starred/duckstation
1
0
Fork 0
mirror of https://github.com/stenzek/duckstation.git synced 2026-02-04 05:04:33 +00:00
Code Issues 1.1k Packages Projects Releases 20 Wiki Activity

SoundEffectManager: Support converting bit depth

Browse Source
This commit is contained in:
Stenzek
2026-01-06 00:51:04 +10:00
parent 2e659d8cf8
commit 4522449156
3 changed files with 183 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

172
src/core/sound_effect_manager.cpp
View File

@@ -129,10 +129,12 @@ static void MixFrames(AudioStream::SampleType* dest, const AudioStream::SampleTy
/// Stops the stream if there are no active sounds.
static void StopStreamIfInactive();
static void ConvertTo16Bit(const void* src, std::span<AudioStream::SampleType> dst, u32 in_bits_per_sample);
static void ConvertToStereo(std::span<const AudioStream::SampleType> src, std::span<AudioStream::SampleType> dst,
u32 in_channels);
static bool ConvertFrames(CachedEffect& effect, u32 in_sample_rate, u32 in_channels, u32 in_frames, Error* error);
static bool ConvertFrames(CachedEffect& effect, const void* frames_in, u32 in_bits_per_sample, u32 in_sample_rate,
u32 in_channels, u32 in_frames, Error* error);
static PlayingResampledEffect CreateResampledStreamedEffect(WAVReader&& reader, Error* error);
@@ -226,14 +228,19 @@ bool SoundEffectManager::LoadCachedEffect(const std::string& resource_name, cons
effect->frames.resize(parsed->num_frames * parsed->num_channels);
if (parsed->num_frames > 0)
{
std::memcpy(effect->frames.data(), parsed->sample_data, parsed->num_frames * parsed->bytes_per_frame);
SpeexResamplerStatePtr resampler_state;
if (parsed->sample_rate != SAMPLE_RATE || parsed->num_channels != NUM_CHANNELS)
if (parsed->bits_per_sample != 16 || parsed->sample_rate != SAMPLE_RATE || parsed->num_channels != NUM_CHANNELS)
{
if (!ConvertFrames(*effect, parsed->sample_data, parsed->bits_per_sample, parsed->sample_rate,
parsed->num_channels, parsed->num_frames, error))
{
if (!ConvertFrames(*effect, parsed->sample_rate, parsed->num_channels, parsed->num_frames, error))
return false;
}
}
else
{
std::memcpy(effect->frames.data(), parsed->sample_data, parsed->num_frames * parsed->bytes_per_frame);
}
DEV_LOG("Loaded effect '{}' with {} frames.", resource_name, effect->frames.size() / NUM_CHANNELS);
}
@@ -285,7 +292,8 @@ void SoundEffectManager::StreamSoundEffect(std::string path)
}
PlayingResampledEffect resampled;
if (reader.GetSampleRate() != SAMPLE_RATE || reader.GetNumChannels() != NUM_CHANNELS)
if (reader.GetSampleRate() != SAMPLE_RATE || reader.GetNumChannels() != NUM_CHANNELS ||
reader.GetBitsPerSample() != 16)
{
resampled = CreateResampledStreamedEffect(std::move(reader), &error);
if (!resampled)
@@ -531,6 +539,95 @@ void SoundEffectManager::MixFrames(AudioStream::SampleType* dest, const AudioStr
}
}
void SoundEffectManager::ConvertTo16Bit(const void* src, std::span<AudioStream::SampleType> dst, u32 in_bits_per_sample)
{
const u32 num_samples = static_cast<u32>(dst.size());
DebugAssert(in_bits_per_sample != 16);
if (in_bits_per_sample == 8)
{
// Convert 8-bit unsigned to 16-bit signed
const u8* src_ptr = static_cast<const u8*>(src);
