mirror of
https://github.com/yuzu-emu/yuzu.git
synced 2024-12-03 18:24:18 +01:00
audio_renderer: Better voice mixing and 6 channel downmixing
Supersedes #3738 and #3321
This commit is contained in:
parent
6ec6cb50dd
commit
9de860a419
@ -37,9 +37,11 @@ public:
|
||||
}
|
||||
|
||||
void SetWaveIndex(std::size_t index);
|
||||
std::vector<s16> DequeueSamples(std::size_t sample_count, Core::Memory::Memory& memory);
|
||||
std::vector<s16> DequeueSamples(std::size_t sample_count, Core::Memory::Memory& memory,
|
||||
std::array<VoiceResourceInformation*, 6> voice_resources);
|
||||
void UpdateState();
|
||||
void RefreshBuffer(Core::Memory::Memory& memory);
|
||||
void RefreshBuffer(Core::Memory::Memory& memory,
|
||||
std::array<VoiceResourceInformation*, 6> voice_resources);
|
||||
|
||||
private:
|
||||
bool is_in_use{};
|
||||
@ -79,7 +81,7 @@ AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory
|
||||
std::shared_ptr<Kernel::WritableEvent> buffer_event,
|
||||
std::size_t instance_number)
|
||||
: worker_params{params}, buffer_event{buffer_event}, voices(params.voice_count),
|
||||
effects(params.effect_count), memory{memory_} {
|
||||
voice_resources(params.voice_count), effects(params.effect_count), memory{memory_} {
|
||||
behavior_info.SetUserRevision(params.revision);
|
||||
audio_out = std::make_unique<AudioCore::AudioOut>();
|
||||
stream = audio_out->OpenStream(core_timing, STREAM_SAMPLE_RATE, STREAM_NUM_CHANNELS,
|
||||
@ -127,6 +129,12 @@ ResultVal<std::vector<u8>> AudioRenderer::UpdateAudioRenderer(const std::vector<
|
||||
input_params.data() + sizeof(UpdateDataHeader) + config.behavior_size,
|
||||
memory_pool_count * sizeof(MemoryPoolInfo));
|
||||
|
||||
// Copy voice resources
|
||||
const std::size_t voice_resource_offset{sizeof(UpdateDataHeader) + config.behavior_size +
|
||||
config.memory_pools_size};
|
||||
std::memcpy(voice_resources.data(), input_params.data() + voice_resource_offset,
|
||||
sizeof(VoiceResourceInformation) * voice_resources.size());
|
||||
|
||||
// Copy VoiceInfo structs
|
||||
std::size_t voice_offset{sizeof(UpdateDataHeader) + config.behavior_size +
|
||||
config.memory_pools_size + config.voice_resource_size};
|
||||
@ -220,14 +228,15 @@ void AudioRenderer::VoiceState::SetWaveIndex(std::size_t index) {
|
||||
is_refresh_pending = true;
|
||||
}
|
||||
|
||||
std::vector<s16> AudioRenderer::VoiceState::DequeueSamples(std::size_t sample_count,
|
||||
Core::Memory::Memory& memory) {
|
||||
std::vector<s16> AudioRenderer::VoiceState::DequeueSamples(
|
||||
std::size_t sample_count, Core::Memory::Memory& memory,
|
||||
std::array<VoiceResourceInformation*, 6> voice_resources) {
|
||||
if (!IsPlaying()) {
|
||||
return {};
|
||||
}
|
||||
|
||||
if (is_refresh_pending) {
|
||||
RefreshBuffer(memory);
|
||||
RefreshBuffer(memory, voice_resources);
|
||||
}
|
||||
|
||||
const std::size_t max_size{samples.size() - offset};
|
||||
@ -271,7 +280,8 @@ void AudioRenderer::VoiceState::UpdateState() {
|
||||
is_in_use = info.is_in_use;
|
||||
}
|
||||
|
||||
void AudioRenderer::VoiceState::RefreshBuffer(Core::Memory::Memory& memory) {
|
||||
void AudioRenderer::VoiceState::RefreshBuffer(
|
||||
Core::Memory::Memory& memory, std::array<VoiceResourceInformation*, 6> voice_resources) {
|
||||
const auto wave_buffer_address = info.wave_buffer[wave_index].buffer_addr;
|
||||
const auto wave_buffer_size = info.wave_buffer[wave_index].buffer_sz;
|
||||
std::vector<s16> new_samples(wave_buffer_size / sizeof(s16));
|
||||
@ -296,17 +306,75 @@ void AudioRenderer::VoiceState::RefreshBuffer(Core::Memory::Memory& memory) {
|
||||
}
|
||||
|
||||
switch (info.channel_count) {
|
||||
case 1:
|
||||
case 1: {
|
||||
// 1 channel is upsampled to 2 channel
|
||||
samples.resize(new_samples.size() * 2);
|
||||
|
||||
for (std::size_t index = 0; index < new_samples.size(); ++index) {
|
||||
samples[index * 2] = new_samples[index];
|
||||
samples[index * 2 + 1] = new_samples[index];
|
||||
auto sample = static_cast<float>(new_samples[index]);
|
||||
if (voice_resources[0]->in_use) {
|
||||
sample *= voice_resources[0]->mix_volumes[0];
|
||||
}
|
||||
|
||||
samples[index * 2] = static_cast<s16>(sample);
|
||||
samples[index * 2 + 1] = static_cast<s16>(sample);
|
||||
}
|
||||
break;
|
||||
}
|
||||
case 2: {
|
||||
// 2 channel is played as is
|
||||
