diff --git a/configure b/configure
index d1d25b3d5725a9a4d3775c8e2cb6fd74dddfbd32..c93ffd75fcc614ec2a64f77a42dafdf879b6a7bc 100755
--- a/configure
+++ b/configure
@@ -247,6 +247,8 @@ EXPERIMENT_LIST="
implicit_segmentation
newbintramodes
comp_interintra_pred
+ tx32x32
+ dwt32x32hybrid
"
CONFIG_LIST="
external_build
diff --git a/test/dct32x32_test.cc b/test/dct32x32_test.cc
new file mode 100644
index 0000000000000000000000000000000000000000..8f089c6839b02a36997c1cb3d69f6c5ea3f4288f
--- /dev/null
+++ b/test/dct32x32_test.cc
@@ -0,0 +1,189 @@
+/*
+ * 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 "vp9/common/vp9_entropy.h"
+#include "./vp9_rtcd.h"
+ void vp9_short_fdct32x32_c(int16_t *input, int16_t *out, int pitch);
+ void vp9_short_idct32x32_c(short *input, short *output, int pitch);
+}
+
+#include "test/acm_random.h"
+#include "vpx/vpx_integer.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {
+
+#if !CONFIG_DWT32X32HYBRID
+static const double kPi = 3.141592653589793238462643383279502884;
+static void reference2_32x32_idct_2d(double *input, double *output) {
+ double x;
+ for (int l = 0; l < 32; ++l) {
+ for (int k = 0; k < 32; ++k) {
+ double s = 0;
+ for (int i = 0; i < 32; ++i) {
+ for (int j = 0; j < 32; ++j) {
+ x = cos(kPi * j * (l + 0.5) / 32.0) *
+ cos(kPi * i * (k + 0.5) / 32.0) * input[i * 32 + j] / 1024;
+ if (i != 0)
+ x *= sqrt(2.0);
+ if (j != 0)
+ x *= sqrt(2.0);
+ s += x;
+ }
+ }
+ output[k * 32 + l] = s / 4;
+ }
+ }
+}
+
+static void reference_32x32_dct_1d(double in[32], double out[32], int stride) {
+ const double kInvSqrt2 = 0.707106781186547524400844362104;
+ for (int k = 0; k < 32; k++) {
+ out[k] = 0.0;
+ for (int n = 0; n < 32; n++)
+ out[k] += in[n] * cos(kPi * (2 * n + 1) * k / 64.0);
+ if (k == 0)
+ out[k] = out[k] * kInvSqrt2;
+ }
+}
+
+static void reference_32x32_dct_2d(int16_t input[32*32], double output[32*32]) {
+ // First transform columns
+ for (int i = 0; i < 32; ++i) {
+ double temp_in[32], temp_out[32];
+ for (int j = 0; j < 32; ++j)
+ temp_in[j] = input[j*32 + i];
+ reference_32x32_dct_1d(temp_in, temp_out, 1);
+ for (int j = 0; j < 32; ++j)
+ output[j * 32 + i] = temp_out[j];
+ }
+ // Then transform rows
+ for (int i = 0; i < 32; ++i) {
+ double temp_in[32], temp_out[32];
+ for (int j = 0; j < 32; ++j)
+ temp_in[j] = output[j + i*32];
+ reference_32x32_dct_1d(temp_in, temp_out, 1);
+ // Scale by some magic number
+ for (int j = 0; j < 32; ++j)
+ output[j + i * 32] = temp_out[j] / 4;
+ }
+}
+
+
+TEST(VP9Idct32x32Test, AccuracyCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = 1000;
+ for (int i = 0; i < count_test_block; ++i) {
+ int16_t in[1024], coeff[1024];
+ int16_t out_c[1024];
+ double out_r[1024];
+
+ // Initialize a test block with input range [-255, 255].
+ for (int j = 0; j < 1024; ++j)
+ in[j] = rnd.Rand8() - rnd.Rand8();
+
+ reference_32x32_dct_2d(in, out_r);
+ for (int j = 0; j < 1024; j++)
+ coeff[j] = round(out_r[j]);
+ vp9_short_idct32x32_c(coeff, out_c, 64);
+ for (int j = 0; j < 1024; ++j) {
+ const int diff = out_c[j] - in[j];
+ const int error = diff * diff;
+ EXPECT_GE(1, error)
+ << "Error: 3x32 IDCT has error " << error
+ << " at index " << j;
+ }
+
+ vp9_short_fdct32x32_c(in, out_c, 64);
+ for (int j = 0; j < 1024; ++j) {
+ const double diff = coeff[j] - out_c[j];
+ const double error = diff * diff;
+ EXPECT_GE(1.0, error)
+ << "Error: 32x32 FDCT has error " << error
+ << " at index " << j;
+ }
+ }
+}
+#else // CONFIG_DWT32X32HYBRID
+ // TODO(rbultje/debargha): add DWT-specific tests
+#endif // CONFIG_DWT32X32HYBRID
+TEST(VP9Fdct32x32Test, AccuracyCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ unsigned int max_error = 0;
+ int64_t total_error = 0;
+ const int count_test_block = 1000;
+ for (int i = 0; i < count_test_block; ++i) {
+ int16_t test_input_block[1024];
+ int16_t test_temp_block[1024];
+ int16_t test_output_block[1024];
+
+ // Initialize a test block with input range [-255, 255].
+ for (int j = 0; j < 1024; ++j)
+ test_input_block[j] = rnd.Rand8() - rnd.Rand8();
+
+ const int pitch = 64;
+ vp9_short_fdct32x32_c(test_input_block, test_temp_block, pitch);
+ vp9_short_idct32x32_c(test_temp_block, test_output_block, pitch);
+
+ for (int j = 0; j < 1024; ++j) {
+ const unsigned diff = test_input_block[j] - test_output_block[j];
+ const unsigned error = diff * diff;
+ if (max_error < error)
+ max_error = error;
+ total_error += error;
+ }
+ }
+
+ EXPECT_GE(1u, max_error)
+ << "Error: 32x32 FDCT/IDCT has an individual roundtrip error > 1";
+
+ EXPECT_GE(count_test_block/10, total_error)
+ << "Error: 32x32 FDCT/IDCT has average roundtrip error > 1/10 per block";
+}
+
+TEST(VP9Fdct32x32Test, CoeffSizeCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = 1000;
+ for (int i = 0; i < count_test_block; ++i) {
+ int16_t input_block[1024], input_extreme_block[1024];
+ int16_t output_block[1024], output_extreme_block[1024];
+
+ // Initialize a test block with input range [-255, 255].
+ for (int j = 0; j < 1024; ++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 < 1024; ++j)
+ input_extreme_block[j] = 255;
+
+ const int pitch = 32;
+ vp9_short_fdct32x32_c(input_block, output_block, pitch);
+ vp9_short_fdct32x32_c(input_extreme_block, output_extreme_block, pitch);
+
+ // The minimum quant value is 4.
+ for (int j = 0; j < 1024; ++j) {
+ EXPECT_GE(4*DCT_MAX_VALUE, abs(output_block[j]))
+ << "Error: 32x32 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
+ EXPECT_GE(4*DCT_MAX_VALUE, abs(output_extreme_block[j]))
+ << "Error: 32x32 FDCT extreme has coefficient larger than "
+ "4*DCT_MAX_VALUE";
+ }
+ }
+}
+} // namespace
diff --git a/test/test.mk b/test/test.mk
index 4fb464e643e93fabb2f3b7b7e9c8d04561907a44..919cf04387bbea5a08115f673bf3302c8b53b808 100644
--- a/test/test.mk
+++ b/test/test.mk
@@ -64,6 +64,9 @@ endif
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct4x4_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct8x8_test.cc
#LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct16x16_test.cc
+ifeq ($(CONFIG_VP9_ENCODER)$(CONFIG_TX32X32),yesyes)
+LIBVPX_TEST_SRCS-yes += dct32x32_test.cc
+endif
LIBVPX_TEST_SRCS-yes += idct8x8_test.cc
LIBVPX_TEST_SRCS-yes += variance_test.cc
endif # VP9
diff --git a/vp9/common/vp9_blockd.h b/vp9/common/vp9_blockd.h
index 27ef22fffd3fe956d37919da701c210ea2c7d20c..11efd44759a25b4d497cdc54bdb3ae6a902471ed 100644
--- a/vp9/common/vp9_blockd.h
+++ b/vp9/common/vp9_blockd.h
@@ -129,7 +129,13 @@ typedef enum {
TX_4X4, // 4x4 dct transform
TX_8X8, // 8x8 dct transform
TX_16X16, // 16x16 dct transform
- TX_SIZE_MAX // Number of different transforms available
+ TX_SIZE_MAX_MB, // Number of transforms available to MBs
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ TX_32X32 = TX_SIZE_MAX_MB, // 32x32 dct transform
+ TX_SIZE_MAX_SB, // Number of transforms available to SBs
+#else
+ TX_SIZE_MAX_SB = TX_SIZE_MAX_MB,
+#endif
} TX_SIZE;
typedef enum {
@@ -302,6 +308,15 @@ typedef struct blockd {
union b_mode_info bmi;
} BLOCKD;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+typedef struct superblockd {
+ /* 32x32 Y and 16x16 U/V. No 2nd order transform yet. */
+ DECLARE_ALIGNED(16, short, diff[32*32+16*16*2]);
+ DECLARE_ALIGNED(16, short, qcoeff[32*32+16*16*2]);
+ DECLARE_ALIGNED(16, short, dqcoeff[32*32+16*16*2]);
+} SUPERBLOCKD;
+#endif
+
typedef struct macroblockd {
DECLARE_ALIGNED(16, short, diff[400]); /* from idct diff */
DECLARE_ALIGNED(16, unsigned char, predictor[384]);
@@ -309,6 +324,10 @@ typedef struct macroblockd {
DECLARE_ALIGNED(16, short, dqcoeff[400]);
DECLARE_ALIGNED(16, unsigned short, eobs[25]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ SUPERBLOCKD sb_coeff_data;
+#endif
+
/* 16 Y blocks, 4 U, 4 V, 1 DC 2nd order block, each with 16 entries. */
BLOCKD block[25];
int fullpixel_mask;
diff --git a/vp9/common/vp9_default_coef_probs.h b/vp9/common/vp9_default_coef_probs.h
index 52fb02f36af34ef5fbcfb8c421fb992544264eb3..31103adb7fe7cd9906d7681eea55a65a6b674c3b 100644
--- a/vp9/common/vp9_default_coef_probs.h
+++ b/vp9/common/vp9_default_coef_probs.h
@@ -1375,3 +1375,5 @@ static const vp9_prob
}
}
};
+
+#define default_coef_probs_32x32 default_coef_probs_16x16
diff --git a/vp9/common/vp9_entropy.c b/vp9/common/vp9_entropy.c
index 4832b4e9c7b973a1b96d9df74c5bbf91e92090fe..321fa8c57504a6e335c6db9dca0ece3ee0880d61 100644
--- a/vp9/common/vp9_entropy.c
+++ b/vp9/common/vp9_entropy.c
@@ -132,6 +132,109 @@ DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_16x16[256]) = {
250, 251, 236, 221, 206, 191, 207, 222, 237, 252, 253, 238, 223, 239, 254, 255,
};
+DECLARE_ALIGNED(16, const int, vp9_coef_bands_32x32[1024]) = {
+ 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,
+
+ 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, 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, 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, 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, vp9_default_zig_zag1d_32x32[1024]) = {
+ 0, 1, 32, 64, 33, 2, 3, 34, 65, 96, 128, 97, 66, 35, 4, 5, 36, 67, 98, 129, 160, 192, 161, 130, 99, 68, 37, 6, 7, 38, 69, 100,
+ 131, 162, 193, 224, 256, 225, 194, 163, 132, 101, 70, 39, 8, 9, 40, 71, 102, 133, 164, 195, 226, 257, 288, 320, 289, 258, 227, 196, 165, 134, 103, 72,
+ 41, 10, 11, 42, 73, 104, 135, 166, 197, 228, 259, 290, 321, 352, 384, 353, 322, 291, 260, 229, 198, 167, 136, 105, 74, 43, 12, 13, 44, 75, 106, 137,
+ 168, 199, 230, 261, 292, 323, 354, 385, 416, 448, 417, 386, 355, 324, 293, 262, 231, 200, 169, 138, 107, 76, 45, 14, 15, 46, 77, 108, 139, 170, 201, 232,
+ 263, 294, 325, 356, 387, 418, 449, 480, 512, 481, 450, 419, 388, 357, 326, 295, 264, 233, 202, 171, 140, 109, 78, 47, 16, 17, 48, 79, 110, 141, 172, 203,
+ 234, 265, 296, 327, 358, 389, 420, 451, 482, 513, 544, 576, 545, 514, 483, 452, 421, 390, 359, 328, 297, 266, 235, 204, 173, 142, 111, 80, 49, 18, 19, 50,
+ 81, 112, 143, 174, 205, 236, 267, 298, 329, 360, 391, 422, 453, 484, 515, 546, 577, 608, 640, 609, 578, 547, 516, 485, 454, 423, 392, 361, 330, 299, 268, 237,
+ 206, 175, 144, 113, 82, 51, 20, 21, 52, 83, 114, 145, 176, 207, 238, 269, 300, 331, 362, 393, 424, 455, 486, 517, 548, 579, 610, 641, 672, 704, 673, 642,
+ 611, 580, 549, 518, 487, 456, 425, 394, 363, 332, 301, 270, 239, 208, 177, 146, 115, 84, 53, 22, 23, 54, 85, 116, 147, 178, 209, 240, 271, 302, 333, 364,
+ 395, 426, 457, 488, 519, 550, 581, 612, 643, 674, 705, 736, 768, 737, 706, 675, 644, 613, 582, 551, 520, 489, 458, 427, 396, 365, 334, 303, 272, 241, 210, 179,
+ 148, 117, 86, 55, 24, 25, 56, 87, 118, 149, 180, 211, 242, 273, 304, 335, 366, 397, 428, 459, 490, 521, 552, 583, 614, 645, 676, 707, 738, 769, 800, 832,
+ 801, 770, 739, 708, 677, 646, 615, 584, 553, 522, 491, 460, 429, 398, 367, 336, 305, 274, 243, 212, 181, 150, 119, 88, 57, 26, 27, 58, 89, 120, 151, 182,
+ 213, 244, 275, 306, 337, 368, 399, 430, 461, 492, 523, 554, 585, 616, 647, 678, 709, 740, 771, 802, 833, 864, 896, 865, 834, 803, 772, 741, 710, 679, 648, 617,
+ 586, 555, 524, 493, 462, 431, 400, 369, 338, 307, 276, 245, 214, 183, 152, 121, 90, 59, 28, 29, 60, 91, 122, 153, 184, 215, 246, 277, 308, 339, 370, 401,
+ 432, 463, 494, 525, 556, 587, 618, 649, 680, 711, 742, 773, 804, 835, 866, 897, 928, 960, 929, 898, 867, 836, 805, 774, 743, 712, 681, 650, 619, 588, 557, 526,
+ 495, 464, 433, 402, 371, 340, 309, 278, 247, 216, 185, 154, 123, 92, 61, 30, 31, 62, 93, 124, 155, 186, 217, 248, 279, 310, 341, 372, 403, 434, 465, 496,
+ 527, 558, 589, 620, 651, 682, 713, 744, 775, 806, 837, 868, 899, 930, 961, 992, 993, 962, 931, 900, 869, 838, 807, 776, 745, 714, 683, 652, 621, 590, 559, 528,
+ 497, 466, 435, 404, 373, 342, 311, 280, 249, 218, 187, 156, 125, 94, 63, 95, 126, 157, 188, 219, 250, 281, 312, 343, 374, 405, 436, 467, 498, 529, 560, 591,
+ 622, 653, 684, 715, 746, 777, 808, 839, 870, 901, 932, 963, 994, 995, 964, 933, 902, 871, 840, 809, 778, 747, 716, 685, 654, 623, 592, 561, 530, 499, 468, 437,
+ 406, 375, 344, 313, 282, 251, 220, 189, 158, 127, 159, 190, 221, 252, 283, 314, 345, 376, 407, 438, 469, 500, 531, 562, 593, 624, 655, 686, 717, 748, 779, 810,
+ 841, 872, 903, 934, 965, 996, 997, 966, 935, 904, 873, 842, 811, 780, 749, 718, 687, 656, 625, 594, 563, 532, 501, 470, 439, 408, 377, 346, 315, 284, 253, 222,
+ 191, 223, 254, 285, 316, 347, 378, 409, 440, 471, 502, 533, 564, 595, 626, 657, 688, 719, 750, 781, 812, 843, 874, 905, 936, 967, 998, 999, 968, 937, 906, 875,
+ 844, 813, 782, 751, 720, 689, 658, 627, 596, 565, 534, 503, 472, 441, 410, 379, 348, 317, 286, 255, 287, 318, 349, 380, 411, 442, 473, 504, 535, 566, 597, 628,
+ 659, 690, 721, 752, 783, 814, 845, 876, 907, 938, 969, 1000, 1001, 970, 939, 908, 877, 846, 815, 784, 753, 722, 691, 660, 629, 598, 567, 536, 505, 474, 443, 412,
+ 381, 350, 319, 351, 382, 413, 444, 475, 506, 537, 568, 599, 630, 661, 692, 723, 754, 785, 816, 847, 878, 909, 940, 971, 1002, 1003, 972, 941, 910, 879, 848, 817,
+ 786, 755, 724, 693, 662, 631, 600, 569, 538, 507, 476, 445, 414, 383, 415, 446, 477, 508, 539, 570, 601, 632, 663, 694, 725, 756, 787, 818, 849, 880, 911, 942,
+ 973, 1004, 1005, 974, 943, 912, 881, 850, 819, 788, 757, 726, 695, 664, 633, 602, 571, 540, 509, 478, 447, 479, 510, 541, 572, 603, 634, 665, 696, 727, 758, 789,
+ 820, 851, 882, 913, 944, 975, 1006, 1007, 976, 945, 914, 883, 852, 821, 790, 759, 728, 697, 666, 635, 604, 573, 542, 511, 543, 574, 605, 636, 667, 698, 729, 760,
+ 791, 822, 853, 884, 915, 946, 977, 1008, 1009, 978, 947, 916, 885, 854, 823, 792, 761, 730, 699, 668, 637, 606, 575, 607, 638, 669, 700, 731, 762, 793, 824, 855,
+ 886, 917, 948, 979, 1010, 1011, 980, 949, 918, 887, 856, 825, 794, 763, 732, 701, 670, 639, 671, 702, 733, 764, 795, 826, 857, 888, 919, 950, 981, 1012, 1013, 982,
+ 951, 920, 889, 858, 827, 796, 765, 734, 703, 735, 766, 797, 828, 859, 890, 921, 952, 983, 1014, 1015, 984, 953, 922, 891, 860, 829, 798, 767, 799, 830, 861, 892,
+ 923, 954, 985, 1016, 1017, 986, 955, 924, 893, 862, 831, 863, 894, 925, 956, 987, 1018, 1019, 988, 957, 926, 895, 927, 958, 989, 1020, 1021, 990, 959, 991, 1022, 1023,
+};
/* Array indices are identical to previously-existing CONTEXT_NODE indices */
@@ -160,10 +263,11 @@ static const Prob Pcat2[] = { 165, 145};
static const Prob Pcat3[] = { 173, 148, 140};
static const Prob Pcat4[] = { 176, 155, 140, 135};
static const Prob Pcat5[] = { 180, 157, 141, 134, 130};
-static const Prob Pcat6[] =
-{ 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129};
+static const Prob Pcat6[] = {
+ 254, 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129
+};
-static vp9_tree_index cat1[2], cat2[4], cat3[6], cat4[8], cat5[10], cat6[26];
+static vp9_tree_index cat1[2], cat2[4], cat3[6], cat4[8], cat5[10], cat6[28];
static void init_bit_tree(vp9_tree_index *p, int n) {
int i = 0;
@@ -182,7 +286,7 @@ static void init_bit_trees() {
init_bit_tree(cat3, 3);
init_bit_tree(cat4, 4);
init_bit_tree(cat5, 5);
- init_bit_tree(cat6, 13);
+ init_bit_tree(cat6, 14);
}
vp9_extra_bit_struct vp9_extra_bits[12] = {
@@ -196,7 +300,7 @@ vp9_extra_bit_struct vp9_extra_bits[12] = {
{ cat3, Pcat3, 3, 11},
{ cat4, Pcat4, 4, 19},
{ cat5, Pcat5, 5, 35},
- { cat6, Pcat6, 13, 67},
+ { cat6, Pcat6, 14, 67},
{ 0, 0, 0, 0}
};
@@ -218,6 +322,11 @@ void vp9_default_coef_probs(VP9_COMMON *pc) {
vpx_memcpy(pc->fc.hybrid_coef_probs_16x16,
default_hybrid_coef_probs_16x16,
sizeof(pc->fc.hybrid_coef_probs_16x16));
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vpx_memcpy(pc->fc.coef_probs_32x32, default_coef_probs_32x32,
+ sizeof(pc->fc.coef_probs_32x32));
+#endif
}
void vp9_coef_tree_initialize() {
@@ -444,4 +553,28 @@ void vp9_adapt_coef_probs(VP9_COMMON *cm) {
else cm->fc.hybrid_coef_probs_16x16[i][j][k][t] = prob;
}
}
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ for (i = 0; i < BLOCK_TYPES_32X32; ++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;
+ vp9_tree_probs_from_distribution(
+ MAX_ENTROPY_TOKENS, vp9_coef_encodings, vp9_coef_tree,
+ coef_probs, branch_ct, cm->fc.coef_counts_32x32[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_32x32[i][j][k][t] *
+ (256 - factor) +
+ (int)coef_probs[t] * factor + 128) >> 8;
+ if (prob <= 0) cm->fc.coef_probs_32x32[i][j][k][t] = 1;
+ else if (prob > 255) cm->fc.coef_probs_32x32[i][j][k][t] = 255;
+ else cm->fc.coef_probs_32x32[i][j][k][t] = prob;
+ }
+ }
+#endif
}
diff --git a/vp9/common/vp9_entropy.h b/vp9/common/vp9_entropy.h
index 3c74de7be9b73dabd5e9e8ddce10b6c45f1e9e1f..96d964448f143b42bf880566445863cb6c3e95a8 100644
--- a/vp9/common/vp9_entropy.h
+++ b/vp9/common/vp9_entropy.h
@@ -55,7 +55,7 @@ extern vp9_extra_bit_struct vp9_extra_bits[12]; /* indexed by token value */
#define PROB_UPDATE_BASELINE_COST 7
#define MAX_PROB 255
-#define DCT_MAX_VALUE 8192
+#define DCT_MAX_VALUE 16384
/* Coefficients are predicted via a 3-dimensional probability table. */
@@ -66,6 +66,10 @@ extern vp9_extra_bit_struct vp9_extra_bits[12]; /* indexed by token value */
#define BLOCK_TYPES_16X16 4
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+#define BLOCK_TYPES_32X32 4
+#endif
