ref: fed8a1837fd6b1e4e36495be8210bd63bfb2bb8f
parent: 2914ab2e8b9fca26a430aed1dea56070621eb71e
author: Daniel Kang <ddkang@google.com>
date: Thu Aug 2 13:03:14 EDT 2012
16x16 DCT blocks. Set on all 16x16 intra/inter modes Features: - Butterfly fDCT/iDCT - Loop filter does not filter internal edges with 16x16 - Optimize coefficient function - Update coefficient probability function - RD - Entropy stats - 16x16 is a config option Have not tested with experiments. hd: 2.60% std-hd: 2.43% yt: 1.32% derf: 0.60% Change-Id: I96fb090517c30c5da84bad4fae602c3ec0c58b1c
--- a/configure
+++ b/configure
@@ -227,6 +227,7 @@
hybridtransform
switchable_interp
htrans8x8
+ tx16x16
"
CONFIG_LIST="
external_build
--- /dev/null
+++ b/test/dct16x16_test.cc
@@ -1,0 +1,356 @@
+/*
+ * Copyright (c) 2012 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"
+
+extern "C" {+#include "vp8/common/entropy.h"
+#include "vp8/common/idct.h"
+#include "vp8/encoder/dct.h"
+}
+
+#include "acm_random.h"
+#include "vpx/vpx_integer.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {+
+const double PI = 3.1415926535898;
+void reference2_16x16_idct_2d(double *input, double *output) {+ double x;
+ for (int l = 0; l < 16; ++l) {+ for (int k = 0; k < 16; ++k) {+ double s = 0;
+ for (int i = 0; i < 16; ++i) {+ for (int j = 0; j < 16; ++j) {+ x=cos(PI*j*(l+0.5)/16.0)*cos(PI*i*(k+0.5)/16.0)*input[i*16+j]/256;
+ if (i != 0)
+ x *= sqrt(2.0);
+ if (j != 0)
+ x *= sqrt(2.0);
+ s += x;
+ }
+ }
+ output[k*16+l] = s;
+ }
+ }
+}
+
+static void butterfly_16x16_dct_1d(double input[16], double output[16]) {+ double step[16];
+ double intermediate[16];
+ double temp1, temp2;
+
+ const double C1 = cos(1*PI/(double)32);
+ const double C2 = cos(2*PI/(double)32);
+ const double C3 = cos(3*PI/(double)32);
+ const double C4 = cos(4*PI/(double)32);
+ const double C5 = cos(5*PI/(double)32);
+ const double C6 = cos(6*PI/(double)32);
+ const double C7 = cos(7*PI/(double)32);
+ const double C8 = cos(8*PI/(double)32);
+ const double C9 = cos(9*PI/(double)32);
+ const double C10 = cos(10*PI/(double)32);
+ const double C11 = cos(11*PI/(double)32);
+ const double C12 = cos(12*PI/(double)32);
+ const double C13 = cos(13*PI/(double)32);
+ const double C14 = cos(14*PI/(double)32);
+ const double C15 = cos(15*PI/(double)32);
+
+ // step 1
+ step[ 0] = input[0] + input[15];
+ step[ 1] = input[1] + input[14];
+ step[ 2] = input[2] + input[13];
+ step[ 3] = input[3] + input[12];
+ step[ 4] = input[4] + input[11];
+ step[ 5] = input[5] + input[10];
+ step[ 6] = input[6] + input[ 9];
+ step[ 7] = input[7] + input[ 8];
+ step[ 8] = input[7] - input[ 8];
+ step[ 9] = input[6] - input[ 9];
+ step[10] = input[5] - input[10];
+ step[11] = input[4] - input[11];
+ step[12] = input[3] - input[12];
+ step[13] = input[2] - input[13];
+ step[14] = input[1] - input[14];
+ step[15] = input[0] - input[15];
+
+ // step 2
+ output[0] = step[0] + step[7];
+ output[1] = step[1] + step[6];
+ output[2] = step[2] + step[5];
+ output[3] = step[3] + step[4];
+ output[4] = step[3] - step[4];
+ output[5] = step[2] - step[5];
+ output[6] = step[1] - step[6];
+ output[7] = step[0] - step[7];
+
+ temp1 = step[ 8]*C7;
+ temp2 = step[15]*C9;
+ output[ 8] = temp1 + temp2;
+
+ temp1 = step[ 9]*C11;
+ temp2 = step[14]*C5;
+ output[ 9] = temp1 - temp2;
+
+ temp1 = step[10]*C3;
+ temp2 = step[13]*C13;
+ output[10] = temp1 + temp2;
+
+ temp1 = step[11]*C15;
+ temp2 = step[12]*C1;
+ output[11] = temp1 - temp2;
+
+ temp1 = step[11]*C1;
+ temp2 = step[12]*C15;
+ output[12] = temp2 + temp1;
+
+ temp1 = step[10]*C13;
+ temp2 = step[13]*C3;
+ output[13] = temp2 - temp1;
+
+ temp1 = step[ 9]*C5;
+ temp2 = step[14]*C11;
+ output[14] = temp2 + temp1;
+
+ temp1 = step[ 8]*C9;
+ temp2 = step[15]*C7;
+ output[15] = temp2 - temp1;
+
+ // step 3
+ step[ 0] = output[0] + output[3];
+ step[ 1] = output[1] + output[2];
+ step[ 2] = output[1] - output[2];
+ step[ 3] = output[0] - output[3];
+
+ temp1 = output[4]*C14;
+ temp2 = output[7]*C2;
+ step[ 4] = temp1 + temp2;
+
+ temp1 = output[5]*C10;
+ temp2 = output[6]*C6;
+ step[ 5] = temp1 + temp2;
+
+ temp1 = output[5]*C6;
+ temp2 = output[6]*C10;
+ step[ 6] = temp2 - temp1;
+
+ temp1 = output[4]*C2;
+ temp2 = output[7]*C14;
+ step[ 7] = temp2 - temp1;
+
+ step[ 8] = output[ 8] + output[11];
+ step[ 9] = output[ 9] + output[10];
+ step[10] = output[ 9] - output[10];
+ step[11] = output[ 8] - output[11];
+
+ step[12] = output[12] + output[15];
+ step[13] = output[13] + output[14];
+ step[14] = output[13] - output[14];
+ step[15] = output[12] - output[15];
+
+ // step 4
+ output[ 0] = (step[ 0] + step[ 1]);
+ output[ 8] = (step[ 0] - step[ 1]);
+
+ temp1 = step[2]*C12;
+ temp2 = step[3]*C4;
+ temp1 = temp1 + temp2;
+ output[ 4] = 2*(temp1*C8);
+
+ temp1 = step[2]*C4;
+ temp2 = step[3]*C12;
+ temp1 = temp2 - temp1;
+ output[12] = 2*(temp1*C8);
+
+ output[ 2] = 2*((step[4] + step[ 5])*C8);
+ output[14] = 2*((step[7] - step[ 6])*C8);
+
+ temp1 = step[4] - step[5];
+ temp2 = step[6] + step[7];
+ output[ 6] = (temp1 + temp2);
+ output[10] = (temp1 - temp2);
+
+ intermediate[8] = step[8] + step[14];
+ intermediate[9] = step[9] + step[15];
+
+ temp1 = intermediate[8]*C12;
+ temp2 = intermediate[9]*C4;
+ temp1 = temp1 - temp2;
+ output[3] = 2*(temp1*C8);
+
+ temp1 = intermediate[8]*C4;
+ temp2 = intermediate[9]*C12;
+ temp1 = temp2 + temp1;
+ output[13] = 2*(temp1*C8);
+
+ output[ 9] = 2*((step[10] + step[11])*C8);
+
+ intermediate[11] = step[10] - step[11];
+ intermediate[12] = step[12] + step[13];
+ intermediate[13] = step[12] - step[13];
+ intermediate[14] = step[ 8] - step[14];
+ intermediate[15] = step[ 9] - step[15];
+
+ output[15] = (intermediate[11] + intermediate[12]);
+ output[ 1] = -(intermediate[11] - intermediate[12]);
+
+ output[ 7] = 2*(intermediate[13]*C8);
+
+ temp1 = intermediate[14]*C12;
+ temp2 = intermediate[15]*C4;
+ temp1 = temp1 - temp2;
+ output[11] = -2*(temp1*C8);
+
+ temp1 = intermediate[14]*C4;
+ temp2 = intermediate[15]*C12;
+ temp1 = temp2 + temp1;
+ output[ 5] = 2*(temp1*C8);
+}
+
+static void reference_16x16_dct_1d(double in[16], double out[16]) {+ const double kPi = 3.141592653589793238462643383279502884;
+ const double kInvSqrt2 = 0.707106781186547524400844362104;
+ for (int k = 0; k < 16; k++) {+ out[k] = 0.0;
+ for (int n = 0; n < 16; n++)
+ out[k] += in[n]*cos(kPi*(2*n+1)*k/32.0);
+ if (k == 0)
+ out[k] = out[k]*kInvSqrt2;
+ }
+}
+
+void reference_16x16_dct_2d(int16_t input[16*16], double output[16*16]) {+ // First transform columns
+ for (int i = 0; i < 16; ++i) {+ double temp_in[16], temp_out[16];
+ for (int j = 0; j < 16; ++j)
+ temp_in[j] = input[j*16 + i];
+ butterfly_16x16_dct_1d(temp_in, temp_out);
+ for (int j = 0; j < 16; ++j)
+ output[j*16 + i] = temp_out[j];
+ }
+ // Then transform rows
+ for (int i = 0; i < 16; ++i) {+ double temp_in[16], temp_out[16];
+ for (int j = 0; j < 16; ++j)
+ temp_in[j] = output[j + i*16];
+ butterfly_16x16_dct_1d(temp_in, temp_out);
+ // Scale by some magic number
+ for (int j = 0; j < 16; ++j)
+ output[j + i*16] = temp_out[j]/2;
+ }
+}
+
+
+TEST(VP8Idct16x16Test, AccuracyCheck) {+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = 1000;
+ for (int i = 0; i < count_test_block; ++i) {+ int16_t in[256], coeff[256];
+ int16_t out_c[256];
+ double out_r[256];
+
+ // Initialize a test block with input range [-255, 255].
+ for (int j = 0; j < 256; ++j)
+ in[j] = rnd.Rand8() - rnd.Rand8();
+
+ reference_16x16_dct_2d(in, out_r);
+ for (int j = 0; j < 256; j++)
+ coeff[j] = round(out_r[j]);
+ vp8_short_idct16x16_c(coeff, out_c, 32);
+ for (int j = 0; j < 256; ++j) {+ const int diff = out_c[j] - in[j];
+ const int error = diff * diff;
+ EXPECT_GE(1, error)
+ << "Error: 16x16 IDCT has error " << error
+ << " at index " << j;
+ }
+
+ vp8_short_fdct16x16_c(in, out_c, 32);
+ for (int j = 0; j < 256; ++j) {+ const double diff = coeff[j] - out_c[j];
+ const double error = diff * diff;
+ EXPECT_GE(1.0, error)
+ << "Error: 16x16 FDCT has error " << error
+ << " at index " << j;
+ }
+ }
+}
+
+TEST(VP8Fdct16x16Test, AccuracyCheck) {+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ int max_error = 0;
+ double total_error = 0;
+ const int count_test_block = 1000;
+ for (int i = 0; i < count_test_block; ++i) {+ int16_t test_input_block[256];
+ int16_t test_temp_block[256];
+ int16_t test_output_block[256];
+
+ // Initialize a test block with input range [-255, 255].
+ for (int j = 0; j < 256; ++j)
+ test_input_block[j] = rnd.Rand8() - rnd.Rand8();
+
+ const int pitch = 32;
+ vp8_short_fdct16x16_c(test_input_block, test_temp_block, pitch);
+ vp8_short_idct16x16_c(test_temp_block, test_output_block, pitch);
+
+ for (int j = 0; j < 256; ++j) {+ const int diff = test_input_block[j] - test_output_block[j];
+ const int error = diff * diff;
+ if (max_error < error)
+ max_error = error;
+ total_error += error;
+ }
+ }
+
+ EXPECT_GE(1, max_error)
+ << "Error: 16x16 FDCT/IDCT has an individual roundtrip error > 1";
+
+ EXPECT_GE(count_test_block/10, total_error)
+ << "Error: 16x16 FDCT/IDCT has average roundtrip error > 1/10 per block";
+}
+
+TEST(VP8Fdct16x16Test, CoeffSizeCheck) {+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = 1000;
+ for (int i = 0; i < count_test_block; ++i) {+ int16_t input_block[256], input_extreme_block[256];
+ int16_t output_block[256], output_extreme_block[256];
+
+ // Initialize a test block with input range [-255, 255].
+ for (int j = 0; j < 256; ++j) {+ input_block[j] = rnd.Rand8() - rnd.Rand8();
+ input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
+ }
+ if (i == 0)
+ for (int j = 0; j < 256; ++j)
+ input_extreme_block[j] = 255;
+
+ const int pitch = 32;
+ vp8_short_fdct16x16_c(input_block, output_block, pitch);
+ vp8_short_fdct16x16_c(input_extreme_block, output_extreme_block, pitch);
+
+ // The minimum quant value is 4.
+ for (int j = 0; j < 256; ++j) {+ EXPECT_GE(4*DCT_MAX_VALUE, abs(output_block[j]))
+ << "Error: 16x16 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
+ EXPECT_GE(4*DCT_MAX_VALUE, abs(output_extreme_block[j]))
+ << "Error: 16x16 FDCT extreme has coefficient larger than 4*DCT_MAX_VALUE";
+ }
+ }
+}
+} // namespace
--- a/test/fdct8x8_test.cc
+++ b/test/fdct8x8_test.cc
@@ -115,8 +115,8 @@
EXPECT_GE(1, max_error)
<< "Error: 8x8 FDCT/IDCT has an individual roundtrip error > 1";
- EXPECT_GE(count_test_block, total_error)
- << "Error: 8x8 FDCT/IDCT has average roundtrip error > 1 per block";
+ EXPECT_GE(count_test_block/5, total_error)
+ << "Error: 8x8 FDCT/IDCT has average roundtrip error > 1/5 per block";
};
TEST(VP8Fdct8x8Test, ExtremalCheck) {@@ -149,9 +149,9 @@
<< "Error: Extremal 8x8 FDCT/IDCT has an"
<< " individual roundtrip error > 1";
- EXPECT_GE(count_test_block, total_error)
+ EXPECT_GE(count_test_block/5, total_error)
<< "Error: Extremal 8x8 FDCT/IDCT has average"
- << " roundtrip error > 1 per block";
+ << " roundtrip error > 1/5 per block";
}
};
--- /dev/null
+++ b/test/idct8x8_test.cc
@@ -1,0 +1,154 @@
+/*
+ * Copyright (c) 2012 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"
+
+extern "C" {+#include "vp8/encoder/dct.h"
+#include "vp8/common/idct.h"
+}
+
+#include "acm_random.h"
+#include "vpx/vpx_integer.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {+
+void reference_dct_1d(double input[8], double output[8]) {+ const double kPi = 3.141592653589793238462643383279502884;
+ const double kInvSqrt2 = 0.707106781186547524400844362104;
+ for (int k = 0; k < 8; k++) {+ output[k] = 0.0;
+ for (int n = 0; n < 8; n++)
+ output[k] += input[n]*cos(kPi*(2*n+1)*k/16.0);
+ if (k == 0)
+ output[k] = output[k]*kInvSqrt2;
+ }
+}
+
+void reference_dct_2d(int16_t input[64], double output[64]) {+ // First transform columns
+ for (int i = 0; i < 8; ++i) {+ double temp_in[8], temp_out[8];
+ for (int j = 0; j < 8; ++j)
+ temp_in[j] = input[j*8 + i];
+ reference_dct_1d(temp_in, temp_out);
+ for (int j = 0; j < 8; ++j)
+ output[j*8 + i] = temp_out[j];
+ }
+ // Then transform rows
+ for (int i = 0; i < 8; ++i) {+ double temp_in[8], temp_out[8];
+ for (int j = 0; j < 8; ++j)
+ temp_in[j] = output[j + i*8];
+ reference_dct_1d(temp_in, temp_out);
+ for (int j = 0; j < 8; ++j)
+ output[j + i*8] = temp_out[j];
+ }
+ // Scale by some magic number
+ for (int i = 0; i < 64; ++i)
+ output[i] *= 2;
+}
+
+void reference_idct_1d(double input[8], double output[8]) {+ const double kPi = 3.141592653589793238462643383279502884;
+ const double kSqrt2 = 1.414213562373095048801688724209698;
+ for (int k = 0; k < 8; k++) {+ output[k] = 0.0;
+ for (int n = 0; n < 8; n++) {+ output[k] += input[n]*cos(kPi*(2*k+1)*n/16.0);
+ if (n == 0)
+ output[k] = output[k]/kSqrt2;
+ }
+ }
+}
+
+void reference_idct_2d(double input[64], int16_t output[64]) {+ double out[64], out2[64];
+ // First transform rows
+ for (int i = 0; i < 8; ++i) {+ double temp_in[8], temp_out[8];
+ for (int j = 0; j < 8; ++j)
+ temp_in[j] = input[j + i*8];
+ reference_idct_1d(temp_in, temp_out);
+ for (int j = 0; j < 8; ++j)
+ out[j + i*8] = temp_out[j];
+ }
+ // Then transform columns
+ for (int i = 0; i < 8; ++i) {+ double temp_in[8], temp_out[8];
+ for (int j = 0; j < 8; ++j)
+ temp_in[j] = out[j*8 + i];
+ reference_idct_1d(temp_in, temp_out);
+ for (int j = 0; j < 8; ++j)
+ out2[j*8 + i] = temp_out[j];
+ }
+ for (int i = 0; i < 64; ++i)
+ output[i] = round(out2[i]/32);
+}
+
+TEST(VP8Idct8x8Test, AccuracyCheck) {+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = 10000;
+ for (int i = 0; i < count_test_block; ++i) {+ int16_t input[64], coeff[64];
+ int16_t output_c[64];
+ double output_r[64];
+
+ // Initialize a test block with input range [-255, 255].
+ for (int j = 0; j < 64; ++j)
+ input[j] = rnd.Rand8() - rnd.Rand8();
+
+ const int pitch = 16;
+ vp8_short_fdct8x8_c(input, output_c, pitch);
+ reference_dct_2d(input, output_r);
+
+ for (int j = 0; j < 64; ++j) {+ const double diff = output_c[j] - output_r[j];
+ const double error = diff * diff;
+ // An error in a DCT coefficient isn't that bad.
+ // We care more about the reconstructed pixels.
+ EXPECT_GE(2.0, error)
+ << "Error: 8x8 FDCT/IDCT has error " << error
+ << " at index " << j;
+ }
+
+#if 0
+ // Tests that the reference iDCT and fDCT match.
+ reference_dct_2d(input, output_r);
+ reference_idct_2d(output_r, output_c);
+ for (int j = 0; j < 64; ++j) {+ const int diff = output_c[j] -input[j];
+ const int error = diff * diff;
+ EXPECT_EQ(0, error)
+ << "Error: 8x8 FDCT/IDCT has error " << error
+ << " at index " << j;
+ }
+#endif
+ reference_dct_2d(input, output_r);
+ for (int j = 0; j < 64; ++j)
+ coeff[j] = round(output_r[j]);
+ vp8_short_idct8x8_c(coeff, output_c, pitch);
+ for (int j = 0; j < 64; ++j) {+ const int diff = output_c[j] -input[j];
+ const int error = diff * diff;
+ EXPECT_GE(1, error)
+ << "Error: 8x8 FDCT/IDCT has error " << error
+ << " at index " << j;
+ }
+ }
+}
+
+} // namespace
--- a/test/test.mk
+++ b/test/test.mk
@@ -1,8 +1,10 @@
LIBVPX_TEST_SRCS-yes += test.mk
LIBVPX_TEST_SRCS-yes += acm_random.h
LIBVPX_TEST_SRCS-yes += boolcoder_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_TX16X16) += dct16x16_test.cc
LIBVPX_TEST_SRCS-yes += fdct4x4_test.cc
LIBVPX_TEST_SRCS-yes += fdct8x8_test.cc
+LIBVPX_TEST_SRCS-yes += idct8x8_test.cc
LIBVPX_TEST_SRCS-yes += test_libvpx.cc
LIBVPX_TEST_DATA-yes += hantro_collage_w352h288.yuv
--- a/vp8/common/alloccommon.c
+++ b/vp8/common/alloccommon.c
@@ -218,7 +218,4 @@
vp8_entropy_mode_init();
vp8_entropy_mv_init();
-
- vp8_init_scan_order_mask();
-
}
--- a/vp8/common/blockd.h
+++ b/vp8/common/blockd.h
@@ -131,11 +131,12 @@
// Segment level features.