samples = std::move(new_samples);
|
||||
const std::size_t sample_count = (samples.size() / 2);
|
||||
for (std::size_t index = 0; index < sample_count; ++index) {
|
||||
const std::size_t index_l = index * 2;
|
||||
const std::size_t index_r = index * 2 + 1;
|
||||
|
||||
auto sample_l = static_cast<float>(samples[index_l]);
|
||||
auto sample_r = static_cast<float>(samples[index_r]);
|
||||
|
||||
if (voice_resources[0]->in_use) {
|
||||
sample_l *= voice_resources[0]->mix_volumes[0];
|
||||
}
|
||||
|
||||
if (voice_resources[1]->in_use) {
|
||||
sample_l *= voice_resources[1]->mix_volumes[1];
|
||||
}
|
||||
|
||||
samples[index_l] = static_cast<s16>(sample_l);
|
||||
samples[index_r] = static_cast<s16>(sample_r);
|
||||
}
|
||||
break;
|
||||
}
|
||||
case 6: {
|
||||
samples.resize((new_samples.size() / 6) * 2);
|
||||
const std::size_t sample_count = samples.size() / 2;
|
||||
|
||||
for (std::size_t index = 0; index < sample_count; ++index) {
|
||||
auto FL = static_cast<float>(new_samples[index * 6]);
|
||||
auto FR = static_cast<float>(new_samples[index * 6 + 1]);
|
||||
auto FC = static_cast<float>(new_samples[index * 6 + 2]);
|
||||
auto BL = static_cast<float>(new_samples[index * 6 + 4]);
|
||||
auto BR = static_cast<float>(new_samples[index * 6 + 5]);
|
||||
|
||||
if (voice_resources[0]->in_use) {
|
||||
FL *= voice_resources[0]->mix_volumes[0];
|
||||
}
|
||||
if (voice_resources[1]->in_use) {
|
||||
FR *= voice_resources[1]->mix_volumes[1];
|
||||
}
|
||||
if (voice_resources[2]->in_use) {
|
||||
FC *= voice_resources[2]->mix_volumes[2];
|
||||
}
|
||||
if (voice_resources[4]->in_use) {
|
||||
BL *= voice_resources[4]->mix_volumes[4];
|
||||
}
|
||||
if (voice_resources[5]->in_use) {
|
||||
BR *= voice_resources[5]->mix_volumes[5];
|
||||
}
|
||||
|
||||
samples[index * 2] = static_cast<s16>(0.3694f * FL + 0.2612f * FC + 0.3694f * BL);
|
||||
samples[index * 2 + 1] = static_cast<s16>(0.3694f * FR + 0.2612f * FC + 0.3694f * BR);
|
||||
}
|
||||
break;
|
||||
}
|
||||
default:
|
||||
@ -352,11 +420,17 @@ void AudioRenderer::QueueMixedBuffer(Buffer::Tag tag) {
|
||||
if (!voice.IsPlaying()) {
|
||||
continue;
|
||||
}
|
||||
std::array<VoiceResourceInformation*, 6> resources{};
|
||||
for (u32 channel = 0; channel < voice.GetInfo().channel_count; channel++) {
|
||||
const auto channel_resource_id = voice.GetInfo().voice_channel_resource_ids[channel];
|
||||
resources[channel] = &voice_resources[channel_resource_id];
|
||||
}
|
||||
|
||||
std::size_t offset{};
|
||||
s64 samples_remaining{BUFFER_SIZE};
|
||||
while (samples_remaining > 0) {
|
||||
const std::vector<s16> samples{voice.DequeueSamples(samples_remaining, memory)};
|
||||
const std::vector<s16> samples{
|
||||
voice.DequeueSamples(samples_remaining, memory, resources)};
|
||||
|
||||
if (samples.empty()) {
|
||||
break;
|
||||
|
@ -9,6 +9,7 @@
|
||||
#include <vector>
|
||||
|
||||
#include "audio_core/behavior_info.h"
|
||||
#include "audio_core/common.h"
|
||||
#include "audio_core/stream.h"
|
||||
#include "common/common_funcs.h"
|
||||
#include "common/common_types.h"
|
||||
@ -116,6 +117,14 @@ struct WaveBuffer {
|
||||
};
|
||||
static_assert(sizeof(WaveBuffer) == 0x38, "WaveBuffer has wrong size");
|
||||
|
||||
struct VoiceResourceInformation {
|
||||
s32_le id{};
|
||||
std::array<float_le, MAX_MIX_BUFFERS> mix_volumes{};
|
||||
bool in_use{};
|
||||
INSERT_PADDING_BYTES(11);
|
||||
};
|
||||
static_assert(sizeof(VoiceResourceInformation) == 0x70, "VoiceResourceInformation has wrong size");
|
||||
|
||||
struct VoiceInfo {
|
||||
u32_le id;
|
||||
u32_le node_id;
|
||||
@ -244,6 +253,7 @@ private:
|
||||
AudioRendererParameter worker_params;
|
||||
std::shared_ptr<Kernel::WritableEvent> buffer_event;
|
||||
std::vector<VoiceState> voices;
|
||||
std::vector<VoiceResourceInformation> voice_resources;
|
||||
std::vector<EffectState> effects;
|
||||
std::unique_ptr<AudioOut> audio_out;
|
||||
StreamPtr stream;
|
||||
|
@ -14,6 +14,7 @@ constexpr ResultCode ERR_INVALID_PARAMETERS{ErrorModule::Audio, 41};
|
||||
}
|
||||
|
||||
constexpr u32_le CURRENT_PROCESS_REVISION = Common::MakeMagic('R', 'E', 'V', '8');
|
||||
constexpr std::size_t MAX_MIX_BUFFERS = 24;
|
||||
|
||||
static constexpr u32 VersionFromRevision(u32_le rev) {
|
||||
// "REV7" -> 7
|
||||
|
Loading…
Reference in New Issue
Block a user