Migrate VideoRtpDepacketizerVp9 implementation from BitBuffer to BitstreamReader
The new version is subjectivly cleaner
and objectively generates smaller binary size
Bug: None
Change-Id: I662596c41f30690f7e81a51804dd7feb263510c3
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/231130
Reviewed-by: Åsa Persson <asapersson@webrtc.org>
Commit-Queue: Danil Chapovalov <danilchap@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#34935}
diff --git a/modules/rtp_rtcp/BUILD.gn b/modules/rtp_rtcp/BUILD.gn
index f9bc1ff..d3f2fe0 100644
--- a/modules/rtp_rtcp/BUILD.gn
+++ b/modules/rtp_rtcp/BUILD.gn
@@ -287,6 +287,7 @@
"../../logging:rtc_event_audio",
"../../logging:rtc_event_rtp_rtcp",
"../../modules/audio_coding:audio_coding_module_typedefs",
+ "../../rtc_base:bitstream_reader",
"../../rtc_base:checks",
"../../rtc_base:divide_round",
"../../rtc_base:gtest_prod",
diff --git a/modules/rtp_rtcp/source/video_rtp_depacketizer_vp9.cc b/modules/rtp_rtcp/source/video_rtp_depacketizer_vp9.cc
index be05009..cd5ebe7 100644
--- a/modules/rtp_rtcp/source/video_rtp_depacketizer_vp9.cc
+++ b/modules/rtp_rtcp/source/video_rtp_depacketizer_vp9.cc
@@ -15,20 +15,13 @@
#include "api/video/video_codec_constants.h"
#include "modules/rtp_rtcp/source/rtp_packet_to_send.h"
#include "modules/video_coding/codecs/interface/common_constants.h"
-#include "rtc_base/bit_buffer.h"
+#include "rtc_base/bitstream_reader.h"
#include "rtc_base/checks.h"
#include "rtc_base/logging.h"
-#define RETURN_FALSE_ON_ERROR(x) \
- if (!(x)) { \
- return false; \
- }
-
namespace webrtc {
namespace {
-constexpr int kFailedToParse = 0;
-
// Picture ID:
//
// +-+-+-+-+-+-+-+-+
@@ -37,66 +30,37 @@
// M: | EXTENDED PID |
// +-+-+-+-+-+-+-+-+
//
-bool ParsePictureId(rtc::BitBuffer* parser, RTPVideoHeaderVP9* vp9) {
- uint32_t picture_id;
- uint32_t m_bit;
- RETURN_FALSE_ON_ERROR(parser->ReadBits(1, m_bit));
- if (m_bit) {
- RETURN_FALSE_ON_ERROR(parser->ReadBits(15, picture_id));
+void ParsePictureId(BitstreamReader& parser, RTPVideoHeaderVP9* vp9) {
+ if (parser.ReadBit()) { // m_bit
+ vp9->picture_id = parser.ReadBits(15);
vp9->max_picture_id = kMaxTwoBytePictureId;
} else {
- RETURN_FALSE_ON_ERROR(parser->ReadBits(7, picture_id));
+ vp9->picture_id = parser.ReadBits(7);
vp9->max_picture_id = kMaxOneBytePictureId;
}
- vp9->picture_id = picture_id;
- return true;
}
-// Layer indices (flexible mode):
+// Layer indices :
//
// +-+-+-+-+-+-+-+-+
// L: | T |U| S |D|
// +-+-+-+-+-+-+-+-+
-//
-bool ParseLayerInfoCommon(rtc::BitBuffer* parser, RTPVideoHeaderVP9* vp9) {
- uint32_t t, u_bit, s, d_bit;
- RETURN_FALSE_ON_ERROR(parser->ReadBits(3, t));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(1, u_bit));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(3, s));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(1, d_bit));
- vp9->temporal_idx = t;
- vp9->temporal_up_switch = u_bit ? true : false;
- if (s >= kMaxSpatialLayers)
- return false;
- vp9->spatial_idx = s;
- vp9->inter_layer_predicted = d_bit ? true : false;
- return true;
-}
-
-// Layer indices (non-flexible mode):
-//
-// +-+-+-+-+-+-+-+-+
-// L: | T |U| S |D|
-// +-+-+-+-+-+-+-+-+
-// | TL0PICIDX |
+// | TL0PICIDX | (non-flexible mode only)
// +-+-+-+-+-+-+-+-+
//
-bool ParseLayerInfoNonFlexibleMode(rtc::BitBuffer* parser,
- RTPVideoHeaderVP9* vp9) {
- uint8_t tl0picidx;
- RETURN_FALSE_ON_ERROR(parser->ReadUInt8(tl0picidx));
- vp9->tl0_pic_idx = tl0picidx;
- return true;
-}
+void ParseLayerInfo(BitstreamReader& parser, RTPVideoHeaderVP9* vp9) {
+ vp9->temporal_idx = parser.ReadBits(3);
+ vp9->temporal_up_switch = parser.Read<bool>();
+ vp9->spatial_idx = parser.ReadBits(3);
+ vp9->inter_layer_predicted = parser.Read<bool>();
+ if (vp9->spatial_idx >= kMaxSpatialLayers) {
+ parser.Invalidate();
+ return;
+ }
-bool ParseLayerInfo(rtc::BitBuffer* parser, RTPVideoHeaderVP9* vp9) {
- if (!ParseLayerInfoCommon(parser, vp9))
- return false;
-
- if (vp9->flexible_mode)
- return true;
-
- return ParseLayerInfoNonFlexibleMode(parser, vp9);
+ if (!vp9->flexible_mode) {
+ vp9->tl0_pic_idx = parser.Read<uint8_t>();
+ }
}
// Reference indices:
@@ -106,19 +70,22 @@
// +-+-+-+-+-+-+-+-+ N=1: An additional P_DIFF follows
// current P_DIFF.