typedef enum {- TX_4X4 = 0, // 4x4 dct transform
- TX_8X8 = 1, // 8x8 dct transform
-
- TX_SIZE_MAX = 2 // Number of differnt transforms avaialble
-
+ TX_4X4, // 4x4 dct transform
+ TX_8X8, // 8x8 dct transform
+#if CONFIG_TX16X16
+ TX_16X16, // 16x16 dct transform
+#endif
+ TX_SIZE_MAX // Number of different transforms available
} TX_SIZE;
#if CONFIG_HYBRIDTRANSFORM
--- a/vp8/common/coefupdateprobs.h
+++ b/vp8/common/coefupdateprobs.h
@@ -13,4 +13,7 @@
Generated file included by entropy.c */
#define COEF_UPDATE_PROB 252
#define COEF_UPDATE_PROB_8X8 252
+#if CONFIG_TX16X16
+#define COEF_UPDATE_PROB_16X16 252
+#endif
--- a/vp8/common/default_coef_probs.h
+++ b/vp8/common/default_coef_probs.h
@@ -488,3 +488,211 @@
}
#endif
};
+#if CONFIG_TX16X16
+static const vp8_prob
+vp8_default_coef_probs_16x16[BLOCK_TYPES_16X16]
+ [COEF_BANDS]
+ [PREV_COEF_CONTEXTS]
+ [ENTROPY_NODES] =
+{+ { /* block Type 0 */+ { /* Coeff Band 0 */+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128}+ },
+ { /* Coeff Band 1 */+ { 60, 140, 195, 255, 212, 214, 128, 128, 128, 128, 128},+ { 75, 221, 231, 255, 203, 255, 128, 128, 128, 128, 128},+ { 9, 212, 196, 251, 197, 207, 255, 185, 128, 128, 128},+ { 9, 212, 196, 251, 197, 207, 255, 185, 128, 128, 128}+ },
+ { /* Coeff Band 2 */+ { 1, 227, 226, 255, 215, 215, 128, 128, 128, 128, 128},+ { 5, 163, 209, 255, 212, 212, 255, 255, 128, 128, 128},+ { 1, 133, 203, 255, 210, 220, 255, 255, 128, 128, 128},+ { 1, 133, 203, 255, 210, 220, 255, 255, 128, 128, 128}+ },
+ { /* Coeff Band 3 */+ { 1, 226, 225, 255, 228, 236, 128, 128, 128, 128, 128},+ { 6, 163, 208, 255, 224, 234, 255, 255, 128, 128, 128},+ { 1, 122, 196, 253, 212, 248, 255, 255, 128, 128, 128},+ { 1, 122, 196, 253, 212, 248, 255, 255, 128, 128, 128}+ },
+ { /* Coeff Band 4 */+ { 1, 222, 197, 254, 193, 216, 255, 236, 128, 128, 128},+ { 7, 140, 163, 251, 195, 211, 255, 238, 128, 128, 128},+ { 1, 91, 152, 249, 181, 197, 255, 239, 128, 128, 128},+ { 1, 91, 152, 249, 181, 197, 255, 239, 128, 128, 128}+ },
+ { /* Coeff Band 5 */+ { 1, 226, 218, 255, 216, 241, 255, 255, 128, 128, 128},+ { 6, 154, 191, 255, 218, 240, 255, 255, 128, 128, 128},+ { 1, 110, 171, 252, 191, 204, 255, 236, 128, 128, 128},+ { 1, 110, 171, 252, 191, 204, 255, 236, 128, 128, 128}+ },
+ { /* Coeff Band 6 */+ { 1, 221, 217, 255, 208, 217, 255, 232, 128, 128, 128},+ { 11, 155, 189, 254, 203, 211, 255, 249, 128, 128, 128},+ { 1, 110, 171, 252, 191, 204, 255, 236, 128, 128, 128},+ { 1, 110, 171, 252, 191, 204, 255, 236, 128, 128, 128}+ },
+ { /* Coeff Band 7 */+ { 1, 207, 235, 255, 232, 240, 128, 128, 128, 128, 128},+ { 58, 161, 216, 255, 229, 235, 255, 255, 128, 128, 128},+ { 8, 133, 204, 255, 219, 231, 255, 255, 128, 128, 128},+ { 8, 133, 204, 255, 219, 231, 255, 255, 128, 128, 128}+ }
+ },
+ { /* block Type 1 */+ { /* Coeff Band 0 */+ { 134, 152, 233, 224, 234, 52, 255, 166, 128, 128, 128},+ { 97, 132, 185, 234, 186, 189, 197, 171, 255, 212, 128},+ { 84, 110, 185, 237, 182, 182, 145, 145, 255, 255, 128}+ },
+ { /* Coeff Band 1 */+ { 1, 124, 213, 247, 192, 212, 255, 255, 128, 128, 128},+ { 88, 111, 178, 254, 189, 211, 255, 255, 128, 128, 128},+ { 12, 59, 129, 236, 150, 179, 239, 195, 255, 255, 128},+ { 12, 59, 129, 236, 150, 179, 239, 195, 255, 255, 128}+ },
+ { /* Coeff Band 2 */+ { 1, 102, 225, 255, 210, 240, 128, 128, 128, 128, 128},+ { 110, 78, 195, 254, 200, 191, 255, 255, 128, 128, 128},+ { 37, 63, 177, 255, 194, 195, 128, 128, 128, 128, 128},+ { 37, 63, 177, 255, 194, 195, 128, 128, 128, 128, 128}+ },
+ { /* Coeff Band 3 */+ { 1, 1, 229, 255, 202, 224, 128, 128, 128, 128, 128},+ { 150, 1, 192, 255, 206, 226, 128, 128, 128, 128, 128},+ { 75, 1, 138, 255, 172, 228, 128, 128, 128, 128, 128},+ { 75, 1, 138, 255, 172, 228, 128, 128, 128, 128, 128}+ },
+ { /* Coeff Band 4 */+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128}+ },
+ { /* Coeff Band 5 */+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128}+ },
+ { /* Coeff Band 6 */+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128}+ },
+ { /* Coeff Band 7 */+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128}+ }
+ },
+ { /* block Type 2 */+ { /* Coeff Band 0 */+ { 11, 181, 226, 199, 183, 255, 255, 255, 128, 128, 128},+ { 2, 147, 185, 248, 163, 180, 255, 236, 128, 128, 128},+ { 1, 123, 157, 238, 154, 176, 255, 226, 255, 255, 128},+ { 1, 123, 157, 238, 154, 176, 255, 226, 255, 255, 128}+ },
+ { /* Coeff Band 1 */+ { 1, 150, 191, 246, 174, 188, 255, 235, 128, 128, 128},+ { 1, 125, 166, 245, 165, 185, 255, 234, 128, 128, 128},+ { 1, 79, 125, 240, 148, 179, 255, 234, 255, 255, 128},+ { 1, 79, 125, 240, 148, 179, 255, 234, 255, 255, 128}+ },
+ { /* Coeff Band 2 */+ { 1, 146, 184, 242, 167, 183, 255, 230, 255, 255, 128},+ { 1, 119, 160, 239, 156, 178, 255, 231, 255, 255, 128},+ { 1, 75, 115, 234, 142, 173, 255, 225, 255, 255, 128},+ { 1, 75, 115, 234, 142, 173, 255, 225, 255, 255, 128}+ },
+ { /* Coeff Band 3 */+ { 1, 150, 188, 244, 169, 183, 255, 233, 255, 255, 128},+ { 1, 123, 162, 243, 161, 180, 255, 233, 128, 128, 128},+ { 1, 76, 120, 238, 148, 178, 255, 230, 255, 255, 128},+ { 1, 76, 120, 238, 148, 178, 255, 230, 255, 255, 128}+ },
+ { /* Coeff Band 4 */+ { 1, 163, 202, 252, 188, 204, 255, 248, 128, 128, 128},+ { 1, 136, 180, 251, 181, 201, 255, 246, 128, 128, 128},+ { 1, 92, 146, 249, 170, 197, 255, 245, 128, 128, 128},+ { 1, 92, 146, 249, 170, 197, 255, 245, 128, 128, 128}+ },
+ { /* Coeff Band 5 */+ { 1, 156, 195, 249, 179, 193, 255, 241, 255, 255, 128},+ { 1, 128, 169, 248, 171, 192, 255, 242, 255, 255, 128},+ { 1, 84, 132, 245, 158, 187, 255, 240, 255, 255, 128},+ { 1, 84, 132, 245, 158, 187, 255, 240, 255, 255, 128}+ },
+ { /* Coeff Band 6 */+ { 1, 36, 71, 251, 192, 201, 255, 243, 255, 255, 128},+ { 1, 49, 185, 250, 184, 199, 255, 242, 128, 128, 128},+ { 1, 95, 147, 247, 168, 190, 255, 239, 255, 255, 128},+ { 1, 95, 147, 247, 168, 190, 255, 239, 255, 255, 128}+ },
+ { /* Coeff Band 7 */+ { 1, 19, 98, 255, 218, 222, 255, 255, 128, 128, 128},+ { 36, 50, 210, 255, 212, 221, 255, 255, 128, 128, 128},+ { 6, 117, 180, 254, 199, 216, 255, 251, 128, 128, 128},+ { 6, 117, 180, 254, 199, 216, 255, 251, 128, 128, 128}+ }
+ },
+ { /* block Type 3 */+ { /* Coeff Band 0 */+ { 17, 105, 227, 195, 164, 170, 168, 137, 221, 160, 184},+ { 6, 92, 166, 193, 158, 169, 179, 142, 236, 175, 200},+ { 2, 68, 118, 193, 147, 168, 187, 149, 241, 178, 247},+ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128}+ },
+ { /* Coeff Band 1 */+ { 1, 193, 221, 246, 198, 194, 244, 176, 255, 192, 128},+ { 112, 160, 209, 244, 196, 194, 243, 175, 255, 209, 128},+ { 45, 123, 175, 240, 184, 195, 239, 178, 255, 218, 255},+ { 16, 53, 75, 169, 119, 152, 209, 146, 255, 219, 255}+ },
+ { /* Coeff Band 2 */+ { 1, 141, 183, 240, 176, 187, 246, 198, 255, 218, 128},+ { 36, 97, 150, 231, 161, 180, 243, 191, 255, 217, 255},+ { 8, 65, 111, 210, 143, 166, 230, 167, 255, 224, 255},+ { 2, 35, 61, 157, 113, 149, 208, 142, 255, 217, 255}+ },
+ { /* Coeff Band 3 */+ { 1, 173, 196, 245, 184, 191, 252, 211, 255, 240, 128},+ { 35, 119, 175, 242, 177, 187, 252, 209, 255, 235, 128},+ { 4, 88, 141, 234, 161, 180, 249, 200, 255, 228, 128},+ { 1, 57, 95, 203, 133, 161, 235, 167, 255, 231, 255}+ },
+ { /* Coeff Band 4 */+ { 1, 208, 227, 249, 209, 204, 248, 188, 255, 248, 128},+ { 28, 162, 211, 247, 203, 200, 252, 188, 255, 232, 128},+ { 5, 114, 174, 238, 182, 189, 245, 184, 255, 238, 128},+ { 1, 61, 100, 205, 136, 164, 235, 163, 255, 239, 128}+ },
+ { /* Coeff Band 5 */+ { 1, 195, 218, 252, 208, 207, 250, 205, 255, 245, 128},+ { 22, 141, 196, 249, 198, 201, 250, 202, 255, 244, 128},+ { 2, 105, 163, 240, 178, 189, 246, 191, 255, 246, 128},+ { 1, 70, 112, 206, 144, 167, 232, 162, 255, 239, 128}+ },
+ { /* Coeff Band 6 */+ { 1, 204, 215, 251, 204, 203, 255, 222, 255, 225, 128},+ { 15, 140, 194, 249, 194, 199, 254, 221, 255, 253, 128},+ { 1, 95, 153, 243, 172, 188, 254, 213, 255, 248, 128},+ { 1, 59, 99, 216, 135, 166, 247, 190, 255, 237, 255}+ },
+ { /* Coeff Band 7 */+ { 1, 7, 231, 255, 227, 223, 255, 240, 255, 255, 128},+ { 15, 157, 217, 255, 218, 219, 255, 239, 255, 255, 128},+ { 1, 114, 182, 252, 198, 207, 255, 235, 255, 255, 128},+ { 1, 71, 122, 238, 154, 181, 255, 216, 255, 255, 128}+ }
+ }
+};
+#endif
--- a/vp8/common/entropy.c
+++ b/vp8/common/entropy.c
@@ -47,7 +47,7 @@
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
-DECLARE_ALIGNED(16, cuchar, vp8_coef_bands[16]) = {+DECLARE_ALIGNED(16, const int, vp8_coef_bands[16]) = {0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 7, 7, 7
};
@@ -79,15 +79,15 @@
#endif
-DECLARE_ALIGNED(64, cuchar, vp8_coef_bands_8x8[64]) = { 0, 1, 2, 3, 5, 4, 4, 5,- 5, 3, 6, 3, 5, 4, 6, 6,
- 6, 5, 5, 6, 6, 6, 6, 6,
- 6, 6, 6, 6, 6, 6, 6, 6,
- 6, 6, 6, 6, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7
- };
+DECLARE_ALIGNED(64, const int, vp8_coef_bands_8x8[64]) = { 0, 1, 2, 3, 5, 4, 4, 5,+ 5, 3, 6, 3, 5, 4, 6, 6,
+ 6, 5, 5, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7
+ };
DECLARE_ALIGNED(64, const int, vp8_default_zig_zag1d_8x8[64]) = {0, 1, 8, 16, 9, 2, 3, 10, 17, 24, 32, 25, 18, 11, 4, 5,
12, 19, 26, 33, 40, 48, 41, 34, 27, 20, 13, 6, 7, 14, 21, 28,
@@ -95,9 +95,46 @@
58, 59, 52, 45, 38, 31, 39, 46, 53, 60, 61, 54, 47, 55, 62, 63,
};
+#if CONFIG_TX16X16
+// Table can be optimized.
+DECLARE_ALIGNED(16, const int, vp8_coef_bands_16x16[256]) = {+ 0, 1, 2, 3, 5, 4, 4, 5, 5, 3, 6, 3, 5, 4, 6, 6,
+ 6, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+};
+DECLARE_ALIGNED(16, const int, vp8_default_zig_zag1d_16x16[256]) = {+ 0, 1, 16, 32, 17, 2, 3, 18, 33, 48, 64, 49, 34, 19, 4, 5,
+ 20, 35, 50, 65, 80, 96, 81, 66, 51, 36, 21, 6, 7, 22, 37, 52,
+ 67, 82, 97, 112, 128, 113, 98, 83, 68, 53, 38, 23, 8, 9, 24, 39,
+ 54, 69, 84, 99, 114, 129, 144, 160, 145, 130, 115, 100, 85, 70, 55, 40,
+ 25, 10, 11, 26, 41, 56, 71, 86, 101, 116, 131, 146, 161, 176, 192, 177,
+ 162, 147, 132, 117, 102, 87, 72, 57, 42, 27, 12, 13, 28, 43, 58, 73,
+ 88, 103, 118, 133, 148, 163, 178, 193, 208, 224, 209, 194, 179, 164, 149, 134,
+ 119, 104, 89, 74, 59, 44, 29, 14, 15, 30, 45, 60, 75, 90, 105, 120,
+ 135, 150, 165, 180, 195, 210, 225, 240, 241, 226, 211, 196, 181, 166, 151, 136,
+ 121, 106, 91, 76, 61, 46, 31, 47, 62, 77, 92, 107, 122, 137, 152, 167,
+ 182, 197, 212, 227, 242, 243, 228, 213, 198, 183, 168, 153, 138, 123, 108, 93,
+ 78, 63, 79, 94, 109, 124, 139, 154, 169, 184, 199, 214, 229, 244, 245, 230,
+ 215, 200, 185, 170, 155, 140, 125, 110, 95, 111, 126, 141, 156, 171, 186, 201,
+ 216, 231, 246, 247, 232, 217, 202, 187, 172, 157, 142, 127, 143, 158, 173, 188,
+ 203, 218, 233, 248, 249, 234, 219, 204, 189, 174, 159, 175, 190, 205, 220, 235,
+ 250, 251, 236, 221, 206, 191, 207, 222, 237, 252, 253, 238, 223, 239, 254, 255,
+};
+#endif
-DECLARE_ALIGNED(16, short, vp8_default_zig_zag_mask[16]);
-DECLARE_ALIGNED(64, short, vp8_default_zig_zag_mask_8x8[64]);// int64_t
/* Array indices are identical to previously-existing CONTEXT_NODE indices */
@@ -131,17 +168,6 @@
static vp8_tree_index cat1[2], cat2[4], cat3[6], cat4[8], cat5[10], cat6[26];
-void vp8_init_scan_order_mask() {- int i;
-
- for (i = 0; i < 16; i++) {- vp8_default_zig_zag_mask[vp8_default_zig_zag1d[i]] = 1 << i;
- }
- for (i = 0; i < 64; i++) {- vp8_default_zig_zag_mask_8x8[vp8_default_zig_zag1d_8x8[i]] = 1 << i;
- }
-}
-
static void init_bit_tree(vp8_tree_index *p, int n) {int i = 0;
@@ -181,11 +207,15 @@
void vp8_default_coef_probs(VP8_COMMON *pc) {vpx_memcpy(pc->fc.coef_probs, default_coef_probs,
- sizeof(default_coef_probs));
+ sizeof(pc->fc.coef_probs));
vpx_memcpy(pc->fc.coef_probs_8x8, vp8_default_coef_probs_8x8,
- sizeof(vp8_default_coef_probs_8x8));
+ sizeof(pc->fc.coef_probs_8x8));
+#if CONFIG_TX16X16
+ vpx_memcpy(pc->fc.coef_probs_16x16, vp8_default_coef_probs_16x16,
+ sizeof(pc->fc.coef_probs_16x16));
+#endif
}
void vp8_coef_tree_initialize() {@@ -304,4 +334,27 @@
else cm->fc.coef_probs_8x8[i][j][k][t] = prob;
}
}
+
+#if CONFIG_TX16X16
+ for (i = 0; i < BLOCK_TYPES_16X16; ++i)
+ for (j = 0; j < COEF_BANDS; ++j)
+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {+ if (k >= 3 && ((i == 0 && j == 1) || (i > 0 && j == 0)))
+ continue;
+ vp8_tree_probs_from_distribution(
+ MAX_ENTROPY_TOKENS, vp8_coef_encodings, vp8_coef_tree,
+ coef_probs, branch_ct, cm->fc.coef_counts_16x16[i][j][k], 256, 1);
+ for (t = 0; t < ENTROPY_NODES; ++t) {+ int prob;
+ count = branch_ct[t][0] + branch_ct[t][1];
+ count = count > count_sat ? count_sat : count;
+ factor = (update_factor * count / count_sat);
+ prob = ((int)cm->fc.pre_coef_probs_16x16[i][j][k][t] * (256 - factor) +
+ (int)coef_probs[t] * factor + 128) >> 8;
+ if (prob <= 0) cm->fc.coef_probs_16x16[i][j][k][t] = 1;
+ else if (prob > 255) cm->fc.coef_probs_16x16[i][j][k][t] = 255;
+ else cm->fc.coef_probs_16x16[i][j][k][t] = prob;
+ }
+ }
+#endif
}
--- a/vp8/common/entropy.h
+++ b/vp8/common/entropy.h
@@ -62,19 +62,22 @@
/* Outside dimension. 0 = Y no DC, 1 = Y2, 2 = UV, 3 = Y with DC */
#define BLOCK_TYPES 4
-
#if CONFIG_HTRANS8X8
#define BLOCK_TYPES_8X8 4
#else
#define BLOCK_TYPES_8X8 3
#endif
+#define BLOCK_TYPES_16X16 4
/* Middle dimension is a coarsening of the coefficient's
position within the 4x4 DCT. */
#define COEF_BANDS 8
-extern DECLARE_ALIGNED(16, const unsigned char, vp8_coef_bands[16]);
-extern DECLARE_ALIGNED(64, const unsigned char, vp8_coef_bands_8x8[64]);
+extern DECLARE_ALIGNED(16, const int, vp8_coef_bands[16]);
+extern DECLARE_ALIGNED(64, const int, vp8_coef_bands_8x8[64]);
+#if CONFIG_TX16X16
+extern DECLARE_ALIGNED(16, const int, vp8_coef_bands_16x16[256]);
+#endif
/* Inside dimension is 3-valued measure of nearby complexity, that is,
the extent to which nearby coefficients are nonzero. For the first
@@ -113,8 +116,11 @@
extern short vp8_default_zig_zag_mask[16];
extern DECLARE_ALIGNED(64, const int, vp8_default_zig_zag1d_8x8[64]);
-extern short vp8_default_zig_zag_mask_8x8[64];// int64_t
void vp8_coef_tree_initialize(void);
+#if CONFIG_TX16X16
+extern DECLARE_ALIGNED(16, const int, vp8_default_zig_zag1d_16x16[256]);
+#endif
void vp8_adapt_coef_probs(struct VP8Common *);
+
#endif
--- a/vp8/common/entropymode.c
+++ b/vp8/common/entropymode.c
@@ -249,11 +249,8 @@
void vp8_init_mbmode_probs(VP8_COMMON *x) {unsigned int bct [VP8_YMODES] [2]; /* num Ymodes > num UV modes */
- vp8_tree_probs_from_distribution(
- VP8_YMODES, vp8_ymode_encodings, vp8_ymode_tree,
- x->fc.ymode_prob, bct, y_mode_cts,
- 256, 1
- );
+ vp8_tree_probs_from_distribution(VP8_YMODES, vp8_ymode_encodings,
+ vp8_ymode_tree, x->fc.ymode_prob, bct, y_mode_cts, 256, 1);
{int i;
for (i = 0; i < 8; i++)
@@ -260,8 +257,7 @@
vp8_tree_probs_from_distribution(
VP8_YMODES, vp8_kf_ymode_encodings, vp8_kf_ymode_tree,
x->kf_ymode_prob[i], bct, kf_y_mode_cts[i],
- 256, 1
- );
+ 256, 1);
}
{int i;
@@ -295,13 +291,9 @@
static void intra_bmode_probs_from_distribution(
vp8_prob p [VP8_BINTRAMODES - 1],
unsigned int branch_ct [VP8_BINTRAMODES - 1] [2],
- const unsigned int events [VP8_BINTRAMODES]
-) {- vp8_tree_probs_from_distribution(
- VP8_BINTRAMODES, vp8_bmode_encodings, vp8_bmode_tree,
- p, branch_ct, events,
- 256, 1
- );
+ const unsigned int events [VP8_BINTRAMODES]) {+ vp8_tree_probs_from_distribution(VP8_BINTRAMODES, vp8_bmode_encodings,
+ vp8_bmode_tree, p, branch_ct, events, 256, 1);
}
void vp8_default_bmode_probs(vp8_prob p [VP8_BINTRAMODES - 1]) {--- a/vp8/common/generic/systemdependent.c
+++ b/vp8/common/generic/systemdependent.c
@@ -32,6 +32,9 @@
rtcd->idct.idct8 = vp8_short_idct8x8_c;
rtcd->idct.idct1_scalar_add_8x8 = vp8_dc_only_idct_add_8x8_c;
rtcd->idct.ihaar2 = vp8_short_ihaar2x2_c;
+#if CONFIG_TX16X16
+ rtcd->idct.idct16x16 = vp8_short_idct16x16_c;
+#endif
rtcd->recon.copy16x16 = vp8_copy_mem16x16_c;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_c;
rtcd->recon.avg16x16 = vp8_avg_mem16x16_c;
--- a/vp8/common/idct.h
+++ b/vp8/common/idct.h
@@ -36,6 +36,13 @@
#define Y2_WHT_UPSCALE_FACTOR 2
#endif
+#if CONFIG_TX16X16
+#ifndef vp8_idct_idct16x16
+#define vp8_idct_idct16x16 vp8_short_idct16x16_c
+#endif
+extern prototype_idct(vp8_idct_idct16x16);
+#endif
+
#ifndef vp8_idct_idct8
#define vp8_idct_idct8 vp8_short_idct8x8_c
#endif
@@ -120,6 +127,10 @@
vp8_idct_scalar_add_fn_t idct1_scalar_add_8x8;
vp8_idct_fn_t ihaar2;
vp8_idct_fn_t ihaar2_1;
+
+#if CONFIG_TX16X16
+ vp8_idct_fn_t idct16x16;
+#endif
} vp8_idct_rtcd_vtable_t;
#if CONFIG_RUNTIME_CPU_DETECT
--- a/vp8/common/idctllm.c
+++ b/vp8/common/idctllm.c
@@ -647,3 +647,275 @@
op[8] = (ip[0] - ip[1] - ip[4] + ip[8]) >> 1;
}
+
+#if CONFIG_TX16X16
+#if 0
+// Keep a really bad float version as reference for now.
