ajm: handle single-frame decode jobs (+mp3 imrovements) (#1520)

* ajm: handle single-frame decode jobs (+mp3 imrovements)

* disable breaking the loop in multi-frame if storage is insufficient

* simplified gapless decoding

src/core/libraries/ajm/ajm_at9.cpp

@@ -40,23 +40,11 @@ void AjmAt9Decoder::Initialize(const void* buffer, u32 buffer_size) {
     const auto params = reinterpret_cast<const AjmDecAt9InitializeParameters*>(buffer);
     std::memcpy(m_config_data, params->config_data, ORBIS_AT9_CONFIG_DATA_SIZE);
     AjmAt9Decoder::Reset();
-    m_pcm_buffer.resize(m_codec_info.frameSamples * m_codec_info.channels * GetPointCodeSize(), 0);
+    m_pcm_buffer.resize(m_codec_info.frameSamples * m_codec_info.channels * GetPCMSize(m_format),
+                        0);
 }
 
-u8 AjmAt9Decoder::GetPointCodeSize() {
+void AjmAt9Decoder::GetInfo(void* out_info) const {
-    switch (m_format) {
-    case AjmFormatEncoding::S16:
-        return sizeof(s16);
-    case AjmFormatEncoding::S32:
-        return sizeof(s32);
-    case AjmFormatEncoding::Float:
-        return sizeof(float);
-    default:
-        UNREACHABLE();
-    }
-}
-
-void AjmAt9Decoder::GetInfo(void* out_info) {
     auto* info = reinterpret_cast<AjmSidebandDecAt9CodecInfo*>(out_info);
     info->super_frame_size = m_codec_info.superframeSize;
     info->frames_in_super_frame = m_codec_info.framesInSuperframe;
@@ -65,8 +53,7 @@ void AjmAt9Decoder::GetInfo(void* out_info) {
 }
 
 std::tuple<u32, u32> AjmAt9Decoder::ProcessData(std::span<u8>& in_buf, SparseOutputBuffer& output,
-                                                AjmSidebandGaplessDecode& gapless,
+                                                AjmInstanceGapless& gapless) {
-                                                std::optional<u32> max_samples_per_channel) {
     int ret = 0;
     int bytes_used = 0;
     switch (m_format) {
@@ -91,32 +78,37 @@ std::tuple<u32, u32> AjmAt9Decoder::ProcessData(std::span<u8>& in_buf, SparseOut
 
     m_superframe_bytes_remain -= bytes_used;
 
-    u32 skipped_samples = 0;
+    u32 skip_samples = 0;
-    if (gapless.skipped_samples < gapless.skip_samples) {
+    if (gapless.current.skip_samples > 0) {
-        skipped_samples = std::min(u32(m_codec_info.frameSamples),
+        skip_samples = std::min(u16(m_codec_info.frameSamples), gapless.current.skip_samples);
-                                   u32(gapless.skip_samples - gapless.skipped_samples));
+        gapless.current.skip_samples -= skip_samples;
-        gapless.skipped_samples += skipped_samples;
     }
 
-    const auto max_samples = max_samples_per_channel.has_value()
+    const auto max_pcm = gapless.init.total_samples != 0
-                                 ? max_samples_per_channel.value() * m_codec_info.channels
+                             ? gapless.current.total_samples * m_codec_info.channels
                              : std::numeric_limits<u32>::max();
 
-    size_t samples_written = 0;
+    size_t pcm_written = 0;
     switch (m_format) {
     case AjmFormatEncoding::S16:
-        samples_written = WriteOutputSamples<s16>(output, skipped_samples, max_samples);
+        pcm_written = WriteOutputSamples<s16>(output, skip_samples, max_pcm);
         break;
     case AjmFormatEncoding::S32:
-        samples_written = WriteOutputSamples<s32>(output, skipped_samples, max_samples);
+        pcm_written = WriteOutputSamples<s32>(output, skip_samples, max_pcm);
         break;
     case AjmFormatEncoding::Float:
-        samples_written = WriteOutputSamples<float>(output, skipped_samples, max_samples);
+        pcm_written = WriteOutputSamples<float>(output, skip_samples, max_pcm);
         break;
     default:
         UNREACHABLE();
     }
 
+    const auto samples_written = pcm_written / m_codec_info.channels;
+    gapless.current.skipped_samples += m_codec_info.frameSamples - samples_written;
+    if (gapless.init.total_samples != 0) {
+        gapless.current.total_samples -= samples_written;
+    }
+
     m_num_frames += 1;
     if ((m_num_frames % m_codec_info.framesInSuperframe) == 0) {
         if (m_superframe_bytes_remain) {
@@ -126,18 +118,28 @@ std::tuple<u32, u32> AjmAt9Decoder::ProcessData(std::span<u8>& in_buf, SparseOut
         m_num_frames = 0;
     }
 
-    return {1, samples_written / m_codec_info.channels};
+    return {1, samples_written};
 }
 
-AjmSidebandFormat AjmAt9Decoder::GetFormat() {
+AjmSidebandFormat AjmAt9Decoder::GetFormat() const {
     return AjmSidebandFormat{
         .num_channels = u32(m_codec_info.channels),
         .channel_mask = GetChannelMask(u32(m_codec_info.channels)),
         .sampl_freq = u32(m_codec_info.samplingRate),
         .sample_encoding = m_format,
-        .bitrate = u32(m_codec_info.samplingRate * GetPointCodeSize() * 8),
+        .bitrate = u32((m_codec_info.samplingRate * m_codec_info.superframeSize * 8) /
+                       (m_codec_info.framesInSuperframe * m_codec_info.frameSamples)),
         .reserved = 0,
     };
 }
 
