);
+
+ if (UNLIKELY((h | w) & 3)) { // if (w <= 2 || h <= 2)
+ vpx_highbd_blend_a64_vmask_c(dst_8, dst_stride,
+ src0_8, src0_stride,
+ src1_8, src1_stride,
+ mask, h, w, bd);
+ } else {
+ uint16_t *const dst = CONVERT_TO_SHORTPTR(dst_8);
+ const uint16_t *const src0 = CONVERT_TO_SHORTPTR(src0_8);
+ const uint16_t *const src1 = CONVERT_TO_SHORTPTR(src1_8);
+
+ blend[bd == 12][(w >> 2) & 1](dst, dst_stride,
+ src0, src0_stride,
+ src1, src1_stride,
+ mask, h, w);
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/x86/blend_sse4.h b/vpx_dsp/x86/blend_sse4.h
new file mode 100644
index 0000000..9b74f90
--- /dev/null
+++ b/vpx_dsp/x86/blend_sse4.h
@@ -0,0 +1,145 @@
+/*
+* 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 VPX_DSP_X86_BLEND_SSE4_H_
+#define VPX_DSP_X86_BLEND_SSE4_H_
+
+#include "vpx_dsp/blend.h"
+#include "vpx_dsp/x86/synonyms.h"
+
+//////////////////////////////////////////////////////////////////////////////
+// Common kernels
+//////////////////////////////////////////////////////////////////////////////
+
+static INLINE __m128i blend_4(const uint8_t *src0, const uint8_t *src1,
+ const __m128i v_m0_w, const __m128i v_m1_w) {
+ const __m128i v_s0_b = xx_loadl_32(src0);
+ const __m128i v_s1_b = xx_loadl_32(src1);
+ const __m128i v_s0_w = _mm_cvtepu8_epi16(v_s0_b);
+ const __m128i v_s1_w = _mm_cvtepu8_epi16(v_s1_b);
+
+ const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
+ const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
+
+ const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
+
+ const __m128i v_res_w = xx_roundn_epu16(v_sum_w, VPX_BLEND_A64_ROUND_BITS);
+
+ return v_res_w;
+}
+
+static INLINE __m128i blend_8(const uint8_t *src0, const uint8_t *src1,
+ const __m128i v_m0_w, const __m128i v_m1_w) {
+ const __m128i v_s0_b = xx_loadl_64(src0);
+ const __m128i v_s1_b = xx_loadl_64(src1);
+ const __m128i v_s0_w = _mm_cvtepu8_epi16(v_s0_b);
+ const __m128i v_s1_w = _mm_cvtepu8_epi16(v_s1_b);
+
+ const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
+ const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
+
+ const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
+
+ const __m128i v_res_w = xx_roundn_epu16(v_sum_w, VPX_BLEND_A64_ROUND_BITS);
+
+ return v_res_w;
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+typedef __m128i (*blend_unit_fn)(const uint16_t *src0, const uint16_t *src1,
+ const __m128i v_m0_w, const __m128i v_m1_w);
+
+static INLINE __m128i blend_4_b10(const uint16_t *src0, const uint16_t *src1,
+ const __m128i v_m0_w, const __m128i v_m1_w) {
+ const __m128i v_s0_w = xx_loadl_64(src0);
+ const __m128i v_s1_w = xx_loadl_64(src1);
+
+ const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
+ const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
+
+ const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
+
+ const __m128i v_res_w = xx_roundn_epu16(v_sum_w, VPX_BLEND_A64_ROUND_BITS);
+
+ return v_res_w;
+}
+
+static INLINE __m128i blend_8_b10(const uint16_t *src0, const uint16_t *src1,
+ const __m128i v_m0_w, const __m128i v_m1_w) {
+ const __m128i v_s0_w = xx_loadu_128(src0);
+ const __m128i v_s1_w = xx_loadu_128(src1);
+
+ const __m128i v_p0_w = _mm_mullo_epi16(v_s0_w, v_m0_w);
+ const __m128i v_p1_w = _mm_mullo_epi16(v_s1_w, v_m1_w);
+
+ const __m128i v_sum_w = _mm_add_epi16(v_p0_w, v_p1_w);
+
+ const __m128i v_res_w = xx_roundn_epu16(v_sum_w, VPX_BLEND_A64_ROUND_BITS);
+
+ return v_res_w;
+}
+
