Merge "Add init and reset functions for RD_COST struct"
diff --git a/examples/vp8cx_set_ref.c b/examples/vp8cx_set_ref.c
index b0961a2..5e29d80 100644
--- a/examples/vp8cx_set_ref.c
+++ b/examples/vp8cx_set_ref.c
@@ -178,7 +178,7 @@
}
// Flush encoder.
- while (encode_frame(&codec, NULL, -1, writer)) {};
+ while (encode_frame(&codec, NULL, -1, writer)) {}
printf("\n");
fclose(infile);
diff --git a/examples/vp9_spatial_svc_encoder.c b/examples/vp9_spatial_svc_encoder.c
index 10d5388..53ede94 100644
--- a/examples/vp9_spatial_svc_encoder.c
+++ b/examples/vp9_spatial_svc_encoder.c
@@ -183,7 +183,7 @@
enc_cfg->kf_min_dist = arg_parse_uint(&arg);
enc_cfg->kf_max_dist = enc_cfg->kf_min_dist;
} else if (arg_match(&arg, &scale_factors_arg, argi)) {
- snprintf(string_options, 1024, "%s scale-factors=%s",
+ snprintf(string_options, sizeof(string_options), "%s scale-factors=%s",
string_options, arg.val);
} else if (arg_match(&arg, &passes_arg, argi)) {
passes = arg_parse_uint(&arg);
@@ -198,10 +198,10 @@
} else if (arg_match(&arg, &fpf_name_arg, argi)) {
fpf_file_name = arg.val;
} else if (arg_match(&arg, &min_q_arg, argi)) {
- snprintf(string_options, 1024, "%s min-quantizers=%s",
+ snprintf(string_options, sizeof(string_options), "%s min-quantizers=%s",
string_options, arg.val);
} else if (arg_match(&arg, &max_q_arg, argi)) {
- snprintf(string_options, 1024, "%s max-quantizers=%s",
+ snprintf(string_options, sizeof(string_options), "%s max-quantizers=%s",
string_options, arg.val);
} else if (arg_match(&arg, &min_bitrate_arg, argi)) {
min_bitrate = arg_parse_uint(&arg);
diff --git a/test/i420_video_source.h b/test/i420_video_source.h
index c3315f9..0a18480 100644
--- a/test/i420_video_source.h
+++ b/test/i420_video_source.h
@@ -13,104 +13,22 @@
#include <cstdlib>
#include <string>
-#include "test/video_source.h"
+#include "test/yuv_video_source.h"
namespace libvpx_test {
// This class extends VideoSource to allow parsing of raw yv12
// so that we can do actual file encodes.
-class I420VideoSource : public VideoSource {
+class I420VideoSource : public YUVVideoSource {
public:
I420VideoSource(const std::string &file_name,
unsigned int width, unsigned int height,
int rate_numerator, int rate_denominator,
unsigned int start, int limit)
- : file_name_(file_name),
- input_file_(NULL),
- img_(NULL),
- start_(start),
- limit_(limit),
- frame_(0),
- width_(0),
- height_(0),
- framerate_numerator_(rate_numerator),
- framerate_denominator_(rate_denominator) {
- // This initializes raw_sz_, width_, height_ and allocates an img.
- SetSize(width, height);
- }
-
- virtual ~I420VideoSource() {
- vpx_img_free(img_);
- if (input_file_)
- fclose(input_file_);
- }
-
- virtual void Begin() {
- if (input_file_)
- fclose(input_file_);
- input_file_ = OpenTestDataFile(file_name_);
- ASSERT_TRUE(input_file_ != NULL) << "Input file open failed. Filename: "
- << file_name_;
- if (start_) {
- fseek(input_file_, static_cast<unsigned>(raw_sz_) * start_, SEEK_SET);
- }
-
- frame_ = start_;
- FillFrame();
- }
-
- virtual void Next() {
- ++frame_;
- FillFrame();
- }
-
- virtual vpx_image_t *img() const { return (frame_ < limit_) ? img_ : NULL; }
-
- // Models a stream where Timebase = 1/FPS, so pts == frame.
- virtual vpx_codec_pts_t pts() const { return frame_; }
-
- virtual unsigned long duration() const { return 1; }
-
- virtual vpx_rational_t timebase() const {
- const vpx_rational_t t = { framerate_denominator_, framerate_numerator_ };
- return t;
- }
-
- virtual unsigned int frame() const { return frame_; }
-
- virtual unsigned int limit() const { return limit_; }
-
- void SetSize(unsigned int width, unsigned int height) {
- if (width != width_ || height != height_) {
- vpx_img_free(img_);
- img_ = vpx_img_alloc(NULL, VPX_IMG_FMT_I420, width, height, 1);
- ASSERT_TRUE(img_ != NULL);
- width_ = width;
- height_ = height;
- raw_sz_ = width * height * 3 / 2;
- }
- }
-
- virtual void FillFrame() {
- ASSERT_TRUE(input_file_ != NULL);
- // Read a frame from input_file.
- if (fread(img_->img_data, raw_sz_, 1, input_file_) == 0) {
- limit_ = frame_;
- }
- }
-
- protected:
- std::string file_name_;
- FILE *input_file_;
- vpx_image_t *img_;
- size_t raw_sz_;
- unsigned int start_;
- unsigned int limit_;
- unsigned int frame_;
- unsigned int width_;
- unsigned int height_;
- int framerate_numerator_;
- int framerate_denominator_;
+ : YUVVideoSource(file_name, VPX_IMG_FMT_I420,
+ width, height,
+ rate_numerator, rate_denominator,
+ start, limit) {}
};
} // namespace libvpx_test
diff --git a/test/svc_test.cc b/test/svc_test.cc
index fdde702..67e83e3 100644
--- a/test/svc_test.cc
+++ b/test/svc_test.cc
@@ -225,10 +225,9 @@
EXPECT_EQ(received_frames, n);
}
- void DropLayersAndMakeItVP9Comaptible(struct vpx_fixed_buf *const inputs,
- const int num_super_frames,
- const int remained_spatial_layers,
- const bool is_multiple_frame_contexts) {
+ void DropEnhancementLayers(struct vpx_fixed_buf *const inputs,
+ const int num_super_frames,
+ const int remained_spatial_layers) {
ASSERT_TRUE(inputs != NULL);
ASSERT_GT(num_super_frames, 0);
ASSERT_GT(remained_spatial_layers, 0);
@@ -250,45 +249,6 @@
if (frame_count == 0) {
// There's no super frame but only a single frame.
ASSERT_EQ(1, remained_spatial_layers);
- if (is_multiple_frame_contexts) {
- // Make a new super frame.
- uint8_t marker = 0xc1;
- unsigned int mask;
- int mag;
-
- // Choose the magnitude.
- for (mag = 0, mask = 0xff; mag < 4; ++mag) {
- if (inputs[i].sz < mask)
- break;
- mask <<= 8;
- mask |= 0xff;
- }
- marker |= mag << 3;
- int index_sz = 2 + (mag + 1) * 2;
-
- inputs[i].buf = realloc(inputs[i].buf, inputs[i].sz + index_sz + 16);
- ASSERT_TRUE(inputs[i].buf != NULL);
- uint8_t *frame_data = static_cast<uint8_t*>(inputs[i].buf);
- frame_data[0] &= ~2; // Set the show_frame flag to 0.
- frame_data += inputs[i].sz;
- // Add an one byte frame with show_existing_frame.
- *frame_data++ = 0x88;
-
- // Write the super frame index.
- *frame_data++ = marker;
-
- frame_sizes[0] = inputs[i].sz;
- frame_sizes[1] = 1;
- for (int j = 0; j < 2; ++j) {
- unsigned int this_sz = frame_sizes[j];
- for (int k = 0; k <= mag; k++) {
- *frame_data++ = this_sz & 0xff;
- this_sz >>= 8;
- }
- }
- *frame_data++ = marker;
- inputs[i].sz += index_sz + 1;
- }
} else {
// Found a super frame.
uint8_t *frame_data = static_cast<uint8_t*>(inputs[i].buf);
@@ -304,16 +264,13 @@
}
ASSERT_LT(frame, frame_count) << "Couldn't find a visible frame. "
<< "remained_spatial_layers: " << remained_spatial_layers
- << " super_frame: " << i
- << " is_multiple_frame_context: " << is_multiple_frame_contexts;
- if (frame == frame_count - 1 && !is_multiple_frame_contexts)
+ << " super_frame: " << i;
+ if (frame == frame_count - 1)
continue;
frame_data += frame_sizes[frame];
// We need to add one more frame for multiple frame contexts.
- if (is_multiple_frame_contexts)
- ++frame;
uint8_t marker =
static_cast<const uint8_t*>(inputs[i].buf)[inputs[i].sz - 1];
const uint32_t mag = ((marker >> 3) & 0x3) + 1;
@@ -323,35 +280,14 @@
marker |= frame;
// Copy existing frame sizes.
- memmove(frame_data + (is_multiple_frame_contexts ? 2 : 1),
- frame_start + inputs[i].sz - index_sz + 1, new_index_sz - 2);
- if (is_multiple_frame_contexts) {
- // Add a one byte frame with flag show_existing_frame.
- *frame_data++ = 0x88 | (remained_spatial_layers - 1);
- }
+ memmove(frame_data + 1, frame_start + inputs[i].sz - index_sz + 1,
+ new_index_sz - 2);
// New marker.
frame_data[0] = marker;
frame_data += (mag * (frame + 1) + 1);
- if (is_multiple_frame_contexts) {
- // Write the frame size for the one byte frame.
- frame_data -= mag;
- *frame_data++ = 1;
- for (uint32_t j = 1; j < mag; ++j) {
- *frame_data++ = 0;
- }
- }
-
*frame_data++ = marker;
inputs[i].sz = frame_data - frame_start;
-
- if (is_multiple_frame_contexts) {
- // Change the show frame flag to 0 for all frames.