+u32 AjmAt9Decoder::GetNextFrameSize(const AjmInstanceGapless& gapless) const {
+    const auto max_samples =
+        gapless.init.total_samples != 0
+            ? std::min(gapless.current.total_samples, u32(m_codec_info.frameSamples))
+            : m_codec_info.frameSamples;
+    const auto skip_samples = std::min(u32(gapless.current.skip_samples), max_samples);
+    return (max_samples - skip_samples) * m_codec_info.channels * GetPCMSize(m_format);
+}
+
 } // namespace Libraries::Ajm

src/core/libraries/ajm/ajm_at9.h

@@ -33,15 +33,13 @@ struct AjmAt9Decoder final : AjmCodec {
 
     void Reset() override;
     void Initialize(const void* buffer, u32 buffer_size) override;
-    void GetInfo(void* out_info) override;
+    void GetInfo(void* out_info) const override;
-    AjmSidebandFormat GetFormat() override;
+    AjmSidebandFormat GetFormat() const override;
+    u32 GetNextFrameSize(const AjmInstanceGapless& gapless) const override;
     std::tuple<u32, u32> ProcessData(std::span<u8>& input, SparseOutputBuffer& output,
-                                     AjmSidebandGaplessDecode& gapless,
+                                     AjmInstanceGapless& gapless) override;
-                                     std::optional<u32> max_samples) override;
 
 private:
-    u8 GetPointCodeSize();
-
     template <class T>
     size_t WriteOutputSamples(SparseOutputBuffer& output, u32 skipped_samples, u32 max_samples) {
         std::span<T> pcm_data{reinterpret_cast<T*>(m_pcm_buffer.data()),

src/core/libraries/ajm/ajm_batch.cpp

@@ -135,7 +135,10 @@ AjmJob AjmJobFromBatchBuffer(u32 instance_id, AjmBatchBuffer batch_buffer) {
         case Identifier::AjmIdentInputControlBuf: {
             ASSERT_MSG(!input_control_buffer.has_value(),
                        "Only one instance of input control buffer is allowed per job");
-            input_control_buffer = batch_buffer.Consume<AjmChunkBuffer>();
+            const auto& buffer = batch_buffer.Consume<AjmChunkBuffer>();
+            if (buffer.p_address != nullptr && buffer.size != 0) {
+                input_control_buffer = buffer;
+            }
             break;
         }
         case Identifier::AjmIdentControlFlags:
@@ -155,19 +158,27 @@ AjmJob AjmJobFromBatchBuffer(u32 instance_id, AjmBatchBuffer batch_buffer) {
         case Identifier::AjmIdentInlineBuf: {
             ASSERT_MSG(!output_control_buffer.has_value(),
                        "Only one instance of inline buffer is allowed per job");
-            inline_buffer = batch_buffer.Consume<AjmChunkBuffer>();
+            const auto& buffer = batch_buffer.Consume<AjmChunkBuffer>();
+            if (buffer.p_address != nullptr && buffer.size != 0) {
+                inline_buffer = buffer;
+            }
             break;
         }
         case Identifier::AjmIdentOutputRunBuf: {
             auto& buffer = batch_buffer.Consume<AjmChunkBuffer>();
             u8* p_begin = reinterpret_cast<u8*>(buffer.p_address);
+            if (p_begin != nullptr && buffer.size != 0) {
                 job.output.buffers.emplace_back(std::span<u8>(p_begin, p_begin + buffer.size));
+            }
             break;
         }
         case Identifier::AjmIdentOutputControlBuf: {
             ASSERT_MSG(!output_control_buffer.has_value(),
                        "Only one instance of output control buffer is allowed per job");
-            output_control_buffer = batch_buffer.Consume<AjmChunkBuffer>();
+            const auto& buffer = batch_buffer.Consume<AjmChunkBuffer>();
+            if (buffer.p_address != nullptr && buffer.size != 0) {
+                output_control_buffer = buffer;
+            }
             break;
         }
         default:

src/core/libraries/ajm/ajm_instance.cpp

@@ -22,6 +22,19 @@ constexpr int ORBIS_AJM_RESULT_PRIORITY_PASSED = 0x00000200;
 constexpr int ORBIS_AJM_RESULT_CODEC_ERROR = 0x40000000;
 constexpr int ORBIS_AJM_RESULT_FATAL = 0x80000000;
 
