early-access version 3384

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 3383.
+This is the source code for early-access 3384.
 
 ## Legal Notice
 

src/audio_core/device/audio_buffers.h

@@ -48,7 +48,7 @@ public:
      *
      * @param out_buffers - The buffers which were registered.
      */
-    void RegisterBuffers(std::span<AudioBuffer> out_buffers, u32& out_size) {
+    void RegisterBuffers(std::vector<AudioBuffer>& out_buffers) {
         std::scoped_lock l{lock};
         const s32 to_register{std::min(std::min(appended_count, BufferAppendLimit),
                                        BufferAppendLimit - registered_count)};
@@ -59,7 +59,7 @@ public:
                 index += N;
             }
 
-            out_buffers[out_size++] = buffers[index];
+            out_buffers.push_back(buffers[index]);
             registered_count++;
             registered_index = (registered_index + 1) % append_limit;
 
@@ -162,7 +162,7 @@ public:
      * @param max_buffers     - Maximum number of buffers to released.
      * @return The number of buffers released.
      */
-    u32 GetRegisteredAppendedBuffers(std::span<AudioBuffer> out_buffers, u32 max_buffers) {
+    u32 GetRegisteredAppendedBuffers(std::vector<AudioBuffer>& buffers_flushed, u32 max_buffers) {
         std::scoped_lock l{lock};
         if (registered_count + appended_count == 0) {
             return 0;
@@ -174,20 +174,19 @@ public:
             return 0;
         }
 
-        u32 buffers_flushed{0};
         while (registered_count > 0) {
             auto index{registered_index - registered_count};
             if (index < 0) {
                 index += N;
             }
 
-            out_buffers[buffers_flushed++] = buffers[index];
+            buffers_flushed.push_back(buffers[index]);
 
             registered_count--;
             released_count++;
             released_index = (released_index + 1) % append_limit;
 
-            if (buffers_flushed >= buffers_to_flush) {
+            if (buffers_flushed.size() >= buffers_to_flush) {
                 break;
             }
         }
@@ -198,18 +197,18 @@ public:
                 index += N;
             }
 
-            out_buffers[buffers_flushed++] = buffers[index];
+            buffers_flushed.push_back(buffers[index]);
 
             appended_count--;
             released_count++;
             released_index = (released_index + 1) % append_limit;
 
-            if (buffers_flushed >= buffers_to_flush) {
+            if (buffers_flushed.size() >= buffers_to_flush) {
                 break;
             }
         }
 
-        return buffers_flushed;
+        return static_cast<u32>(buffers_flushed.size());
     }
 
     /**
@@ -271,9 +270,8 @@ public:
      */
     bool FlushBuffers(u32& buffers_released) {
         std::scoped_lock l{lock};
+        std::vector<AudioBuffer> buffers_flushed{};
 
-        static Common::ScratchBuffer<AudioBuffer> buffers_flushed{};
-        buffers_flushed.resize_destructive(append_limit);
         buffers_released = GetRegisteredAppendedBuffers(buffers_flushed, append_limit);
 
         if (registered_count > 0) {

src/audio_core/device/device_session.cpp

@@ -6,7 +6,6 @@
 #include "audio_core/device/audio_buffer.h"
 #include "audio_core/device/device_session.h"
 #include "audio_core/sink/sink_stream.h"
-#include "common/scratch_buffer.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/memory.h"
@@ -80,22 +79,21 @@ void DeviceSession::ClearBuffers() {
     }
 }
 
-void DeviceSession::AppendBuffers(std::span<const AudioBuffer> buffers, u32 out_size) const {
+void DeviceSession::AppendBuffers(std::span<const AudioBuffer> buffers) const {
-    static Common::ScratchBuffer<s16> samples{};
+    for (const auto& buffer : buffers) {
-
-    for (u32 i = 0; i < out_size; i++) {
         Sink::SinkBuffer new_buffer{
-            .frames = buffers[i].size / (channel_count * sizeof(s16)),
+            .frames = buffer.size / (channel_count * sizeof(s16)),
             .frames_played = 0,
-            .tag = buffers[i].tag,
+            .tag = buffer.tag,
             .consumed = false,
         };
 
