audio: let the AJM Opus decoder write part of a packet (#1914)

ajm: buffer and resample Opus output
This commit is contained in:
Devran Cosmo Uenal
2026-10-28 21:31:16 +01:00
committed by GitHub
parent c67352d5d6
commit c528324d27
5 changed files with 621 additions and 18 deletions
+55 -10
View File
@@ -6,6 +6,7 @@
#include "libs/ajm/decoder.h"
#include "libs/ajm/mp3_decoder.h"
#include "libs/ajm/opus_decoder.h"
#include "libs/ajm/resampler.h"
#include "libs/audio.h"
#include "libs/errno.h"
#include "libs/libs.h"
@@ -63,12 +64,6 @@ struct AjmSidebandMFrame {
uint32_t reserved;
};
struct AjmSidebandResampleInfo {
float ratio;
int32_t num_samples;
uint32_t reserved[8];
};
struct AjmGetStatisticsResult {
AjmSidebandResult result;
float engine_usage_batch;
@@ -104,6 +99,8 @@ constexpr uint64_t AJM_FLAG_RUN_GET_CODEC_INFO = 1ull << 11u;
constexpr uint64_t AJM_FLAG_RUN_MULTIPLE_FRAMES = 1ull << 12u;
constexpr uint64_t AJM_FLAG_CONTROL_RESET = 1ull << 13u;
constexpr uint64_t AJM_FLAG_CONTROL_INITIALIZE = 1ull << 14u;
constexpr uint64_t AJM_FLAG_CONTROL_RESAMPLE = 1ull << 15u;
constexpr uint64_t AJM_FLAG_CONTROL_RESAMPLE_EX = 1ull << 16u;
constexpr uint64_t AJM_FLAG_SIDEBAND_RESAMPLE_INFO = 1ull << 44u;
constexpr uint64_t AJM_FLAG_SIDEBAND_GAPLESS = 1ull << 45u;
constexpr uint64_t AJM_FLAG_SIDEBAND_FORMAT = 1ull << 46u;
@@ -177,6 +174,7 @@ struct AjmInstanceState {
uint64_t flags = 0;
AjmGaplessState gapless;
std::unique_ptr<AjmDecoder> decoder;
AjmResampler resampler;
};
static std::mutex g_ajm_instances_mutex;
@@ -358,7 +356,7 @@ static void AjmWriteSideband(uint64_t flags, void* sideband_output, size_t sideb
offset + AJM_SIDEBAND_RESAMPLE_INFO_SIZE <= sideband_output_size) {
auto* info = reinterpret_cast<AjmSidebandResampleInfo*>(
static_cast<uint8_t*>(sideband_output) + offset);
info->ratio = 1.0f;
*info = decode_result.resample_info;
offset += AJM_SIDEBAND_RESAMPLE_INFO_SIZE;
}
@@ -425,6 +423,12 @@ static AjmDecodeResult AjmDecodeInstance(uint32_t instance, const void* input, s
return result;
}
// The resampler requires one complete decoded packet per decoder call.