+u8 GetPCMSize(AjmFormatEncoding format) {
+    switch (format) {
+    case AjmFormatEncoding::S16:
+        return sizeof(s16);
+    case AjmFormatEncoding::S32:
+        return sizeof(s32);
+    case AjmFormatEncoding::Float:
+        return sizeof(float);
+    default:
+        UNREACHABLE();
+    }
+}
+
 AjmInstance::AjmInstance(AjmCodecType codec_type, AjmInstanceFlags flags) : m_flags(flags) {
     switch (codec_type) {
     case AjmCodecType::At9Dec: {
@@ -30,7 +43,8 @@ AjmInstance::AjmInstance(AjmCodecType codec_type, AjmInstanceFlags flags) : m_fl
         break;
     }
     case AjmCodecType::Mp3Dec: {
-        m_codec = std::make_unique<AjmMp3Decoder>(AjmFormatEncoding(flags.format));
+        m_codec = std::make_unique<AjmMp3Decoder>(AjmFormatEncoding(flags.format),
+                                                  AjmMp3CodecFlags(flags.codec));
         break;
     }
     default:
@@ -45,7 +59,6 @@ void AjmInstance::ExecuteJob(AjmJob& job) {
         m_format = {};
         m_gapless = {};
         m_resample_parameters = {};
-        m_gapless_samples = 0;
         m_total_samples = 0;
         m_codec->Reset();
     }
@@ -64,27 +77,47 @@ void AjmInstance::ExecuteJob(AjmJob& job) {
     }
     if (job.input.gapless_decode.has_value()) {
         auto& params = job.input.gapless_decode.value();
-        m_gapless.total_samples = params.total_samples;
+        if (params.total_samples != 0) {
-        m_gapless.skip_samples = params.skip_samples;
+            const auto max = std::max(params.total_samples, m_gapless.init.total_samples);
+            m_gapless.current.total_samples += max - m_gapless.init.total_samples;
+            m_gapless.init.total_samples = max;
+        }
+        if (params.skip_samples != 0) {
+            const auto max = std::max(params.skip_samples, m_gapless.init.skip_samples);
+            m_gapless.current.skip_samples += max - m_gapless.init.skip_samples;
+            m_gapless.init.skip_samples = max;
+        }
     }
 
     if (!job.input.buffer.empty() && !job.output.buffers.empty()) {
-        u32 frames_decoded = 0;
         std::span<u8> in_buf(job.input.buffer);
         SparseOutputBuffer out_buf(job.output.buffers);
 
+        u32 frames_decoded = 0;
         auto in_size = in_buf.size();
         auto out_size = out_buf.Size();
-        while (!in_buf.empty() && !out_buf.IsEmpty() && !IsGaplessEnd()) {
+        while (!in_buf.empty() && !out_buf.IsEmpty() && !m_gapless.IsEnd()) {
-            const auto samples_remain =
+            if (!HasEnoughSpace(out_buf)) {
-                m_gapless.total_samples != 0
+                if (job.output.p_mframe == nullptr || frames_decoded == 0) {
-                    ? std::optional<u32>{m_gapless.total_samples - m_gapless_samples}
+                    job.output.p_result->result = ORBIS_AJM_RESULT_NOT_ENOUGH_ROOM;
-                    : std::optional<u32>{};
+                    break;
-            const auto [nframes, nsamples] =
+                }
-                m_codec->ProcessData(in_buf, out_buf, m_gapless, samples_remain);
+            }
+
+            const auto [nframes, nsamples] = m_codec->ProcessData(in_buf, out_buf, m_gapless);
             frames_decoded += nframes;
             m_total_samples += nsamples;
-            m_gapless_samples += nsamples;
+
+            if (False(job.flags.run_flags & AjmJobRunFlags::MultipleFrames)) {
+                break;
+            }
+        }
+
+        if (m_gapless.IsEnd()) {
+            in_buf = in_buf.subspan(in_buf.size());
+            m_gapless.current.total_samples = m_gapless.init.total_samples;
+            m_gapless.current.skip_samples = m_gapless.init.skip_samples;
+            m_codec->Reset();
         }
         if (job.output.p_mframe) {
             job.output.p_mframe->num_frames = frames_decoded;
@@ -96,25 +129,19 @@ void AjmInstance::ExecuteJob(AjmJob& job) {
         }
     }
 
-    if (m_flags.gapless_loop && m_gapless.total_samples != 0 &&
-        m_gapless_samples >= m_gapless.total_samples) {
-        m_gapless_samples = 0;
-        m_gapless.skipped_samples = 0;
-        m_codec->Reset();
-    }
     if (job.output.p_format != nullptr) {
         *job.output.p_format = m_codec->GetFormat();
     }
     if (job.output.p_gapless_decode != nullptr) {
-        *job.output.p_gapless_decode = m_gapless;
+        *job.output.p_gapless_decode = m_gapless.current;
     }
     if (job.output.p_codec_info != nullptr) {
         m_codec->GetInfo(job.output.p_codec_info);
     }
 }
 
-bool AjmInstance::IsGaplessEnd() {
+bool AjmInstance::HasEnoughSpace(const SparseOutputBuffer& output) const {
-    return m_gapless.total_samples != 0 && m_gapless_samples >= m_gapless.total_samples;
+    return output.Size() >= m_codec->GetNextFrameSize(m_gapless);
 }
 
 } // namespace Libraries::Ajm

src/core/libraries/ajm/ajm_instance.h

@@ -14,6 +14,8 @@
 
 namespace Libraries::Ajm {
 
+u8 GetPCMSize(AjmFormatEncoding format);
+
 class SparseOutputBuffer {
 public:
     SparseOutputBuffer(std::span<std::span<u8>> chunks)
@@ -33,14 +35,17 @@ public:
                 ++m_current;
             }
         }
+        if (!pcm.empty()) {
+            LOG_ERROR(Lib_Ajm, "Could not write {} samples", pcm.size());
+        }
         return samples_written;
     }
 
-    bool IsEmpty() {
+    bool IsEmpty() const {
         return m_current == m_chunks.end();
     }
 
