yuzu/src/audio_core/audio_renderer.cpp

269 lines
11 KiB
C++
Raw Normal View History

// Copyright 2018 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <vector>
#include "audio_core/audio_out.h"
#include "audio_core/audio_renderer.h"
#include "audio_core/common.h"
#include "audio_core/info_updater.h"
#include "audio_core/voice_context.h"
#include "common/logging/log.h"
#include "core/hle/kernel/writable_event.h"
#include "core/memory.h"
#include "core/settings.h"
namespace AudioCore {
AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory::Memory& memory_,
AudioCommon::AudioRendererParameter params,
std::shared_ptr<Kernel::WritableEvent> buffer_event,
2019-07-12 15:35:40 +00:00
std::size_t instance_number)
: worker_params{params}, buffer_event{buffer_event},
memory_pool_info(params.effect_count + params.voice_count * 4),
voice_context(params.voice_count), effect_context(params.effect_count), mix_context(),
sink_context(params.sink_count), splitter_context(),
voices(params.voice_count), memory{memory_},
2020-08-16 15:23:55 +00:00
command_generator(worker_params, voice_context, mix_context, splitter_context, effect_context,
memory),
temp_mix_buffer(AudioCommon::TOTAL_TEMP_MIX_SIZE) {
behavior_info.SetUserRevision(params.revision);
splitter_context.Initialize(behavior_info, params.splitter_count,
params.num_splitter_send_channels);
2020-08-16 15:23:55 +00:00
mix_context.Initialize(behavior_info, params.submix_count + 1, params.effect_count);
audio_out = std::make_unique<AudioCore::AudioOut>();
stream =
audio_out->OpenStream(core_timing, params.sample_rate, AudioCommon::STREAM_NUM_CHANNELS,
fmt::format("AudioRenderer-Instance{}", instance_number),
[=]() { buffer_event->Signal(); });
audio_out->StartStream(stream);
QueueMixedBuffer(0);
QueueMixedBuffer(1);
QueueMixedBuffer(2);
2020-07-12 13:56:24 +00:00
QueueMixedBuffer(3);
}
AudioRenderer::~AudioRenderer() = default;
u32 AudioRenderer::GetSampleRate() const {
return worker_params.sample_rate;
}
u32 AudioRenderer::GetSampleCount() const {
return worker_params.sample_count;
}
u32 AudioRenderer::GetMixBufferCount() const {
return worker_params.mix_buffer_count;
}
Stream::State AudioRenderer::GetStreamState() const {
2018-09-23 12:32:01 +00:00
return stream->GetState();
}
static constexpr s16 ClampToS16(s32 value) {
return static_cast<s16>(std::clamp(value, -32768, 32767));
}
ResultCode AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_params,
std::vector<u8>& output_params) {
InfoUpdater info_updater{input_params, output_params, behavior_info};
if (!info_updater.UpdateBehaviorInfo(behavior_info)) {
LOG_ERROR(Audio, "Failed to update behavior info input parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
if (!info_updater.UpdateMemoryPools(memory_pool_info)) {
LOG_ERROR(Audio, "Failed to update memory pool parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
if (!info_updater.UpdateVoiceChannelResources(voice_context)) {
LOG_ERROR(Audio, "Failed to update voice channel resource parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
if (!info_updater.UpdateVoices(voice_context, memory_pool_info, 0)) {
LOG_ERROR(Audio, "Failed to update voice parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
// TODO(ogniK): Deal with stopped audio renderer but updates still taking place
if (!info_updater.UpdateEffects(effect_context, true)) {
LOG_ERROR(Audio, "Failed to update effect parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
if (behavior_info.IsSplitterSupported()) {
if (!info_updater.UpdateSplitterInfo(splitter_context)) {
LOG_ERROR(Audio, "Failed to update splitter parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
}
2020-08-16 15:23:55 +00:00
auto mix_result = info_updater.UpdateMixes(mix_context, worker_params.mix_buffer_count,
splitter_context, effect_context);
if (mix_result.IsError()) {
LOG_ERROR(Audio, "Failed to update mix parameters");
return mix_result;
}
// TODO(ogniK): Sinks
if (!info_updater.UpdateSinks(sink_context)) {
LOG_ERROR(Audio, "Failed to update sink parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
// TODO(ogniK): Performance buffer
if (!info_updater.UpdatePerformanceBuffer()) {
LOG_ERROR(Audio, "Failed to update performance buffer parameters");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
if (!info_updater.UpdateErrorInfo(behavior_info)) {
LOG_ERROR(Audio, "Failed to update error info");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
if (behavior_info.IsElapsedFrameCountSupported()) {
if (!info_updater.UpdateRendererInfo(elapsed_frame_count)) {
