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 authored and GitHub committed 2026-10-28 21:31:16 +01:00
1 parent c67352d5d6
commit c528324d27
5 files changed
+621 -18

No files matched your search

+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