-    size_t Size() {
+    size_t Size() const {
         size_t result = 0;
         for (auto it = m_current; it != m_chunks.end(); ++it) {
             result += it->size();
@@ -53,17 +58,26 @@ private:
     std::span<std::span<u8>>::iterator m_current;
 };
 
+struct AjmInstanceGapless {
+    AjmSidebandGaplessDecode init{};
+    AjmSidebandGaplessDecode current{};
+
+    bool IsEnd() const {
+        return init.total_samples != 0 && current.total_samples == 0;
+    }
+};
+
 class AjmCodec {
 public:
     virtual ~AjmCodec() = default;
 
     virtual void Initialize(const void* buffer, u32 buffer_size) = 0;
     virtual void Reset() = 0;
-    virtual void GetInfo(void* out_info) = 0;
+    virtual void GetInfo(void* out_info) const = 0;
-    virtual AjmSidebandFormat GetFormat() = 0;
+    virtual AjmSidebandFormat GetFormat() const = 0;
+    virtual u32 GetNextFrameSize(const AjmInstanceGapless& gapless) const = 0;
     virtual std::tuple<u32, u32> ProcessData(std::span<u8>& input, SparseOutputBuffer& output,
-                                             AjmSidebandGaplessDecode& gapless,
+                                             AjmInstanceGapless& gapless) = 0;
-                                             std::optional<u32> max_samples_per_channel) = 0;
 };
 
 class AjmInstance {
@@ -73,16 +87,14 @@ public:
     void ExecuteJob(AjmJob& job);
 
 private:
-    bool IsGaplessEnd();
+    bool HasEnoughSpace(const SparseOutputBuffer& output) const;
+    std::optional<u32> GetNumRemainingSamples() const;
 
     AjmInstanceFlags m_flags{};
     AjmSidebandFormat m_format{};
-    AjmSidebandGaplessDecode m_gapless{};
+    AjmInstanceGapless m_gapless{};
     AjmSidebandResampleParameters m_resample_parameters{};
-
-    u32 m_gapless_samples{};
     u32 m_total_samples{};
-
     std::unique_ptr<AjmCodec> m_codec;
 };
 

src/core/libraries/ajm/ajm_mp3.cpp

@@ -15,18 +15,50 @@ extern "C" {
 namespace Libraries::Ajm {
 
 // Following tables have been reversed from AJM library
-static constexpr std::array<std::array<s32, 3>, 3> SamplerateTable = {{
+static constexpr std::array<std::array<s32, 4>, 4> Mp3SampleRateTable = {
-    {0x5622, 0x5DC0, 0x3E80},
+    std::array<s32, 4>{11025, 12000, 8000, 0},
-    {0xAC44, 0xBB80, 0x7D00},
+    std::array<s32, 4>{0, 0, 0, 0},
-    {0x2B11, 0x2EE0, 0x1F40},
+    std::array<s32, 4>{22050, 24000, 16000, 0},
-}};
+    std::array<s32, 4>{44100, 48000, 32000, 0},
+};
 
-static constexpr std::array<std::array<s32, 15>, 2> BitrateTable = {{
+static constexpr std::array<std::array<s32, 16>, 4> Mp3BitRateTable = {
-    {0, 0x20, 0x28, 0x30, 0x38, 0x40, 0x50, 0x60, 0x70, 0x80, 0xA0, 0xC0, 0xE0, 0x100, 0x140},
+    std::array<s32, 16>{0, 8, 16, 24, 32, 40, 48, 56, 64, 0, 0, 0, 0, 0, 0, 0},
-    {0, 0x8, 0x10, 0x18, 0x20, 0x28, 0x30, 0x38, 0x40, 0x50, 0x60, 0x70, 0x80, 0x90, 0xA0},
+    std::array<s32, 16>{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
-}};
+    std::array<s32, 16>{0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, 0},
+    std::array<s32, 16>{0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 0},
+};
 
-static constexpr std::array<s32, 2> UnkTable = {0x48, 0x90};
+enum class Mp3AudioVersion : u32 {
+    V2_5 = 0,
+    Reserved = 1,
+    V2 = 2,
+    V1 = 3,
+};
+
+enum class Mp3ChannelMode : u32 {
+    Stereo = 0,
+    JointStereo = 1,
+    DualChannel = 2,
+    SingleChannel = 3,
+};
+
+struct Mp3Header {
+    u32 emphasis : 2;
+    u32 original : 1;
+    u32 copyright : 1;
+    u32 mode_ext_idx : 2;
+    Mp3ChannelMode channel_mode : 2;
+    u32 : 1;
+    u32 padding : 1;
+    u32 sampling_rate_idx : 2;
+    u32 bitrate_idx : 4;
+    u32 protection_type : 1;
+    u32 layer_type : 2;
+    Mp3AudioVersion version : 2;
+    u32 sync : 11;
+};
+static_assert(sizeof(Mp3Header) == sizeof(u32));
 
 static AVSampleFormat AjmToAVSampleFormat(AjmFormatEncoding format) {
     switch (format) {
@@ -62,7 +94,7 @@ AVFrame* AjmMp3Decoder::ConvertAudioFrame(AVFrame* frame) {
     swr_init(m_swr_context);
     const auto res = swr_convert_frame(m_swr_context, new_frame, frame);
     if (res < 0) {
-        LOG_ERROR(Lib_AvPlayer, "Could not convert to S16: {}", av_err2str(res));
+        LOG_ERROR(Lib_AvPlayer, "Could not convert frame: {}", av_err2str(res));
         av_frame_free(&new_frame);
         av_frame_free(&frame);
         return nullptr;
@@ -71,48 +103,57 @@ AVFrame* AjmMp3Decoder::ConvertAudioFrame(AVFrame* frame) {
     return new_frame;
 }
 
