Ryujinx/Ryujinx.Audio/Renderers/SoundIo/SoundIoAudioTrack.cs

647 lines
24 KiB
C#
Raw Normal View History

Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
using SoundIOSharp;
using System;
using System.Collections.Concurrent;
using System.Linq;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
namespace Ryujinx.Audio.SoundIo
{
internal class SoundIoAudioTrack : IDisposable
{
/// <summary>
/// The audio track ring buffer
/// </summary>
private SoundIoRingBuffer m_Buffer;
/// <summary>
/// A list of buffers currently pending writeback to the audio backend
/// </summary>
private ConcurrentQueue<SoundIoBuffer> m_ReservedBuffers;
/// <summary>
/// Occurs when a buffer has been released by the audio backend
/// </summary>
private event ReleaseCallback BufferReleased;
/// <summary>
/// The track ID of this <see cref="SoundIoAudioTrack"/>
/// </summary>
public int TrackID { get; private set; }
/// <summary>
/// The current playback state
/// </summary>
public PlaybackState State { get; private set; }
/// <summary>
/// The <see cref="SoundIO"/> audio context this track belongs to
/// </summary>
public SoundIO AudioContext { get; private set; }
/// <summary>
/// The <see cref="SoundIODevice"/> this track belongs to
/// </summary>
public SoundIODevice AudioDevice { get; private set; }
/// <summary>
/// The audio output stream of this track
/// </summary>
public SoundIOOutStream AudioStream { get; private set; }
/// <summary>
/// Released buffers the track is no longer holding
/// </summary>
public ConcurrentQueue<long> ReleasedBuffers { get; private set; }
/// <summary>
/// Buffer count of the track
/// </summary>
public uint BufferCount => (uint)m_ReservedBuffers.Count;
/// <summary>
/// Played sample count of the track
/// </summary>
public ulong PlayedSampleCount { get; private set; }
private int _hardwareChannels;
private int _virtualChannels;
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
/// <summary>
/// Constructs a new instance of a <see cref="SoundIoAudioTrack"/>
/// </summary>
/// <param name="trackId">The track ID</param>
/// <param name="audioContext">The SoundIO audio context</param>
/// <param name="audioDevice">The SoundIO audio device</param>
public SoundIoAudioTrack(int trackId, SoundIO audioContext, SoundIODevice audioDevice)
{
TrackID = trackId;
AudioContext = audioContext;
AudioDevice = audioDevice;
State = PlaybackState.Stopped;
ReleasedBuffers = new ConcurrentQueue<long>();
m_Buffer = new SoundIoRingBuffer();
m_ReservedBuffers = new ConcurrentQueue<SoundIoBuffer>();
}
/// <summary>
/// Opens the audio track with the specified parameters
/// </summary>
/// <param name="sampleRate">The requested sample rate of the track</param>
/// <param name="hardwareChannels">The requested hardware channels</param>
/// <param name="virtualChannels">The requested virtual channels</param>
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
/// <param name="callback">A <see cref="ReleaseCallback" /> that represents the delegate to invoke when a buffer has been released by the audio track</param>
/// <param name="format">The requested sample format of the track</param>
public void Open(
int sampleRate,
int hardwareChannels,
int virtualChannels,
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
ReleaseCallback callback,
SoundIOFormat format = SoundIOFormat.S16LE)
{
// Close any existing audio streams
if (AudioStream != null)
{
Close();
}
if (!AudioDevice.SupportsSampleRate(sampleRate))
{
throw new InvalidOperationException($"This sound device does not support a sample rate of {sampleRate}Hz");
}
if (!AudioDevice.SupportsFormat(format))
{
throw new InvalidOperationException($"This sound device does not support SoundIOFormat.{Enum.GetName(typeof(SoundIOFormat), format)}");
}
if (!AudioDevice.SupportsChannelCount(hardwareChannels))
{
throw new InvalidOperationException($"This sound device does not support channel count {hardwareChannels}");
}
_hardwareChannels = hardwareChannels;
_virtualChannels = virtualChannels;
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
AudioStream = AudioDevice.CreateOutStream();
AudioStream.Name = $"SwitchAudioTrack_{TrackID}";
AudioStream.Layout = SoundIOChannelLayout.GetDefault(hardwareChannels);
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
AudioStream.Format = format;
AudioStream.SampleRate = sampleRate;
AudioStream.WriteCallback = WriteCallback;
BufferReleased += callback;
AudioStream.Open();
}
/// <summary>
/// This callback occurs when the sound device is ready to buffer more frames
/// </summary>
/// <param name="minFrameCount">The minimum amount of frames expected by the audio backend</param>
/// <param name="maxFrameCount">The maximum amount of frames that can be written to the audio backend</param>
private unsafe void WriteCallback(int minFrameCount, int maxFrameCount)
{
int bytesPerFrame = AudioStream.BytesPerFrame;
uint bytesPerSample = (uint)AudioStream.BytesPerSample;
int bufferedFrames = m_Buffer.Length / bytesPerFrame;
long bufferedSamples = m_Buffer.Length / bytesPerSample;
int frameCount = Math.Min(bufferedFrames, maxFrameCount);
if (frameCount == 0)
{
return;
}
SoundIOChannelAreas areas = AudioStream.BeginWrite(ref frameCount);
int channelCount = areas.ChannelCount;
byte[] samples = new byte[frameCount * bytesPerFrame];
m_Buffer.Read(samples, 0, samples.Length);
// This is a huge ugly block of code, but we save
// a significant amount of time over the generic
// loop that handles other channel counts.