+
/* Middle dimension is a coarsening of the coefficient's
position within the 4x4 DCT. */
@@ -73,6 +77,9 @@ extern vp9_extra_bit_struct vp9_extra_bits[12]; /* indexed by token value */
extern DECLARE_ALIGNED(16, const int, vp9_coef_bands[16]);
extern DECLARE_ALIGNED(64, const int, vp9_coef_bands_8x8[64]);
extern DECLARE_ALIGNED(16, const int, vp9_coef_bands_16x16[256]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+extern DECLARE_ALIGNED(16, const int, vp9_coef_bands_32x32[1024]);
+#endif
/* Inside dimension is 3-valued measure of nearby complexity, that is,
the extent to which nearby coefficients are nonzero. For the first
@@ -106,9 +113,13 @@ extern DECLARE_ALIGNED(16, const int, vp9_col_scan[16]);
extern DECLARE_ALIGNED(16, const int, vp9_row_scan[16]);
extern DECLARE_ALIGNED(64, const int, vp9_default_zig_zag1d_8x8[64]);
+extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_16x16[256]);
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_32x32[1024]);
+#endif
+
void vp9_coef_tree_initialize(void);
-extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_16x16[256]);
void vp9_adapt_coef_probs(struct VP9Common *);
#endif
diff --git a/vp9/common/vp9_idctllm.c b/vp9/common/vp9_idctllm.c
index 9622dfdee1335b596afa7b88f53b60b77b630903..cc685b99e68a3e52387b7c65c37a1827d43b31e5 100644
--- a/vp9/common/vp9_idctllm.c
+++ b/vp9/common/vp9_idctllm.c
@@ -1774,3 +1774,465 @@ void vp9_short_idct10_16x16_c(int16_t *input, int16_t *output, int pitch) {
#undef RIGHT_SHIFT
#undef RIGHT_ROUNDING
#endif
+
+#if CONFIG_TX32X32
+#if !CONFIG_DWT32X32HYBRID
+#define DownshiftMultiplyBy2(x) x * 2
+#define DownshiftMultiply(x) x
+static void idct16(double *input, double *output, int stride) {
+ static const double C1 = 0.995184726672197;
+ static const double C2 = 0.98078528040323;
+ static const double C3 = 0.956940335732209;
+ static const double C4 = 0.923879532511287;
+ static const double C5 = 0.881921264348355;
+ static const double C6 = 0.831469612302545;
+ static const double C7 = 0.773010453362737;
+ static const double C8 = 0.707106781186548;
+ static const double C9 = 0.634393284163646;
+ static const double C10 = 0.555570233019602;
+ static const double C11 = 0.471396736825998;
+ static const double C12 = 0.38268343236509;
+ static const double C13 = 0.290284677254462;
+ static const double C14 = 0.195090322016128;
+ static const double C15 = 0.098017140329561;
+
+ double step[16];
+ double intermediate[16];
+ double temp1, temp2;
+
+ // step 1 and 2
+ step[ 0] = input[stride*0] + input[stride*8];
+ step[ 1] = input[stride*0] - input[stride*8];
+
+ temp1 = input[stride*4]*C12;
+ temp2 = input[stride*12]*C4;
+
+ temp1 -= temp2;
+ temp1 = DownshiftMultiply(temp1);
+ temp1 *= C8;
+
+ step[ 2] = DownshiftMultiplyBy2(temp1);
+
+ temp1 = input[stride*4]*C4;
+ temp2 = input[stride*12]*C12;
+ temp1 += temp2;
+ temp1 = DownshiftMultiply(temp1);
+ temp1 *= C8;
+ step[ 3] = DownshiftMultiplyBy2(temp1);
+
+ temp1 = input[stride*2]*C8;
+ temp1 = DownshiftMultiplyBy2(temp1);
+ temp2 = input[stride*6] + input[stride*10];
+
+ step[ 4] = temp1 + temp2;
+ step[ 5] = temp1 - temp2;
+
+ temp1 = input[stride*14]*C8;
+ temp1 = DownshiftMultiplyBy2(temp1);
+ temp2 = input[stride*6] - input[stride*10];
+
+ step[ 6] = temp2 - temp1;
+ step[ 7] = temp2 + temp1;
+
+ // for odd input
+ temp1 = input[stride*3]*C12;
+ temp2 = input[stride*13]*C4;
+ temp1 += temp2;
+ temp1 = DownshiftMultiply(temp1);
+ temp1 *= C8;
+ intermediate[ 8] = DownshiftMultiplyBy2(temp1);
+
+ temp1 = input[stride*3]*C4;
+ temp2 = input[stride*13]*C12;
+ temp2 -= temp1;
+ temp2 = DownshiftMultiply(temp2);
+ temp2 *= C8;
+ intermediate[ 9] = DownshiftMultiplyBy2(temp2);
+
+ intermediate[10] = DownshiftMultiplyBy2(input[stride*9]*C8);
+ intermediate[11] = input[stride*15] - input[stride*1];
+ intermediate[12] = input[stride*15] + input[stride*1];
+ intermediate[13] = DownshiftMultiplyBy2((input[stride*7]*C8));
+
+ temp1 = input[stride*11]*C12;
+ temp2 = input[stride*5]*C4;
+ temp2 -= temp1;
+ temp2 = DownshiftMultiply(temp2);
+ temp2 *= C8;
+ intermediate[14] = DownshiftMultiplyBy2(temp2);
+
+ temp1 = input[stride*11]*C4;
+ temp2 = input[stride*5]*C12;
+ temp1 += temp2;
+ temp1 = DownshiftMultiply(temp1);
+ temp1 *= C8;
+ intermediate[15] = DownshiftMultiplyBy2(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[stride*0] = step[ 0] + step[ 3];
+ output[stride*1] = step[ 1] + step[ 2];
+ output[stride*2] = step[ 1] - step[ 2];
+ output[stride*3] = step[ 0] - step[ 3];
+
+ temp1 = step[ 4]*C14;
+ temp2 = step[ 7]*C2;
+ temp1 -= temp2;
+ output[stride*4] = DownshiftMultiply(temp1);
+
+ temp1 = step[ 4]*C2;
+ temp2 = step[ 7]*C14;
+ temp1 += temp2;
+ output[stride*7] = DownshiftMultiply(temp1);
+
+ temp1 = step[ 5]*C10;
+ temp2 = step[ 6]*C6;
+ temp1 -= temp2;
+ output[stride*5] = DownshiftMultiply(temp1);
+
+ temp1 = step[ 5]*C6;
+ temp2 = step[ 6]*C10;
+ temp1 += temp2;
+ output[stride*6] = DownshiftMultiply(temp1);
+
+ output[stride*8] = step[ 8] + step[11];
+ output[stride*9] = step[ 9] + step[10];
+ output[stride*10] = step[ 9] - step[10];
+ output[stride*11] = step[ 8] - step[11];
+ output[stride*12] = step[12] + step[15];
+ output[stride*13] = step[13] + step[14];
+ output[stride*14] = step[13] - step[14];
+ output[stride*15] = step[12] - step[15];
+
+ // output 4
+ step[ 0] = output[stride*0] + output[stride*7];
+ step[ 1] = output[stride*1] + output[stride*6];
+ step[ 2] = output[stride*2] + output[stride*5];
+ step[ 3] = output[stride*3] + output[stride*4];
+ step[ 4] = output[stride*3] - output[stride*4];
+ step[ 5] = output[stride*2] - output[stride*5];
+ step[ 6] = output[stride*1] - output[stride*6];
+ step[ 7] = output[stride*0] - output[stride*7];
+
+ temp1 = output[stride*8]*C7;
+ temp2 = output[stride*15]*C9;
+ temp1 -= temp2;
+ step[ 8] = DownshiftMultiply(temp1);
+
+ temp1 = output[stride*9]*C11;
+ temp2 = output[stride*14]*C5;
+ temp1 += temp2;
+ step[ 9] = DownshiftMultiply(temp1);
+
+ temp1 = output[stride*10]*C3;
+ temp2 = output[stride*13]*C13;
+ temp1 -= temp2;
+ step[10] = DownshiftMultiply(temp1);
+
+ temp1 = output[stride*11]*C15;
+ temp2 = output[stride*12]*C1;
+ temp1 += temp2;
+ step[11] = DownshiftMultiply(temp1);
+
+ temp1 = output[stride*11]*C1;
+ temp2 = output[stride*12]*C15;
+ temp2 -= temp1;
+ step[12] = DownshiftMultiply(temp2);
+
+ temp1 = output[stride*10]*C13;
+ temp2 = output[stride*13]*C3;
+ temp1 += temp2;
+ step[13] = DownshiftMultiply(temp1);
+
+ temp1 = output[stride*9]*C5;
+ temp2 = output[stride*14]*C11;
+ temp2 -= temp1;
+ step[14] = DownshiftMultiply(temp2);
+
+ temp1 = output[stride*8]*C9;
+ temp2 = output[stride*15]*C7;
+ temp1 += temp2;
+ step[15] = DownshiftMultiply(temp1);
+
+ // step 5
+ output[stride*0] = step[0] + step[15];
+ output[stride*1] = step[1] + step[14];
+ output[stride*2] = step[2] + step[13];
+ output[stride*3] = step[3] + step[12];
+ output[stride*4] = step[4] + step[11];
+ output[stride*5] = step[5] + step[10];
+ output[stride*6] = step[6] + step[ 9];
+ output[stride*7] = step[7] + step[ 8];
+
+ output[stride*15] = step[0] - step[15];
+ output[stride*14] = step[1] - step[14];
+ output[stride*13] = step[2] - step[13];
+ output[stride*12] = step[3] - step[12];
+ output[stride*11] = step[4] - step[11];
+ output[stride*10] = step[5] - step[10];
+ output[stride*9] = step[6] - step[ 9];
+ output[stride*8] = step[7] - step[ 8];
+}
+static void butterfly_32_idct_1d(double *input, double *output, int stride) {
+ static const double C1 = 0.998795456205; // cos(pi * 1 / 64)
+ static const double C3 = 0.989176509965; // cos(pi * 3 / 64)
+ static const double C5 = 0.970031253195; // cos(pi * 5 / 64)
+ static const double C7 = 0.941544065183; // cos(pi * 7 / 64)
+ static const double C9 = 0.903989293123; // cos(pi * 9 / 64)
+ static const double C11 = 0.857728610000; // cos(pi * 11 / 64)
+ static const double C13 = 0.803207531481; // cos(pi * 13 / 64)
+ static const double C15 = 0.740951125355; // cos(pi * 15 / 64)
+ static const double C16 = 0.707106781187; // cos(pi * 16 / 64)
+ static const double C17 = 0.671558954847; // cos(pi * 17 / 64)
+ static const double C19 = 0.595699304492; // cos(pi * 19 / 64)
+ static const double C21 = 0.514102744193; // cos(pi * 21 / 64)
+ static const double C23 = 0.427555093430; // cos(pi * 23 / 64)
+ static const double C25 = 0.336889853392; // cos(pi * 25 / 64)
+ static const double C27 = 0.242980179903; // cos(pi * 27 / 64)
+ static const double C29 = 0.146730474455; // cos(pi * 29 / 64)
+ static const double C31 = 0.049067674327; // cos(pi * 31 / 64)
+
+ double step1[32];
+ double step2[32];
+
+ step1[ 0] = input[stride*0];
+ step1[ 1] = input[stride*2];
+ step1[ 2] = input[stride*4];
+ step1[ 3] = input[stride*6];
+ step1[ 4] = input[stride*8];
+ step1[ 5] = input[stride*10];
+ step1[ 6] = input[stride*12];
+ step1[ 7] = input[stride*14];
+ step1[ 8] = input[stride*16];
+ step1[ 9] = input[stride*18];
+ step1[10] = input[stride*20];
+ step1[11] = input[stride*22];
+ step1[12] = input[stride*24];
+ step1[13] = input[stride*26];
+ step1[14] = input[stride*28];
+ step1[15] = input[stride*30];
+
+ step1[16] = DownshiftMultiplyBy2(input[stride*1]*C16);
+ step1[17] = (input[stride*3] + input[stride*1]);
+ step1[18] = (input[stride*5] + input[stride*3]);
+ step1[19] = (input[stride*7] + input[stride*5]);
+ step1[20] = (input[stride*9] + input[stride*7]);
+ step1[21] = (input[stride*11] + input[stride*9]);
+ step1[22] = (input[stride*13] + input[stride*11]);
+ step1[23] = (input[stride*15] + input[stride*13]);
+ step1[24] = (input[stride*17] + input[stride*15]);
+ step1[25] = (input[stride*19] + input[stride*17]);
+ step1[26] = (input[stride*21] + input[stride*19]);
+ step1[27] = (input[stride*23] + input[stride*21]);
+ step1[28] = (input[stride*25] + input[stride*23]);
+ step1[29] = (input[stride*27] + input[stride*25]);
+ step1[30] = (input[stride*29] + input[stride*27]);
+ step1[31] = (input[stride*31] + input[stride*29]);
+
+ idct16(step1, step2, 1);
+ idct16(step1 + 16, step2 + 16, 1);
+
+ step2[16] = DownshiftMultiply(step2[16] / (2*C1));
+ step2[17] = DownshiftMultiply(step2[17] / (2*C3));
+ step2[18] = DownshiftMultiply(step2[18] / (2*C5));
+ step2[19] = DownshiftMultiply(step2[19] / (2*C7));
+ step2[20] = DownshiftMultiply(step2[20] / (2*C9));
+ step2[21] = DownshiftMultiply(step2[21] / (2*C11));
+ step2[22] = DownshiftMultiply(step2[22] / (2*C13));
+ step2[23] = DownshiftMultiply(step2[23] / (2*C15));
+ step2[24] = DownshiftMultiply(step2[24] / (2*C17));
+ step2[25] = DownshiftMultiply(step2[25] / (2*C19));
+ step2[26] = DownshiftMultiply(step2[26] / (2*C21));
+ step2[27] = DownshiftMultiply(step2[27] / (2*C23));
+ step2[28] = DownshiftMultiply(step2[28] / (2*C25));
+ step2[29] = DownshiftMultiply(step2[29] / (2*C27));
+ step2[30] = DownshiftMultiply(step2[30] / (2*C29));
+ step2[31] = DownshiftMultiply(step2[31] / (2*C31));
+
+ output[stride* 0] = step2[ 0] + step2[16];
+ output[stride* 1] = step2[ 1] + step2[17];
+ output[stride* 2] = step2[ 2] + step2[18];
+ output[stride* 3] = step2[ 3] + step2[19];
+ output[stride* 4] = step2[ 4] + step2[20];
+ output[stride* 5] = step2[ 5] + step2[21];
+ output[stride* 6] = step2[ 6] + step2[22];
+ output[stride* 7] = step2[ 7] + step2[23];
+ output[stride* 8] = step2[ 8] + step2[24];
+ output[stride* 9] = step2[ 9] + step2[25];
+ output[stride*10] = step2[10] + step2[26];
+ output[stride*11] = step2[11] + step2[27];
+ output[stride*12] = step2[12] + step2[28];
+ output[stride*13] = step2[13] + step2[29];
+ output[stride*14] = step2[14] + step2[30];
+ output[stride*15] = step2[15] + step2[31];
+ output[stride*16] = step2[15] - step2[(31 - 0)];
+ output[stride*17] = step2[14] - step2[(31 - 1)];
+ output[stride*18] = step2[13] - step2[(31 - 2)];
+ output[stride*19] = step2[12] - step2[(31 - 3)];
+ output[stride*20] = step2[11] - step2[(31 - 4)];
+ output[stride*21] = step2[10] - step2[(31 - 5)];
+ output[stride*22] = step2[ 9] - step2[(31 - 6)];
+ output[stride*23] = step2[ 8] - step2[(31 - 7)];
+ output[stride*24] = step2[ 7] - step2[(31 - 8)];
+ output[stride*25] = step2[ 6] - step2[(31 - 9)];
+ output[stride*26] = step2[ 5] - step2[(31 - 10)];
+ output[stride*27] = step2[ 4] - step2[(31 - 11)];
+ output[stride*28] = step2[ 3] - step2[(31 - 12)];
+ output[stride*29] = step2[ 2] - step2[(31 - 13)];
+ output[stride*30] = step2[ 1] - step2[(31 - 14)];
+ output[stride*31] = step2[ 0] - step2[(31 - 15)];
+}
+
+void vp9_short_idct32x32_c(short *input, short *output, int pitch) {
+ vp9_clear_system_state(); // Make it simd safe : __asm emms;
+ {
+ double out[32*32], out2[32*32];
+ const int short_pitch = pitch >> 1;
+ int i, j;
+ // First transform rows
+ for (i = 0; i < 32; ++i) {
+ double temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j)
+ temp_in[j] = input[j + i*short_pitch];
+ butterfly_32_idct_1d(temp_in, temp_out, 1);
+ for (j = 0; j < 32; ++j)
+ out[j + i*32] = temp_out[j];
+ }
+ // Then transform columns
+ for (i = 0; i < 32; ++i) {
+ double temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j)
+ temp_in[j] = out[j*32 + i];
+ butterfly_32_idct_1d(temp_in, temp_out, 1);
+ for (j = 0; j < 32; ++j)
+ out2[j*32 + i] = temp_out[j];
+ }
+ for (i = 0; i < 32*32; ++i)
+ output[i] = round(out2[i]/128);
+ }
+ vp9_clear_system_state(); // Make it simd safe : __asm emms;
+}
+#else // CONFIG_DWT32X32HYBRID
+
+#define MAX_BLOCK_LENGTH 64
+#define ENH_PRECISION_BITS 1
+#define ENH_PRECISION_RND ((1 << ENH_PRECISION_BITS) / 2)
+
+// Note: block length must be even for this implementation
+static void synthesis_53_row(int length, short *lowpass, short *highpass,
+ short *x) {
+ short r, * a, * b;
+ int n;
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ r = *highpass;
+ while (n--) {
+ *a++ -= (r + (*b) + 1) >> 1;
+ r = *b++;
+ }
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ while (--n) {
+ *x++ = ((r = *a++) + 1) >> 1;
+ *x++ = *b++ + ((r + (*a) + 2) >> 2);
+ }
+ *x++ = ((r = *a) + 1)>>1;
+ *x++ = *b + ((r+1)>>1);
+}
+
+static void synthesis_53_col(int length, short *lowpass, short *highpass,
+ short *x) {
+ short r, * a, * b;
+ int n;
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ r = *highpass;
+ while (n--) {
+ *a++ -= (r + (*b) + 1) >> 1;
+ r = *b++;
+ }
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ while (--n) {
+ *x++ = r = *a++;
+ *x++ = ((*b++) << 1) + ((r + (*a) + 1) >> 1);
+ }
+ *x++ = r = *a;
+ *x++ = ((*b) << 1) + r;
+}
+
+// NOTE: Using a 5/3 integer wavelet for now. Explore using a wavelet
+// with a better response later
+void dyadic_synthesize(int levels, int width, int height, short *c, int pitch_c,
+ short *x, int pitch_x) {
+ int th[16], tw[16], lv, i, j, nh, nw, hh = height, hw = width;
+ short buffer[2 * MAX_BLOCK_LENGTH];
+
+ th[0] = hh;
+ tw[0] = hw;
+ for (i = 1; i <= levels; i++) {
+ th[i] = (th[i - 1] + 1) >> 1;
+ tw[i] = (tw[i - 1] + 1) >> 1;
+ }
+ for (lv = levels - 1; lv >= 0; lv--) {
+ nh = th[lv];
+ nw = tw[lv];
+ hh = th[lv + 1];
+ hw = tw[lv + 1];
+ if ((nh < 2) || (nw < 2)) continue;
+ for (j = 0; j < nw; j++) {
+ for (i = 0; i < nh; i++)
+ buffer[i] = c[i * pitch_c + j];
+ synthesis_53_col(nh, buffer, buffer + hh, buffer + nh);
+ for (i = 0; i < nh; i++)
+ c[i * pitch_c + j] = buffer[i + nh];
+ }
+ for (i = 0; i < nh; i++) {
+ memcpy(buffer, &c[i * pitch_c], nw * sizeof(short));
+ synthesis_53_row(nw, buffer, buffer + hw, &c[i * pitch_c]);
+ }
+ }
+ for (i = 0; i < height; i++)
+ for (j = 0; j < width; j++)
+ x[i * pitch_x + j] = (c[i * pitch_c + j] + ENH_PRECISION_RND) >>
+ ENH_PRECISION_BITS;
+}
+
+void vp9_short_idct32x32_c(short *input, short *output, int pitch) {
+ // assume out is a 32x32 buffer
+ short buffer[16 * 16];
+ short buffer2[32 * 32];
+ const int short_pitch = pitch >> 1;
+ int i;
+ // TODO(debargha): Implement more efficiently by adding output pitch
+ // argument to the idct16x16 function
+ vp9_short_idct16x16_c(input, buffer, pitch);
+ for (i = 0; i < 16; ++i) {
+ vpx_memcpy(buffer2 + i * 32, buffer + i * 16, sizeof(short) * 16);
+ vpx_memcpy(buffer2 + i * 32 + 16, input + i * short_pitch + 16,
+ sizeof(short) * 16);
+ }
+ for (; i < 32; ++i) {
+ vpx_memcpy(buffer2 + i * 32, input + i * short_pitch,
+ sizeof(short) * 32);
+ }
+ dyadic_synthesize(1, 32, 32, buffer2, 32, output, 32);
+}
+#endif // CONFIG_DWT32X32HYBRID
+#endif // CONFIG_TX32X32
diff --git a/vp9/common/vp9_invtrans.c b/vp9/common/vp9_invtrans.c
index c78f1ad3c6ab26801d54627c7ad206e08787922e..3abf328940edcad0eede716a125a569f048b614b 100644
--- a/vp9/common/vp9_invtrans.c
+++ b/vp9/common/vp9_invtrans.c
@@ -143,3 +143,16 @@ void vp9_inverse_transform_mb_16x16(MACROBLOCKD *xd) {
vp9_inverse_transform_mby_16x16(xd);
vp9_inverse_transform_mbuv_8x8(xd);
}
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+void vp9_inverse_transform_sby_32x32(SUPERBLOCKD *xd_sb) {
+ vp9_short_idct32x32(xd_sb->dqcoeff, xd_sb->diff, 64);
+}
+
+void vp9_inverse_transform_sbuv_16x16(SUPERBLOCKD *xd_sb) {
+ vp9_inverse_transform_b_16x16(xd_sb->dqcoeff + 1024,
+ xd_sb->diff + 1024, 32);