-AjmMp3Decoder::AjmMp3Decoder(AjmFormatEncoding format)
+AjmMp3Decoder::AjmMp3Decoder(AjmFormatEncoding format, AjmMp3CodecFlags flags)
-    : m_format(format), m_codec(avcodec_find_decoder(AV_CODEC_ID_MP3)),
+    : m_format(format), m_flags(flags), m_codec(avcodec_find_decoder(AV_CODEC_ID_MP3)),
-      m_parser(av_parser_init(m_codec->id)) {
+      m_codec_context(avcodec_alloc_context3(m_codec)), m_parser(av_parser_init(m_codec->id)) {
-    AjmMp3Decoder::Reset();
-}
-
-AjmMp3Decoder::~AjmMp3Decoder() {
-    swr_free(&m_swr_context);
-    avcodec_free_context(&m_codec_context);
-}
-
-void AjmMp3Decoder::Reset() {
-    if (m_codec_context) {
-        avcodec_free_context(&m_codec_context);
-    }
-    m_codec_context = avcodec_alloc_context3(m_codec);
-    ASSERT_MSG(m_codec_context, "Could not allocate audio m_codec context");
     int ret = avcodec_open2(m_codec_context, m_codec, nullptr);
     ASSERT_MSG(ret >= 0, "Could not open m_codec");
 }
 
-void AjmMp3Decoder::GetInfo(void* out_info) {
+AjmMp3Decoder::~AjmMp3Decoder() {
+    swr_free(&m_swr_context);
+    av_parser_close(m_parser);
+    avcodec_free_context(&m_codec_context);
+}
+
+void AjmMp3Decoder::Reset() {
+    avcodec_flush_buffers(m_codec_context);
+    m_header.reset();
+    m_frame_samples = 0;
+}
+
+void AjmMp3Decoder::GetInfo(void* out_info) const {
     auto* info = reinterpret_cast<AjmSidebandDecMp3CodecInfo*>(out_info);
+    if (m_header.has_value()) {
+        auto* header = reinterpret_cast<const Mp3Header*>(&m_header.value());
+        info->header = std::byteswap(m_header.value());
+        info->has_crc = header->protection_type;
+        info->channel_mode = static_cast<ChannelMode>(header->channel_mode);
+        info->mode_extension = header->mode_ext_idx;
+        info->copyright = header->copyright;
+        info->original = header->original;
+        info->emphasis = header->emphasis;
+    }
 }
 
 std::tuple<u32, u32> AjmMp3Decoder::ProcessData(std::span<u8>& in_buf, SparseOutputBuffer& output,
-                                                AjmSidebandGaplessDecode& gapless,
+                                                AjmInstanceGapless& gapless) {
-                                                std::optional<u32> max_samples_per_channel) {
     AVPacket* pkt = av_packet_alloc();
 
+    if ((!m_header.has_value() || m_frame_samples == 0) && in_buf.size() >= 4) {
+        m_header = std::byteswap(*reinterpret_cast<u32*>(in_buf.data()));
+        AjmDecMp3ParseFrame info{};
+        ParseMp3Header(in_buf.data(), in_buf.size(), false, &info);
+        m_frame_samples = info.samples_per_channel;
+    }
+
     int ret = av_parser_parse2(m_parser, m_codec_context, &pkt->data, &pkt->size, in_buf.data(),
                                in_buf.size(), AV_NOPTS_VALUE, AV_NOPTS_VALUE, 0);
     ASSERT_MSG(ret >= 0, "Error while parsing {}", ret);
     in_buf = in_buf.subspan(ret);
 
     u32 frames_decoded = 0;
-    u32 samples_decoded = 0;
+    u32 samples_written = 0;
-
-    auto max_samples =
-        max_samples_per_channel.has_value()
-            ? max_samples_per_channel.value() * m_codec_context->ch_layout.nb_channels
-            : std::numeric_limits<u32>::max();
 
     if (pkt->size) {
         // Send the packet with the compressed data to the decoder
@@ -135,31 +176,38 @@ std::tuple<u32, u32> AjmMp3Decoder::ProcessData(std::span<u8>& in_buf, SparseOut
             frame = ConvertAudioFrame(frame);
 
             frames_decoded += 1;
-            u32 skipped_samples = 0;
+            u32 skip_samples = 0;
-            if (gapless.skipped_samples < gapless.skip_samples) {
+            if (gapless.current.skip_samples > 0) {
-                skipped_samples = std::min(u32(frame->nb_samples),
+                skip_samples = std::min(u16(frame->nb_samples), gapless.current.skip_samples);
-                                           u32(gapless.skip_samples - gapless.skipped_samples));
+                gapless.current.skip_samples -= skip_samples;
-                gapless.skipped_samples += skipped_samples;
             }
 