if (state->codec == static_cast<uint32_t>(AjmCodec::DecOpus) &&
(state->flags & AJM_INSTANCE_FLAG_RESAMPLE) != 0) {
return state->resampler.Decode(*state->decoder, state->flags, input, input_size, output,
output_size, multiple_frames, &state->gapless);
}
return state->decoder->Decode(input, input_size, output, output_size, multiple_frames,
&state->gapless);
}
@@ -514,6 +518,32 @@ static AjmDecodeResult AjmControlInstance(uint32_t instance, uint64_t flags,
sideband_input = static_cast<const uint8_t*>(sideband_input) + AJM_SIDEBAND_GAPLESS_SIZE;
sideband_input_size -= AJM_SIDEBAND_GAPLESS_SIZE;
}
for (const auto flag : {AJM_FLAG_CONTROL_RESAMPLE, AJM_FLAG_CONTROL_RESAMPLE_EX}) {
if ((flags & flag) == 0) {
continue;
}
const bool extended = flag == AJM_FLAG_CONTROL_RESAMPLE_EX;
const size_t size = extended ? 12 : 8;
if (sideband_input == nullptr || sideband_input_size < size ||
(state->flags & AJM_INSTANCE_FLAG_RESAMPLE) == 0) {
result.result = AJM_RESULT_INVALID_PARAMETER;
return result;
}
float ratio, change = 0;
uint32_t resample_flags;
const auto* data = static_cast<const uint8_t*>(sideband_input);
std::memcpy(&ratio, data, sizeof(ratio));
if (extended) {
std::memcpy(&change, data + sizeof(ratio), sizeof(change));
}
std::memcpy(&resample_flags, data + size - sizeof(resample_flags), sizeof(resample_flags));
result.result = state->resampler.SetParameters(ratio, change, resample_flags);
if (result.result != OK) {
return result;
}
sideband_input = data + size;
sideband_input_size -= size;
}
if ((flags & AJM_FLAG_CONTROL_RESET) != 0) {
state->gapless.Reset();
decoder->Reset();
@@ -524,6 +554,7 @@ static AjmDecodeResult AjmControlInstance(uint32_t instance, uint64_t flags,
result = decoder->Initialize(sideband_input, sideband_input_size);
}
result.resample_info = state->resampler.GetInfo();
return result;
}
@@ -572,6 +603,7 @@ static AjmSidebandFormat AjmGetInstanceFormat(uint32_t instance,
if (out_result != nullptr) {
*out_result = state->decoder->MakeResult();
out_result->format = state->decoder->GetFormat();
out_result->resample_info = state->resampler.GetInfo();
}
return state->decoder->GetFormat();
}
@@ -921,7 +953,13 @@ int KYTY_SYSV_ABI AjmBatchJobSetResampleParameters(AjmBatchInfo* info, uint32_t
float ratio, uint32_t flags, void* result) {
PRINT_NAME();
EXIT_NOT_IMPLEMENTED(info == nullptr);
AjmWriteResult(result, AJM_SIDEBAND_RESULT_SIZE);
const struct {
float ratio;
uint32_t flags;
} params {ratio, flags};
const auto decoded = AjmControlInstance(instance, AJM_FLAG_CONTROL_RESAMPLE, &params,
sizeof(params), nullptr, nullptr);
AjmWriteResult(result, AJM_SIDEBAND_RESULT_SIZE, decoded);
LOGF("\t instance = 0x%08" PRIx32 ", ratio = %f, flags = 0x%08" PRIx32 "\n", instance,
static_cast<double>(ratio), flags);
return AjmAppendJob(info, AJM_JOB_SET_RESAMPLE_PARAMETERS_SIZE, "set-resample");
@@ -933,7 +971,14 @@ int KYTY_SYSV_ABI AjmBatchJobSetResampleParametersEx(AjmBatchInfo* info, uint32_
void* result) {
PRINT_NAME();
EXIT_NOT_IMPLEMENTED(info == nullptr);
AjmWriteResult(result, AJM_SIDEBAND_RESULT_SIZE);
const struct {
float ratio;
float change;
uint32_t flags;
} params {ratio_start, ratio_change_per_sample, flags};
const auto decoded = AjmControlInstance(instance, AJM_FLAG_CONTROL_RESAMPLE_EX, &params,
sizeof(params), nullptr, nullptr);
AjmWriteResult(result, AJM_SIDEBAND_RESULT_SIZE, decoded);
LOGF("\t instance = 0x%08" PRIx32 ", ratio_start = %f, ratio_change = %f, flags = 0x%08" PRIx32
"\n",
instance, static_cast<double>(ratio_start), static_cast<double>(ratio_change_per_sample),
@@ -952,7 +997,7 @@ int KYTY_SYSV_ABI AjmBatchJobGetResampleInfo(AjmBatchInfo* info, uint32_t instan
if (result != nullptr) {