         if (type == Sink::StreamType::In) {
+            std::vector<s16> samples{};
             stream->AppendBuffer(new_buffer, samples);
         } else {
-            samples.resize_destructive(buffers[i].size / sizeof(s16));
+            std::vector<s16> samples(buffer.size / sizeof(s16));
-            system.Memory().ReadBlockUnsafe(buffers[i].samples, samples.data(), buffers[i].size);
+            system.Memory().ReadBlockUnsafe(buffer.samples, samples.data(), buffer.size);
             stream->AppendBuffer(new_buffer, samples);
         }
     }

src/audio_core/device/device_session.h

@@ -62,7 +62,7 @@ public:
      *
      * @param buffers - The buffers to play.
      */
-    void AppendBuffers(std::span<const AudioBuffer> buffers, u32 out_size) const;
+    void AppendBuffers(std::span<const AudioBuffer> buffers) const;
 
     /**
      * (Audio In only) Pop samples from the backend, and write them back to this buffer's address.

src/audio_core/in/audio_in_system.cpp

@@ -89,10 +89,9 @@ Result System::Start() {
     session->Start();
     state = State::Started;
 
-    std::array<AudioBuffer, BufferAppendLimit> buffers_to_flush{};
+    std::vector<AudioBuffer> buffers_to_flush{};
-    u32 out_size{0};
+    buffers.RegisterBuffers(buffers_to_flush);
-    buffers.RegisterBuffers(buffers_to_flush, out_size);
+    session->AppendBuffers(buffers_to_flush);
-    session->AppendBuffers(buffers_to_flush, out_size);
     session->SetRingSize(static_cast<u32>(buffers_to_flush.size()));
 
     return ResultSuccess;
@@ -135,10 +134,9 @@ bool System::AppendBuffer(const AudioInBuffer& buffer, const u64 tag) {
 
 void System::RegisterBuffers() {
     if (state == State::Started) {
-        std::array<AudioBuffer, BufferAppendLimit> registered_buffers{};
+        std::vector<AudioBuffer> registered_buffers{};
-        u32 out_size{0};
+        buffers.RegisterBuffers(registered_buffers);
-        buffers.RegisterBuffers(registered_buffers, out_size);
+        session->AppendBuffers(registered_buffers);
-        session->AppendBuffers(registered_buffers, out_size);
     }
 }
 

src/audio_core/out/audio_out_system.cpp

@@ -89,10 +89,9 @@ Result System::Start() {
     session->Start();
     state = State::Started;
 
-    std::array<AudioBuffer, BufferAppendLimit> buffers_to_flush{};
+    std::vector<AudioBuffer> buffers_to_flush{};
-    u32 out_size{0};
+    buffers.RegisterBuffers(buffers_to_flush);
-    buffers.RegisterBuffers(buffers_to_flush, out_size);
+    session->AppendBuffers(buffers_to_flush);
-    session->AppendBuffers(buffers_to_flush, out_size);
     session->SetRingSize(static_cast<u32>(buffers_to_flush.size()));
 
     return ResultSuccess;
@@ -135,10 +134,9 @@ bool System::AppendBuffer(const AudioOutBuffer& buffer, u64 tag) {
 
 void System::RegisterBuffers() {
     if (state == State::Started) {
-        std::array<AudioBuffer, BufferAppendLimit> registered_buffers{};
+        std::vector<AudioBuffer> registered_buffers{};
-        u32 out_size{0};
+        buffers.RegisterBuffers(registered_buffers);
-        buffers.RegisterBuffers(registered_buffers, out_size);
+        session->AppendBuffers(registered_buffers);
-        session->AppendBuffers(registered_buffers, out_size);
     }
 }
 

src/audio_core/renderer/command/command_generator.cpp

@@ -46,7 +46,7 @@ void CommandGenerator::GenerateDataSourceCommand(VoiceInfo& voice_info,
             while (destination != nullptr) {
                 if (destination->IsConfigured()) {
                     auto mix_id{destination->GetMixId()};
-                    if (mix_id < mix_context.GetCount() && mix_id != -1) {
+                    if (mix_id < mix_context.GetCount() && mix_id != UnusedSplitterId) {
                         auto mix_info{mix_context.GetInfo(mix_id)};
                         command_buffer.GenerateDepopPrepareCommand(
                             voice_info.node_id, voice_state, render_context.depop_buffer,