+            const auto max_pcm =
+                gapless.init.total_samples != 0
+                    ? gapless.current.total_samples * m_codec_context->ch_layout.nb_channels
+                    : std::numeric_limits<u32>::max();
+
+            u32 pcm_written = 0;
             switch (m_format) {
             case AjmFormatEncoding::S16:
-                samples_decoded +=
+                pcm_written = WriteOutputPCM<s16>(frame, output, skip_samples, max_pcm);
-                    WriteOutputSamples<s16>(frame, output, skipped_samples, max_samples);
                 break;
             case AjmFormatEncoding::S32:
-                samples_decoded +=
+                pcm_written = WriteOutputPCM<s32>(frame, output, skip_samples, max_pcm);
-                    WriteOutputSamples<s32>(frame, output, skipped_samples, max_samples);
                 break;
             case AjmFormatEncoding::Float:
-                samples_decoded +=
+                pcm_written = WriteOutputPCM<float>(frame, output, skip_samples, max_pcm);
-                    WriteOutputSamples<float>(frame, output, skipped_samples, max_samples);
                 break;
             default:
                 UNREACHABLE();
             }
 
-            max_samples -= samples_decoded;
+            const auto samples = pcm_written / m_codec_context->ch_layout.nb_channels;
+            samples_written += samples;
+            gapless.current.skipped_samples += frame->nb_samples - samples;
+            if (gapless.init.total_samples != 0) {
+                gapless.current.total_samples -= samples;
+            }
 
             av_frame_free(&frame);
         }
@@ -167,38 +215,221 @@ std::tuple<u32, u32> AjmMp3Decoder::ProcessData(std::span<u8>& in_buf, SparseOut
 
     av_packet_free(&pkt);
 
-    return {frames_decoded, samples_decoded};
+    return {frames_decoded, samples_written};
 }
 
-int AjmMp3Decoder::ParseMp3Header(const u8* buf, u32 stream_size, int parse_ofl,
+u32 AjmMp3Decoder::GetNextFrameSize(const AjmInstanceGapless& gapless) const {
+    const auto max_samples = gapless.init.total_samples != 0
+                                 ? std::min(gapless.current.total_samples, m_frame_samples)
+                                 : m_frame_samples;
+    const auto skip_samples = std::min(u32(gapless.current.skip_samples), max_samples);
+    return (max_samples - skip_samples) * m_codec_context->ch_layout.nb_channels *
+           GetPCMSize(m_format);
+}
+
+class BitReader {
+public:
+    BitReader(const u8* data) : m_data(data) {}
+
+    template <class T>
+    T Read(u32 const nbits) {
+        T accumulator = 0;
+        for (unsigned i = 0; i < nbits; ++i) {
+            accumulator = (accumulator << 1) + GetBit();
+        }
+        return accumulator;
+    }
+
+    void Skip(size_t nbits) {
+        m_bit_offset += nbits;
+    }
+
+    size_t GetCurrentOffset() {
+        return m_bit_offset;
+    }
+
+private:
+    u8 GetBit() {
+        const auto bit = (m_data[m_bit_offset / 8] >> (7 - (m_bit_offset % 8))) & 1;
+        m_bit_offset += 1;
+        return bit;
+    }
+
+    const u8* m_data;
+    size_t m_bit_offset = 0;
+};
+
+int AjmMp3Decoder::ParseMp3Header(const u8* p_begin, u32 stream_size, int parse_ofl,
                                   AjmDecMp3ParseFrame* frame) {
     LOG_INFO(Lib_Ajm, "called stream_size = {} parse_ofl = {}", stream_size, parse_ofl);
-    if (buf == nullptr || stream_size < 4 || frame == nullptr) {
+
-        return ORBIS_AJM_ERROR_INVALID_PARAMETER;
+    if (p_begin == nullptr || stream_size < 4 || frame == nullptr) {
-    }
-    if ((buf[0] & SYNCWORDH) != SYNCWORDH || (buf[1] & SYNCWORDL) != SYNCWORDL) {
         return ORBIS_AJM_ERROR_INVALID_PARAMETER;
     }
 
-    const u32 unk_idx = buf[1] >> 3 & 1;
+    const auto* p_current = p_begin;
-    const s32 version_idx = (buf[1] >> 3 & 3) ^ 2;
+
-    const s32 sr_idx = buf[2] >> 2 & 3;
+    auto bytes = std::byteswap(*reinterpret_cast<const u32*>(p_current));
-    const s32 br_idx = (buf[2] >> 4) & 0xf;
+    p_current += 4;
-    const s32 padding_bit = (buf[2] >> 1) & 0x1;
+    auto header = reinterpret_cast<const Mp3Header*>(&bytes);
+    if (header->sync != 0x7FF) {
+        return ORBIS_AJM_ERROR_INVALID_PARAMETER;
+    }
+
+    frame->sample_rate = Mp3SampleRateTable[u32(header->version)][header->sampling_rate_idx];
+    frame->bitrate = Mp3BitRateTable[u32(header->version)][header->bitrate_idx] * 1000;
+    frame->num_channels = header->channel_mode == Mp3ChannelMode::SingleChannel ? 1 : 2;
+    if (header->version == Mp3AudioVersion::V1) {
+        frame->frame_size = (144 * frame->bitrate) / frame->sample_rate + header->padding;
+        frame->samples_per_channel = 1152;
+    } else {
+        frame->frame_size = (72 * frame->bitrate) / frame->sample_rate + header->padding;
+        frame->samples_per_channel = 576;
+    }
 
