Mirrors/yuzu
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Use spans over guest memory where possible instead of copying data.

Browse source
This commit is contained in:
Kelebek1 2023-05-29 00:35:51 +01:00
parent 95ceae40e6
commit 6f7cb69c94
22 changed files with 462 additions and 233 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/audio_core/device/device_session.cpp
View file

@ -92,9 +92,9 @@ void DeviceSession::AppendBuffers(std::span<const AudioBuffer> buffers) {
if (type == Sink::StreamType::In) {
stream->AppendBuffer(new_buffer, tmp_samples);
} else {
system.ApplicationMemory().ReadBlockUnsafe(buffer.samples, tmp_samples.data(),
buffer.size);
stream->AppendBuffer(new_buffer, tmp_samples);
Core::Memory::CpuGuestMemory<s16, Core::Memory::GuestMemoryFlags::UnsafeRead> samples(
system.ApplicationMemory(), buffer.samples, buffer.size / sizeof(s16));
stream->AppendBuffer(new_buffer, samples);
}
}
}

21
src/audio_core/renderer/command/data_source/decode.cpp
View file

@ -28,7 +28,6 @@ constexpr std::array<u8, 3> PitchBySrcQuality = {4, 8, 4};
template <typename T>
static u32 DecodePcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
const DecodeArg& req) {
std::array<T, TempBufferSize> tmp_samples{};
constexpr s32 min{std::numeric_limits<s16>::min()};
constexpr s32 max{std::numeric_limits<s16>::max()};
@ -49,19 +48,18 @@ static u32 DecodePcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
const VAddr source{req.buffer +
(((req.start_offset + req.offset) * channel_count) * sizeof(T))};
const u64 size{channel_count * samples_to_decode};
const u64 size_bytes{size * sizeof(T)};
memory.ReadBlockUnsafe(source, tmp_samples.data(), size_bytes);
Core::Memory::CpuGuestMemory<T, Core::Memory::GuestMemoryFlags::UnsafeRead> samples(
memory, source, size);
if constexpr (std::is_floating_point_v<T>) {
for (u32 i = 0; i < samples_to_decode; i++) {
auto sample{static_cast<s32>(tmp_samples[i * channel_count + req.target_channel] *
auto sample{static_cast<s32>(samples[i * channel_count + req.target_channel] *
std::numeric_limits<s16>::max())};
out_buffer[i] = static_cast<s16>(std::clamp(sample, min, max));
}
} else {
for (u32 i = 0; i < samples_to_decode; i++) {
out_buffer[i] = tmp_samples[i * channel_count + req.target_channel];
out_buffer[i] = samples[i * channel_count + req.target_channel];
}
}
} break;
@ -74,16 +72,17 @@ static u32 DecodePcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
}
const VAddr source{req.buffer + ((req.start_offset + req.offset) * sizeof(T))};
memory.ReadBlockUnsafe(source, tmp_samples.data(), samples_to_decode * sizeof(T));
Core::Memory::CpuGuestMemory<T, Core::Memory::GuestMemoryFlags::UnsafeRead> samples(
memory, source, samples_to_decode);
if constexpr (std::is_floating_point_v<T>) {
for (u32 i = 0; i < samples_to_decode; i++) {
auto sample{static_cast<s32>(tmp_samples[i * channel_count + req.target_channel] *
auto sample{static_cast<s32>(samples[i * channel_count + req.target_channel] *
std::numeric_limits<s16>::max())};
out_buffer[i] = static_cast<s16>(std::clamp(sample, min, max));
}
} else {
std::memcpy(out_buffer.data(), tmp_samples.data(), samples_to_decode * sizeof(s16));
std::memcpy(out_buffer.data(), samples.data(), samples_to_decode * sizeof(s16));
}
break;
}
@ -101,7 +100,6 @@ static u32 DecodePcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
*/
static u32 DecodeAdpcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
const DecodeArg& req) {
std::array<u8, TempBufferSize> wavebuffer{};
constexpr u32 SamplesPerFrame{14};
constexpr u32 NibblesPerFrame{16};
@ -139,7 +137,8 @@ static u32 DecodeAdpcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
}
const auto size{std::max((samples_to_process / 8U) * SamplesPerFrame, 8U)};
memory.ReadBlockUnsafe(req.buffer + position_in_frame / 2, wavebuffer.data(), size);
Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> wavebuffer(
memory, req.buffer + position_in_frame / 2, size);
auto context{req.adpcm_context};
auto header{context->header};

