/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ /* This Source Code Form is subject to the terms of the Mozilla Public * License, v. 2.0. If a copy of the MPL was not distributed with this * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ #if !defined(AudioConverter_h) #define AudioConverter_h #include "MediaInfo.h" // Forward declaration typedef struct SpeexResamplerState_ SpeexResamplerState; namespace mozilla { template struct AudioDataBufferTypeChooser; template <> struct AudioDataBufferTypeChooser { typedef uint8_t Type; }; template <> struct AudioDataBufferTypeChooser { typedef int16_t Type; }; template <> struct AudioDataBufferTypeChooser { typedef int32_t Type; }; template <> struct AudioDataBufferTypeChooser { typedef int32_t Type; }; template <> struct AudioDataBufferTypeChooser { typedef int32_t Type; }; template <> struct AudioDataBufferTypeChooser { typedef float Type; }; // 'Value' is the type used externally to deal with stored value. // AudioDataBuffer can perform conversion between different SampleFormat content. template ::Type> class AudioDataBuffer { public: AudioDataBuffer() {} AudioDataBuffer(Value* aBuffer, size_t aLength) : mBuffer(aBuffer, aLength) {} explicit AudioDataBuffer(const AudioDataBuffer& aOther) : mBuffer(aOther.mBuffer) {} AudioDataBuffer(AudioDataBuffer&& aOther) : mBuffer(Move(aOther.mBuffer)) {} template explicit AudioDataBuffer(const AudioDataBuffer& other) { // TODO: Convert from different type, may use asm routines. MOZ_CRASH("Conversion not implemented yet"); } // A u8, s16 and float aligned buffer can only be treated as // FORMAT_U8, FORMAT_S16 and FORMAT_FLT respectively. // So allow them as copy and move constructors. explicit AudioDataBuffer(const AlignedByteBuffer& aBuffer) : mBuffer(aBuffer) { static_assert(Format == AudioConfig::FORMAT_U8, "Conversion not implemented yet"); } explicit AudioDataBuffer(const AlignedShortBuffer& aBuffer) : mBuffer(aBuffer) { static_assert(Format == AudioConfig::FORMAT_S16, "Conversion not implemented yet"); } explicit AudioDataBuffer(const AlignedFloatBuffer& aBuffer) : mBuffer(aBuffer) { static_assert(Format == AudioConfig::FORMAT_FLT, "Conversion not implemented yet"); } explicit AudioDataBuffer(AlignedByteBuffer&& aBuffer) : mBuffer(Move(aBuffer)) { static_assert(Format == AudioConfig::FORMAT_U8, "Conversion not implemented yet"); } explicit AudioDataBuffer(AlignedShortBuffer&& aBuffer) : mBuffer(Move(aBuffer)) { static_assert(Format == AudioConfig::FORMAT_S16, "Conversion not implemented yet"); } explicit AudioDataBuffer(AlignedFloatBuffer&& aBuffer) : mBuffer(Move(aBuffer)) { static_assert(Format == AudioConfig::FORMAT_FLT, "Conversion not implemented yet"); } AudioDataBuffer& operator=(AudioDataBuffer&& aOther) { mBuffer = Move(aOther.mBuffer); return *this; } AudioDataBuffer& operator=(const AudioDataBuffer& aOther) { mBuffer = aOther.mBuffer; return *this; } Value* Data() const { return mBuffer.Data(); } size_t Length() const { return mBuffer.Length(); } size_t Size() const { return mBuffer.Size(); } AlignedBuffer Forget() { // Correct type -> Just give values as-is. return Move(mBuffer); } private: AlignedBuffer mBuffer; }; typedef AudioDataBuffer AudioSampleBuffer; class AudioConverter { public: AudioConverter(const AudioConfig& aIn, const AudioConfig& aOut); ~AudioConverter(); // Convert the AudioDataBuffer. // Conversion will be done in place if possible. Otherwise a new buffer will // be returned. // Providing an empty buffer and resampling is expected, the resampler // will be drained. template AudioDataBuffer