+void vp8_short_idct16x16_c(short *input, short *output, int pitch) {+ double x;
+ const int short_pitch = pitch >> 1;
+ int i, j, k, l;
+ for (l = 0; l < 16; ++l) {+ for (k = 0; k < 16; ++k) {+ double s = 0;
+ for (i = 0; i < 16; ++i) {+ for (j = 0; j < 16; ++j) {+ x=cos(PI*j*(l+0.5)/16.0)*cos(PI*i*(k+0.5)/16.0)*input[i*16+j]/32;
+ if (i != 0)
+ x *= sqrt(2.0);
+ if (j != 0)
+ x *= sqrt(2.0);
+ s += x;
+ }
+ }
+ output[k*short_pitch+l] = (short)round(s);
+ }
+ }
+}
+#endif
+
+static void butterfly_16x16_idct_1d(double input[16], double output[16]) {+ double step[16];
+ double intermediate[16];
+ double temp1, temp2;
+
+ const double PI = M_PI;
+ const double C1 = cos(1*PI/(double)32);
+ const double C2 = cos(2*PI/(double)32);
+ const double C3 = cos(3*PI/(double)32);
+ const double C4 = cos(4*PI/(double)32);
+ const double C5 = cos(5*PI/(double)32);
+ const double C6 = cos(6*PI/(double)32);
+ const double C7 = cos(7*PI/(double)32);
+ const double C8 = cos(8*PI/(double)32);
+ const double C9 = cos(9*PI/(double)32);
+ const double C10 = cos(10*PI/(double)32);
+ const double C11 = cos(11*PI/(double)32);
+ const double C12 = cos(12*PI/(double)32);
+ const double C13 = cos(13*PI/(double)32);
+ const double C14 = cos(14*PI/(double)32);
+ const double C15 = cos(15*PI/(double)32);
+
+ // step 1 and 2
+ step[ 0] = input[0] + input[8];
+ step[ 1] = input[0] - input[8];
+
+ temp1 = input[4]*C12;
+ temp2 = input[12]*C4;
+
+ temp1 -= temp2;
+ temp1 *= C8;
+
+ step[ 2] = 2*(temp1);
+
+ temp1 = input[4]*C4;
+ temp2 = input[12]*C12;
+ temp1 += temp2;
+ temp1 = (temp1);
+ temp1 *= C8;
+ step[ 3] = 2*(temp1);
+
+ temp1 = input[2]*C8;
+ temp1 = 2*(temp1);
+ temp2 = input[6] + input[10];
+
+ step[ 4] = temp1 + temp2;
+ step[ 5] = temp1 - temp2;
+
+ temp1 = input[14]*C8;
+ temp1 = 2*(temp1);
+ temp2 = input[6] - input[10];
+
+ step[ 6] = temp2 - temp1;
+ step[ 7] = temp2 + temp1;
+
+ // for odd input
+ temp1 = input[3]*C12;
+ temp2 = input[13]*C4;
+ temp1 += temp2;
+ temp1 = (temp1);
+ temp1 *= C8;
+ intermediate[ 8] = 2*(temp1);
+
+ temp1 = input[3]*C4;
+ temp2 = input[13]*C12;
+ temp2 -= temp1;
+ temp2 = (temp2);
+ temp2 *= C8;
+ intermediate[ 9] = 2*(temp2);
+
+ intermediate[10] = 2*(input[9]*C8);
+ intermediate[11] = input[15] - input[1];
+ intermediate[12] = input[15] + input[1];
+ intermediate[13] = 2*((input[7]*C8));
+
+ temp1 = input[11]*C12;
+ temp2 = input[5]*C4;
+ temp2 -= temp1;
+ temp2 = (temp2);
+ temp2 *= C8;
+ intermediate[14] = 2*(temp2);
+
+ temp1 = input[11]*C4;
+ temp2 = input[5]*C12;
+ temp1 += temp2;
+ temp1 = (temp1);
+ temp1 *= C8;
+ intermediate[15] = 2*(temp1);
+
+ step[ 8] = intermediate[ 8] + intermediate[14];
+ step[ 9] = intermediate[ 9] + intermediate[15];
+ step[10] = intermediate[10] + intermediate[11];
+ step[11] = intermediate[10] - intermediate[11];
+ step[12] = intermediate[12] + intermediate[13];
+ step[13] = intermediate[12] - intermediate[13];
+ step[14] = intermediate[ 8] - intermediate[14];
+ step[15] = intermediate[ 9] - intermediate[15];
+
+ // step 3
+ output[0] = step[ 0] + step[ 3];
+ output[1] = step[ 1] + step[ 2];
+ output[2] = step[ 1] - step[ 2];
+ output[3] = step[ 0] - step[ 3];
+
+ temp1 = step[ 4]*C14;
+ temp2 = step[ 7]*C2;
+ temp1 -= temp2;
+ output[4] = (temp1);
+
+ temp1 = step[ 4]*C2;
+ temp2 = step[ 7]*C14;
+ temp1 += temp2;
+ output[7] = (temp1);
+
+ temp1 = step[ 5]*C10;
+ temp2 = step[ 6]*C6;
+ temp1 -= temp2;
+ output[5] = (temp1);
+
+ temp1 = step[ 5]*C6;
+ temp2 = step[ 6]*C10;
+ temp1 += temp2;
+ output[6] = (temp1);
+
+ output[8] = step[ 8] + step[11];
+ output[9] = step[ 9] + step[10];
+ output[10] = step[ 9] - step[10];
+ output[11] = step[ 8] - step[11];
+ output[12] = step[12] + step[15];
+ output[13] = step[13] + step[14];
+ output[14] = step[13] - step[14];
+ output[15] = step[12] - step[15];
+
+ // output 4
+ step[ 0] = output[0] + output[7];
+ step[ 1] = output[1] + output[6];
+ step[ 2] = output[2] + output[5];
+ step[ 3] = output[3] + output[4];
+ step[ 4] = output[3] - output[4];
+ step[ 5] = output[2] - output[5];
+ step[ 6] = output[1] - output[6];
+ step[ 7] = output[0] - output[7];
+
+ temp1 = output[8]*C7;
+ temp2 = output[15]*C9;
+ temp1 -= temp2;
+ step[ 8] = (temp1);
+
+ temp1 = output[9]*C11;
+ temp2 = output[14]*C5;
+ temp1 += temp2;
+ step[ 9] = (temp1);
+
+ temp1 = output[10]*C3;
+ temp2 = output[13]*C13;
+ temp1 -= temp2;
+ step[10] = (temp1);
+
+ temp1 = output[11]*C15;
+ temp2 = output[12]*C1;
+ temp1 += temp2;
+ step[11] = (temp1);
+
+ temp1 = output[11]*C1;
+ temp2 = output[12]*C15;
+ temp2 -= temp1;
+ step[12] = (temp2);
+
+ temp1 = output[10]*C13;
+ temp2 = output[13]*C3;
+ temp1 += temp2;
+ step[13] = (temp1);
+
+ temp1 = output[9]*C5;
+ temp2 = output[14]*C11;
+ temp2 -= temp1;
+ step[14] = (temp2);
+
+ temp1 = output[8]*C9;
+ temp2 = output[15]*C7;
+ temp1 += temp2;
+ step[15] = (temp1);
+
+ // step 5
+ output[0] = (step[0] + step[15]);
+ output[1] = (step[1] + step[14]);
+ output[2] = (step[2] + step[13]);
+ output[3] = (step[3] + step[12]);
+ output[4] = (step[4] + step[11]);
+ output[5] = (step[5] + step[10]);
+ output[6] = (step[6] + step[ 9]);
+ output[7] = (step[7] + step[ 8]);
+
+ output[15] = (step[0] - step[15]);
+ output[14] = (step[1] - step[14]);
+ output[13] = (step[2] - step[13]);
+ output[12] = (step[3] - step[12]);
+ output[11] = (step[4] - step[11]);
+ output[10] = (step[5] - step[10]);
+ output[9] = (step[6] - step[ 9]);
+ output[8] = (step[7] - step[ 8]);
+}
+
+// Remove once an int version of iDCT is written
+#if 0
+void reference_16x16_idct_1d(double input[16], double output[16]) {+ const double kPi = 3.141592653589793238462643383279502884;
+ const double kSqrt2 = 1.414213562373095048801688724209698;
+ for (int k = 0; k < 16; k++) {+ output[k] = 0.0;
+ for (int n = 0; n < 16; n++) {+ output[k] += input[n]*cos(kPi*(2*k+1)*n/32.0);
+ if (n == 0)
+ output[k] = output[k]/kSqrt2;
+ }
+ }
+}
+#endif
+
+void vp8_short_idct16x16_c(short *input, short *output, int pitch) {+ double out[16*16], out2[16*16];
+ const int short_pitch = pitch >> 1;
+ int i, j;
+ // First transform rows
+ for (i = 0; i < 16; ++i) {+ double temp_in[16], temp_out[16];
+ for (j = 0; j < 16; ++j)
+ temp_in[j] = input[j + i*short_pitch];
+ butterfly_16x16_idct_1d(temp_in, temp_out);
+ for (j = 0; j < 16; ++j)
+ out[j + i*16] = temp_out[j];
+ }
+ // Then transform columns
+ for (i = 0; i < 16; ++i) {+ double temp_in[16], temp_out[16];
+ for (j = 0; j < 16; ++j)
+ temp_in[j] = out[j*16 + i];
+ butterfly_16x16_idct_1d(temp_in, temp_out);
+ for (j = 0; j < 16; ++j)
+ out2[j*16 + i] = temp_out[j];
+ }
+ for (i = 0; i < 16*16; ++i)
+ output[i] = round(out2[i]/128);
+}
+#endif
--- a/vp8/common/invtrans.c
+++ b/vp8/common/invtrans.c
@@ -153,3 +153,33 @@
}
+#if CONFIG_TX16X16
+void vp8_inverse_transform_b_16x16(const vp8_idct_rtcd_vtable_t *rtcd,
+ short *input_dqcoeff,
+ short *output_coeff, int pitch) {+ IDCT_INVOKE(rtcd, idct16x16)(input_dqcoeff, output_coeff, pitch);
+}
+
+void vp8_inverse_transform_mby_16x16(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x) {+ vp8_inverse_transform_b_16x16(rtcd, &x->block[0].dqcoeff[0], &x->block[0].diff[0], 32);
+}
+
+// U,V blocks are 8x8 per macroblock, so just run 8x8
+void vp8_inverse_transform_mbuv_16x16(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x) {+ int i;
+ for (i = 16; i < 24; i += 4)
+ vp8_inverse_transform_b_8x8(rtcd, &x->block[i].dqcoeff[0], &x->block[i].diff[0], 16);
+}
+
+void vp8_inverse_transform_mb_16x16(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x) {+ int i;
+
+ // Luma
+ vp8_inverse_transform_b_16x16(rtcd, &x->block[0].dqcoeff[0], &x->block[0].diff[0], 32);
+
+ // U, V
+ // Chroma blocks are downscaled, so run an 8x8 on them.
+ for (i = 16; i < 24; i+= 4)
+ vp8_inverse_transform_b_8x8(rtcd, &x->block[i].dqcoeff[0], &x->block[i].diff[0], 16);
+}
+#endif
--- a/vp8/common/invtrans.h
+++ b/vp8/common/invtrans.h
@@ -30,4 +30,12 @@
extern void vp8_inverse_transform_mby_8x8(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x);
extern void vp8_inverse_transform_mbuv_8x8(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x);
+#if CONFIG_TX16X16
+extern void vp8_inverse_transform_b_16x16(const vp8_idct_rtcd_vtable_t *rtcd,
+ short *input_dqcoeff, short *output_coeff,
+ int pitch);
+extern void vp8_inverse_transform_mb_16x16(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x);
+extern void vp8_inverse_transform_mby_16x16(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x);
+extern void vp8_inverse_transform_mbuv_16x16(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x);
+#endif
#endif
--- a/vp8/common/loopfilter.c
+++ b/vp8/common/loopfilter.c
@@ -329,7 +329,11 @@
vp8_loop_filter_mbv_c
(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
- if (!skip_lf) {+ if (!skip_lf
+#if CONFIG_TX16X16
+ && tx_type != TX_16X16
+#endif
+ ) {if (tx_type == TX_8X8)
vp8_loop_filter_bv8x8_c
(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
@@ -344,7 +348,11 @@
vp8_loop_filter_mbh_c
(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
- if (!skip_lf) {+ if (!skip_lf
+#if CONFIG_TX16X16
+ && tx_type != TX_16X16
+#endif
+ ) {if (tx_type == TX_8X8)
vp8_loop_filter_bh8x8_c
(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
@@ -353,6 +361,7 @@
(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
}
} else {+ // FIXME: Not 8x8 aware
if (mb_col > 0)
LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
(y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
@@ -431,7 +440,6 @@
const int seg = mode_info_context->mbmi.segment_id;
const int ref_frame = mode_info_context->mbmi.ref_frame;
int tx_type = mode_info_context->mbmi.txfm_size;
-
filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
if (filter_level) {@@ -446,7 +454,11 @@
vp8_loop_filter_mbv_c
(y_ptr, 0, 0, post->y_stride, 0, &lfi);
- if (!skip_lf) {+ if (!skip_lf
+#if CONFIG_TX16X16
+ && tx_type != TX_16X16
+#endif
+ ) {if (tx_type == TX_8X8)
vp8_loop_filter_bv8x8_c
(y_ptr, 0, 0, post->y_stride, 0, &lfi);
@@ -460,7 +472,11 @@
vp8_loop_filter_mbh_c
(y_ptr, 0, 0, post->y_stride, 0, &lfi);
- if (!skip_lf) {+ if (!skip_lf
+#if CONFIG_TX16X16
+ && tx_type != TX_16X16
+#endif
+ ) {if (tx_type == TX_8X8)
vp8_loop_filter_bh8x8_c
(y_ptr, 0, 0, post->y_stride, 0, &lfi);
@@ -469,6 +485,7 @@
(y_ptr, 0, 0, post->y_stride, 0, &lfi);
}
} else {+ // FIXME: Not 8x8 aware
if (mb_col > 0)
LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
(y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
--- a/vp8/common/onyxc_int.h
+++ b/vp8/common/onyxc_int.h
@@ -52,6 +52,9 @@
vp8_prob mbsplit_prob [VP8_NUMMBSPLITS - 1];
vp8_prob coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp8_prob coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#if CONFIG_TX16X16
+ vp8_prob coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#endif
MV_CONTEXT mvc[2];
#if CONFIG_HIGH_PRECISION_MV
MV_CONTEXT_HP mvc_hp[2];
@@ -73,12 +76,22 @@
unsigned int sub_mv_ref_counts [SUBMVREF_COUNT][VP8_SUBMVREFS];
unsigned int mbsplit_counts [VP8_NUMMBSPLITS];
- vp8_prob pre_coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
- vp8_prob pre_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+ vp8_prob pre_coef_probs [BLOCK_TYPES] [COEF_BANDS]
+ [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+ vp8_prob pre_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS]
+ [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#if CONFIG_TX16X16
+ vp8_prob pre_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
+ [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#endif
unsigned int coef_counts [BLOCK_TYPES] [COEF_BANDS]
- [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
+ [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
unsigned int coef_counts_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS]
- [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
+ [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
+#if CONFIG_TX16X16
+ unsigned int coef_counts_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
+ [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
+#endif
unsigned int MVcount [2] [MVvals];
#if CONFIG_HIGH_PRECISION_MV
unsigned int MVcount_hp [2] [MVvals_hp];
--- a/vp8/decoder/decodframe.c
+++ b/vp8/decoder/decodframe.c
@@ -217,22 +217,46 @@
#endif
if (pbi->common.frame_type == KEY_FRAME) {- if (pbi->common.txfm_mode == ALLOW_8X8 &&
+#if CONFIG_TX16X16
+ if (xd->mode_info_context->mbmi.mode <= TM_PRED ||
+ xd->mode_info_context->mbmi.mode == NEWMV ||
+ xd->mode_info_context->mbmi.mode == ZEROMV ||
+ xd->mode_info_context->mbmi.mode == NEARMV ||
+ xd->mode_info_context->mbmi.mode == NEARESTMV)
+ xd->mode_info_context->mbmi.txfm_size = TX_16X16;
+ else if (pbi->common.txfm_mode == ALLOW_8X8 &&
xd->mode_info_context->mbmi.mode != I8X8_PRED &&
xd->mode_info_context->mbmi.mode != B_PRED)
+#else
+ if (pbi->common.txfm_mode == ALLOW_8X8 &&
+ xd->mode_info_context->mbmi.mode != I8X8_PRED &&
+ xd->mode_info_context->mbmi.mode != B_PRED)
+#endif
xd->mode_info_context->mbmi.txfm_size = TX_8X8;
else
xd->mode_info_context->mbmi.txfm_size = TX_4X4;
} else {- if (pbi->common.txfm_mode == ONLY_4X4) {+#if CONFIG_TX16X16
+ if (xd->mode_info_context->mbmi.mode <= TM_PRED ||
+ xd->mode_info_context->mbmi.mode == NEWMV ||
+ xd->mode_info_context->mbmi.mode == ZEROMV ||
+ xd->mode_info_context->mbmi.mode == NEARMV ||
+ xd->mode_info_context->mbmi.mode == NEARESTMV) {+ xd->mode_info_context->mbmi.txfm_size = TX_16X16;
+ } else if (pbi->common.txfm_mode == ALLOW_8X8 &&
+ xd->mode_info_context->mbmi.mode != I8X8_PRED &&
+ xd->mode_info_context->mbmi.mode != B_PRED &&
+ xd->mode_info_context->mbmi.mode != SPLITMV) {+#else
+ if (pbi->common.txfm_mode == ALLOW_8X8 &&
+ xd->mode_info_context->mbmi.mode != I8X8_PRED &&
+ xd->mode_info_context->mbmi.mode != B_PRED &&
+ xd->mode_info_context->mbmi.mode != SPLITMV) {+#endif
+ xd->mode_info_context->mbmi.txfm_size = TX_8X8;
+ }
+ else {xd->mode_info_context->mbmi.txfm_size = TX_4X4;
- } else if (pbi->common.txfm_mode == ALLOW_8X8) {- if (xd->mode_info_context->mbmi.mode == B_PRED
- || xd->mode_info_context->mbmi.mode == I8X8_PRED
- || xd->mode_info_context->mbmi.mode == SPLITMV)
- xd->mode_info_context->mbmi.txfm_size = TX_4X4;
- else
- xd->mode_info_context->mbmi.txfm_size = TX_8X8;
}
}
@@ -251,6 +275,11 @@
xd->block[i].eob = 0;
xd->eobs[i] = 0;
}
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ eobtotal = vp8_decode_mb_tokens_16x16(pbi, xd);
+ else
+#endif
if (tx_type == TX_8X8)
eobtotal = vp8_decode_mb_tokens_8x8(pbi, xd);
else
@@ -462,6 +491,15 @@
xd->dst.y_stride, xd->eobs);
} else {BLOCKD *b = &xd->block[24];
+
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16) {+ vp8_dequant_idct_add_16x16_c(xd->qcoeff, xd->block[0].dequant,
+ xd->predictor, xd->dst.y_buffer,
+ 16, xd->dst.y_stride);
+ }
+ else
+#endif
if (tx_type == TX_8X8) {DEQUANT_INVOKE(&pbi->dequant, block_2x2)(b);
#ifdef DEC_DEBUG
@@ -511,7 +549,11 @@
}
}
- if (tx_type == TX_8X8)
+ if (tx_type == TX_8X8
+#if CONFIG_TX16X16
+ || tx_type == TX_16X16
+#endif
+ )
DEQUANT_INVOKE(&pbi->dequant, idct_add_uv_block_8x8) //
(xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->predictor + 16 * 16, xd->dst.u_buffer, xd->dst.v_buffer,
@@ -904,7 +946,7 @@
}
}
}
- }
+ }
}
if (pbi->common.txfm_mode == ALLOW_8X8 && vp8_read_bit(bc)) {@@ -925,6 +967,28 @@
}
}
}
+
+#if CONFIG_TX16X16
+ // 16x16
+ if (vp8_read_bit(bc)) {+ // read coef probability tree
+ for (i = 0; i < BLOCK_TYPES_16X16; ++i)
+ for (j = !i; j < COEF_BANDS; ++j)
+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {+ if (k >= 3 && ((i == 0 && j == 1) ||
+ (i > 0 && j == 0)))
+ continue;
+ for (l = 0; l < ENTROPY_NODES; ++l) {+
+ vp8_prob *const p = pc->fc.coef_probs_16x16[i][j][k] + l;
+
+ if (vp8_read(bc, COEF_UPDATE_PROB_16X16)) {+ *p = read_prob_diff_update(bc, *p);
+ }
+ }
+ }
+ }
+#endif
}
int vp8_decode_frame(VP8D_COMP *pbi) {@@ -1287,6 +1351,9 @@
vp8_copy(pbi->common.fc.pre_coef_probs, pbi->common.fc.coef_probs);
vp8_copy(pbi->common.fc.pre_coef_probs_8x8, pbi->common.fc.coef_probs_8x8);
+#if CONFIG_TX16X16
+ vp8_copy(pbi->common.fc.pre_coef_probs_16x16, pbi->common.fc.coef_probs_16x16);
+#endif
vp8_copy(pbi->common.fc.pre_ymode_prob, pbi->common.fc.ymode_prob);
vp8_copy(pbi->common.fc.pre_uv_mode_prob, pbi->common.fc.uv_mode_prob);
vp8_copy(pbi->common.fc.pre_bmode_prob, pbi->common.fc.bmode_prob);
@@ -1299,6 +1366,9 @@
#endif
vp8_zero(pbi->common.fc.coef_counts);
vp8_zero(pbi->common.fc.coef_counts_8x8);
+#if CONFIG_TX16X16
+ vp8_zero(pbi->common.fc.coef_counts_16x16);
+#endif
vp8_zero(pbi->common.fc.ymode_counts);
vp8_zero(pbi->common.fc.uv_mode_counts);
vp8_zero(pbi->common.fc.bmode_counts);
--- a/vp8/decoder/dequantize.c
+++ b/vp8/decoder/dequantize.c
@@ -422,3 +422,39 @@
#endif
}
+#if CONFIG_TX16X16
+void vp8_dequant_idct_add_16x16_c(short *input, short *dq, unsigned char *pred,
+ unsigned char *dest, int pitch, int stride) {+ short output[256];
+ short *diff_ptr = output;
+ int r, c, i;
+
+ input[0]= input[0] * dq[0];
+
+ // recover quantizer for 4 4x4 blocks
+ for (i = 1; i < 256; i++)
+ input[i] = input[i] * dq[1];
+
+ // the idct halves ( >> 1) the pitch
+ vp8_short_idct16x16_c(input, output, 32);