src/audio_core/renderer/command/data_source/decode.cpp

@@ -8,7 +8,6 @@
 #include "audio_core/renderer/command/resample/resample.h"
 #include "common/fixed_point.h"
 #include "common/logging/log.h"
-#include "common/scratch_buffer.h"
 #include "core/memory.h"
 
 namespace AudioCore::AudioRenderer {
@@ -30,7 +29,6 @@ static u32 DecodePcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
                      const DecodeArg& req) {
     constexpr s32 min{std::numeric_limits<s16>::min()};
     constexpr s32 max{std::numeric_limits<s16>::max()};
-    static Common::ScratchBuffer<T> samples;
 
     if (req.buffer == 0 || req.buffer_size == 0) {
         return 0;
@@ -51,7 +49,7 @@ static u32 DecodePcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
         const u64 size{channel_count * samples_to_decode};
         const u64 size_bytes{size * sizeof(T)};
 
-        samples.resize_destructive(size);
+        std::vector<T> samples(size);
         memory.ReadBlockUnsafe(source, samples.data(), size_bytes);
 
         if constexpr (std::is_floating_point_v<T>) {
@@ -75,7 +73,7 @@ static u32 DecodePcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
         }
 
         const VAddr source{req.buffer + ((req.start_offset + req.offset) * sizeof(T))};
-        samples.resize_destructive(samples_to_decode);
+        std::vector<T> samples(samples_to_decode);
         memory.ReadBlockUnsafe(source, samples.data(), samples_to_decode * sizeof(T));
 
         if constexpr (std::is_floating_point_v<T>) {
@@ -105,7 +103,6 @@ static u32 DecodeAdpcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
                        const DecodeArg& req) {
     constexpr u32 SamplesPerFrame{14};
     constexpr u32 NibblesPerFrame{16};
-    static Common::ScratchBuffer<u8> wavebuffer;
 
     if (req.buffer == 0 || req.buffer_size == 0) {
         return 0;
@@ -141,7 +138,7 @@ static u32 DecodeAdpcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
     }
 
     const auto size{std::max((samples_to_process / 8U) * SamplesPerFrame, 8U)};
-    wavebuffer.resize_destructive(size);
+    std::vector<u8> wavebuffer(size);
     memory.ReadBlockUnsafe(req.buffer + position_in_frame / 2, wavebuffer.data(),
                            wavebuffer.size());
 
@@ -230,8 +227,6 @@ static u32 DecodeAdpcm(Core::Memory::Memory& memory, std::span<s16> out_buffer,
  * @param args   - The wavebuffer data, and information for how to decode it.
  */
 void DecodeFromWaveBuffers(Core::Memory::Memory& memory, const DecodeFromWaveBuffersArgs& args) {
-    Common::ScratchBuffer<s16> temp_buffer(TempBufferSize);
-
     auto& voice_state{*args.voice_state};
     auto remaining_sample_count{args.sample_count};
     auto fraction{voice_state.fraction};
@@ -261,8 +256,9 @@ void DecodeFromWaveBuffers(Core::Memory::Memory& memory, const DecodeFromWaveBuf
 
     bool is_buffer_starved{false};
     u32 offset{voice_state.offset};
-    std::memset(temp_buffer.data(), 0, temp_buffer.size() * sizeof(s16));
+
     auto output_buffer{args.output};
+    std::vector<s16> temp_buffer(TempBufferSize, 0);
 
     while (remaining_sample_count > 0) {
         const auto samples_to_write{std::min(remaining_sample_count, max_remaining_sample_count)};

src/audio_core/renderer/command/effect/aux_.cpp

@@ -41,11 +41,10 @@ static void ResetAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr aux_in
  * @param update_count - If non-zero, send_info_ will be updated.
  * @return Number of samples written.
  */
-static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr send_info_,
+static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr send_info_,
-                             [[maybe_unused]] u32 sample_count, const CpuAddr send_buffer,
+                             [[maybe_unused]] u32 sample_count, CpuAddr send_buffer, u32 count_max,
-                             const u32 count_max, std::span<const s32> input,
+                             std::span<const s32> input, u32 write_count_, u32 write_offset,
-                             const u32 write_count_, const u32 write_offset,
+                             u32 update_count) {
-                             const u32 update_count) {
     if (write_count_ > count_max) {
         LOG_ERROR(Service_Audio,
                   "write_count must be smaller than count_max! write_count {}, count_max {}",
@@ -53,6 +52,11 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr send_in
         return 0;
     }
 