+ vp9_inverse_transform_b_16x16(xd_sb->dqcoeff + 1280,
+ xd_sb->diff + 1280, 32);
+}
+#endif
diff --git a/vp9/common/vp9_invtrans.h b/vp9/common/vp9_invtrans.h
index b012834f37337e4d93111d2738aa1e7afbe16bf6..94593f8cc60eb440d6889bd76de273934daab4da 100644
--- a/vp9/common/vp9_invtrans.h
+++ b/vp9/common/vp9_invtrans.h
@@ -38,4 +38,9 @@ extern void vp9_inverse_transform_mb_16x16(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mby_16x16(MACROBLOCKD *xd);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+extern void vp9_inverse_transform_sby_32x32(SUPERBLOCKD *xd_sb);
+extern void vp9_inverse_transform_sbuv_16x16(SUPERBLOCKD *xd_sb);
+#endif
+
#endif // __INC_INVTRANS_H
diff --git a/vp9/common/vp9_loopfilter.c b/vp9/common/vp9_loopfilter.c
index 8d4d014ba7cfc912697b3ae2739b5f3bc0b54bef..1139fb5d13161b0fb56fab80682758263eb8f0c8 100644
--- a/vp9/common/vp9_loopfilter.c
+++ b/vp9/common/vp9_loopfilter.c
@@ -192,6 +192,9 @@ void vp9_loop_filter_frame(VP9_COMMON *cm, MACROBLOCKD *xd) {
/* Point at base of Mb MODE_INFO list */
const MODE_INFO *mode_info_context = cm->mi;
+#if CONFIG_SUPERBLOCKS
+ const int mis = cm->mode_info_stride;
+#endif
/* Initialize the loop filter for this frame. */
vp9_loop_filter_frame_init(cm, xd, cm->filter_level);
@@ -226,14 +229,18 @@ void vp9_loop_filter_frame(VP9_COMMON *cm, MACROBLOCKD *xd) {
if (mb_col > 0
#if CONFIG_SUPERBLOCKS
&& !((mb_col & 1) && mode_info_context->mbmi.encoded_as_sb &&
- mode_info_context[0].mbmi.mb_skip_coeff &&
- mode_info_context[-1].mbmi.mb_skip_coeff)
+ ((mode_info_context[0].mbmi.mb_skip_coeff &&
+ mode_info_context[-1].mbmi.mb_skip_coeff)
+#if CONFIG_TX32X32
+ || mode_info_context[-1].mbmi.txfm_size == TX_32X32
+#endif
+ ))
#endif
)
vp9_loop_filter_mbv(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
- if (!skip_lf && tx_type != TX_16X16) {
+ if (!skip_lf && tx_type < TX_16X16) {
if (tx_type == TX_8X8)
vp9_loop_filter_bv8x8(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
@@ -247,14 +254,18 @@ void vp9_loop_filter_frame(VP9_COMMON *cm, MACROBLOCKD *xd) {
if (mb_row > 0
#if CONFIG_SUPERBLOCKS
&& !((mb_row & 1) && mode_info_context->mbmi.encoded_as_sb &&
- mode_info_context[0].mbmi.mb_skip_coeff &&
- mode_info_context[-cm->mode_info_stride].mbmi.mb_skip_coeff)
+ ((mode_info_context[0].mbmi.mb_skip_coeff &&
+ mode_info_context[-mis].mbmi.mb_skip_coeff)
+#if CONFIG_TX32X32
+ || mode_info_context[-mis].mbmi.txfm_size == TX_32X32
+#endif
+ ))
#endif
)
vp9_loop_filter_mbh(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
- if (!skip_lf && tx_type != TX_16X16) {
+ if (!skip_lf && tx_type < TX_16X16) {
if (tx_type == TX_8X8)
vp9_loop_filter_bh8x8(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
diff --git a/vp9/common/vp9_onyxc_int.h b/vp9/common/vp9_onyxc_int.h
index 0b6de7f82a2360e742b2c902b64c8f725c7bd206..d80498df1ad4e18dd0fd32a46e9547e607adea50 100644
--- a/vp9/common/vp9_onyxc_int.h
+++ b/vp9/common/vp9_onyxc_int.h
@@ -58,6 +58,9 @@ typedef struct frame_contexts {
vp9_prob hybrid_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob hybrid_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_prob coef_probs_32x32 [BLOCK_TYPES_32X32] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#endif
nmv_context nmvc;
nmv_context pre_nmvc;
@@ -95,6 +98,11 @@ typedef struct frame_contexts {
vp9_prob pre_hybrid_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_prob pre_coef_probs_32x32 [BLOCK_TYPES_32X32] [COEF_BANDS]
+ [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+#endif
+
unsigned int coef_counts [BLOCK_TYPES] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
unsigned int hybrid_coef_counts [BLOCK_TYPES] [COEF_BANDS]
@@ -110,6 +118,11 @@ typedef struct frame_contexts {
unsigned int hybrid_coef_counts_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ unsigned int coef_counts_32x32 [BLOCK_TYPES_32X32] [COEF_BANDS]
+ [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
+#endif
+
nmv_context_counts NMVcount;
vp9_prob switchable_interp_prob[VP9_SWITCHABLE_FILTERS + 1]
[VP9_SWITCHABLE_FILTERS - 1];
@@ -139,8 +152,11 @@ typedef enum {
ONLY_4X4 = 0,
ALLOW_8X8 = 1,
ALLOW_16X16 = 2,
- TX_MODE_SELECT = 3,
- NB_TXFM_MODES = 4,
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ ALLOW_32X32 = 3,
+#endif
+ TX_MODE_SELECT = 3 + (CONFIG_TX32X32 && CONFIG_SUPERBLOCKS),
+ NB_TXFM_MODES = 4 + (CONFIG_TX32X32 && CONFIG_SUPERBLOCKS),
} TXFM_MODE;
typedef struct VP9Common {
@@ -268,7 +284,7 @@ typedef struct VP9Common {
vp9_prob prob_comppred[COMP_PRED_CONTEXTS];
// FIXME contextualize
- vp9_prob prob_tx[TX_SIZE_MAX - 1];
+ vp9_prob prob_tx[TX_SIZE_MAX_SB - 1];
vp9_prob mbskip_pred_probs[MBSKIP_CONTEXTS];
diff --git a/vp9/common/vp9_recon.c b/vp9/common/vp9_recon.c
index 1f8dfce34b2f2e6be18293a2ba4d36e5b9d3fc69..e567bac8d54cec696bd7c52d09155ec8d5bf920b 100644
--- a/vp9/common/vp9_recon.c
+++ b/vp9/common/vp9_recon.c
@@ -168,6 +168,53 @@ void vp9_recon_mbuv_s_c(MACROBLOCKD *xd, uint8_t *udst, uint8_t *vdst) {
}
}
}
+
+#if CONFIG_TX32X32
+void vp9_recon_sby_s_c(MACROBLOCKD *xd, uint8_t *dst) {
+ int x, y, stride = xd->block[0].dst_stride;
+ short *diff = xd->sb_coeff_data.diff;
+
+ for (y = 0; y < 32; y++) {
+ for (x = 0; x < 32; x++) {
+ int a = dst[x] + diff[x];
+ if (a < 0)
+ a = 0;
+ else if (a > 255)
+ a = 255;
+ dst[x] = a;
+ }
+ dst += stride;
+ diff += 32;
+ }
+}
+
+void vp9_recon_sbuv_s_c(MACROBLOCKD *xd, uint8_t *udst, uint8_t *vdst) {
+ int x, y, stride = xd->block[16].dst_stride;
+ short *udiff = xd->sb_coeff_data.diff + 1024;
+ short *vdiff = xd->sb_coeff_data.diff + 1280;
+
+ for (y = 0; y < 16; y++) {
+ for (x = 0; x < 16; x++) {
+ int u = udst[x] + udiff[x];
+ int v = vdst[x] + vdiff[x];
+ if (u < 0)
+ u = 0;
+ else if (u > 255)
+ u = 255;
+ if (v < 0)
+ v = 0;
+ else if (v > 255)
+ v = 255;
+ udst[x] = u;
+ vdst[x] = v;
+ }
+ udst += stride;
+ vdst += stride;
+ udiff += 16;
+ vdiff += 16;
+ }
+}
+#endif
#endif
void vp9_recon_mby_c(MACROBLOCKD *xd) {
diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh
index 5b7af100b636f9efb2665cec789b45b63b79b5c1..49a3a8595ed7b02871b4c1b9e56edae7ec6aa80a 100644
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -361,6 +361,9 @@ specialize vp9_short_idct16x16
prototype void vp9_short_idct10_16x16 "short *input, short *output, int pitch"
specialize vp9_short_idct10_16x16
+prototype void vp9_short_idct32x32 "short *input, short *output, int pitch"
+specialize vp9_short_idct32x32
+
prototype void vp9_ihtllm "const short *input, short *output, int pitch, int tx_type, int tx_dim"
specialize vp9_ihtllm
@@ -640,6 +643,9 @@ specialize vp9_short_fdct8x4
prototype void vp9_short_walsh4x4 "short *InputData, short *OutputData, int pitch"
specialize vp9_short_walsh4x4
+prototype void vp9_short_fdct32x32 "short *InputData, short *OutputData, int pitch"
+specialize vp9_short_fdct32x32
+
prototype void vp9_short_fdct16x16 "short *InputData, short *OutputData, int pitch"
specialize vp9_short_fdct16x16
diff --git a/vp9/common/vp9_seg_common.c b/vp9/common/vp9_seg_common.c
index 46a6ee45444f25203b7a6142f5e7ac6a72699fee..89c1e458dd851b3fc4aa80b247dc34f8b3470409 100644
--- a/vp9/common/vp9_seg_common.c
+++ b/vp9/common/vp9_seg_common.c
@@ -14,7 +14,7 @@
static const int segfeaturedata_signed[SEG_LVL_MAX] = { 1, 1, 0, 0, 0, 0 };
static const int seg_feature_data_max[SEG_LVL_MAX] =
- { MAXQ, 63, 0xf, MB_MODE_COUNT - 1, 255, TX_SIZE_MAX - 1};
+ { MAXQ, 63, 0xf, MB_MODE_COUNT - 1, 255, TX_SIZE_MAX_SB - 1};
// These functions provide access to new segment level features.
// Eventually these function may be "optimized out" but for the moment,
diff --git a/vp9/decoder/vp9_decodemv.c b/vp9/decoder/vp9_decodemv.c
index 7e53884f78fd4282e27d7715c344e93a16d85a03..b9f411dd2959cbcb5ba65a8681168390980b9403 100644
--- a/vp9/decoder/vp9_decodemv.c
+++ b/vp9/decoder/vp9_decodemv.c
@@ -209,8 +209,17 @@ static void kfread_modes(VP9D_COMP *pbi,
m->mbmi.mode <= I8X8_PRED) {
// FIXME(rbultje) code ternary symbol once all experiments are merged
m->mbmi.txfm_size = vp9_read(bc, cm->prob_tx[0]);
- if (m->mbmi.txfm_size != TX_4X4 && m->mbmi.mode != I8X8_PRED)
+ if (m->mbmi.txfm_size != TX_4X4 && m->mbmi.mode != I8X8_PRED) {
m->mbmi.txfm_size += vp9_read(bc, cm->prob_tx[1]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (m->mbmi.txfm_size != TX_8X8 && m->mbmi.encoded_as_sb)
+ m->mbmi.txfm_size += vp9_read(bc, cm->prob_tx[2]);
+#endif
+ }
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ } else if (cm->txfm_mode >= ALLOW_32X32 && m->mbmi.encoded_as_sb) {
+ m->mbmi.txfm_size = TX_32X32;
+#endif
} else if (cm->txfm_mode >= ALLOW_16X16 && m->mbmi.mode <= TM_PRED) {
m->mbmi.txfm_size = TX_16X16;
} else if (cm->txfm_mode >= ALLOW_8X8 && m->mbmi.mode != B_PRED) {
@@ -1219,8 +1228,17 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
// FIXME(rbultje) code ternary symbol once all experiments are merged
mbmi->txfm_size = vp9_read(bc, cm->prob_tx[0]);
if (mbmi->txfm_size != TX_4X4 && mbmi->mode != I8X8_PRED &&
- mbmi->mode != SPLITMV)
+ mbmi->mode != SPLITMV) {
mbmi->txfm_size += vp9_read(bc, cm->prob_tx[1]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (mbmi->encoded_as_sb && mbmi->txfm_size != TX_8X8)
+ mbmi->txfm_size += vp9_read(bc, cm->prob_tx[2]);
+#endif
+ }
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ } else if (mbmi->encoded_as_sb && cm->txfm_mode >= ALLOW_32X32) {
+ mbmi->txfm_size = TX_32X32;
+#endif
} else if (cm->txfm_mode >= ALLOW_16X16 &&
((mbmi->ref_frame == INTRA_FRAME && mbmi->mode <= TM_PRED) ||
(mbmi->ref_frame != INTRA_FRAME && mbmi->mode != SPLITMV))) {
diff --git a/vp9/decoder/vp9_decodframe.c b/vp9/decoder/vp9_decodframe.c
index 76349ad86737c01359808754a23656000acad37d..7f851a18a48cde74e24e7d73f458789a6c55f0d7 100644
--- a/vp9/decoder/vp9_decodframe.c
+++ b/vp9/decoder/vp9_decodframe.c
@@ -693,6 +693,7 @@ static void decode_superblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
TX_SIZE tx_size = xd->mode_info_context->mbmi.txfm_size;
VP9_COMMON *const pc = &pbi->common;
MODE_INFO *orig_mi = xd->mode_info_context;
+ const int mis = pc->mode_info_stride;
assert(xd->mode_info_context->mbmi.encoded_as_sb);
@@ -733,6 +734,30 @@ static void decode_superblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
}
/* dequantization and idct */
+#if CONFIG_TX32X32
+ if (xd->mode_info_context->mbmi.txfm_size == TX_32X32) {
+ eobtotal = vp9_decode_sb_tokens(pbi, xd, bc);
+ if (eobtotal == 0) { // skip loopfilter
+ xd->mode_info_context->mbmi.mb_skip_coeff = 1;
+ if (mb_col + 1 < pc->mb_cols)
+ xd->mode_info_context[1].mbmi.mb_skip_coeff = 1;
+ if (mb_row + 1 < pc->mb_rows) {
+ xd->mode_info_context[mis].mbmi.mb_skip_coeff = 1;
+ if (mb_col + 1 < pc->mb_cols)
+ xd->mode_info_context[mis + 1].mbmi.mb_skip_coeff = 1;
+ }
+ } else {
+ vp9_dequant_idct_add_32x32(xd->sb_coeff_data.qcoeff, xd->block[0].dequant,
+ xd->dst.y_buffer, xd->dst.y_buffer,
+ xd->dst.y_stride, xd->dst.y_stride,
+ xd->eobs[0]);
+ vp9_dequant_idct_add_uv_block_16x16_c(xd->sb_coeff_data.qcoeff + 1024,
+ xd->block[16].dequant,
+ xd->dst.u_buffer, xd->dst.v_buffer,
+ xd->dst.uv_stride, xd->eobs + 16);
+ }
+ } else {
+#endif
for (n = 0; n < 4; n++) {
int x_idx = n & 1, y_idx = n >> 1;
@@ -742,7 +767,7 @@ static void decode_superblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
xd->above_context = pc->above_context + mb_col + x_idx;
xd->left_context = pc->left_context + y_idx;
- xd->mode_info_context = orig_mi + x_idx + y_idx * pc->mode_info_stride;
+ xd->mode_info_context = orig_mi + x_idx + y_idx * mis;
for (i = 0; i < 25; i++) {
xd->block[i].eob = 0;
xd->eobs[i] = 0;
@@ -766,6 +791,9 @@ static void decode_superblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
xd->above_context = pc->above_context + mb_col;
xd->left_context = pc->left_context;
xd->mode_info_context = orig_mi;
+#if CONFIG_TX32X32
+ }
+#endif
}
#endif
@@ -1244,6 +1272,11 @@ static void read_coef_probs(VP9D_COMP *pbi, BOOL_DECODER* const bc) {
read_coef_probs_common(bc, pc->fc.coef_probs_16x16);
read_coef_probs_common(bc, pc->fc.hybrid_coef_probs_16x16);
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (pbi->common.txfm_mode > ALLOW_16X16) {
+ read_coef_probs_common(bc, pc->fc.coef_probs_32x32);
+ }
+#endif
}
int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
@@ -1433,9 +1466,16 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
/* Read the loop filter level and type */
pc->txfm_mode = vp9_read_literal(&header_bc, 2);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (pc->txfm_mode == 3)
+ pc->txfm_mode += vp9_read_bit(&header_bc);
+#endif
if (pc->txfm_mode == TX_MODE_SELECT) {
pc->prob_tx[0] = vp9_read_literal(&header_bc, 8);
pc->prob_tx[1] = vp9_read_literal(&header_bc, 8);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ pc->prob_tx[2] = vp9_read_literal(&header_bc, 8);
+#endif
}
pc->filter_type = (LOOPFILTERTYPE) vp9_read_bit(&header_bc);
@@ -1591,6 +1631,10 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
pbi->common.fc.coef_probs_16x16);
vp9_copy(pbi->common.fc.pre_hybrid_coef_probs_16x16,
pbi->common.fc.hybrid_coef_probs_16x16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_copy(pbi->common.fc.pre_coef_probs_32x32,
+ pbi->common.fc.coef_probs_32x32);
+#endif
vp9_copy(pbi->common.fc.pre_ymode_prob, pbi->common.fc.ymode_prob);
#if CONFIG_SUPERBLOCKS
vp9_copy(pbi->common.fc.pre_sb_ymode_prob, pbi->common.fc.sb_ymode_prob);
@@ -1610,6 +1654,9 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
vp9_zero(pbi->common.fc.hybrid_coef_counts_8x8);
vp9_zero(pbi->common.fc.coef_counts_16x16);
vp9_zero(pbi->common.fc.hybrid_coef_counts_16x16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_zero(pbi->common.fc.coef_counts_32x32);
+#endif
vp9_zero(pbi->common.fc.ymode_counts);
#if CONFIG_SUPERBLOCKS
vp9_zero(pbi->common.fc.sb_ymode_counts);
diff --git a/vp9/decoder/vp9_dequantize.c b/vp9/decoder/vp9_dequantize.c
index 79114d58ca1a3cfb2e9ee426f5d43887f7410650..22a66716fd2de62325d43f80eba1276f88f3e1cb 100644
--- a/vp9/decoder/vp9_dequantize.c
+++ b/vp9/decoder/vp9_dequantize.c
@@ -352,3 +352,30 @@ void vp9_dequant_idct_add_16x16_c(int16_t *input, const int16_t *dq,
add_residual(diff_ptr, pred, pitch, dest, stride, 16, 16);
}
}
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+void vp9_dequant_idct_add_32x32(int16_t *input, const int16_t *dq,
+ uint8_t *pred, uint8_t *dest, int pitch,
+ int stride, uint16_t eobs) {
+ short output[1024];
+ int i;
+
+ input[0]= input[0] * dq[0] / 2;
+ for (i = 1; i < 1024; i++)
+ input[i] = input[i] * dq[1] / 2;
+ vp9_short_idct32x32_c(input, output, 64);
+ vpx_memset(input, 0, 2048);
+
+ add_residual(output, pred, pitch, dest, stride, 32, 32);
+}
+
+void vp9_dequant_idct_add_uv_block_16x16_c(short *q, const short *dq,
+ unsigned char *dstu,
+ unsigned char *dstv,
+ int stride,
+ unsigned short *eobs) {
+ vp9_dequant_idct_add_16x16_c(q, dq, dstu, dstu, stride, stride, eobs[0]);
+ vp9_dequant_idct_add_16x16_c(q + 256, dq,
+ dstv, dstv, stride, stride, eobs[4]);
+}
+#endif
diff --git a/vp9/decoder/vp9_detokenize.c b/vp9/decoder/vp9_detokenize.c
index 897ad5204ac0dbbc162357c8a126a7695dea197c..35a26477ad6818eef10cf03e8aa41488f04a8484 100644
--- a/vp9/decoder/vp9_detokenize.c
+++ b/vp9/decoder/vp9_detokenize.c
@@ -55,8 +55,9 @@
#define CAT5_PROB3 157
#define CAT5_PROB4 180
-static const unsigned char cat6_prob[14] =
-{ 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0 };
+static const unsigned char cat6_prob[15] = {
+ 254, 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0
+};
void vp9_reset_mb_tokens_context(MACROBLOCKD* const xd) {
/* Clear entropy contexts */
@@ -161,6 +162,12 @@ static int decode_coefs(VP9D_COMP *dx, const MACROBLOCKD *xd,
coef_counts = fc->hybrid_coef_counts_16x16[type];
}
break;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ case TX_32X32:
+ coef_probs = fc->coef_probs_32x32[type];
+ coef_counts = fc->coef_counts_32x32[type];
+ break;
+#endif
}
VP9_COMBINEENTROPYCONTEXTS(pt, *a, *l);
@@ -256,6 +263,54 @@ static int get_eob(MACROBLOCKD* const xd, int segment_id, int eob_max) {
return eob;
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+int vp9_decode_sb_tokens(VP9D_COMP* const pbi,
+ MACROBLOCKD* const xd,
+ BOOL_DECODER* const bc) {
+ ENTROPY_CONTEXT* const A = (ENTROPY_CONTEXT *)xd->above_context;
+ ENTROPY_CONTEXT* const L = (ENTROPY_CONTEXT *)xd->left_context;
+ unsigned short* const eobs = xd->eobs;
+ const int segment_id = xd->mode_info_context->mbmi.segment_id;
+ int c, i, eobtotal = 0, seg_eob;
+
+ // Luma block
+ eobs[0] = c = decode_coefs(pbi, xd, bc, A, L, PLANE_TYPE_Y_WITH_DC,
+ DCT_DCT, get_eob(xd, segment_id, 1024),
+ xd->sb_coeff_data.qcoeff,
+ vp9_default_zig_zag1d_32x32,
+ TX_32X32, vp9_coef_bands_32x32);
+ A[1] = A[2] = A[3] = A[0];
+ L[1] = L[2] = L[3] = L[0];
+ eobtotal += c;
+
+ // 16x16 chroma blocks
+ seg_eob = get_eob(xd, segment_id, 256);
+ for (i = 16; i < 24; i += 4) {
+ ENTROPY_CONTEXT* const a = A + vp9_block2above_8x8[i];
+ ENTROPY_CONTEXT* const l = L + vp9_block2left_8x8[i];
+
+ eobs[i] = c = decode_coefs(pbi, xd, bc, a, l, PLANE_TYPE_UV,