- for (int j = 0; j < frame; ++j) {
- frame_start[0] &= ~2;
- frame_start += frame_sizes[j];
- }
- }
}
}
}
@@ -555,7 +491,7 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, false);
+ DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -573,13 +509,13 @@
Pass2EncodeNFrames(&stats_buf, 10, 5, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 4, false);
+ DropEnhancementLayers(&outputs[0], 10, 4);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 3, false);
+ DropEnhancementLayers(&outputs[0], 10, 3);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 2, false);
+ DropEnhancementLayers(&outputs[0], 10, 2);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, false);
+ DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
@@ -616,9 +552,9 @@
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 20, 3, &outputs[0]);
DecodeNFrames(&outputs[0], 20);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 20, 2, false);
+ DropEnhancementLayers(&outputs[0], 20, 2);
DecodeNFrames(&outputs[0], 20);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 20, 1, false);
+ DropEnhancementLayers(&outputs[0], 20, 1);
DecodeNFrames(&outputs[0], 20);
FreeBitstreamBuffers(&outputs[0], 20);
@@ -649,7 +585,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 2, true);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -667,7 +602,7 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, true);
+ DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -686,7 +621,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 2, true);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -707,32 +641,13 @@
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 3, &outputs[0]);
- vpx_fixed_buf outputs_new[10];
- for (int i = 0; i < 10; ++i) {
- outputs_new[i].buf = malloc(outputs[i].sz + 16);
- ASSERT_TRUE(outputs_new[i].buf != NULL);
- memcpy(outputs_new[i].buf, outputs[i].buf, outputs[i].sz);
- outputs_new[i].sz = outputs[i].sz;
- }
- DropLayersAndMakeItVP9Comaptible(&outputs_new[0], 10, 3, true);
- DecodeNFrames(&outputs_new[0], 10);
-
- for (int i = 0; i < 10; ++i) {
- memcpy(outputs_new[i].buf, outputs[i].buf, outputs[i].sz);
- outputs_new[i].sz = outputs[i].sz;
- }
- DropLayersAndMakeItVP9Comaptible(&outputs_new[0], 10, 2, true);
- DecodeNFrames(&outputs_new[0], 10);
-
- for (int i = 0; i < 10; ++i) {
- memcpy(outputs_new[i].buf, outputs[i].buf, outputs[i].sz);
- outputs_new[i].sz = outputs[i].sz;
- }
- DropLayersAndMakeItVP9Comaptible(&outputs_new[0], 10, 1, true);
- DecodeNFrames(&outputs_new[0], 10);
+ DecodeNFrames(&outputs[0], 10);
+ DropEnhancementLayers(&outputs[0], 10, 2);
+ DecodeNFrames(&outputs[0], 10);
+ DropEnhancementLayers(&outputs[0], 10, 1);
+ DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
- FreeBitstreamBuffers(&outputs_new[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2TemporalLayers) {
@@ -769,7 +684,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, true);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -814,7 +728,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, true);
vpx_fixed_buf base_layer[5];
for (int i = 0; i < 5; ++i)
diff --git a/test/test-data.mk b/test/test-data.mk
index 404c166..e4dae3a 100644
--- a/test/test-data.mk
+++ b/test/test-data.mk
@@ -7,12 +7,15 @@
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_420.y4m
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_422.y4m
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_444.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_440.yuv
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_420.y4m
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_422.y4m
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_444.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_440.yuv
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_420.y4m
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_422.y4m
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_444.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_440.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += rush_hour_444.y4m
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += screendata.y4m
@@ -644,10 +647,12 @@
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-17-show-existing-frame.webm.md5
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-18-resize.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-18-resize.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm.md5
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-02.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-02.webm.md5
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp91-2-04-yuv444.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp91-2-04-yuv444.webm.md5
diff --git a/test/test-data.sha1 b/test/test-data.sha1
index e6114ab..69d1d2f 100644
--- a/test/test-data.sha1
+++ b/test/test-data.sha1
@@ -17,12 +17,15 @@
a432f96ff0a787268e2f94a8092ab161a18d1b06 park_joy_90p_10_420.y4m
0b194cc312c3a2e84d156a221b0a5eb615dfddc5 park_joy_90p_10_422.y4m
ff0e0a21dc2adc95b8c1b37902713700655ced17 park_joy_90p_10_444.y4m
+c934da6fb8cc54ee2a8c17c54cf6076dac37ead0 park_joy_90p_10_440.yuv
614c32ae1eca391e867c70d19974f0d62664dd99 park_joy_90p_12_420.y4m
c92825f1ea25c5c37855083a69faac6ac4641a9e park_joy_90p_12_422.y4m
b592189b885b6cc85db55cc98512a197d73d3b34 park_joy_90p_12_444.y4m
+82c1bfcca368c2f22bad7d693d690d5499ecdd11 park_joy_90p_12_440.yuv
4e0eb61e76f0684188d9bc9f3ce61f6b6b77bb2c park_joy_90p_8_420.y4m
7a193ff7dfeb96ba5f82b2afd7afa9e1fe83d947 park_joy_90p_8_422.y4m
bdb7856e6bc93599bdda05c2e773a9f22b6c6d03 park_joy_90p_8_444.y4m
+81e1f3843748438b8f2e71db484eb22daf72e939 park_joy_90p_8_440.yuv
b1f1c3ec79114b9a0651af24ce634afb44a9a419 rush_hour_444.y4m
5184c46ddca8b1fadd16742e8500115bc8f749da vp80-00-comprehensive-001.ivf
65bf1bbbced81b97bd030f376d1b7f61a224793f vp80-00-comprehensive-002.ivf
@@ -691,6 +694,8 @@
368dccdde5288c13c25695d2eacdc7402cadf613 vp90-2-19-skip.webm.md5
ffe460282df2b0e7d4603c2158653ad96f574b02 vp90-2-19-skip-01.webm
bd21bc9eda4a4a36b221d71ede3a139fc3c7bd85 vp90-2-19-skip-01.webm.md5
+178f5bd239e38cc1cc2657a7a5e1a9f52ad2d3fe vp90-2-19-skip-02.webm
+9020d5e260bd7df08e2b3d4b86f8623cee3daea2 vp90-2-19-skip-02.webm.md5
b03c408cf23158638da18dbc3323b99a1635c68a invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf
0a3884edb3fd8f9d9b500223e650f7de257b67d8 invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf.res
5e67e24e7f53fd189e565513cef8519b1bd6c712 invalid-vp90-2-05-resize.ivf.s59293_r01-05_b6-.ivf
diff --git a/test/test.mk b/test/test.mk
index 1bd732d..30c13a1 100644
--- a/test/test.mk
+++ b/test/test.mk
@@ -23,6 +23,7 @@
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += error_resilience_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += i420_video_source.h
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += y4m_video_source.h
+LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += yuv_video_source.h
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += altref_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += config_test.cc
@@ -38,6 +39,7 @@
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += frame_size_tests.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += resize_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_lossless_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_end_to_end_test.cc
LIBVPX_TEST_SRCS-yes += decode_test_driver.cc
LIBVPX_TEST_SRCS-yes += decode_test_driver.h
diff --git a/test/test_vectors.cc b/test/test_vectors.cc
index cccebf8..7efa8c0 100644
--- a/test/test_vectors.cc
+++ b/test/test_vectors.cc
@@ -182,7 +182,8 @@
"vp90-2-15-segkey.webm", "vp90-2-15-segkey_adpq.webm",
"vp90-2-16-intra-only.webm", "vp90-2-17-show-existing-frame.webm",
"vp90-2-18-resize.ivf", "vp90-2-19-skip.webm",
- "vp90-2-19-skip-01.webm", "vp91-2-04-yuv444.webm",
+ "vp90-2-19-skip-01.webm", "vp90-2-19-skip-02.webm",
+ "vp91-2-04-yuv444.webm",
};
const int kNumVP9TestVectors = NELEMENTS(kVP9TestVectors);
#endif // CONFIG_VP9_DECODER
diff --git a/test/vp8_denoiser_sse2_test.cc b/test/vp8_denoiser_sse2_test.cc
index d04a505..d4abdad 100644
--- a/test/vp8_denoiser_sse2_test.cc
+++ b/test/vp8_denoiser_sse2_test.cc
@@ -28,19 +28,18 @@
namespace {
const int kNumPixels = 16 * 16;
-class VP8DenoiserTest
- : public ::testing::TestWithParam<int> {
+class VP8DenoiserTest : public ::testing::TestWithParam<int> {
public:
virtual ~VP8DenoiserTest() {}
virtual void SetUp() {
- increase_denoising = GetParam();
+ increase_denoising_ = GetParam();
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
- int increase_denoising;
+ int increase_denoising_;
};
TEST_P(VP8DenoiserTest, BitexactCheck) {
@@ -63,8 +62,8 @@
for (int i = 0; i < count_test_block; ++i) {
// Generate random motion magnitude, 20% of which exceed the threshold.
- uint8_t motion_magnitude_random
- = rnd.Rand8() % (uint8_t)(MOTION_MAGNITUDE_THRESHOLD * 1.2);
+ const int motion_magnitude_ran =
+ rnd.Rand8() % static_cast<int>(MOTION_MAGNITUDE_THRESHOLD * 1.2);
// Initialize a test block with random number in range [0, 255].
for (int j = 0; j < kNumPixels; ++j) {
@@ -72,20 +71,20 @@
sig_block_sse2[j] = sig_block_c[j] = rnd.Rand8();
// The pixels in mc_avg_block are generated by adding a random
// number in range [-19, 19] to corresponding pixels in sig_block.
- temp = sig_block_c[j] + (rnd.Rand8() % 2 == 0? -1 : 1) *
- (rnd.Rand8()%20);
+ temp = sig_block_c[j] + (rnd.Rand8() % 2 == 0 ? -1 : 1) *
+ (rnd.Rand8() % 20);
// Clip.
- mc_avg_block[j] = (temp < 0? 0 : (temp > 255? 255 : temp));
+ mc_avg_block[j] = (temp < 0) ? 0 : ((temp > 255) ? 255 : temp);
}
// Test denosiser on Y component.
- ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_c(mc_avg_block, stride,
- avg_block_c, stride, sig_block_c, stride,
- motion_magnitude_random, increase_denoising));
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_c(
+ mc_avg_block, stride, avg_block_c, stride, sig_block_c, stride,
+ motion_magnitude_ran, increase_denoising_));
- ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_sse2(mc_avg_block, stride,
- avg_block_sse2, stride, sig_block_sse2, stride,
- motion_magnitude_random, increase_denoising));
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_sse2(
+ mc_avg_block, stride, avg_block_sse2, stride, sig_block_sse2, stride,
+ motion_magnitude_ran, increase_denoising_));
// Check bitexactness.
for (int h = 0; h < 16; ++h) {
@@ -94,14 +93,14 @@
}
}
- // Test denosiser on UV component.
- ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_uv_c(mc_avg_block, stride,
- avg_block_c, stride, sig_block_c, stride,
- motion_magnitude_random, increase_denoising));
+ // Test denoiser on UV component.
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_uv_c(
+ mc_avg_block, stride, avg_block_c, stride, sig_block_c, stride,
+ motion_magnitude_ran, increase_denoising_));
- ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_uv_sse2(mc_avg_block, stride,
- avg_block_sse2, stride, sig_block_sse2, stride,
- motion_magnitude_random, increase_denoising));
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_uv_sse2(
+ mc_avg_block, stride, avg_block_sse2, stride, sig_block_sse2, stride,
+ motion_magnitude_ran, increase_denoising_));
// Check bitexactness.
for (int h = 0; h < 16; ++h) {
@@ -113,7 +112,5 @@
}
// Test for all block size.
-INSTANTIATE_TEST_CASE_P(
- SSE2, VP8DenoiserTest,
- ::testing::Values(0, 1));
+INSTANTIATE_TEST_CASE_P(SSE2, VP8DenoiserTest, ::testing::Values(0, 1));
} // namespace
diff --git a/test/vp9_denoiser_sse2_test.cc b/test/vp9_denoiser_sse2_test.cc
index edebc83..0ecba07 100644
--- a/test/vp9_denoiser_sse2_test.cc
+++ b/test/vp9_denoiser_sse2_test.cc
@@ -29,19 +29,18 @@
namespace {
const int kNumPixels = 64 * 64;
-class VP9DenoiserTest
- : public ::testing::TestWithParam<int> {
+class VP9DenoiserTest : public ::testing::TestWithParam<BLOCK_SIZE> {
public:
virtual ~VP9DenoiserTest() {}
virtual void SetUp() {
- bs = (BLOCK_SIZE)GetParam();
+ bs_ = GetParam();
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
- BLOCK_SIZE bs;
+ BLOCK_SIZE bs_;
};
TEST_P(VP9DenoiserTest, BitexactCheck) {
@@ -60,8 +59,8 @@
for (int i = 0; i < count_test_block; ++i) {
// Generate random motion magnitude, 20% of which exceed the threshold.
- uint8_t motion_magnitude_random
- = rnd.Rand8() % (uint8_t)(MOTION_MAGNITUDE_THRESHOLD * 1.2);
+ const int motion_magnitude_random =
+ rnd.Rand8() % static_cast<int>(MOTION_MAGNITUDE_THRESHOLD * 1.2);
// Initialize a test block with random number in range [0, 255].
for (int j = 0; j < kNumPixels; ++j) {
@@ -69,23 +68,23 @@
sig_block[j] = rnd.Rand8();
// The pixels in mc_avg_block are generated by adding a random
// number in range [-19, 19] to corresponding pixels in sig_block.
- temp = sig_block[j] + (rnd.Rand8() % 2 == 0? -1 : 1) *
- (rnd.Rand8()%20);
+ temp = sig_block[j] + ((rnd.Rand8() % 2 == 0) ? -1 : 1) *
+ (rnd.Rand8() % 20);
// Clip.
- mc_avg_block[j] = (temp < 0? 0 : (temp > 255? 255 : temp));
+ mc_avg_block[j] = (temp < 0) ? 0 : ((temp > 255) ? 255 : temp);
}
- ASM_REGISTER_STATE_CHECK(vp9_denoiser_filter_c(sig_block, 64,
- mc_avg_block, 64, avg_block_c, 64,
- 0, bs, motion_magnitude_random));
+ ASM_REGISTER_STATE_CHECK(vp9_denoiser_filter_c(
+ sig_block, 64, mc_avg_block, 64, avg_block_c,
+ 64, 0, bs_, motion_magnitude_random));
- ASM_REGISTER_STATE_CHECK(vp9_denoiser_filter_sse2(sig_block, 64,
- mc_avg_block, 64, avg_block_sse2, 64,
- 0, bs, motion_magnitude_random));
+ ASM_REGISTER_STATE_CHECK(vp9_denoiser_filter_sse2(
+ sig_block, 64, mc_avg_block, 64, avg_block_sse2,
+ 64, 0, bs_, motion_magnitude_random));
// Test bitexactness.