+    if (send_info_ == 0) {
+        LOG_ERROR(Service_Audio, "send_info_ is 0!");
+        return 0;
+    }
+
     if (input.empty()) {
         LOG_ERROR(Service_Audio, "input buffer is empty!");
         return 0;
@@ -67,35 +71,30 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr send_in
         return 0;
     }
 
-    AuxInfo::AuxInfoDsp send_info{};
+    auto send_info{reinterpret_cast<AuxInfo::AuxInfoDsp*>(memory.GetPointer(send_info_))};
-    memory.ReadBlockUnsafe(send_info_, &send_info, sizeof(AuxInfo::AuxInfoDsp));
 
-    u32 target_write_offset{send_info.write_offset + write_offset};
+    u32 target_write_offset{send_info->write_offset + write_offset};
-    if (target_write_offset > count_max || write_count_ == 0) {
+    if (target_write_offset > count_max) {
         return 0;
     }
 
     u32 write_count{write_count_};
-    u32 write_pos{0};
+    u32 read_pos{0};
     while (write_count > 0) {
         u32 to_write{std::min(count_max - target_write_offset, write_count)};
-
+        if (to_write) {
-        if (to_write > 0) {
             memory.WriteBlockUnsafe(send_buffer + target_write_offset * sizeof(s32),
-                                    &input[write_pos], to_write * sizeof(s32));
+                                    &input[read_pos], to_write * sizeof(s32));
         }
-
         target_write_offset = (target_write_offset + to_write) % count_max;
         write_count -= to_write;
-        write_pos += to_write;
+        read_pos += to_write;
     }
 
     if (update_count) {
-        send_info.write_offset = (send_info.write_offset + update_count) % count_max;
+        send_info->write_offset = (send_info->write_offset + update_count) % count_max;
     }
 
-    memory.WriteBlockUnsafe(send_info_, &send_info, sizeof(AuxInfo::AuxInfoDsp));
-
     return write_count_;
 }
 
@@ -103,7 +102,7 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr send_in
  * Read the given memory at return_buffer into the output mix buffer, and update return_info_ if
  * update_count is set, to notify the game that an update happened.
  *
- * @param memory        - Core memory for writing.
+ * @param memory        - Core memory for reading.
  * @param return_info_  - Meta information for where to read the mix buffer.
  * @param return_buffer - Memory address to read the samples from.
  * @param count_max     - Maximum number of samples in the receiving buffer.
@@ -113,16 +112,21 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr send_in
  * @param update_count  - If non-zero, send_info_ will be updated.
  * @return Number of samples read.
  */
-static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr return_info_,
+static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr return_info_,
-                            const CpuAddr return_buffer, const u32 count_max, std::span<s32> output,
+                            CpuAddr return_buffer, u32 count_max, std::span<s32> output,
-                            const u32 count_, const u32 read_offset, const u32 update_count) {
+                            u32 read_count_, u32 read_offset, u32 update_count) {
     if (count_max == 0) {
         return 0;
     }
 
-    if (count_ > count_max) {
+    if (read_count_ > count_max) {
         LOG_ERROR(Service_Audio, "count must be smaller than count_max! count {}, count_max {}",
-                  count_, count_max);
+                  read_count_, count_max);
+        return 0;
+    }
+
+    if (return_info_ == 0) {
+        LOG_ERROR(Service_Audio, "return_info_ is 0!");
         return 0;
     }
 
@@ -136,36 +140,31 @@ static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr return_i
         return 0;
     }
 
-    AuxInfo::AuxInfoDsp return_info{};
+    auto return_info{reinterpret_cast<AuxInfo::AuxInfoDsp*>(memory.GetPointer(return_info_))};
-    memory.ReadBlockUnsafe(return_info_, &return_info, sizeof(AuxInfo::AuxInfoDsp));
 