// Mono
if (channelCount == 1)
{
SoundIOChannelArea area = areas.GetArea(0);
fixed (byte* srcptr = samples)
{
if (bytesPerSample == 1)
{
for (int frame = 0; frame < frameCount; frame++)
{
((byte*)area.Pointer)[0] = srcptr[frame * bytesPerFrame];
area.Pointer += area.Step;
}
}
else if (bytesPerSample == 2)
{
for (int frame = 0; frame < frameCount; frame++)
{
((short*)area.Pointer)[0] = ((short*)srcptr)[frame * bytesPerFrame >> 1];
area.Pointer += area.Step;
}
}
else if (bytesPerSample == 4)
{
for (int frame = 0; frame < frameCount; frame++)
{
((int*)area.Pointer)[0] = ((int*)srcptr)[frame * bytesPerFrame >> 2];
area.Pointer += area.Step;
}
}
else
{
for (int frame = 0; frame < frameCount; frame++)
{
Unsafe.CopyBlockUnaligned((byte*)area.Pointer, srcptr + (frame * bytesPerFrame), bytesPerSample);
area.Pointer += area.Step;
}
}
}
}
// Stereo
else if (channelCount == 2)
{
SoundIOChannelArea area1 = areas.GetArea(0);
SoundIOChannelArea area2 = areas.GetArea(1);
fixed (byte* srcptr = samples)
{
if (bytesPerSample == 1)
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
((byte*)area1.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 0];
// Channel 2
((byte*)area2.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 1];
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
}
}
else if (bytesPerSample == 2)
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
((short*)area1.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 0];
// Channel 2
((short*)area2.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 1];
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
}
}
else if (bytesPerSample == 4)
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
((int*)area1.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 0];
// Channel 2
((int*)area2.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 1];
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
}
}
else
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
Unsafe.CopyBlockUnaligned((byte*)area1.Pointer, srcptr + (frame * bytesPerFrame) + (0 * bytesPerSample), bytesPerSample);
// Channel 2
Unsafe.CopyBlockUnaligned((byte*)area2.Pointer, srcptr + (frame * bytesPerFrame) + (1 * bytesPerSample), bytesPerSample);
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
}
}
}
}
// Surround
else if (channelCount == 6)
{
SoundIOChannelArea area1 = areas.GetArea(0);
SoundIOChannelArea area2 = areas.GetArea(1);
SoundIOChannelArea area3 = areas.GetArea(2);
SoundIOChannelArea area4 = areas.GetArea(3);
SoundIOChannelArea area5 = areas.GetArea(4);
SoundIOChannelArea area6 = areas.GetArea(5);
fixed (byte* srcptr = samples)
{
if (bytesPerSample == 1)
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
((byte*)area1.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 0];
// Channel 2
((byte*)area2.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 1];
// Channel 3
((byte*)area3.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 2];
// Channel 4
((byte*)area4.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 3];
// Channel 5
((byte*)area5.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 4];
// Channel 6
((byte*)area6.Pointer)[0] = srcptr[(frame * bytesPerFrame) + 5];
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
area3.Pointer += area3.Step;
area4.Pointer += area4.Step;
area5.Pointer += area5.Step;
area6.Pointer += area6.Step;
}
}
else if (bytesPerSample == 2)
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
((short*)area1.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 0];
// Channel 2
((short*)area2.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 1];
// Channel 3
((short*)area3.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 2];
// Channel 4
((short*)area4.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 3];
// Channel 5
((short*)area5.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 4];
// Channel 6
((short*)area6.Pointer)[0] = ((short*)srcptr)[(frame * bytesPerFrame >> 1) + 5];
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
area3.Pointer += area3.Step;
area4.Pointer += area4.Step;
area5.Pointer += area5.Step;
area6.Pointer += area6.Step;
}
}
else if (bytesPerSample == 4)