AudioStream::SampleType* dst_ptr = dst.data();
u32 i = 0;
#ifdef CPU_ARCH_SIMD
static constexpr u32 SAMPLES_PER_VEC = 16;
const u32 num_samples_aligned = Common::AlignDownPow2(num_samples, SAMPLES_PER_VEC);
for (; i < num_samples_aligned; i += SAMPLES_PER_VEC)
{
const GSVector4i vsrc =
GSVector4i::load<false>(src_ptr) ^ GSVector4i::cxpr8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128,
-128, -128, -128, -128, -128, -128);
const GSVector4i vlow = vsrc.upl8(vsrc).sll16<8>();
const GSVector4i vhi = vsrc.uph8(vsrc).sll16<8>();
GSVector4i::store<false>(dst_ptr, vlow);
GSVector4i::store<false>(dst_ptr + 8, vhi);
src_ptr += SAMPLES_PER_VEC;
dst_ptr += SAMPLES_PER_VEC;
}
#endif
for (; i < num_samples; i++)
{
const s16 sample = (static_cast<s16>(*(src_ptr++)) ^ 0x80) << 8;
*(dst_ptr++) = sample;
}
}
else if (in_bits_per_sample == 24)
{
// Convert 24-bit signed to 16-bit signed
const u8* src_ptr = static_cast<const u8*>(src);
AudioStream::SampleType* dst_ptr = dst.data();
for (u32 i = 0; i < num_samples; i++)
{
const s32 sample =
(static_cast<s32>(src_ptr[0]) | (static_cast<s32>(src_ptr[1]) << 8) | (static_cast<s32>(src_ptr[2]) << 16));
// sign extend
const s32 signed_sample = (sample << 8) >> 8;
*(dst_ptr++) = static_cast<s16>(signed_sample >> 8);
src_ptr += 3;
}
}
else if (in_bits_per_sample == 32)
{
// Convert 32-bit signed to 16-bit signed
const s32* src_ptr = static_cast<const s32*>(src);
AudioStream::SampleType* dst_ptr = dst.data();
u32 i = 0;
#ifdef CPU_ARCH_SIMD
static constexpr u32 SAMPLES_PER_VEC = 8;
const u32 num_samples_aligned = Common::AlignDownPow2(num_samples, SAMPLES_PER_VEC);
for (; i < num_samples_aligned; i += SAMPLES_PER_VEC)
{
const GSVector4i vsrclow = GSVector4i::load<false>(src_ptr).sra32<16>();
src_ptr += 4;
const GSVector4i vsrchigh = GSVector4i::load<false>(src_ptr).sra32<16>();
src_ptr += 4;
const GSVector4i vdest = vsrclow.ps32(vsrchigh);
GSVector4i::store<false>(dst_ptr, vdest);
dst_ptr += SAMPLES_PER_VEC;
}
#endif
i = 0;
src_ptr = static_cast<const s32*>(src);
dst_ptr = dst.data();
for (; i < num_samples; i++)
{
const s32 sample = *(src_ptr++);
*(dst_ptr++) = static_cast<s16>(sample >> 16);
}
}
else
{
ERROR_LOG("Unsupported bits per sample: {}", in_bits_per_sample);
std::memset(dst.data(), 0, num_samples * sizeof(AudioStream::SampleType));
}
}
void SoundEffectManager::ConvertToStereo(std::span<const AudioStream::SampleType> src,
std::span<AudioStream::SampleType> dst, u32 in_channels)
{
@@ -593,16 +690,31 @@ void SoundEffectManager::SpeexResamplerStateDeleter::operator()(SpeexResamplerSt
speex_resampler_destroy(state);
}
bool SoundEffectManager::ConvertFrames(CachedEffect& effect, u32 in_sample_rate, u32 in_channels, u32 in_frames,
Error* error)
bool SoundEffectManager::ConvertFrames(CachedEffect& effect, const void* frames_in, u32 in_bits_per_sample,
u32 in_sample_rate, u32 in_channels, u32 in_frames, Error* error)
{
DynamicHeapArray<AudioStream::SampleType> temp_frames;
std::span<const s16> frames_in_span;
if (in_bits_per_sample != 16)
{
temp_frames.resize(in_frames * in_channels);
ConvertTo16Bit(frames_in, temp_frames, in_bits_per_sample);
