Merge "Extends ext-tx to support 32x32 masked transforms" into nextgenv2
diff --git a/test/avg_test.cc b/test/avg_test.cc
index 7d5380f..44d8dd7 100644
--- a/test/avg_test.cc
+++ b/test/avg_test.cc
@@ -55,19 +55,19 @@
}
// Sum Pixels
- unsigned int ReferenceAverage8x8(const uint8_t* source, int pitch ) {
+ unsigned int ReferenceAverage8x8(const uint8_t* source, int pitch) {
unsigned int average = 0;
for (int h = 0; h < 8; ++h)
for (int w = 0; w < 8; ++w)
- average += source[h * source_stride_ + w];
+ average += source[h * pitch + w];
return ((average + 32) >> 6);
}
- unsigned int ReferenceAverage4x4(const uint8_t* source, int pitch ) {
+ unsigned int ReferenceAverage4x4(const uint8_t* source, int pitch) {
unsigned int average = 0;
for (int h = 0; h < 4; ++h)
for (int w = 0; w < 4; ++w)
- average += source[h * source_stride_ + w];
+ average += source[h * pitch + w];
return ((average + 8) >> 4);
}
diff --git a/test/convolve_test.cc b/test/convolve_test.cc
index 0826788..12022be 100644
--- a/test/convolve_test.cc
+++ b/test/convolve_test.cc
@@ -279,8 +279,7 @@
uint16_t *output_ptr,
unsigned int output_stride,
unsigned int output_width,
- unsigned int output_height,
- int bd) {
+ unsigned int output_height) {
unsigned int i, j;
for (i = 0; i < output_height; ++i) {
for (j = 0; j < output_width; ++j) {
@@ -306,7 +305,7 @@
highbd_filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 64,
output_width, output_height, bd);
highbd_block2d_average_c(tmp, 64, dst_ptr, dst_stride,
- output_width, output_height, bd);
+ output_width, output_height);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/test/datarate_test.cc b/test/datarate_test.cc
index b98f8c8..9d5074e 100644
--- a/test/datarate_test.cc
+++ b/test/datarate_test.cc
@@ -850,8 +850,7 @@
const vpx_svc_extra_cfg_t *svc_params,
int spatial_layers,
int temporal_layers,
- int temporal_layering_mode,
- unsigned int total_rate) {
+ int temporal_layering_mode) {
int sl, spatial_layer_target;
float total = 0;
float alloc_ratio[VPX_MAX_LAYERS] = {0};
@@ -914,8 +913,7 @@
cfg_.rc_target_bitrate = i;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
- cfg_.ts_number_layers, cfg_.temporal_layering_mode,
- cfg_.rc_target_bitrate);
+ cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, effective_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
@@ -953,8 +951,7 @@
cfg_.rc_target_bitrate = 800;
ResetModel();
assign_layer_bitrates(&cfg_, &svc_params_, cfg_.ss_number_layers,
- cfg_.ts_number_layers, cfg_.temporal_layering_mode,
- cfg_.rc_target_bitrate);
+ cfg_.ts_number_layers, cfg_.temporal_layering_mode);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(cfg_.rc_target_bitrate, effective_datarate_ * 0.85)
<< " The datarate for the file exceeds the target by too much!";
diff --git a/test/dct16x16_test.cc b/test/dct16x16_test.cc
index 0449b52..d6cc5e4 100644
--- a/test/dct16x16_test.cc
+++ b/test/dct16x16_test.cc
@@ -276,12 +276,12 @@
}
void idct16x16_10_ref(const tran_low_t *in, uint8_t *out, int stride,
- int tx_type) {
+ int /*tx_type*/) {
idct16x16_10(in, out, stride);
}
void idct16x16_12_ref(const tran_low_t *in, uint8_t *out, int stride,
- int tx_type) {
+ int /*tx_type*/) {
idct16x16_12(in, out, stride);
}
@@ -778,7 +778,7 @@
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
- void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) {}
+ void RunFwdTxfm(int16_t * /*in*/, tran_low_t * /*out*/, int /*stride*/) {}
void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
inv_txfm_(out, dst, stride);
}
diff --git a/test/error_resilience_test.cc b/test/error_resilience_test.cc
index 9a2ad2f..cd0dca2 100644
--- a/test/error_resilience_test.cc
+++ b/test/error_resilience_test.cc
@@ -100,7 +100,7 @@
}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
- ::libvpx_test::Encoder *encoder) {
+ ::libvpx_test::Encoder * /*encoder*/) {
frame_flags_ &= ~(VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF);
diff --git a/test/fdct4x4_test.cc b/test/fdct4x4_test.cc
index 3f6b738..0c91aee 100644
--- a/test/fdct4x4_test.cc
+++ b/test/fdct4x4_test.cc
@@ -40,7 +40,7 @@
typedef std::tr1::tuple<FhtFunc, IhtFunc, int, vpx_bit_depth_t> Ht4x4Param;
void fdct4x4_ref(const int16_t *in, tran_low_t *out, int stride,
- int tx_type) {
+ int /*tx_type*/) {
vpx_fdct4x4_c(in, out, stride);
}
@@ -49,7 +49,7 @@
}
void fwht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
- int tx_type) {
+ int /*tx_type*/) {
vp9_fwht4x4_c(in, out, stride);
}
diff --git a/test/fdct8x8_test.cc b/test/fdct8x8_test.cc
index 72d2aed..edf4682 100644
--- a/test/fdct8x8_test.cc
+++ b/test/fdct8x8_test.cc
@@ -47,7 +47,7 @@