+ DCT_DCT, seg_eob,
+ xd->sb_coeff_data.qcoeff + 1024 + (i - 16) * 64,
+ vp9_default_zig_zag1d_16x16,
+ TX_16X16, vp9_coef_bands_16x16);
+ a[1] = a[0];
+ l[1] = l[0];
+ eobtotal += c;
+ }
+
+ // no Y2 block
+ vpx_memset(&A[8], 0, sizeof(A[8]));
+ vpx_memset(&L[8], 0, sizeof(L[8]));
+
+ vpx_memcpy(xd->above_context + 1, xd->above_context,
+ sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(xd->left_context + 1, xd->left_context,
+ sizeof(ENTROPY_CONTEXT_PLANES));
+
+ return eobtotal;
+}
+#endif
static int vp9_decode_mb_tokens_16x16(VP9D_COMP* const pbi,
MACROBLOCKD* const xd,
diff --git a/vp9/decoder/vp9_detokenize.h b/vp9/decoder/vp9_detokenize.h
index 9b319d4a951c856940a10ddffcc79b763d32c8c2..09d354ea6c5ddefdd151e6e86e60590948a63b33 100644
--- a/vp9/decoder/vp9_detokenize.h
+++ b/vp9/decoder/vp9_detokenize.h
@@ -23,6 +23,12 @@ int vp9_decode_coefs_4x4(VP9D_COMP *dx, MACROBLOCKD *xd,
int vp9_decode_mb_tokens(VP9D_COMP* const, MACROBLOCKD* const,
BOOL_DECODER* const);
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+int vp9_decode_sb_tokens(VP9D_COMP* const pbi,
+ MACROBLOCKD* const xd,
+ BOOL_DECODER* const bc);
+#endif
+
int vp9_decode_mb_tokens_4x4_uv(VP9D_COMP* const dx, MACROBLOCKD* const xd,
BOOL_DECODER* const bc);
diff --git a/vp9/encoder/vp9_bitstream.c b/vp9/encoder/vp9_bitstream.c
index 0adaeee0a14d6ee008b1bef942a320dbd3cb60b2..847815f50849dbff0c37d9446d72f1c8c4410748 100644
--- a/vp9/encoder/vp9_bitstream.c
+++ b/vp9/encoder/vp9_bitstream.c
@@ -1200,8 +1200,13 @@ static void pack_inter_mode_mvs(VP9_COMP *const cpi, vp9_writer *const bc) {
TX_SIZE sz = mi->txfm_size;
// FIXME(rbultje) code ternary symbol once all experiments are merged
vp9_write(bc, sz != TX_4X4, pc->prob_tx[0]);
- if (sz != TX_4X4 && mode != I8X8_PRED && mode != SPLITMV)
+ if (sz != TX_4X4 && mode != I8X8_PRED && mode != SPLITMV) {
vp9_write(bc, sz != TX_8X8, pc->prob_tx[1]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (mi->encoded_as_sb && sz != TX_8X8)
+ vp9_write(bc, sz != TX_16X16, pc->prob_tx[2]);
+#endif
+ }
}
#ifdef ENTROPY_STATS
@@ -1337,8 +1342,13 @@ static void write_mb_modes_kf(const VP9_COMMON *c,
TX_SIZE sz = m->mbmi.txfm_size;
// FIXME(rbultje) code ternary symbol once all experiments are merged
vp9_write(bc, sz != TX_4X4, c->prob_tx[0]);
- if (sz != TX_4X4 && ym <= TM_PRED)
+ if (sz != TX_4X4 && ym <= TM_PRED) {
vp9_write(bc, sz != TX_8X8, c->prob_tx[1]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (m->mbmi.encoded_as_sb && sz != TX_8X8)
+ vp9_write(bc, sz != TX_16X16, c->prob_tx[2]);
+#endif
+ }
}
}
@@ -1547,29 +1557,54 @@ static void build_coeff_contexts(VP9_COMP *cpi) {
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
+ }
+ }
+ }
+ 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;
+ vp9_tree_probs_from_distribution(
+ MAX_ENTROPY_TOKENS, vp9_coef_encodings, vp9_coef_tree,
+ cpi->frame_hybrid_coef_probs_16x16[i][j][k],
+ cpi->frame_hybrid_branch_ct_16x16[i][j][k],
+ cpi->hybrid_coef_counts_16x16[i][j][k], 256, 1);
+#ifdef ENTROPY_STATS
+ if (!cpi->dummy_packing)
+ for (t = 0; t < MAX_ENTROPY_TOKENS; ++t)
+ hybrid_context_counters_16x16[i][j][k][t] +=
+ cpi->hybrid_coef_counts_16x16[i][j][k][t];
#endif
}
}
}
}
- 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;
- vp9_tree_probs_from_distribution(
- MAX_ENTROPY_TOKENS, vp9_coef_encodings, vp9_coef_tree,
- cpi->frame_hybrid_coef_probs_16x16[i][j][k],
- cpi->frame_hybrid_branch_ct_16x16[i][j][k],
- cpi->hybrid_coef_counts_16x16[i][j][k], 256, 1);
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (cpi->common.txfm_mode > ALLOW_16X16) {
+ for (i = 0; i < BLOCK_TYPES_32X32; ++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;
+ vp9_tree_probs_from_distribution(
+ MAX_ENTROPY_TOKENS, vp9_coef_encodings, vp9_coef_tree,
+ cpi->frame_coef_probs_32x32[i][j][k],
+ cpi->frame_branch_ct_32x32[i][j][k],
+ cpi->coef_counts_32x32[i][j][k], 256, 1);
#ifdef ENTROPY_STATS
- if (!cpi->dummy_packing)
- for (t = 0; t < MAX_ENTROPY_TOKENS; ++t)
- hybrid_context_counters_16x16[i][j][k][t] += cpi->hybrid_coef_counts_16x16[i][j][k][t];
+ if (!cpi->dummy_packing)
+ for (t = 0; t < MAX_ENTROPY_TOKENS; ++t)
+ context_counters_32x32[i][j][k][t] +=
+ cpi->coef_counts_32x32[i][j][k][t];
#endif
+ }
}
}
}
+#endif
}
static void update_coef_probs_common(
@@ -1714,6 +1749,15 @@ static void update_coef_probs(VP9_COMP* const cpi, vp9_writer* const bc) {
cpi->common.fc.hybrid_coef_probs_16x16,
cpi->frame_hybrid_branch_ct_16x16);
}
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (cpi->common.txfm_mode > ALLOW_16X16) {
+ update_coef_probs_common(bc,
+ cpi->frame_coef_probs_32x32,
+ cpi->common.fc.coef_probs_32x32,
+ cpi->frame_branch_ct_32x32);
+ }
+#endif
}
#ifdef PACKET_TESTING
@@ -1955,18 +1999,53 @@ void vp9_pack_bitstream(VP9_COMP *cpi, unsigned char *dest,
{
if (pc->txfm_mode == TX_MODE_SELECT) {
- pc->prob_tx[0] = get_prob(cpi->txfm_count[0] + cpi->txfm_count_8x8p[0],
- cpi->txfm_count[0] + cpi->txfm_count[1] + cpi->txfm_count[2] +
- cpi->txfm_count_8x8p[0] + cpi->txfm_count_8x8p[1]);
- pc->prob_tx[1] = get_prob(cpi->txfm_count[1], cpi->txfm_count[1] + cpi->txfm_count[2]);
+ pc->prob_tx[0] = get_prob(cpi->txfm_count_32x32p[TX_4X4] +
+ cpi->txfm_count_16x16p[TX_4X4] +
+ cpi->txfm_count_8x8p[TX_4X4],
+ cpi->txfm_count_32x32p[TX_4X4] +
+ cpi->txfm_count_32x32p[TX_8X8] +
+ cpi->txfm_count_32x32p[TX_16X16] +
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ cpi->txfm_count_32x32p[TX_32X32] +
+#endif
+ cpi->txfm_count_16x16p[TX_4X4] +
+ cpi->txfm_count_16x16p[TX_8X8] +
+ cpi->txfm_count_16x16p[TX_16X16] +
+ cpi->txfm_count_8x8p[TX_4X4] +
+ cpi->txfm_count_8x8p[TX_8X8]);
+ pc->prob_tx[1] = get_prob(cpi->txfm_count_32x32p[TX_8X8] +
+ cpi->txfm_count_16x16p[TX_8X8],
+ cpi->txfm_count_32x32p[TX_8X8] +
+ cpi->txfm_count_32x32p[TX_16X16] +
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ cpi->txfm_count_32x32p[TX_32X32] +
+#endif
+ cpi->txfm_count_16x16p[TX_8X8] +
+ cpi->txfm_count_16x16p[TX_16X16]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ pc->prob_tx[2] = get_prob(cpi->txfm_count_32x32p[TX_16X16],
+ cpi->txfm_count_32x32p[TX_16X16] +
+ cpi->txfm_count_32x32p[TX_32X32]);
+#endif
} else {
pc->prob_tx[0] = 128;
pc->prob_tx[1] = 128;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ pc->prob_tx[2] = 128;
+#endif
+ }
+ vp9_write_literal(&header_bc, pc->txfm_mode <= 3 ? pc->txfm_mode : 3, 2);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (pc->txfm_mode > ALLOW_16X16) {
+ vp9_write_bit(&header_bc, pc->txfm_mode == TX_MODE_SELECT);
}
- vp9_write_literal(&header_bc, pc->txfm_mode, 2);
+#endif
if (pc->txfm_mode == TX_MODE_SELECT) {
vp9_write_literal(&header_bc, pc->prob_tx[0], 8);
vp9_write_literal(&header_bc, pc->prob_tx[1], 8);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_write_literal(&header_bc, pc->prob_tx[2], 8);
+#endif
}
}
@@ -2150,6 +2229,10 @@ void vp9_pack_bitstream(VP9_COMP *cpi, unsigned char *dest,
vp9_copy(cpi->common.fc.pre_hybrid_coef_probs_8x8, cpi->common.fc.hybrid_coef_probs_8x8);
vp9_copy(cpi->common.fc.pre_coef_probs_16x16, cpi->common.fc.coef_probs_16x16);
vp9_copy(cpi->common.fc.pre_hybrid_coef_probs_16x16, cpi->common.fc.hybrid_coef_probs_16x16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_copy(cpi->common.fc.pre_coef_probs_32x32,
+ cpi->common.fc.coef_probs_32x32);
+#endif
#if CONFIG_SUPERBLOCKS
vp9_copy(cpi->common.fc.pre_sb_ymode_prob, cpi->common.fc.sb_ymode_prob);
#endif
diff --git a/vp9/encoder/vp9_block.h b/vp9/encoder/vp9_block.h
index 4669d2ed6945dbccff7d75213c748f30c699cf37..82dc5edc1670b227db75def2cca47b73b0a10d87 100644
--- a/vp9/encoder/vp9_block.h
+++ b/vp9/encoder/vp9_block.h
@@ -36,9 +36,15 @@ typedef struct block {
short *zbin;
short *zbin_8x8;
short *zbin_16x16;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ short *zbin_32x32;
+#endif
short *zrun_zbin_boost;
short *zrun_zbin_boost_8x8;
short *zrun_zbin_boost_16x16;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ short *zrun_zbin_boost_32x32;
+#endif
short *round;
// Zbin Over Quant value
@@ -52,6 +58,9 @@ typedef struct block {
int eob_max_offset;
int eob_max_offset_8x8;
int eob_max_offset_16x16;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ int eob_max_offset_32x32;
+#endif
} BLOCK;
typedef struct {
@@ -83,6 +92,13 @@ typedef struct {
int64_t txfm_rd_diff[NB_TXFM_MODES];
} PICK_MODE_CONTEXT;
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+typedef struct superblock {
+ DECLARE_ALIGNED(16, short, src_diff[32*32+16*16*2]);
+ DECLARE_ALIGNED(16, short, coeff[32*32+16*16*2]);
+} SUPERBLOCK;
+#endif
+
typedef struct macroblock {
DECLARE_ALIGNED(16, short, src_diff[400]); // 16x16 Y 8x8 U 8x8 V 4x4 2nd Y
DECLARE_ALIGNED(16, short, coeff[400]); // 16x16 Y 8x8 U 8x8 V 4x4 2nd Y
@@ -95,6 +111,10 @@ typedef struct macroblock {
// 1 DC 2nd order block each with 16 entries
BLOCK block[25];
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+ SUPERBLOCK sb_coeff_data;
+#endif
+
YV12_BUFFER_CONFIG src;
MACROBLOCKD e_mbd;
@@ -153,9 +173,9 @@ typedef struct macroblock {
unsigned char *active_ptr;
- unsigned int token_costs[TX_SIZE_MAX][BLOCK_TYPES][COEF_BANDS]
+ unsigned int token_costs[TX_SIZE_MAX_SB][BLOCK_TYPES][COEF_BANDS]
[PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS];
- unsigned int hybrid_token_costs[TX_SIZE_MAX][BLOCK_TYPES][COEF_BANDS]
+ unsigned int hybrid_token_costs[TX_SIZE_MAX_SB][BLOCK_TYPES][COEF_BANDS]
[PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS];
int optimize;
diff --git a/vp9/encoder/vp9_dct.c b/vp9/encoder/vp9_dct.c
index 6753f2462883d9c9b272efb8387f79b999b5c5c4..0fc8fa35edb416c177395f372ea2151dc6e7ece5 100644
--- a/vp9/encoder/vp9_dct.c
+++ b/vp9/encoder/vp9_dct.c
@@ -1330,3 +1330,461 @@ void vp9_short_fdct16x16_c(int16_t *input, int16_t *out, int pitch) {
#undef RIGHT_SHIFT
#undef ROUNDING
#endif
+
+#if CONFIG_TX32X32
+#if !CONFIG_DWT32X32HYBRID
+static void dct32_1d(double *input, double *output, int stride) {
+ static const double C1 = 0.998795456205; // cos(pi * 1 / 64)
+ static const double C2 = 0.995184726672; // cos(pi * 2 / 64)
+ static const double C3 = 0.989176509965; // cos(pi * 3 / 64)
+ static const double C4 = 0.980785280403; // cos(pi * 4 / 64)
+ static const double C5 = 0.970031253195; // cos(pi * 5 / 64)
+ static const double C6 = 0.956940335732; // cos(pi * 6 / 64)
+ static const double C7 = 0.941544065183; // cos(pi * 7 / 64)
+ static const double C8 = 0.923879532511; // cos(pi * 8 / 64)
+ static const double C9 = 0.903989293123; // cos(pi * 9 / 64)
+ static const double C10 = 0.881921264348; // cos(pi * 10 / 64)
+ static const double C11 = 0.857728610000; // cos(pi * 11 / 64)
+ static const double C12 = 0.831469612303; // cos(pi * 12 / 64)
+ static const double C13 = 0.803207531481; // cos(pi * 13 / 64)
+ static const double C14 = 0.773010453363; // cos(pi * 14 / 64)
+ static const double C15 = 0.740951125355; // cos(pi * 15 / 64)
+ static const double C16 = 0.707106781187; // cos(pi * 16 / 64)
+ static const double C17 = 0.671558954847; // cos(pi * 17 / 64)
+ static const double C18 = 0.634393284164; // cos(pi * 18 / 64)
+ static const double C19 = 0.595699304492; // cos(pi * 19 / 64)
+ static const double C20 = 0.555570233020; // cos(pi * 20 / 64)
+ static const double C21 = 0.514102744193; // cos(pi * 21 / 64)
+ static const double C22 = 0.471396736826; // cos(pi * 22 / 64)
+ static const double C23 = 0.427555093430; // cos(pi * 23 / 64)
+ static const double C24 = 0.382683432365; // cos(pi * 24 / 64)
+ static const double C25 = 0.336889853392; // cos(pi * 25 / 64)
+ static const double C26 = 0.290284677254; // cos(pi * 26 / 64)
+ static const double C27 = 0.242980179903; // cos(pi * 27 / 64)
+ static const double C28 = 0.195090322016; // cos(pi * 28 / 64)
+ static const double C29 = 0.146730474455; // cos(pi * 29 / 64)
+ static const double C30 = 0.098017140330; // cos(pi * 30 / 64)
+ static const double C31 = 0.049067674327; // cos(pi * 31 / 64)
+
+ double step[32];
+
+ // Stage 1
+ step[0] = input[stride*0] + input[stride*(32 - 1)];
+ step[1] = input[stride*1] + input[stride*(32 - 2)];
+ step[2] = input[stride*2] + input[stride*(32 - 3)];
+ step[3] = input[stride*3] + input[stride*(32 - 4)];
+ step[4] = input[stride*4] + input[stride*(32 - 5)];
+ step[5] = input[stride*5] + input[stride*(32 - 6)];
+ step[6] = input[stride*6] + input[stride*(32 - 7)];
+ step[7] = input[stride*7] + input[stride*(32 - 8)];
+ step[8] = input[stride*8] + input[stride*(32 - 9)];
+ step[9] = input[stride*9] + input[stride*(32 - 10)];
+ step[10] = input[stride*10] + input[stride*(32 - 11)];
+ step[11] = input[stride*11] + input[stride*(32 - 12)];
+ step[12] = input[stride*12] + input[stride*(32 - 13)];
+ step[13] = input[stride*13] + input[stride*(32 - 14)];
+ step[14] = input[stride*14] + input[stride*(32 - 15)];
+ step[15] = input[stride*15] + input[stride*(32 - 16)];
+ step[16] = -input[stride*16] + input[stride*(32 - 17)];
+ step[17] = -input[stride*17] + input[stride*(32 - 18)];
+ step[18] = -input[stride*18] + input[stride*(32 - 19)];
+ step[19] = -input[stride*19] + input[stride*(32 - 20)];
+ step[20] = -input[stride*20] + input[stride*(32 - 21)];
+ step[21] = -input[stride*21] + input[stride*(32 - 22)];
+ step[22] = -input[stride*22] + input[stride*(32 - 23)];
+ step[23] = -input[stride*23] + input[stride*(32 - 24)];
+ step[24] = -input[stride*24] + input[stride*(32 - 25)];
+ step[25] = -input[stride*25] + input[stride*(32 - 26)];
+ step[26] = -input[stride*26] + input[stride*(32 - 27)];
+ step[27] = -input[stride*27] + input[stride*(32 - 28)];
+ step[28] = -input[stride*28] + input[stride*(32 - 29)];
+ step[29] = -input[stride*29] + input[stride*(32 - 30)];
+ step[30] = -input[stride*30] + input[stride*(32 - 31)];
+ step[31] = -input[stride*31] + input[stride*(32 - 32)];
+
+ // Stage 2
+ output[stride*0] = step[0] + step[16 - 1];
+ output[stride*1] = step[1] + step[16 - 2];
+ output[stride*2] = step[2] + step[16 - 3];
+ output[stride*3] = step[3] + step[16 - 4];
+ output[stride*4] = step[4] + step[16 - 5];
+ output[stride*5] = step[5] + step[16 - 6];
+ output[stride*6] = step[6] + step[16 - 7];
+ output[stride*7] = step[7] + step[16 - 8];
+ output[stride*8] = -step[8] + step[16 - 9];
+ output[stride*9] = -step[9] + step[16 - 10];
+ output[stride*10] = -step[10] + step[16 - 11];
+ output[stride*11] = -step[11] + step[16 - 12];
+ output[stride*12] = -step[12] + step[16 - 13];
+ output[stride*13] = -step[13] + step[16 - 14];
+ output[stride*14] = -step[14] + step[16 - 15];
+ output[stride*15] = -step[15] + step[16 - 16];
+
+ output[stride*16] = step[16];
+ output[stride*17] = step[17];
+ output[stride*18] = step[18];
+ output[stride*19] = step[19];
+
+ output[stride*20] = (-step[20] + step[27])*C16;
+ output[stride*21] = (-step[21] + step[26])*C16;
+ output[stride*22] = (-step[22] + step[25])*C16;
+ output[stride*23] = (-step[23] + step[24])*C16;
+
+ output[stride*24] = (step[24] + step[23])*C16;
+ output[stride*25] = (step[25] + step[22])*C16;
+ output[stride*26] = (step[26] + step[21])*C16;
+ output[stride*27] = (step[27] + step[20])*C16;
+
+ output[stride*28] = step[28];
+ output[stride*29] = step[29];
+ output[stride*30] = step[30];
+ output[stride*31] = step[31];
+
+ // Stage 3
+ step[0] = output[stride*0] + output[stride*(8 - 1)];
+ step[1] = output[stride*1] + output[stride*(8 - 2)];
+ step[2] = output[stride*2] + output[stride*(8 - 3)];
+ step[3] = output[stride*3] + output[stride*(8 - 4)];
+ step[4] = -output[stride*4] + output[stride*(8 - 5)];
+ step[5] = -output[stride*5] + output[stride*(8 - 6)];
+ step[6] = -output[stride*6] + output[stride*(8 - 7)];
+ step[7] = -output[stride*7] + output[stride*(8 - 8)];
+ step[8] = output[stride*8];
+ step[9] = output[stride*9];
+ step[10] = (-output[stride*10] + output[stride*13])*C16;
+ step[11] = (-output[stride*11] + output[stride*12])*C16;
+ step[12] = (output[stride*12] + output[stride*11])*C16;
+ step[13] = (output[stride*13] + output[stride*10])*C16;
+ step[14] = output[stride*14];
+ step[15] = output[stride*15];
+
+ step[16] = output[stride*16] + output[stride*23];
+ step[17] = output[stride*17] + output[stride*22];
+ step[18] = output[stride*18] + output[stride*21];
+ step[19] = output[stride*19] + output[stride*20];
+ step[20] = -output[stride*20] + output[stride*19];
+ step[21] = -output[stride*21] + output[stride*18];
+ step[22] = -output[stride*22] + output[stride*17];
+ step[23] = -output[stride*23] + output[stride*16];
+ step[24] = -output[stride*24] + output[stride*31];
+ step[25] = -output[stride*25] + output[stride*30];
+ step[26] = -output[stride*26] + output[stride*29];
+ step[27] = -output[stride*27] + output[stride*28];
+ step[28] = output[stride*28] + output[stride*27];
+ step[29] = output[stride*29] + output[stride*26];
+ step[30] = output[stride*30] + output[stride*25];
+ step[31] = output[stride*31] + output[stride*24];
+
+ // Stage 4
+ output[stride*0] = step[0] + step[3];
+ output[stride*1] = step[1] + step[2];
+ output[stride*2] = -step[2] + step[1];
+ output[stride*3] = -step[3] + step[0];