-    frame->sample_rate = SamplerateTable[version_idx][sr_idx];
-    frame->bitrate = BitrateTable[version_idx != 1][br_idx] * 1000;
-    frame->num_channels = (buf[3] < 0xc0) + 1;
-    frame->frame_size = (UnkTable[unk_idx] * frame->bitrate) / frame->sample_rate + padding_bit;
-    frame->samples_per_channel = UnkTable[unk_idx] * 8;
     frame->encoder_delay = 0;
+    frame->num_frames = 0;
+    frame->total_samples = 0;
+    frame->ofl_type = AjmDecMp3OflType::None;
+
+    if (!parse_ofl) {
+        return ORBIS_OK;
+    }
+
+    BitReader reader(p_current);
+    if (header->protection_type == 0) {
+        reader.Skip(16); // crc = reader.Read<u16>(16);
+    }
+
+    if (header->version == Mp3AudioVersion::V1) {
+        // main_data_begin = reader.Read<u16>(9);
+        // if (header->channel_mode == Mp3ChannelMode::SingleChannel) {
+        //     private_bits = reader.Read<u8>(5);
+        // } else {
+        //     private_bits = reader.Read<u8>(3);
+        // }
+        // for (u32 ch = 0; ch < frame->num_channels; ++ch) {
+        //     for (u8 band = 0; band < 4; ++band) {
+        //         scfsi[ch][band] = reader.Read<bool>(1);
+        //     }
+        // }
+        if (header->channel_mode == Mp3ChannelMode::SingleChannel) {
+            reader.Skip(18);
+        } else {
+            reader.Skip(20);
+        }
+    } else {
+        // main_data_begin = reader.Read<u16>(8);
+        // if (header->channel_mode == Mp3ChannelMode::SingleChannel) {
+        //     private_bits = reader.Read<u8>(1);
+        // } else {
+        //     private_bits = reader.Read<u8>(2);
+        // }
+        if (header->channel_mode == Mp3ChannelMode::SingleChannel) {
+            reader.Skip(9);
+        } else {
+            reader.Skip(10);
+        }
+    }
+
+    u32 part2_3_length = 0;
+    // Number of granules (18x32 sub-band samples)
+    const u8 ngr = header->version == Mp3AudioVersion::V1 ? 2 : 1;
+    for (u8 gr = 0; gr < ngr; ++gr) {
+        for (u32 ch = 0; ch < frame->num_channels; ++ch) {
+            // part2_3_length[gr][ch] = reader.Read<u16>(12);
+            part2_3_length += reader.Read<u16>(12);
+            // big_values[gr][ch] = reader.Read<u16>(9);
+            // global_main[gr][ch] = reader.Read<u8>(8);
+            // if (header->version == Mp3AudioVersion::V1) {
+            //     scalefac_compress[gr][ch] = reader.Read<u16>(4);
+            // } else {
+            //     scalefac_compress[gr][ch] = reader.Read<u16>(9);
+            // }
+            // window_switching_flag = reader.Read<bool>(1);
+            // if (window_switching_flag) {
+            //     block_type[gr][ch] = reader.Read<u8>(2);
+            //     mixed_block_flag[gr][ch] = reader.Read<bool>(1);
+            //     for (u8 region = 0; region < 2; ++region) {
+            //         table_select[gr][ch][region] = reader.Read<u8>(5);
+            //     }
+            //     for (u8 window = 0; window < 3; ++window) {
+            //         subblock_gain[gr][ch][window] = reader.Read<u8>(3);
+            //     }
+            // } else {
+            //     for (u8 region = 0; region < 3; ++region) {
+            //         table_select[gr][ch][region] = reader.Read<u8>(5);
+            //     }
+            //     region0_count[gr][ch] = reader.Read<u8>(4);
+            //     region1_count[gr][ch] = reader.Read<u8>(3);
+            // }
+            // if (header->version == Mp3AudioVersion::V1) {
+            //     preflag[gr][ch] = reader.Read<bool>(1);
+            // }
+            // scalefac_scale[gr][ch] = reader.Read<bool>(1);
+            // count1table_select[gr][ch] = reader.Read<bool>(1);
+            if (header->version == Mp3AudioVersion::V1) {
+                reader.Skip(47);
+            } else {
+                reader.Skip(51);
+            }
+        }
+    }
+    reader.Skip(part2_3_length);
+
+    p_current += ((reader.GetCurrentOffset() + 7) / 8);
+
+    const auto* p_end = p_begin + frame->frame_size;
+    if (memcmp(p_current, "Xing", 4) == 0 || memcmp(p_current, "Info", 4) == 0) {
+        // TODO: Parse Xing/Lame header
+        LOG_ERROR(Lib_Ajm, "Xing/Lame header is not implemented.");
+    } else if (memcmp(p_current, "VBRI", 4) == 0) {
+        // TODO: Parse VBRI header
+        LOG_ERROR(Lib_Ajm, "VBRI header is not implemented.");
+    } else {
+        // Parse FGH header
+        constexpr auto fgh_indicator = 0xB4;
+        while ((p_current + 9) < p_end && *p_current != fgh_indicator) {
+            ++p_current;
+        }
+        auto p_fgh = p_current;
+        if ((p_current + 9) < p_end && *p_current == fgh_indicator) {
+            u8 crc = 0xFF;
+            auto crc_func = [](u8 c, u8 v, u8 s) {
+                if (((c >> 7) & 1) != ((v >> s) & 1)) {
+                    return c * 2;
+                }
+                return (c * 2) ^ 0x45;
+            };
+            for (u8 i = 0; i < 9; ++i, ++p_current) {
+                for (u8 j = 0; j < 8; ++j) {
+                    crc = crc_func(crc, *p_current, 7 - j);
+                }
+            }
+            if (p_fgh[9] == crc) {
+                frame->encoder_delay = std::byteswap(*reinterpret_cast<const u16*>(p_fgh + 1));
+                frame->total_samples = std::byteswap(*reinterpret_cast<const u32*>(p_fgh + 3));
+                frame->ofl_type = AjmDecMp3OflType::Fgh;
+            } else {
+                LOG_ERROR(Lib_Ajm, "FGH header CRC is incorrect.");
+            }
+        } else {
+            LOG_ERROR(Lib_Ajm, "Could not find vendor header.");
+        }
+    }
 