typedef std::tr1::tuple<FhtFunc, IhtFunc, int, vpx_bit_depth_t> Ht8x8Param;
typedef std::tr1::tuple<IdctFunc, IdctFunc, int, vpx_bit_depth_t> Idct8x8Param;
-void reference_8x8_dct_1d(const double in[8], double out[8], int stride) {
+void reference_8x8_dct_1d(const double in[8], double out[8]) {
const double kInvSqrt2 = 0.707106781186547524400844362104;
for (int k = 0; k < 8; k++) {
out[k] = 0.0;
@@ -65,7 +65,7 @@
double temp_in[8], temp_out[8];
for (int j = 0; j < 8; ++j)
temp_in[j] = input[j*8 + i];
- reference_8x8_dct_1d(temp_in, temp_out, 1);
+ reference_8x8_dct_1d(temp_in, temp_out);
for (int j = 0; j < 8; ++j)
output[j * 8 + i] = temp_out[j];
}
@@ -74,7 +74,7 @@
double temp_in[8], temp_out[8];
for (int j = 0; j < 8; ++j)
temp_in[j] = output[j + i*8];
- reference_8x8_dct_1d(temp_in, temp_out, 1);
+ reference_8x8_dct_1d(temp_in, temp_out);
// Scale by some magic number
for (int j = 0; j < 8; ++j)
output[j + i * 8] = temp_out[j] * 2;
@@ -82,7 +82,8 @@
}
-void fdct8x8_ref(const int16_t *in, tran_low_t *out, int stride, int tx_type) {
+void fdct8x8_ref(const int16_t *in, tran_low_t *out, int stride,
+ int /*tx_type*/) {
vpx_fdct8x8_c(in, out, stride);
}
@@ -642,7 +643,7 @@
void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
inv_txfm_(out, dst, stride);
}
- void RunFwdTxfm(int16_t *out, tran_low_t *dst, int stride) {}
+ void RunFwdTxfm(int16_t * /*out*/, tran_low_t * /*dst*/, int /*stride*/) {}
IdctFunc ref_txfm_;
IdctFunc inv_txfm_;
diff --git a/test/resize_test.cc b/test/resize_test.cc
index c5f05f3..0177308 100644
--- a/test/resize_test.cc
+++ b/test/resize_test.cc
@@ -90,74 +90,178 @@
unsigned int h;
};
-unsigned int ScaleForFrameNumber(unsigned int frame, unsigned int val) {
- if (frame < 10)
- return val;
- if (frame < 20)
- return val * 3 / 4;
- if (frame < 30)
- return val / 2;
- if (frame < 40)
- return val;
- if (frame < 50)
- return val * 3 / 4;
- if (frame < 60)
- return val / 2;
- if (frame < 70)
- return val * 3 / 4;
- if (frame < 80)
- return val;
- if (frame < 90)
- return val * 3 / 4;
- if (frame < 100)
- return val / 2;
- if (frame < 110)
- return val * 3 / 4;
- if (frame < 120)
- return val;
- if (frame < 130)
- return val * 3 / 4;
- if (frame < 140)
- return val / 2;
- if (frame < 150)
- return val * 3 / 4;
- if (frame < 160)
- return val;
- if (frame < 170)
- return val / 2;
- if (frame < 180)
- return val * 3 / 4;
- if (frame < 190)
- return val;
- if (frame < 200)
- return val * 3 / 4;
- if (frame < 210)
- return val / 2;
- if (frame < 220)
- return val * 3 / 4;
- if (frame < 230)
- return val;
- if (frame < 240)
- return val / 2;
- if (frame < 250)
- return val * 3 / 4;
- return val;
+void ScaleForFrameNumber(unsigned int frame,
+ unsigned int initial_w,
+ unsigned int initial_h,
+ unsigned int *w,
+ unsigned int *h,
+ int flag_codec) {
+ if (frame < 10) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ if (frame < 20) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 30) {
+ *w = initial_w / 2;
+ *h = initial_h / 2;
+ return;
+ }
+ if (frame < 40) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ if (frame < 50) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 60) {
+ *w = initial_w / 2;
+ *h = initial_h / 2;
+ return;
+ }
+ if (frame < 70) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ if (frame < 80) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 90) {
+ *w = initial_w / 2;
+ *h = initial_h / 2;
+ return;
+ }
+ if (frame < 100) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 110) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ if (frame < 120) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 130) {
+ *w = initial_w / 2;
+ *h = initial_h / 2;
+ return;
+ }
+ if (frame < 140) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 150) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ if (frame < 160) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 170) {
+ *w = initial_w / 2;
+ *h = initial_h / 2;
+ return;
+ }
+ if (frame < 180) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 190) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ if (frame < 200) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 210) {
+ *w = initial_w / 2;
+ *h = initial_h / 2;
+ return;
+ }
+ if (frame < 220) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 230) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ if (frame < 240) {
+ *w = initial_w * 3 / 4;
+ *h = initial_h * 3 / 4;
+ return;
+ }
+ if (frame < 250) {
+ *w = initial_w / 2;
+ *h = initial_h / 2;
+ return;
+ }
+ if (frame < 260) {
+ *w = initial_w;
+ *h = initial_h;
+ return;
+ }
+ // Go down very low.