+static INLINE __m128i blend_4_b12(const uint16_t *src0, const uint16_t *src1,
+ const __m128i v_m0_w, const __m128i v_m1_w) {
+ const __m128i v_s0_w = xx_loadl_64(src0);
+ const __m128i v_s1_w = xx_loadl_64(src1);
+
+ // Interleave
+ const __m128i v_m01_w = _mm_unpacklo_epi16(v_m0_w, v_m1_w);
+ const __m128i v_s01_w = _mm_unpacklo_epi16(v_s0_w, v_s1_w);
+
+ // Multiply-Add
+ const __m128i v_sum_d = _mm_madd_epi16(v_s01_w, v_m01_w);
+
+ // Scale
+ const __m128i v_ssum_d = _mm_srli_epi32(v_sum_d,
+ VPX_BLEND_A64_ROUND_BITS - 1);
+
+ // Pack
+ const __m128i v_pssum_d = _mm_packs_epi32(v_ssum_d, v_ssum_d);
+
+ // Round
+ const __m128i v_res_w = xx_round_epu16(v_pssum_d);
+
+ return v_res_w;
+}
+
+static INLINE __m128i blend_8_b12(const uint16_t *src0, const uint16_t *src1,
+ const __m128i v_m0_w, const __m128i v_m1_w) {
+ const __m128i v_s0_w = xx_loadu_128(src0);
+ const __m128i v_s1_w = xx_loadu_128(src1);
+
+ // Interleave
+ const __m128i v_m01l_w = _mm_unpacklo_epi16(v_m0_w, v_m1_w);
+ const __m128i v_m01h_w = _mm_unpackhi_epi16(v_m0_w, v_m1_w);
+ const __m128i v_s01l_w = _mm_unpacklo_epi16(v_s0_w, v_s1_w);
+ const __m128i v_s01h_w = _mm_unpackhi_epi16(v_s0_w, v_s1_w);
+
+ // Multiply-Add
+ const __m128i v_suml_d = _mm_madd_epi16(v_s01l_w, v_m01l_w);
+ const __m128i v_sumh_d = _mm_madd_epi16(v_s01h_w, v_m01h_w);
+
+ // Scale
+ const __m128i v_ssuml_d = _mm_srli_epi32(v_suml_d,
+ VPX_BLEND_A64_ROUND_BITS - 1);
+ const __m128i v_ssumh_d = _mm_srli_epi32(v_sumh_d,
+ VPX_BLEND_A64_ROUND_BITS - 1);
+
+ // Pack
+ const __m128i v_pssum_d = _mm_packs_epi32(v_ssuml_d, v_ssumh_d);
+
+ // Round
+ const __m128i v_res_w = xx_round_epu16(v_pssum_d);
+
+ return v_res_w;
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+#endif // VPX_DSP_X86_BLEND_SSE4_H_
diff --git a/vpx_dsp/x86/highbd_quantize_intrin_sse2.c b/vpx_dsp/x86/highbd_quantize_intrin_sse2.c
index fd46bef..164ffcf 100644
--- a/vpx_dsp/x86/highbd_quantize_intrin_sse2.c
+++ b/vpx_dsp/x86/highbd_quantize_intrin_sse2.c
@@ -10,6 +10,7 @@
#include <emmintrin.h>
+#include "./vpx_dsp_rtcd.h"
#include "vpx_dsp/vpx_dsp_common.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
diff --git a/vpx_dsp/x86/obmc_sad_sse4.c b/vpx_dsp/x86/obmc_sad_sse4.c
index 57e1428..de12e1d 100644
--- a/vpx_dsp/x86/obmc_sad_sse4.c
+++ b/vpx_dsp/x86/obmc_sad_sse4.c
@@ -21,26 +21,28 @@
// 8 bit
////////////////////////////////////////////////////////////////////////////////
-static INLINE unsigned int obmc_sad_w4(const uint8_t *a, const int a_stride,
- const int32_t *b, const int32_t *m,
+static INLINE unsigned int obmc_sad_w4(const uint8_t *pre,
+ const int pre_stride,
+ const int32_t *wsrc,
+ const int32_t *mask,
const int height) {
- const int a_step = a_stride - 4;
+ const int pre_step = pre_stride - 4;
int n = 0;
__m128i v_sad_d = _mm_setzero_si128();
do {
- const __m128i v_a_b = xx_loadl_32(a + n);
- const __m128i v_m_d = xx_load_128(m + n);
- const __m128i v_b_d = xx_load_128(b + n);
+ const __m128i v_p_b = xx_loadl_32(pre + n);
+ const __m128i v_m_d = xx_load_128(mask + n);
+ const __m128i v_w_d = xx_load_128(wsrc + n);
- const __m128i v_a_d = _mm_cvtepu8_epi32(v_a_b);
+ const __m128i v_p_d = _mm_cvtepu8_epi32(v_p_b);
- // Values in both a and m fit in 15 bits, and are packed at 32 bit
+ // Values in both pre and mask fit in 15 bits, and are packed at 32 bit
// boundaries. We use pmaddwd, as it has lower latency on Haswell
// than pmulld but produces the same result with these inputs.