- for (int h = 0; h < (4 << b_height_log2_lookup[bs]); ++h) {
- for (int w = 0; w < (4 << b_width_log2_lookup[bs]); ++w) {
+ for (int h = 0; h < (4 << b_height_log2_lookup[bs_]); ++h) {
+ for (int w = 0; w < (4 << b_width_log2_lookup[bs_]); ++w) {
EXPECT_EQ(avg_block_c[h * 64 + w], avg_block_sse2[h * 64 + w]);
}
}
diff --git a/test/vp9_end_to_end_test.cc b/test/vp9_end_to_end_test.cc
new file mode 100644
index 0000000..a8f6793
--- /dev/null
+++ b/test/vp9_end_to_end_test.cc
@@ -0,0 +1,155 @@
+/*
+ * Copyright (c) 2014 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 "test/codec_factory.h"
+#include "test/encode_test_driver.h"
+#include "test/y4m_video_source.h"
+#include "test/yuv_video_source.h"
+#include "test/util.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+namespace {
+
+const unsigned int kWidth = 160;
+const unsigned int kHeight = 90;
+const unsigned int kFramerate = 50;
+const unsigned int kFrames = 10;
+const int kBitrate = 500;
+const int kCpuUsed = 2;
+const double psnr_threshold = 35.0;
+
+typedef struct {
+ const char *filename;
+ unsigned int input_bit_depth;
+ vpx_img_fmt fmt;
+ vpx_bit_depth_t bit_depth;
+ unsigned int profile;
+} TestVideoParam;
+
+const TestVideoParam TestVectors[] = {
+ {"park_joy_90p_8_420.y4m", 8, VPX_IMG_FMT_I420, VPX_BITS_8, 0},
+ {"park_joy_90p_8_422.y4m", 8, VPX_IMG_FMT_I422, VPX_BITS_8, 1},
+ {"park_joy_90p_8_444.y4m", 8, VPX_IMG_FMT_I444, VPX_BITS_8, 1},
+ {"park_joy_90p_8_440.yuv", 8, VPX_IMG_FMT_I440, VPX_BITS_8, 1},
+#if CONFIG_VP9_HIGHBITDEPTH
+ {"park_joy_90p_10_420.y4m", 10, VPX_IMG_FMT_I42016, VPX_BITS_10, 2},
+ {"park_joy_90p_10_422.y4m", 10, VPX_IMG_FMT_I42216, VPX_BITS_10, 3},
+ {"park_joy_90p_10_444.y4m", 10, VPX_IMG_FMT_I44416, VPX_BITS_10, 3},
+ {"park_joy_90p_10_440.yuv", 10, VPX_IMG_FMT_I44016, VPX_BITS_10, 3},
+ {"park_joy_90p_12_420.y4m", 12, VPX_IMG_FMT_I42016, VPX_BITS_12, 2},
+ {"park_joy_90p_12_422.y4m", 12, VPX_IMG_FMT_I42216, VPX_BITS_12, 3},
+ {"park_joy_90p_12_444.y4m", 12, VPX_IMG_FMT_I44416, VPX_BITS_12, 3},
+ {"park_joy_90p_12_440.yuv", 12, VPX_IMG_FMT_I44016, VPX_BITS_12, 3},
+#endif // CONFIG_VP9_HIGHBITDEPTH
+};
+
+int is_extension_y4m(const char *filename) {
+ const char *dot = strrchr(filename, '.');
+ if (!dot || dot == filename)
+ return 0;
+ else
+ return !strcmp(dot, ".y4m");
+}
+
+class EndToEndTestLarge
+ : public ::libvpx_test::EncoderTest,
+ public ::libvpx_test::CodecTestWith2Params<libvpx_test::TestMode, \
+ TestVideoParam> {
+ protected:
+ EndToEndTestLarge()
+ : EncoderTest(GET_PARAM(0)),
+ psnr_(0.0),
+ nframes_(0),
+ encoding_mode_(GET_PARAM(1)) {
+ }
+
+ virtual ~EndToEndTestLarge() {}
+
+ virtual void SetUp() {
+ InitializeConfig();
+ SetMode(encoding_mode_);
+ if (encoding_mode_ != ::libvpx_test::kRealTime) {
+ cfg_.g_lag_in_frames = 5;
+ cfg_.rc_end_usage = VPX_VBR;
+ } else {
+ cfg_.g_lag_in_frames = 0;
+ cfg_.rc_end_usage = VPX_CBR;
+ }
+ test_video_param_ = GET_PARAM(2);
+ }
+
+ virtual void BeginPassHook(unsigned int) {
+ psnr_ = 0.0;
+ nframes_ = 0;
+ }
+
+ virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {
+ psnr_ += pkt->data.psnr.psnr[0];
+ nframes_++;
+ }
+
+ virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
+ ::libvpx_test::Encoder *encoder) {
+ if (video->frame() == 1) {
+ encoder->Control(VP8E_SET_CPUUSED, kCpuUsed);
+ if (encoding_mode_ != ::libvpx_test::kRealTime) {
+ encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
+ encoder->Control(VP8E_SET_ARNR_MAXFRAMES, 7);
+ encoder->Control(VP8E_SET_ARNR_STRENGTH, 5);
+ encoder->Control(VP8E_SET_ARNR_TYPE, 3);
+ }
+ }
+ }
+
+ double GetAveragePsnr() const {
+ if (nframes_)
+ return psnr_ / nframes_;
+ return 0.0;
+ }
+
+ TestVideoParam test_video_param_;
+
+ private:
+ double psnr_;
+ unsigned int nframes_;
+ libvpx_test::TestMode encoding_mode_;
+};
+
+TEST_P(EndToEndTestLarge, EndtoEndPSNRTest) {
+ cfg_.rc_target_bitrate = kBitrate;
+ cfg_.g_error_resilient = 0;
+ cfg_.g_profile = test_video_param_.profile;
+ cfg_.g_input_bit_depth = test_video_param_.input_bit_depth;
+ cfg_.g_bit_depth = test_video_param_.bit_depth;
+ init_flags_ = VPX_CODEC_USE_PSNR;
+
+ libvpx_test::VideoSource *video;
+ if (is_extension_y4m(test_video_param_.filename)) {
+ video = new libvpx_test::Y4mVideoSource(test_video_param_.filename,
+ 0, kFrames);
+ } else {
+ video = new libvpx_test::YUVVideoSource(test_video_param_.filename,
+ test_video_param_.fmt,
+ kWidth, kHeight,
+ kFramerate, 1, 0, kFrames);
+ }
+
+ ASSERT_NO_FATAL_FAILURE(RunLoop(video));
+ const double psnr = GetAveragePsnr();
+ EXPECT_GT(psnr, psnr_threshold);
+ delete(video);
+}
+
+VP9_INSTANTIATE_TEST_CASE(
+ EndToEndTestLarge,
+ ::testing::Values(::libvpx_test::kTwoPassGood, ::libvpx_test::kOnePassGood),
+ ::testing::ValuesIn(TestVectors));
+
+} // namespace
diff --git a/test/yuv_video_source.h b/test/yuv_video_source.h
new file mode 100644
index 0000000..3c852b2
--- /dev/null
+++ b/test/yuv_video_source.h
@@ -0,0 +1,151 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+#ifndef TEST_YUV_VIDEO_SOURCE_H_
+#define TEST_YUV_VIDEO_SOURCE_H_
+
+#include <cstdio>
+#include <cstdlib>
+#include <string>
+
+#include "test/video_source.h"
+#include "vpx/vpx_image.h"
+
+namespace libvpx_test {
+
+// This class extends VideoSource to allow parsing of raw YUV
+// formats of various color sampling and bit-depths so that we can
+// do actual file encodes.
+class YUVVideoSource : public VideoSource {
+ public:
+ YUVVideoSource(const std::string &file_name, vpx_img_fmt format,
+ unsigned int width, unsigned int height,
+ int rate_numerator, int rate_denominator,
+ unsigned int start, int limit)
+ : file_name_(file_name),
+ input_file_(NULL),
+ img_(NULL),
+ start_(start),
+ limit_(limit),
+ frame_(0),
+ width_(0),
+ height_(0),
+ format_(VPX_IMG_FMT_NONE),
+ framerate_numerator_(rate_numerator),
+ framerate_denominator_(rate_denominator) {
+ // This initializes format_, raw_size_, width_, height_ and allocates img.
+ SetSize(width, height, format);
+ }
+
+ virtual ~YUVVideoSource() {
+ vpx_img_free(img_);
+ if (input_file_)
+ fclose(input_file_);
+ }
+
+ virtual void Begin() {
+ if (input_file_)
+ fclose(input_file_);
+ input_file_ = OpenTestDataFile(file_name_);
+ ASSERT_TRUE(input_file_ != NULL) << "Input file open failed. Filename: "
+ << file_name_;
+ if (start_)
+ fseek(input_file_, static_cast<unsigned>(raw_size_) * start_, SEEK_SET);
+
+ frame_ = start_;
+ FillFrame();
+ }
+
+ virtual void Next() {
+ ++frame_;
+ FillFrame();
+ }
+
+ virtual vpx_image_t *img() const { return (frame_ < limit_) ? img_ : NULL; }
+
+ // Models a stream where Timebase = 1/FPS, so pts == frame.
+ virtual vpx_codec_pts_t pts() const { return frame_; }
+
+ virtual unsigned long duration() const { return 1; }
+
+ virtual vpx_rational_t timebase() const {
+ const vpx_rational_t t = { framerate_denominator_, framerate_numerator_ };
+ return t;
+ }
+
+ virtual unsigned int frame() const { return frame_; }
+
+ virtual unsigned int limit() const { return limit_; }
+
+ virtual void SetSize(unsigned int width, unsigned int height,
+ vpx_img_fmt format) {
+ if (width != width_ || height != height_ || format != format_) {
+ vpx_img_free(img_);
+ img_ = vpx_img_alloc(NULL, format, width, height, 1);
+ ASSERT_TRUE(img_ != NULL);
+ width_ = width;
+ height_ = height;
+ format_ = format;
+ switch (format) {
+ case VPX_IMG_FMT_I420:
+ raw_size_ = width * height * 3 / 2;
+ break;
+ case VPX_IMG_FMT_I422:
+ raw_size_ = width * height * 2;
+ break;
+ case VPX_IMG_FMT_I440:
+ raw_size_ = width * height * 2;
+ break;
+ case VPX_IMG_FMT_I444:
+ raw_size_ = width * height * 3;
+ break;
+ case VPX_IMG_FMT_I42016:
+ raw_size_ = width * height * 3;
+ break;
+ case VPX_IMG_FMT_I42216:
+ raw_size_ = width * height * 4;
+ break;
+ case VPX_IMG_FMT_I44016:
+ raw_size_ = width * height * 4;
+ break;
+ case VPX_IMG_FMT_I44416:
+ raw_size_ = width * height * 6;
+ break;
+ default:
+ ASSERT_TRUE(0);
+ }
+ }
+ }
+
+ virtual void FillFrame() {
+ ASSERT_TRUE(input_file_ != NULL);
+ // Read a frame from input_file.
+ if (fread(img_->img_data, raw_size_, 1, input_file_) == 0) {
+ limit_ = frame_;
+ }
+ }
+
+ protected:
+ std::string file_name_;
+ FILE *input_file_;
+ vpx_image_t *img_;
+ size_t raw_size_;
+ unsigned int start_;
+ unsigned int limit_;
+ unsigned int frame_;
+ unsigned int width_;
+ unsigned int height_;
+ vpx_img_fmt format_;
+ int framerate_numerator_;
+ int framerate_denominator_;
+};
+
+} // namespace libvpx_test
+
+#endif // TEST_YUV_VIDEO_SOURCE_H_
diff --git a/tools_common.c b/tools_common.c
index 30fd352..e243a91 100644
--- a/tools_common.c
+++ b/tools_common.c
@@ -276,7 +276,7 @@
// Note the offset is 1 less than half.
const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
int plane;
- if (dst->w != src->w || dst->h != src->h ||
+ if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
dst->x_chroma_shift != src->x_chroma_shift ||
dst->y_chroma_shift != src->y_chroma_shift ||
dst->fmt != src->fmt || input_shift < 0) {
@@ -293,12 +293,12 @@
break;
}
for (plane = 0; plane < 3; plane++) {
- int w = src->w;
- int h = src->h;
+ int w = src->d_w;
+ int h = src->d_h;
int x, y;
if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
+ w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
+ h = (h + src->y_chroma_shift) >> src->y_chroma_shift;
}
for (y = 0; y < h; y++) {
uint16_t *p_src =
@@ -316,7 +316,7 @@
// Note the offset is 1 less than half.
const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
int plane;
- if (dst->w != src->w || dst->h != src->h ||
+ if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
dst->x_chroma_shift != src->x_chroma_shift ||
dst->y_chroma_shift != src->y_chroma_shift ||
dst->fmt != src->fmt + VPX_IMG_FMT_HIGHBITDEPTH ||
@@ -334,8 +334,8 @@
break;
}
for (plane = 0; plane < 3; plane++) {
- int w = src->w;
- int h = src->h;
+ int w = src->d_w;
+ int h = src->d_h;
int x, y;
if (plane) {
w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
@@ -384,8 +384,8 @@
int h = src->d_h;
int x, y;
if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
+ w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
+ h = (h + src->y_chroma_shift) >> src->y_chroma_shift;
}
for (y = 0; y < h; y++) {
uint16_t *p_src =
diff --git a/vp8/encoder/pickinter.c b/vp8/encoder/pickinter.c
index ea3b46e..9d5556d 100644
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -516,9 +516,8 @@
// Adjust rd for ZEROMV and LAST, if LAST is the closest reference frame.
if (this_mode == ZEROMV &&
x->e_mbd.mode_info_context->mbmi.ref_frame == LAST_FRAME &&
- (denoise_aggressive || cpi->closest_reference_frame == LAST_FRAME))
- {
- this_rd = ((int64_t)this_rd) * rd_adj / 100;
+ (denoise_aggressive || cpi->closest_reference_frame == LAST_FRAME)) {
+ this_rd = ((int64_t)this_rd) * rd_adj / 100;
}
check_for_encode_breakout(*sse, x);
diff --git a/vp8/vp8_dx_iface.c b/vp8/vp8_dx_iface.c
index 3ab8ed0..5aa274d 100644
--- a/vp8/vp8_dx_iface.c
+++ b/vp8/vp8_dx_iface.c
@@ -112,22 +112,19 @@
* structure. More memory may be required at the time the stream
* information becomes known.