Process(AudioDataBuffer&& aBuffer) { MOZ_DIAGNOSTIC_ASSERT(mIn.Format() == mOut.Format() && mIn.Format() == Format); AudioDataBuffer buffer = Move(aBuffer); if (CanWorkInPlace()) { size_t frames = SamplesInToFrames(buffer.Length()); frames = ProcessInternal(buffer.Data(), buffer.Data(), frames); if (frames && mIn.Rate() != mOut.Rate()) { frames = ResampleAudio(buffer.Data(), buffer.Data(), frames); } AlignedBuffer temp = buffer.Forget(); temp.SetLength(FramesOutToSamples(frames)); return AudioDataBuffer(Move(temp));; } return Process(buffer); } template AudioDataBuffer Process(const AudioDataBuffer& aBuffer) { MOZ_DIAGNOSTIC_ASSERT(mIn.Format() == mOut.Format() && mIn.Format() == Format); // Perform the downmixing / reordering in temporary buffer. size_t frames = SamplesInToFrames(aBuffer.Length()); AlignedBuffer temp1; if (!temp1.SetLength(FramesOutToSamples(frames))) { return AudioDataBuffer(Move(temp1)); } frames = ProcessInternal(temp1.Data(), aBuffer.Data(), frames); if (mIn.Rate() == mOut.Rate()) { MOZ_ALWAYS_TRUE(temp1.SetLength(FramesOutToSamples(frames))); return AudioDataBuffer(Move(temp1)); } // At this point, temp1 contains the buffer reordered and downmixed. // If we are downsampling we can re-use it. AlignedBuffer* outputBuffer = &temp1; AlignedBuffer temp2; if (!frames || mOut.Rate() > mIn.Rate()) { // We are upsampling or about to drain, we can't work in place. // Allocate another temporary buffer where the upsampling will occur. if (!temp2.SetLength(FramesOutToSamples(ResampleRecipientFrames(frames)))) { return AudioDataBuffer(Move(temp2)); } outputBuffer = &temp2; } if (!frames) { frames = DrainResampler(outputBuffer->Data()); } else { frames = ResampleAudio(outputBuffer->Data(), temp1.Data(), frames); } MOZ_ALWAYS_TRUE(outputBuffer->SetLength(FramesOutToSamples(frames))); return AudioDataBuffer(Move(*outputBuffer)); } // Attempt to convert the AudioDataBuffer in place. // Will return 0 if the conversion wasn't possible. template size_t Process(Value* aBuffer, size_t aFrames) { MOZ_DIAGNOSTIC_ASSERT(mIn.Format() == mOut.Format()); if (!CanWorkInPlace()) { return 0; } size_t frames = ProcessInternal(aBuffer, aBuffer, aFrames); if (frames && mIn.Rate() != mOut.Rate()) { frames = ResampleAudio(aBuffer, aBuffer, aFrames); } return frames; } bool CanWorkInPlace() const; bool CanReorderAudio() const { return mIn.Layout().MappingTable(mOut.Layout()); } const AudioConfig& InputConfig() const { return mIn; } const AudioConfig& OutputConfig() const { return mOut; } private: const AudioConfig mIn; const AudioConfig mOut; uint8_t mChannelOrderMap[MAX_AUDIO_CHANNELS]; /** * ProcessInternal * Parameters: * aOut : destination buffer where converted samples will be copied * aIn : source buffer * aSamples: number of frames in source buffer * * Return Value: number of frames converted or 0 if error */ size_t ProcessInternal(void* aOut, const void* aIn, size_t aFrames); void ReOrderInterleavedChannels(void* aOut, const void* aIn, size_t aFrames) const; size_t DownmixAudio(void* aOut, const void* aIn, size_t aFrames) const; size_t UpmixAudio(void* aOut, const void* aIn, size_t aFrames) const; size_t FramesOutToSamples(size_t aFrames) const; size_t SamplesInToFrames(size_t aSamples) const; size_t FramesOutToBytes(size_t aFrames) const; // Resampler context. SpeexResamplerState* mResampler; size_t ResampleAudio(void* aOut, const void* aIn, size_t aFrames); size_t ResampleRecipientFrames(size_t aFrames) const; void RecreateResampler(); size_t DrainResampler(void* aOut); }; } // namespace mozilla #endif /* AudioConverter_h */