+
+ vpx_memset(input, 0, 512);
+
+ for (r = 0; r < 16; r++) {+ for (c = 0; c < 16; c++) {+ int a = diff_ptr[c] + pred[c];
+
+ if (a < 0)
+ a = 0;
+ else if (a > 255)
+ a = 255;
+
+ dest[c] = (unsigned char) a;
+ }
+
+ dest += stride;
+ diff_ptr += 16;
+ pred += pitch;
+ }
+}
+#endif
--- a/vp8/decoder/dequantize.h
+++ b/vp8/decoder/dequantize.h
@@ -145,6 +145,12 @@
#endif
extern prototype_dequant_idct_add_uv_block_8x8(vp8_dequant_idct_add_uv_block_8x8);
+#if CONFIG_TX16X16
+#ifndef vp8_dequant_idct_add_16x16
+#define vp8_dequant_idct_add_16x16 vp8_dequant_idct_add_16x16_c
+#endif
+extern prototype_dequant_idct_add(vp8_dequant_idct_add_16x16);
+#endif
typedef prototype_dequant_block((*vp8_dequant_block_fn_t));
@@ -178,6 +184,9 @@
vp8_dequant_dc_idct_add_y_block_fn_t_8x8 dc_idct_add_y_block_8x8;
vp8_dequant_idct_add_y_block_fn_t_8x8 idct_add_y_block_8x8;
vp8_dequant_idct_add_uv_block_fn_t_8x8 idct_add_uv_block_8x8;
+#if CONFIG_TX16X16
+ vp8_dequant_idct_add_fn_t idct_add_16x16;
+#endif
} vp8_dequant_rtcd_vtable_t;
#if CONFIG_RUNTIME_CPU_DETECT
--- a/vp8/decoder/detokenize.c
+++ b/vp8/decoder/detokenize.c
@@ -22,13 +22,13 @@
#define OCB_X PREV_COEF_CONTEXTS * ENTROPY_NODES
-DECLARE_ALIGNED(16, int, coef_bands_x[16]) = {+DECLARE_ALIGNED(16, const int, coef_bands_x[16]) = {0 * OCB_X, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X,
6 * OCB_X, 4 * OCB_X, 5 * OCB_X, 6 * OCB_X,
6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X,
6 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X
};
-DECLARE_ALIGNED(16, int, coef_bands_x_8x8[64]) = {+DECLARE_ALIGNED(16, const int, coef_bands_x_8x8[64]) = {0 * OCB_X, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X, 5 * OCB_X, 4 * OCB_X, 4 * OCB_X, 5 * OCB_X,
5 * OCB_X, 3 * OCB_X, 6 * OCB_X, 3 * OCB_X, 5 * OCB_X, 4 * OCB_X, 6 * OCB_X, 6 * OCB_X,
6 * OCB_X, 5 * OCB_X, 5 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X,
@@ -39,6 +39,27 @@
7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
};
+#if CONFIG_TX16X16
+DECLARE_ALIGNED(16, const int, coef_bands_x_16x16[256]) = {+ 0 * OCB_X, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X, 5 * OCB_X, 4 * OCB_X, 4 * OCB_X, 5 * OCB_X, 5 * OCB_X, 3 * OCB_X, 6 * OCB_X, 3 * OCB_X, 5 * OCB_X, 4 * OCB_X, 6 * OCB_X, 6 * OCB_X,
+ 6 * OCB_X, 5 * OCB_X, 5 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X,
+ 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X,
+ 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X
+};
+#endif
+
#define EOB_CONTEXT_NODE 0
#define ZERO_CONTEXT_NODE 1
#define ONE_CONTEXT_NODE 2
@@ -81,9 +102,13 @@
void vp8_reset_mb_tokens_context(MACROBLOCKD *x) {/* Clear entropy contexts for Y2 blocks */
- if (x->mode_info_context->mbmi.mode != B_PRED &&
+ if ((x->mode_info_context->mbmi.mode != B_PRED &&
x->mode_info_context->mbmi.mode != I8X8_PRED &&
- x->mode_info_context->mbmi.mode != SPLITMV) {+ x->mode_info_context->mbmi.mode != SPLITMV)
+#if CONFIG_TX16X16
+ || x->mode_info_context->mbmi.txfm_size == TX_16X16
+#endif
+ ) {vpx_memset(x->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
vpx_memset(x->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
} else {@@ -200,7 +225,28 @@
}
}
+#if CONFIG_TX16X16
+void static count_tokens_16x16(INT16 *qcoeff_ptr, int block, int type,
+ ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
+ int eob, int seg_eob, FRAME_CONTEXT *fc) {+ int c, pt, token;
+ VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l);
+ for (c = !type; c < eob; ++c) {+ int rc = vp8_default_zig_zag1d_16x16[c];
+ int v = qcoeff_ptr[rc];
+ int band = vp8_coef_bands_16x16[c];
+ token = get_token(v);
+ fc->coef_counts_16x16[type][band][pt][token]++;
+ pt = vp8_prev_token_class[token];
+ }
+ if (eob < seg_eob) {+ int band = vp8_coef_bands_16x16[c];
+ fc->coef_counts_16x16[type][band][pt][DCT_EOB_TOKEN]++;
+ }
+}
+#endif
+
static int vp8_get_signed(BOOL_DECODER *br, int value_to_sign) {const int split = (br->range + 1) >> 1;
const VP8_BD_VALUE bigsplit = (VP8_BD_VALUE)split << (VP8_BD_VALUE_SIZE - 8);
@@ -224,16 +270,16 @@
return v;
}
-#define WRITE_COEF_CONTINUE(val) \
- { \- Prob = coef_probs + (ENTROPY_NODES*PREV_CONTEXT_INC(val));\
+#define WRITE_COEF_CONTINUE(val) \
+ { \+ prob = coef_probs + (ENTROPY_NODES*PREV_CONTEXT_INC(val));\
qcoeff_ptr[scan[c]] = (INT16) vp8_get_signed(br, val); \
c++; \
continue; \
}
-#define ADJUST_COEF(prob, bits_count) \
- do { \+#define ADJUST_COEF(prob, bits_count) \
+ do { \if (vp8_read(br, prob)) \
val += (UINT16)(1 << bits_count);\
} while (0);
@@ -246,48 +292,59 @@
FRAME_CONTEXT *const fc = &dx->common.fc;
BOOL_DECODER *br = xd->current_bc;
int tmp, c = (type == 0);
- const vp8_prob *Prob;
- const vp8_prob *coef_probs =
- (block_type == TX_4X4) ? fc->coef_probs[type][0][0]
- : fc->coef_probs_8x8[type][0][0];
+ const vp8_prob *prob, *coef_probs;
+ switch (block_type) {+ case TX_4X4:
+ coef_probs = fc->coef_probs[type][0][0];
+ break;
+ case TX_8X8:
+ coef_probs = fc->coef_probs_8x8[type][0][0];
+ break;
+#if CONFIG_TX16X16
+ default:
+ coef_probs = fc->coef_probs_16x16[type][0][0];
+ break;
+#endif
+ }
+
VP8_COMBINEENTROPYCONTEXTS(tmp, *a, *l);
- Prob = coef_probs + tmp * ENTROPY_NODES;
+ prob = coef_probs + tmp * ENTROPY_NODES;
while (1) {int val;
const uint8_t *cat6 = cat6_prob;
if (c == seg_eob) break;
- Prob += coef_bands[c];
- if (!vp8_read(br, Prob[EOB_CONTEXT_NODE]))
+ prob += coef_bands[c];
+ if (!vp8_read(br, prob[EOB_CONTEXT_NODE]))
break;
- SKIP_START:
+SKIP_START:
if (c == seg_eob) break;
- if (!vp8_read(br, Prob[ZERO_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[ZERO_CONTEXT_NODE])) {++c;
- Prob = coef_probs + coef_bands[c];
+ prob = coef_probs + coef_bands[c];
goto SKIP_START;
}
// ONE_CONTEXT_NODE_0_
- if (!vp8_read(br, Prob[ONE_CONTEXT_NODE])) {- Prob = coef_probs + ENTROPY_NODES;
+ if (!vp8_read(br, prob[ONE_CONTEXT_NODE])) {+ prob = coef_probs + ENTROPY_NODES;
qcoeff_ptr[scan[c]] = (INT16) vp8_get_signed(br, 1);
++c;
continue;
}
// LOW_VAL_CONTEXT_NODE_0_
- if (!vp8_read(br, Prob[LOW_VAL_CONTEXT_NODE])) {- if (!vp8_read(br, Prob[TWO_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[LOW_VAL_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[TWO_CONTEXT_NODE])) {WRITE_COEF_CONTINUE(2);
}
- if (!vp8_read(br, Prob[THREE_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[THREE_CONTEXT_NODE])) {WRITE_COEF_CONTINUE(3);
}
WRITE_COEF_CONTINUE(4);
}
// HIGH_LOW_CONTEXT_NODE_0_
- if (!vp8_read(br, Prob[HIGH_LOW_CONTEXT_NODE])) {- if (!vp8_read(br, Prob[CAT_ONE_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[HIGH_LOW_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[CAT_ONE_CONTEXT_NODE])) {val = CAT1_MIN_VAL;
ADJUST_COEF(CAT1_PROB0, 0);
WRITE_COEF_CONTINUE(val);
@@ -298,8 +355,8 @@
WRITE_COEF_CONTINUE(val);
}
// CAT_THREEFOUR_CONTEXT_NODE_0_
- if (!vp8_read(br, Prob[CAT_THREEFOUR_CONTEXT_NODE])) {- if (!vp8_read(br, Prob[CAT_THREE_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[CAT_THREEFOUR_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[CAT_THREE_CONTEXT_NODE])) {val = CAT3_MIN_VAL;
ADJUST_COEF(CAT3_PROB2, 2);
ADJUST_COEF(CAT3_PROB1, 1);
@@ -314,7 +371,7 @@
WRITE_COEF_CONTINUE(val);
}
// CAT_FIVE_CONTEXT_NODE_0_:
- if (!vp8_read(br, Prob[CAT_FIVE_CONTEXT_NODE])) {+ if (!vp8_read(br, prob[CAT_FIVE_CONTEXT_NODE])) {val = CAT5_MIN_VAL;
ADJUST_COEF(CAT5_PROB4, 4);
ADJUST_COEF(CAT5_PROB3, 3);
@@ -331,18 +388,81 @@
WRITE_COEF_CONTINUE(val);
}
- if (block_type == TX_4X4)
+ if (block_type == TX_4X4) {#if CONFIG_HYBRIDTRANSFORM
count_tokens_adaptive_scan(xd, qcoeff_ptr, i, type, a, l, c, seg_eob, fc);
#else
count_tokens(qcoeff_ptr, i, type, a, l, c, seg_eob, fc);
#endif
-
- else
+ }
+ else if (block_type == TX_8X8)
count_tokens_8x8(qcoeff_ptr, i, type, a, l, c, seg_eob, fc);
+#if CONFIG_TX16X16
+ else
+ count_tokens_16x16(qcoeff_ptr, i, type, a, l, c, seg_eob, fc);
+#endif
return c;
}
+#if CONFIG_TX16X16
+int vp8_decode_mb_tokens_16x16(VP8D_COMP *pbi, MACROBLOCKD *xd) {+ ENTROPY_CONTEXT* const A = (ENTROPY_CONTEXT *)xd->above_context;
+ ENTROPY_CONTEXT* const L = (ENTROPY_CONTEXT *)xd->left_context;
+
+ char* const eobs = xd->eobs;
+ int c, i, type, eobtotal = 0, seg_eob;
+ const int segment_id = xd->mode_info_context->mbmi.segment_id;
+ const int seg_active = segfeature_active(xd, segment_id, SEG_LVL_EOB);
+ INT16 *qcoeff_ptr = &xd->qcoeff[0];
+
+ type = PLANE_TYPE_Y_WITH_DC;
+ if (seg_active)
+ seg_eob = get_segdata(xd, segment_id, SEG_LVL_EOB);
+ else
+ seg_eob = 256;
+
+ // Luma block
+ {+ const int* const scan = vp8_default_zig_zag1d_16x16;
+ c = vp8_decode_coefs(pbi, xd, A, L, type, seg_eob, qcoeff_ptr,
+ 0, scan, TX_16X16, coef_bands_x_16x16);
+ eobs[0] = c;
+ *A = *L = (c != !type);
+ for (i = 1; i < 16; i++) {+ *(A + vp8_block2above[i]) = *(A);
+ *(L + vp8_block2left[i]) = *(L);
+ }
+ eobtotal += c;
+ }
+
+ // 8x8 chroma blocks
+ qcoeff_ptr += 256;
+ type = PLANE_TYPE_UV;
+ if (seg_active)
+ seg_eob = get_segdata(xd, segment_id, SEG_LVL_EOB);
+ else
+ seg_eob = 64;
+ for (i = 16; i < 24; i += 4) {+ ENTROPY_CONTEXT* const a = A + vp8_block2above_8x8[i];
+ ENTROPY_CONTEXT* const l = L + vp8_block2left_8x8[i];
+ const int* const scan = vp8_default_zig_zag1d_8x8;
+
+ c = vp8_decode_coefs(pbi, xd, a, l, type, seg_eob, qcoeff_ptr,
+ i, scan, TX_8X8, coef_bands_x_8x8);
+ a[0] = l[0] = ((eobs[i] = c) != !type);
+ a[1] = a[0];
+ l[1] = l[0];
+
+ eobtotal += c;
+ qcoeff_ptr += 64;
+ }
+ vpx_memset(&A[8], 0, sizeof(A[8]));
+ vpx_memset(&L[8], 0, sizeof(L[8]));
+ return eobtotal;
+}
+#endif
+
+
int vp8_decode_mb_tokens_8x8(VP8D_COMP *pbi, MACROBLOCKD *xd) {ENTROPY_CONTEXT *const A = (ENTROPY_CONTEXT *)xd->above_context;
ENTROPY_CONTEXT *const L = (ENTROPY_CONTEXT *)xd->left_context;
@@ -464,7 +584,6 @@
c = vp8_decode_coefs(dx, xd, a, l, type, seg_eob, qcoeff_ptr + 24 * 16, 24,
scan, TX_4X4, coef_bands_x);
a[0] = l[0] = ((eobs[24] = c) != !type);
-
eobtotal += c - 16;
type = PLANE_TYPE_Y_NO_DC;
--- a/vp8/decoder/detokenize.h
+++ b/vp8/decoder/detokenize.h
@@ -17,5 +17,8 @@
void vp8_reset_mb_tokens_context(MACROBLOCKD *x);
int vp8_decode_mb_tokens(VP8D_COMP *, MACROBLOCKD *);
int vp8_decode_mb_tokens_8x8(VP8D_COMP *, MACROBLOCKD *);
+#if CONFIG_TX16X16
+int vp8_decode_mb_tokens_16x16(VP8D_COMP *, MACROBLOCKD *);
+#endif
#endif /* DETOKENIZE_H */
--- a/vp8/decoder/generic/dsystemdependent.c
+++ b/vp8/decoder/generic/dsystemdependent.c
@@ -22,6 +22,9 @@
pbi->mb.rtcd = &pbi->common.rtcd;
pbi->dequant.block_2x2 = vp8_dequantize_b_2x2_c;
pbi->dequant.idct_add_8x8 = vp8_dequant_idct_add_8x8_c;
+#if CONFIG_TX16X16
+ pbi->dequant.idct_add_16x16 = vp8_dequant_idct_add_16x16_c;
+#endif
pbi->dequant.dc_idct_add_8x8 = vp8_dequant_dc_idct_add_8x8_c;
pbi->dequant.dc_idct_add_y_block_8x8 = vp8_dequant_dc_idct_add_y_block_8x8_c;
pbi->dequant.idct_add_y_block_8x8 = vp8_dequant_idct_add_y_block_8x8_c;
--- a/vp8/decoder/onyxd_int.h
+++ b/vp8/decoder/onyxd_int.h
@@ -54,6 +54,9 @@
vp8_prob const *coef_probs[BLOCK_TYPES];
vp8_prob const *coef_probs_8x8[BLOCK_TYPES_8X8];
+#if CONFIG_TX16X16
+ vp8_prob const *coef_probs_16X16[BLOCK_TYPES_16X16];
+#endif
UINT8 eob[25];
--- a/vp8/encoder/bitstream.c
+++ b/vp8/encoder/bitstream.c
@@ -42,6 +42,12 @@
[COEF_BANDS]
[PREV_COEF_CONTEXTS]
[ENTROPY_NODES] [2];
+#if CONFIG_TX16X16
+unsigned int tree_update_hist_16x16 [BLOCK_TYPES_16X16]
+ [COEF_BANDS]
+ [PREV_COEF_CONTEXTS]
+ [ENTROPY_NODES] [2];
+#endif
extern unsigned int active_section;
#endif
@@ -1283,15 +1289,13 @@
void build_coeff_contexts(VP8_COMP *cpi) {- int i = 0;
- do {- int j = 0;
- do {- int k = 0;
- do {+ int i = 0, j, k;
#ifdef ENTROPY_STATS
- int t;
+ int t = 0;
#endif
+ for (i = 0; i < BLOCK_TYPES; ++i) {+ for (j = 0; j < COEF_BANDS; ++j) {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {if (k >= 3 && ((i == 0 && j == 1) || (i > 0 && j == 0)))
continue;
vp8_tree_probs_from_distribution(
@@ -1302,33 +1306,23 @@
256, 1
);
#ifdef ENTROPY_STATS
- if (!cpi->dummy_packing) {- t = 0;
- do {- context_counters [i][j][k][t] +=
- cpi->coef_counts [i][j][k][t];
- } while (++t < MAX_ENTROPY_TOKENS);
- }
+ if (!cpi->dummy_packing)
+ for (t = 0; t < MAX_ENTROPY_TOKENS; ++t)
+ context_counters[i][j][k][t] += cpi->coef_counts[i][j][k][t];
#endif
- } while (++k < PREV_COEF_CONTEXTS);
- } while (++j < COEF_BANDS);
- } while (++i < BLOCK_TYPES);
+ }
+ }
+ }
- i = 0;
if (cpi->common.txfm_mode == ALLOW_8X8) {- do {- int j = 0; /* token/prob index */
- do {- int k = 0;
- do {+ for (i = 0; i < BLOCK_TYPES_8X8; ++i) {+ for (j = 0; j < COEF_BANDS; ++j) {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {/* at every context */
/* calc probs and branch cts for this frame only */
// vp8_prob new_p [ENTROPY_NODES];
// unsigned int branch_ct [ENTROPY_NODES] [2];
-#ifdef ENTROPY_STATS
- int t = 0; /* token/prob index */
-#endif
if (k >= 3 && ((i == 0 && j == 1) || (i > 0 && j == 0)))
continue;
vp8_tree_probs_from_distribution(
@@ -1339,20 +1333,36 @@
256, 1
);
#ifdef ENTROPY_STATS
- if (!cpi->dummy_packing) {- t = 0;
- do {- context_counters_8x8 [i][j][k][t] +=
- cpi->coef_counts_8x8 [i][j][k][t];
- } while (++t < MAX_ENTROPY_TOKENS);
- }
+ if (!cpi->dummy_packing)
+ for (t = 0; t < MAX_ENTROPY_TOKENS; ++t)
+ context_counters_8x8[i][j][k][t] += cpi->coef_counts_8x8[i][j][k][t];
#endif
-
- } while (++k < PREV_COEF_CONTEXTS);
- } while (++j < COEF_BANDS);
- } while (++i < BLOCK_TYPES_8X8);
+ }
+ }
+ }
}
+#if CONFIG_TX16X16
+ //16x16
+ for (i = 0; i < BLOCK_TYPES_16X16; ++i) {+ for (j = 0; j < COEF_BANDS; ++j) {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {+ if (k >= 3 && ((i == 0 && j == 1) || (i > 0 && j == 0)))
+ continue;
+ vp8_tree_probs_from_distribution(
+ MAX_ENTROPY_TOKENS, vp8_coef_encodings, vp8_coef_tree,
+ cpi->frame_coef_probs_16x16[i][j][k],
+ cpi->frame_branch_ct_16x16[i][j][k],
+ cpi->coef_counts_16x16[i][j][k], 256, 1);
+#ifdef ENTROPY_STATS
+ if (!cpi->dummy_packing)
+ for (t = 0; t < MAX_ENTROPY_TOKENS; ++t)
+ context_counters_16x16[i][j][k][t] += cpi->coef_counts_16x16[i][j][k][t];
+#endif
+ }
+ }
+ }
+#endif
}
static void update_coef_probs3(VP8_COMP *cpi) {@@ -1696,7 +1706,7 @@
}
static void update_coef_probs(VP8_COMP *cpi) {- int i = 0;
+ int i, j, k, t;
vp8_writer *const w = & cpi->bc;
int update[2] = {0, 0};int savings;
@@ -1704,7 +1714,6 @@
vp8_clear_system_state(); // __asm emms;
// Build the cofficient contexts based on counts collected in encode loop
-
build_coeff_contexts(cpi);
// vp8_prob bestupd = find_coef_update_prob(cpi);
@@ -1711,14 +1720,11 @@
/* dry run to see if there is any udpate at all needed */
savings = 0;
- do {- int j = !i;
- do {- int k = 0;
+ for (i = 0; i < BLOCK_TYPES; ++i) {+ for (j = !i; j < COEF_BANDS; ++j) { int prev_coef_savings[ENTROPY_NODES] = {0};- do {- int t = 0; /* token/prob index */
- do {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {+ for (t = 0; t < ENTROPY_NODES; ++t) {vp8_prob newp = cpi->frame_coef_probs [i][j][k][t];
vp8_prob *Pold = cpi->common.fc.coef_probs [i][j][k] + t;
const vp8_prob upd = COEF_UPDATE_PROB;
@@ -1747,29 +1753,23 @@
#endif
update[u]++;
- } while (++t < ENTROPY_NODES);
- } while (++k < PREV_COEF_CONTEXTS);
- } while (++j < COEF_BANDS);
- } while (++i < BLOCK_TYPES);
+ }
+ }
+ }
+ }
// printf("Update %d %d, savings %d\n", update[0], update[1], savings);/* Is coef updated at all */
if (update[1] == 0 || savings < 0)
- {vp8_write_bit(w, 0);
- } else {+ else {vp8_write_bit(w, 1);
- i = 0;
- do {- int j = !i;
- do {- int k = 0;
+ for (i = 0; i < BLOCK_TYPES; ++i) {+ for (j = !i; j < COEF_BANDS; ++j) { int prev_coef_savings[ENTROPY_NODES] = {0};-
- do {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {// calc probs and branch cts for this frame only
- int t = 0; /* token/prob index */
- do {+ for (t = 0; t < ENTROPY_NODES; ++t) {vp8_prob newp = cpi->frame_coef_probs [i][j][k][t];
vp8_prob *Pold = cpi->common.fc.coef_probs [i][j][k] + t;
const vp8_prob upd = COEF_UPDATE_PROB;
@@ -1791,8 +1791,6 @@
if (s > 0)
u = 1;
#endif
-
-
vp8_write(w, u, upd);
#ifdef ENTROPY_STATS
if (!cpi->dummy_packing)
@@ -1803,28 +1801,23 @@
write_prob_diff_update(w, newp, *Pold);
*Pold = newp;
}
- } while (++t < ENTROPY_NODES);
-
- } while (++k < PREV_COEF_CONTEXTS);
- } while (++j < COEF_BANDS);
- } while (++i < BLOCK_TYPES);
+ }
+ }
+ }
+ }
}
- /* do not do this if not evena allowed */
+ /* do not do this if not even allowed */
if (cpi->common.txfm_mode == ALLOW_8X8) {/* dry run to see if update is necessary */
update[0] = update[1] = 0;
savings = 0;
- i = 0;
- do {- int j = !i;
- do {- int k = 0;
- do {+ for (i = 0; i < BLOCK_TYPES_8X8; ++i) {+ for (j = !i; j < COEF_BANDS; ++j) {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {// calc probs and branch cts for this frame only
- int t = 0; /* token/prob index */
- do {+ for (t = 0; t < ENTROPY_NODES; ++t) {const unsigned int *ct = cpi->frame_branch_ct_8x8 [i][j][k][t];
vp8_prob newp = cpi->frame_coef_probs_8x8 [i][j][k][t];
vp8_prob *Pold = cpi->common.fc.coef_probs_8x8 [i][j][k] + t;
@@ -1846,26 +1839,20 @@
if (u)
savings += s;
#endif
-
update[u]++;
- } while (++t < MAX_ENTROPY_TOKENS - 1);
- } while (++k < PREV_COEF_CONTEXTS);
- } while (++j < COEF_BANDS);
- } while (++i < BLOCK_TYPES_8X8);
+ }
+ }
+ }
+ }