-    u32 target_read_offset{return_info.read_offset + read_offset};
+    u32 target_read_offset{return_info->read_offset + read_offset};
     if (target_read_offset > count_max) {
         return 0;
     }
 
-    u32 read_count{count_};
+    u32 read_count{read_count_};
-    u32 read_pos{0};
+    u32 write_pos{0};
     while (read_count > 0) {
         u32 to_read{std::min(count_max - target_read_offset, read_count)};
-
+        if (to_read) {
-        if (to_read > 0) {
             memory.ReadBlockUnsafe(return_buffer + target_read_offset * sizeof(s32),
-                                   &output[read_pos], to_read * sizeof(s32));
+                                   &output[write_pos], to_read * sizeof(s32));
         }
-
         target_read_offset = (target_read_offset + to_read) % count_max;
         read_count -= to_read;
-        read_pos += to_read;
+        write_pos += to_read;
     }
 
     if (update_count) {
-        return_info.read_offset = (return_info.read_offset + update_count) % count_max;
+        return_info->read_offset = (return_info->read_offset + update_count) % count_max;
     }
 
-    memory.WriteBlockUnsafe(return_info_, &return_info, sizeof(AuxInfo::AuxInfoDsp));
+    return read_count_;
-
-    return count_;
 }
 
 void AuxCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor& processor,
@@ -175,19 +174,6 @@ void AuxCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor& process
 }
 
 void AuxCommand::Process(const ADSP::CommandListProcessor& processor) {
-    // HACK!
-    // Ignore aux for Super Mario Odyssey and Metroid Prime Remastered.
-    // For some reason these games receive output samples, and then send them back in as input
-    // again. Problem is the data being sent back in is slightly offset from the current output by
-    // 240 or 480 samples, leading to a very fast echoing effect, which should not be there.
-    // Timing issue or some bug in the code?
-    // We can't disable this unconditionally as some games rely on it for synchronisation and will
-    // softlock without it (Age of Calamity).
-    const auto program_id = processor.system->GetCurrentProcessProgramID();
-    if (program_id == 0x0100000000010000ull || program_id == 0x010012101468C000ull) {
-        return;
-    }
-
     auto input_buffer{
         processor.mix_buffers.subspan(input * processor.sample_count, processor.sample_count)};
     auto output_buffer{
@@ -203,7 +189,7 @@ void AuxCommand::Process(const ADSP::CommandListProcessor& processor) {
                                    update_count)};
 
         if (read != processor.sample_count) {
-            std::memset(&output_buffer[read], 0, processor.sample_count - read);
+            std::memset(&output_buffer[read], 0, (processor.sample_count - read) * sizeof(s32));
         }
     } else {
         ResetAuxBufferDsp(*processor.memory, send_buffer_info);

src/audio_core/renderer/command/effect/compressor.cpp

@@ -8,7 +8,6 @@
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "audio_core/renderer/command/effect/compressor.h"
 #include "audio_core/renderer/effect/compressor.h"
-#include "common/scratch_buffer.h"
 
 namespace AudioCore::AudioRenderer {
 
@@ -45,8 +44,8 @@ static void InitializeCompressorEffect(const CompressorInfo::ParameterVersion2&
 
 static void ApplyCompressorEffect(const CompressorInfo::ParameterVersion2& params,
                                   CompressorInfo::State& state, bool enabled,
-                                  std::span<std::span<const s32>> input_buffers,
+                                  std::vector<std::span<const s32>> input_buffers,
-                                  std::span<std::span<s32>> output_buffers, u32 sample_count) {
+                                  std::vector<std::span<s32>> output_buffers, u32 sample_count) {
     if (enabled) {
         auto state_00{state.unk_00};
         auto state_04{state.unk_04};
@@ -125,10 +124,8 @@ void CompressorCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor&
 }
 