*/
- if (!ctx->priv)
- {
- vp8_init_ctx(ctx);
- priv = (vpx_codec_alg_priv_t *)ctx->priv;
+ if (!ctx->priv) {
+ vp8_init_ctx(ctx);
+ priv = (vpx_codec_alg_priv_t *)ctx->priv;
- /* initialize number of fragments to zero */
- priv->fragments.count = 0;
- /* is input fragments enabled? */
- priv->fragments.enabled =
- (priv->base.init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS);
+ /* initialize number of fragments to zero */
+ priv->fragments.count = 0;
+ /* is input fragments enabled? */
+ priv->fragments.enabled =
+ (priv->base.init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS);
- /*post processing level initialized to do nothing */
- }
- else
- {
- priv = (vpx_codec_alg_priv_t *)ctx->priv;
+ /*post processing level initialized to do nothing */
+ } else {
+ priv = (vpx_codec_alg_priv_t *)ctx->priv;
}
priv->yv12_frame_buffers.use_frame_threads =
@@ -138,11 +135,10 @@
if (priv->yv12_frame_buffers.use_frame_threads &&
((ctx->priv->init_flags & VPX_CODEC_USE_ERROR_CONCEALMENT) ||
- (ctx->priv->init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS)))
- {
- /* row-based threading, error concealment, and input fragments will
- * not be supported when using frame-based threading */
- res = VPX_CODEC_INVALID_PARAM;
+ (ctx->priv->init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS))) {
+ /* row-based threading, error concealment, and input fragments will
+ * not be supported when using frame-based threading */
+ res = VPX_CODEC_INVALID_PARAM;
}
return res;
diff --git a/vp9/common/vp9_blockd.c b/vp9/common/vp9_blockd.c
index e13445f..7094a01 100644
--- a/vp9/common/vp9_blockd.c
+++ b/vp9/common/vp9_blockd.c
@@ -92,7 +92,7 @@
void *arg) {
int plane;
- for (plane = 0; plane < MAX_MB_PLANE; plane++)
+ for (plane = 0; plane < MAX_MB_PLANE; ++plane)
vp9_foreach_transformed_block_in_plane(xd, bsize, plane, visit, arg);
}
diff --git a/vp9/common/vp9_mvref_common.c b/vp9/common/vp9_mvref_common.c
index fbb3d4b..3b34050 100644
--- a/vp9/common/vp9_mvref_common.c
+++ b/vp9/common/vp9_mvref_common.c
@@ -24,10 +24,7 @@
? cm->prev_mi[mi_row * xd->mi_stride + mi_col].src_mi
: NULL;
const MB_MODE_INFO *const prev_mbmi = prev_mi ? &prev_mi->src_mi->mbmi : NULL;
-
-
const POSITION *const mv_ref_search = mv_ref_blocks[mi->mbmi.sb_type];
-
int different_ref_found = 0;
int context_counter = 0;
@@ -127,7 +124,6 @@
}
}
-
void vp9_find_best_ref_mvs(MACROBLOCKD *xd, int allow_hp,
int_mv *mvlist, int_mv *nearest, int_mv *near) {
int i;
diff --git a/vp9/common/vp9_onyxc_int.h b/vp9/common/vp9_onyxc_int.h
index c28f156..f1eda91 100644
--- a/vp9/common/vp9_onyxc_int.h
+++ b/vp9/common/vp9_onyxc_int.h
@@ -143,7 +143,6 @@
int prev_mi_idx;
int mi_alloc_size;
MODE_INFO *mip_array[2];
- MODE_INFO **mi_grid_base_array[2];
MODE_INFO *mip; /* Base of allocated array */
MODE_INFO *mi; /* Corresponds to upper left visible macroblock */
diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl
index c2a9181..de389e7 100644
--- a/vp9/common/vp9_rtcd_defs.pl
+++ b/vp9/common/vp9_rtcd_defs.pl
@@ -1155,7 +1155,7 @@
specialize qw/vp9_quantize_fp_32x32/, "$ssse3_x86_64";
add_proto qw/void vp9_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
- specialize qw/vp9_quantize_b/, "$ssse3_x86_64";
+ specialize qw/vp9_quantize_b sse2/, "$ssse3_x86_64";
add_proto qw/void vp9_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
specialize qw/vp9_quantize_b_32x32/, "$ssse3_x86_64";
diff --git a/vp9/decoder/vp9_decodeframe.c b/vp9/decoder/vp9_decodeframe.c
index 2690f48..dc712f0 100644
--- a/vp9/decoder/vp9_decodeframe.c
+++ b/vp9/decoder/vp9_decodeframe.c
@@ -656,10 +656,8 @@
void vp9_read_frame_size(struct vp9_read_bit_buffer *rb,
int *width, int *height) {
- const int w = vp9_rb_read_literal(rb, 16) + 1;
- const int h = vp9_rb_read_literal(rb, 16) + 1;
- *width = w;
- *height = h;
+ *width = vp9_rb_read_literal(rb, 16) + 1;
+ *height = vp9_rb_read_literal(rb, 16) + 1;
}
static void setup_display_size(VP9_COMMON *cm, struct vp9_read_bit_buffer *rb) {
@@ -749,7 +747,7 @@
if (!found)
vp9_read_frame_size(rb, &width, &height);
- if (width <=0 || height <= 0)
+ if (width <= 0 || height <= 0)
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
"Invalid frame size");
diff --git a/vp9/decoder/vp9_decodemv.c b/vp9/decoder/vp9_decodemv.c
index 3c8e7cc..a01fe84 100644
--- a/vp9/decoder/vp9_decodemv.c
+++ b/vp9/decoder/vp9_decodemv.c
@@ -223,7 +223,6 @@
fr = vp9_read_tree(r, vp9_mv_fp_tree, class0 ? mvcomp->class0_fp[d]
: mvcomp->fp);
-
// High precision part (if hp is not used, the default value of the hp is 1)
hp = usehp ? vp9_read(r, class0 ? mvcomp->class0_hp : mvcomp->hp)
: 1;
diff --git a/vp9/encoder/vp9_bitstream.c b/vp9/encoder/vp9_bitstream.c
index 694cac7..3954fe6 100644
--- a/vp9/encoder/vp9_bitstream.c
+++ b/vp9/encoder/vp9_bitstream.c
@@ -1013,7 +1013,11 @@
((cpi->svc.number_temporal_layers > 1 &&
cpi->oxcf.rc_mode == VPX_CBR) ||
(cpi->svc.number_spatial_layers > 1 &&
- cpi->svc.layer_context[cpi->svc.spatial_layer_id].is_key_frame))) {
+ cpi->svc.layer_context[cpi->svc.spatial_layer_id].is_key_frame) ||
+ (is_two_pass_svc(cpi) &&
+ cpi->svc.encode_empty_frame_state == ENCODING &&
+ cpi->svc.layer_context[0].frames_from_key_frame <
+ cpi->svc.number_temporal_layers + 1))) {
found = 0;
}
vp9_wb_write_bit(wb, found);
@@ -1105,8 +1109,7 @@
// will change to show_frame flag to 0, then add an one byte frame with
// show_existing_frame flag which tells the decoder which frame we want to
// show.
- if (!cm->show_frame ||
- (is_two_pass_svc(cpi) && cm->error_resilient_mode == 0))
+ if (!cm->show_frame)
vp9_wb_write_bit(wb, cm->intra_only);
if (!cm->error_resilient_mode)
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index a867cf9..b4fe70a 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -412,29 +412,47 @@
return 1;
}
- // Vertical split is available on all but the bottom border.
- if (mi_row + block_height / 2 < cm->mi_rows &&
- vt.part_variances->vert[0].variance < threshold &&
- vt.part_variances->vert[1].variance < threshold) {
- BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_VERT);
- set_block_size(cpi, mi_row, mi_col, subsize);
- set_block_size(cpi, mi_row, mi_col + block_width / 2, subsize);
- return 1;
+ // Only allow split for blocks above 16x16.
+ if (bsize > BLOCK_16X16) {
+ // Vertical split is available on all but the bottom border.
+ if (mi_row + block_height / 2 < cm->mi_rows &&
+ vt.part_variances->vert[0].variance < threshold &&
+ vt.part_variances->vert[1].variance < threshold) {
+ BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_VERT);
+ set_block_size(cpi, mi_row, mi_col, subsize);
+ set_block_size(cpi, mi_row, mi_col + block_width / 2, subsize);
+ return 1;
+ }
+
+ // Horizontal split is available on all but the right border.
+ if (mi_col + block_width / 2 < cm->mi_cols &&
+ vt.part_variances->horz[0].variance < threshold &&
+ vt.part_variances->horz[1].variance < threshold) {
+ BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_HORZ);
+ set_block_size(cpi, mi_row, mi_col, subsize);
+ set_block_size(cpi, mi_row + block_height / 2, mi_col, subsize);
+ return 1;
+ }
}
- // Horizontal split is available on all but the right border.
- if (mi_col + block_width / 2 < cm->mi_cols &&
- vt.part_variances->horz[0].variance < threshold &&
- vt.part_variances->horz[1].variance < threshold) {
- BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_HORZ);
- set_block_size(cpi, mi_row, mi_col, subsize);
- set_block_size(cpi, mi_row + block_height / 2, mi_col, subsize);
- return 1;
+ // This will only allow 8x8 if the 16x16 variance is very large.
+ if (bsize == BLOCK_16X16) {
+ if (mi_col + block_width / 2 < cm->mi_cols &&
+ mi_row + block_height / 2 < cm->mi_rows &&
+ vt.part_variances->none.variance < (threshold << 6)) {
+ set_block_size(cpi, mi_row, mi_col, bsize);
+ return 1;
+ }
}
return 0;
}
-// TODO(debargha): Fix this function and make it work as expected.
+// This function chooses partitioning based on the variance
+// between source and reconstructed last, where variance is
+// computed for 8x8 downsampled inputs. Some things to check:
+// using the last source rather than reconstructed last, and
+// allowing for small downsampling (4x4 or 2x2) for selection
+// of smaller block sizes (i.e., < 16x16).
static void choose_partitioning(VP9_COMP *cpi,
const TileInfo *const tile,
int mi_row, int mi_col) {
@@ -549,27 +567,11 @@
for (j = 0; j < 4; ++j) {
const int x16_idx = ((j & 1) << 1);
const int y16_idx = ((j >> 1) << 1);
- // NOTE: This is a temporary hack to disable 8x8 partitions,
- // since it works really bad - possibly due to a bug
-#define DISABLE_8X8_VAR_BASED_PARTITION
-#ifdef DISABLE_8X8_VAR_BASED_PARTITION
- if (mi_row + y32_idx + y16_idx + 1 < cm->mi_rows &&
- mi_row + x32_idx + x16_idx + 1 < cm->mi_cols) {
- set_block_size(cpi,
- (mi_row + y32_idx + y16_idx),
- (mi_col + x32_idx + x16_idx),
- BLOCK_16X16);
- } else {
- for (k = 0; k < 4; ++k) {
- const int x8_idx = (k & 1);
- const int y8_idx = (k >> 1);
- set_block_size(cpi,
- (mi_row + y32_idx + y16_idx + y8_idx),
- (mi_col + x32_idx + x16_idx + x8_idx),
- BLOCK_8X8);
- }
- }
-#else
+ // NOTE: Since this uses 8x8 downsampling for variance calculation
+ // we cannot really select block size 8x8 (or even 8x16/16x8),
+ // since we do not sufficient samples for variance.
+ // For now, 8x8 partition is only set if the variance of the 16x16
+ // block is very high. This is controlled in set_vt_partitioning.
if (!set_vt_partitioning(cpi, &vt.split[i].split[j],
BLOCK_16X16,
mi_row + y32_idx + y16_idx,
@@ -583,7 +585,6 @@
BLOCK_8X8);
}
}
-#endif
}
}
}
@@ -3156,7 +3157,6 @@
int64_t dummy_dist = 0;
const int idx_str = cm->mi_stride * mi_row + mi_col;
MODE_INFO *mi = cm->mi + idx_str;
- MODE_INFO *prev_mi = (cm->prev_mip + cm->mi_stride + 1 + idx_str)->src_mi;
BLOCK_SIZE bsize;
x->in_static_area = 0;
x->source_variance = UINT_MAX;
@@ -3194,7 +3194,7 @@
&dummy_rate, &dummy_dist, 1, INT64_MAX,
cpi->pc_root);
} else {
- copy_partitioning(cm, mi, prev_mi);
+ choose_partitioning(cpi, tile, mi_row, mi_col);
nonrd_use_partition(cpi, tile, mi, tp, mi_row, mi_col,
BLOCK_64X64, 1, &dummy_rate, &dummy_dist,
cpi->pc_root);
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index 8096a90..1758e3f 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -225,6 +225,9 @@
}
vpx_memset(&cpi->svc.scaled_frames[0], 0,
MAX_LAG_BUFFERS * sizeof(cpi->svc.scaled_frames[0]));
+
+ vp9_free_frame_buffer(&cpi->svc.empty_frame.img);
+ vpx_memset(&cpi->svc.empty_frame, 0, sizeof(cpi->svc.empty_frame));
}
static void save_coding_context(VP9_COMP *cpi) {
@@ -585,8 +588,6 @@
cpi->ref_frame_flags = 0;
init_buffer_indices(cpi);
-
- set_tile_limits(cpi);
}
static void set_rc_buffer_sizes(RATE_CONTROL *rc,
@@ -2981,7 +2982,9 @@
}
}
if (is_two_pass_svc(cpi) && cm->error_resilient_mode == 0) {
+ // Use the last frame context for the empty frame.
cm->frame_context_idx =
+ (cpi->svc.encode_empty_frame_state == ENCODING) ? FRAME_CONTEXTS - 1 :
cpi->svc.spatial_layer_id * cpi->svc.number_temporal_layers +
cpi->svc.temporal_layer_id;
@@ -3162,7 +3165,9 @@
cpi->ref_frame_flags = get_ref_frame_flags(cpi);
cm->last_frame_type = cm->frame_type;
- vp9_rc_postencode_update(cpi, *size);
+
+ if (!(is_two_pass_svc(cpi) && cpi->svc.encode_empty_frame_state == ENCODING))
+ vp9_rc_postencode_update(cpi, *size);
#if 0
output_frame_level_debug_stats(cpi);
@@ -3186,12 +3191,8 @@
cm->last_height = cm->height;
// reset to normal state now that we are done.