if (update[1] == 0 || savings < 0)
- {vp8_write_bit(w, 0);
- } else {+ else {vp8_write_bit(w, 1);
- i = 0;
- do {- int j = !i;
- do {- int k = 0;
- do {- int t = 0; /* token/prob index */
- do {+ for (i = 0; i < BLOCK_TYPES_8X8; ++i) {+ for (j = !i; j < COEF_BANDS; ++j) {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {+ for (t = 0; t < ENTROPY_NODES; ++t) {const unsigned int *ct = cpi->frame_branch_ct_8x8 [i][j][k][t];
vp8_prob newp = cpi->frame_coef_probs_8x8 [i][j][k][t];
vp8_prob *Pold = cpi->common.fc.coef_probs_8x8 [i][j][k] + t;
@@ -1892,12 +1879,90 @@
write_prob_diff_update(w, newp, oldp);
*Pold = newp;
}
- } while (++t < MAX_ENTROPY_TOKENS - 1);
- } while (++k < PREV_COEF_CONTEXTS);
- } while (++j < COEF_BANDS);
- } while (++i < BLOCK_TYPES_8X8);
+ }
+ }
+ }
+ }
}
}
+
+#if CONFIG_TX16X16
+ // 16x16
+ /* dry run to see if update is necessary */
+ update[0] = update[1] = 0;
+ savings = 0;
+ for (i = 0; i < BLOCK_TYPES_16X16; ++i) {+ for (j = !i; j < COEF_BANDS; ++j) {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {+ // calc probs and branch cts for this frame only
+ for (t = 0; t < ENTROPY_NODES; ++t) {+ const unsigned int *ct = cpi->frame_branch_ct_16x16[i][j][k][t];
+ vp8_prob newp = cpi->frame_coef_probs_16x16[i][j][k][t];
+ vp8_prob *Pold = cpi->common.fc.coef_probs_16x16[i][j][k] + t;
+ const vp8_prob oldp = *Pold;
+ int s, u;
+ const vp8_prob upd = COEF_UPDATE_PROB_16X16;
+ if (k >= 3 && ((i == 0 && j == 1) || (i > 0 && j == 0)))
+ continue;
+#if defined(SEARCH_NEWP)
+ s = prob_diff_update_savings_search(ct, oldp, &newp, upd);
+ u = s > 0 && newp != oldp ? 1 : 0;
+ if (u)
+ savings += s - (int)(vp8_cost_zero(upd));
+ else
+ savings -= (int)(vp8_cost_zero(upd));
+#else
+ s = prob_update_savings(ct, oldp, newp, upd);
+ u = s > 0 ? 1 : 0;
+ if (u)
+ savings += s;
+#endif
+ update[u]++;
+ }
+ }
+ }
+ }
+
+ if (update[1] == 0 || savings < 0)
+ vp8_write_bit(w, 0);
+ else {+ vp8_write_bit(w, 1);
+ for (i = 0; i < BLOCK_TYPES_16X16; ++i) {+ for (j = !i; j < COEF_BANDS; ++j) {+ for (k = 0; k < PREV_COEF_CONTEXTS; ++k) {+ for (t = 0; t < ENTROPY_NODES; ++t) {+ const unsigned int *ct = cpi->frame_branch_ct_16x16[i][j][k][t];
+ vp8_prob newp = cpi->frame_coef_probs_16x16[i][j][k][t];
+ vp8_prob *Pold = cpi->common.fc.coef_probs_16x16[i][j][k] + t;
+ const vp8_prob oldp = *Pold;
+ const vp8_prob upd = COEF_UPDATE_PROB_16X16;
+ int s, u;
+ if (k >= 3 && ((i == 0 && j == 1) ||
+ (i > 0 && j == 0)))
+ continue;
+#if defined(SEARCH_NEWP)
+ s = prob_diff_update_savings_search(ct, oldp, &newp, upd);
+ u = s > 0 && newp != oldp ? 1 : 0;
+#else
+ s = prob_update_savings(ct, oldp, newp, upd);
+ u = s > 0 ? 1 : 0;
+#endif
+ vp8_write(w, u, upd);
+#ifdef ENTROPY_STATS
+ if (!cpi->dummy_packing)
+ ++tree_update_hist_16x16[i][j][k][t][u];
+#endif
+ if (u) {+ /* send/use new probability */
+ write_prob_diff_update(w, newp, oldp);
+ *Pold = newp;
+ }
+ }
+ }
+ }
+ }
+ }
+#endif
}
#ifdef PACKET_TESTING
@@ -2310,12 +2375,10 @@
vp8_write_bit(bc, pc->refresh_last_frame);
#ifdef ENTROPY_STATS
-
if (pc->frame_type == INTER_FRAME)
active_section = 0;
else
active_section = 7;
-
#endif
vp8_clear_system_state(); // __asm emms;
@@ -2322,6 +2385,9 @@
vp8_copy(cpi->common.fc.pre_coef_probs, cpi->common.fc.coef_probs);
vp8_copy(cpi->common.fc.pre_coef_probs_8x8, cpi->common.fc.coef_probs_8x8);
+#if CONFIG_TX16X16
+ vp8_copy(cpi->common.fc.pre_coef_probs_16x16, cpi->common.fc.coef_probs_16x16);
+#endif
vp8_copy(cpi->common.fc.pre_ymode_prob, cpi->common.fc.ymode_prob);
vp8_copy(cpi->common.fc.pre_uv_mode_prob, cpi->common.fc.uv_mode_prob);
vp8_copy(cpi->common.fc.pre_bmode_prob, cpi->common.fc.bmode_prob);
@@ -2401,24 +2467,20 @@
FILE *f = fopen("coefupdprob.h", "w");int Sum;
fprintf(f, "\n/* Update probabilities for token entropy tree. */\n\n");
+
fprintf(f, "const vp8_prob\n"
"vp8_coef_update_probs[BLOCK_TYPES]\n"
" [COEF_BANDS]\n"
" [PREV_COEF_CONTEXTS]\n"
" [ENTROPY_NODES] = {\n");-
for (i = 0; i < BLOCK_TYPES; i++) { fprintf(f, " { \n");-
for (j = 0; j < COEF_BANDS; j++) { fprintf(f, " {\n");-
for (k = 0; k < PREV_COEF_CONTEXTS; k++) { fprintf(f, " {");-
for (l = 0; l < ENTROPY_NODES; l++) {Sum = tree_update_hist[i][j][k][l][0] + tree_update_hist[i][j][k][l][1];
-
if (Sum > 0) {if (((tree_update_hist[i][j][k][l][0] * 255) / Sum) > 0)
fprintf(f, "%3ld, ", (tree_update_hist[i][j][k][l][0] * 255) / Sum);
@@ -2427,16 +2489,12 @@
} else
fprintf(f, "%3ld, ", 128);
}
-
fprintf(f, "},\n");
}
-
fprintf(f, " },\n");
}
-
fprintf(f, " },\n");
}
-
fprintf(f, "};\n");
fprintf(f, "const vp8_prob\n"
@@ -2444,20 +2502,14 @@
" [COEF_BANDS]\n"
" [PREV_COEF_CONTEXTS]\n"
" [ENTROPY_NODES] = {\n");-
-
for (i = 0; i < BLOCK_TYPES_8X8; i++) { fprintf(f, " { \n");-
for (j = 0; j < COEF_BANDS; j++) { fprintf(f, " {\n");-
for (k = 0; k < PREV_COEF_CONTEXTS; k++) { fprintf(f, " {");-
for (l = 0; l < MAX_ENTROPY_TOKENS - 1; l++) {Sum = tree_update_hist_8x8[i][j][k][l][0] + tree_update_hist_8x8[i][j][k][l][1];
-
if (Sum > 0) {if (((tree_update_hist_8x8[i][j][k][l][0] * 255) / Sum) > 0)
fprintf(f, "%3ld, ", (tree_update_hist_8x8[i][j][k][l][0] * 255) / Sum);
@@ -2466,20 +2518,50 @@
} else
fprintf(f, "%3ld, ", 128);
}
-
fprintf(f, "},\n");
}
-
fprintf(f, " },\n");
}
+ fprintf(f, " },\n");
+ }
+#if CONFIG_TX16X16
+ fprintf(f, "const vp8_prob\n"
+ "vp8_coef_update_probs_16x16[BLOCK_TYPES_16X16]\n"
+ " [COEF_BANDS]\n"
+ " [PREV_COEF_CONTEXTS]\n"
+ " [ENTROPY_NODES] = {\n");+ for (i = 0; i < BLOCK_TYPES_16X16; i++) {+ fprintf(f, " { \n");+ for (j = 0; j < COEF_BANDS; j++) {+ fprintf(f, " {\n");+ for (k = 0; k < PREV_COEF_CONTEXTS; k++) {+ fprintf(f, " {");+ for (l = 0; l < MAX_ENTROPY_TOKENS - 1; l++) {+ Sum = tree_update_hist_16x16[i][j][k][l][0] + tree_update_hist_16x16[i][j][k][l][1];
+ if (Sum > 0) {+ if (((tree_update_hist_16x16[i][j][k][l][0] * 255) / Sum) > 0)
+ fprintf(f, "%3ld, ", (tree_update_hist_16x16[i][j][k][l][0] * 255) / Sum);
+ else
+ fprintf(f, "%3ld, ", 1);
+ } else
+ fprintf(f, "%3ld, ", 128);
+ }
+ fprintf(f, "},\n");
+ }
+ fprintf(f, " },\n");
+ }
fprintf(f, " },\n");
}
+#endif
+
fclose(f);
f = fopen("treeupdate.bin", "wb");fwrite(tree_update_hist, sizeof(tree_update_hist), 1, f);
fwrite(tree_update_hist_8x8, sizeof(tree_update_hist_8x8), 1, f);
+#if CONFIG_TX16X16
+ fwrite(tree_update_hist_16x16, sizeof(tree_update_hist_16x16), 1, f);
+#endif
fclose(f);
-
}
#endif
--- a/vp8/encoder/block.h
+++ b/vp8/encoder/block.h
@@ -35,8 +35,14 @@
unsigned char *quant_shift;
short *zbin;
short *zbin_8x8;
+#if CONFIG_TX16X16
+ short *zbin_16x16;
+#endif
short *zrun_zbin_boost;
short *zrun_zbin_boost_8x8;
+#if CONFIG_TX16X16
+ short *zrun_zbin_boost_16x16;
+#endif
short *round;
// Zbin Over Quant value
@@ -49,7 +55,9 @@
int eob_max_offset;
int eob_max_offset_8x8;
-
+#if CONFIG_TX16X16
+ int eob_max_offset_16x16;
+#endif
} BLOCK;
typedef struct {@@ -153,9 +161,13 @@
#endif
unsigned int token_costs[BLOCK_TYPES] [COEF_BANDS]
- [PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS];
+ [PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS];
unsigned int token_costs_8x8[BLOCK_TYPES_8X8] [COEF_BANDS]
- [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
+ [PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS];
+#if CONFIG_TX16X16
+ unsigned int token_costs_16x16[BLOCK_TYPES_16X16] [COEF_BANDS]
+ [PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS];
+#endif
int optimize;
int q_index;
@@ -176,7 +188,13 @@
void (*quantize_b)(BLOCK *b, BLOCKD *d);
void (*quantize_b_pair)(BLOCK *b1, BLOCK *b2, BLOCKD *d0, BLOCKD *d1);
void (*vp8_short_fdct8x8)(short *input, short *output, int pitch);
+#if CONFIG_TX16X16
+ void (*vp8_short_fdct16x16)(short *input, short *output, int pitch);
+#endif
void (*short_fhaar2x2)(short *input, short *output, int pitch);
+#if CONFIG_TX16X16
+ void (*quantize_b_16x16)(BLOCK *b, BLOCKD *d);
+#endif
void (*quantize_b_8x8)(BLOCK *b, BLOCKD *d);
void (*quantize_b_2x2)(BLOCK *b, BLOCKD *d);
--- a/vp8/encoder/dct.c
+++ b/vp8/encoder/dct.c
@@ -575,3 +575,205 @@
vp8_short_walsh4x4_x8_c(input + 4, output + 16, pitch);
}
#endif
+
+#if CONFIG_TX16X16
+static void dct16x16_1d(double input[16], double output[16]) {+ double step[16];
+ double intermediate[16];
+ double temp1, temp2;
+
+ const double PI = 3.1415926535898;
+ const double C1 = cos(1*PI/(double)32);
+ const double C2 = cos(2*PI/(double)32);
+ const double C3 = cos(3*PI/(double)32);
+ const double C4 = cos(4*PI/(double)32);
+ const double C5 = cos(5*PI/(double)32);
+ const double C6 = cos(6*PI/(double)32);
+ const double C7 = cos(7*PI/(double)32);
+ const double C8 = cos(8*PI/(double)32);
+ const double C9 = cos(9*PI/(double)32);
+ const double C10 = cos(10*PI/(double)32);
+ const double C11 = cos(11*PI/(double)32);
+ const double C12 = cos(12*PI/(double)32);
+ const double C13 = cos(13*PI/(double)32);
+ const double C14 = cos(14*PI/(double)32);
+ const double C15 = cos(15*PI/(double)32);
+
+ // step 1
+ step[ 0] = input[0] + input[15];
+ step[ 1] = input[1] + input[14];
+ step[ 2] = input[2] + input[13];
+ step[ 3] = input[3] + input[12];
+ step[ 4] = input[4] + input[11];
+ step[ 5] = input[5] + input[10];
+ step[ 6] = input[6] + input[ 9];
+ step[ 7] = input[7] + input[ 8];
+ step[ 8] = input[7] - input[ 8];
+ step[ 9] = input[6] - input[ 9];
+ step[10] = input[5] - input[10];
+ step[11] = input[4] - input[11];
+ step[12] = input[3] - input[12];
+ step[13] = input[2] - input[13];
+ step[14] = input[1] - input[14];
+ step[15] = input[0] - input[15];
+
+ // step 2
+ output[0] = step[0] + step[7];
+ output[1] = step[1] + step[6];
+ output[2] = step[2] + step[5];
+ output[3] = step[3] + step[4];
+ output[4] = step[3] - step[4];
+ output[5] = step[2] - step[5];
+ output[6] = step[1] - step[6];
+ output[7] = step[0] - step[7];
+
+ temp1 = step[ 8]*C7;
+ temp2 = step[15]*C9;
+ output[ 8] = temp1 + temp2;
+
+ temp1 = step[ 9]*C11;
+ temp2 = step[14]*C5;
+ output[ 9] = temp1 - temp2;
+
+ temp1 = step[10]*C3;
+ temp2 = step[13]*C13;
+ output[10] = temp1 + temp2;
+
+ temp1 = step[11]*C15;
+ temp2 = step[12]*C1;
+ output[11] = temp1 - temp2;
+
+ temp1 = step[11]*C1;
+ temp2 = step[12]*C15;
+ output[12] = temp2 + temp1;
+
+ temp1 = step[10]*C13;
+ temp2 = step[13]*C3;
+ output[13] = temp2 - temp1;
+
+ temp1 = step[ 9]*C5;
+ temp2 = step[14]*C11;
+ output[14] = temp2 + temp1;
+
+ temp1 = step[ 8]*C9;
+ temp2 = step[15]*C7;
+ output[15] = temp2 - temp1;
+
+ // step 3
+ step[ 0] = output[0] + output[3];
+ step[ 1] = output[1] + output[2];
+ step[ 2] = output[1] - output[2];
+ step[ 3] = output[0] - output[3];
+
+ temp1 = output[4]*C14;
+ temp2 = output[7]*C2;
+ step[ 4] = temp1 + temp2;
+
+ temp1 = output[5]*C10;
+ temp2 = output[6]*C6;
+ step[ 5] = temp1 + temp2;
+
+ temp1 = output[5]*C6;
+ temp2 = output[6]*C10;
+ step[ 6] = temp2 - temp1;
+
+ temp1 = output[4]*C2;
+ temp2 = output[7]*C14;
+ step[ 7] = temp2 - temp1;
+
+ step[ 8] = output[ 8] + output[11];
+ step[ 9] = output[ 9] + output[10];
+ step[10] = output[ 9] - output[10];
+ step[11] = output[ 8] - output[11];
+
+ step[12] = output[12] + output[15];
+ step[13] = output[13] + output[14];
+ step[14] = output[13] - output[14];
+ step[15] = output[12] - output[15];
+
+ // step 4
+ output[ 0] = (step[ 0] + step[ 1]);
+ output[ 8] = (step[ 0] - step[ 1]);
+
+ temp1 = step[2]*C12;
+ temp2 = step[3]*C4;
+ temp1 = temp1 + temp2;
+ output[ 4] = 2*(temp1*C8);
+
+ temp1 = step[2]*C4;
+ temp2 = step[3]*C12;
+ temp1 = temp2 - temp1;
+ output[12] = 2*(temp1*C8);
+
+ output[ 2] = 2*((step[4] + step[ 5])*C8);
+ output[14] = 2*((step[7] - step[ 6])*C8);
+
+ temp1 = step[4] - step[5];
+ temp2 = step[6] + step[7];
+ output[ 6] = (temp1 + temp2);
+ output[10] = (temp1 - temp2);
+
+ intermediate[8] = step[8] + step[14];
+ intermediate[9] = step[9] + step[15];
+
+ temp1 = intermediate[8]*C12;
+ temp2 = intermediate[9]*C4;
+ temp1 = temp1 - temp2;
+ output[3] = 2*(temp1*C8);
+
+ temp1 = intermediate[8]*C4;
+ temp2 = intermediate[9]*C12;
+ temp1 = temp2 + temp1;
+ output[13] = 2*(temp1*C8);
+
+ output[ 9] = 2*((step[10] + step[11])*C8);
+
+ intermediate[11] = step[10] - step[11];
+ intermediate[12] = step[12] + step[13];
+ intermediate[13] = step[12] - step[13];
+ intermediate[14] = step[ 8] - step[14];
+ intermediate[15] = step[ 9] - step[15];
+
+ output[15] = (intermediate[11] + intermediate[12]);
+ output[ 1] = -(intermediate[11] - intermediate[12]);
+
+ output[ 7] = 2*(intermediate[13]*C8);
+
+ temp1 = intermediate[14]*C12;
+ temp2 = intermediate[15]*C4;
+ temp1 = temp1 - temp2;
+ output[11] = -2*(temp1*C8);
+
+ temp1 = intermediate[14]*C4;
+ temp2 = intermediate[15]*C12;
+ temp1 = temp2 + temp1;
+ output[ 5] = 2*(temp1*C8);
+}
+
+void vp8_short_fdct16x16_c(short *input, short *out, int pitch) {+ int shortpitch = pitch >> 1;
+ int i, j;
+ double output[256];
+ // First transform columns
+ for (i = 0; i < 16; i++) {+ double temp_in[16], temp_out[16];
+ for (j = 0; j < 16; j++)
+ temp_in[j] = input[j*shortpitch + i];
+ dct16x16_1d(temp_in, temp_out);
+ for (j = 0; j < 16; j++)
+ output[j*16 + i] = temp_out[j];
+ }
+ // Then transform rows
+ for (i = 0; i < 16; ++i) {+ double temp_in[16], temp_out[16];
+ for (j = 0; j < 16; ++j)
+ temp_in[j] = output[j + i*16];
+ dct16x16_1d(temp_in, temp_out);
+ for (j = 0; j < 16; ++j)
+ output[j + i*16] = temp_out[j];
+ }
+ // Scale by some magic number
+ for (i = 0; i < 256; i++)
+ out[i] = (short)round(output[i]/2);
+}
+#endif
--- a/vp8/encoder/dct.h
+++ b/vp8/encoder/dct.h
@@ -28,6 +28,13 @@
void vp8_fht8x4_c(short *input, short *output, int pitch, TX_TYPE tx_type);
#endif
+#if CONFIG_TX16X16
+#ifndef vp8_fdct_short16x16
+#define vp8_fdct_short16x16 vp8_short_fdct16x16_c
+#endif
+extern prototype_fdct(vp8_fdct_short16x16);
+#endif
+
#ifndef vp8_fdct_short8x8
#define vp8_fdct_short8x8 vp8_short_fdct8x8_c
#endif
@@ -71,6 +78,9 @@
typedef prototype_fdct(*vp8_fdct_fn_t);
typedef struct {+#if CONFIG_TX16X16
+ vp8_fdct_fn_t short16x16;
+#endif
vp8_fdct_fn_t short8x8;
vp8_fdct_fn_t haar_short2x2;
vp8_fdct_fn_t short4x4;
--- a/vp8/encoder/encodeframe.c
+++ b/vp8/encoder/encodeframe.c
@@ -1132,6 +1132,9 @@
#endif
vp8_zero(cpi->coef_counts);
vp8_zero(cpi->coef_counts_8x8);
+#if CONFIG_TX16X16
+ vp8_zero(cpi->coef_counts_16x16);
+#endif
vp8cx_frame_init_quantizer(cpi);
@@ -1437,6 +1440,13 @@
}
/* test code: set transform size based on mode selection */
+#if CONFIG_TX16X16
+ if (x->e_mbd.mode_info_context->mbmi.mode <= TM_PRED) {+ x->e_mbd.mode_info_context->mbmi.txfm_size = TX_16X16;
+ cpi->t16x16_count++;
+ }
+ else
+#endif
if (cpi->common.txfm_mode == ALLOW_8X8
&& x->e_mbd.mode_info_context->mbmi.mode != I8X8_PRED
&& x->e_mbd.mode_info_context->mbmi.mode != B_PRED) {@@ -1470,12 +1480,9 @@
extern void vp8_fix_contexts(MACROBLOCKD *x);
-void vp8cx_encode_inter_macroblock
-(
- VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t,
- int recon_yoffset, int recon_uvoffset,
- int output_enabled
-) {+void vp8cx_encode_inter_macroblock (VP8_COMP *cpi, MACROBLOCK *x,
+ TOKENEXTRA **t, int recon_yoffset,
+ int recon_uvoffset, int output_enabled) {VP8_COMMON *cm = &cpi->common;
MACROBLOCKD *const xd = &x->e_mbd;
unsigned char *segment_id = &xd->mode_info_context->mbmi.segment_id;
@@ -1523,6 +1530,16 @@
set_pred_flag(xd, PRED_REF, ref_pred_flag);
/* test code: set transform size based on mode selection */
+#if CONFIG_TX16X16
+ if (x->e_mbd.mode_info_context->mbmi.mode <= TM_PRED ||
+ x->e_mbd.mode_info_context->mbmi.mode == NEWMV ||
+ x->e_mbd.mode_info_context->mbmi.mode == ZEROMV ||
+ x->e_mbd.mode_info_context->mbmi.mode == NEARMV ||
+ x->e_mbd.mode_info_context->mbmi.mode == NEARESTMV) {+ x->e_mbd.mode_info_context->mbmi.txfm_size = TX_16X16;
+ cpi->t16x16_count++;
+ } else
+#endif
if (cpi->common.txfm_mode == ALLOW_8X8
&& x->e_mbd.mode_info_context->mbmi.mode != I8X8_PRED
&& x->e_mbd.mode_info_context->mbmi.mode != B_PRED
--- a/vp8/encoder/encodeintra.c
+++ b/vp8/encoder/encodeintra.c
@@ -160,11 +160,21 @@
ENCODEMB_INVOKE(&rtcd->encodemb, submby)(x->src_diff, *(b->base_src), x->e_mbd.predictor, b->src_stride);
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_transform_intra_mby_16x16(x);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_transform_intra_mby_8x8(x);
else
vp8_transform_intra_mby(x);
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_quantize_mby_16x16(x);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_quantize_mby_8x8(x);
else
@@ -171,6 +181,11 @@
vp8_quantize_mby(x);
if (x->optimize) {+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_optimize_mby_16x16(x, rtcd);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_optimize_mby_8x8(x, rtcd);
else
@@ -177,6 +192,11 @@
vp8_optimize_mby(x, rtcd);
}
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_inverse_transform_mby_16x16(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_inverse_transform_mby_8x8(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
else
@@ -220,6 +240,9 @@
void vp8_encode_intra16x16mbuv(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x) {int tx_type = x->e_mbd.mode_info_context->mbmi.txfm_size;
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16) tx_type = TX_8X8; // 16x16 for U and V should default to 8x8 behavior.