 void CompressorCommand::Process(const ADSP::CommandListProcessor& processor) {
-    static Common::ScratchBuffer<std::span<const s32>> input_buffers{};
+    std::vector<std::span<const s32>> input_buffers(parameter.channel_count);
-    static Common::ScratchBuffer<std::span<s32>> output_buffers{};
+    std::vector<std::span<s32>> output_buffers(parameter.channel_count);
-    input_buffers.resize_destructive(parameter.channel_count);
-    output_buffers.resize_destructive(parameter.channel_count);
 
     for (s16 i = 0; i < parameter.channel_count; i++) {
         input_buffers[i] = processor.mix_buffers.subspan(inputs[i] * processor.sample_count,

src/audio_core/renderer/command/effect/delay.cpp

@@ -3,7 +3,6 @@
 
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "audio_core/renderer/command/effect/delay.h"
-#include "common/scratch_buffer.h"
 
 namespace AudioCore::AudioRenderer {
 /**
@@ -75,8 +74,8 @@ static void InitializeDelayEffect(const DelayInfo::ParameterVersion1& params,
  */
 template <size_t NumChannels>
 static void ApplyDelay(const DelayInfo::ParameterVersion1& params, DelayInfo::State& state,
-                       std::span<std::span<const s32>> inputs, std::span<std::span<s32>> outputs,
+                       std::vector<std::span<const s32>>& inputs,
-                       const u32 sample_count) {
+                       std::vector<std::span<s32>>& outputs, const u32 sample_count) {
     for (u32 sample_index = 0; sample_index < sample_count; sample_index++) {
         std::array<Common::FixedPoint<50, 14>, NumChannels> input_samples{};
         for (u32 channel = 0; channel < NumChannels; channel++) {
@@ -154,8 +153,8 @@ static void ApplyDelay(const DelayInfo::ParameterVersion1& params, DelayInfo::St
  * @param sample_count - Number of samples to process.
  */
 static void ApplyDelayEffect(const DelayInfo::ParameterVersion1& params, DelayInfo::State& state,
-                             const bool enabled, std::span<std::span<const s32>> inputs,
+                             const bool enabled, std::vector<std::span<const s32>>& inputs,
-                             std::span<std::span<s32>> outputs, const u32 sample_count) {
+                             std::vector<std::span<s32>>& outputs, const u32 sample_count) {
 
     if (!IsChannelCountValid(params.channel_count)) {
         LOG_ERROR(Service_Audio, "Invalid delay channels {}", params.channel_count);
@@ -209,10 +208,8 @@ void DelayCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor& proce
 }
 
 void DelayCommand::Process(const ADSP::CommandListProcessor& processor) {
-    static Common::ScratchBuffer<std::span<const s32>> input_buffers{};
+    std::vector<std::span<const s32>> input_buffers(parameter.channel_count);
-    static Common::ScratchBuffer<std::span<s32>> output_buffers{};
+    std::vector<std::span<s32>> output_buffers(parameter.channel_count);
-    input_buffers.resize_destructive(parameter.channel_count);
-    output_buffers.resize_destructive(parameter.channel_count);
 
     for (s16 i = 0; i < parameter.channel_count; i++) {
         input_buffers[i] = processor.mix_buffers.subspan(inputs[i] * processor.sample_count,

src/audio_core/renderer/command/effect/i3dl2_reverb.cpp

@@ -6,7 +6,6 @@
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "audio_core/renderer/command/effect/i3dl2_reverb.h"
 #include "common/polyfill_ranges.h"
-#include "common/scratch_buffer.h"
 
 namespace AudioCore::AudioRenderer {
 
@@ -409,10 +408,8 @@ void I3dl2ReverbCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor&
 }
 
 void I3dl2ReverbCommand::Process(const ADSP::CommandListProcessor& processor) {
-    static Common::ScratchBuffer<std::span<const s32>> input_buffers{};
+    std::vector<std::span<const s32>> input_buffers(parameter.channel_count);
-    static Common::ScratchBuffer<std::span<s32>> output_buffers{};
+    std::vector<std::span<s32>> output_buffers(parameter.channel_count);
-    input_buffers.resize_destructive(parameter.channel_count);
-    output_buffers.resize_destructive(parameter.channel_count);
 
     for (u32 i = 0; i < parameter.channel_count; i++) {
         input_buffers[i] = processor.mix_buffers.subspan(inputs[i] * processor.sample_count,