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
((int*)area1.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 0];
// Channel 2
((int*)area2.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 1];
// Channel 3
((int*)area3.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 2];
// Channel 4
((int*)area4.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 3];
// Channel 5
((int*)area5.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 4];
// Channel 6
((int*)area6.Pointer)[0] = ((int*)srcptr)[(frame * bytesPerFrame >> 2) + 5];
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
area3.Pointer += area3.Step;
area4.Pointer += area4.Step;
area5.Pointer += area5.Step;
area6.Pointer += area6.Step;
}
}
else
{
for (int frame = 0; frame < frameCount; frame++)
{
// Channel 1
Unsafe.CopyBlockUnaligned((byte*)area1.Pointer, srcptr + (frame * bytesPerFrame) + (0 * bytesPerSample), bytesPerSample);
// Channel 2
Unsafe.CopyBlockUnaligned((byte*)area2.Pointer, srcptr + (frame * bytesPerFrame) + (1 * bytesPerSample), bytesPerSample);
// Channel 3
Unsafe.CopyBlockUnaligned((byte*)area3.Pointer, srcptr + (frame * bytesPerFrame) + (2 * bytesPerSample), bytesPerSample);
// Channel 4
Unsafe.CopyBlockUnaligned((byte*)area4.Pointer, srcptr + (frame * bytesPerFrame) + (3 * bytesPerSample), bytesPerSample);
// Channel 5
Unsafe.CopyBlockUnaligned((byte*)area5.Pointer, srcptr + (frame * bytesPerFrame) + (4 * bytesPerSample), bytesPerSample);
// Channel 6
Unsafe.CopyBlockUnaligned((byte*)area6.Pointer, srcptr + (frame * bytesPerFrame) + (5 * bytesPerSample), bytesPerSample);
area1.Pointer += area1.Step;
area2.Pointer += area2.Step;
area3.Pointer += area3.Step;
area4.Pointer += area4.Step;
area5.Pointer += area5.Step;
area6.Pointer += area6.Step;
}
}
}
}
// Every other channel count
else
{
SoundIOChannelArea[] channels = new SoundIOChannelArea[channelCount];
// Obtain the channel area for each channel
for (int i = 0; i < channelCount; i++)
{
channels[i] = areas.GetArea(i);
}
fixed (byte* srcptr = samples)
{
for (int frame = 0; frame < frameCount; frame++)
for (int channel = 0; channel < areas.ChannelCount; channel++)
{
// Copy channel by channel, frame by frame. This is slow!
Unsafe.CopyBlockUnaligned((byte*)channels[channel].Pointer, srcptr + (frame * bytesPerFrame) + (channel * bytesPerSample), bytesPerSample);
channels[channel].Pointer += channels[channel].Step;
}
}
}
AudioStream.EndWrite();
PlayedSampleCount += (ulong)samples.Length;
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
UpdateReleasedBuffers(samples.Length);
}
/// <summary>
/// Releases any buffers that have been fully written to the output device
/// </summary>
/// <param name="bytesRead">The amount of bytes written in the last device write</param>
private void UpdateReleasedBuffers(int bytesRead)
{
bool bufferReleased = false;
while (bytesRead > 0)
{
if (m_ReservedBuffers.TryPeek(out SoundIoBuffer buffer))
{
if (buffer.Length > bytesRead)
{
buffer.Length -= bytesRead;
bytesRead = 0;
}
else
{
bufferReleased = true;
bytesRead -= buffer.Length;
m_ReservedBuffers.TryDequeue(out buffer);
ReleasedBuffers.Enqueue(buffer.Tag);
}
}
}
if (bufferReleased)
{
OnBufferReleased();
}
}
/// <summary>
/// Starts audio playback
/// </summary>
public void Start()
{
if (AudioStream == null)
{
return;
}
AudioStream.Start();
AudioStream.Pause(false);
AudioContext.FlushEvents();
State = PlaybackState.Playing;
}
/// <summary>
/// Stops audio playback
/// </summary>
public void Stop()
{
if (AudioStream == null)
{
return;
}
AudioStream.Pause(true);
AudioContext.FlushEvents();
State = PlaybackState.Stopped;
}
/// <summary>
/// Appends an audio buffer to the tracks internal ring buffer
/// </summary>
/// <typeparam name="T">The audio sample type</typeparam>
/// <param name="bufferTag">The unqiue tag of the buffer being appended</param>
/// <param name="buffer">The buffer to append</param>
public void AppendBuffer<T>(long bufferTag, T[] buffer) where T: struct
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
{
if (AudioStream == null)
{
return;
}
int sampleSize = Unsafe.SizeOf<T>();
int targetSize = sampleSize * buffer.Length;
// Do we need to downmix?