- const __m128i v_am_d = _mm_madd_epi16(v_a_d, v_m_d);
+ const __m128i v_pm_d = _mm_madd_epi16(v_p_d, v_m_d);
- const __m128i v_diff_d = _mm_sub_epi32(v_b_d, v_am_d);
+ const __m128i v_diff_d = _mm_sub_epi32(v_w_d, v_pm_d);
const __m128i v_absdiff_d = _mm_abs_epi32(v_diff_d);
// Rounded absolute difference
@@ -51,39 +53,42 @@
n += 4;
if (n % 4 == 0)
- a += a_step;
+ pre += pre_step;
} while (n < 4 * height);
return xx_hsum_epi32_si32(v_sad_d);
}
-static INLINE unsigned int obmc_sad_w8n(const uint8_t *a, const int a_stride,
- const int32_t *b, const int32_t *m,
- const int width, const int height) {
- const int a_step = a_stride - width;
+static INLINE unsigned int obmc_sad_w8n(const uint8_t *pre,
+ const int pre_stride,
+ const int32_t *wsrc,
+ const int32_t *mask,
+ const int width,
+ const int height) {
+ const int pre_step = pre_stride - width;
int n = 0;
__m128i v_sad_d = _mm_setzero_si128();
assert(width >= 8 && (width & (width - 1)) == 0);
do {
- const __m128i v_a1_b = xx_loadl_32(a + n + 4);
- const __m128i v_m1_d = xx_load_128(m + n + 4);
- const __m128i v_b1_d = xx_load_128(b + n + 4);
- const __m128i v_a0_b = xx_loadl_32(a + n);
- const __m128i v_m0_d = xx_load_128(m + n);
- const __m128i v_b0_d = xx_load_128(b + n);
+ const __m128i v_p1_b = xx_loadl_32(pre + n + 4);
+ const __m128i v_m1_d = xx_load_128(mask + n + 4);
+ const __m128i v_w1_d = xx_load_128(wsrc + n + 4);
+ const __m128i v_p0_b = xx_loadl_32(pre + n);
+ const __m128i v_m0_d = xx_load_128(mask + n);
+ const __m128i v_w0_d = xx_load_128(wsrc + n);
- const __m128i v_a0_d = _mm_cvtepu8_epi32(v_a0_b);
- const __m128i v_a1_d = _mm_cvtepu8_epi32(v_a1_b);
+ const __m128i v_p0_d = _mm_cvtepu8_epi32(v_p0_b);
+ const __m128i v_p1_d = _mm_cvtepu8_epi32(v_p1_b);
- // Values in both a and m fit in 15 bits, and are packed at 32 bit
+ // Values in both pre and mask fit in 15 bits, and are packed at 32 bit
// boundaries. We use pmaddwd, as it has lower latency on Haswell
// than pmulld but produces the same result with these inputs.