+ if (frame < 270) {
+ *w = initial_w / 4;
+ *h = initial_h / 4;
+ return;
+ }
+ if (flag_codec == 1) {
+ // Cases that only works for VP9.
+ // For VP9: Swap width and height of original.
+ if (frame < 320) {
+ *w = initial_h;
+ *h = initial_w;
+ return;
+ }
+ }
+ *w = initial_w;
+ *h = initial_h;
}
class ResizingVideoSource : public ::libvpx_test::DummyVideoSource {
public:
ResizingVideoSource() {
SetSize(kInitialWidth, kInitialHeight);
- limit_ = 300;
+ limit_ = 350;
}
-
+ int flag_codec_;
virtual ~ResizingVideoSource() {}
protected:
virtual void Next() {
++frame_;
- SetSize(ScaleForFrameNumber(frame_, kInitialWidth),
- ScaleForFrameNumber(frame_, kInitialHeight));
+ unsigned int width;
+ unsigned int height;
+ ScaleForFrameNumber(frame_, kInitialWidth, kInitialHeight, &width, &height,
+ flag_codec_);
+ SetSize(width, height);
FillFrame();
}
};
@@ -184,15 +288,17 @@
TEST_P(ResizeTest, TestExternalResizeWorks) {
ResizingVideoSource video;
+ video.flag_codec_ = 0;
cfg_.g_lag_in_frames = 0;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
for (std::vector<FrameInfo>::const_iterator info = frame_info_list_.begin();
info != frame_info_list_.end(); ++info) {
const unsigned int frame = static_cast<unsigned>(info->pts);
- const unsigned int expected_w = ScaleForFrameNumber(frame, kInitialWidth);
- const unsigned int expected_h = ScaleForFrameNumber(frame, kInitialHeight);
-
+ unsigned int expected_w;
+ unsigned int expected_h;
+ ScaleForFrameNumber(frame, kInitialWidth, kInitialHeight,
+ &expected_w, &expected_h, 0);
EXPECT_EQ(expected_w, info->w)
<< "Frame " << frame << " had unexpected width";
EXPECT_EQ(expected_h, info->h)
@@ -386,6 +492,7 @@
TEST_P(ResizeRealtimeTest, TestExternalResizeWorks) {
ResizingVideoSource video;
+ video.flag_codec_ = 1;
DefaultConfig();
// Disable internal resize for this test.
cfg_.rc_resize_allowed = 0;
@@ -395,9 +502,10 @@
for (std::vector<FrameInfo>::const_iterator info = frame_info_list_.begin();
info != frame_info_list_.end(); ++info) {
const unsigned int frame = static_cast<unsigned>(info->pts);
- const unsigned int expected_w = ScaleForFrameNumber(frame, kInitialWidth);
- const unsigned int expected_h = ScaleForFrameNumber(frame, kInitialHeight);
-
+ unsigned int expected_w;
+ unsigned int expected_h;
+ ScaleForFrameNumber(frame, kInitialWidth, kInitialHeight,
+ &expected_w, &expected_h, 1);
EXPECT_EQ(expected_w, info->w)
<< "Frame " << frame << " had unexpected width";
EXPECT_EQ(expected_h, info->h)
diff --git a/test/test_vector_test.cc b/test/test_vector_test.cc
index 6f0cbdf..f1aa4d7 100644
--- a/test/test_vector_test.cc
+++ b/test/test_vector_test.cc
@@ -103,7 +103,7 @@
const int mode = std::tr1::get<kDecodeMode>(input);
libvpx_test::CompressedVideoSource *video = NULL;
vpx_codec_flags_t flags = 0;
- vpx_codec_dec_cfg_t cfg = {0};
+ vpx_codec_dec_cfg_t cfg = vpx_codec_dec_cfg_t();
char str[256];
if (mode == kFrameParallelMode) {
diff --git a/test/vp9_encoder_parms_get_to_decoder.cc b/test/vp9_encoder_parms_get_to_decoder.cc
index 3ef6022..bd84098 100644
--- a/test/vp9_encoder_parms_get_to_decoder.cc
+++ b/test/vp9_encoder_parms_get_to_decoder.cc
@@ -45,9 +45,9 @@
};
const EncodeParameters kVP9EncodeParameterSet[] = {
- {0, 0, 0, 1, 0, VPX_CR_STUDIO_RANGE, VPX_CS_BT_601},
- {0, 0, 0, 0, 0, VPX_CR_FULL_RANGE, VPX_CS_BT_709},
- {0, 0, 1, 0, 0, VPX_CR_FULL_RANGE, VPX_CS_BT_2020},
+ {0, 0, 0, 1, 0, VPX_CR_STUDIO_RANGE, VPX_CS_BT_601, { 0, 0 }},
+ {0, 0, 0, 0, 0, VPX_CR_FULL_RANGE, VPX_CS_BT_709, { 0, 0 }},
+ {0, 0, 1, 0, 0, VPX_CR_FULL_RANGE, VPX_CS_BT_2020, { 0, 0 }},
{0, 2, 0, 0, 1, VPX_CR_STUDIO_RANGE, VPX_CS_UNKNOWN, { 640, 480 }},
// TODO(JBB): Test profiles (requires more work).