//
-bool ParseRefIndices(rtc::BitBuffer* parser, RTPVideoHeaderVP9* vp9) {
- if (vp9->picture_id == kNoPictureId)
- return false;
+void ParseRefIndices(BitstreamReader& parser, RTPVideoHeaderVP9* vp9) {
+ if (vp9->picture_id == kNoPictureId) {
+ parser.Invalidate();
+ return;
+ }
vp9->num_ref_pics = 0;
- uint32_t n_bit;
+ bool n_bit;
do {
- if (vp9->num_ref_pics == kMaxVp9RefPics)
- return false;
+ if (vp9->num_ref_pics == kMaxVp9RefPics) {
+ parser.Invalidate();
+ return;
+ }
- uint32_t p_diff;
- RETURN_FALSE_ON_ERROR(parser->ReadBits(7, p_diff));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(1, n_bit));
+ uint8_t p_diff = parser.ReadBits(7);
+ n_bit = parser.Read<bool>();
vp9->pid_diff[vp9->num_ref_pics] = p_diff;
uint32_t scaled_pid = vp9->picture_id;
@@ -128,8 +95,6 @@
}
vp9->ref_picture_id[vp9->num_ref_pics++] = scaled_pid - p_diff;
} while (n_bit);
-
- return true;
}
// Scalability structure (SS).
@@ -152,54 +117,40 @@
// | P_DIFF | (OPTIONAL) . R times .
// +-+-+-+-+-+-+-+-+ -| -|
//
-bool ParseSsData(rtc::BitBuffer* parser, RTPVideoHeaderVP9* vp9) {
- uint32_t n_s, y_bit, g_bit;
- RETURN_FALSE_ON_ERROR(parser->ReadBits(3, n_s));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(1, y_bit));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(1, g_bit));
- RETURN_FALSE_ON_ERROR(parser->ConsumeBits(3));
- vp9->num_spatial_layers = n_s + 1;
- vp9->spatial_layer_resolution_present = y_bit ? true : false;
+void ParseSsData(BitstreamReader& parser, RTPVideoHeaderVP9* vp9) {
+ vp9->num_spatial_layers = parser.ReadBits(3) + 1;
+ vp9->spatial_layer_resolution_present = parser.Read<bool>();
+ bool g_bit = parser.Read<bool>();
+ parser.ConsumeBits(3);
vp9->gof.num_frames_in_gof = 0;
- if (y_bit) {
+ if (vp9->spatial_layer_resolution_present) {
for (size_t i = 0; i < vp9->num_spatial_layers; ++i) {
- RETURN_FALSE_ON_ERROR(parser->ReadUInt16(vp9->width[i]));
- RETURN_FALSE_ON_ERROR(parser->ReadUInt16(vp9->height[i]));
+ vp9->width[i] = parser.Read<uint16_t>();
+ vp9->height[i] = parser.Read<uint16_t>();
}
}
if (g_bit) {
- uint8_t n_g;
- RETURN_FALSE_ON_ERROR(parser->ReadUInt8(n_g));
- vp9->gof.num_frames_in_gof = n_g;
+ vp9->gof.num_frames_in_gof = parser.Read<uint8_t>();
}
for (size_t i = 0; i < vp9->gof.num_frames_in_gof; ++i) {
- uint32_t t, u_bit, r;
- RETURN_FALSE_ON_ERROR(parser->ReadBits(3, t));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(1, u_bit));
- RETURN_FALSE_ON_ERROR(parser->ReadBits(2, r));
- RETURN_FALSE_ON_ERROR(parser->ConsumeBits(2));
- vp9->gof.temporal_idx[i] = t;
- vp9->gof.temporal_up_switch[i] = u_bit ? true : false;
- vp9->gof.num_ref_pics[i] = r;
+ vp9->gof.temporal_idx[i] = parser.ReadBits(3);
+ vp9->gof.temporal_up_switch[i] = parser.Read<bool>();
+ vp9->gof.num_ref_pics[i] = parser.ReadBits(2);
+ parser.ConsumeBits(2);
for (uint8_t p = 0; p < vp9->gof.num_ref_pics[i]; ++p) {
- uint8_t p_diff;
- RETURN_FALSE_ON_ERROR(parser->ReadUInt8(p_diff));
- vp9->gof.pid_diff[i][p] = p_diff;