- const __m128i v_am0_d = _mm_madd_epi16(v_a0_d, v_m0_d);
- const __m128i v_am1_d = _mm_madd_epi16(v_a1_d, v_m1_d);
+ const __m128i v_pm0_d = _mm_madd_epi16(v_p0_d, v_m0_d);
+ const __m128i v_pm1_d = _mm_madd_epi16(v_p1_d, v_m1_d);
- const __m128i v_diff0_d = _mm_sub_epi32(v_b0_d, v_am0_d);
- const __m128i v_diff1_d = _mm_sub_epi32(v_b1_d, v_am1_d);
+ const __m128i v_diff0_d = _mm_sub_epi32(v_w0_d, v_pm0_d);
+ const __m128i v_diff1_d = _mm_sub_epi32(v_w1_d, v_pm1_d);
const __m128i v_absdiff0_d = _mm_abs_epi32(v_diff0_d);
const __m128i v_absdiff1_d = _mm_abs_epi32(v_diff1_d);
@@ -97,21 +102,21 @@
n += 8;
if (n % width == 0)
- a += a_step;
+ pre += pre_step;
} while (n < width * height);
return xx_hsum_epi32_si32(v_sad_d);
}
#define OBMCSADWXH(w, h) \
-unsigned int vpx_obmc_sad##w##x##h##_sse4_1(const uint8_t *ref, \
- int ref_stride, \
+unsigned int vpx_obmc_sad##w##x##h##_sse4_1(const uint8_t *pre, \
+ int pre_stride, \
const int32_t *wsrc, \
const int32_t *msk) { \
if (w == 4) \
- return obmc_sad_w4(ref, ref_stride, wsrc, msk, h); \
+ return obmc_sad_w4(pre, pre_stride, wsrc, msk, h); \
else \
- return obmc_sad_w8n(ref, ref_stride, wsrc, msk, w, h); \
+ return obmc_sad_w8n(pre, pre_stride, wsrc, msk, w, h); \
}
#if CONFIG_EXT_PARTITION
@@ -138,28 +143,29 @@
////////////////////////////////////////////////////////////////////////////////
#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE unsigned int hbd_obmc_sad_w4(const uint8_t *a8,
- const int a_stride,
- const int32_t *b, const int32_t *m,
+static INLINE unsigned int hbd_obmc_sad_w4(const uint8_t *pre8,
+ const int pre_stride,
+ const int32_t *wsrc,
+ const int32_t *mask,
const int height) {
- const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
- const int a_step = a_stride - 4;
+ const uint16_t *pre = CONVERT_TO_SHORTPTR(pre8);
+ const int pre_step = pre_stride - 4;
int n = 0;
__m128i v_sad_d = _mm_setzero_si128();
do {
- const __m128i v_a_w = xx_loadl_64(a + n);
- const __m128i v_m_d = xx_load_128(m + n);
- const __m128i v_b_d = xx_load_128(b + n);
+ const __m128i v_p_w = xx_loadl_64(pre + n);
+ const __m128i v_m_d = xx_load_128(mask + n);
+ const __m128i v_w_d = xx_load_128(wsrc + n);
- const __m128i v_a_d = _mm_cvtepu16_epi32(v_a_w);
+ const __m128i v_p_d = _mm_cvtepu16_epi32(v_p_w);
- // Values in both a and m fit in 15 bits, and are packed at 32 bit
+ // Values in both pre and mask fit in 15 bits, and are packed at 32 bit
// boundaries. We use pmaddwd, as it has lower latency on Haswell
// than pmulld but produces the same result with these inputs.
- const __m128i v_am_d = _mm_madd_epi16(v_a_d, v_m_d);
+ const __m128i v_pm_d = _mm_madd_epi16(v_p_d, v_m_d);
- const __m128i v_diff_d = _mm_sub_epi32(v_b_d, v_am_d);
+ const __m128i v_diff_d = _mm_sub_epi32(v_w_d, v_pm_d);
const __m128i v_absdiff_d = _mm_abs_epi32(v_diff_d);
// Rounded absolute difference
@@ -170,41 +176,43 @@
n += 4;
if (n % 4 == 0)
- a += a_step;
+ pre += pre_step;
} while (n < 4 * height);
return xx_hsum_epi32_si32(v_sad_d);
}
-static INLINE unsigned int hbd_obmc_sad_w8n(const uint8_t *a8,
- const int a_stride,
- const int32_t *b, const int32_t *m,
- const int width, const int height) {
- const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
- const int a_step = a_stride - width;
+static INLINE unsigned int hbd_obmc_sad_w8n(const uint8_t *pre8,
+ const int pre_stride,
+ const int32_t *wsrc,
+ const int32_t *mask,
+ const int width,
+ const int height) {
+ const uint16_t *pre = CONVERT_TO_SHORTPTR(pre8);
+ const int pre_step = pre_stride - width;
int n = 0;
__m128i v_sad_d = _mm_setzero_si128();
assert(width >= 8 && (width & (width - 1)) == 0);
do {
- const __m128i v_a1_w = xx_loadl_64(a + n + 4);
- const __m128i v_m1_d = xx_load_128(m + n + 4);
- const __m128i v_b1_d = xx_load_128(b + n + 4);
- const __m128i v_a0_w = xx_loadl_64(a + n);
- const __m128i v_m0_d = xx_load_128(m + n);
- const __m128i v_b0_d = xx_load_128(b + n);
+ const __m128i v_p1_w = xx_loadl_64(pre + n + 4);
+ const __m128i v_m1_d = xx_load_128(mask + n + 4);
+ const __m128i v_w1_d = xx_load_128(wsrc + n + 4);
+ const __m128i v_p0_w = xx_loadl_64(pre + n);
+ const __m128i v_m0_d = xx_load_128(mask + n);
+ const __m128i v_w0_d = xx_load_128(wsrc + n);
- const __m128i v_a0_d = _mm_cvtepu16_epi32(v_a0_w);
- const __m128i v_a1_d = _mm_cvtepu16_epi32(v_a1_w);
+ const __m128i v_p0_d = _mm_cvtepu16_epi32(v_p0_w);
+ const __m128i v_p1_d = _mm_cvtepu16_epi32(v_p1_w);
- // Values in both a and m fit in 15 bits, and are packed at 32 bit
+ // Values in both pre and mask fit in 15 bits, and are packed at 32 bit
// boundaries. We use pmaddwd, as it has lower latency on Haswell
// than pmulld but produces the same result with these inputs.