};
@@ -93,7 +93,7 @@
}
virtual bool HandleDecodeResult(const vpx_codec_err_t res_dec,
- const libvpx_test::VideoSource &video,
+ const libvpx_test::VideoSource & /*video*/,
libvpx_test::Decoder *decoder) {
vpx_codec_ctx_t *const vp9_decoder = decoder->GetDecoder();
vpx_codec_alg_priv_t *const priv =
diff --git a/test/vp9_error_block_test.cc b/test/vp9_error_block_test.cc
index 77b12ea..23a249e 100644
--- a/test/vp9_error_block_test.cc
+++ b/test/vp9_error_block_test.cc
@@ -164,7 +164,7 @@
const tran_low_t *dqcoeff,
intptr_t block_size,
int64_t *ssz, int bps) {
- assert(bps == 8);
+ EXPECT_EQ(8, bps);
return vp9_highbd_block_error_8bit_c(coeff, dqcoeff, block_size, ssz);
}
@@ -173,7 +173,7 @@
const tran_low_t *dqcoeff,
intptr_t block_size,
int64_t *ssz, int bps) {
- assert(bps == 8);
+ EXPECT_EQ(8, bps);
return vp9_highbd_block_error_8bit_sse2(coeff, dqcoeff, block_size, ssz);
}
@@ -195,7 +195,7 @@
const tran_low_t *dqcoeff,
intptr_t block_size,
int64_t *ssz, int bps) {
- assert(bps == 8);
+ EXPECT_EQ(8, bps);
return vp9_highbd_block_error_8bit_avx(coeff, dqcoeff, block_size, ssz);
}
diff --git a/test/vp9_ethread_test.cc b/test/vp9_ethread_test.cc
index 8ac5c33..29a653f 100644
--- a/test/vp9_ethread_test.cc
+++ b/test/vp9_ethread_test.cc
@@ -62,7 +62,7 @@
encoder_initialized_ = false;
}
- virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
+ virtual void PreEncodeFrameHook(::libvpx_test::VideoSource * /*video*/,
::libvpx_test::Encoder *encoder) {
if (!encoder_initialized_) {
// Encode 4 column tiles.
diff --git a/test/vp9_intrapred_test.cc b/test/vp9_intrapred_test.cc
index e6198af..416f3c3 100644
--- a/test/vp9_intrapred_test.cc
+++ b/test/vp9_intrapred_test.cc
@@ -34,7 +34,7 @@
virtual ~VP9IntraPredBase() { libvpx_test::ClearSystemState(); }
protected:
- virtual void Predict(PREDICTION_MODE mode) = 0;
+ virtual void Predict() = 0;
void CheckPrediction(int test_case_number, int *error_count) const {
// For each pixel ensure that the calculated value is the same as reference.
@@ -73,7 +73,7 @@
left_col_[y] = rnd.Rand16() & mask_;
}
}
- Predict(DC_PRED);
+ Predict();
CheckPrediction(i, &error_count);
}
ASSERT_EQ(0, error_count);
@@ -106,7 +106,7 @@
mask_ = (1 << bit_depth_) - 1;
}
- virtual void Predict(PREDICTION_MODE mode) {
+ virtual void Predict() {
const uint16_t *const_above_row = above_row_;
const uint16_t *const_left_col = left_col_;
ref_fn_(ref_dst_, stride_, const_above_row, const_left_col, bit_depth_);
diff --git a/vp10/common/reconinter.c b/vp10/common/reconinter.c
index efc1ee1..c1cb69d 100644
--- a/vp10/common/reconinter.c
+++ b/vp10/common/reconinter.c
@@ -654,8 +654,6 @@
uint8_t *dst = use_tmp_dst_buf ?
&final_buf[plane][(i * 8) >> pd->subsampling_x] :
&pd->dst.buf[(i * 8) >> pd->subsampling_x];
- int bmc_stride = pd->dst.stride;
- uint8_t *bmc = &pd->dst.buf[(i * 8) >> pd->subsampling_x];
int tmp_stride = tmp_stride1[plane];
uint8_t *tmp = &tmp_buf1[plane][(i * 8) >> pd->subsampling_x];
const uint8_t *mask[2];
@@ -665,27 +663,22 @@
#if CONFIG_VP9_HIGHBITDEPTH
if (is_hbd) {
uint16_t *dst16 = CONVERT_TO_SHORTPTR(dst);
- uint16_t *bmc16 = CONVERT_TO_SHORTPTR(bmc);
uint16_t *tmp16 = CONVERT_TO_SHORTPTR(tmp);
for (row = 0; row < bh; ++row) {
- for (col = 0; col < bw; ++col) {
- dst16[col] = (mask[0][row] * bmc16[col] + mask[1][row] * tmp16[col]
+ for (col = 0; col < bw; ++col)
+ dst16[col] = (mask[0][row] * dst16[col] + mask[1][row] * tmp16[col]
+ 32) >> 6;
- }
dst16 += dst_stride;
- bmc16 += bmc_stride;
tmp16 += tmp_stride;
}
} else {
#endif // CONFIG_VP9_HIGHBITDEPTH
for (row = 0; row < bh; ++row) {
- for (col = 0; col < bw; ++col) {
- dst[col] = (mask[0][row] * bmc[col] + mask[1][row] * tmp[col] + 32)
+ for (col = 0; col < bw; ++col)
+ dst[col] = (mask[0][row] * dst[col] + mask[1][row] * tmp[col] + 32)
>> 6;
- }
dst += dst_stride;
- bmc += bmc_stride;
tmp += tmp_stride;
}
#if CONFIG_VP9_HIGHBITDEPTH
@@ -727,8 +720,6 @@
uint8_t *dst = use_tmp_dst_buf ?