effect.frames.swap(temp_frames);
frames_in_span = effect.frames;
}
else
{
frames_in_span = std::span<const s16>(static_cast<const s16*>(frames_in), in_frames * in_channels);
}
if (in_channels != NUM_CHANNELS)
{
temp_frames.resize(in_frames * NUM_CHANNELS);
ConvertToStereo(effect.frames, temp_frames, in_channels);
ConvertToStereo(frames_in_span, temp_frames, in_channels);
effect.frames.swap(temp_frames);
frames_in_span = effect.frames;
}
if (in_sample_rate != SAMPLE_RATE)
@@ -643,7 +755,7 @@ bool SoundEffectManager::ConvertFrames(CachedEffect& effect, u32 in_sample_rate,
unsigned int frames_processed = in_frames - input_frames_count;
unsigned int frames_generated = (static_cast<u32>(temp_frames.size() / NUM_CHANNELS) - output_frame_count);
const int ret = speex_resampler_process_interleaved_int(
resampler_state.get(), (frames_processed > 0) ? &effect.frames[input_frames_count * NUM_CHANNELS] : nullptr,
resampler_state.get(), (frames_processed > 0) ? &frames_in_span[input_frames_count * NUM_CHANNELS] : nullptr,
&frames_processed, &temp_frames[output_frame_count * NUM_CHANNELS], &frames_generated);
if (ret != RESAMPLER_ERR_SUCCESS)
{
@@ -700,16 +812,25 @@ u32 SoundEffectManager::ReadEntryFrames(PlayingResampledEffect& effect, AudioStr
if (effect->input_buffer_pos == effect->input_buffer_size && effect->reader.GetRemainingFrames() > 0)
{
const bool needs_upmix = (effect->reader.GetNumChannels() != NUM_CHANNELS);
if (needs_upmix)
const bool needs_shift = (effect->reader.GetBitsPerSample() != 16);
if (needs_upmix || needs_shift)
{
const u32 required_buffer_space = ResampledStreamedEffect::INPUT_BUFFER_SIZE * effect->reader.GetNumChannels();
u32 required_buffer_space = ((ResampledStreamedEffect::INPUT_BUFFER_SIZE * effect->reader.GetNumChannels() *
effect->reader.GetBytesPerFrame()) +
(sizeof(s16) - 1)) /
sizeof(s16);
// use second half of buffer as scratch when doing both
if (needs_upmix && needs_shift)
required_buffer_space += ResampledStreamedEffect::INPUT_BUFFER_SIZE * effect->reader.GetNumChannels();
if (required_buffer_space > s_locals.temp_buffer.size())
s_locals.temp_buffer.resize(required_buffer_space);
}
Error error;
const std::optional<u32> frames =
effect->reader.ReadFrames(needs_upmix ? s_locals.temp_buffer.data() : effect->input_buffer.data(),
const std::optional<u32> frames = effect->reader.ReadFrames(
(needs_upmix || needs_shift) ? s_locals.temp_buffer.data() : effect->input_buffer.data(),
ResampledStreamedEffect::INPUT_BUFFER_SIZE, &error);
if (!frames.has_value())
{
@@ -717,11 +838,26 @@ u32 SoundEffectManager::ReadEntryFrames(PlayingResampledEffect& effect, AudioStr
break;
}
if (needs_upmix && frames.value() > 0)
if (frames.value() > 0)
{
ConvertToStereo(s_locals.temp_buffer.cspan(0, frames.value() * effect->reader.GetNumChannels()),
std::span(effect->input_buffer).subspan(0, frames.value() * NUM_CHANNELS),
effect->reader.GetNumChannels());
const u32 total_samples = frames.value() * effect->reader.GetNumChannels();
const std::span<AudioStream::SampleType> final_frames_out =
std::span(effect->input_buffer).subspan(0, frames.value() * NUM_CHANNELS);
std::span<AudioStream::SampleType> frames_in;