- const __m128i v_am0_d = _mm_madd_epi16(v_a0_d, v_m0_d);
- const __m128i v_am1_d = _mm_madd_epi16(v_a1_d, v_m1_d);
+ const __m128i v_pm0_d = _mm_madd_epi16(v_p0_d, v_m0_d);
+ const __m128i v_pm1_d = _mm_madd_epi16(v_p1_d, v_m1_d);
- const __m128i v_diff0_d = _mm_sub_epi32(v_b0_d, v_am0_d);
- const __m128i v_diff1_d = _mm_sub_epi32(v_b1_d, v_am1_d);
+ const __m128i v_diff0_d = _mm_sub_epi32(v_w0_d, v_pm0_d);
+ const __m128i v_diff1_d = _mm_sub_epi32(v_w1_d, v_pm1_d);
const __m128i v_absdiff0_d = _mm_abs_epi32(v_diff0_d);
const __m128i v_absdiff1_d = _mm_abs_epi32(v_diff1_d);
@@ -218,21 +226,21 @@
n += 8;
if (n % width == 0)
- a += a_step;
+ pre += pre_step;
} while (n < width * height);
return xx_hsum_epi32_si32(v_sad_d);
}
#define HBD_OBMCSADWXH(w, h) \
-unsigned int vpx_highbd_obmc_sad##w##x##h##_sse4_1(const uint8_t *ref, \
- int ref_stride, \
+unsigned int vpx_highbd_obmc_sad##w##x##h##_sse4_1(const uint8_t *pre, \
+ int pre_stride, \
const int32_t *wsrc, \
- const int32_t *msk) { \
+ const int32_t *mask) { \
if (w == 4) \
- return hbd_obmc_sad_w4(ref, ref_stride, wsrc, msk, h); \
+ return hbd_obmc_sad_w4(pre, pre_stride, wsrc, mask, h); \
else \
- return hbd_obmc_sad_w8n(ref, ref_stride, wsrc, msk, w, h); \
+ return hbd_obmc_sad_w8n(pre, pre_stride, wsrc, mask, w, h); \
}
#if CONFIG_EXT_PARTITION
diff --git a/vpx_ports/mem.h b/vpx_ports/mem.h
index 1dca1a8..48549ce 100644
--- a/vpx_ports/mem.h
+++ b/vpx_ports/mem.h
@@ -38,24 +38,15 @@
#define __builtin_prefetch(x)
#endif
-/* Shift down with rounding for use when n > 0 */
+/* Shift down with rounding for use when n >= 0, value >= 0 */
#define ROUND_POWER_OF_TWO(value, n) \
- (((value) + (1 << ((n) - 1))) >> (n))
+ (((value) + (((1 << (n)) >> 1))) >> (n))
-/* Shift down with rounding for use when n >= 0 */
-#define ROUNDZ_POWER_OF_TWO(value, n) \
- ((n) ? (((value) + (1 << ((n) - 1))) >> (n)) : (value))
-
-/* Shift down with rounding for signed integers, for use when n > 0 */
+/* Shift down with rounding for signed integers, for use when n >= 0 */
#define ROUND_POWER_OF_TWO_SIGNED(value, n) \
(((value) < 0) ? -ROUND_POWER_OF_TWO(-(value), (n)) \
: ROUND_POWER_OF_TWO((value), (n)))
-/* Shift down with rounding for signed integers, for use when n >= 0 */
-#define ROUNDZ_POWER_OF_TWO_SIGNED(value, n) \
- (((value) < 0) ? -ROUNDZ_POWER_OF_TWO(-(value), (n)) \
- : ROUNDZ_POWER_OF_TWO((value), (n)))
-
#define ALIGN_POWER_OF_TWO(value, n) \
(((value) + ((1 << (n)) - 1)) & ~((1 << (n)) - 1))