src/audio_core/renderer/command/effect/light_limiter.cpp

@@ -3,7 +3,6 @@
 
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "audio_core/renderer/command/effect/light_limiter.h"
-#include "common/scratch_buffer.h"
 
 namespace AudioCore::AudioRenderer {
 /**
@@ -48,8 +47,8 @@ static void InitializeLightLimiterEffect(const LightLimiterInfo::ParameterVersio
  */
 static void ApplyLightLimiterEffect(const LightLimiterInfo::ParameterVersion2& params,
                                     LightLimiterInfo::State& state, const bool enabled,
-                                    std::span<std::span<const s32>> inputs,
+                                    std::vector<std::span<const s32>>& inputs,
-                                    std::span<std::span<s32>> outputs, const u32 sample_count,
+                                    std::vector<std::span<s32>>& outputs, const u32 sample_count,
                                     LightLimiterInfo::StatisticsInternal* statistics) {
     constexpr s64 min{std::numeric_limits<s32>::min()};
     constexpr s64 max{std::numeric_limits<s32>::max()};
@@ -148,10 +147,8 @@ void LightLimiterVersion1Command::Dump([[maybe_unused]] const ADSP::CommandListP
 }
 
 void LightLimiterVersion1Command::Process(const ADSP::CommandListProcessor& processor) {
-    static Common::ScratchBuffer<std::span<const s32>> input_buffers{};
+    std::vector<std::span<const s32>> input_buffers(parameter.channel_count);
-    static Common::ScratchBuffer<std::span<s32>> output_buffers{};
+    std::vector<std::span<s32>> output_buffers(parameter.channel_count);
-    input_buffers.resize_destructive(parameter.channel_count);
-    output_buffers.resize_destructive(parameter.channel_count);
 
     for (u32 i = 0; i < parameter.channel_count; i++) {
         input_buffers[i] = processor.mix_buffers.subspan(inputs[i] * processor.sample_count,
@@ -193,10 +190,8 @@ void LightLimiterVersion2Command::Dump([[maybe_unused]] const ADSP::CommandListP
 }
 
 void LightLimiterVersion2Command::Process(const ADSP::CommandListProcessor& processor) {
-    static Common::ScratchBuffer<std::span<const s32>> input_buffers{};
+    std::vector<std::span<const s32>> input_buffers(parameter.channel_count);
-    static Common::ScratchBuffer<std::span<s32>> output_buffers{};
+    std::vector<std::span<s32>> output_buffers(parameter.channel_count);
-    input_buffers.resize_destructive(parameter.channel_count);
-    output_buffers.resize_destructive(parameter.channel_count);
 
     for (u32 i = 0; i < parameter.channel_count; i++) {
         input_buffers[i] = processor.mix_buffers.subspan(inputs[i] * processor.sample_count,

src/audio_core/renderer/command/effect/reverb.cpp

@@ -7,7 +7,6 @@
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "audio_core/renderer/command/effect/reverb.h"
 #include "common/polyfill_ranges.h"
-#include "common/scratch_buffer.h"
 
 namespace AudioCore::AudioRenderer {
 
@@ -251,8 +250,8 @@ static Common::FixedPoint<50, 14> Axfx2AllPassTick(ReverbInfo::ReverbDelayLine&
  */
 template <size_t NumChannels>
 static void ApplyReverbEffect(const ReverbInfo::ParameterVersion2& params, ReverbInfo::State& state,
-                              std::span<std::span<const s32>> inputs,
+                              std::vector<std::span<const s32>>& inputs,
-                              std::span<std::span<s32>> outputs, const u32 sample_count) {
+                              std::vector<std::span<s32>>& outputs, const u32 sample_count) {
     constexpr std::array<u8, ReverbInfo::MaxDelayTaps> OutTapIndexes1Ch{
         0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
     };
@@ -369,8 +368,8 @@ static void ApplyReverbEffect(const ReverbInfo::ParameterVersion2& params, Rever
  * @param sample_count - Number of samples to process.
  */
 static void ApplyReverbEffect(const ReverbInfo::ParameterVersion2& params, ReverbInfo::State& state,
-                              const bool enabled, std::span<std::span<const s32>> inputs,
+                              const bool enabled, std::vector<std::span<const s32>>& inputs,
-                              std::span<std::span<s32>> outputs, const u32 sample_count) {
+                              std::vector<std::span<s32>>& outputs, const u32 sample_count) {
     if (enabled) {
         switch (params.channel_count) {
         case 0:
@@ -412,10 +411,8 @@ void ReverbCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor& proc
 }
 