&final_buf[plane][(i * 8 * dst_stride) >> pd->subsampling_y] :
&pd->dst.buf[(i * 8 * dst_stride) >> pd->subsampling_y];
- int bmc_stride = pd->dst.stride;
- uint8_t *bmc = &pd->dst.buf[(i * 8 * bmc_stride) >> pd->subsampling_y];
int tmp_stride = tmp_stride2[plane];
uint8_t *tmp = &tmp_buf2[plane]
[(i * 8 * tmp_stride) >> pd->subsampling_y];
@@ -739,27 +730,22 @@
#if CONFIG_VP9_HIGHBITDEPTH
if (is_hbd) {
uint16_t *dst16 = CONVERT_TO_SHORTPTR(dst);
- uint16_t *bmc16 = CONVERT_TO_SHORTPTR(bmc);
uint16_t *tmp16 = CONVERT_TO_SHORTPTR(tmp);
for (row = 0; row < bh; ++row) {
- for (col = 0; col < bw; ++col) {
- dst16[col] = (mask[0][row] * bmc16[col] + mask[1][row] * tmp16[col]
+ for (col = 0; col < bw; ++col)
+ dst16[col] = (mask[0][row] * dst16[col] + mask[1][row] * tmp16[col]
+ 32) >> 6;
- }
dst16 += dst_stride;
- bmc16 += bmc_stride;
tmp16 += tmp_stride;
}
} else {
#endif // CONFIG_VP9_HIGHBITDEPTH
for (row = 0; row < bh; ++row) {
- for (col = 0; col < bw; ++col) {
- dst[col] = (mask[0][col] * bmc[col] + mask[1][col] * tmp[col] + 32)
+ for (col = 0; col < bw; ++col)
+ dst[col] = (mask[0][col] * dst[col] + mask[1][col] * tmp[col] + 32)
>> 6;
- }
dst += dst_stride;
- bmc += bmc_stride;
tmp += tmp_stride;
}
#if CONFIG_VP9_HIGHBITDEPTH
diff --git a/vp10/encoder/encodeframe.c b/vp10/encoder/encodeframe.c
index 755c33b..2b96a86 100644
--- a/vp10/encoder/encodeframe.c
+++ b/vp10/encoder/encodeframe.c
@@ -2065,6 +2065,11 @@
if (!x->skip) {
// TODO(geza.lore): Investigate if this can be relaxed
x->skip_recode = 0;
+ memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
+
+ x->skip_optimize = 0;
+ x->use_lp32x32fdct = cpi->sf.use_lp32x32fdct;
+
vp10_encode_sb_supertx(x, bsize);
vp10_tokenize_sb_supertx(cpi, td, tp, !output_enabled, bsize);
} else {
diff --git a/vp10/encoder/encoder.c b/vp10/encoder/encoder.c
index 4c4261e..4f97c79 100644
--- a/vp10/encoder/encoder.c
+++ b/vp10/encoder/encoder.c
@@ -3993,13 +3993,22 @@
int vp10_receive_raw_frame(VP10_COMP *cpi, unsigned int frame_flags,
YV12_BUFFER_CONFIG *sd, int64_t time_stamp,
int64_t end_time) {
- VP10_COMMON *cm = &cpi->common;
+ VP10_COMMON *volatile const cm = &cpi->common;
struct vpx_usec_timer timer;
- int res = 0;
+ volatile int res = 0;
const int subsampling_x = sd->subsampling_x;
const int subsampling_y = sd->subsampling_y;
#if CONFIG_VP9_HIGHBITDEPTH
const int use_highbitdepth = (sd->flags & YV12_FLAG_HIGHBITDEPTH) != 0;
+#endif
+
+ if (setjmp(cm->error.jmp)) {
+ cm->error.setjmp = 0;
+ return -1;
+ }
+ cm->error.setjmp = 1;
+
+#if CONFIG_VP9_HIGHBITDEPTH
check_initial_width(cpi, use_highbitdepth, subsampling_x, subsampling_y);
#else
check_initial_width(cpi, subsampling_x, subsampling_y);
@@ -4032,6 +4041,7 @@
res = -1;
}
+ cm->error.setjmp = 0;
return res;
}
diff --git a/vp10/encoder/rdopt.c b/vp10/encoder/rdopt.c
index 30b8406..736adbb 100644
--- a/vp10/encoder/rdopt.c