if (_hardwareChannels != _virtualChannels)
{
if (sampleSize != sizeof(short))
{
throw new NotImplementedException("Downmixing formats other than PCM16 is not supported!");
}
short[] downmixedBuffer;
ReadOnlySpan<short> bufferPCM16 = MemoryMarshal.Cast<T, short>(buffer);
if (_virtualChannels == 6)
{
downmixedBuffer = Downmixing.DownMixSurroundToStereo(bufferPCM16);
if (_hardwareChannels == 1)
{
downmixedBuffer = Downmixing.DownMixStereoToMono(downmixedBuffer);
}
}
else if (_virtualChannels == 2)
{
downmixedBuffer = Downmixing.DownMixStereoToMono(bufferPCM16);
}
else
{
throw new NotImplementedException($"Downmixing from {_virtualChannels} to {_hardwareChannels} not implemented!");
}
targetSize = sampleSize * downmixedBuffer.Length;
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
// Copy the memory to our ring buffer
m_Buffer.Write(downmixedBuffer, 0, targetSize);
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
// Keep track of "buffered" buffers
m_ReservedBuffers.Enqueue(new SoundIoBuffer(bufferTag, targetSize));
}
else
{
// Copy the memory to our ring buffer
m_Buffer.Write(buffer, 0, targetSize);
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
// Keep track of "buffered" buffers
m_ReservedBuffers.Enqueue(new SoundIoBuffer(bufferTag, targetSize));
}
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
}
/// <summary>
/// Returns a value indicating whether the specified buffer is currently reserved by the track
/// </summary>
/// <param name="bufferTag">The buffer tag to check</param>
public bool ContainsBuffer(long bufferTag)
{
return m_ReservedBuffers.Any(x => x.Tag == bufferTag);
}
/// <summary>
/// Flush all track buffers
/// </summary>
public bool FlushBuffers()
{
m_Buffer.Clear();
if (m_ReservedBuffers.Count > 0)
{
foreach (var buffer in m_ReservedBuffers)
{
ReleasedBuffers.Enqueue(buffer.Tag);
}
OnBufferReleased();
return true;
}
return false;
}
Implement libsoundio as an alternative audio backend (#406) * Audio: Implement libsoundio as an alternative audio backend libsoundio will be preferred over OpenAL if it is available on the machine. If neither are available, it will fallback to a dummy audio renderer that outputs no sound. * Audio: Fix SoundIoRingBuffer documentation * Audio: Unroll and optimize the audio write callback Copying one sample at a time is slow, this unrolls the most common audio channel layouts and manually copies the bytes between source and destination. This is over 2x faster than calling CopyBlockUnaligned every sample. * Audio: Optimize the write callback further This dramatically reduces the audio buffer copy time. When the sample size is one of handled sample sizes the buffer copy operation is almost 10x faster than CopyBlockAligned. This works by copying full samples at a time, rather than the individual bytes that make up the sample. This allows for 2x or 4x faster copy operations depending on sample size. * Audio: Fix typo in Stereo write callback * Audio: Fix Surround (5.1) audio write callback * Audio: Update Documentation * Audio: Use built-in Unsafe.SizeOf<T>() Built-in `SizeOf<T>()` is 10x faster than our `TypeSize<T>` helper. This also helps reduce code surface area. * Audio: Keep fixed buffer style consistent * Audio: Address styling nits * Audio: More style nits * Audio: Add additional documentation * Audio: Move libsoundio bindings internal As per discussion, moving the libsoundio native bindings into Ryujinx.Audio * Audio: Bump Target Framework back up to .NET Core 2.1 * Audio: Remove voice mixing optimizations. Leaves Saturation optimizations in place.
2018-11-14 18:22:50 -08:00
/// <summary>
/// Closes the <see cref="SoundIoAudioTrack"/>
/// </summary>
public void Close()
{
if (AudioStream != null)
{
AudioStream.Pause(true);
AudioStream.Dispose();
}
m_Buffer.Clear();
OnBufferReleased();
ReleasedBuffers.Clear();
State = PlaybackState.Stopped;
AudioStream = null;
BufferReleased = null;
}
private void OnBufferReleased()
{
BufferReleased?.Invoke();
}
/// <summary>
/// Releases the unmanaged resources used by the <see cref="SoundIoAudioTrack" />
/// </summary>
public void Dispose()
{
Close();
}
~SoundIoAudioTrack()
{
Dispose();
}
}
}