+ vp9->gof.pid_diff[i][p] = parser.Read<uint8_t>();
}
}
- return true;
}
} // namespace
absl::optional<VideoRtpDepacketizer::ParsedRtpPayload>
VideoRtpDepacketizerVp9::Parse(rtc::CopyOnWriteBuffer rtp_payload) {
- rtc::ArrayView<const uint8_t> payload(rtp_payload.cdata(),
- rtp_payload.size());
absl::optional<ParsedRtpPayload> result(absl::in_place);
- int offset = ParseRtpPayload(payload, &result->video_header);
- if (offset == kFailedToParse)
+ int offset = ParseRtpPayload(rtp_payload, &result->video_header);
+ if (offset == 0)
return absl::nullopt;
RTC_DCHECK_LT(offset, rtp_payload.size());
result->video_payload =
@@ -212,12 +163,8 @@
RTPVideoHeader* video_header) {
RTC_DCHECK(video_header);
// Parse mandatory first byte of payload descriptor.
- rtc::BitBuffer parser(rtp_payload.data(), rtp_payload.size());
- uint8_t first_byte;
- if (!parser.ReadUInt8(first_byte)) {
- RTC_LOG(LS_ERROR) << "Payload length is zero.";
- return kFailedToParse;
- }
+ BitstreamReader parser(rtp_payload);
+ uint8_t first_byte = parser.Read<uint8_t>();
bool i_bit = first_byte & 0b1000'0000; // PictureId present .
bool p_bit = first_byte & 0b0100'0000; // Inter-picture predicted.
bool l_bit = first_byte & 0b0010'0000; // Layer indices present.
@@ -247,23 +194,17 @@
vp9_header.non_ref_for_inter_layer_pred = z_bit;
// Parse fields that are present.
- if (i_bit && !ParsePictureId(&parser, &vp9_header)) {
- RTC_LOG(LS_ERROR) << "Failed parsing VP9 picture id.";
- return kFailedToParse;
+ if (i_bit) {
+ ParsePictureId(parser, &vp9_header);
}
- if (l_bit && !ParseLayerInfo(&parser, &vp9_header)) {
- RTC_LOG(LS_ERROR) << "Failed parsing VP9 layer info.";
- return kFailedToParse;
+ if (l_bit) {
+ ParseLayerInfo(parser, &vp9_header);
}
- if (p_bit && f_bit && !ParseRefIndices(&parser, &vp9_header)) {
- RTC_LOG(LS_ERROR) << "Failed parsing VP9 ref indices.";
- return kFailedToParse;
+ if (p_bit && f_bit) {
+ ParseRefIndices(parser, &vp9_header);
}
if (v_bit) {
- if (!ParseSsData(&parser, &vp9_header)) {
- RTC_LOG(LS_ERROR) << "Failed parsing VP9 SS data.";
- return kFailedToParse;
- }
+ ParseSsData(parser, &vp9_header);
if (vp9_header.spatial_layer_resolution_present) {
// TODO(asapersson): Add support for spatial layers.
video_header->width = vp9_header.width[0];
@@ -273,15 +214,13 @@
video_header->is_first_packet_in_frame =
b_bit && (!l_bit || !vp9_header.inter_layer_predicted);
- size_t byte_offset;
- size_t bit_offset;
- parser.GetCurrentOffset(&byte_offset, &bit_offset);
- RTC_DCHECK_EQ(bit_offset, 0);
- if (byte_offset == rtp_payload.size()) {
- // Empty vp9 payload data.
- return kFailedToParse;
+ int num_remaining_bits = parser.RemainingBitCount();
+ if (num_remaining_bits <= 0) {
+ // Failed to parse or empty vp9 payload data.
+ return 0;
}
-
- return byte_offset;
+ // vp9 descriptor is byte aligned.
+ RTC_DCHECK_EQ(num_remaining_bits % 8, 0);
+ return rtp_payload.size() - num_remaining_bits / 8;
}
} // namespace webrtc