+++ b/vp10/encoder/rdopt.c
@@ -5089,6 +5089,8 @@
DECLARE_ALIGNED(16, uint8_t, tmp_buf[MAX_MB_PLANE * 64 * 64]);
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_OBMC
+ int allow_obmc = is_obmc_allowed(mbmi);
+ int best_obmc_flag = 0;
#if CONFIG_VP9_HIGHBITDEPTH
DECLARE_ALIGNED(16, uint16_t, tmp_buf1_16[MAX_MB_PLANE * 64 * 64]);
uint8_t *tmp_buf1;
@@ -5098,13 +5100,11 @@
uint8_t *obmc_tmp_buf[3] = {tmp_buf1, tmp_buf1 + 4096, tmp_buf1 + 8192};
#endif // CONFIG_VP9_HIGHBITDEPTH
int obmc_tmp_stride[3] = {64, 64, 64};
- int best_obmc_flag = 0;
uint8_t tmp_skip_txfm[MAX_MB_PLANE << 2] = {0};
int64_t tmp_bsse[MAX_MB_PLANE << 2] = {0};
int64_t rdobmc;
int skip_txfm_sb_obmc = 0;
int64_t skip_sse_sb_obmc = INT64_MAX;
- int allow_obmc = is_obmc_allowed(mbmi);
#endif // CONFIG_OBMC
int pred_exists = 0;
int intpel_mv;
@@ -5587,6 +5587,7 @@
#if CONFIG_OBMC
int tmp_rate_obmc;
int64_t tmp_dist_obmc;
+ restore_dst_buf(xd, orig_dst, orig_dst_stride);
#endif // CONFIG_OBMC
// Handles the special case when a filter that is not in the
// switchable list (ex. bilinear) is indicated at the frame level, or
@@ -5594,19 +5595,14 @@
vp10_build_inter_predictors_sb(xd, mi_row, mi_col, bsize);
#if CONFIG_OBMC
if (mbmi->obmc) {
- vp10_build_obmc_inter_prediction(cm, xd, mi_row, mi_col, 1,
- obmc_tmp_buf, obmc_tmp_stride,
+ vp10_build_obmc_inter_prediction(cm, xd, mi_row, mi_col, 0,
+ NULL, NULL,
dst_buf1, dst_stride1,
dst_buf2, dst_stride2);
- for (i = 0; i < MAX_MB_PLANE; ++i) {
- xd->plane[i].dst.buf = obmc_tmp_buf[i];
- xd->plane[i].dst.stride = obmc_tmp_stride[i];
- }
model_rd_for_sb(cpi, bsize, x, xd, &tmp_rate, &tmp_dist,
&skip_txfm_sb, &skip_sse_sb);
rd = RDCOST(x->rdmult, x->rddiv,
- rs + tmp_rate + cpi->obmc_cost[bsize][1],
- tmp_dist);
+ rs + tmp_rate + cpi->obmc_cost[bsize][1], tmp_dist);
} else {
#endif // CONFIG_OBMC
model_rd_for_sb(cpi, bsize, x, xd, &tmp_rate, &tmp_dist,
diff --git a/vp9/common/vp9_alloccommon.c b/vp9/common/vp9_alloccommon.c
index 24c6c54..7dd1005 100644
--- a/vp9/common/vp9_alloccommon.c
+++ b/vp9/common/vp9_alloccommon.c
@@ -119,6 +119,20 @@
cm->lf.lfm = NULL;
}
+
+int vp9_alloc_loop_filter(VP9_COMMON *cm) {
+ vpx_free(cm->lf.lfm);
+ // Each lfm holds bit masks for all the 8x8 blocks in a 64x64 region. The
+ // stride and rows are rounded up / truncated to a multiple of 8.
+ cm->lf.lfm_stride = (cm->mi_cols + (MI_BLOCK_SIZE - 1)) >> 3;
+ cm->lf.lfm = (LOOP_FILTER_MASK *)vpx_calloc(
+ ((cm->mi_rows + (MI_BLOCK_SIZE - 1)) >> 3) * cm->lf.lfm_stride,
+ sizeof(*cm->lf.lfm));
+ if (!cm->lf.lfm)
+ return 1;
+ return 0;
+}
+
int vp9_alloc_context_buffers(VP9_COMMON *cm, int width, int height) {
int new_mi_size;
@@ -151,15 +165,8 @@
cm->above_context_alloc_cols = cm->mi_cols;
}
- vpx_free(cm->lf.lfm);
-
- // Each lfm holds bit masks for all the 8x8 blocks in a 64x64 region. The
- // stride and rows are rounded up / truncated to a multiple of 8.