82
src/audio_core/renderer/command/effect/aux_.cpp
View file

@ -21,23 +21,13 @@ static void ResetAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr aux_in
}
AuxInfo::AuxInfoDsp info{};
auto info_ptr{&info};
bool host_safe{(aux_info & Core::Memory::YUZU_PAGEMASK) <=
(Core::Memory::YUZU_PAGESIZE - sizeof(AuxInfo::AuxInfoDsp))};
memory.ReadBlockUnsafe(aux_info, &info, sizeof(AuxInfo::AuxInfoDsp));
if (host_safe) [[likely]] {
info_ptr = memory.GetPointer<AuxInfo::AuxInfoDsp>(aux_info);
} else {
memory.ReadBlockUnsafe(aux_info, info_ptr, sizeof(AuxInfo::AuxInfoDsp));
}
info.read_offset = 0;
info.write_offset = 0;
info.total_sample_count = 0;
info_ptr->read_offset = 0;
info_ptr->write_offset = 0;
info_ptr->total_sample_count = 0;
if (!host_safe) [[unlikely]] {
memory.WriteBlockUnsafe(aux_info, info_ptr, sizeof(AuxInfo::AuxInfoDsp));
}
memory.WriteBlockUnsafe(aux_info, &info, sizeof(AuxInfo::AuxInfoDsp));
}
/**
@ -86,17 +76,9 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr send_info_,
}
AuxInfo::AuxInfoDsp send_info{};
auto send_ptr = &send_info;
bool host_safe = (send_info_ & Core::Memory::YUZU_PAGEMASK) <=
(Core::Memory::YUZU_PAGESIZE - sizeof(AuxInfo::AuxInfoDsp));
memory.ReadBlockUnsafe(send_info_, &send_info, sizeof(AuxInfo::AuxInfoDsp));
if (host_safe) [[likely]] {
send_ptr = memory.GetPointer<AuxInfo::AuxInfoDsp>(send_info_);
} else {
memory.ReadBlockUnsafe(send_info_, send_ptr, sizeof(AuxInfo::AuxInfoDsp));
}
u32 target_write_offset{send_ptr->write_offset + write_offset};
u32 target_write_offset{send_info.write_offset + write_offset};
if (target_write_offset > count_max) {
return 0;
}
@ -105,15 +87,9 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr send_info_,
u32 read_pos{0};
while (write_count > 0) {
u32 to_write{std::min(count_max - target_write_offset, write_count)};
const auto write_addr = send_buffer + target_write_offset * sizeof(s32);
bool write_safe{(write_addr & Core::Memory::YUZU_PAGEMASK) <=
(Core::Memory::YUZU_PAGESIZE - (write_addr + to_write * sizeof(s32)))};
if (write_safe) [[likely]] {
auto ptr = memory.GetPointer(write_addr);
std::memcpy(ptr, &input[read_pos], to_write * sizeof(s32));
} else {
memory.WriteBlockUnsafe(send_buffer + target_write_offset * sizeof(s32),
&input[read_pos], to_write * sizeof(s32));
if (to_write > 0) {
const auto write_addr = send_buffer + target_write_offset * sizeof(s32);
memory.WriteBlockUnsafe(write_addr, &input[read_pos], to_write * sizeof(s32));
}
target_write_offset = (target_write_offset + to_write) % count_max;
write_count -= to_write;
@ -121,13 +97,10 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr send_info_,
}
if (update_count) {
send_ptr->write_offset = (send_ptr->write_offset + update_count) % count_max;
}
if (!host_safe) [[unlikely]] {
memory.WriteBlockUnsafe(send_info_, send_ptr, sizeof(AuxInfo::AuxInfoDsp));
send_info.write_offset = (send_info.write_offset + update_count) % count_max;
}
memory.WriteBlockUnsafe(send_info_, &send_info, sizeof(AuxInfo::AuxInfoDsp));
return write_count_;
}
@ -174,17 +147,9 @@ static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr return_info_,
}
AuxInfo::AuxInfoDsp return_info{};
auto return_ptr = &return_info;
bool host_safe = (return_info_ & Core::Memory::YUZU_PAGEMASK) <=
(Core::Memory::YUZU_PAGESIZE - sizeof(AuxInfo::AuxInfoDsp));
memory.ReadBlockUnsafe(return_info_, &return_info, sizeof(AuxInfo::AuxInfoDsp));
if (host_safe) [[likely]] {
return_ptr = memory.GetPointer<AuxInfo::AuxInfoDsp>(return_info_);
} else {
memory.ReadBlockUnsafe(return_info_, return_ptr, sizeof(AuxInfo::AuxInfoDsp));
}
u32 target_read_offset{return_ptr->read_offset + read_offset};
u32 target_read_offset{return_info.read_offset + read_offset};
if (target_read_offset > count_max) {
return 0;
}
@ -193,15 +158,9 @@ static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr return_info_,
u32 write_pos{0};
while (read_count > 0) {
u32 to_read{std::min(count_max - target_read_offset, read_count)};
const auto read_addr = return_buffer + target_read_offset * sizeof(s32);
bool read_safe{(read_addr & Core::Memory::YUZU_PAGEMASK) <=
(Core::Memory::YUZU_PAGESIZE - (read_addr + to_read * sizeof(s32)))};
if (read_safe) [[likely]] {
auto ptr = memory.GetPointer(read_addr);
std::memcpy(&output[write_pos], ptr, to_read * sizeof(s32));
} else {
memory.ReadBlockUnsafe(return_buffer + target_read_offset * sizeof(s32),
&output[write_pos], to_read * sizeof(s32));
if (to_read > 0) {
const auto read_addr = return_buffer + target_read_offset * sizeof(s32);
memory.ReadBlockUnsafe(read_addr, &output[write_pos], to_read * sizeof(s32));
}
target_read_offset = (target_read_offset + to_read) % count_max;
read_count -= to_read;
@ -209,13 +168,10 @@ static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr return_info_,
}
if (update_count) {
return_ptr->read_offset = (return_ptr->read_offset + update_count) % count_max;
}
if (!host_safe) [[unlikely]] {
memory.WriteBlockUnsafe(return_info_, return_ptr, sizeof(AuxInfo::AuxInfoDsp));
return_info.read_offset = (return_info.read_offset + update_count) % count_max;
}
memory.WriteBlockUnsafe(return_info_, &return_info, sizeof(AuxInfo::AuxInfoDsp));
return read_count_;
}