+ output[stride*4] = step[4];
+ output[stride*5] = (-step[5] + step[6])*C16;
+ output[stride*6] = (step[6] + step[5])*C16;
+ output[stride*7] = step[7];
+ output[stride*8] = step[8] + step[11];
+ output[stride*9] = step[9] + step[10];
+ output[stride*10] = -step[10] + step[9];
+ output[stride*11] = -step[11] + step[8];
+ output[stride*12] = -step[12] + step[15];
+ output[stride*13] = -step[13] + step[14];
+ output[stride*14] = step[14] + step[13];
+ output[stride*15] = step[15] + step[12];
+
+ output[stride*16] = step[16];
+ output[stride*17] = step[17];
+ output[stride*18] = step[18]*-C8 + step[29]*C24;
+ output[stride*19] = step[19]*-C8 + step[28]*C24;
+ output[stride*20] = step[20]*-C24 + step[27]*-C8;
+ output[stride*21] = step[21]*-C24 + step[26]*-C8;
+ output[stride*22] = step[22];
+ output[stride*23] = step[23];
+ output[stride*24] = step[24];
+ output[stride*25] = step[25];
+ output[stride*26] = step[26]*C24 + step[21]*-C8;
+ output[stride*27] = step[27]*C24 + step[20]*-C8;
+ output[stride*28] = step[28]*C8 + step[19]*C24;
+ output[stride*29] = step[29]*C8 + step[18]*C24;
+ output[stride*30] = step[30];
+ output[stride*31] = step[31];
+
+ // Stage 5
+ step[0] = (output[stride*0] + output[stride*1]) * C16;
+ step[1] = (-output[stride*1] + output[stride*0]) * C16;
+ step[2] = output[stride*2]*C24 + output[stride*3] * C8;
+ step[3] = output[stride*3]*C24 - output[stride*2] * C8;
+ step[4] = output[stride*4] + output[stride*5];
+ step[5] = -output[stride*5] + output[stride*4];
+ step[6] = -output[stride*6] + output[stride*7];
+ step[7] = output[stride*7] + output[stride*6];
+ step[8] = output[stride*8];
+ step[9] = output[stride*9]*-C8 + output[stride*14]*C24;
+ step[10] = output[stride*10]*-C24 + output[stride*13]*-C8;
+ step[11] = output[stride*11];
+ step[12] = output[stride*12];
+ step[13] = output[stride*13]*C24 + output[stride*10]*-C8;
+ step[14] = output[stride*14]*C8 + output[stride*9]*C24;
+ step[15] = output[stride*15];
+
+ step[16] = output[stride*16] + output[stride*19];
+ step[17] = output[stride*17] + output[stride*18];
+ step[18] = -output[stride*18] + output[stride*17];
+ step[19] = -output[stride*19] + output[stride*16];
+ step[20] = -output[stride*20] + output[stride*23];
+ step[21] = -output[stride*21] + output[stride*22];
+ step[22] = output[stride*22] + output[stride*21];
+ step[23] = output[stride*23] + output[stride*20];
+ step[24] = output[stride*24] + output[stride*27];
+ step[25] = output[stride*25] + output[stride*26];
+ step[26] = -output[stride*26] + output[stride*25];
+ step[27] = -output[stride*27] + output[stride*24];
+ step[28] = -output[stride*28] + output[stride*31];
+ step[29] = -output[stride*29] + output[stride*30];
+ step[30] = output[stride*30] + output[stride*29];
+ step[31] = output[stride*31] + output[stride*28];
+
+ // Stage 6
+ output[stride*0] = step[0];
+ output[stride*1] = step[1];
+ output[stride*2] = step[2];
+ output[stride*3] = step[3];
+ output[stride*4] = step[4]*C28 + step[7]*C4;
+ output[stride*5] = step[5]*C12 + step[6]*C20;
+ output[stride*6] = step[6]*C12 + step[5]*-C20;
+ output[stride*7] = step[7]*C28 + step[4]*-C4;
+ output[stride*8] = step[8] + step[9];
+ output[stride*9] = -step[9] + step[8];
+ output[stride*10] = -step[10] + step[11];
+ output[stride*11] = step[11] + step[10];
+ output[stride*12] = step[12] + step[13];
+ output[stride*13] = -step[13] + step[12];
+ output[stride*14] = -step[14] + step[15];
+ output[stride*15] = step[15] + step[14];
+
+ output[stride*16] = step[16];
+ output[stride*17] = step[17]*-C4 + step[30]*C28;
+ output[stride*18] = step[18]*-C28 + step[29]*-C4;
+ output[stride*19] = step[19];
+ output[stride*20] = step[20];
+ output[stride*21] = step[21]*-C20 + step[26]*C12;
+ output[stride*22] = step[22]*-C12 + step[25]*-C20;
+ output[stride*23] = step[23];
+ output[stride*24] = step[24];
+ output[stride*25] = step[25]*C12 + step[22]*-C20;
+ output[stride*26] = step[26]*C20 + step[21]*C12;
+ output[stride*27] = step[27];
+ output[stride*28] = step[28];
+ output[stride*29] = step[29]*C28 + step[18]*-C4;
+ output[stride*30] = step[30]*C4 + step[17]*C28;
+ output[stride*31] = step[31];
+
+ // Stage 7
+ step[0] = output[stride*0];
+ step[1] = output[stride*1];
+ step[2] = output[stride*2];
+ step[3] = output[stride*3];
+ step[4] = output[stride*4];
+ step[5] = output[stride*5];
+ step[6] = output[stride*6];
+ step[7] = output[stride*7];
+ step[8] = output[stride*8]*C30 + output[stride*15]*C2;
+ step[9] = output[stride*9]*C14 + output[stride*14]*C18;
+ step[10] = output[stride*10]*C22 + output[stride*13]*C10;
+ step[11] = output[stride*11]*C6 + output[stride*12]*C26;
+ step[12] = output[stride*12]*C6 + output[stride*11]*-C26;
+ step[13] = output[stride*13]*C22 + output[stride*10]*-C10;
+ step[14] = output[stride*14]*C14 + output[stride*9]*-C18;
+ step[15] = output[stride*15]*C30 + output[stride*8]*-C2;
+
+ step[16] = output[stride*16] + output[stride*17];
+ step[17] = -output[stride*17] + output[stride*16];
+ step[18] = -output[stride*18] + output[stride*19];
+ step[19] = output[stride*19] + output[stride*18];
+ step[20] = output[stride*20] + output[stride*21];
+ step[21] = -output[stride*21] + output[stride*20];
+ step[22] = -output[stride*22] + output[stride*23];
+ step[23] = output[stride*23] + output[stride*22];
+ step[24] = output[stride*24] + output[stride*25];
+ step[25] = -output[stride*25] + output[stride*24];
+ step[26] = -output[stride*26] + output[stride*27];
+ step[27] = output[stride*27] + output[stride*26];
+ step[28] = output[stride*28] + output[stride*29];
+ step[29] = -output[stride*29] + output[stride*28];
+ step[30] = -output[stride*30] + output[stride*31];
+ step[31] = output[stride*31] + output[stride*30];
+
+ // Final stage --- outputs indices are bit-reversed.
+ output[stride*0] = step[0];
+ output[stride*16] = step[1];
+ output[stride*8] = step[2];
+ output[stride*24] = step[3];
+ output[stride*4] = step[4];
+ output[stride*20] = step[5];
+ output[stride*12] = step[6];
+ output[stride*28] = step[7];
+ output[stride*2] = step[8];
+ output[stride*18] = step[9];
+ output[stride*10] = step[10];
+ output[stride*26] = step[11];
+ output[stride*6] = step[12];
+ output[stride*22] = step[13];
+ output[stride*14] = step[14];
+ output[stride*30] = step[15];
+
+ output[stride*1] = step[16]*C31 + step[31]*C1;
+ output[stride*17] = step[17]*C15 + step[30]*C17;
+ output[stride*9] = step[18]*C23 + step[29]*C9;
+ output[stride*25] = step[19]*C7 + step[28]*C25;
+ output[stride*5] = step[20]*C27 + step[27]*C5;
+ output[stride*21] = step[21]*C11 + step[26]*C21;
+ output[stride*13] = step[22]*C19 + step[25]*C13;
+ output[stride*29] = step[23]*C3 + step[24]*C29;
+ output[stride*3] = step[24]*C3 + step[23]*-C29;
+ output[stride*19] = step[25]*C19 + step[22]*-C13;
+ output[stride*11] = step[26]*C11 + step[21]*-C21;
+ output[stride*27] = step[27]*C27 + step[20]*-C5;
+ output[stride*7] = step[28]*C7 + step[19]*-C25;
+ output[stride*23] = step[29]*C23 + step[18]*-C9;
+ output[stride*15] = step[30]*C15 + step[17]*-C17;
+ output[stride*31] = step[31]*C31 + step[16]*-C1;
+}
+
+void vp9_short_fdct32x32_c(int16_t *input, int16_t *out, int pitch) {
+ vp9_clear_system_state(); // Make it simd safe : __asm emms;
+ {
+ int shortpitch = pitch >> 1;
+ int i, j;
+ double output[1024];
+ // First transform columns
+ for (i = 0; i < 32; i++) {
+ double temp_in[32], temp_out[32];
+ for (j = 0; j < 32; j++)
+ temp_in[j] = input[j*shortpitch + i];
+ dct32_1d(temp_in, temp_out, 1);
+ for (j = 0; j < 32; j++)
+ output[j*32 + i] = temp_out[j];
+ }
+ // Then transform rows
+ for (i = 0; i < 32; ++i) {
+ double temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j)
+ temp_in[j] = output[j + i*32];
+ dct32_1d(temp_in, temp_out, 1);
+ for (j = 0; j < 32; ++j)
+ output[j + i*32] = temp_out[j];
+ }
+ // Scale by some magic number
+ for (i = 0; i < 1024; i++) {
+ out[i] = (short)round(output[i]/4);
+ }
+ }
+
+ vp9_clear_system_state(); // Make it simd safe : __asm emms;
+}
+
+#else // CONFIG_DWT32X32HYBRID
+
+#define MAX_BLOCK_LENGTH 64
+#define ENH_PRECISION_BITS 1
+#define ENH_PRECISION_RND ((1 << ENH_PRECISION_BITS) / 2)
+
+// Note: block length must be even for this implementation
+static void analysis_53_row(int length, short *x,
+ short *lowpass, short *highpass) {
+ int n;
+ short r, * a, * b;
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ while (--n) {
+ *a++ = (r = *x++) << 1;
+ *b++ = *x - ((r + x[1] + 1) >> 1);
+ x++;
+ }
+ *a = (r = *x++) << 1;
+ *b = *x - r;
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ r = *highpass;
+ while (n--) {
+ *a++ += (r + (*b) + 1) >> 1;
+ r = *b++;
+ }
+}
+
+static void analysis_53_col(int length, short *x,
+ short *lowpass, short *highpass) {
+ int n;
+ short r, * a, * b;
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ while (--n) {
+ *a++ = (r = *x++);
+ *b++ = (((*x) << 1) - (r + x[1]) + 2) >> 2;
+ x++;
+ }
+ *a = (r = *x++);
+ *b = (*x - r + 1) >> 1;
+
+ n = length >> 1;
+ b = highpass;
+ a = lowpass;
+ r = *highpass;
+ while (n--) {
+ *a++ += (r + (*b) + 1) >> 1;
+ r = *b++;
+ }
+}
+
+// NOTE: Using a 5/3 integer wavelet for now. Explore using a wavelet
+// with a better response later
+static void dyadic_analyze(int levels, int width, int height,
+ short *x, int pitch_x, short *c, int pitch_c) {
+ int lv, i, j, nh, nw, hh = height, hw = width;
+ short buffer[2 * MAX_BLOCK_LENGTH];
+ for (i = 0; i < height; i++) {
+ for (j = 0; j < width; j++) {
+ c[i * pitch_c + j] = x[i * pitch_x + j] << ENH_PRECISION_BITS;
+ }
+ }
+ for (lv = 0; lv < levels; lv++) {
+ nh = hh;
+ hh = (hh + 1) >> 1;
+ nw = hw;
+ hw = (hw + 1) >> 1;
+ if ((nh < 2) || (nw < 2)) return;
+ for (i = 0; i < nh; i++) {
+ memcpy(buffer, &c[i * pitch_c], nw * sizeof(short));
+ analysis_53_row(nw, buffer, &c[i * pitch_c], &c[i * pitch_c] + hw);
+ }
+ for (j = 0; j < nw; j++) {
+ for (i = 0; i < nh; i++)
+ buffer[i + nh] = c[i * pitch_c + j];
+ analysis_53_col(nh, buffer + nh, buffer, buffer + hh);
+ for (i = 0; i < nh; i++)
+ c[i * pitch_c + j] = buffer[i];
+ }
+ }
+}
+
+void vp9_short_fdct32x32_c(short *input, short *out, int pitch) {
+ // assume out is a 32x32 buffer
+ short buffer[16 * 16];
+ int i;
+ const int short_pitch = pitch >> 1;
+ dyadic_analyze(1, 32, 32, input, short_pitch, out, 32);
+ // TODO(debargha): Implement more efficiently by adding output pitch
+ // argument to the dct16x16 function
+ vp9_short_fdct16x16_c(out, buffer, 64);
+ for (i = 0; i < 16; ++i)
+ vpx_memcpy(out + i * 32, buffer + i * 16, sizeof(short) * 16);
+}
+#endif // CONFIG_DWT32X32HYBRID
+#endif // CONFIG_TX32X32
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index 6ab23cae0b2f43480dea341bd2e1ea362e5b1703..f504fc53c73bd47be8e252bb7ff6365291160d10 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -456,6 +456,10 @@ static void update_state(VP9_COMP *cpi, MACROBLOCK *x,
if (xd->mb_to_right_edge >= 0)
vpx_memcpy(xd->mode_info_context + mis + 1, mi, sizeof(MODE_INFO));
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ } else {
+ ctx->txfm_rd_diff[ALLOW_32X32] = ctx->txfm_rd_diff[ALLOW_16X16];
+#endif
}
#endif
@@ -1487,6 +1491,9 @@ static void encode_frame_internal(VP9_COMP *cpi) {
vp9_zero(cpi->hybrid_coef_counts_8x8);
vp9_zero(cpi->coef_counts_16x16);
vp9_zero(cpi->hybrid_coef_counts_16x16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_zero(cpi->coef_counts_32x32);
+#endif
vp9_frame_init_quantizer(cpi);
@@ -1507,7 +1514,8 @@ static void encode_frame_internal(VP9_COMP *cpi) {
vpx_memset(cpi->rd_comp_pred_diff, 0, sizeof(cpi->rd_comp_pred_diff));
vpx_memset(cpi->single_pred_count, 0, sizeof(cpi->single_pred_count));
vpx_memset(cpi->comp_pred_count, 0, sizeof(cpi->comp_pred_count));
- vpx_memset(cpi->txfm_count, 0, sizeof(cpi->txfm_count));
+ vpx_memset(cpi->txfm_count_32x32p, 0, sizeof(cpi->txfm_count_32x32p));
+ vpx_memset(cpi->txfm_count_16x16p, 0, sizeof(cpi->txfm_count_16x16p));
vpx_memset(cpi->txfm_count_8x8p, 0, sizeof(cpi->txfm_count_8x8p));
vpx_memset(cpi->rd_tx_select_diff, 0, sizeof(cpi->rd_tx_select_diff));
{
@@ -1700,7 +1708,11 @@ void vp9_encode_frame(VP9_COMP *cpi) {
* keyframe's probabilities as an estimate of what the current keyframe's
* coefficient cost distributions may look like. */
if (frame_type == 0) {
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ txfm_type = ALLOW_32X32;
+#else
txfm_type = ALLOW_16X16;
+#endif
} else
#if 0
/* FIXME (rbultje)
@@ -1731,9 +1743,15 @@ void vp9_encode_frame(VP9_COMP *cpi) {
} else
txfm_type = ALLOW_8X8;
#else
- txfm_type = cpi->rd_tx_select_threshes[frame_type][ALLOW_16X16] >=
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ txfm_type = cpi->rd_tx_select_threshes[frame_type][ALLOW_32X32] >=
cpi->rd_tx_select_threshes[frame_type][TX_MODE_SELECT] ?
+ ALLOW_32X32 : TX_MODE_SELECT;
+#else
+ txfm_type = cpi->rd_tx_select_threshes[frame_type][ALLOW_16X16] >=
+ cpi->rd_tx_select_threshes[frame_type][TX_MODE_SELECT] ?
ALLOW_16X16 : TX_MODE_SELECT;
+#endif
#endif
cpi->common.txfm_mode = txfm_type;
if (txfm_type != TX_MODE_SELECT) {
@@ -1753,7 +1771,8 @@ void vp9_encode_frame(VP9_COMP *cpi) {
int64_t pd = cpi->rd_tx_select_diff[i];
int diff;
if (i == TX_MODE_SELECT)
- pd -= RDCOST(cpi->mb.rdmult, cpi->mb.rddiv, 2048 * (TX_SIZE_MAX - 1), 0);
+ pd -= RDCOST(cpi->mb.rdmult, cpi->mb.rddiv,
+ 2048 * (TX_SIZE_MAX_SB - 1), 0);
diff = (int)(pd / cpi->common.MBs);
cpi->rd_tx_select_threshes[frame_type][i] += diff;
cpi->rd_tx_select_threshes[frame_type][i] /= 2;
@@ -1776,19 +1795,37 @@ void vp9_encode_frame(VP9_COMP *cpi) {
}
if (cpi->common.txfm_mode == TX_MODE_SELECT) {
- const int count4x4 = cpi->txfm_count[TX_4X4] + cpi->txfm_count_8x8p[TX_4X4];
- const int count8x8 = cpi->txfm_count[TX_8X8];
+ const int count4x4 = cpi->txfm_count_16x16p[TX_4X4] +
+ cpi->txfm_count_32x32p[TX_4X4] +
+ cpi->txfm_count_8x8p[TX_4X4];
+ const int count8x8_lp = cpi->txfm_count_32x32p[TX_8X8] +
+ cpi->txfm_count_16x16p[TX_8X8];
const int count8x8_8x8p = cpi->txfm_count_8x8p[TX_8X8];
- const int count16x16 = cpi->txfm_count[TX_16X16];
+ const int count16x16_16x16p = cpi->txfm_count_16x16p[TX_16X16];
+ const int count16x16_lp = cpi->txfm_count_32x32p[TX_16X16];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ const int count32x32 = cpi->txfm_count_32x32p[TX_32X32];
+#else
+ const int count32x32 = 0;
+#endif
- if (count4x4 == 0 && count16x16 == 0) {
+ if (count4x4 == 0 && count16x16_lp == 0 && count16x16_16x16p == 0 &&
+ count32x32 == 0) {
cpi->common.txfm_mode = ALLOW_8X8;
reset_skip_txfm_size(cpi, TX_8X8);
- } else if (count8x8 == 0 && count16x16 == 0 && count8x8_8x8p == 0) {
+ } else if (count8x8_8x8p == 0 && count16x16_16x16p == 0 &&
+ count8x8_lp == 0 && count16x16_lp == 0 && count32x32 == 0) {
cpi->common.txfm_mode = ONLY_4X4;
reset_skip_txfm_size(cpi, TX_4X4);
- } else if (count8x8 == 0 && count4x4 == 0) {
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ } else if (count8x8_lp == 0 && count16x16_lp == 0 && count4x4 == 0) {
+ cpi->common.txfm_mode = ALLOW_32X32;
+#endif
+ } else if (count32x32 == 0 && count8x8_lp == 0 && count4x4 == 0) {
cpi->common.txfm_mode = ALLOW_16X16;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ reset_skip_txfm_size(cpi, TX_16X16);
+#endif
}
}
} else {
@@ -2087,6 +2124,7 @@ static void encode_macroblock(VP9_COMP *cpi, MACROBLOCK *x,
vp9_set_pred_flag(xd, PRED_REF, ref_pred_flag);
}
+ assert(mbmi->txfm_size <= TX_16X16);
if (mbmi->ref_frame == INTRA_FRAME) {
#ifdef ENC_DEBUG
if (enc_debug) {
@@ -2266,7 +2304,7 @@ static void encode_macroblock(VP9_COMP *cpi, MACROBLOCK *x,
vp9_get_segdata(&x->e_mbd, segment_id, SEG_LVL_EOB) == 0))) {
if (mbmi->mode != B_PRED && mbmi->mode != I8X8_PRED &&
mbmi->mode != SPLITMV) {
- cpi->txfm_count[mbmi->txfm_size]++;
+ cpi->txfm_count_16x16p[mbmi->txfm_size]++;
} else if (mbmi->mode == I8X8_PRED ||
(mbmi->mode == SPLITMV &&
mbmi->partitioning != PARTITIONING_4X4)) {
@@ -2308,6 +2346,7 @@ static void encode_superblock(VP9_COMP *cpi, MACROBLOCK *x,
MODE_INFO *mi = x->e_mbd.mode_info_context;
unsigned int segment_id = mi->mbmi.segment_id;
ENTROPY_CONTEXT_PLANES ta[4], tl[4];
+ const int mis = cm->mode_info_stride;
x->skip = 0;
@@ -2397,6 +2436,53 @@ static void encode_superblock(VP9_COMP *cpi, MACROBLOCK *x,
xd->dst.y_stride, xd->dst.uv_stride);
}
+#if CONFIG_TX32X32
+ if (xd->mode_info_context->mbmi.txfm_size == TX_32X32) {
+ vp9_subtract_sby_s_c(x->sb_coeff_data.src_diff, src, src_y_stride,
+ dst, dst_y_stride);
+ vp9_subtract_sbuv_s_c(x->sb_coeff_data.src_diff,
+ usrc, vsrc, src_uv_stride,
+ udst, vdst, dst_uv_stride);
+ vp9_transform_sby_32x32(x);
+ vp9_transform_sbuv_16x16(x);
+ vp9_quantize_sby_32x32(x);
+ vp9_quantize_sbuv_16x16(x);
+ // TODO(rbultje): trellis optimize
+ vp9_inverse_transform_sbuv_16x16(&x->e_mbd.sb_coeff_data);
+ vp9_inverse_transform_sby_32x32(&x->e_mbd.sb_coeff_data);
+ vp9_recon_sby_s_c(&x->e_mbd, dst);
+ vp9_recon_sbuv_s_c(&x->e_mbd, udst, vdst);
+
+ if (!x->skip) {
+ vp9_tokenize_sb(cpi, &x->e_mbd, t, 0);
+ } else {
+ int mb_skip_context =
+ cpi->common.mb_no_coeff_skip ?