- if (!cm->show_existing_frame) {
- if (is_two_pass_svc(cpi) && cm->error_resilient_mode == 0)
- cm->last_show_frame = 0;
- else
- cm->last_show_frame = cm->show_frame;
- }
+ if (!cm->show_existing_frame)
+ cm->last_show_frame = cm->show_frame;
if (cm->show_frame) {
vp9_swap_mi_and_prev_mi(cm);
@@ -3228,7 +3229,9 @@
uint8_t *dest, unsigned int *frame_flags) {
cpi->allow_encode_breakout = ENCODE_BREAKOUT_ENABLED;
encode_frame_to_data_rate(cpi, size, dest, frame_flags);
- vp9_twopass_postencode_update(cpi);
+
+ if (!(is_two_pass_svc(cpi) && cpi->svc.encode_empty_frame_state == ENCODING))
+ vp9_twopass_postencode_update(cpi);
}
static void init_motion_estimation(VP9_COMP *cpi) {
@@ -3416,6 +3419,9 @@
if (is_two_pass_svc(cpi)) {
#if CONFIG_SPATIAL_SVC
vp9_svc_start_frame(cpi);
+ // Use a small empty frame instead of a real frame
+ if (cpi->svc.encode_empty_frame_state == ENCODING)
+ source = &cpi->svc.empty_frame;
#endif
if (oxcf->pass == 2)
vp9_restore_layer_context(cpi);
@@ -3434,6 +3440,11 @@
// Should we encode an arf frame.
arf_src_index = get_arf_src_index(cpi);
+
+ // Skip alt frame if we encode the empty frame
+ if (is_two_pass_svc(cpi) && source != NULL)
+ arf_src_index = 0;
+
if (arf_src_index) {
assert(arf_src_index <= rc->frames_to_key);
@@ -3544,7 +3555,10 @@
// For two pass encodes analyse the first pass stats and determine
// the bit allocation and other parameters for this frame / group of frames.
- if ((oxcf->pass == 2) && (!cpi->use_svc || is_two_pass_svc(cpi))) {
+ if ((oxcf->pass == 2) &&
+ (!cpi->use_svc ||
+ (is_two_pass_svc(cpi) &&
+ cpi->svc.encode_empty_frame_state != ENCODING))) {
vp9_rc_get_second_pass_params(cpi);
}
@@ -3773,10 +3787,18 @@
#endif
- if (is_two_pass_svc(cpi) && cm->show_frame) {
- ++cpi->svc.spatial_layer_to_encode;
- if (cpi->svc.spatial_layer_to_encode >= cpi->svc.number_spatial_layers)
- cpi->svc.spatial_layer_to_encode = 0;
+ if (is_two_pass_svc(cpi)) {
+ if (cpi->svc.encode_empty_frame_state == ENCODING)
+ cpi->svc.encode_empty_frame_state = ENCODED;
+
+ if (cm->show_frame) {
+ ++cpi->svc.spatial_layer_to_encode;
+ if (cpi->svc.spatial_layer_to_encode >= cpi->svc.number_spatial_layers)
+ cpi->svc.spatial_layer_to_encode = 0;
+
+ // May need the empty frame after an visible frame.
+ cpi->svc.encode_empty_frame_state = NEED_TO_ENCODE;
+ }
}
return 0;
}
@@ -3867,10 +3889,6 @@
if (width) {
cm->width = width;
- if (cm->width * 5 < cpi->initial_width) {
- cm->width = cpi->initial_width / 5 + 1;
- printf("Warning: Desired width too small, changed to %d\n", cm->width);
- }
if (cm->width > cpi->initial_width) {
cm->width = cpi->initial_width;
printf("Warning: Desired width too large, changed to %d\n", cm->width);
@@ -3879,10 +3897,6 @@
if (height) {
cm->height = height;
- if (cm->height * 5 < cpi->initial_height) {
- cm->height = cpi->initial_height / 5 + 1;
- printf("Warning: Desired height too small, changed to %d\n", cm->height);
- }
if (cm->height > cpi->initial_height) {
cm->height = cpi->initial_height;
printf("Warning: Desired height too large, changed to %d\n", cm->height);
diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c
index c9588a3..96c3e0a 100644
--- a/vp9/encoder/vp9_firstpass.c
+++ b/vp9/encoder/vp9_firstpass.c
@@ -2405,6 +2405,9 @@
cpi->ref_frame_flags &=
(~VP9_LAST_FLAG & ~VP9_GOLD_FLAG & ~VP9_ALT_FLAG);
lc->frames_from_key_frame = 0;
+ // Reset the empty frame resolution since we have a key frame.
+ cpi->svc.empty_frame_width = cm->width;
+ cpi->svc.empty_frame_height = cm->height;
}
} else {
cm->frame_type = INTER_FRAME;
@@ -2478,6 +2481,7 @@
if (rc->total_actual_bits) {
rc->rate_error_estimate =
(int)((rc->vbr_bits_off_target * 100) / rc->total_actual_bits);
+ rc->rate_error_estimate = clamp(rc->rate_error_estimate, -100, 100);
} else {
rc->rate_error_estimate = 0;
}
diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c
index 42f4691..b74b2dd 100644
--- a/vp9/encoder/vp9_pickmode.c
+++ b/vp9/encoder/vp9_pickmode.c
@@ -486,8 +486,12 @@
// var_y and sse_y are saved to be used in skipping checking
unsigned int var_y = UINT_MAX;
unsigned int sse_y = UINT_MAX;
+ // Reduce the intra cost penalty for small blocks (<=16x16).
+ const int reduction_fac =
+ (cpi->sf.partition_search_type == VAR_BASED_PARTITION &&
+ bsize <= BLOCK_16X16) ? 4 : 1;
const int intra_cost_penalty = vp9_get_intra_cost_penalty(
- cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
+ cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth) / reduction_fac;
const int64_t inter_mode_thresh = RDCOST(x->rdmult, x->rddiv,
intra_cost_penalty, 0);
const int intra_mode_cost = 50;
diff --git a/vp9/encoder/vp9_ratectrl.c b/vp9/encoder/vp9_ratectrl.c
index ef32fe1..65bca66 100644
--- a/vp9/encoder/vp9_ratectrl.c
+++ b/vp9/encoder/vp9_ratectrl.c
@@ -1500,9 +1500,7 @@
rc->max_gf_interval = 16;
// Extended interval for genuinely static scenes
- rc->static_scene_max_gf_interval = oxcf->key_freq >> 1;
- if (rc->static_scene_max_gf_interval > (MAX_LAG_BUFFERS * 2))
- rc->static_scene_max_gf_interval = MAX_LAG_BUFFERS * 2;
+ rc->static_scene_max_gf_interval = MAX_LAG_BUFFERS * 2;
if (is_altref_enabled(cpi)) {
if (rc->static_scene_max_gf_interval > oxcf->lag_in_frames - 1)
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 7565cc5..eca8e58 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -1140,12 +1140,14 @@
*sse = 0;
*skippable = 1;
- for (plane = 1; plane < MAX_MB_PLANE && is_cost_valid; ++plane) {
+ for (plane = 1; plane < MAX_MB_PLANE; ++plane) {
txfm_rd_in_plane(x, &pnrate, &pndist, &pnskip, &pnsse,
ref_best_rd, plane, bsize, uv_tx_size,
cpi->sf.use_fast_coef_costing);
- if (pnrate == INT_MAX)
+ if (pnrate == INT_MAX) {
is_cost_valid = 0;
+ break;
+ }
*rate += pnrate;
*distortion += pndist;
*sse += pnsse;
@@ -3392,6 +3394,7 @@
}
if (best_mode_index < 0 || best_rd >= best_rd_so_far) {
+ rd_cost->rate = INT_MAX;
rd_cost->rdcost = INT64_MAX;
return;
}
@@ -3562,6 +3565,7 @@
rd_cost->rdcost = this_rd;
if (this_rd >= best_rd_so_far) {
+ rd_cost->rate = INT_MAX;
rd_cost->rdcost = INT64_MAX;
return;
}
@@ -4113,6 +4117,7 @@
}
if (best_rd >= best_rd_so_far) {
+ rd_cost->rate = INT_MAX;
rd_cost->rdcost = INT64_MAX;
return;
}
diff --git a/vp9/encoder/vp9_speed_features.c b/vp9/encoder/vp9_speed_features.c
index 8788be6..bec77d7 100644
--- a/vp9/encoder/vp9_speed_features.c
+++ b/vp9/encoder/vp9_speed_features.c
@@ -271,6 +271,10 @@
sf->partition_search_type = REFERENCE_PARTITION;
sf->use_nonrd_pick_mode = 1;
sf->allow_skip_recode = 0;
+ sf->inter_mode_mask[BLOCK_32X32] = INTER_NEAREST_NEW_ZERO;
+ sf->inter_mode_mask[BLOCK_32X64] = INTER_NEAREST_NEW_ZERO;
+ sf->inter_mode_mask[BLOCK_64X32] = INTER_NEAREST_NEW_ZERO;
+ sf->inter_mode_mask[BLOCK_64X64] = INTER_NEAREST_NEW_ZERO;
}
if (speed >= 6) {
@@ -285,10 +289,6 @@
sf->partition_search_type = VAR_BASED_PARTITION;
sf->search_type_check_frequency = 50;
sf->mv.search_method = NSTEP;
- sf->inter_mode_mask[BLOCK_32X32] = INTER_NEAREST_NEW_ZERO;
- sf->inter_mode_mask[BLOCK_32X64] = INTER_NEAREST_NEW_ZERO;
- sf->inter_mode_mask[BLOCK_64X32] = INTER_NEAREST_NEW_ZERO;
- sf->inter_mode_mask[BLOCK_64X64] = INTER_NEAREST_NEW_ZERO;
sf->tx_size_search_method = is_keyframe ? USE_LARGESTALL : USE_TX_8X8;
diff --git a/vp9/encoder/vp9_svc_layercontext.c b/vp9/encoder/vp9_svc_layercontext.c
index eed681c..1573557 100644
--- a/vp9/encoder/vp9_svc_layercontext.c
+++ b/vp9/encoder/vp9_svc_layercontext.c
@@ -14,6 +14,8 @@
#include "vp9/encoder/vp9_svc_layercontext.h"
#include "vp9/encoder/vp9_extend.h"
+#define SMALL_FRAME_FB_IDX 7
+
void vp9_init_layer_context(VP9_COMP *const cpi) {
SVC *const svc = &cpi->svc;
const VP9EncoderConfig *const oxcf = &cpi->oxcf;
@@ -28,6 +30,25 @@
layer_end = svc->number_temporal_layers;
} else {
layer_end = svc->number_spatial_layers;
+
+ if (cpi->oxcf.error_resilient_mode == 0 && cpi->oxcf.pass == 2) {
+ if (vp9_realloc_frame_buffer(&cpi->svc.empty_frame.img,
+ cpi->common.width, cpi->common.height,
+ cpi->common.subsampling_x,
+ cpi->common.subsampling_y,
+#if CONFIG_VP9_HIGHBITDEPTH
+ cpi->common.use_highbitdepth,
+#endif
+ VP9_ENC_BORDER_IN_PIXELS, NULL, NULL, NULL))
+ vpx_internal_error(&cpi->common.error, VPX_CODEC_MEM_ERROR,
+ "Failed to allocate empty frame for multiple frame "
+ "contexts");
+
+ vpx_memset(cpi->svc.empty_frame.img.buffer_alloc, 0x80,
+ cpi->svc.empty_frame.img.buffer_alloc_sz);
+ cpi->svc.empty_frame_width = cpi->common.width;
+ cpi->svc.empty_frame_height = cpi->common.height;
+ }
}
for (layer = 0; layer < layer_end; ++layer) {
@@ -310,6 +331,47 @@
get_layer_resolution(cpi->oxcf.width, cpi->oxcf.height,
lc->scaling_factor_num, lc->scaling_factor_den,
&width, &height);
+
+ // Workaround for multiple frame contexts. In some frames we can't use prev_mi
+ // since its previous frame could be changed during decoding time. The idea is
+ // we put a empty invisible frame in front of them, then we will not use
+ // prev_mi when encoding these frames.
+ if (cpi->oxcf.error_resilient_mode == 0 && cpi->oxcf.pass == 2 &&
+ cpi->svc.encode_empty_frame_state == NEED_TO_ENCODE) {
+ if ((cpi->svc.number_temporal_layers > 1 &&
+ cpi->svc.temporal_layer_id < cpi->svc.number_temporal_layers - 1) ||
+ (cpi->svc.number_spatial_layers > 1 &&
+ cpi->svc.spatial_layer_id == 0)) {
+ struct lookahead_entry *buf = vp9_lookahead_peek(cpi->lookahead, 0);
+
+ if (buf != NULL) {
+ cpi->svc.empty_frame.ts_start = buf->ts_start;
+ cpi->svc.empty_frame.ts_end = buf->ts_end;
+ cpi->svc.encode_empty_frame_state = ENCODING;
+ cpi->common.show_frame = 0;
+ cpi->ref_frame_flags = 0;
+ cpi->common.frame_type = INTER_FRAME;
+ cpi->lst_fb_idx =
+ cpi->gld_fb_idx = cpi->alt_fb_idx = SMALL_FRAME_FB_IDX;
+
+ // Gradually make the empty frame smaller to save bits. Make it half of
+ // its previous size because of the scaling factor restriction.