+#endif
#if CONFIG_COMP_INTRA_PRED
if (x->e_mbd.mode_info_context->mbmi.second_uv_mode == (MB_PREDICTION_MODE)(DC_PRED - 1)) {#endif
--- a/vp8/encoder/encodemb.c
+++ b/vp8/encoder/encodemb.c
@@ -282,7 +282,43 @@
}
}
+#if CONFIG_TX16X16
+void vp8_transform_mbuv_16x16(MACROBLOCK *x) {+ int i;
+ vp8_clear_system_state();
+ // Default to the 8x8
+ for (i = 16; i < 24; i += 4)
+ x->vp8_short_fdct8x8(&x->block[i].src_diff[0],
+ &x->block[i].coeff[0], 16);
+}
+
+
+void vp8_transform_intra_mby_16x16(MACROBLOCK *x) {+ vp8_clear_system_state();
+ x->vp8_short_fdct16x16(&x->block[0].src_diff[0],
+ &x->block[0].coeff[0], 32);
+}
+
+
+void vp8_transform_mb_16x16(MACROBLOCK *x) {+ int i;
+ vp8_clear_system_state();
+ x->vp8_short_fdct16x16(&x->block[0].src_diff[0],
+ &x->block[0].coeff[0], 32);
+
+ for (i = 16; i < 24; i += 4) {+ x->vp8_short_fdct8x8(&x->block[i].src_diff[0],
+ &x->block[i].coeff[0], 16);
+ }
+}
+
+void vp8_transform_mby_16x16(MACROBLOCK *x) {+ vp8_clear_system_state();
+ x->vp8_short_fdct16x16(&x->block[0].src_diff[0], &x->block[0].coeff[0], 32);
+}
+#endif
+
#define RDTRUNC(RM,DM,R,D) ( (128+(R)*(RM)) & 0xFF )
#define RDTRUNC_8x8(RM,DM,R,D) ( (128+(R)*(RM)) & 0xFF )
typedef struct vp8_token_state vp8_token_state;
@@ -290,7 +326,7 @@
struct vp8_token_state {int rate;
int error;
- signed char next;
+ int next;
signed char token;
short qc;
};
@@ -1017,6 +1053,237 @@
}
+
+
+#if CONFIG_TX16X16
+#define UPDATE_RD_COST()\
+{\+ rd_cost0 = RDCOST(rdmult, rddiv, rate0, error0);\
+ rd_cost1 = RDCOST(rdmult, rddiv, rate1, error1);\
+ if (rd_cost0 == rd_cost1) {\+ rd_cost0 = RDTRUNC(rdmult, rddiv, rate0, error0);\
+ rd_cost1 = RDTRUNC(rdmult, rddiv, rate1, error1);\
+ }\
+}
+
+void optimize_b_16x16(MACROBLOCK *mb, int i, int type,
+ ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
+ const VP8_ENCODER_RTCD *rtcd) {+ BLOCK *b = &mb->block[i];
+ BLOCKD *d = &mb->e_mbd.block[i];
+ vp8_token_state tokens[257][2];
+ unsigned best_index[257][2];
+ const short *dequant_ptr = d->dequant, *coeff_ptr = b->coeff;
+ short *qcoeff_ptr = qcoeff_ptr = d->qcoeff;
+ short *dqcoeff_ptr = dqcoeff_ptr = d->dqcoeff;
+ int eob = d->eob, final_eob, sz = 0;
+ int rc, x, next;
+ int64_t rdmult, rddiv, rd_cost0, rd_cost1;
+ int rate0, rate1, error0, error1, t0, t1;
+ int best, band, pt;
+ int err_mult = plane_rd_mult[type];
+
+ /* Now set up a Viterbi trellis to evaluate alternative roundings. */
+ rdmult = mb->rdmult * err_mult;
+ if (mb->e_mbd.mode_info_context->mbmi.ref_frame == INTRA_FRAME)
+ rdmult = (rdmult * 9)>>4;
+ rddiv = mb->rddiv;
+ memset(best_index, 0, sizeof(best_index));
+ /* Initialize the sentinel node of the trellis. */
+ tokens[eob][0].rate = 0;
+ tokens[eob][0].error = 0;
+ tokens[eob][0].next = 256;
+ tokens[eob][0].token = DCT_EOB_TOKEN;
+ tokens[eob][0].qc = 0;
+ *(tokens[eob] + 1) = *(tokens[eob] + 0);
+ next = eob;
+ for (i = eob; i-- > 0;) {+ int base_bits, d2, dx;
+
+ rc = vp8_default_zig_zag1d_16x16[i];
+ x = qcoeff_ptr[rc];
+ /* Only add a trellis state for non-zero coefficients. */
+ if (x) {+ int shortcut = 0;
+ error0 = tokens[next][0].error;
+ error1 = tokens[next][1].error;
+ /* Evaluate the first possibility for this state. */
+ rate0 = tokens[next][0].rate;
+ rate1 = tokens[next][1].rate;
+ t0 = (vp8_dct_value_tokens_ptr + x)->Token;
+ /* Consider both possible successor states. */
+ if (next < 256) {+ band = vp8_coef_bands_16x16[i + 1];
+ pt = vp8_prev_token_class[t0];
+ rate0 += mb->token_costs_16x16[type][band][pt][tokens[next][0].token];
+ rate1 += mb->token_costs_16x16[type][band][pt][tokens[next][1].token];
+ }
+ UPDATE_RD_COST();
+ /* And pick the best. */
+ best = rd_cost1 < rd_cost0;
+ base_bits = *(vp8_dct_value_cost_ptr + x);
+ dx = dqcoeff_ptr[rc] - coeff_ptr[rc];
+ d2 = dx*dx;
+ tokens[i][0].rate = base_bits + (best ? rate1 : rate0);
+ tokens[i][0].error = d2 + (best ? error1 : error0);
+ tokens[i][0].next = next;
+ tokens[i][0].token = t0;
+ tokens[i][0].qc = x;
+ best_index[i][0] = best;
+ /* Evaluate the second possibility for this state. */
+ rate0 = tokens[next][0].rate;
+ rate1 = tokens[next][1].rate;
+
+ if((abs(x)*dequant_ptr[rc!=0]>abs(coeff_ptr[rc])) &&
+ (abs(x)*dequant_ptr[rc!=0]<abs(coeff_ptr[rc])+dequant_ptr[rc!=0]))
+ shortcut = 1;
+ else
+ shortcut = 0;
+
+ if (shortcut) {+ sz = -(x < 0);
+ x -= 2*sz + 1;
+ }
+
+ /* Consider both possible successor states. */
+ if (!x) {+ /* If we reduced this coefficient to zero, check to see if
+ * we need to move the EOB back here.
+ */
+ t0 = tokens[next][0].token == DCT_EOB_TOKEN ?
+ DCT_EOB_TOKEN : ZERO_TOKEN;
+ t1 = tokens[next][1].token == DCT_EOB_TOKEN ?
+ DCT_EOB_TOKEN : ZERO_TOKEN;
+ }
+ else
+ t0=t1 = (vp8_dct_value_tokens_ptr + x)->Token;
+ if (next < 256) {+ band = vp8_coef_bands_16x16[i + 1];
+ if (t0 != DCT_EOB_TOKEN) {+ pt = vp8_prev_token_class[t0];
+ rate0 += mb->token_costs_16x16[type][band][pt]
+ [tokens[next][0].token];
+ }
+ if (t1!=DCT_EOB_TOKEN) {+ pt = vp8_prev_token_class[t1];
+ rate1 += mb->token_costs_16x16[type][band][pt]
+ [tokens[next][1].token];
+ }
+ }
+ UPDATE_RD_COST();
+ /* And pick the best. */
+ best = rd_cost1 < rd_cost0;
+ base_bits = *(vp8_dct_value_cost_ptr + x);
+
+ if(shortcut) {+ dx -= (dequant_ptr[rc!=0] + sz) ^ sz;
+ d2 = dx*dx;
+ }
+ tokens[i][1].rate = base_bits + (best ? rate1 : rate0);
+ tokens[i][1].error = d2 + (best ? error1 : error0);
+ tokens[i][1].next = next;
+ tokens[i][1].token = best ? t1 : t0;
+ tokens[i][1].qc = x;
+ best_index[i][1] = best;
+ /* Finally, make this the new head of the trellis. */
+ next = i;
+ }
+ /* There's no choice to make for a zero coefficient, so we don't
+ * add a new trellis node, but we do need to update the costs.
+ */
+ else {+ band = vp8_coef_bands_16x16[i + 1];
+ t0 = tokens[next][0].token;
+ t1 = tokens[next][1].token;
+ /* Update the cost of each path if we're past the EOB token. */
+ if (t0 != DCT_EOB_TOKEN) {+ tokens[next][0].rate += mb->token_costs_16x16[type][band][0][t0];
+ tokens[next][0].token = ZERO_TOKEN;
+ }
+ if (t1 != DCT_EOB_TOKEN) {+ tokens[next][1].rate += mb->token_costs_16x16[type][band][0][t1];
+ tokens[next][1].token = ZERO_TOKEN;
+ }
+ /* Don't update next, because we didn't add a new node. */
+ }
+ }
+
+ /* Now pick the best path through the whole trellis. */
+ band = vp8_coef_bands_16x16[i + 1];
+ VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l);
+ rate0 = tokens[next][0].rate;
+ rate1 = tokens[next][1].rate;
+ error0 = tokens[next][0].error;
+ error1 = tokens[next][1].error;
+ t0 = tokens[next][0].token;
+ t1 = tokens[next][1].token;
+ rate0 += mb->token_costs_16x16[type][band][pt][t0];
+ rate1 += mb->token_costs_16x16[type][band][pt][t1];
+ UPDATE_RD_COST();
+ best = rd_cost1 < rd_cost0;
+ final_eob = -1;
+
+ for (i = next; i < eob; i = next) {+ x = tokens[i][best].qc;
+ if (x)
+ final_eob = i;
+ rc = vp8_default_zig_zag1d_16x16[i];
+ qcoeff_ptr[rc] = x;
+ dqcoeff_ptr[rc] = (x * dequant_ptr[rc!=0]);
+
+ next = tokens[i][best].next;
+ best = best_index[i][best];
+ }
+ final_eob++;
+
+ d->eob = final_eob;
+ *a = *l = (d->eob != !type);
+}
+
+void vp8_optimize_mby_16x16(MACROBLOCK *x, const VP8_ENCODER_RTCD *rtcd) {+ ENTROPY_CONTEXT_PLANES t_above, t_left;
+ ENTROPY_CONTEXT *ta, *tl;
+
+ if (!x->e_mbd.above_context)
+ return;
+ if (!x->e_mbd.left_context)
+ return;
+
+ vpx_memcpy(&t_above, x->e_mbd.above_context, sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(&t_left, x->e_mbd.left_context, sizeof(ENTROPY_CONTEXT_PLANES));
+
+ ta = (ENTROPY_CONTEXT *)&t_above;
+ tl = (ENTROPY_CONTEXT *)&t_left;
+ optimize_b_16x16(x, 0, PLANE_TYPE_Y_WITH_DC, ta, tl, rtcd);
+ *(ta + 1) = *ta;
+ *(tl + 1) = *tl;
+}
+
+void optimize_mb_16x16(MACROBLOCK *x, const VP8_ENCODER_RTCD *rtcd) {+ int b;
+ ENTROPY_CONTEXT_PLANES t_above, t_left;
+ ENTROPY_CONTEXT *ta, *tl;
+
+ vpx_memcpy(&t_above, x->e_mbd.above_context, sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(&t_left, x->e_mbd.left_context, sizeof(ENTROPY_CONTEXT_PLANES));
+
+ ta = (ENTROPY_CONTEXT *)&t_above;
+ tl = (ENTROPY_CONTEXT *)&t_left;
+
+ optimize_b_16x16(x, 0, PLANE_TYPE_Y_WITH_DC, ta, tl, rtcd);
+ *(ta + 1) = *ta;
+ *(tl + 1) = *tl;
+
+ for (b = 16; b < 24; b += 4) {+ optimize_b_8x8(x, b, PLANE_TYPE_UV,
+ ta + vp8_block2above_8x8[b], tl + vp8_block2left_8x8[b],
+ rtcd);
+ *(ta + vp8_block2above_8x8[b] + 1) = *(ta + vp8_block2above_8x8[b]);
+ *(tl + vp8_block2left_8x8[b] + 1) = *(tl + vp8_block2left_8x8[b]);
+ }
+}
+#endif
+
void vp8_encode_inter16x16(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x) {int tx_type = x->e_mbd.mode_info_context->mbmi.txfm_size;
vp8_build_inter_predictors_mb(&x->e_mbd);
@@ -1023,11 +1290,21 @@
vp8_subtract_mb(rtcd, x);
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_transform_mb_16x16(x);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_transform_mb_8x8(x);
else
transform_mb(x);
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_quantize_mb_16x16(x);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_quantize_mb_8x8(x);
else
@@ -1034,6 +1311,11 @@
vp8_quantize_mb(x);
if (x->optimize) {+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ optimize_mb_16x16(x, rtcd);
+ else
+#endif
if (tx_type == TX_8X8)
optimize_mb_8x8(x, rtcd);
else
@@ -1040,6 +1322,11 @@
optimize_mb(x, rtcd);
}
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_inverse_transform_mb_16x16(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_inverse_transform_mb_8x8(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
else
@@ -1111,6 +1398,11 @@
ENCODEMB_INVOKE(&rtcd->encodemb, submby)(x->src_diff, *(b->base_src), x->e_mbd.predictor, b->src_stride);
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_transform_mby_16x16(x);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_transform_mby_8x8(x);
else
@@ -1118,6 +1410,11 @@
vp8_quantize_mby(x);
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_inverse_transform_mby_16x16(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_inverse_transform_mby_8x8(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
else
@@ -1126,3 +1423,4 @@
RECON_INVOKE(&rtcd->common->recon, recon_mby)
(IF_RTCD(&rtcd->common->recon), &x->e_mbd);
}
+
--- a/vp8/encoder/encodemb.h
+++ b/vp8/encoder/encodemb.h
@@ -121,6 +121,15 @@
void vp8_optimize_mby_8x8(MACROBLOCK *x, const struct VP8_ENCODER_RTCD *rtcd);
void vp8_optimize_mbuv_8x8(MACROBLOCK *x, const struct VP8_ENCODER_RTCD *rtcd);
+#if CONFIG_TX16X16
+void vp8_transform_mb_16x16(MACROBLOCK *mb);
+void vp8_transform_mby_16x16(MACROBLOCK *x);
+void vp8_transform_mbuv_16x16(MACROBLOCK *x);
+void vp8_transform_intra_mby_16x16(MACROBLOCK *x);
+void vp8_build_dcblock_16x16(MACROBLOCK *b);
+void vp8_optimize_mby_16x16(MACROBLOCK *x, const struct VP8_ENCODER_RTCD *rtcd);
+#endif
+
void vp8_subtract_4b_c(BLOCK *be, BLOCKD *bd, int pitch);
#endif
--- a/vp8/encoder/generic/csystemdependent.c
+++ b/vp8/encoder/generic/csystemdependent.c
@@ -69,6 +69,9 @@
cpi->rtcd.variance.getmbss = vp8_get_mb_ss_c;
cpi->rtcd.fdct.short8x8 = vp8_short_fdct8x8_c;
+#if CONFIG_TX16X16
+ cpi->rtcd.fdct.short16x16 = vp8_short_fdct16x16_c;
+#endif
cpi->rtcd.fdct.haar_short2x2 = vp8_short_fhaar2x2_c;
cpi->rtcd.fdct.short4x4 = vp8_short_fdct4x4_c;
cpi->rtcd.fdct.short8x4 = vp8_short_fdct8x4_c;
--- a/vp8/encoder/onyx_if.c
+++ b/vp8/encoder/onyx_if.c
@@ -1161,10 +1161,16 @@
}
if (cpi->sf.improved_dct) {+#if CONFIG_TX16X16
+ cpi->mb.vp8_short_fdct16x16 = FDCT_INVOKE(&cpi->rtcd.fdct, short16x16);
+#endif
cpi->mb.vp8_short_fdct8x8 = FDCT_INVOKE(&cpi->rtcd.fdct, short8x8);
cpi->mb.vp8_short_fdct8x4 = FDCT_INVOKE(&cpi->rtcd.fdct, short8x4);
cpi->mb.vp8_short_fdct4x4 = FDCT_INVOKE(&cpi->rtcd.fdct, short4x4);
} else {+#if CONFIG_TX16X16
+ cpi->mb.vp8_short_fdct16x16 = FDCT_INVOKE(&cpi->rtcd.fdct, short16x16);
+#endif
cpi->mb.vp8_short_fdct8x8 = FDCT_INVOKE(&cpi->rtcd.fdct, short8x8);
cpi->mb.vp8_short_fdct8x4 = FDCT_INVOKE(&cpi->rtcd.fdct, fast8x4);
cpi->mb.vp8_short_fdct4x4 = FDCT_INVOKE(&cpi->rtcd.fdct, fast4x4);
@@ -1177,6 +1183,9 @@
cpi->mb.quantize_b = vp8_regular_quantize_b;
cpi->mb.quantize_b_pair = vp8_regular_quantize_b_pair;
cpi->mb.quantize_b_8x8 = vp8_regular_quantize_b_8x8;
+#if CONFIG_TX16X16
+ cpi->mb.quantize_b_16x16= vp8_regular_quantize_b_16x16;
+#endif
cpi->mb.quantize_b_2x2 = vp8_regular_quantize_b_2x2;
vp8cx_init_quantizer(cpi);
@@ -3641,6 +3650,9 @@
update_reference_frames(cm);
vp8_copy(cpi->common.fc.coef_counts, cpi->coef_counts);
vp8_copy(cpi->common.fc.coef_counts_8x8, cpi->coef_counts_8x8);
+#if CONFIG_TX16X16
+ vp8_copy(cpi->common.fc.coef_counts_16x16, cpi->coef_counts_16x16);
+#endif
vp8_adapt_coef_probs(&cpi->common);
if (cpi->common.frame_type != KEY_FRAME) {vp8_copy(cpi->common.fc.ymode_counts, cpi->ymode_count);
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -91,9 +91,13 @@
signed char last_mode_lf_deltas[MAX_MODE_LF_DELTAS];
vp8_prob coef_probs[BLOCK_TYPES]
- [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES];
+ [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES];
vp8_prob coef_probs_8x8[BLOCK_TYPES_8X8]
- [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES];
+ [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES];
+#if CONFIG_TX16X16
+ vp8_prob coef_probs_16x16[BLOCK_TYPES_16X16]
+ [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES];
+#endif
vp8_prob ymode_prob [VP8_YMODES - 1]; /* interframe intra mode probs */
vp8_prob uv_mode_prob [VP8_YMODES][VP8_UV_MODES - 1];
@@ -390,6 +394,15 @@
DECLARE_ALIGNED(64, short, zrun_zbin_boost_y2_8x8[QINDEX_RANGE][64]);
DECLARE_ALIGNED(64, short, zrun_zbin_boost_uv_8x8[QINDEX_RANGE][64]);
+#if CONFIG_TX16X16
+ DECLARE_ALIGNED(16, short, Y1zbin_16x16[QINDEX_RANGE][256]);
+ DECLARE_ALIGNED(16, short, Y2zbin_16x16[QINDEX_RANGE][256]);
+ DECLARE_ALIGNED(16, short, UVzbin_16x16[QINDEX_RANGE][256]);
+ DECLARE_ALIGNED(16, short, zrun_zbin_boost_y1_16x16[QINDEX_RANGE][256]);
+ DECLARE_ALIGNED(16, short, zrun_zbin_boost_y2_16x16[QINDEX_RANGE][256]);
+ DECLARE_ALIGNED(16, short, zrun_zbin_boost_uv_16x16[QINDEX_RANGE][256]);
+#endif
+
MACROBLOCK mb;
VP8_COMMON common;
vp8_writer bc, bc2;
@@ -540,6 +553,11 @@
unsigned int coef_counts_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */
vp8_prob frame_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
unsigned int frame_branch_ct_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2];
+#if CONFIG_TX16X16
+ unsigned int coef_counts_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */
+ vp8_prob frame_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+ unsigned int frame_branch_ct_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2];
+#endif
int gfu_boost;
int last_boost;
@@ -598,6 +616,9 @@
int skip_false_count[3];
int t4x4_count;
int t8x8_count;
+#if CONFIG_TX16X16
+ int t16x16_count;
+#endif
unsigned char *segmentation_map;
--- a/vp8/encoder/quantize.c
+++ b/vp8/encoder/quantize.c
@@ -302,9 +302,8 @@
void vp8_quantize_mbuv_8x8(MACROBLOCK *x) {int i;
- for (i = 16; i < 24; i ++) {+ for (i = 16; i < 24; i ++)
x->e_mbd.block[i].eob = 0;
- }
for (i = 16; i < 24; i += 4)
x->quantize_b_8x8(&x->block[i], &x->e_mbd.block[i]);
}
@@ -311,6 +310,85 @@
+
+#if CONFIG_TX16X16
+void vp8_quantize_mby_16x16(MACROBLOCK *x) {+ int i;
+ for (i = 0; i < 16; i++)
+ x->e_mbd.block[i].eob = 0;
+ x->e_mbd.block[24].eob = 0;
+ x->quantize_b_16x16(&x->block[0], &x->e_mbd.block[0]);
+}
+
+void vp8_quantize_mb_16x16(MACROBLOCK *x) {+ int i;
+ for(i = 0; i < 25; i++)
+ x->e_mbd.block[i].eob = 0;
+ x->quantize_b_16x16(&x->block[0], &x->e_mbd.block[0]);
+ for (i = 16; i < 24; i += 4)
+ x->quantize_b_8x8(&x->block[i], &x->e_mbd.block[i]);
+}
+
+// U and V should use 8x8
+void vp8_quantize_mbuv_16x16(MACROBLOCK *x) {+ int i;
+
+ for(i = 16; i < 24; i++)
+ x->e_mbd.block[i].eob = 0;
+ for (i = 16; i < 24; i += 4)
+ x->quantize_b_8x8(&x->block[i], &x->e_mbd.block[i]);
+}
+
+void vp8_regular_quantize_b_16x16(BLOCK *b, BLOCKD *d) {+ int i, rc, eob;
+ int zbin;
+ int x, y, z, sz;
+ short *zbin_boost_ptr = b->zrun_zbin_boost_16x16;
+ short *coeff_ptr = b->coeff;
+ short *zbin_ptr = b->zbin_16x16;
+ short *round_ptr = b->round;
+ short *quant_ptr = b->quant;
+ unsigned char *quant_shift_ptr = b->quant_shift;
+ short *qcoeff_ptr = d->qcoeff;
+ short *dqcoeff_ptr = d->dqcoeff;
+ short *dequant_ptr = d->dequant;
+ short zbin_oq_value = b->zbin_extra;
+
+ vpx_memset(qcoeff_ptr, 0, 256*sizeof(short));
+ vpx_memset(dqcoeff_ptr, 0, 256*sizeof(short));
+
+ eob = -1;
+ for (i = 0; i < b->eob_max_offset_16x16; i++) {+ rc = vp8_default_zig_zag1d_16x16[i];
+ z = coeff_ptr[rc];
+
+ zbin = (zbin_ptr[rc!=0] + *zbin_boost_ptr + zbin_oq_value);
+ zbin_boost_ptr ++;
+
+ sz = (z >> 31); // sign of z
+ x = (z ^ sz) - sz; // x = abs(z)
+
+ if (x >= zbin) {+ x += (round_ptr[rc!=0]);
+ y = ((int)(((int)(x * quant_ptr[rc!=0]) >> 16) + x))
+ >> quant_shift_ptr[rc!=0]; // quantize (x)
+ x = (y ^ sz) - sz; // get the sign back
+ qcoeff_ptr[rc] = x; // write to destination
+ dqcoeff_ptr[rc] = x * dequant_ptr[rc!=0]; // dequantized value
+
+ if (y) {+ eob = i; // last nonzero coeffs
+ zbin_boost_ptr = b->zrun_zbin_boost_16x16;
+ }
+ }
+ }
+
+ d->eob = eob + 1;
+}
+#endif
+
+
+
/* quantize_b_pair function pointer in MACROBLOCK structure is set to one of
* these two C functions if corresponding optimized routine is not available.