+ (mi - 1)->mbmi.mb_skip_coeff +
+ (mi - mis)->mbmi.mb_skip_coeff :
+ 0;
+ mi->mbmi.mb_skip_coeff = 1;
+ if (cm->mb_no_coeff_skip) {
+ cpi->skip_true_count[mb_skip_context]++;
+ vp9_fix_contexts_sb(xd);
+ } else {
+ vp9_stuff_sb(cpi, xd, t, 0);
+ cpi->skip_false_count[mb_skip_context]++;
+ }
+ }
+
+ // copy skip flag on all mb_mode_info contexts in this SB
+ // if this was a skip at this txfm size
+ if (mb_col < cm->mb_cols - 1)
+ mi[1].mbmi.mb_skip_coeff = mi->mbmi.mb_skip_coeff;
+ if (mb_row < cm->mb_rows - 1) {
+ mi[mis].mbmi.mb_skip_coeff = mi->mbmi.mb_skip_coeff;
+ if (mb_col < cm->mb_cols - 1)
+ mi[mis + 1].mbmi.mb_skip_coeff = mi->mbmi.mb_skip_coeff;
+ }
+ skip[0] = skip[2] = skip[1] = skip[3] = mi->mbmi.mb_skip_coeff;
+ } else {
+#endif
for (n = 0; n < 4; n++) {
int x_idx = n & 1, y_idx = n >> 1;
@@ -2405,7 +2491,7 @@ static void encode_superblock(VP9_COMP *cpi, MACROBLOCK *x,
memcpy(&ta[n], xd->above_context, sizeof(ta[n]));
memcpy(&tl[n], xd->left_context, sizeof(tl[n]));
tp[n] = *t;
- xd->mode_info_context = mi + x_idx + y_idx * cm->mode_info_stride;
+ xd->mode_info_context = mi + x_idx + y_idx * mis;
vp9_subtract_mby_s_c(x->src_diff,
src + x_idx * 16 + y_idx * 16 * src_y_stride,
@@ -2433,7 +2519,7 @@ static void encode_superblock(VP9_COMP *cpi, MACROBLOCK *x,
int mb_skip_context =
cpi->common.mb_no_coeff_skip ?
(x->e_mbd.mode_info_context - 1)->mbmi.mb_skip_coeff +
- (x->e_mbd.mode_info_context - cpi->common.mode_info_stride)->mbmi.mb_skip_coeff :
+ (x->e_mbd.mode_info_context - mis)->mbmi.mb_skip_coeff :
0;
xd->mode_info_context->mbmi.mb_skip_coeff = skip[n] = 1;
if (cpi->common.mb_no_coeff_skip) {
@@ -2450,20 +2536,29 @@ static void encode_superblock(VP9_COMP *cpi, MACROBLOCK *x,
xd->mode_info_context = mi;
update_sb_skip_coeff_state(cpi, x, ta, tl, tp, t, skip);
+#if CONFIG_TX32X32
+ }
+#endif
if (cm->txfm_mode == TX_MODE_SELECT &&
!((cm->mb_no_coeff_skip && skip[0] && skip[1] && skip[2] && skip[3]) ||
(vp9_segfeature_active(xd, segment_id, SEG_LVL_EOB) &&
vp9_get_segdata(xd, segment_id, SEG_LVL_EOB) == 0))) {
- cpi->txfm_count[mi->mbmi.txfm_size]++;
+ cpi->txfm_count_32x32p[mi->mbmi.txfm_size]++;
} else {
- TX_SIZE sz = (cm->txfm_mode == TX_MODE_SELECT) ? TX_16X16 : cm->txfm_mode;
+ TX_SIZE sz = (cm->txfm_mode == TX_MODE_SELECT) ?
+#if CONFIG_TX32X32
+ TX_32X32 :
+#else
+ TX_16X16 :
+#endif
+ cm->txfm_mode;
mi->mbmi.txfm_size = sz;
if (mb_col < cm->mb_cols - 1)
mi[1].mbmi.txfm_size = sz;
if (mb_row < cm->mb_rows - 1) {
- mi[cm->mode_info_stride].mbmi.txfm_size = sz;
+ mi[mis].mbmi.txfm_size = sz;
if (mb_col < cm->mb_cols - 1)
- mi[cm->mode_info_stride + 1].mbmi.txfm_size = sz;
+ mi[mis + 1].mbmi.txfm_size = sz;
}
}
}
diff --git a/vp9/encoder/vp9_encodemb.c b/vp9/encoder/vp9_encodemb.c
index 091f2f0fe9b09053cfa53e2a3f603938ababc6d8..46087c28e53e41ece34856bad43c3e64a416c1d3 100644
--- a/vp9/encoder/vp9_encodemb.c
+++ b/vp9/encoder/vp9_encodemb.c
@@ -108,6 +108,52 @@ void vp9_subtract_mby_s_c(short *diff, const unsigned char *src, int src_stride,
}
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+void vp9_subtract_sby_s_c(short *diff, const unsigned char *src, int src_stride,
+ const unsigned char *pred, int dst_stride) {
+ int r, c;
+
+ for (r = 0; r < 32; r++) {
+ for (c = 0; c < 32; c++) {
+ diff[c] = src[c] - pred[c];
+ }
+
+ diff += 32;
+ pred += dst_stride;
+ src += src_stride;
+ }
+}
+
+void vp9_subtract_sbuv_s_c(short *diff, const unsigned char *usrc,
+ const unsigned char *vsrc, int src_stride,
+ const unsigned char *upred,
+ const unsigned char *vpred, int dst_stride) {
+ short *udiff = diff + 1024;
+ short *vdiff = diff + 1024 + 256;
+ int r, c;
+
+ for (r = 0; r < 16; r++) {
+ for (c = 0; c < 16; c++) {
+ udiff[c] = usrc[c] - upred[c];
+ }
+
+ udiff += 16;
+ upred += dst_stride;
+ usrc += src_stride;
+ }
+
+ for (r = 0; r < 16; r++) {
+ for (c = 0; c < 16; c++) {
+ vdiff[c] = vsrc[c] - vpred[c];
+ }
+
+ vdiff += 16;
+ vpred += dst_stride;
+ vsrc += src_stride;
+ }
+}
+#endif
+
void vp9_subtract_mby_c(short *diff, unsigned char *src,
unsigned char *pred, int stride) {
vp9_subtract_mby_s_c(diff, src, stride, pred, 16);
@@ -265,6 +311,22 @@ void vp9_transform_mb_16x16(MACROBLOCK *x) {
vp9_transform_mbuv_8x8(x);
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+void vp9_transform_sby_32x32(MACROBLOCK *x) {
+ SUPERBLOCK * const x_sb = &x->sb_coeff_data;
+ vp9_short_fdct32x32(x_sb->src_diff, x_sb->coeff, 64);
+}
+
+void vp9_transform_sbuv_16x16(MACROBLOCK *x) {
+ SUPERBLOCK * const x_sb = &x->sb_coeff_data;
+ vp9_clear_system_state();
+ x->vp9_short_fdct16x16(x_sb->src_diff + 1024,
+ x_sb->coeff + 1024, 32);
+ x->vp9_short_fdct16x16(x_sb->src_diff + 1280,
+ x_sb->coeff + 1280, 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 vp9_token_state vp9_token_state;
diff --git a/vp9/encoder/vp9_encodemb.h b/vp9/encoder/vp9_encodemb.h
index 4f49647a2c76127c52f14ddaaca513eef423ac4a..3c0a0a5a2376d711ae4f02fcbcb5e35a247b178f 100644
--- a/vp9/encoder/vp9_encodemb.h
+++ b/vp9/encoder/vp9_encodemb.h
@@ -47,6 +47,11 @@ void vp9_transform_mb_16x16(MACROBLOCK *mb);
void vp9_transform_mby_16x16(MACROBLOCK *x);
void vp9_optimize_mby_16x16(MACROBLOCK *x);
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+void vp9_transform_sby_32x32(MACROBLOCK *x);
+void vp9_transform_sbuv_16x16(MACROBLOCK *x);
+#endif
+
void vp9_fidct_mb(MACROBLOCK *x);
void vp9_subtract_4b_c(BLOCK *be, BLOCKD *bd, int pitch);
@@ -59,6 +64,14 @@ void vp9_subtract_mbuv_s_c(short *diff, const unsigned char *usrc,
void vp9_subtract_mby_s_c(short *diff, const unsigned char *src,
int src_stride, const unsigned char *pred,
int dst_stride);
+#if CONFIG_TX32X32
+void vp9_subtract_sby_s_c(short *diff, const unsigned char *src, int src_stride,
+ const unsigned char *pred, int dst_stride);
+void vp9_subtract_sbuv_s_c(short *diff, const unsigned char *usrc,
+ const unsigned char *vsrc, int src_stride,
+ const unsigned char *upred,
+ const unsigned char *vpred, int dst_stride);
+#endif
#endif
#endif
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index 0fe7a14b6e1852a3efeabdb14167b35aa3496e0e..779534bac1b1d9ac5399fbda05729ac16beab90a 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -1810,7 +1810,7 @@ VP9_PTR vp9_create_compressor(VP9_CONFIG *oxcf) {
#endif
for (i = 0; i < COMP_PRED_CONTEXTS; i++)
cm->prob_comppred[i] = 128;
- for (i = 0; i < TX_SIZE_MAX - 1; i++)
+ for (i = 0; i < TX_SIZE_MAX_SB - 1; i++)
cm->prob_tx[i] = 128;
// Prime the recent reference frame useage counters.
@@ -3698,6 +3698,9 @@ static void encode_frame_to_data_rate
vp9_copy(cpi->common.fc.coef_counts_16x16, cpi->coef_counts_16x16);
vp9_copy(cpi->common.fc.hybrid_coef_counts_16x16,
cpi->hybrid_coef_counts_16x16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_copy(cpi->common.fc.coef_counts_32x32, cpi->coef_counts_32x32);
+#endif
vp9_adapt_coef_probs(&cpi->common);
if (cpi->common.frame_type != KEY_FRAME) {
#if CONFIG_SUPERBLOCKS
diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index 511e62f1c85776012cef5da467438502977f9465..28acc96d47a2a7a09446ef978db6707a66be2a2b 100644
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -109,6 +109,11 @@ typedef struct {
vp9_prob hybrid_coef_probs_16x16[BLOCK_TYPES_16X16]
[COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_prob coef_probs_32x32[BLOCK_TYPES_32X32]
+ [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES];
+#endif
+
#if CONFIG_SUPERBLOCKS
vp9_prob sb_ymode_prob[VP9_I32X32_MODES - 1];
#endif
@@ -435,6 +440,15 @@ typedef struct VP9_COMP {
DECLARE_ALIGNED(16, short, zrun_zbin_boost_y2_16x16[QINDEX_RANGE][256]);
DECLARE_ALIGNED(16, short, zrun_zbin_boost_uv_16x16[QINDEX_RANGE][256]);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ DECLARE_ALIGNED(16, short, Y1zbin_32x32[QINDEX_RANGE][1024]);
+ DECLARE_ALIGNED(16, short, Y2zbin_32x32[QINDEX_RANGE][1024]);
+ DECLARE_ALIGNED(16, short, UVzbin_32x32[QINDEX_RANGE][1024]);
+ DECLARE_ALIGNED(16, short, zrun_zbin_boost_y1_32x32[QINDEX_RANGE][1024]);
+ DECLARE_ALIGNED(16, short, zrun_zbin_boost_y2_32x32[QINDEX_RANGE][1024]);
+ DECLARE_ALIGNED(16, short, zrun_zbin_boost_uv_32x32[QINDEX_RANGE][1024]);
+#endif
+
MACROBLOCK mb;
VP9_COMMON common;
VP9_CONFIG oxcf;
@@ -483,8 +497,9 @@ typedef struct VP9_COMP {
int comp_pred_count[COMP_PRED_CONTEXTS];
int single_pred_count[COMP_PRED_CONTEXTS];
// FIXME contextualize
- int txfm_count[TX_SIZE_MAX];
- int txfm_count_8x8p[TX_SIZE_MAX - 1];
+ int txfm_count_32x32p[TX_SIZE_MAX_SB];
+ int txfm_count_16x16p[TX_SIZE_MAX_MB];
+ int txfm_count_8x8p[TX_SIZE_MAX_MB - 1];
int64_t rd_tx_select_diff[NB_TXFM_MODES];
int rd_tx_select_threshes[4][NB_TXFM_MODES];
@@ -604,6 +619,12 @@ typedef struct VP9_COMP {
vp9_prob frame_hybrid_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
unsigned int frame_hybrid_branch_ct_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2];
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+ unsigned int coef_counts_32x32 [BLOCK_TYPES_32X32] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */
+ vp9_prob frame_coef_probs_32x32 [BLOCK_TYPES_32X32] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
+ unsigned int frame_branch_ct_32x32 [BLOCK_TYPES_32X32] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2];
+#endif
+
int gfu_boost;
int last_boost;
int kf_boost;
diff --git a/vp9/encoder/vp9_quantize.c b/vp9/encoder/vp9_quantize.c
index f160edb565e112907345c9321e99b938e9de9587..fcc7d2948b00d011f6fe702e87d8e3d505159f0a 100644
--- a/vp9/encoder/vp9_quantize.c
+++ b/vp9/encoder/vp9_quantize.c
@@ -323,28 +323,25 @@ void vp9_quantize_mb_16x16(MACROBLOCK *x) {
vp9_quantize_mbuv_8x8(x);
}
-void vp9_regular_quantize_b_16x16(BLOCK *b, BLOCKD *d) {
+static void quantize(short *zbin_boost_orig_ptr,
+ short *coeff_ptr, int n_coeffs, int max_coeffs,
+ short *zbin_ptr, short *round_ptr, short *quant_ptr,
+ unsigned char *quant_shift_ptr,
+ short *qcoeff_ptr, short *dqcoeff_ptr,
+ short *dequant_ptr, short zbin_oq_value,
+ int *eob_ptr, const int *scan, int mul) {
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;
+ short *zbin_boost_ptr = zbin_boost_orig_ptr;
- vpx_memset(qcoeff_ptr, 0, 256*sizeof(short));
- vpx_memset(dqcoeff_ptr, 0, 256*sizeof(short));
+ vpx_memset(qcoeff_ptr, 0, n_coeffs*sizeof(short));
+ vpx_memset(dqcoeff_ptr, 0, n_coeffs*sizeof(short));
eob = -1;
- for (i = 0; i < b->eob_max_offset_16x16; i++) {
- rc = vp9_default_zig_zag1d_16x16[i];
- z = coeff_ptr[rc];
+ for (i = 0; i < max_coeffs; i++) {
+ rc = scan[i];
+ z = coeff_ptr[rc] * mul;
zbin = (zbin_ptr[rc!=0] + *zbin_boost_ptr + zbin_oq_value);
zbin_boost_ptr ++;
@@ -354,22 +351,70 @@ void vp9_regular_quantize_b_16x16(BLOCK *b, BLOCKD *d) {
if (x >= zbin) {
x += (round_ptr[rc!=0]);
- y = ((int)(((int)(x * quant_ptr[rc!=0]) >> 16) + x))
+ 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
+ dqcoeff_ptr[rc] = x * dequant_ptr[rc != 0] / mul; // dequantized value
if (y) {
eob = i; // last nonzero coeffs
- zbin_boost_ptr = b->zrun_zbin_boost_16x16;
+ zbin_boost_ptr = zbin_boost_orig_ptr;
}
}
}
- d->eob = eob + 1;
+ *eob_ptr = eob + 1;
+}
+
+void vp9_regular_quantize_b_16x16(BLOCK *b, BLOCKD *d) {
+ quantize(b->zrun_zbin_boost_16x16,
+ b->coeff,
+ 256, b->eob_max_offset_16x16,
+ b->zbin_16x16, b->round, b->quant, b->quant_shift,
+ d->qcoeff,
+ d->dqcoeff,
+ d->dequant,
+ b->zbin_extra,
+ &d->eob, vp9_default_zig_zag1d_16x16, 1);
+}
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+void vp9_quantize_sby_32x32(MACROBLOCK *x) {
+ x->e_mbd.block[0].eob = 0;
+ quantize(x->block[0].zrun_zbin_boost_32x32,
+ x->sb_coeff_data.coeff,
+ 1024, x->block[0].eob_max_offset_32x32,
+ x->block[0].zbin_32x32,
+ x->block[0].round, x->block[0].quant, x->block[0].quant_shift,
+ x->e_mbd.sb_coeff_data.qcoeff,
+ x->e_mbd.sb_coeff_data.dqcoeff,
+ x->e_mbd.block[0].dequant,
+ x->block[0].zbin_extra,
+ &x->e_mbd.block[0].eob,
+ vp9_default_zig_zag1d_32x32, 2);
}
+void vp9_quantize_sbuv_16x16(MACROBLOCK *x) {
+ int i;
+
+ x->e_mbd.block[16].eob = 0;
+ x->e_mbd.block[20].eob = 0;
+ for (i = 16; i < 24; i += 4)
+ quantize(x->block[i].zrun_zbin_boost_16x16,
+ x->sb_coeff_data.coeff + 1024 + (i - 16) * 64,
+ 256, x->block[i].eob_max_offset_16x16,
+ x->block[i].zbin_16x16,
+ x->block[i].round, x->block[0].quant, x->block[i].quant_shift,
+ x->e_mbd.sb_coeff_data.qcoeff + 1024 + (i - 16) * 64,
+ x->e_mbd.sb_coeff_data.dqcoeff + 1024 + (i - 16) * 64,
+ x->e_mbd.block[i].dequant,
+ x->block[i].zbin_extra,
+ &x->e_mbd.block[i].eob,
+ vp9_default_zig_zag1d_16x16, 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
@@ -427,6 +472,74 @@ void vp9_init_quantizer(VP9_COMP *cpi) {
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_TX32X32 && CONFIG_SUPERBLOCKS
+ static const int zbin_boost_32x32[1024] = {
+ 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,
+ 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, 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, 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, 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;
@@ -454,7 +567,13 @@ void vp9_init_quantizer(VP9_COMP *cpi) {
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;
- cpi->zrun_zbin_boost_y1_16x16[Q][0] = ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
+ cpi->zrun_zbin_boost_y1_16x16[Q][0] =
+ ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ cpi->Y1zbin_32x32[Q][0] = ((qzbin_factor * quant_val) + 64) >> 7;
+ cpi->zrun_zbin_boost_y1_32x32[Q][0] =
+ ((quant_val * zbin_boost_32x32[0]) + 64) >> 7;
+#endif
quant_val = vp9_dc2quant(Q, cpi->common.y2dc_delta_q);
@@ -468,7 +587,8 @@ void vp9_init_quantizer(VP9_COMP *cpi) {
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;
- cpi->zrun_zbin_boost_y2_16x16[Q][0] = ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
+ cpi->zrun_zbin_boost_y2_16x16[Q][0] =
+ ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
quant_val = vp9_dc_uv_quant(Q, cpi->common.uvdc_delta_q);
invert_quant(cpi->UVquant[Q] + 0,
@@ -481,7 +601,8 @@ void vp9_init_quantizer(VP9_COMP *cpi) {
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;
- cpi->zrun_zbin_boost_uv_16x16[Q][0] = ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
+ cpi->zrun_zbin_boost_uv_16x16[Q][0] =
+ ((quant_val * zbin_boost_16x16[0]) + 64) >> 7;
// all the 4x4 ac values =;
for (i = 1; i < 16; i++) {
@@ -543,16 +664,30 @@ void vp9_init_quantizer(VP9_COMP *cpi) {
quant_val = vp9_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;
+ cpi->zrun_zbin_boost_y1_16x16[Q][i] =
+ ((quant_val * zbin_boost_16x16[i]) + 64) >> 7;
quant_val = vp9_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;
+ cpi->zrun_zbin_boost_y2_16x16[Q][i] =
+ ((quant_val * zbin_boost_16x16[i]) + 64) >> 7;
quant_val = vp9_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;
+ cpi->zrun_zbin_boost_uv_16x16[Q][i] =
+ ((quant_val * zbin_boost_16x16[i]) + 64) >> 7;
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ // 32x32 structures. Same comment above applies.