+ cpi->svc.empty_frame_width >>= 1;
+ cpi->svc.empty_frame_width = (cpi->svc.empty_frame_width + 1) & ~1;
+ if (cpi->svc.empty_frame_width < 16)
+ cpi->svc.empty_frame_width = 16;
+
+ cpi->svc.empty_frame_height >>= 1;
+ cpi->svc.empty_frame_height = (cpi->svc.empty_frame_height + 1) & ~1;
+ if (cpi->svc.empty_frame_height < 16)
+ cpi->svc.empty_frame_height = 16;
+
+ width = cpi->svc.empty_frame_width;
+ height = cpi->svc.empty_frame_height;
+ }
+ }
+ }
+
if (vp9_set_size_literal(cpi, width, height) != 0)
return VPX_CODEC_INVALID_PARAM;
@@ -317,7 +379,6 @@
cpi->oxcf.best_allowed_q = vp9_quantizer_to_qindex(lc->min_q);
vp9_change_config(cpi, &cpi->oxcf);
-
vp9_set_high_precision_mv(cpi, 1);
cpi->alt_ref_source = get_layer_context(cpi)->alt_ref_source;
diff --git a/vp9/encoder/vp9_svc_layercontext.h b/vp9/encoder/vp9_svc_layercontext.h
index 47a5456..e9645ce 100644
--- a/vp9/encoder/vp9_svc_layercontext.h
+++ b/vp9/encoder/vp9_svc_layercontext.h
@@ -50,6 +50,16 @@
int spatial_layer_to_encode;
+ // Workaround for multiple frame contexts
+ enum {
+ ENCODED = 0,
+ ENCODING,
+ NEED_TO_ENCODE
+ }encode_empty_frame_state;
+ struct lookahead_entry empty_frame;
+ int empty_frame_width;
+ int empty_frame_height;
+
// Store scaled source frames to be used for temporal filter to generate
// a alt ref frame.
YV12_BUFFER_CONFIG scaled_frames[MAX_LAG_BUFFERS];
diff --git a/vp9/encoder/vp9_temporal_filter.c b/vp9/encoder/vp9_temporal_filter.c
index 9ae81e7..5599227 100644
--- a/vp9/encoder/vp9_temporal_filter.c
+++ b/vp9/encoder/vp9_temporal_filter.c
@@ -719,6 +719,9 @@
++frame_used;
}
}
+ cm->mi = cm->mip + cm->mi_stride + 1;
+ cpi->mb.e_mbd.mi = cm->mi;
+ cpi->mb.e_mbd.mi[0].src_mi = &cpi->mb.e_mbd.mi[0];
} else {
// ARF is produced at the native frame size and resized when coded.
#if CONFIG_VP9_HIGHBITDEPTH
diff --git a/vp9/encoder/x86/vp9_denoiser_sse2.c b/vp9/encoder/x86/vp9_denoiser_sse2.c
index bf400d3..0a120e4 100644
--- a/vp9/encoder/x86/vp9_denoiser_sse2.c
+++ b/vp9/encoder/x86/vp9_denoiser_sse2.c
@@ -50,9 +50,9 @@
const __m128i l21,
__m128i acc_diff) {
// Calculate differences
- const __m128i v_sig = _mm_loadu_si128((__m128i *)(&sig[0]));
+ const __m128i v_sig = _mm_loadu_si128((const __m128i *)(&sig[0]));
const __m128i v_mc_running_avg_y = _mm_loadu_si128(
- (__m128i *)(&mc_running_avg_y[0]));
+ (const __m128i *)(&mc_running_avg_y[0]));
__m128i v_running_avg_y;
const __m128i pdiff = _mm_subs_epu8(v_mc_running_avg_y, v_sig);
const __m128i ndiff = _mm_subs_epu8(v_sig, v_mc_running_avg_y);
@@ -103,9 +103,9 @@
__m128i acc_diff) {
__m128i v_running_avg_y = _mm_loadu_si128((__m128i *)(&running_avg_y[0]));
// Calculate differences.
- const __m128i v_sig = _mm_loadu_si128((__m128i *)(&sig[0]));
+ const __m128i v_sig = _mm_loadu_si128((const __m128i *)(&sig[0]));
const __m128i v_mc_running_avg_y =
- _mm_loadu_si128((__m128i *)(&mc_running_avg_y[0]));
+ _mm_loadu_si128((const __m128i *)(&mc_running_avg_y[0]));
const __m128i pdiff = _mm_subs_epu8(v_mc_running_avg_y, v_sig);
const __m128i ndiff = _mm_subs_epu8(v_sig, v_mc_running_avg_y);
// Obtain the sign. FF if diff is negative.
@@ -357,9 +357,9 @@
const __m128i l21 = _mm_set1_epi8(1);
int sum_diff = 0;
- for (int i = 0; i < 4; i++) {
- for (int j = 0; j < 4; j++) {
- acc_diff[i][j] = _mm_setzero_si128();
+ for (c = 0; c < 4; ++c) {
+ for (r = 0; r < 4; ++r) {
+ acc_diff[c][r] = _mm_setzero_si128();
}
}
diff --git a/vp9/encoder/x86/vp9_quantize_sse2.c b/vp9/encoder/x86/vp9_quantize_sse2.c
new file mode 100644
index 0000000..7c1c884
--- /dev/null
+++ b/vp9/encoder/x86/vp9_quantize_sse2.c
@@ -0,0 +1,225 @@
+/*
+ * Copyright (c) 2014 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 <emmintrin.h>
+#include <xmmintrin.h>
+
+#include "vpx/vpx_integer.h"
+
+void vp9_quantize_b_sse2(const int16_t* coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t* zbin_ptr,
+ const int16_t* round_ptr, const int16_t* quant_ptr,
+ const int16_t* quant_shift_ptr, int16_t* qcoeff_ptr,
+ int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
+ int zbin_oq_value, uint16_t* eob_ptr,
+ const int16_t* scan_ptr,
+ const int16_t* iscan_ptr) {
+ __m128i zero;
+ (void)scan_ptr;
+
+ coeff_ptr += n_coeffs;
+ iscan_ptr += n_coeffs;
+ qcoeff_ptr += n_coeffs;
+ dqcoeff_ptr += n_coeffs;
+ n_coeffs = -n_coeffs;
+ zero = _mm_setzero_si128();
+ if (!skip_block) {
+ __m128i eob;
+ __m128i zbin;
+ __m128i round, quant, dequant, shift;
+ {
+ __m128i coeff0, coeff1;
+
+ // Setup global values
+ {
+ __m128i zbin_oq;
+ __m128i pw_1;
+ zbin_oq = _mm_set1_epi16(zbin_oq_value);
+ zbin = _mm_load_si128((const __m128i*)zbin_ptr);
+ round = _mm_load_si128((const __m128i*)round_ptr);
+ quant = _mm_load_si128((const __m128i*)quant_ptr);
+ zbin = _mm_add_epi16(zbin, zbin_oq);
+ pw_1 = _mm_set1_epi16(1);
+ zbin = _mm_sub_epi16(zbin, pw_1);
+ dequant = _mm_load_si128((const __m128i*)dequant_ptr);
+ shift = _mm_load_si128((const __m128i*)quant_shift_ptr);
+ }
+
+ {
+ __m128i coeff0_sign, coeff1_sign;
+ __m128i qcoeff0, qcoeff1;
+ __m128i qtmp0, qtmp1;
+ __m128i cmp_mask0, cmp_mask1;
+ // Do DC and first 15 AC
+ coeff0 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs));
+ coeff1 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs) + 1);
+
+ // Poor man's sign extract
+ coeff0_sign = _mm_srai_epi16(coeff0, 15);
+ coeff1_sign = _mm_srai_epi16(coeff1, 15);
+ qcoeff0 = _mm_xor_si128(coeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(coeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ cmp_mask0 = _mm_cmpgt_epi16(qcoeff0, zbin);
+ zbin = _mm_unpackhi_epi64(zbin, zbin); // Switch DC to AC
+ cmp_mask1 = _mm_cmpgt_epi16(qcoeff1, zbin);
+ qcoeff0 = _mm_adds_epi16(qcoeff0, round);
+ round = _mm_unpackhi_epi64(round, round);
+ qcoeff1 = _mm_adds_epi16(qcoeff1, round);
+ qtmp0 = _mm_mulhi_epi16(qcoeff0, quant);
+ quant = _mm_unpackhi_epi64(quant, quant);
+ qtmp1 = _mm_mulhi_epi16(qcoeff1, quant);
+ qtmp0 = _mm_add_epi16(qtmp0, qcoeff0);
+ qtmp1 = _mm_add_epi16(qtmp1, qcoeff1);
+ qcoeff0 = _mm_mulhi_epi16(qtmp0, shift);
+ shift = _mm_unpackhi_epi64(shift, shift);
+ qcoeff1 = _mm_mulhi_epi16(qtmp1, shift);
+
+ // Reinsert signs
+ qcoeff0 = _mm_xor_si128(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(qcoeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ // Mask out zbin threshold coeffs
+ qcoeff0 = _mm_and_si128(qcoeff0, cmp_mask0);
+ qcoeff1 = _mm_and_si128(qcoeff1, cmp_mask1);
+
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), qcoeff0);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
+
+ coeff0 = _mm_mullo_epi16(qcoeff0, dequant);
+ dequant = _mm_unpackhi_epi64(dequant, dequant);
+ coeff1 = _mm_mullo_epi16(qcoeff1, dequant);
+
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), coeff0);
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
+ }
+
+ {
+ // Scan for eob
+ __m128i zero_coeff0, zero_coeff1;
+ __m128i nzero_coeff0, nzero_coeff1;
+ __m128i iscan0, iscan1;
+ __m128i eob1;
+ zero_coeff0 = _mm_cmpeq_epi16(coeff0, zero);
+ zero_coeff1 = _mm_cmpeq_epi16(coeff1, zero);
+ nzero_coeff0 = _mm_cmpeq_epi16(zero_coeff0, zero);
+ nzero_coeff1 = _mm_cmpeq_epi16(zero_coeff1, zero);
+ iscan0 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs));
+ iscan1 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs) + 1);
+ // Add one to convert from indices to counts
+ iscan0 = _mm_sub_epi16(iscan0, nzero_coeff0);
+ iscan1 = _mm_sub_epi16(iscan1, nzero_coeff1);
+ eob = _mm_and_si128(iscan0, nzero_coeff0);
+ eob1 = _mm_and_si128(iscan1, nzero_coeff1);
+ eob = _mm_max_epi16(eob, eob1);
+ }
+ n_coeffs += 8 * 2;
+ }
+
+ // AC only loop
+ while (n_coeffs < 0) {
+ __m128i coeff0, coeff1;
+ {
+ __m128i coeff0_sign, coeff1_sign;
+ __m128i qcoeff0, qcoeff1;
+ __m128i qtmp0, qtmp1;
+ __m128i cmp_mask0, cmp_mask1;
+
+ coeff0 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs));
+ coeff1 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs) + 1);
+
+ // Poor man's sign extract
+ coeff0_sign = _mm_srai_epi16(coeff0, 15);
+ coeff1_sign = _mm_srai_epi16(coeff1, 15);
+ qcoeff0 = _mm_xor_si128(coeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(coeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ cmp_mask0 = _mm_cmpgt_epi16(qcoeff0, zbin);
+ cmp_mask1 = _mm_cmpgt_epi16(qcoeff1, zbin);
+ qcoeff0 = _mm_adds_epi16(qcoeff0, round);
+ qcoeff1 = _mm_adds_epi16(qcoeff1, round);
+ qtmp0 = _mm_mulhi_epi16(qcoeff0, quant);
+ qtmp1 = _mm_mulhi_epi16(qcoeff1, quant);
+ qtmp0 = _mm_add_epi16(qtmp0, qcoeff0);
+ qtmp1 = _mm_add_epi16(qtmp1, qcoeff1);
+ qcoeff0 = _mm_mulhi_epi16(qtmp0, shift);
+ qcoeff1 = _mm_mulhi_epi16(qtmp1, shift);
+
+ // Reinsert signs
+ qcoeff0 = _mm_xor_si128(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(qcoeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ // Mask out zbin threshold coeffs
+ qcoeff0 = _mm_and_si128(qcoeff0, cmp_mask0);
+ qcoeff1 = _mm_and_si128(qcoeff1, cmp_mask1);
+
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), qcoeff0);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
+
+ coeff0 = _mm_mullo_epi16(qcoeff0, dequant);
+ coeff1 = _mm_mullo_epi16(qcoeff1, dequant);
+
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), coeff0);
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
+ }
+
+ {
+ // Scan for eob
+ __m128i zero_coeff0, zero_coeff1;
+ __m128i nzero_coeff0, nzero_coeff1;
+ __m128i iscan0, iscan1;
+ __m128i eob0, eob1;
+ zero_coeff0 = _mm_cmpeq_epi16(coeff0, zero);
+ zero_coeff1 = _mm_cmpeq_epi16(coeff1, zero);
+ nzero_coeff0 = _mm_cmpeq_epi16(zero_coeff0, zero);
+ nzero_coeff1 = _mm_cmpeq_epi16(zero_coeff1, zero);
+ iscan0 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs));
+ iscan1 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs) + 1);
+ // Add one to convert from indices to counts
+ iscan0 = _mm_sub_epi16(iscan0, nzero_coeff0);
+ iscan1 = _mm_sub_epi16(iscan1, nzero_coeff1);
+ eob0 = _mm_and_si128(iscan0, nzero_coeff0);
+ eob1 = _mm_and_si128(iscan1, nzero_coeff1);
+ eob0 = _mm_max_epi16(eob0, eob1);
+ eob = _mm_max_epi16(eob, eob0);
+ }
+ n_coeffs += 8 * 2;
+ }
+
+ // Accumulate EOB
+ {
+ __m128i eob_shuffled;
+ eob_shuffled = _mm_shuffle_epi32(eob, 0xe);
+ eob = _mm_max_epi16(eob, eob_shuffled);
+ eob_shuffled = _mm_shufflelo_epi16(eob, 0xe);
+ eob = _mm_max_epi16(eob, eob_shuffled);
+ eob_shuffled = _mm_shufflelo_epi16(eob, 0x1);
+ eob = _mm_max_epi16(eob, eob_shuffled);
+ *eob_ptr = _mm_extract_epi16(eob, 1);
+ }
+ } else {
+ do {
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), zero);
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, zero);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), zero);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, zero);
+ n_coeffs += 8 * 2;
+ } while (n_coeffs < 0);
+ *eob_ptr = 0;
+ }
+}
diff --git a/vp9/vp9_cx_iface.c b/vp9/vp9_cx_iface.c
index 041ba27..d0ca524 100644
--- a/vp9/vp9_cx_iface.c
+++ b/vp9/vp9_cx_iface.c
@@ -188,11 +188,9 @@
}
if (alt_ref_sum > REF_FRAMES - cfg->ss_number_layers)
ERROR("Not enough ref buffers for svc alt ref frames");
- if ((cfg->ss_number_layers > 3 ||
- cfg->ss_number_layers * cfg->ts_number_layers > 4) &&
+ if (cfg->ss_number_layers * cfg->ts_number_layers > 3 &&
cfg->g_error_resilient == 0)
- ERROR("Multiple frame context are not supported for more than 3 spatial "
- "layers or more than 4 spatial x temporal layers");
+ ERROR("Multiple frame context are not supported for more than 3 layers");
}
#endif
diff --git a/vp9/vp9cx.mk b/vp9/vp9cx.mk
index a2e3cda..ad76722 100644
--- a/vp9/vp9cx.mk
+++ b/vp9/vp9cx.mk
@@ -101,6 +101,7 @@
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_subpel_variance_impl_intrin_avx2.c
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_temporal_filter_apply_sse2.asm
VP9_CX_SRCS-$(HAVE_SSE3) += encoder/x86/vp9_sad_sse3.asm
+VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_quantize_sse2.c
ifeq ($(CONFIG_USE_X86INC),yes)
VP9_CX_SRCS-$(HAVE_MMX) += encoder/x86/vp9_dct_mmx.asm