- cm->lf.lfm_stride = (cm->mi_cols + (MI_BLOCK_SIZE - 1)) >> 3;
- cm->lf.lfm = (LOOP_FILTER_MASK *)vpx_calloc(
- ((cm->mi_rows + (MI_BLOCK_SIZE - 1)) >> 3) * cm->lf.lfm_stride,
- sizeof(*cm->lf.lfm));
- if (!cm->lf.lfm) goto fail;
+ if (vp9_alloc_loop_filter(cm))
+ goto fail;
return 0;
diff --git a/vp9/common/vp9_alloccommon.h b/vp9/common/vp9_alloccommon.h
index c0e51a6..e53955b 100644
--- a/vp9/common/vp9_alloccommon.h
+++ b/vp9/common/vp9_alloccommon.h
@@ -23,6 +23,7 @@
void vp9_remove_common(struct VP9Common *cm);
+int vp9_alloc_loop_filter(struct VP9Common *cm);
int vp9_alloc_context_buffers(struct VP9Common *cm, int width, int height);
void vp9_init_context_buffers(struct VP9Common *cm);
void vp9_free_context_buffers(struct VP9Common *cm);
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index a2445b0..cf1fe81 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -3031,10 +3031,24 @@
TileInfo *const tile_info = &tile_data->tile_info;
MACROBLOCKD *const xd = &x->e_mbd;
MODE_INFO *mi;
+ ENTROPY_CONTEXT l[16 * MAX_MB_PLANE], a[16 * MAX_MB_PLANE];
+ BLOCK_SIZE bs = VPXMAX(bsize, BLOCK_8X8); // processing unit block size
+ const int num_4x4_blocks_wide = num_4x4_blocks_wide_lookup[bs];
+ const int num_4x4_blocks_high = num_4x4_blocks_high_lookup[bs];
+ int plane;
+
set_offsets(cpi, tile_info, x, mi_row, mi_col, bsize);
mi = xd->mi[0];
mi->sb_type = bsize;
+ for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
+ struct macroblockd_plane *pd = &xd->plane[plane];
+ memcpy(a + num_4x4_blocks_wide * plane, pd->above_context,
+ (sizeof(a[0]) * num_4x4_blocks_wide) >> pd->subsampling_x);
+ memcpy(l + num_4x4_blocks_high * plane, pd->left_context,
+ (sizeof(l[0]) * num_4x4_blocks_high) >> pd->subsampling_y);
+ }
+
if (cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ && cm->seg.enabled)
if (cyclic_refresh_segment_id_boosted(mi->segment_id))
x->rdmult = vp9_cyclic_refresh_get_rdmult(cpi->cyclic_refresh);
@@ -3052,6 +3066,14 @@
duplicate_mode_info_in_sb(cm, xd, mi_row, mi_col, bsize);
+ for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
+ struct macroblockd_plane *pd = &xd->plane[plane];
+ memcpy(pd->above_context, a + num_4x4_blocks_wide * plane,
+ (sizeof(a[0]) * num_4x4_blocks_wide) >> pd->subsampling_x);
+ memcpy(pd->left_context, l + num_4x4_blocks_high * plane,
+ (sizeof(l[0]) * num_4x4_blocks_high) >> pd->subsampling_y);
+ }
+
if (rd_cost->rate == INT_MAX)
vp9_rd_cost_reset(rd_cost);
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index f3147e9..69d5fa7 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -1538,8 +1538,12 @@
realloc_segmentation_maps(cpi);
cpi->initial_width = cpi->initial_height = 0;
cpi->external_resize = 0;
+ } else if (cm->mi_alloc_size == new_mi_size &&
+ (cpi->oxcf.width > last_w || cpi->oxcf.height > last_h)) {
+ vp9_alloc_loop_filter(cm);
}
}
+
update_frame_size(cpi);
if ((last_w != cpi->oxcf.width || last_h != cpi->oxcf.height) &&
@@ -4141,13 +4145,22 @@
int vp9_receive_raw_frame(VP9_COMP *cpi, unsigned int frame_flags,
YV12_BUFFER_CONFIG *sd, int64_t time_stamp,
int64_t end_time) {
- VP9_COMMON *cm = &cpi->common;
+ VP9_COMMON *volatile const cm = &cpi->common;
struct vpx_usec_timer timer;
- int res = 0;
+ volatile int res = 0;
const int subsampling_x = sd->subsampling_x;
const int subsampling_y = sd->subsampling_y;
#if CONFIG_VP9_HIGHBITDEPTH
const int use_highbitdepth = (sd->flags & YV12_FLAG_HIGHBITDEPTH) != 0;
+#endif
+
+ if (setjmp(cm->error.jmp)) {
+ cm->error.setjmp = 0;
+ return -1;
+ }
+ cm->error.setjmp = 1;
+
+#if CONFIG_VP9_HIGHBITDEPTH
check_initial_width(cpi, use_highbitdepth, subsampling_x, subsampling_y);
#else
check_initial_width(cpi, subsampling_x, subsampling_y);
@@ -4180,6 +4193,7 @@
res = -1;
}
+ cm->error.setjmp = 0;
return res;
}
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 1480ea4..193c9d3 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -787,9 +787,9 @@
#if CONFIG_VP9_HIGHBITDEPTH
uint16_t best_dst16[8 * 8];
#endif
+ memcpy(ta, a, num_4x4_blocks_wide * sizeof(a[0]));
+ memcpy(tl, l, num_4x4_blocks_high * sizeof(l[0]));
- memcpy(ta, a, sizeof(ta));
- memcpy(tl, l, sizeof(tl));
xd->mi[0]->tx_size = TX_4X4;
#if CONFIG_VP9_HIGHBITDEPTH
@@ -810,8 +810,8 @@
continue;
}
- memcpy(tempa, ta, sizeof(ta));
- memcpy(templ, tl, sizeof(tl));
+ memcpy(tempa, ta, num_4x4_blocks_wide * sizeof(ta[0]));