     return ORBIS_OK;
 }
 
-AjmSidebandFormat AjmMp3Decoder::GetFormat() {
+AjmSidebandFormat AjmMp3Decoder::GetFormat() const {
-    LOG_ERROR(Lib_Ajm, "Unimplemented");
+    return AjmSidebandFormat{
-    return AjmSidebandFormat{};
+        .num_channels = u32(m_codec_context->ch_layout.nb_channels),
+        .channel_mask = GetChannelMask(u32(m_codec_context->ch_layout.nb_channels)),
+        .sampl_freq = u32(m_codec_context->sample_rate),
+        .sample_encoding = m_format,
+        .bitrate = u32(m_codec_context->bit_rate),
+        .reserved = 0,
+    };
 };
 
 } // namespace Libraries::Ajm

src/core/libraries/ajm/ajm_mp3.h

@@ -13,7 +13,19 @@ struct SwrContext;
 
 namespace Libraries::Ajm {
 
-enum class AjmDecMp3OflType : u32 { None = 0, Lame = 1, Vbri = 2, Fgh = 3, VbriAndFgh = 4 };
+enum class AjmDecMp3OflType : u32 {
+    None = 0,
+    Lame = 1,
+    Vbri = 2,
+    Fgh = 3,
+    VbriAndFgh = 4,
+};
+
+enum AjmMp3CodecFlags : u32 {
+    IgnoreOfl = 1 << 0,
+    VlcRewind = 1 << 8,
+};
+DECLARE_ENUM_FLAG_OPERATORS(AjmMp3CodecFlags)
 
 // 11-bit syncword if MPEG 2.5 extensions are enabled
 static constexpr u8 SYNCWORDH = 0xff;
@@ -51,39 +63,40 @@ struct AjmSidebandDecMp3CodecInfo {
 
 class AjmMp3Decoder : public AjmCodec {
 public:
-    explicit AjmMp3Decoder(AjmFormatEncoding format);
+    explicit AjmMp3Decoder(AjmFormatEncoding format, AjmMp3CodecFlags flags);
     ~AjmMp3Decoder() override;
 
     void Reset() override;
     void Initialize(const void* buffer, u32 buffer_size) override {}
-    void GetInfo(void* out_info) override;
+    void GetInfo(void* out_info) const override;
-    AjmSidebandFormat GetFormat() override;
+    AjmSidebandFormat GetFormat() const override;
+    u32 GetNextFrameSize(const AjmInstanceGapless& gapless) const override;
     std::tuple<u32, u32> ProcessData(std::span<u8>& input, SparseOutputBuffer& output,
-                                     AjmSidebandGaplessDecode& gapless,
+                                     AjmInstanceGapless& gapless) override;
-                                     std::optional<u32> max_samples_per_channel) override;
 
     static int ParseMp3Header(const u8* buf, u32 stream_size, int parse_ofl,
                               AjmDecMp3ParseFrame* frame);
 
 private:
     template <class T>
-    size_t WriteOutputSamples(AVFrame* frame, SparseOutputBuffer& output, u32 skipped_samples,
+    size_t WriteOutputPCM(AVFrame* frame, SparseOutputBuffer& output, u32 skipped_samples,
-                              u32 max_samples) {
+                          u32 max_pcm) {
-        const auto size = frame->ch_layout.nb_channels * frame->nb_samples * sizeof(T);
+        std::span<T> pcm_data(reinterpret_cast<T*>(frame->data[0]),
-        std::span<T> pcm_data(reinterpret_cast<T*>(frame->data[0]), size >> 1);
+                              frame->nb_samples * frame->ch_layout.nb_channels);
         pcm_data = pcm_data.subspan(skipped_samples * frame->ch_layout.nb_channels);
-        const auto pcm_size = std::min(u32(pcm_data.size()), max_samples);
+        return output.Write(pcm_data.subspan(0, std::min(u32(pcm_data.size()), max_pcm)));
-        const auto samples_written = output.Write(pcm_data.subspan(0, pcm_size));
-        return samples_written / frame->ch_layout.nb_channels;
     }
 
     AVFrame* ConvertAudioFrame(AVFrame* frame);
 
     const AjmFormatEncoding m_format;
+    const AjmMp3CodecFlags m_flags;
     const AVCodec* m_codec = nullptr;
     AVCodecContext* m_codec_context = nullptr;
     AVCodecParserContext* m_parser = nullptr;
     SwrContext* m_swr_context = nullptr;
+    std::optional<u32> m_header;
+    u32 m_frame_samples = 0;
 };
 
 } // namespace Libraries::Ajm