if (needs_shift)
{
std::span<AudioStream::SampleType> frames_out =
needs_upmix ? s_locals.temp_buffer.span(s_locals.temp_buffer.size() - total_samples) : final_frames_out;
ConvertTo16Bit(s_locals.temp_buffer.data(), frames_out, effect->reader.GetBitsPerSample());
frames_in = frames_out;
}
else
{
frames_in = s_locals.temp_buffer.span(0, total_samples);
}
if (needs_upmix)
ConvertToStereo(frames_in, final_frames_out, effect->reader.GetNumChannels());
}
effect->input_buffer_pos = 0;

33
src/util/wav_reader_writer.cpp
View File

@@ -64,9 +64,10 @@ WAVReader::WAVReader(WAVReader&& move)
{
m_file = std::exchange(move.m_file, nullptr);
m_frames_start = std::exchange(move.m_frames_start, 0);
m_bits_per_sample = std::exchange(move.m_bits_per_sample, static_cast<u8>(0));
m_num_channels = std::exchange(move.m_num_channels, static_cast<u8>(0));
m_bytes_per_frame = std::exchange(move.m_bytes_per_frame, static_cast<u16>(0));
m_sample_rate = std::exchange(move.m_sample_rate, 0);
m_num_channels = std::exchange(move.m_num_channels, static_cast<u16>(0));
m_num_frames = std::exchange(move.m_num_frames, 0);
m_current_frame = std::exchange(move.m_current_frame, 0);
}
@@ -81,8 +82,9 @@ WAVReader& WAVReader::operator=(WAVReader&& move)
{
m_file = std::exchange(move.m_file, nullptr);
m_frames_start = std::exchange(move.m_frames_start, 0);
m_bits_per_sample = std::exchange(move.m_bits_per_sample, static_cast<u8>(0));
m_num_channels = std::exchange(move.m_num_channels, static_cast<u8>(0));
m_bytes_per_frame = std::exchange(move.m_bytes_per_frame, static_cast<u16>(0));
m_num_channels = std::exchange(move.m_num_channels, static_cast<u16>(0));
m_sample_rate = std::exchange(move.m_sample_rate, 0);
m_num_frames = std::exchange(move.m_num_frames, 0);
m_current_frame = std::exchange(move.m_current_frame, 0);
@@ -162,7 +164,9 @@ bool WAVReader::Open(const char* path, Error* error /*= nullptr*/)
return false;
}
if (format.sample_rate == 0 || format.num_channels == 0 || format.bits_per_sample != 16)
if (format.sample_rate == 0 || format.num_channels == 0 ||
(format.bits_per_sample != 8 && format.bits_per_sample != 16 && format.bits_per_sample != 24 &&
format.bits_per_sample != 32))
{
Error::SetStringFmt(error, "Unsupported file format samplerate={} channels={} bits={}", format.sample_rate,
format.num_channels, format.bits_per_sample);
@@ -176,7 +180,8 @@ bool WAVReader::Open(const char* path, Error* error /*= nullptr*/)
return false;
}
const u32 num_frames = data.chunk_size / (sizeof(s16) * format.num_channels);
const u32 bytes_per_frame = (format.bits_per_sample / 8) * format.num_channels;
const u32 num_frames = (bytes_per_frame > 0) ? (data.chunk_size / bytes_per_frame) : 0;
if (num_frames == 0)
{
Error::SetStringFmt(error, "File has no frames");
@@ -185,9 +190,10 @@ bool WAVReader::Open(const char* path, Error* error /*= nullptr*/)
m_file = fp.release();
m_frames_start = FileSystem::FTell64(m_file);
m_bits_per_sample = static_cast<u8>(format.bits_per_sample);
m_num_channels = static_cast<u8>(format.num_channels);
m_bytes_per_frame = static_cast<s16>(bytes_per_frame);
m_sample_rate = format.sample_rate;