+ for (i = 1; i < 1024; i++) {
+ int rc = vp9_default_zig_zag1d_32x32[i];
+
+ quant_val = vp9_ac_yquant(Q);
+ cpi->Y1zbin_32x32[Q][rc] = ((qzbin_factor * quant_val) + 64) >> 7;
+ cpi->zrun_zbin_boost_y1_32x32[Q][i] =
+ ((quant_val * zbin_boost_32x32[i]) + 64) >> 7;
+ }
+#endif
}
}
@@ -592,11 +727,17 @@ void vp9_mb_init_quantizer(VP9_COMP *cpi, MACROBLOCK *x) {
x->block[i].zbin = cpi->Y1zbin[QIndex];
x->block[i].zbin_8x8 = cpi->Y1zbin_8x8[QIndex];
x->block[i].zbin_16x16 = cpi->Y1zbin_16x16[QIndex];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ x->block[i].zbin_32x32 = cpi->Y1zbin_32x32[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];
x->block[i].zrun_zbin_boost_16x16 = cpi->zrun_zbin_boost_y1_16x16[QIndex];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ x->block[i].zrun_zbin_boost_32x32 = cpi->zrun_zbin_boost_y1_32x32[QIndex];
+#endif
x->block[i].zbin_extra = (short)zbin_extra;
// Segment max eob offset feature.
@@ -607,10 +748,17 @@ void vp9_mb_init_quantizer(VP9_COMP *cpi, MACROBLOCK *x) {
vp9_get_segdata(xd, segment_id, SEG_LVL_EOB);
x->block[i].eob_max_offset_16x16 =
vp9_get_segdata(xd, segment_id, SEG_LVL_EOB);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ x->block[i].eob_max_offset_32x32 =
+ vp9_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;
x->block[i].eob_max_offset_16x16 = 256;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ x->block[i].eob_max_offset_32x32 = 1024;
+#endif
}
}
@@ -640,9 +788,12 @@ void vp9_mb_init_quantizer(VP9_COMP *cpi, MACROBLOCK *x) {
vp9_get_segdata(xd, segment_id, SEG_LVL_EOB);
x->block[i].eob_max_offset_8x8 =
vp9_get_segdata(xd, segment_id, SEG_LVL_EOB);
+ x->block[i].eob_max_offset_16x16 =
+ vp9_get_segdata(xd, segment_id, SEG_LVL_EOB);
} else {
x->block[i].eob_max_offset = 16;
x->block[i].eob_max_offset_8x8 = 64;
+ x->block[i].eob_max_offset_16x16 = 256;
}
}
diff --git a/vp9/encoder/vp9_quantize.h b/vp9/encoder/vp9_quantize.h
index dd11e75bab5f0bd2309100a5733d28b6b6462329..832a486f505b4125a21ad026d12dec6fb975dcde 100644
--- a/vp9/encoder/vp9_quantize.h
+++ b/vp9/encoder/vp9_quantize.h
@@ -78,6 +78,11 @@ void vp9_quantize_mb_16x16(MACROBLOCK *x);
extern prototype_quantize_block(vp9_quantize_quantb_16x16);
extern prototype_quantize_mb(vp9_quantize_mby_16x16);
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+void vp9_quantize_sby_32x32(MACROBLOCK *x);
+void vp9_quantize_sbuv_16x16(MACROBLOCK *x);
+#endif
+
struct VP9_COMP;
extern void vp9_set_quantizer(struct VP9_COMP *cpi, int Q);
diff --git a/vp9/encoder/vp9_ratectrl.c b/vp9/encoder/vp9_ratectrl.c
index b589243e8207398f2bd03d60a999bd12bc638b28..c896e41b1eb9cfd14423ae0d27598bdf9d0278cf 100644
--- a/vp9/encoder/vp9_ratectrl.c
+++ b/vp9/encoder/vp9_ratectrl.c
@@ -175,6 +175,9 @@ void vp9_save_coding_context(VP9_COMP *cpi) {
vp9_copy(cc->hybrid_coef_probs_8x8, cm->fc.hybrid_coef_probs_8x8);
vp9_copy(cc->coef_probs_16x16, cm->fc.coef_probs_16x16);
vp9_copy(cc->hybrid_coef_probs_16x16, cm->fc.hybrid_coef_probs_16x16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_copy(cc->coef_probs_32x32, cm->fc.coef_probs_32x32);
+#endif
vp9_copy(cc->switchable_interp_prob, cm->fc.switchable_interp_prob);
#if CONFIG_COMP_INTERINTRA_PRED
cc->interintra_prob = cm->fc.interintra_prob;
@@ -234,6 +237,9 @@ void vp9_restore_coding_context(VP9_COMP *cpi) {
vp9_copy(cm->fc.hybrid_coef_probs_8x8, cc->hybrid_coef_probs_8x8);
vp9_copy(cm->fc.coef_probs_16x16, cc->coef_probs_16x16);
vp9_copy(cm->fc.hybrid_coef_probs_16x16, cc->hybrid_coef_probs_16x16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ vp9_copy(cm->fc.coef_probs_32x32, cc->coef_probs_32x32);
+#endif
vp9_copy(cm->fc.switchable_interp_prob, cc->switchable_interp_prob);
#if CONFIG_COMP_INTERINTRA_PRED
cm->fc.interintra_prob = cc->interintra_prob;
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 9cea189698b47bab3ccbd653fdd947435962f968..60f14f8feaf60e7475cd09bbf5e82d5417a05a67 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -400,12 +400,18 @@ void vp9_initialize_rd_consts(VP9_COMP *cpi, int QIndex) {
cpi->common.fc.hybrid_coef_probs_16x16,
BLOCK_TYPES_16X16);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ fill_token_costs(
+ cpi->mb.token_costs[TX_32X32],
+ (const vp9_prob(*)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.coef_probs_32x32,
+ BLOCK_TYPES_32X32);
+#endif
+
/*rough estimate for costing*/
cpi->common.kf_ymode_probs_index = cpi->common.base_qindex >> 4;
vp9_init_mode_costs(cpi);
- if (cpi->common.frame_type != KEY_FRAME)
- {
+ if (cpi->common.frame_type != KEY_FRAME) {
vp9_build_nmv_cost_table(
cpi->mb.nmvjointcost,
cpi->mb.e_mbd.allow_high_precision_mv ?
@@ -556,7 +562,7 @@ static int cost_coeffs_2x2(MACROBLOCK *mb,
static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type,
ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
- int tx_size) {
+ TX_SIZE tx_size) {
const int eob = b->eob;
int c = (type == PLANE_TYPE_Y_NO_DC); /* start at coef 0, unless Y with Y2 */
int cost = 0, default_eob, seg_eob;
@@ -613,9 +619,24 @@ static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type,
default_eob = 256;
if (type == PLANE_TYPE_Y_WITH_DC) {
tx_type = get_tx_type_16x16(xd, b);
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ } else if (type == PLANE_TYPE_UV) {
+ int ib = (int)(b - xd->block) - 16;
+
+ qcoeff_ptr = xd->sb_coeff_data.qcoeff + 1024 + 64 * ib;
+#endif
}
break;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ case TX_32X32:
+ scan = vp9_default_zig_zag1d_32x32;
+ band = vp9_coef_bands_32x32;
+ default_eob = 1024;
+ qcoeff_ptr = xd->sb_coeff_data.qcoeff;
+ break;
+#endif
default:
+ abort();
break;
}
if (vp9_segfeature_active(&mb->e_mbd, segment_id, SEG_LVL_EOB))
@@ -813,23 +834,28 @@ static void macro_block_yrd_16x16(MACROBLOCK *mb, int *Rate, int *Distortion,
}
static void choose_txfm_size_from_rd(VP9_COMP *cpi, MACROBLOCK *x,
- int r[2][TX_SIZE_MAX], int *rate,
- int d[TX_SIZE_MAX], int *distortion,
- int s[TX_SIZE_MAX], int *skip,
- int64_t txfm_cache[NB_TXFM_MODES]) {
+ int (*r)[2], int *rate,
+ int *d, int *distortion,
+ int *s, int *skip,
+ int64_t txfm_cache[NB_TXFM_MODES],
+ TX_SIZE max_txfm_size) {
VP9_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mode_info_context->mbmi;
vp9_prob skip_prob = cm->mb_no_coeff_skip ?
vp9_get_pred_prob(cm, xd, PRED_MBSKIP) : 128;
- int64_t rd[2][TX_SIZE_MAX];
- int n;
-
- r[1][TX_16X16] = r[0][TX_16X16] + vp9_cost_one(cm->prob_tx[0]) +
- vp9_cost_one(cm->prob_tx[1]);
- r[1][TX_8X8] = r[0][TX_8X8] + vp9_cost_one(cm->prob_tx[0]) +
- vp9_cost_zero(cm->prob_tx[1]);
- r[1][TX_4X4] = r[0][TX_4X4] + vp9_cost_zero(cm->prob_tx[0]);
+ int64_t rd[TX_SIZE_MAX_SB][2];
+ int n, m;
+
+ for (n = TX_4X4; n <= max_txfm_size; n++) {
+ r[n][1] = r[n][0];
+ for (m = 0; m <= n - (n == max_txfm_size); m++) {
+ if (m == n)
+ r[n][1] += vp9_cost_zero(cm->prob_tx[m]);
+ else
+ r[n][1] += vp9_cost_one(cm->prob_tx[m]);
+ }
+ }
if (cm->mb_no_coeff_skip) {
int s0, s1;
@@ -838,64 +864,82 @@ static void choose_txfm_size_from_rd(VP9_COMP *cpi, MACROBLOCK *x,
s0 = vp9_cost_bit(skip_prob, 0);
s1 = vp9_cost_bit(skip_prob, 1);
- for (n = TX_4X4; n <= TX_16X16; n++) {
+ for (n = TX_4X4; n <= max_txfm_size; n++) {
if (s[n]) {
- rd[0][n] = rd[1][n] = RDCOST(x->rdmult, x->rddiv, s1, d[n]);
+ rd[n][0] = rd[n][1] = RDCOST(x->rdmult, x->rddiv, s1, d[n]);
} else {
- rd[0][n] = RDCOST(x->rdmult, x->rddiv, r[0][n] + s0, d[n]);
- rd[1][n] = RDCOST(x->rdmult, x->rddiv, r[1][n] + s0, d[n]);
+ rd[n][0] = RDCOST(x->rdmult, x->rddiv, r[n][0] + s0, d[n]);
+ rd[n][1] = RDCOST(x->rdmult, x->rddiv, r[n][1] + s0, d[n]);
}
}
} else {
- for (n = TX_4X4; n <= TX_16X16; n++) {
- rd[0][n] = RDCOST(x->rdmult, x->rddiv, r[0][n], d[n]);
- rd[1][n] = RDCOST(x->rdmult, x->rddiv, r[1][n], d[n]);
+ for (n = TX_4X4; n <= max_txfm_size; n++) {
+ rd[n][0] = RDCOST(x->rdmult, x->rddiv, r[n][0], d[n]);
+ rd[n][1] = RDCOST(x->rdmult, x->rddiv, r[n][1], d[n]);
}
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ if (max_txfm_size == TX_32X32 &&
+ (cm->txfm_mode == ALLOW_32X32 ||
+ (cm->txfm_mode == TX_MODE_SELECT &&
+ rd[TX_32X32][1] < rd[TX_16X16][1] && rd[TX_32X32][1] < rd[TX_8X8][1] &&
+ rd[TX_32X32][1] < rd[TX_4X4][1]))) {
+ mbmi->txfm_size = TX_32X32;
+ } else
+#endif
if ( cm->txfm_mode == ALLOW_16X16 ||
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ (max_txfm_size == TX_16X16 && cm->txfm_mode == ALLOW_32X32) ||
+#endif
(cm->txfm_mode == TX_MODE_SELECT &&
- rd[1][TX_16X16] < rd[1][TX_8X8] && rd[1][TX_16X16] < rd[1][TX_4X4])) {
+ rd[TX_16X16][1] < rd[TX_8X8][1] && rd[TX_16X16][1] < rd[TX_4X4][1])) {
mbmi->txfm_size = TX_16X16;
} else if (cm->txfm_mode == ALLOW_8X8 ||
- (cm->txfm_mode == TX_MODE_SELECT && rd[1][TX_8X8] < rd[1][TX_4X4])) {
+ (cm->txfm_mode == TX_MODE_SELECT && rd[TX_8X8][1] < rd[TX_4X4][1])) {
mbmi->txfm_size = TX_8X8;
} else {
- assert(cm->txfm_mode == ONLY_4X4 ||
- (cm->txfm_mode == TX_MODE_SELECT && rd[1][TX_4X4] <= rd[1][TX_8X8]));
+ assert(cm->txfm_mode == ONLY_4X4 || cm->txfm_mode == TX_MODE_SELECT);
mbmi->txfm_size = TX_4X4;
}
*distortion = d[mbmi->txfm_size];
- *rate = r[cm->txfm_mode == TX_MODE_SELECT][mbmi->txfm_size];
+ *rate = r[mbmi->txfm_size][cm->txfm_mode == TX_MODE_SELECT];
*skip = s[mbmi->txfm_size];
- txfm_cache[ONLY_4X4] = rd[0][TX_4X4];
- txfm_cache[ALLOW_8X8] = rd[0][TX_8X8];
- txfm_cache[ALLOW_16X16] = rd[0][TX_16X16];
- if (rd[1][TX_16X16] < rd[1][TX_8X8] && rd[1][TX_16X16] < rd[1][TX_4X4])
- txfm_cache[TX_MODE_SELECT] = rd[1][TX_16X16];
+ txfm_cache[ONLY_4X4] = rd[TX_4X4][0];
+ txfm_cache[ALLOW_8X8] = rd[TX_8X8][0];
+ txfm_cache[ALLOW_16X16] = rd[TX_16X16][0];
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ txfm_cache[ALLOW_32X32] = rd[max_txfm_size][0];
+ if (max_txfm_size == TX_32X32 &&
+ rd[TX_32X32][1] < rd[TX_16X16][1] && rd[TX_32X32][1] < rd[TX_8X8][1] &&
+ rd[TX_32X32][1] < rd[TX_4X4][1])
+ txfm_cache[TX_MODE_SELECT] = rd[TX_32X32][1];
+ else
+#endif
+ if (rd[TX_16X16][1] < rd[TX_8X8][1] && rd[TX_16X16][1] < rd[TX_4X4][1])
+ txfm_cache[TX_MODE_SELECT] = rd[TX_16X16][1];
else
- txfm_cache[TX_MODE_SELECT] = rd[1][TX_4X4] < rd[1][TX_8X8] ?
- rd[1][TX_4X4] : rd[1][TX_8X8];
+ txfm_cache[TX_MODE_SELECT] = rd[TX_4X4][1] < rd[TX_8X8][1] ?
+ rd[TX_4X4][1] : rd[TX_8X8][1];
}
static void macro_block_yrd(VP9_COMP *cpi, MACROBLOCK *x, int *rate,
int *distortion, int *skippable,
int64_t txfm_cache[NB_TXFM_MODES]) {
MACROBLOCKD *const xd = &x->e_mbd;
- int r[2][TX_SIZE_MAX], d[TX_SIZE_MAX], s[TX_SIZE_MAX];
+ int r[TX_SIZE_MAX_MB][2], d[TX_SIZE_MAX_MB], s[TX_SIZE_MAX_MB];
vp9_subtract_mby(x->src_diff, *(x->block[0].base_src), xd->predictor,
x->block[0].src_stride);
- macro_block_yrd_16x16(x, &r[0][TX_16X16], &d[TX_16X16],
- &s[TX_16X16], 1);
- macro_block_yrd_8x8(x, &r[0][TX_8X8], &d[TX_8X8], &s[TX_8X8], 1);
- macro_block_yrd_4x4(x, &r[0][TX_4X4], &d[TX_4X4], &s[TX_4X4], 1);
+ macro_block_yrd_16x16(x, &r[TX_16X16][0], &d[TX_16X16], &s[TX_16X16], 1);
+ macro_block_yrd_8x8(x, &r[TX_8X8][0], &d[TX_8X8], &s[TX_8X8], 1);
+ macro_block_yrd_4x4(x, &r[TX_4X4][0], &d[TX_4X4], &s[TX_4X4], 1);
choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s, skippable,
- txfm_cache);
+ txfm_cache, TX_16X16);
}
static void copy_predictor(unsigned char *dst, const unsigned char *predictor) {
@@ -908,25 +952,91 @@ static void copy_predictor(unsigned char *dst, const unsigned char *predictor) {
}
#if CONFIG_SUPERBLOCKS
+#if CONFIG_TX32X32
+static int rdcost_sby_32x32(MACROBLOCK *x) {
+ MACROBLOCKD * const xd = &x->e_mbd;
+ ENTROPY_CONTEXT_PLANES t_above, t_left;
+ ENTROPY_CONTEXT *ta = (ENTROPY_CONTEXT *) &t_above,
+ *tl = (ENTROPY_CONTEXT *) &t_left;
+
+ vpx_memcpy(&t_above, xd->above_context, sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(&t_left, xd->left_context, sizeof(ENTROPY_CONTEXT_PLANES));
+
+ return cost_coeffs(x, xd->block, PLANE_TYPE_Y_WITH_DC, ta, tl, TX_32X32);
+}
+
+static int vp9_sb_block_error_c(short *coeff, short *dqcoeff, int block_size) {
+ int i;
+ int64_t error = 0;
+
+ for (i = 0; i < block_size; i++) {
+ unsigned int this_diff = coeff[i] - dqcoeff[i];
+ error += this_diff * this_diff;
+ }
+
+ return error > INT_MAX ? INT_MAX : error;
+}
+
+#define DEBUG_ERROR 0
+static void super_block_yrd_32x32(MACROBLOCK *x,
+ int *rate, int *distortion, int *skippable) {
+ SUPERBLOCK * const x_sb = &x->sb_coeff_data;
+ MACROBLOCKD * const xd = &x->e_mbd;
+ SUPERBLOCKD * const xd_sb = &xd->sb_coeff_data;
+#if DEBUG_ERROR || CONFIG_DWT32X32HYBRID
+ short out[1024];
+#endif
+
+ vp9_transform_sby_32x32(x);
+ vp9_quantize_sby_32x32(x);
+#if DEBUG_ERROR || CONFIG_DWT32X32HYBRID
+ vp9_short_idct32x32(xd_sb->dqcoeff, out, 64);
+#endif
+
+#if !CONFIG_DWT32X32HYBRID
+ *distortion = vp9_sb_block_error_c(x_sb->coeff, xd_sb->dqcoeff, 1024);
+#else
+ *distortion = vp9_block_error_c(x_sb->src_diff, out, 1024) << 4;
+#endif
+#if DEBUG_ERROR
+ printf("IDCT/FDCT error 32x32: %d (d: %d)\n",
+ vp9_block_error_c(x_sb->src_diff, out, 1024), *distortion);
+#endif
+ *rate = rdcost_sby_32x32(x);
+ *skippable = vp9_sby_is_skippable_32x32(&x->e_mbd);
+}
+#endif
+
static void super_block_yrd(VP9_COMP *cpi,
MACROBLOCK *x, int *rate, int *distortion,
int *skip,
int64_t txfm_cache[NB_TXFM_MODES]) {
MACROBLOCKD *const xd = &x->e_mbd;
- int r[2][TX_SIZE_MAX], d[TX_SIZE_MAX], s[TX_SIZE_MAX], n;
+ int r[TX_SIZE_MAX_SB][2], d[TX_SIZE_MAX_SB], s[TX_SIZE_MAX_SB], n;
const uint8_t *src = x->src.y_buffer, *dst = xd->dst.y_buffer;
int src_y_stride = x->src.y_stride, dst_y_stride = xd->dst.y_stride;
- ENTROPY_CONTEXT_PLANES t_above[3][2], *orig_above = xd->above_context;
- ENTROPY_CONTEXT_PLANES t_left[3][2], *orig_left = xd->left_context;
+ ENTROPY_CONTEXT_PLANES t_above[TX_SIZE_MAX_MB][2],
+ *orig_above = xd->above_context;
+ ENTROPY_CONTEXT_PLANES t_left[TX_SIZE_MAX_MB][2],
+ *orig_left = xd->left_context;
- for (n = TX_4X4; n <= TX_16X16; n++) {
+ for (n = TX_4X4; n < TX_SIZE_MAX_MB; n++) {
vpx_memcpy(t_above[n], xd->above_context, sizeof(t_above[n]));
vpx_memcpy(t_left[n], xd->left_context, sizeof(t_left[n]));
- r[0][n] = 0;
+ r[n][0] = 0;
d[n] = 0;
s[n] = 1;
}
+#if CONFIG_TX32X32
+ vp9_subtract_sby_s_c(x->sb_coeff_data.src_diff, src, src_y_stride,
+ dst, dst_y_stride);
+ super_block_yrd_32x32(x, &r[TX_32X32][0], &d[TX_32X32], &s[TX_32X32]);
+#endif
+
+#if DEBUG_ERROR
+ int err[3] = { 0, 0, 0 };
+#endif
for (n = 0; n < 4; n++) {
int x_idx = n & 1, y_idx = n >> 1;
int r_tmp, d_tmp, s_tmp;
@@ -941,25 +1051,42 @@ static void super_block_yrd(VP9_COMP *cpi,
xd->left_context = &t_left[TX_16X16][y_idx];
macro_block_yrd_16x16(x, &r_tmp, &d_tmp, &s_tmp, 0);
d[TX_16X16] += d_tmp;
- r[0][TX_16X16] += r_tmp;
+ r[TX_16X16][0] += r_tmp;
s[TX_16X16] = s[TX_16X16] && s_tmp;
+#if DEBUG_ERROR
+ vp9_inverse_transform_mby_16x16(xd);
+ err[2] += vp9_block_error_c(xd->diff, x->src_diff, 256);
+#endif
xd->above_context = &t_above[TX_4X4][x_idx];
xd->left_context = &t_left[TX_4X4][y_idx];
macro_block_yrd_4x4(x, &r_tmp, &d_tmp, &s_tmp, 0);
d[TX_4X4] += d_tmp;
- r[0][TX_4X4] += r_tmp;
+ r[TX_4X4][0] += r_tmp;
s[TX_4X4] = s[TX_4X4] && s_tmp;
+#if DEBUG_ERROR
+ vp9_inverse_transform_mby_4x4(xd);
+ err[0] += vp9_block_error_c(xd->diff, x->src_diff, 256);
+#endif
xd->above_context = &t_above[TX_8X8][x_idx];
xd->left_context = &t_left[TX_8X8][y_idx];
macro_block_yrd_8x8(x, &r_tmp, &d_tmp, &s_tmp, 0);
d[TX_8X8] += d_tmp;
- r[0][TX_8X8] += r_tmp;
+ r[TX_8X8][0] += r_tmp;
s[TX_8X8] = s[TX_8X8] && s_tmp;
+#if DEBUG_ERROR
+ vp9_inverse_transform_mby_8x8(xd);
+ err[1] += vp9_block_error_c(xd->diff, x->src_diff, 256);
+#endif
}
-
- choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s, skip, txfm_cache);
+#if DEBUG_ERROR
+ printf("IDCT/FDCT error 16x16: %d (d: %d)\n", err[2], d[2]);
+ printf("IDCT/FDCT error 8x8: %d (d: %d)\n", err[1], d[1]);
+ printf("IDCT/FDCT error 4x4: %d (d: %d)\n", err[0], d[0]);
+#endif
+ choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s, skip, txfm_cache,
+ TX_SIZE_MAX_SB - 1);
xd->above_context = orig_above;
xd->left_context = orig_left;
@@ -1632,14 +1759,59 @@ static int64_t rd_inter16x16_uv_8x8(VP9_COMP *cpi, MACROBLOCK *x, int *rate,
}
#if CONFIG_SUPERBLOCKS
+#if CONFIG_TX32X32
+static int rd_cost_sbuv_16x16(MACROBLOCK *x) {
+ int b;
+ int cost = 0;
+ MACROBLOCKD *const xd = &x->e_mbd;
+ ENTROPY_CONTEXT_PLANES t_above, t_left;
+ ENTROPY_CONTEXT *ta, *tl;
+
+ vpx_memcpy(&t_above, xd->above_context, sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(&t_left, xd->left_context, sizeof(ENTROPY_CONTEXT_PLANES));
+
+ ta = (ENTROPY_CONTEXT *) &t_above;
+ tl = (ENTROPY_CONTEXT *) &t_left;
+
+ for (b = 16; b < 24; b += 4)
+ cost += cost_coeffs(x, xd->block + b, PLANE_TYPE_UV,
+ ta + vp9_block2above_8x8[b],
+ tl + vp9_block2left_8x8[b], TX_16X16);
+
+ return cost;
+}
+
+static void rd_inter32x32_uv_16x16(MACROBLOCK *x, int *rate,
+ int *distortion, int *skip) {
+ MACROBLOCKD *const xd = &x->e_mbd;
+
+ vp9_transform_sbuv_16x16(x);
+ vp9_quantize_sbuv_16x16(x);
+
+ *rate = rd_cost_sbuv_16x16(x);
+ *distortion = vp9_block_error_c(x->sb_coeff_data.coeff + 1024,
+ xd->sb_coeff_data.dqcoeff + 1024, 512) >> 2;
+ *skip = vp9_sbuv_is_skippable_16x16(xd);
+}
+#endif
+
static int64_t rd_inter32x32_uv(VP9_COMP *cpi, MACROBLOCK *x, int *rate,
int *distortion, int fullpixel, int *skip) {
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mode_info_context->mbmi;
- int n, r = 0, d = 0;
const uint8_t *usrc = x->src.u_buffer, *udst = xd->dst.u_buffer;
const uint8_t *vsrc = x->src.v_buffer, *vdst = xd->dst.v_buffer;
int src_uv_stride = x->src.uv_stride, dst_uv_stride = xd->dst.uv_stride;
+
+#if CONFIG_TX32X32
+ if (mbmi->txfm_size == TX_32X32) {
+ vp9_subtract_sbuv_s_c(x->sb_coeff_data.src_diff,
+ usrc, vsrc, src_uv_stride,
+ udst, vdst, dst_uv_stride);
+ rd_inter32x32_uv_16x16(x, rate, distortion, skip);
+ } else {
+#endif
+ int n, r = 0, d = 0;
int skippable = 1;
ENTROPY_CONTEXT_PLANES t_above[2], t_left[2];
ENTROPY_CONTEXT_PLANES *ta = xd->above_context;
@@ -1680,8 +1852,11 @@ static int64_t rd_inter32x32_uv(VP9_COMP *cpi, MACROBLOCK *x, int *rate,
xd->above_context = ta;
memcpy(xd->above_context, t_above, sizeof(t_above));
memcpy(xd->left_context, t_left, sizeof(t_left));
+#if CONFIG_TX32X32
+ }
+#endif
- return RDCOST(x->rdmult, x->rddiv, r, d);
+ return RDCOST(x->rdmult, x->rddiv, *rate, *distortion);
}
#endif
@@ -1818,15 +1993,26 @@ static void rd_pick_intra_mbuv_mode_8x8(VP9_COMP *cpi,
}
#if CONFIG_SUPERBLOCKS
-static void super_block_uvrd_8x8(MACROBLOCK *x,
- int *rate,
- int *distortion,
- int *skippable) {
+// TODO(rbultje) very similar to rd_inter32x32_uv(), merge?