* NEON optimized version implements currently the fast quantization for pair
@@ -337,20 +415,39 @@
int i;
int quant_val;
int Q;
- int zbin_boost[16] = { 0, 0, 8, 10, 12, 14, 16, 20,- 24, 28, 32, 36, 40, 44, 44, 44
- };
+ static const int zbin_boost[16] = { 0, 0, 8, 10, 12, 14, 16, 20,+ 24, 28, 32, 36, 40, 44, 44, 44
+ };
- int zbin_boost_8x8[64] = { 0, 0, 0, 8, 8, 8, 10, 12,- 14, 16, 18, 20, 22, 24, 26, 28,
- 30, 32, 34, 36, 38, 40, 42, 44,
- 46, 48, 48, 48, 48, 48, 48, 48,
- 48, 48, 48, 48, 48, 48, 48, 48,
- 48, 48, 48, 48, 48, 48, 48, 48,
- 48, 48, 48, 48, 48, 48, 48, 48,
- 48, 48, 48, 48, 48, 48, 48, 48
- };
-
+ static const int zbin_boost_8x8[64] = { 0, 0, 0, 8, 8, 8, 10, 12,+ 14, 16, 18, 20, 22, 24, 26, 28,
+ 30, 32, 34, 36, 38, 40, 42, 44,
+ 46, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48
+ };
+#if CONFIG_TX16X16
+ static const int zbin_boost_16x16[256] = {+ 0, 0, 0, 8, 8, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28,
+ 30, 32, 34, 36, 38, 40, 42, 44, 46, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+ };
+#endif
int qrounding_factor = 48;
@@ -372,33 +469,52 @@
cpi->Y1quant_shift[Q] + 0, quant_val);
cpi->Y1zbin[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
cpi->Y1zbin_8x8[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+#if CONFIG_TX16X16
+ cpi->Y1zbin_16x16[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+#endif
cpi->Y1round[Q][0] = (qrounding_factor * quant_val) >> 7;
cpi->common.Y1dequant[Q][0] = quant_val;
cpi->zrun_zbin_boost_y1[Q][0] = (quant_val * zbin_boost[0]) >> 7;
cpi->zrun_zbin_boost_y1_8x8[Q][0] =
((quant_val * zbin_boost_8x8[0]) + 64) >> 7;
+#if CONFIG_TX16X16
+ cpi->zrun_zbin_boost_y1_16x16[Q][0] = ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
+#endif
+
quant_val = vp8_dc2quant(Q, cpi->common.y2dc_delta_q);
invert_quant(cpi->Y2quant[Q] + 0,
cpi->Y2quant_shift[Q] + 0, quant_val);
cpi->Y2zbin[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
cpi->Y2zbin_8x8[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+#if CONFIG_TX16X16
+ cpi->Y2zbin_16x16[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+#endif
cpi->Y2round[Q][0] = (qrounding_factor * quant_val) >> 7;
cpi->common.Y2dequant[Q][0] = quant_val;
cpi->zrun_zbin_boost_y2[Q][0] = (quant_val * zbin_boost[0]) >> 7;
cpi->zrun_zbin_boost_y2_8x8[Q][0] =
((quant_val * zbin_boost_8x8[0]) + 64) >> 7;
+#if CONFIG_TX16X16
+ cpi->zrun_zbin_boost_y2_16x16[Q][0] = ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
+#endif
quant_val = vp8_dc_uv_quant(Q, cpi->common.uvdc_delta_q);
invert_quant(cpi->UVquant[Q] + 0,
cpi->UVquant_shift[Q] + 0, quant_val);
- cpi->UVzbin[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;;
- cpi->UVzbin_8x8[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;;
+ cpi->UVzbin[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+ cpi->UVzbin_8x8[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+#if CONFIG_TX16X16
+ cpi->UVzbin_16x16[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+#endif
cpi->UVround[Q][0] = (qrounding_factor * quant_val) >> 7;
cpi->common.UVdequant[Q][0] = quant_val;
cpi->zrun_zbin_boost_uv[Q][0] = (quant_val * zbin_boost[0]) >> 7;
cpi->zrun_zbin_boost_uv_8x8[Q][0] =
((quant_val * zbin_boost_8x8[0]) + 64) >> 7;
+#if CONFIG_TX16X16
+ cpi->zrun_zbin_boost_uv_16x16[Q][0] = ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
+#endif
// all the 4x4 ac values =;
for (i = 1; i < 16; i++) {@@ -453,6 +569,25 @@
cpi->zrun_zbin_boost_uv_8x8[Q][i] =
((quant_val * zbin_boost_8x8[i]) + 64) >> 7;
}
+
+#if CONFIG_TX16X16
+ // 16x16 structures. Same comment above applies.
+ for (i = 1; i < 256; i++) {+ int rc = vp8_default_zig_zag1d_16x16[i];
+
+ quant_val = vp8_ac_yquant(Q);
+ cpi->Y1zbin_16x16[Q][rc] = ((qzbin_factor * quant_val) + 64) >> 7;
+ cpi->zrun_zbin_boost_y1_16x16[Q][i] = ((quant_val * zbin_boost_16x16[i]) + 64) >> 7;
+
+ quant_val = vp8_ac2quant(Q, cpi->common.y2ac_delta_q);
+ cpi->Y2zbin_16x16[Q][rc] = ((qzbin_factor * quant_val) + 64) >> 7;
+ cpi->zrun_zbin_boost_y2_16x16[Q][i] = ((quant_val * zbin_boost_16x16[i]) + 64) >> 7;
+
+ quant_val = vp8_ac_uv_quant(Q, cpi->common.uvac_delta_q);
+ cpi->UVzbin_16x16[Q][rc] = ((qzbin_factor * quant_val) + 64) >> 7;
+ cpi->zrun_zbin_boost_uv_16x16[Q][i] = ((quant_val * zbin_boost_16x16[i]) + 64) >> 7;
+ }
+#endif
}
}
@@ -491,10 +626,16 @@
x->block[i].quant_shift = cpi->Y1quant_shift[QIndex];
x->block[i].zbin = cpi->Y1zbin[QIndex];
x->block[i].zbin_8x8 = cpi->Y1zbin_8x8[QIndex];
+#if CONFIG_TX16X16
+ x->block[i].zbin_16x16 = cpi->Y1zbin_16x16[QIndex];
+#endif
x->block[i].round = cpi->Y1round[QIndex];
x->e_mbd.block[i].dequant = cpi->common.Y1dequant[QIndex];
x->block[i].zrun_zbin_boost = cpi->zrun_zbin_boost_y1[QIndex];
x->block[i].zrun_zbin_boost_8x8 = cpi->zrun_zbin_boost_y1_8x8[QIndex];
+#if CONFIG_TX16X16
+ x->block[i].zrun_zbin_boost_16x16 = cpi->zrun_zbin_boost_y1_16x16[QIndex];
+#endif
x->block[i].zbin_extra = (short)zbin_extra;
// Segment max eob offset feature.
@@ -503,9 +644,16 @@
get_segdata(xd, segment_id, SEG_LVL_EOB);
x->block[i].eob_max_offset_8x8 =
get_segdata(xd, segment_id, SEG_LVL_EOB);
+#if CONFIG_TX16X16
+ x->block[i].eob_max_offset_16x16 =
+ get_segdata(xd, segment_id, SEG_LVL_EOB);
+#endif
} else {x->block[i].eob_max_offset = 16;
x->block[i].eob_max_offset_8x8 = 64;
+#if CONFIG_TX16X16
+ x->block[i].eob_max_offset_16x16 = 256;
+#endif
}
}
@@ -520,10 +668,16 @@
x->block[i].quant_shift = cpi->UVquant_shift[QIndex];
x->block[i].zbin = cpi->UVzbin[QIndex];
x->block[i].zbin_8x8 = cpi->UVzbin_8x8[QIndex];
+#if CONFIG_TX16X16
+ x->block[i].zbin_16x16 = cpi->UVzbin_16x16[QIndex];
+#endif
x->block[i].round = cpi->UVround[QIndex];
x->e_mbd.block[i].dequant = cpi->common.UVdequant[QIndex];
x->block[i].zrun_zbin_boost = cpi->zrun_zbin_boost_uv[QIndex];
x->block[i].zrun_zbin_boost_8x8 = cpi->zrun_zbin_boost_uv_8x8[QIndex];
+#if CONFIG_TX16X16
+ x->block[i].zrun_zbin_boost_16x16 = cpi->zrun_zbin_boost_uv_16x16[QIndex];
+#endif
x->block[i].zbin_extra = (short)zbin_extra;
@@ -549,10 +703,16 @@
x->block[24].quant_shift = cpi->Y2quant_shift[QIndex];
x->block[24].zbin = cpi->Y2zbin[QIndex];
x->block[24].zbin_8x8 = cpi->Y2zbin_8x8[QIndex];
+#if CONFIG_TX16X16
+ x->block[24].zbin_16x16 = cpi->Y2zbin_16x16[QIndex];
+#endif
x->block[24].round = cpi->Y2round[QIndex];
x->e_mbd.block[24].dequant = cpi->common.Y2dequant[QIndex];
x->block[24].zrun_zbin_boost = cpi->zrun_zbin_boost_y2[QIndex];
x->block[24].zrun_zbin_boost_8x8 = cpi->zrun_zbin_boost_y2_8x8[QIndex];
+#if CONFIG_TX16X16
+ x->block[24].zrun_zbin_boost_16x16 = cpi->zrun_zbin_boost_y2_16x16[QIndex];
+#endif
x->block[24].zbin_extra = (short)zbin_extra;
// TBD perhaps not use for Y2
--- a/vp8/encoder/quantize.h
+++ b/vp8/encoder/quantize.h
@@ -46,6 +46,13 @@
#endif
extern prototype_quantize_block(vp8_quantize_quantb_8x8);
+#if CONFIG_TX16X16
+#ifndef vp8_quantize_quantb_16x16
+#define vp8_quantize_quantb_16x16 vp8_regular_quantize_b_16x16
+#endif
+extern prototype_quantize_block(vp8_quantize_quantb_16x16);
+#endif
+
#ifndef vp8_quantize_quantb_2x2
#define vp8_quantize_quantb_2x2 vp8_regular_quantize_b_2x2
#endif
@@ -69,6 +76,13 @@
extern prototype_quantize_mb(vp8_quantize_mby_8x8);
extern prototype_quantize_mb(vp8_quantize_mbuv_8x8);
+
+#if CONFIG_TX16X16
+void vp8_quantize_mb_16x16(MACROBLOCK *x);
+extern prototype_quantize_block(vp8_quantize_quantb_16x16);
+extern prototype_quantize_mb(vp8_quantize_mby_16x16);
+extern prototype_quantize_mb(vp8_quantize_mbuv_16x16);
+#endif
struct VP8_COMP;
extern void vp8_set_quantizer(struct VP8_COMP *cpi, int Q);
--- a/vp8/encoder/ratectrl.c
+++ b/vp8/encoder/ratectrl.c
@@ -177,6 +177,9 @@
#if CONFIG_SWITCHABLE_INTERP
vp8_copy(cc->switchable_interp_prob, cm->fc.switchable_interp_prob);
#endif
+#if CONFIG_TX16X16
+ vp8_copy(cc->coef_probs_16x16, cm->fc.coef_probs_16x16);
+#endif
}
void vp8_restore_coding_context(VP8_COMP *cpi) {@@ -232,6 +235,9 @@
vp8_copy(cm->fc.coef_probs_8x8, cc->coef_probs_8x8);
#if CONFIG_SWITCHABLE_INTERP
vp8_copy(cm->fc.switchable_interp_prob, cc->switchable_interp_prob);
+#endif
+#if CONFIG_TX16X16
+ vp8_copy(cm->fc.coef_probs_16x16, cc->coef_probs_16x16);
#endif
}
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -366,6 +366,13 @@
(const vp8_prob( *)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.coef_probs_8x8,
BLOCK_TYPES_8X8);
+#if CONFIG_TX16X16
+ fill_token_costs(
+ cpi->mb.token_costs_16x16,
+ (const vp8_prob(*)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.coef_probs_16x16,
+ BLOCK_TYPES_16X16);
+#endif
+
/*rough estimate for costing*/
cpi->common.kf_ymode_probs_index = cpi->common.base_qindex >> 4;
vp8_init_mode_costs(cpi);
@@ -809,6 +816,72 @@
*Rate = vp8_rdcost_mby_8x8(mb);
}
+#if CONFIG_TX16X16
+static int cost_coeffs_16x16(MACROBLOCK *mb, BLOCKD *b, int type,
+ ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l) {+ const int eob = b->eob;
+ int c = !type; /* start at coef 0, unless Y with Y2 */
+ int cost = 0, pt; /* surrounding block/prev coef predictor */
+ short *qcoeff_ptr = b->qcoeff;
+
+ VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l);
+
+# define QC16X16(I) ( qcoeff_ptr [vp8_default_zig_zag1d_16x16[I]] )
+
+ for (; c < eob; c++) {+ int v = QC16X16(c);
+ int t = vp8_dct_value_tokens_ptr[v].Token;
+ cost += mb->token_costs_16x16[type][vp8_coef_bands_16x16[c]][pt][t];
+ cost += vp8_dct_value_cost_ptr[v];
+ pt = vp8_prev_token_class[t];
+ }
+
+# undef QC16X16
+ if (c < 256)
+ cost += mb->token_costs_16x16[type][vp8_coef_bands_16x16[c]]
+ [pt][DCT_EOB_TOKEN];
+
+ pt = (c != !type); // is eob first coefficient;
+ *a = *l = pt;
+ return cost;
+}
+
+static int vp8_rdcost_mby_16x16(MACROBLOCK *mb) {+ int cost;
+ MACROBLOCKD *x = &mb->e_mbd;
+ ENTROPY_CONTEXT_PLANES t_above, t_left;
+ ENTROPY_CONTEXT *ta, *tl;
+
+ vpx_memcpy(&t_above, mb->e_mbd.above_context, sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(&t_left, mb->e_mbd.left_context, sizeof(ENTROPY_CONTEXT_PLANES));
+
+ ta = (ENTROPY_CONTEXT *)&t_above;
+ tl = (ENTROPY_CONTEXT *)&t_left;
+
+ cost = cost_coeffs_16x16(mb, x->block, PLANE_TYPE_Y_WITH_DC, ta, tl);
+ return cost;
+}
+
+static void macro_block_yrd_16x16(MACROBLOCK *mb, int *Rate, int *Distortion,
+ const VP8_ENCODER_RTCD *rtcd) {+ int d;
+
+ ENCODEMB_INVOKE(&rtcd->encodemb, submby)(
+ mb->src_diff,
+ *(mb->block[0].base_src),
+ mb->e_mbd.predictor,
+ mb->block[0].src_stride);
+
+ vp8_transform_mby_16x16(mb);
+ vp8_quantize_mby_16x16(mb);
+ d = ENCODEMB_INVOKE(&rtcd->encodemb, mberr)(mb, 0);
+
+ *Distortion = (d >> 2);
+ // rate
+ *Rate = vp8_rdcost_mby_16x16(mb);
+}
+#endif
+
static void copy_predictor(unsigned char *dst, const unsigned char *predictor) {const unsigned int *p = (const unsigned int *)predictor;
unsigned int *d = (unsigned int *)dst;
@@ -1121,7 +1194,12 @@
}
#endif
- macro_block_yrd_8x8(x, &ratey, &distortion, IF_RTCD(&cpi->rtcd));
+#if CONFIG_TX16X16
+ if (mode <= TM_PRED)
+ macro_block_yrd_16x16(x, &ratey, &distortion, IF_RTCD(&cpi->rtcd));
+ else
+#endif
+ macro_block_yrd_8x8(x, &ratey, &distortion, IF_RTCD(&cpi->rtcd));
// FIXME add compoundmode cost
// FIXME add rate for mode2
rate = ratey + x->mbmode_cost[x->e_mbd.frame_type]
@@ -3081,6 +3159,10 @@
vp8_cost_bit(get_pred_prob(cm, xd, PRED_COMP), 0);
}
break;
+ case DC_PRED:
+ case V_PRED:
+ case H_PRED:
+ case TM_PRED:
case D45_PRED:
case D135_PRED:
case D117_PRED:
@@ -3087,13 +3169,26 @@
case D153_PRED:
case D27_PRED:
case D63_PRED:
- case DC_PRED:
- case V_PRED:
- case H_PRED:
- case TM_PRED:
+#if CONFIG_TX16X16
+ // FIXME: breaks lossless since 4x4 isn't allowed
x->e_mbd.mode_info_context->mbmi.ref_frame = INTRA_FRAME;
// FIXME compound intra prediction
RECON_INVOKE(&cpi->common.rtcd.recon, build_intra_predictors_mby)
+ (&x->e_mbd);
+ macro_block_yrd_16x16(x, &rate_y, &distortion,
+ IF_RTCD(&cpi->rtcd));
+ rate2 += rate_y;
+ distortion2 += distortion;
+ rate2 += x->mbmode_cost[x->e_mbd.frame_type][x->e_mbd.mode_info_context->mbmi.mode];
+ rate2 += uv_intra_rate_8x8;
+ rate_uv = uv_intra_rate_tokenonly_8x8;
+ distortion2 += uv_intra_distortion_8x8;
+ distortion_uv = uv_intra_distortion_8x8;
+ break;
+#else
+ x->e_mbd.mode_info_context->mbmi.ref_frame = INTRA_FRAME;
+ // FIXME compound intra prediction
+ RECON_INVOKE(&cpi->common.rtcd.recon, build_intra_predictors_mby)
(&x->e_mbd);
if (cpi->common.txfm_mode == ALLOW_8X8)
macro_block_yrd_8x8(x, &rate_y, &distortion,
@@ -3116,6 +3211,7 @@
distortion_uv = uv_intra_distortion;
}
break;
+#endif
case NEWMV: {int thissme;
@@ -3269,7 +3365,6 @@
}
case ZEROMV:
-
// Trap vectors that reach beyond the UMV borders
// Note that ALL New MV, Nearest MV Near MV and Zero MV code drops through to this point
// because of the lack of break statements in the previous two cases.