m_bytes_per_frame = sizeof(s16) * format.num_channels;
m_num_channels = format.num_channels;
m_num_frames = num_frames;
m_current_frame = 0;
return true;
@@ -201,8 +207,9 @@ void WAVReader::Close()
std::fclose(m_file);
m_file = nullptr;
m_frames_start = 0;
m_bytes_per_frame = 0;
m_bits_per_sample = 0;
m_num_channels = 0;
m_bytes_per_frame = 0;
m_sample_rate = 0;
m_num_frames = 0;
m_current_frame = 0;
@@ -213,7 +220,9 @@ std::FILE* WAVReader::TakeFile()
std::FILE* ret = std::exchange(m_file, nullptr);
m_frames_start = 0;
m_bytes_per_frame = 0;
m_bits_per_sample = 0;
m_num_channels = 0;
m_bytes_per_frame = 0;
m_sample_rate = 0;
m_num_frames = 0;
m_current_frame = 0;
@@ -349,7 +358,9 @@ std::optional<WAVReader::MemoryParseResult> WAVReader::ParseMemory(const void* d
return result;
}
if (format.sample_rate == 0 || format.num_channels == 0 || format.bits_per_sample != 16)
if (format.sample_rate == 0 || format.num_channels == 0 ||
(format.bits_per_sample != 8 && format.bits_per_sample != 16 && format.bits_per_sample != 24 &&
format.bits_per_sample != 32))
{
Error::SetStringFmt(error, "Unsupported file format samplerate={} channels={} bits={}", format.sample_rate,
format.num_channels, format.bits_per_sample);
@@ -364,7 +375,8 @@ std::optional<WAVReader::MemoryParseResult> WAVReader::ParseMemory(const void* d
return result;
}
const u32 num_frames = static_cast<u32>(sample_data->size() / (sizeof(s16) * format.num_channels));
const u32 bytes_per_frame = (format.bits_per_sample / 8) * format.num_channels;
const u32 num_frames = (bytes_per_frame > 0) ? (static_cast<u32>(sample_data->size() / bytes_per_frame)) : 0;
if (num_frames == 0)
{
Error::SetStringFmt(error, "File has no frames");
@@ -372,8 +384,9 @@ std::optional<WAVReader::MemoryParseResult> WAVReader::ParseMemory(const void* d
}
result.emplace();
result->bytes_per_frame = sizeof(s16) * format.num_channels;
result->bits_per_sample = format.bits_per_sample;
result->sample_rate = format.sample_rate;
result->bytes_per_frame = bytes_per_frame;
result->num_channels = format.num_channels;
result->num_frames = num_frames;
result->sample_data = sample_data->data();

7
src/util/wav_reader_writer.h
View File

@@ -25,6 +25,7 @@ public:
ALWAYS_INLINE u32 GetSampleRate() const { return m_sample_rate; }
ALWAYS_INLINE u32 GetNumChannels() const { return m_num_channels; }
ALWAYS_INLINE u32 GetNumFrames() const { return m_num_frames; }
ALWAYS_INLINE u32 GetBitsPerSample() const { return m_bits_per_sample; }
ALWAYS_INLINE u32 GetBytesPerFrame() const { return m_bytes_per_frame; }
ALWAYS_INLINE u64 GetFramesStartOffset() const { return m_frames_start; }
ALWAYS_INLINE bool IsOpen() const { return (m_file != nullptr); }
@@ -43,8 +44,9 @@ public:
struct MemoryParseResult
{
u32 bytes_per_frame;
u32 bits_per_sample;
u32 sample_rate;
u32 bytes_per_frame;
u32 num_channels;
u32 num_frames;
const void* sample_data;
@@ -57,8 +59,9 @@ private:
std::FILE* m_file = nullptr;
s64 m_frames_start = 0;
u8 m_bits_per_sample = 0;
u8 m_num_channels = 0;
u16 m_bytes_per_frame = 0;
u16 m_num_channels = 0;
u32 m_sample_rate = 0;
u32 m_num_frames = 0;
u32 m_current_frame = 0;