+static void super_block_uvrd(MACROBLOCK *x,
+ int *rate,
+ int *distortion,
+ int *skippable) {
MACROBLOCKD *const xd = &x->e_mbd;
- int d = 0, r = 0, n, s = 1;
+ MB_MODE_INFO *const mbmi = &xd->mode_info_context->mbmi;
const uint8_t *usrc = x->src.u_buffer, *udst = xd->dst.u_buffer;
const uint8_t *vsrc = x->src.v_buffer, *vdst = xd->dst.v_buffer;
int src_uv_stride = x->src.uv_stride, dst_uv_stride = xd->dst.uv_stride;
+
+#if CONFIG_TX32X32
+ if (mbmi->txfm_size == TX_32X32) {
+ vp9_subtract_sbuv_s_c(x->sb_coeff_data.src_diff,
+ usrc, vsrc, src_uv_stride,
+ udst, vdst, dst_uv_stride);
+ rd_inter32x32_uv_16x16(x, rate, distortion, skippable);
+ } else {
+#endif
+ int d = 0, r = 0, n, s = 1;
ENTROPY_CONTEXT_PLANES t_above[2], t_left[2];
ENTROPY_CONTEXT_PLANES *ta = xd->above_context;
ENTROPY_CONTEXT_PLANES *tl = xd->left_context;
@@ -1844,9 +2030,15 @@ static void super_block_uvrd_8x8(MACROBLOCK *x,
udst + x_idx * 8 + y_idx * 8 * dst_uv_stride,
vdst + x_idx * 8 + y_idx * 8 * dst_uv_stride,
dst_uv_stride);
- vp9_transform_mbuv_8x8(x);
- vp9_quantize_mbuv_8x8(x);
- s &= vp9_mbuv_is_skippable_8x8(xd);
+ if (mbmi->txfm_size == TX_4X4) {
+ vp9_transform_mbuv_4x4(x);
+ vp9_quantize_mbuv_4x4(x);
+ s &= vp9_mbuv_is_skippable_4x4(xd);
+ } else {
+ vp9_transform_mbuv_8x8(x);
+ vp9_quantize_mbuv_8x8(x);
+ s &= vp9_mbuv_is_skippable_8x8(xd);
+ }
d += vp9_mbuverror(x) >> 2;
xd->above_context = ta + x_idx;
@@ -1864,6 +2056,9 @@ static void super_block_uvrd_8x8(MACROBLOCK *x,
xd->above_context = ta;
memcpy(xd->above_context, t_above, sizeof(t_above));
memcpy(xd->left_context, t_left, sizeof(t_left));
+#if CONFIG_TX32X32
+ }
+#endif
}
static int64_t rd_pick_intra_sbuv_mode(VP9_COMP *cpi,
@@ -1882,8 +2077,8 @@ static int64_t rd_pick_intra_sbuv_mode(VP9_COMP *cpi,
x->e_mbd.mode_info_context->mbmi.uv_mode = mode;
vp9_build_intra_predictors_sbuv_s(&x->e_mbd);
- super_block_uvrd_8x8(x, &this_rate_tokenonly,
- &this_distortion, &s);
+ super_block_uvrd(x, &this_rate_tokenonly,
+ &this_distortion, &s);
this_rate = this_rate_tokenonly +
x->intra_uv_mode_cost[x->e_mbd.frame_type][mode];
this_rd = RDCOST(x->rdmult, x->rddiv, this_rate, this_distortion);
@@ -4141,8 +4336,6 @@ void vp9_rd_pick_intra_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
int y_skip, uv_skip;
int64_t txfm_cache[NB_TXFM_MODES];
- xd->mode_info_context->mbmi.txfm_size = TX_8X8;
-
error_y = rd_pick_intra_sby_mode(cpi, x, &rate_y, &rate_y_tokenonly,
&dist_y, &y_skip, txfm_cache);
error_uv = rd_pick_intra_sbuv_mode(cpi, x, &rate_uv, &rate_uv_tokenonly,
@@ -4362,6 +4555,11 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
int dist_uv_4x4 = 0, dist_uv_8x8 = 0, uv_skip_4x4 = 0, uv_skip_8x8 = 0;
MB_PREDICTION_MODE mode_uv_4x4 = NEARESTMV, mode_uv_8x8 = NEARESTMV;
int switchable_filter_index = 0;
+#if CONFIG_TX32X32
+ int rate_uv_16x16 = 0, rate_uv_tokenonly_16x16 = 0;
+ int dist_uv_16x16 = 0, uv_skip_16x16 = 0;
+ MB_PREDICTION_MODE mode_uv_16x16;
+#endif
x->skip = 0;
xd->mode_info_context->mbmi.segment_id = segment_id;
@@ -4397,6 +4595,14 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
&dist_uv_8x8, &uv_skip_8x8);
mode_uv_8x8 = mbmi->uv_mode;
}
+#if CONFIG_TX32X32
+ if (cm->txfm_mode >= ALLOW_32X32) {
+ mbmi->txfm_size = TX_32X32;
+ rd_pick_intra_sbuv_mode(cpi, x, &rate_uv_16x16, &rate_uv_tokenonly_16x16,
+ &dist_uv_16x16, &uv_skip_16x16);
+ mode_uv_16x16 = mbmi->uv_mode;
+ }
+#endif
for (mode_index = 0; mode_index < MAX_MODES;
mode_index += (!switchable_filter_index)) {
@@ -4524,6 +4730,13 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
distortion_uv = dist_uv_4x4;
skippable = skippable && uv_skip_4x4;
mbmi->uv_mode = mode_uv_4x4;
+#if CONFIG_TX32X32
+ } else if (mbmi->txfm_size == TX_32X32) {
+ rate_uv = rate_uv_16x16;
+ distortion_uv = dist_uv_16x16;
+ skippable = skippable && uv_skip_16x16;
+ mbmi->uv_mode = mode_uv_16x16;
+#endif
} else {
rate_uv = rate_uv_8x8;
distortion_uv = dist_uv_8x8;
diff --git a/vp9/encoder/vp9_tokenize.c b/vp9/encoder/vp9_tokenize.c
index 44963b2237dfa050ec8351ce3f4c7a713fe195b2..a662e048edae44ebbae2a729000e848f7b6c7630 100644
--- a/vp9/encoder/vp9_tokenize.c
+++ b/vp9/encoder/vp9_tokenize.c
@@ -117,7 +117,7 @@ static void tokenize_b(VP9_COMP *cpi,
int dry_run) {
int pt; /* near block/prev token context index */
int c = (type == PLANE_TYPE_Y_NO_DC) ? 1 : 0;
- const int eob = b->eob; /* one beyond last nonzero coeff */
+ int eob = b->eob; /* one beyond last nonzero coeff */
TOKENEXTRA *t = *tp; /* store tokens starting here */
const short *qcoeff_ptr = b->qcoeff;
int seg_eob;
@@ -177,7 +177,23 @@ static void tokenize_b(VP9_COMP *cpi,
counts = cpi->coef_counts_16x16;
probs = cpi->common.fc.coef_probs_16x16;
}
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+ if (type == PLANE_TYPE_UV) {
+ int uv_idx = (((int) (b - xd->block)) - 16) >> 2;
+ qcoeff_ptr = xd->sb_coeff_data.qcoeff + 1024 + 256 * uv_idx;
+ }
+#endif
+ break;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ case TX_32X32:
+ seg_eob = 1024;
+ bands = vp9_coef_bands_32x32;
+ scan = vp9_default_zig_zag1d_32x32;
+ counts = cpi->coef_counts_32x32;
+ probs = cpi->common.fc.coef_probs_32x32;
+ qcoeff_ptr = xd->sb_coeff_data.qcoeff;
break;
+#endif
}
if (vp9_segfeature_active(xd, segment_id, SEG_LVL_EOB))
@@ -283,6 +299,79 @@ static int mb_is_skippable_16x16(MACROBLOCKD *xd) {
return (vp9_mby_is_skippable_16x16(xd) & vp9_mbuv_is_skippable_8x8(xd));
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+int vp9_sby_is_skippable_32x32(MACROBLOCKD *xd) {
+ int skip = 1;
+ skip &= !xd->block[0].eob;
+ return skip;
+}
+
+int vp9_sbuv_is_skippable_16x16(MACROBLOCKD *xd) {
+ return (!xd->block[16].eob) & (!xd->block[20].eob);
+}
+
+static int sb_is_skippable_32x32(MACROBLOCKD *xd) {
+ return vp9_sby_is_skippable_32x32(xd) &&
+ vp9_sbuv_is_skippable_16x16(xd);
+}
+
+void vp9_tokenize_sb(VP9_COMP *cpi,
+ MACROBLOCKD *xd,
+ TOKENEXTRA **t,
+ int dry_run) {
+ VP9_COMMON * const cm = &cpi->common;
+ MB_MODE_INFO * const mbmi = &xd->mode_info_context->mbmi;
+ TOKENEXTRA *t_backup = *t;
+ ENTROPY_CONTEXT *A[2] = { (ENTROPY_CONTEXT *) (xd->above_context + 0),
+ (ENTROPY_CONTEXT *) (xd->above_context + 1), };
+ ENTROPY_CONTEXT *L[2] = { (ENTROPY_CONTEXT *) (xd->left_context + 0),
+ (ENTROPY_CONTEXT *) (xd->left_context + 1), };
+ const int mb_skip_context = vp9_get_pred_context(cm, xd, PRED_MBSKIP);
+ const int segment_id = mbmi->segment_id;
+ const int skip_inc = !vp9_segfeature_active(xd, segment_id, SEG_LVL_EOB) ||
+ (vp9_get_segdata(xd, segment_id, SEG_LVL_EOB) != 0);
+ int b;
+
+ mbmi->mb_skip_coeff = sb_is_skippable_32x32(xd);
+
+ if (mbmi->mb_skip_coeff) {
+ if (!dry_run)
+ cpi->skip_true_count[mb_skip_context] += skip_inc;
+ if (!cm->mb_no_coeff_skip) {
+ vp9_stuff_sb(cpi, xd, t, dry_run);
+ } else {
+ vp9_fix_contexts_sb(xd);
+ }
+ if (dry_run)
+ *t = t_backup;
+ return;
+ }
+
+ if (!dry_run)
+ cpi->skip_false_count[mb_skip_context] += skip_inc;
+
+ tokenize_b(cpi, xd, xd->block, t, PLANE_TYPE_Y_WITH_DC,
+ A[0], L[0], TX_32X32, dry_run);
+ A[0][1] = A[0][2] = A[0][3] = A[0][0];
+ L[0][1] = L[0][2] = L[0][3] = L[0][0];
+
+ for (b = 16; b < 24; b += 4) {
+ tokenize_b(cpi, xd, xd->block + b, t, PLANE_TYPE_UV,
+ A[0] + vp9_block2above_8x8[b], L[0] + vp9_block2left_8x8[b],
+ TX_16X16, dry_run);
+ A[0][vp9_block2above_8x8[b] + 1] = A[0][vp9_block2above_8x8[b]];
+ L[0][vp9_block2left_8x8[b] + 1] = L[0][vp9_block2left_8x8[b]];
+ }
+ vpx_memset(&A[0][8], 0, sizeof(A[0][8]));
+ vpx_memset(&L[0][8], 0, sizeof(L[0][8]));
+ vpx_memcpy(A[1], A[0], sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(L[1], L[0], sizeof(ENTROPY_CONTEXT_PLANES));
+
+ if (dry_run)
+ *t = t_backup;
+}
+#endif
+
void vp9_tokenize_mb(VP9_COMP *cpi,
MACROBLOCKD *xd,
TOKENEXTRA **t,
@@ -717,6 +806,13 @@ static __inline void stuff_b(VP9_COMP *cpi,
probs = cpi->common.fc.coef_probs_16x16;
}
break;
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+ case TX_32X32:
+ bands = vp9_coef_bands_32x32;
+ counts = cpi->coef_counts_32x32;
+ probs = cpi->common.fc.coef_probs_32x32;
+ break;
+#endif
}
band = bands[(type == PLANE_TYPE_Y_NO_DC) ? 1 : 0];
t->Token = DCT_EOB_TOKEN;
@@ -775,7 +871,8 @@ static void stuff_mb_16x16(VP9_COMP *cpi, MACROBLOCKD *xd,
A[1] = A[2] = A[3] = A[0];
L[1] = L[2] = L[3] = L[0];
for (b = 16; b < 24; b += 4) {
- stuff_b(cpi, xd, xd->block + b, t, PLANE_TYPE_UV, A + vp9_block2above[b],
+ stuff_b(cpi, xd, xd->block + b, t, PLANE_TYPE_UV,
+ A + vp9_block2above_8x8[b],
L + vp9_block2above_8x8[b], TX_8X8, dry_run);
A[vp9_block2above_8x8[b] + 1] = A[vp9_block2above_8x8[b]];
L[vp9_block2left_8x8[b] + 1] = L[vp9_block2left_8x8[b]];
@@ -869,6 +966,43 @@ void vp9_stuff_mb(VP9_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) {
}
}
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+static void stuff_sb_32x32(VP9_COMP *cpi, MACROBLOCKD *xd,
+ TOKENEXTRA **t, int dry_run) {
+ ENTROPY_CONTEXT *A[2] = { (ENTROPY_CONTEXT *) (xd->above_context + 0),
+ (ENTROPY_CONTEXT *) (xd->above_context + 1), };
+ ENTROPY_CONTEXT *L[2] = { (ENTROPY_CONTEXT *) (xd->left_context + 0),
+ (ENTROPY_CONTEXT *) (xd->left_context + 1), };
+ int b;
+
+ stuff_b(cpi, xd, xd->block, t, PLANE_TYPE_Y_WITH_DC,
+ A[0], L[0], TX_32X32, dry_run);
+ A[0][1] = A[0][2] = A[0][3] = A[0][0];
+ L[0][1] = L[0][2] = L[0][3] = L[0][0];
+ for (b = 16; b < 24; b += 4) {
+ stuff_b(cpi, xd, xd->block + b, t, PLANE_TYPE_UV,
+ A[0] + vp9_block2above_8x8[b],
+ L[0] + vp9_block2above_8x8[b], TX_16X16, dry_run);
+ A[0][vp9_block2above_8x8[b] + 1] = A[0][vp9_block2above_8x8[b]];
+ L[0][vp9_block2left_8x8[b] + 1] = L[0][vp9_block2left_8x8[b]];
+ }
+ vpx_memset(&A[0][8], 0, sizeof(A[0][8]));
+ vpx_memset(&L[0][8], 0, sizeof(L[0][8]));
+ vpx_memcpy(A[1], A[0], sizeof(ENTROPY_CONTEXT_PLANES));
+ vpx_memcpy(L[1], L[0], sizeof(ENTROPY_CONTEXT_PLANES));
+}
+
+void vp9_stuff_sb(VP9_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) {
+ TOKENEXTRA * const t_backup = *t;
+
+ stuff_sb_32x32(cpi, xd, t, dry_run);
+
+ if (dry_run) {
+ *t = t_backup;
+ }
+}
+#endif
+
void vp9_fix_contexts(MACROBLOCKD *xd) {
/* Clear entropy contexts for blocks */
if ((xd->mode_info_context->mbmi.mode != B_PRED
@@ -885,3 +1019,10 @@ void vp9_fix_contexts(MACROBLOCKD *xd) {
xd->left_context->y2 = 1;
}
}
+
+#if CONFIG_TX32X32 && CONFIG_SUPERBLOCKS
+void vp9_fix_contexts_sb(MACROBLOCKD *xd) {
+ vpx_memset(xd->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * 2);
+ vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * 2);
+}
+#endif
diff --git a/vp9/encoder/vp9_tokenize.h b/vp9/encoder/vp9_tokenize.h
index 868909be39fe8ff32a1f319df830c446c7b7d496..cfd5db694b420e2aa718561dde600ae9c7a65bd8 100644
--- a/vp9/encoder/vp9_tokenize.h
+++ b/vp9/encoder/vp9_tokenize.h
@@ -34,16 +34,29 @@ extern int vp9_mbuv_is_skippable_4x4(MACROBLOCKD *xd);
extern int vp9_mby_is_skippable_8x8(MACROBLOCKD *xd, int has_y2_block);
extern int vp9_mbuv_is_skippable_8x8(MACROBLOCKD *xd);
extern int vp9_mby_is_skippable_16x16(MACROBLOCKD *xd);
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+extern int vp9_sby_is_skippable_32x32(MACROBLOCKD *xd);
+extern int vp9_sbuv_is_skippable_16x16(MACROBLOCKD *xd);
+#endif
struct VP9_COMP;
extern void vp9_tokenize_mb(struct VP9_COMP *cpi, MACROBLOCKD *xd,
TOKENEXTRA **t, int dry_run);
+extern void vp9_tokenize_sb(struct VP9_COMP *cpi, MACROBLOCKD *xd,
+ TOKENEXTRA **t, int dry_run);
extern void vp9_stuff_mb(struct VP9_COMP *cpi, MACROBLOCKD *xd,
TOKENEXTRA **t, int dry_run);
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+extern void vp9_stuff_sb(struct VP9_COMP *cpi, MACROBLOCKD *xd,
+ TOKENEXTRA **t, int dry_run);
+#endif
extern void vp9_fix_contexts(MACROBLOCKD *xd);
+#if CONFIG_SUPERBLOCKS && CONFIG_TX32X32
+extern void vp9_fix_contexts_sb(MACROBLOCKD *xd);
+#endif
#ifdef ENTROPY_STATS
void init_context_counters();