+ memcpy(templ, tl, num_4x4_blocks_high * sizeof(tl[0]));
for (idy = 0; idy < num_4x4_blocks_high; ++idy) {
for (idx = 0; idx < num_4x4_blocks_wide; ++idx) {
@@ -874,8 +874,8 @@
*bestdistortion = distortion;
best_rd = this_rd;
*best_mode = mode;
- memcpy(a, tempa, sizeof(tempa));
- memcpy(l, templ, sizeof(templ));
+ memcpy(a, tempa, num_4x4_blocks_wide * sizeof(tempa[0]));
+ memcpy(l, templ, num_4x4_blocks_high * sizeof(templ[0]));
for (idy = 0; idy < num_4x4_blocks_high * 4; ++idy) {
memcpy(best_dst16 + idy * 8,
CONVERT_TO_SHORTPTR(dst_init + idy * dst_stride),
@@ -914,8 +914,8 @@
continue;
}
- memcpy(tempa, ta, sizeof(ta));
- memcpy(templ, tl, sizeof(tl));
+ memcpy(tempa, ta, num_4x4_blocks_wide * sizeof(ta[0]));
+ memcpy(templ, tl, num_4x4_blocks_high * sizeof(tl[0]));
for (idy = 0; idy < num_4x4_blocks_high; ++idy) {
for (idx = 0; idx < num_4x4_blocks_wide; ++idx) {
@@ -976,8 +976,8 @@
*bestdistortion = distortion;
best_rd = this_rd;
*best_mode = mode;
- memcpy(a, tempa, sizeof(tempa));
- memcpy(l, templ, sizeof(templ));
+ memcpy(a, tempa, num_4x4_blocks_wide * sizeof(tempa[0]));
+ memcpy(l, templ, num_4x4_blocks_high * sizeof(templ[0]));
for (idy = 0; idy < num_4x4_blocks_high * 4; ++idy)
memcpy(best_dst + idy * 8, dst_init + idy * dst_stride,
num_4x4_blocks_wide * 4);
@@ -1013,12 +1013,8 @@
int64_t total_distortion = 0;
int tot_rate_y = 0;
int64_t total_rd = 0;
- ENTROPY_CONTEXT t_above[4], t_left[4];
const int *bmode_costs = cpi->mbmode_cost;
- memcpy(t_above, xd->plane[0].above_context, sizeof(t_above));
- memcpy(t_left, xd->plane[0].left_context, sizeof(t_left));
-
// Pick modes for each sub-block (of size 4x4, 4x8, or 8x4) in an 8x8 block.
for (idy = 0; idy < 2; idy += num_4x4_blocks_high) {
for (idx = 0; idx < 2; idx += num_4x4_blocks_wide) {
@@ -1034,8 +1030,11 @@
}
this_rd = rd_pick_intra4x4block(cpi, mb, idy, idx, &best_mode,
- bmode_costs, t_above + idx, t_left + idy,
+ bmode_costs,
+ xd->plane[0].above_context + idx,
+ xd->plane[0].left_context + idy,
&r, &ry, &d, bsize, best_rd - total_rd);
+
if (this_rd >= best_rd - total_rd)
return INT64_MAX;
diff --git a/vp9/encoder/vp9_speed_features.c b/vp9/encoder/vp9_speed_features.c
index 8a34fd9..f684507 100644
--- a/vp9/encoder/vp9_speed_features.c
+++ b/vp9/encoder/vp9_speed_features.c
@@ -303,14 +303,26 @@
FLAG_SKIP_INTRA_LOWVAR;
sf->adaptive_pred_interp_filter = 2;
- // Disable reference masking if using spatial scaling or for dynamic
- // resizing (internal or external) since pred_mv_sad will not be set
- // (since vp9_mv_pred will not be called).
- // TODO(marpan): Fix this condition to allow reference masking for when
- // all references have same resolution as source frame.
- sf->reference_masking = (cpi->external_resize == 0 &&
- cpi->oxcf.resize_mode != RESIZE_DYNAMIC &&
- cpi->svc.number_spatial_layers == 1) ? 1 : 0;
+ // Reference masking only enabled for 1 spatial layer, and if none of the
+ // references have been scaled. The latter condition needs to be checked
+ // for external or internal dynamic resize.
+ sf->reference_masking = (cpi->svc.number_spatial_layers == 1);
+ if (sf->reference_masking == 1 &&
+ (cpi->external_resize == 1 ||
+ cpi->oxcf.resize_mode == RESIZE_DYNAMIC)) {
+ MV_REFERENCE_FRAME ref_frame;
+ static const int flag_list[4] =
+ {0, VP9_LAST_FLAG, VP9_GOLD_FLAG, VP9_ALT_FLAG};
+ for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {
+ const YV12_BUFFER_CONFIG *yv12 = get_ref_frame_buffer(cpi, ref_frame);
+ if (yv12 != NULL && (cpi->ref_frame_flags & flag_list[ref_frame])) {
+ const struct scale_factors *const scale_fac =
+ &cm->frame_refs[ref_frame - 1].sf;
+ if (vp9_is_scaled(scale_fac))
+ sf->reference_masking = 0;
+ }
+ }
+ }
sf->disable_filter_search_var_thresh = 50;
sf->comp_inter_joint_search_thresh = BLOCK_SIZES;
diff --git a/vpx_dsp/intrapred.c b/vpx_dsp/intrapred.c
index dcc9b30..b1076f8 100644
--- a/vpx_dsp/intrapred.c
+++ b/vpx_dsp/intrapred.c
@@ -320,6 +320,7 @@
const int K = above[2];
const int L = above[3];
const int M = above[4];
+ (void)left;
dst[0] = AVG3(H, I, J);
dst[1] = AVG3(I, J, K);
diff --git a/vpxenc.c b/vpxenc.c
index c61d060..f14470a 100644
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -1470,6 +1470,8 @@
global->codec->fourcc,
pixel_aspect_ratio);
}
+#else
+ (void)pixel_aspect_ratio;
#endif
if (!stream->config.write_webm) {