auto* resample = reinterpret_cast<AjmSidebandResampleInfo*>(static_cast<uint8_t*>(result) +
AJM_SIDEBAND_RESULT_SIZE);
resample->ratio = 1.0f;
*resample = decode_result.resample_info;
}
LOGF("\t instance = 0x%08" PRIx32 "\n", instance);
return AjmAppendJob(info, AJM_JOB_RUN_SIZE, "get-resample-info");
+15 -8
View File
@@ -56,15 +56,22 @@ enum class AjmSampleEncoding : uint32_t {
Float = 2,
};
struct AjmSidebandResampleInfo {
float ratio = 1.0f;
int32_t num_samples = 0;
uint32_t reserved[8] {};
};
struct AjmDecodeResult {
int32_t result = OK;
int32_t internal_result = OK;
size_t input_consumed = 0;
size_t output_written = 0;
uint64_t total_decoded_samples = 0;
uint32_t frames = 0;
uint32_t frames_per_packet = 0;
AjmSidebandFormat format {};
int32_t result = OK;
int32_t internal_result = OK;
size_t input_consumed = 0;
size_t output_written = 0;
uint64_t total_decoded_samples = 0;
uint32_t frames = 0;
uint32_t frames_per_packet = 0;
AjmSidebandFormat format {};
AjmSidebandResampleInfo resample_info {};
};
constexpr int32_t AJM_RESULT_NOT_INITIALIZED = 0x00000001;
+182
View File
@@ -0,0 +1,182 @@
#pragma once
#include "libs/ajm/decoder.h"
#include <algorithm>
#include <cmath>
#include <cstring>
#include <vector>
namespace Libs::Audio::Ajm {
constexpr uint64_t AJM_INSTANCE_FLAG_GAPLESS_LOOP = 1ull << 10u;
constexpr uint64_t AJM_INSTANCE_FLAG_RESAMPLE = 1ull << 12u;
constexpr uint64_t AJM_INSTANCE_FLAG_SEAMLESS_LOOP = 1ull << 13u;
constexpr uint32_t AJM_RESAMPLE_FLAG_RESET = 1u;
constexpr uint32_t AJM_RESAMPLE_FLAG_END = 2u;
constexpr uint32_t AJM_RESAMPLE_FLAG_IGNORE_RATIO_PARAMETER = 8u;
class AjmResampler {
public:
int32_t SetParameters(float ratio, float change, uint32_t flags) {
if ((flags & AJM_RESAMPLE_FLAG_IGNORE_RATIO_PARAMETER) == 0 &&
(!std::isfinite(ratio) || ratio < 0 || ratio > 4 || !std::isfinite(change))) {
return AJM_RESULT_INVALID_PARAMETER;
}
if ((flags & AJM_RESAMPLE_FLAG_RESET) != 0) {
m_pcm.clear();
m_position = 0;
m_end = false;
}
if ((flags & AJM_RESAMPLE_FLAG_IGNORE_RATIO_PARAMETER) == 0) {
m_ratio = ratio;
m_change = std::abs(change) < 0.000001f ? 0 : change;
}
m_end |= (flags & AJM_RESAMPLE_FLAG_END) != 0;
return OK;
}
[[nodiscard]] AjmSidebandResampleInfo GetInfo() const {
AjmSidebandResampleInfo info;
info.ratio = static_cast<float>(m_ratio);
info.num_samples = m_end && m_position >= Samples()
? -1
: static_cast<int32_t>(std::ceil(Samples() - m_position));
return info;
}
AjmDecodeResult Decode(AjmDecoder& decoder, uint64_t flags, const void* input,
size_t input_size, void* output, size_t output_size,
bool multiple_frames, AjmGaplessState* gapless) {
auto result = decoder.MakeResult();
if (result.format.channel_num == 0) {
return decoder.Decode(input, input_size, output, output_size, multiple_frames, gapless);
}
bool stop_decoding = false;
for (;;) {
const auto format = decoder.GetFormat();
const auto encoding = static_cast<AjmSampleEncoding>(format.sample_encoding);
const auto sample_bytes = format.channel_num * AjmBytesPerSample(encoding);
if (!m_pcm.empty() && sample_bytes != m_sample_bytes) {
result.result = AJM_RESULT_INVALID_PARAMETER;
break;
}
if (output == nullptr || output_size - result.output_written < sample_bytes) {
if (m_position < Samples() || result.input_consumed < input_size) {
result.result |= AJM_RESULT_NOT_ENOUGH_ROOM;
}
break;
}
const auto index = static_cast<size_t>(m_position);
const auto fraction = m_position - static_cast<double>(index);
const bool need_next = fraction != 0 && index + 1 >= Samples();
if (index >= Samples() || need_next) {
Compact();
if (!stop_decoding && input != nullptr && result.input_consumed < input_size &&
(gapless == nullptr || !gapless->IsEnd())) {
// A decode job can produce up to six 20 ms Opus frames in one packet.