diff --git a/vpx/src/svc_encodeframe.c b/vpx/src/svc_encodeframe.c
index 773087d..fa3409c 100644
--- a/vpx/src/svc_encodeframe.c
+++ b/vpx/src/svc_encodeframe.c
@@ -350,7 +350,7 @@
}
}
- for (i = 0; i < svc_ctx->spatial_layers; ++i) {
+ for (i = 0; i < VPX_SS_MAX_LAYERS; ++i) {
if (total > 0) {
enc_cfg->ss_target_bitrate[i] = (unsigned int)
(enc_cfg->rc_target_bitrate * alloc_ratio[i] / total);
diff --git a/vpx/vpx_encoder.h b/vpx/vpx_encoder.h
index c6c7d08..044243d 100644
--- a/vpx/vpx_encoder.h
+++ b/vpx/vpx_encoder.h
@@ -161,7 +161,9 @@
VPX_CODEC_STATS_PKT, /**< Two-pass statistics for this frame */
VPX_CODEC_FPMB_STATS_PKT, /**< first pass mb statistics for this frame */
VPX_CODEC_PSNR_PKT, /**< PSNR statistics for this frame */
-#if CONFIG_SPATIAL_SVC
+ // TODO(minghai): This is for testing purporses. The released library can't
+ // depend on vpx_config.h
+#if defined(CONFIG_SPATIAL_SVC) && CONFIG_SPATIAL_SVC
VPX_CODEC_SPATIAL_SVC_LAYER_SIZES, /**< Sizes for each layer in this frame*/
VPX_CODEC_SPATIAL_SVC_LAYER_PSNR, /**< PSNR for each layer in this frame*/
#endif
@@ -201,7 +203,9 @@
double psnr[4]; /**< PSNR, total/y/u/v */
} psnr; /**< data for PSNR packet */
vpx_fixed_buf_t raw; /**< data for arbitrary packets */
-#if CONFIG_SPATIAL_SVC
+ // TODO(minghai): This is for testing purporses. The released library
+ // can't depend on vpx_config.h
+#if defined(CONFIG_SPATIAL_SVC) && CONFIG_SPATIAL_SVC
size_t layer_sizes[VPX_SS_MAX_LAYERS];
struct vpx_psnr_pkt layer_psnr[VPX_SS_MAX_LAYERS];
#endif
diff --git a/vpxdec.c b/vpxdec.c
index 0469071..2afdb71 100644
--- a/vpxdec.c
+++ b/vpxdec.c
@@ -47,52 +47,49 @@
struct WebmInputContext *webm_ctx;
};
-static const arg_def_t looparg = ARG_DEF(NULL, "loops", 1,
- "Number of times to decode the file");
-static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
- "Codec to use");
-static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
- "Output raw YV12 frames");
-static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
- "Output raw I420 frames");
-static const arg_def_t flipuvarg = ARG_DEF(NULL, "flipuv", 0,
- "Flip the chroma planes in the output");
-static const arg_def_t rawvideo = ARG_DEF(NULL, "rawvideo", 0,
- "Output raw YUV frames");
-static const arg_def_t noblitarg = ARG_DEF(NULL, "noblit", 0,
- "Don't process the decoded frames");
-static const arg_def_t progressarg = ARG_DEF(NULL, "progress", 0,
- "Show progress after each frame decodes");
-static const arg_def_t limitarg = ARG_DEF(NULL, "limit", 1,
- "Stop decoding after n frames");
-static const arg_def_t skiparg = ARG_DEF(NULL, "skip", 1,
- "Skip the first n input frames");
-static const arg_def_t postprocarg = ARG_DEF(NULL, "postproc", 0,
- "Postprocess decoded frames");
-static const arg_def_t summaryarg = ARG_DEF(NULL, "summary", 0,
- "Show timing summary");
-static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
- "Output file name pattern (see below)");
-static const arg_def_t threadsarg = ARG_DEF("t", "threads", 1,
- "Max threads to use");
-static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
- "Show version string");
-static const arg_def_t error_concealment = ARG_DEF(NULL, "error-concealment", 0,
- "Enable decoder error-concealment");
-static const arg_def_t scalearg = ARG_DEF("S", "scale", 0,
- "Scale output frames uniformly");
-static const arg_def_t continuearg =
- ARG_DEF("k", "keep-going", 0, "(debug) Continue decoding after error");
-
-static const arg_def_t fb_arg =
- ARG_DEF(NULL, "frame-buffers", 1, "Number of frame buffers to use");
-
-static const arg_def_t md5arg = ARG_DEF(NULL, "md5", 0,
- "Compute the MD5 sum of the decoded frame");
+static const arg_def_t looparg = ARG_DEF(
+ NULL, "loops", 1, "Number of times to decode the file");
+static const arg_def_t codecarg = ARG_DEF(
+ NULL, "codec", 1, "Codec to use");
+static const arg_def_t use_yv12 = ARG_DEF(
+ NULL, "yv12", 0, "Output raw YV12 frames");
+static const arg_def_t use_i420 = ARG_DEF(
+ NULL, "i420", 0, "Output raw I420 frames");
+static const arg_def_t flipuvarg = ARG_DEF(
+ NULL, "flipuv", 0, "Flip the chroma planes in the output");
+static const arg_def_t rawvideo = ARG_DEF(
+ NULL, "rawvideo", 0, "Output raw YUV frames");
+static const arg_def_t noblitarg = ARG_DEF(
+ NULL, "noblit", 0, "Don't process the decoded frames");
+static const arg_def_t progressarg = ARG_DEF(
+ NULL, "progress", 0, "Show progress after each frame decodes");
+static const arg_def_t limitarg = ARG_DEF(
+ NULL, "limit", 1, "Stop decoding after n frames");
+static const arg_def_t skiparg = ARG_DEF(
+ NULL, "skip", 1, "Skip the first n input frames");
+static const arg_def_t postprocarg = ARG_DEF(
+ NULL, "postproc", 0, "Postprocess decoded frames");
+static const arg_def_t summaryarg = ARG_DEF(
+ NULL, "summary", 0, "Show timing summary");
+static const arg_def_t outputfile = ARG_DEF(
+ "o", "output", 1, "Output file name pattern (see below)");
+static const arg_def_t threadsarg = ARG_DEF(
+ "t", "threads", 1, "Max threads to use");
+static const arg_def_t verbosearg = ARG_DEF(
+ "v", "verbose", 0, "Show version string");
+static const arg_def_t error_concealment = ARG_DEF(
+ NULL, "error-concealment", 0, "Enable decoder error-concealment");
+static const arg_def_t scalearg = ARG_DEF(
+ "S", "scale", 0, "Scale output frames uniformly");
+static const arg_def_t continuearg = ARG_DEF(
+ "k", "keep-going", 0, "(debug) Continue decoding after error");
+static const arg_def_t fb_arg = ARG_DEF(
+ NULL, "frame-buffers", 1, "Number of frame buffers to use");
+static const arg_def_t md5arg = ARG_DEF(
+ NULL, "md5", 0, "Compute the MD5 sum of the decoded frame");
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
static const arg_def_t outbitdeptharg = ARG_DEF(
- NULL, "output-bit-depth", 1,
- "Output bit-depth for decoded frames");
+ NULL, "output-bit-depth", 1, "Output bit-depth for decoded frames");
#endif
static const arg_def_t *all_args[] = {
diff --git a/vpxenc.c b/vpxenc.c
index cc81f4a..0a0c071 100644
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -128,50 +128,50 @@
return 0;
}
-static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
- "Debug mode (makes output deterministic)");
-static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
- "Output filename");
-static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
- "Input file is YV12 ");
-static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
- "Input file is I420 (default)");
-static const arg_def_t use_i422 = ARG_DEF(NULL, "i422", 0,
- "Input file is I422");
-static const arg_def_t use_i444 = ARG_DEF(NULL, "i444", 0,
- "Input file is I444");
-static const arg_def_t use_i440 = ARG_DEF(NULL, "i440", 0,
- "Input file is I440");
-static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
- "Codec to use");
-static const arg_def_t passes = ARG_DEF("p", "passes", 1,
- "Number of passes (1/2)");
-static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
- "Pass to execute (1/2)");
-static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
- "First pass statistics file name");
+static const arg_def_t debugmode = ARG_DEF(
+ "D", "debug", 0, "Debug mode (makes output deterministic)");
+static const arg_def_t outputfile = ARG_DEF(
+ "o", "output", 1, "Output filename");
+static const arg_def_t use_yv12 = ARG_DEF(
+ NULL, "yv12", 0, "Input file is YV12 ");
+static const arg_def_t use_i420 = ARG_DEF(
+ NULL, "i420", 0, "Input file is I420 (default)");
+static const arg_def_t use_i422 = ARG_DEF(
+ NULL, "i422", 0, "Input file is I422");
+static const arg_def_t use_i444 = ARG_DEF(
+ NULL, "i444", 0, "Input file is I444");
+static const arg_def_t use_i440 = ARG_DEF(
+ NULL, "i440", 0, "Input file is I440");
+static const arg_def_t codecarg = ARG_DEF(
+ NULL, "codec", 1, "Codec to use");
+static const arg_def_t passes = ARG_DEF(
+ "p", "passes", 1, "Number of passes (1/2)");
+static const arg_def_t pass_arg = ARG_DEF(
+ NULL, "pass", 1, "Pass to execute (1/2)");
+static const arg_def_t fpf_name = ARG_DEF(
+ NULL, "fpf", 1, "First pass statistics file name");
#if CONFIG_FP_MB_STATS
-static const arg_def_t fpmbf_name = ARG_DEF(NULL, "fpmbf", 1,
- "First pass block statistics file name");
+static const arg_def_t fpmbf_name = ARG_DEF(
+ NULL, "fpmbf", 1, "First pass block statistics file name");
#endif
-static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
- "Stop encoding after n input frames");
-static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
- "Skip the first n input frames");
-static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
- "Deadline per frame (usec)");
-static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
- "Use Best Quality Deadline");
-static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
- "Use Good Quality Deadline");
-static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
- "Use Realtime Quality Deadline");
-static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
- "Do not print encode progress");
-static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
- "Show encoder parameters");
-static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
- "Show PSNR in status line");
+static const arg_def_t limit = ARG_DEF(
+ NULL, "limit", 1, "Stop encoding after n input frames");
+static const arg_def_t skip = ARG_DEF(
+ NULL, "skip", 1, "Skip the first n input frames");
+static const arg_def_t deadline = ARG_DEF(
+ "d", "deadline", 1, "Deadline per frame (usec)");
+static const arg_def_t best_dl = ARG_DEF(
+ NULL, "best", 0, "Use Best Quality Deadline");
+static const arg_def_t good_dl = ARG_DEF(
+ NULL, "good", 0, "Use Good Quality Deadline");
+static const arg_def_t rt_dl = ARG_DEF(
+ NULL, "rt", 0, "Use Realtime Quality Deadline");
+static const arg_def_t quietarg = ARG_DEF(
+ "q", "quiet", 0, "Do not print encode progress");
+static const arg_def_t verbosearg = ARG_DEF(
+ "v", "verbose", 0, "Show encoder parameters");
+static const arg_def_t psnrarg = ARG_DEF(
+ NULL, "psnr", 0, "Show PSNR in status line");
static const struct arg_enum_list test_decode_enum[] = {
{"off", TEST_DECODE_OFF},
@@ -179,25 +179,25 @@
{"warn", TEST_DECODE_WARN},
{NULL, 0}
};
-static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
- "Test encode/decode mismatch",
- test_decode_enum);
-static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
- "Stream frame rate (rate/scale)");
-static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
- "Output IVF (default is WebM if WebM IO is enabled)");
-static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
- "Makes encoder output partitions. Requires IVF output!");
-static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
- "Show quantizer histogram (n-buckets)");
-static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
- "Show rate histogram (n-buckets)");
-static const arg_def_t disable_warnings =
- ARG_DEF(NULL, "disable-warnings", 0,
- "Disable warnings about potentially incorrect encode settings.");
-static const arg_def_t disable_warning_prompt =
- ARG_DEF("y", "disable-warning-prompt", 0,
- "Display warnings, but do not prompt user to continue.");
+static const arg_def_t recontest = ARG_DEF_ENUM(
+ NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
+static const arg_def_t framerate = ARG_DEF(
+ NULL, "fps", 1, "Stream frame rate (rate/scale)");
+static const arg_def_t use_ivf = ARG_DEF(
+ NULL, "ivf", 0, "Output IVF (default is WebM if WebM IO is enabled)");
+static const arg_def_t out_part = ARG_DEF(
+ "P", "output-partitions", 0,
+ "Makes encoder output partitions. Requires IVF output!");
+static const arg_def_t q_hist_n = ARG_DEF(
+ NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
+static const arg_def_t rate_hist_n = ARG_DEF(
+ NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
+static const arg_def_t disable_warnings = ARG_DEF(
+ NULL, "disable-warnings", 0,
+ "Disable warnings about potentially incorrect encode settings.");