@@ -3348,12 +3443,23 @@
rate2 += vp8_cost_mv_ref(cpi, this_mode, mdcounts);
// Y cost and distortion
- if (cpi->common.txfm_mode == ALLOW_8X8)
- macro_block_yrd_8x8(x, &rate_y, &distortion,
- IF_RTCD(&cpi->rtcd));
- else
- macro_block_yrd(x, &rate_y, &distortion,
- IF_RTCD(&cpi->rtcd));
+#if CONFIG_TX16X16
+ if (this_mode == ZEROMV ||
+ this_mode == NEARESTMV ||
+ this_mode == NEARMV ||
+ this_mode == NEWMV)
+ macro_block_yrd_16x16(x, &rate_y, &distortion, IF_RTCD(&cpi->rtcd));
+ else {+#endif
+ if (cpi->common.txfm_mode == ALLOW_8X8)
+ macro_block_yrd_8x8(x, &rate_y, &distortion,
+ IF_RTCD(&cpi->rtcd));
+ else
+ macro_block_yrd(x, &rate_y, &distortion,
+ IF_RTCD(&cpi->rtcd));
+#if CONFIG_TX16X16
+ }
+#endif
rate2 += rate_y;
distortion2 += distortion;
@@ -3361,7 +3467,14 @@
// UV cost and distortion
vp8_build_inter16x16_predictors_mbuv(&x->e_mbd);
- if (cpi->common.txfm_mode == ALLOW_8X8)
+ if (cpi->common.txfm_mode == ALLOW_8X8
+#if CONFIG_TX16X16
+ || this_mode == ZEROMV ||
+ this_mode == NEARESTMV ||
+ this_mode == NEARMV ||
+ this_mode == NEWMV
+#endif
+ )
rd_inter16x16_uv_8x8(cpi, x, &rate_uv,
&distortion_uv,
cpi->common.full_pixel);
@@ -3487,9 +3600,21 @@
&x->e_mbd.predictor[320], 16, 8);
/* Y cost and distortion */
- if (cpi->common.txfm_mode == ALLOW_8X8)
+ if (cpi->common.txfm_mode == ALLOW_8X8
+#if CONFIG_TX16X16
+ || this_mode == ZEROMV ||
+ this_mode == NEARESTMV ||
+ this_mode == NEARMV ||
+ this_mode == NEWMV
+#endif
+ )
+#if CONFIG_TX16X16
+ macro_block_yrd_16x16(x, &rate_y, &distortion,
+ IF_RTCD(&cpi->rtcd));
+#else
macro_block_yrd_8x8(x, &rate_y, &distortion,
IF_RTCD(&cpi->rtcd));
+#endif
else
macro_block_yrd(x, &rate_y, &distortion,
IF_RTCD(&cpi->rtcd));
@@ -3498,7 +3623,14 @@
distortion2 += distortion;
/* UV cost and distortion */
- if (cpi->common.txfm_mode == ALLOW_8X8)
+ if (cpi->common.txfm_mode == ALLOW_8X8
+#if CONFIG_TX16X16
+ || this_mode == ZEROMV ||
+ this_mode == NEARESTMV ||
+ this_mode == NEARMV ||
+ this_mode == NEWMV
+#endif
+ )
rd_inter16x16_uv_8x8(cpi, x, &rate_uv,
&distortion_uv,
cpi->common.full_pixel);
@@ -3541,6 +3673,15 @@
&& this_mode != B_PRED
&& this_mode != I8X8_PRED);
+#if CONFIGURE_TX16X16
+ if (this_mode <= TM_PRED ||
+ this_mode == NEWMV ||
+ this_mode == ZEROMV ||
+ this_mode == NEARESTMV ||
+ this_mode == NEARMV)
+ mb_skippable = mb_is_skippable_16x16(&x->e_mbd);
+ else
+#endif
if ((cpi->common.txfm_mode == ALLOW_8X8) && has_y2) {if (x->e_mbd.mode_info_context->mbmi.ref_frame != INTRA_FRAME)
mb_skippable = mb_is_skippable_8x8(&x->e_mbd);
@@ -4002,10 +4143,25 @@
}
/* test code: set transform size based on mode selection */
+#if CONFIG_TX16X16
+ if (xd->mode_info_context->mbmi.mode <= TM_PRED ||
+ xd->mode_info_context->mbmi.mode == NEWMV ||
+ xd->mode_info_context->mbmi.mode == ZEROMV ||
+ xd->mode_info_context->mbmi.mode == NEARMV ||
+ xd->mode_info_context->mbmi.mode == NEARESTMV) {+ xd->mode_info_context->mbmi.txfm_size = TX_16X16;
+ cpi->t16x16_count++;
+ }
+ else if (cpi->common.txfm_mode == ALLOW_8X8
+ && xd->mode_info_context->mbmi.mode != I8X8_PRED
+ && xd->mode_info_context->mbmi.mode != B_PRED
+ && xd->mode_info_context->mbmi.mode != SPLITMV) {+#else
if (cpi->common.txfm_mode == ALLOW_8X8
&& xd->mode_info_context->mbmi.mode != I8X8_PRED
&& xd->mode_info_context->mbmi.mode != B_PRED
&& xd->mode_info_context->mbmi.mode != SPLITMV) {+#endif
xd->mode_info_context->mbmi.txfm_size = TX_8X8;
cpi->t8x8_count++;
} else {--- a/vp8/encoder/tokenize.c
+++ b/vp8/encoder/tokenize.c
@@ -26,17 +26,23 @@
#ifdef ENTROPY_STATS
INT64 context_counters[BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
INT64 context_counters_8x8[BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
-extern unsigned int tree_update_hist [BLOCK_TYPES]
-[COEF_BANDS]
-[PREV_COEF_CONTEXTS]
-[ENTROPY_NODES][2];
-extern unsigned int tree_update_hist_8x8 [BLOCK_TYPES_8X8]
-[COEF_BANDS]
-[PREV_COEF_CONTEXTS]
-[ENTROPY_NODES] [2];
+#if CONFIG_TX16X16
+INT64 context_counters_16x16[BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
#endif
+extern unsigned int tree_update_hist[BLOCK_TYPES][COEF_BANDS]
+ [PREV_COEF_CONTEXTS][ENTROPY_NODES][2];
+extern unsigned int tree_update_hist_8x8[BLOCK_TYPES_8X8][COEF_BANDS]
+ [PREV_COEF_CONTEXTS][ENTROPY_NODES] [2];
+#if CONFIG_TX16X16
+extern unsigned int tree_update_hist_16x16[BLOCK_TYPES_16X16][COEF_BANDS]
+ [PREV_COEF_CONTEXTS][ENTROPY_NODES] [2];
+#endif
+#endif
void vp8_stuff_mb(VP8_COMP *cpi, MACROBLOCKD *x, TOKENEXTRA **t);
void vp8_stuff_mb_8x8(VP8_COMP *cpi, MACROBLOCKD *x, TOKENEXTRA **t);
+#if CONFIG_TX16X16
+void vp8_stuff_mb_16x16(VP8_COMP *cpi, MACROBLOCKD *x, TOKENEXTRA **t);
+#endif
void vp8_fix_contexts(MACROBLOCKD *x);
static TOKENVALUE dct_value_tokens[DCT_MAX_VALUE * 2];
@@ -103,6 +109,54 @@
vp8_dct_value_cost_ptr = dct_value_cost + DCT_MAX_VALUE;
}
+#if CONFIG_TX16X16
+static void tokenize1st_order_b_16x16(MACROBLOCKD *xd, const BLOCKD *const b, TOKENEXTRA **tp,
+ const int type, const FRAME_TYPE frametype, ENTROPY_CONTEXT *a,
+ ENTROPY_CONTEXT *l, VP8_COMP *cpi) {+ int pt; /* near block/prev token context index */
+ int c = 0; /* start at DC unless type 0 */
+ const int eob = b->eob; /* one beyond last nonzero coeff */
+ TOKENEXTRA *t = *tp; /* store tokens starting here */
+ int x;
+ const short *qcoeff_ptr = b->qcoeff;
+
+ int seg_eob = 256;
+ int segment_id = xd->mode_info_context->mbmi.segment_id;
+
+ if (segfeature_active(xd, segment_id, SEG_LVL_EOB))
+ seg_eob = get_segdata(xd, segment_id, SEG_LVL_EOB);
+
+ VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l);
+
+ do {+ const int band = vp8_coef_bands_16x16[c];
+ int v;
+
+ x = DCT_EOB_TOKEN;
+ if (c < eob) {+ int rc = vp8_default_zig_zag1d_16x16[c];
+ v = qcoeff_ptr[rc];
+
+ assert(-DCT_MAX_VALUE <= v && v < (DCT_MAX_VALUE));
+
+ t->Extra = vp8_dct_value_tokens_ptr[v].Extra;
+ x = vp8_dct_value_tokens_ptr[v].Token;
+ }
+
+ t->Token = x;
+ t->context_tree = cpi->common.fc.coef_probs_16x16[type][band][pt];
+
+ t->skip_eob_node = pt == 0 && ((band > 0 && type > 0) || (band > 1 && type == 0));
+
+ ++cpi->coef_counts_16x16[type][band][pt][x];
+ } while (pt = vp8_prev_token_class[x], ++t, c < eob && ++c < seg_eob);
+
+ *tp = t;
+ pt = (c != !type); /* 0 <-> all coeff data is zero */
+ *a = *l = pt;
+}
+#endif
+
static void tokenize2nd_order_b_8x8
(
MACROBLOCKD *xd,
@@ -170,12 +224,8 @@
}
-static void tokenize2nd_order_b
-(
- MACROBLOCKD *xd,
- TOKENEXTRA **tp,
- VP8_COMP *cpi
-) {+static void tokenize2nd_order_b(MACROBLOCKD *xd, TOKENEXTRA **tp,
+ VP8_COMP *cpi) {int pt; /* near block/prev token context index */
int c; /* start at DC */
TOKENEXTRA *t = *tp;/* store tokens starting here */
@@ -188,9 +238,8 @@
int seg_eob = 16;
int segment_id = xd->mode_info_context->mbmi.segment_id;
- if (segfeature_active(xd, segment_id, SEG_LVL_EOB)) {+ if (segfeature_active(xd, segment_id, SEG_LVL_EOB))
seg_eob = get_segdata(xd, segment_id, SEG_LVL_EOB);
- }
b = xd->block + 24;
qcoeff_ptr = b->qcoeff;
@@ -542,14 +591,10 @@
unsigned int block;
const BLOCKD *b;
int pt; /* near block/prev token context index */
- int c;
- int token;
+ int band, rc, v, c, token;
TOKENEXTRA *t = *tp;/* store tokens starting here */
const short *qcoeff_ptr;
- ENTROPY_CONTEXT *a;
- ENTROPY_CONTEXT *l;
- int band, rc, v;
- int tmp1, tmp2;
+ ENTROPY_CONTEXT *a, *l;
int seg_eob = 16;
int segment_id = xd->mode_info_context->mbmi.segment_id;
@@ -561,11 +606,9 @@
b = xd->block;
/* Luma */
for (block = 0; block < 16; block++, b++) {- tmp1 = vp8_block2above[block];
- tmp2 = vp8_block2left[block];
qcoeff_ptr = b->qcoeff;
- a = (ENTROPY_CONTEXT *)xd->above_context + tmp1;
- l = (ENTROPY_CONTEXT *)xd->left_context + tmp2;
+ a = (ENTROPY_CONTEXT *)xd->above_context + vp8_block2above[block];
+ l = (ENTROPY_CONTEXT *)xd->left_context + vp8_block2left[block];
VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l);
c = type ? 0 : 1;
@@ -609,11 +652,9 @@
}
/* Chroma */
for (block = 16; block < 24; block++, b++) {- tmp1 = vp8_block2above[block];
- tmp2 = vp8_block2left[block];
qcoeff_ptr = b->qcoeff;
- a = (ENTROPY_CONTEXT *)xd->above_context + tmp1;
- l = (ENTROPY_CONTEXT *)xd->left_context + tmp2;
+ a = (ENTROPY_CONTEXT *)xd->above_context + vp8_block2above[block];
+ l = (ENTROPY_CONTEXT *)xd->left_context + vp8_block2left[block];
VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l);
@@ -701,7 +742,21 @@
return (mby_is_skippable_8x8(x) & mbuv_is_skippable_8x8(x));
}
+#if CONFIG_TX16X16
+int mby_is_skippable_16x16(MACROBLOCKD *x) {+ int skip = 1;
+ //skip &= (x->block[0].eob < 2); // I think this should be commented? No second order == DC must be coded
+ //skip &= (x->block[0].eob < 1);
+ //skip &= (!x->block[24].eob);
+ skip &= !x->block[0].eob;
+ return skip;
+}
+int mb_is_skippable_16x16(MACROBLOCKD *x) {+ return (mby_is_skippable_16x16(x) & mbuv_is_skippable_8x8(x));
+}
+#endif
+
void vp8_tokenize_mb(VP8_COMP *cpi, MACROBLOCKD *x, TOKENEXTRA **t) {int plane_type;
int has_y2_block;
@@ -730,16 +785,32 @@
has_y2_block = (x->mode_info_context->mbmi.mode != B_PRED
&& x->mode_info_context->mbmi.mode != I8X8_PRED
&& x->mode_info_context->mbmi.mode != SPLITMV);
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16) has_y2_block = 0; // Because of inter frames
+#endif
- x->mode_info_context->mbmi.mb_skip_coeff =
- ((tx_type == TX_8X8) ?
- mb_is_skippable_8x8(x) :
- mb_is_skippable(x, has_y2_block));
+ switch (tx_type) {+#if CONFIG_TX16X16
+ case TX_16X16:
+ x->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_16x16(x);
+ break;
+#endif
+ case TX_8X8:
+ x->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_8x8(x);
+ break;
+ default:
+ x->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable(x, has_y2_block);
+ break;
+ }
if (x->mode_info_context->mbmi.mb_skip_coeff) {cpi->skip_true_count[mb_skip_context] += skip_inc;
-
if (!cpi->common.mb_no_coeff_skip) {+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16)
+ vp8_stuff_mb_16x16(cpi, x, t);
+ else
+#endif
if (tx_type == TX_8X8)
vp8_stuff_mb_8x8(cpi, x, t);
else
@@ -766,9 +837,28 @@
tokenize2nd_order_b(x, t, cpi);
plane_type = 0;
-
}
+#if CONFIG_TX16X16
+ if (tx_type == TX_16X16) {+ ENTROPY_CONTEXT * A = (ENTROPY_CONTEXT *)x->above_context;
+ ENTROPY_CONTEXT * L = (ENTROPY_CONTEXT *)x->left_context;
+ tokenize1st_order_b_16x16(x, x->block, t, 3, x->frame_type, A, L, cpi);
+ for (b = 1; b < 16; b++) {+ *(A + vp8_block2above[b]) = *(A);
+ *(L + vp8_block2left[b] ) = *(L);
+ }
+ for (b = 16; b < 24; b += 4) {+ tokenize1st_order_b_8x8(x, x->block + b, t, 2, x->frame_type,
+ A + vp8_block2above_8x8[b], L + vp8_block2left_8x8[b], cpi);
+ *(A + vp8_block2above_8x8[b]+1) = *(A + vp8_block2above_8x8[b]);
+ *(L + vp8_block2left_8x8[b]+1 ) = *(L + vp8_block2left_8x8[b]);
+ }
+ vpx_memset(&A[8], 0, sizeof(A[8]));
+ vpx_memset(&L[8], 0, sizeof(L[8]));
+ }
+ else
+#endif
if (tx_type == TX_8X8) {ENTROPY_CONTEXT *A = (ENTROPY_CONTEXT *)x->above_context;
ENTROPY_CONTEXT *L = (ENTROPY_CONTEXT *)x->left_context;
@@ -827,15 +917,20 @@
#ifdef ENTROPY_STATS
-
void init_context_counters(void) { FILE *f = fopen("context.bin", "rb"); if (!f) {vpx_memset(context_counters, 0, sizeof(context_counters));
vpx_memset(context_counters_8x8, 0, sizeof(context_counters_8x8));
+#if CONFIG_TX16X16
+ vpx_memset(context_counters_16x16, 0, sizeof(context_counters_16x16));
+#endif
} else {fread(context_counters, sizeof(context_counters), 1, f);
fread(context_counters_8x8, sizeof(context_counters_8x8), 1, f);
+#if CONFIG_TX16X16
+ fread(context_counters_16x16, sizeof(context_counters_16x16), 1, f);
+#endif
fclose(f);
}
@@ -843,15 +938,20 @@
if (!f) {vpx_memset(tree_update_hist, 0, sizeof(tree_update_hist));
vpx_memset(tree_update_hist_8x8, 0, sizeof(tree_update_hist_8x8));
+#if CONFIG_TX16X16
+ vpx_memset(tree_update_hist_16x16, 0, sizeof(tree_update_hist_16x16));
+#endif
} else {fread(tree_update_hist, sizeof(tree_update_hist), 1, f);
fread(tree_update_hist_8x8, sizeof(tree_update_hist_8x8), 1, f);
+#if CONFIG_TX16X16
+ fread(tree_update_hist_16x16, sizeof(tree_update_hist_16x16), 1, f);
+#endif
fclose(f);
}
}
void print_context_counters() {-
int type, band, pt, t;
FILE *f = fopen("context.c", "w");@@ -892,7 +992,6 @@
fprintf(f, "static const unsigned int\nvp8_default_coef_counts_8x8"
"[BLOCK_TYPES_8X8] [COEF_BANDS]"
"[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS] = {");-
type = 0;
do { fprintf(f, "%s\n { /* block Type %d */", Comma(type), type);@@ -921,26 +1020,54 @@
fprintf(f, "\n }");
} while (++type < BLOCK_TYPES_8X8);
+ fprintf(f, "\n};\n");
+#if CONFIG_TX16X16
+ fprintf(f, "static const unsigned int\nvp8_default_coef_counts_16x16"
+ "[BLOCK_TYPES_16X16] [COEF_BANDS]"
+ "[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS] = {");+ type = 0;
+ do {+ fprintf(f, "%s\n { /* block Type %d */", Comma(type), type);+ band = 0;
+ do {+ fprintf(f, "%s\n { /* Coeff Band %d */", Comma(band), band);+ pt = 0;
+ do {+ fprintf(f, "%s\n {", Comma(pt));+ t = 0;
+ do {+ const INT64 x = context_counters_16x16 [type] [band] [pt] [t];
+ const int y = (int) x;
+
+ assert(x == (INT64) y); /* no overflow handling yet */
+ fprintf(f, "%s %d", Comma(t), y);
+
+ } while (++t < MAX_ENTROPY_TOKENS);
+
+ fprintf(f, "}");
+ } while (++pt < PREV_COEF_CONTEXTS);
+
+ fprintf(f, "\n }");
+
+ } while (++band < COEF_BANDS);
+
+ fprintf(f, "\n }");
+ } while (++type < BLOCK_TYPES_16X16);
fprintf(f, "\n};\n");
+#endif
fprintf(f, "static const vp8_prob\n"
"vp8_default_coef_probs[BLOCK_TYPES] [COEF_BANDS] \n"
"[PREV_COEF_CONTEXTS] [ENTROPY_NODES] = {");type = 0;
-
do { fprintf(f, "%s\n { /* block Type %d */", Comma(type), type);-
band = 0;
-
do { fprintf(f, "%s\n { /* Coeff Band %d */", Comma(band), band);-
pt = 0;
-
do {-
unsigned int branch_ct [ENTROPY_NODES] [2];
unsigned int coef_counts[MAX_ENTROPY_TOKENS];
vp8_prob coef_probs[ENTROPY_NODES];
@@ -952,7 +1079,6 @@
fprintf(f, "%s\n {", Comma(pt));t = 0;
-
do {fprintf(f, "%s %d", Comma(t), coef_probs[t]);
@@ -960,11 +1086,8 @@
fprintf(f, "}");
} while (++pt < PREV_COEF_CONTEXTS);
-
fprintf(f, "\n }");
-
} while (++band < COEF_BANDS);
-
fprintf(f, "\n }");
} while (++type < BLOCK_TYPES);
fprintf(f, "\n};\n");
@@ -973,19 +1096,13 @@
"vp8_default_coef_probs_8x8[BLOCK_TYPES_8X8] [COEF_BANDS]\n"
"[PREV_COEF_CONTEXTS] [ENTROPY_NODES] = {");type = 0;
-
do { fprintf(f, "%s\n { /* block Type %d */", Comma(type), type);-
band = 0;
-
do { fprintf(f, "%s\n { /* Coeff Band %d */", Comma(band), band);-
pt = 0;
-
do {-
unsigned int branch_ct [ENTROPY_NODES] [2];
unsigned int coef_counts[MAX_ENTROPY_TOKENS];
vp8_prob coef_probs[ENTROPY_NODES];
@@ -994,34 +1111,65 @@
vp8_tree_probs_from_distribution(
MAX_ENTROPY_TOKENS, vp8_coef_encodings, vp8_coef_tree,
coef_probs, branch_ct, coef_counts, 256, 1);
-
fprintf(f, "%s\n {", Comma(pt));- t = 0;
+ t = 0;
do {fprintf(f, "%s %d", Comma(t), coef_probs[t]);
-
} while (++t < ENTROPY_NODES);
-
fprintf(f, "}");
} while (++pt < PREV_COEF_CONTEXTS);
-
fprintf(f, "\n }");
-
} while (++band < COEF_BANDS);
-
fprintf(f, "\n }");
} while (++type < BLOCK_TYPES_8X8);
fprintf(f, "\n};\n");
+#if CONFIG_TX16X16
+ fprintf(f, "static const vp8_prob\n"
+ "vp8_default_coef_probs_16x16[BLOCK_TYPES_16X16] [COEF_BANDS]\n"
+ "[PREV_COEF_CONTEXTS] [ENTROPY_NODES] = {");+ type = 0;
+ do {+ fprintf(f, "%s\n { /* block Type %d */", Comma(type), type);+ band = 0;
+ do {+ fprintf(f, "%s\n { /* Coeff Band %d */", Comma(band), band);+ pt = 0;
+ do {+ unsigned int branch_ct [ENTROPY_NODES] [2];
+ unsigned int coef_counts[MAX_ENTROPY_TOKENS];
+ vp8_prob coef_probs[ENTROPY_NODES];
+ for (t = 0; t < MAX_ENTROPY_TOKENS; ++t)
+ coef_counts[t] = context_counters_16x16[type] [band] [pt] [t];
+ vp8_tree_probs_from_distribution(
+ MAX_ENTROPY_TOKENS, vp8_coef_encodings, vp8_coef_tree,
+ coef_probs, branch_ct, coef_counts, 256, 1);
+ fprintf(f, "%s\n {", Comma(pt));+
+ t = 0;
+ do {+ fprintf(f, "%s %d", Comma(t), coef_probs[t]);
+ } while (++t < ENTROPY_NODES);
+ fprintf(f, "}");
+ } while (++pt < PREV_COEF_CONTEXTS);
+ fprintf(f, "\n }");
+ } while (++band < COEF_BANDS);
+ fprintf(f, "\n }");
+ } while (++type < BLOCK_TYPES_16X16);
+ fprintf(f, "\n};\n");
+#endif
+
fclose(f);
f = fopen("context.bin", "wb");fwrite(context_counters, sizeof(context_counters), 1, f);
fwrite(context_counters_8x8, sizeof(context_counters_8x8), 1, f);
+#if CONFIG_TX16X16
+ fwrite(context_counters_16x16, sizeof(context_counters_16x16), 1, f);
+#endif
fclose(f);
}
-
#endif
@@ -1151,6 +1299,50 @@
}
+#if CONFIG_TX16X16
+static __inline
+void stuff1st_order_b_16x16(const BLOCKD *const b, TOKENEXTRA **tp, const FRAME_TYPE frametype,
+ ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l, VP8_COMP *cpi)
+{+ int pt; /* near block/prev token context index */
+ TOKENEXTRA *t = *tp; /* store tokens starting here */
+ VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l);
+ (void) frametype;
+ (void) b;
+
+ t->Token = DCT_EOB_TOKEN;
+ t->context_tree = cpi->common.fc.coef_probs_16x16[3][1][pt];
+ t->skip_eob_node = 0;
+ ++t;
+ *tp = t;
+ ++cpi->coef_counts_16x16[3][1][pt][DCT_EOB_TOKEN];
+ pt = 0; /* 0 <-> all coeff data is zero */
+ *a = *l = pt;
+}
+
+void vp8_stuff_mb_16x16(VP8_COMP *cpi, MACROBLOCKD *x, TOKENEXTRA **t) {+ ENTROPY_CONTEXT * A = (ENTROPY_CONTEXT *)x->above_context;
+ ENTROPY_CONTEXT * L = (ENTROPY_CONTEXT *)x->left_context;
+ int b, i;
+
+ stuff1st_order_b_16x16(x->block, t, x->frame_type, A, L, cpi);
+ for (i = 1; i < 16; i++) {+ *(A + vp8_block2above[i]) = *(A);
+ *(L + vp8_block2left[i]) = *(L);
+ }
+ for (b = 16; b < 24; b += 4) {+ stuff1st_order_buv_8x8(x->block + b, t, 2, x->frame_type,
+ A + vp8_block2above[b],
+ L + vp8_block2left[b],
+ cpi);
+ *(A + vp8_block2above_8x8[b]+1) = *(A + vp8_block2above_8x8[b]);
+ *(L + vp8_block2left_8x8[b]+1 ) = *(L + vp8_block2left_8x8[b]);
+ }
+ vpx_memset(&A[8], 0, sizeof(A[8]));
+ vpx_memset(&L[8], 0, sizeof(L[8]));
+}
+#endif
+
static __inline void stuff2nd_order_b
(
TOKENEXTRA **tp,
@@ -1215,7 +1407,6 @@
++cpi->coef_counts[2] [0] [pt] [DCT_EOB_TOKEN];
pt = 0; /* 0 <-> all coeff data is zero */
*a = *l = pt;
-
}
void vp8_stuff_mb(VP8_COMP *cpi, MACROBLOCKD *x, TOKENEXTRA **t) {@@ -1241,9 +1432,13 @@
}
void vp8_fix_contexts(MACROBLOCKD *x) {/* Clear entropy contexts for Y2 blocks */
- if (x->mode_info_context->mbmi.mode != B_PRED
+ if ((x->mode_info_context->mbmi.mode != B_PRED
&& x->mode_info_context->mbmi.mode != I8X8_PRED
- && x->mode_info_context->mbmi.mode != SPLITMV) {+ && x->mode_info_context->mbmi.mode != SPLITMV)
+#if CONFIG_TX16X16
+ || x->mode_info_context->mbmi.txfm_size == TX_16X16
+#endif
+ ) {vpx_memset(x->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
vpx_memset(x->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
} else {--- a/vp8/encoder/tokenize.h
+++ b/vp8/encoder/tokenize.h
@@ -44,8 +44,11 @@
extern INT64 context_counters[BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
extern INT64 context_counters_8x8[BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
-
+#if CONFIG_TX16X16
+extern INT64 context_counters_16x16[BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
#endif
+#endif
+
extern const int *vp8_dct_value_cost_ptr;
/* TODO: The Token field should be broken out into a separate char array to
* improve cache locality, since it's needed for costing when the rest of the
--
⑨