LOG_ERROR(Audio, "Failed to update renderer info");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
}
// TODO(ogniK): Statistics
if (!info_updater.WriteOutputHeader()) {
LOG_ERROR(Audio, "Failed to write output header");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
}
2018-08-12 18:32:39 +00:00
// TODO(ogniK): Check when all sections are implemented
if (!info_updater.CheckConsumedSize()) {
LOG_ERROR(Audio, "Audio buffers were not consumed!");
return AudioCommon::Audren::ERR_INVALID_PARAMETERS;
2018-10-07 03:14:09 +00:00
}
ReleaseAndQueueBuffers();
return RESULT_SUCCESS;
}
void AudioRenderer::QueueMixedBuffer(Buffer::Tag tag) {
command_generator.PreCommand();
// Clear mix buffers before our next operation
command_generator.ClearMixBuffers();
// If the splitter is not in use, sort our mixes
if (!splitter_context.UsingSplitter()) {
mix_context.SortInfo();
}
// Sort our voices
voice_context.SortInfo();
// Handle samples
command_generator.GenerateVoiceCommands();
command_generator.GenerateSubMixCommands();
command_generator.GenerateFinalMixCommands();
command_generator.PostCommand();
// Base sample size
std::size_t BUFFER_SIZE{worker_params.sample_count};
// Samples
std::vector<s16> buffer(BUFFER_SIZE * stream->GetNumChannels());
// Make sure to clear our samples
std::memset(buffer.data(), 0, buffer.size() * sizeof(s16));
if (sink_context.InUse()) {
const auto stream_channel_count = stream->GetNumChannels();
const auto buffer_offsets = sink_context.OutputBuffers();
const auto channel_count = buffer_offsets.size();
const auto& final_mix = mix_context.GetFinalMixInfo();
const auto& in_params = final_mix.GetInParams();
std::vector<s32*> mix_buffers(channel_count);
for (std::size_t i = 0; i < channel_count; i++) {
2020-10-21 02:07:39 +00:00
mix_buffers[i] =
command_generator.GetMixBuffer(in_params.buffer_offset + buffer_offsets[i]);
}
for (std::size_t i = 0; i < BUFFER_SIZE; i++) {
if (channel_count == 1) {
const auto sample = ClampToS16(mix_buffers[0][i]);
buffer[i * stream_channel_count + 0] = sample;
if (stream_channel_count > 1) {
buffer[i * stream_channel_count + 1] = sample;
}
if (stream_channel_count == 6) {
buffer[i * stream_channel_count + 2] = sample;
buffer[i * stream_channel_count + 4] = sample;
buffer[i * stream_channel_count + 5] = sample;
}
} else if (channel_count == 2) {
const auto l_sample = ClampToS16(mix_buffers[0][i]);
const auto r_sample = ClampToS16(mix_buffers[1][i]);
if (stream_channel_count == 1) {
buffer[i * stream_channel_count + 0] = l_sample;
} else if (stream_channel_count == 2) {
buffer[i * stream_channel_count + 0] = l_sample;
buffer[i * stream_channel_count + 1] = r_sample;
} else if (stream_channel_count == 6) {
buffer[i * stream_channel_count + 0] = l_sample;
buffer[i * stream_channel_count + 1] = r_sample;
buffer[i * stream_channel_count + 2] =
ClampToS16((static_cast<s32>(l_sample) + static_cast<s32>(r_sample)) / 2);
buffer[i * stream_channel_count + 4] = l_sample;
buffer[i * stream_channel_count + 5] = r_sample;
}
} else if (channel_count == 6) {
const auto fl_sample = ClampToS16(mix_buffers[0][i]);
const auto fr_sample = ClampToS16(mix_buffers[1][i]);
const auto fc_sample = ClampToS16(mix_buffers[2][i]);
const auto lf_sample = ClampToS16(mix_buffers[3][i]);
const auto bl_sample = ClampToS16(mix_buffers[4][i]);
const auto br_sample = ClampToS16(mix_buffers[5][i]);
if (stream_channel_count == 1) {
buffer[i * stream_channel_count + 0] = fc_sample;
} else if (stream_channel_count == 2) {
buffer[i * stream_channel_count + 0] =
static_cast<s16>(0.3694f * static_cast<float>(fl_sample) +
0.2612f * static_cast<float>(fc_sample) +
0.3694f * static_cast<float>(bl_sample));
buffer[i * stream_channel_count + 1] =
static_cast<s16>(0.3694f * static_cast<float>(fr_sample) +
0.2612f * static_cast<float>(fc_sample) +
0.3694f * static_cast<float>(br_sample));
} else if (stream_channel_count == 6) {
buffer[i * stream_channel_count + 0] = fl_sample;
buffer[i * stream_channel_count + 1] = fr_sample;
buffer[i * stream_channel_count + 2] = fc_sample;
buffer[i * stream_channel_count + 3] = lf_sample;
buffer[i * stream_channel_count + 4] = bl_sample;
buffer[i * stream_channel_count + 5] = br_sample;
}
}
}
}
audio_out->QueueBuffer(stream, tag, std::move(buffer));
elapsed_frame_count++;
voice_context.UpdateStateByDspShared();
}
void AudioRenderer::ReleaseAndQueueBuffers() {
const auto released_buffers{audio_out->GetTagsAndReleaseBuffers(stream, 2)};
for (const auto& tag : released_buffers) {
QueueMixedBuffer(tag);
}
}
} // namespace AudioCore