const auto channels = std::max<uint32_t>(format.channel_num, flags & 0x7fu);
m_frame.resize(5760 * channels * AjmBytesPerSample(encoding));
auto decoded =
decoder.Decode(static_cast<const uint8_t*>(input) + result.input_consumed,
input_size - result.input_consumed, m_frame.data(),
m_frame.size(), false, gapless);
result.input_consumed += decoded.input_consumed;
result.frames += decoded.frames;
result.frames_per_packet = decoded.frames_per_packet;
result.total_decoded_samples = decoded.total_decoded_samples;
result.format = decoded.format;
result.result |= decoded.result;
result.internal_result = decoded.internal_result;
m_sample_bytes = decoded.format.channel_num * AjmBytesPerSample(encoding);
m_pcm.insert(m_pcm.end(), m_frame.begin(),
m_frame.begin() + decoded.output_written);
stop_decoding =
!multiple_frames || decoded.result != OK || decoded.input_consumed == 0;
if (gapless != nullptr && gapless->IsEnd() &&
(flags & AJM_INSTANCE_FLAG_GAPLESS_LOOP) != 0) {
gapless->Reset();
stop_decoding |= (flags & AJM_INSTANCE_FLAG_SEAMLESS_LOOP) == 0;
}
continue;
}
if (!m_end || m_position >= Samples()) {
if (!m_end && !stop_decoding && (gapless == nullptr || !gapless->IsEnd())) {
result.result |= AJM_RESULT_PARTIAL_INPUT;
}
break;
}
}
const auto current = static_cast<size_t>(m_position);
const auto next = std::min(current + 1, Samples() - 1);
auto* dst = static_cast<uint8_t*>(output) + result.output_written;
const auto* a = m_pcm.data() + current * sample_bytes;
const auto* b = m_pcm.data() + next * sample_bytes;
const auto weight = m_position - static_cast<double>(current);
if (weight == 0) {
std::memcpy(dst, a, sample_bytes);
} else {
switch (encoding) {
case AjmSampleEncoding::S16:
Interpolate<int16_t>(dst, a, b, format.channel_num, weight);
break;
case AjmSampleEncoding::S32:
Interpolate<int32_t>(dst, a, b, format.channel_num, weight);
break;
case AjmSampleEncoding::Float:
Interpolate<float>(dst, a, b, format.channel_num, weight);
break;
}
}
result.output_written += sample_bytes;
m_position += m_ratio;
m_ratio = std::clamp(m_ratio + m_change, 0.0, 4.0);
if (m_ratio == 0 || m_ratio == 4) {
m_change = 0;
}
}
Compact();
result.resample_info = GetInfo();
return result;
}
private:
template <typename T>
static void Interpolate(uint8_t* dst, const uint8_t* a, const uint8_t* b, uint32_t channels,
double weight) {
for (uint32_t channel = 0; channel < channels; ++channel) {
T first, second;
std::memcpy(&first, a + channel * sizeof(T), sizeof(T));
std::memcpy(&second, b + channel * sizeof(T), sizeof(T));
const auto value =
static_cast<T>(first + (static_cast<double>(second) - first) * weight);
std::memcpy(dst + channel * sizeof(T), &value, sizeof(T));
}
}
[[nodiscard]] size_t Samples() const {
return m_sample_bytes == 0 ? 0 : m_pcm.size() / m_sample_bytes;
}
void Compact() {
const auto consumed = std::min(static_cast<size_t>(m_position), Samples());
const auto bytes = consumed * m_sample_bytes;
if (bytes != 0) {
std::memmove(m_pcm.data(), m_pcm.data() + bytes, m_pcm.size() - bytes);
m_pcm.resize(m_pcm.size() - bytes);
m_position -= static_cast<double>(consumed);
}
}
std::vector<uint8_t> m_pcm;
std::vector<uint8_t> m_frame;
size_t m_sample_bytes = 0;
double m_position = 0;
double m_ratio = 1;
double m_change = 0;
bool m_end = false;
};
} // namespace Libs::Audio::Ajm