+static const arg_def_t disable_warning_prompt = ARG_DEF(
+ "y", "disable-warning-prompt", 0,
+ "Display warnings, but do not prompt user to continue.");
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
static const arg_def_t test16bitinternalarg = ARG_DEF(
@@ -213,16 +213,14 @@
NULL
};
-static const arg_def_t usage = ARG_DEF("u", "usage", 1,
- "Usage profile number to use");
-static const arg_def_t threads = ARG_DEF("t", "threads", 1,
- "Max number of threads to use");
-static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
- "Bitstream profile number to use");
-static const arg_def_t width = ARG_DEF("w", "width", 1,
- "Frame width");
-static const arg_def_t height = ARG_DEF("h", "height", 1,
- "Frame height");
+static const arg_def_t usage = ARG_DEF(
+ "u", "usage", 1, "Usage profile number to use");
+static const arg_def_t threads = ARG_DEF(
+ "t", "threads", 1, "Max number of threads to use");
+static const arg_def_t profile = ARG_DEF(
+ NULL, "profile", 1, "Bitstream profile number to use");
+static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
+static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
#if CONFIG_WEBM_IO
static const struct arg_enum_list stereo_mode_enum[] = {
{"mono", STEREO_FORMAT_MONO},
@@ -232,15 +230,15 @@
{"right-left", STEREO_FORMAT_RIGHT_LEFT},
{NULL, 0}
};
-static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
- "Stereo 3D video format", stereo_mode_enum);
+static const arg_def_t stereo_mode = ARG_DEF_ENUM(
+ NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
#endif
-static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
- "Output timestamp precision (fractional seconds)");
-static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
- "Enable error resiliency features");
-static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
- "Max number of frames to lag");
+static const arg_def_t timebase = ARG_DEF(
+ NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
+static const arg_def_t error_resilient = ARG_DEF(
+ NULL, "error-resilient", 1, "Enable error resiliency features");
+static const arg_def_t lag_in_frames = ARG_DEF(
+ NULL, "lag-in-frames", 1, "Max number of frames to lag");
static const arg_def_t *global_args[] = {
&use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
@@ -257,18 +255,18 @@
&lag_in_frames, NULL
};
-static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
- "Temporal resampling threshold (buf %)");
-static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
- "Spatial resampling enabled (bool)");
-static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
- "Width of encoded frame");
-static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
- "Height of encoded frame");
-static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
- "Upscale threshold (buf %)");
-static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
- "Downscale threshold (buf %)");
+static const arg_def_t dropframe_thresh = ARG_DEF(
+ NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
+static const arg_def_t resize_allowed = ARG_DEF(
+ NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
+static const arg_def_t resize_width = ARG_DEF(
+ NULL, "resize-width", 1, "Width of encoded frame");
+static const arg_def_t resize_height = ARG_DEF(
+ NULL, "resize-height", 1, "Height of encoded frame");
+static const arg_def_t resize_up_thresh = ARG_DEF(
+ NULL, "resize-up", 1, "Upscale threshold (buf %)");
+static const arg_def_t resize_down_thresh = ARG_DEF(
+ NULL, "resize-down", 1, "Downscale threshold (buf %)");
static const struct arg_enum_list end_usage_enum[] = {
{"vbr", VPX_VBR},
{"cbr", VPX_CBR},
@@ -276,24 +274,24 @@
{"q", VPX_Q},
{NULL, 0}
};
-static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
- "Rate control mode", end_usage_enum);
-static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
- "Bitrate (kbps)");
-static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
- "Minimum (best) quantizer");
-static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
- "Maximum (worst) quantizer");
-static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
- "Datarate undershoot (min) target (%)");
-static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
- "Datarate overshoot (max) target (%)");
-static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
- "Client buffer size (ms)");
-static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
- "Client initial buffer size (ms)");
-static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
- "Client optimal buffer size (ms)");
+static const arg_def_t end_usage = ARG_DEF_ENUM(
+ NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
+static const arg_def_t target_bitrate = ARG_DEF(
+ NULL, "target-bitrate", 1, "Bitrate (kbps)");
+static const arg_def_t min_quantizer = ARG_DEF(
+ NULL, "min-q", 1, "Minimum (best) quantizer");
+static const arg_def_t max_quantizer = ARG_DEF(
+ NULL, "max-q", 1, "Maximum (worst) quantizer");
+static const arg_def_t undershoot_pct = ARG_DEF(
+ NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
+static const arg_def_t overshoot_pct = ARG_DEF(
+ NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
+static const arg_def_t buf_sz = ARG_DEF(
+ NULL, "buf-sz", 1, "Client buffer size (ms)");
+static const arg_def_t buf_initial_sz = ARG_DEF(
+ NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
+static const arg_def_t buf_optimal_sz = ARG_DEF(
+ NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
static const arg_def_t *rc_args[] = {
&dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
&resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
@@ -302,59 +300,59 @@
};
-static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
- "CBR/VBR bias (0=CBR, 100=VBR)");
-static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
- "GOP min bitrate (% of target)");
-static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
- "GOP max bitrate (% of target)");
+static const arg_def_t bias_pct = ARG_DEF(
+ NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
+static const arg_def_t minsection_pct = ARG_DEF(
+ NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
+static const arg_def_t maxsection_pct = ARG_DEF(
+ NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
static const arg_def_t *rc_twopass_args[] = {
&bias_pct, &minsection_pct, &maxsection_pct, NULL
};
-static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
- "Minimum keyframe interval (frames)");
-static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
- "Maximum keyframe interval (frames)");
-static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
- "Disable keyframe placement");
+static const arg_def_t kf_min_dist = ARG_DEF(
+ NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
+static const arg_def_t kf_max_dist = ARG_DEF(
+ NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
+static const arg_def_t kf_disabled = ARG_DEF(
+ NULL, "disable-kf", 0, "Disable keyframe placement");
static const arg_def_t *kf_args[] = {
&kf_min_dist, &kf_max_dist, &kf_disabled, NULL
};
-static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
- "Noise sensitivity (frames to blur)");
-static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
- "Loop filter sharpness (0..7)");
-static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
- "Motion detection threshold");
-static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
- "CPU Used (-16..16)");
-static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
- "Enable automatic alt reference frames");
-static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
- "AltRef max frames (0..15)");
-static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
- "AltRef filter strength (0..6)");
-static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
- "AltRef type");
+static const arg_def_t noise_sens = ARG_DEF(
+ NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
+static const arg_def_t sharpness = ARG_DEF(
+ NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
+static const arg_def_t static_thresh = ARG_DEF(
+ NULL, "static-thresh", 1, "Motion detection threshold");
+static const arg_def_t cpu_used = ARG_DEF(
+ NULL, "cpu-used", 1, "CPU Used (-16..16)");
+static const arg_def_t auto_altref = ARG_DEF(
+ NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
+static const arg_def_t arnr_maxframes = ARG_DEF(
+ NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
+static const arg_def_t arnr_strength = ARG_DEF(
+ NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
+static const arg_def_t arnr_type = ARG_DEF(
+ NULL, "arnr-type", 1, "AltRef type");
static const struct arg_enum_list tuning_enum[] = {
{"psnr", VP8_TUNE_PSNR},
{"ssim", VP8_TUNE_SSIM},
{NULL, 0}
};
-static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
- "Material to favor", tuning_enum);
-static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
- "Constant/Constrained Quality level");
-static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
- "Max I-frame bitrate (pct)");
+static const arg_def_t tune_ssim = ARG_DEF_ENUM(
+ NULL, "tune", 1, "Material to favor", tuning_enum);
+static const arg_def_t cq_level = ARG_DEF(
+ NULL, "cq-level", 1, "Constant/Constrained Quality level");
+static const arg_def_t max_intra_rate_pct = ARG_DEF(
+ NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
#if CONFIG_VP8_ENCODER
-static const arg_def_t token_parts =
- ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
+static const arg_def_t token_parts = ARG_DEF(
+ NULL, "token-parts", 1, "Number of token partitions to use, log2");
static const arg_def_t *vp8_args[] = {
&cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
&token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
@@ -372,11 +370,12 @@
#endif
#if CONFIG_VP9_ENCODER
-static const arg_def_t tile_cols =
- ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
-static const arg_def_t tile_rows =
- ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
-static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
+static const arg_def_t tile_cols = ARG_DEF(
+ NULL, "tile-columns", 1, "Number of tile columns to use, log2");
+static const arg_def_t tile_rows = ARG_DEF(
+ NULL, "tile-rows", 1, "Number of tile rows to use, log2");
+static const arg_def_t lossless = ARG_DEF(
+ NULL, "lossless", 1, "Lossless mode");
static const arg_def_t frame_parallel_decoding = ARG_DEF(
NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
static const arg_def_t aq_mode = ARG_DEF(
@@ -395,13 +394,12 @@
{NULL, 0}
};
-static const arg_def_t bitdeptharg = ARG_DEF_ENUM("b", "bit-depth", 1,
- "Bit depth for codec "
- "(8 for version <=1, "
- "10 or 12 for version 2)",
- bitdepth_enum);
-static const arg_def_t inbitdeptharg = ARG_DEF(NULL, "input-bit-depth", 1,
- "Bit depth of input");
+static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
+ "b", "bit-depth", 1,
+ "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
+ bitdepth_enum);
+static const arg_def_t inbitdeptharg = ARG_DEF(
+ NULL, "input-bit-depth", 1, "Bit depth of input");
#endif
static const struct arg_enum_list tune_content_enum[] = {
@@ -1161,7 +1159,6 @@
if (j == config->arg_ctrl_cnt)
config->arg_ctrl_cnt++;
}
-
}
}
if (!match)
@@ -1917,14 +1914,15 @@
/* If the input file doesn't specify its w/h (raw files), try to get
* the data from the first stream's configuration.
*/
- if (!input.width || !input.height)
- FOREACH_STREAM( {
- if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
- input.width = stream->config.cfg.g_w;
- input.height = stream->config.cfg.g_h;
- break;
- }
- });
+ if (!input.width || !input.height) {
+ FOREACH_STREAM({
+ if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
+ input.width = stream->config.cfg.g_w;
+ input.height = stream->config.cfg.g_h;
+ break;
+ }
+ });
+ }
/* Update stream configurations from the input file's parameters */
if (!input.width || !input.height)