 void ReverbCommand::Process(const ADSP::CommandListProcessor& processor) {
-    static Common::ScratchBuffer<std::span<const s32>> input_buffers{};
+    std::vector<std::span<const s32>> input_buffers(parameter.channel_count);
-    static Common::ScratchBuffer<std::span<s32>> output_buffers{};
+    std::vector<std::span<s32>> output_buffers(parameter.channel_count);
-    input_buffers.resize_destructive(parameter.channel_count);
-    output_buffers.resize_destructive(parameter.channel_count);
 
     for (u32 i = 0; i < parameter.channel_count; i++) {
         input_buffers[i] = processor.mix_buffers.subspan(inputs[i] * processor.sample_count,

src/audio_core/renderer/command/sink/circular_buffer.cpp

@@ -5,7 +5,6 @@
 
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "audio_core/renderer/command/sink/circular_buffer.h"
-#include "common/scratch_buffer.h"
 #include "core/memory.h"
 
 namespace AudioCore::AudioRenderer {
@@ -25,9 +24,7 @@ void CircularBufferSinkCommand::Process(const ADSP::CommandListProcessor& proces
     constexpr s32 min{std::numeric_limits<s16>::min()};
     constexpr s32 max{std::numeric_limits<s16>::max()};
 
-    static Common::ScratchBuffer<s16> output{};
+    std::vector<s16> output(processor.sample_count);
-    output.resize_destructive(processor.sample_count);
-
     for (u32 channel = 0; channel < input_count; channel++) {
         auto input{processor.mix_buffers.subspan(inputs[channel] * processor.sample_count,
                                                  processor.sample_count)};

src/audio_core/renderer/command/sink/device.cpp

@@ -33,8 +33,7 @@ void DeviceSinkCommand::Process(const ADSP::CommandListProcessor& processor) {
         .consumed{false},
     };
 
-    static Common::ScratchBuffer<s16> samples{};
+    std::vector<s16> samples(out_buffer.frames * input_count);
-    samples.resize_destructive(out_buffer.frames * input_count);
 
     for (u32 channel = 0; channel < input_count; channel++) {
         const auto offset{inputs[channel] * out_buffer.frames};

src/audio_core/sink/null_sink.h

@@ -9,7 +9,6 @@
 
 #include "audio_core/sink/sink.h"
 #include "audio_core/sink/sink_stream.h"
-#include "common/scratch_buffer.h"
 
 namespace Core {
 class System;
@@ -21,7 +20,7 @@ public:
     explicit NullSinkStreamImpl(Core::System& system_, StreamType type_)
         : SinkStream{system_, type_} {}
     ~NullSinkStreamImpl() override {}
-    void AppendBuffer(SinkBuffer&, Common::ScratchBuffer<s16>&) override {}
+    void AppendBuffer(SinkBuffer&, std::vector<s16>&) override {}
     std::vector<s16> ReleaseBuffer(u64) override {
         return {};
     }

src/audio_core/sink/sink_stream.cpp

@@ -17,7 +17,7 @@
 
 namespace AudioCore::Sink {
 
-void SinkStream::AppendBuffer(SinkBuffer& buffer, Common::ScratchBuffer<s16>& samples) {
+void SinkStream::AppendBuffer(SinkBuffer& buffer, std::vector<s16>& samples) {
     if (type == StreamType::In) {
         queue.enqueue(buffer);
         queued_buffers++;
@@ -71,9 +71,7 @@ void SinkStream::AppendBuffer(SinkBuffer& buffer, Common::ScratchBuffer<s16>& sa
         // We need moar samples! Not all games will provide 6 channel audio.
         // TODO: Implement some upmixing here. Currently just passthrough, with other
         // channels left as silence.
-        static Common::ScratchBuffer<s16> new_samples{};
+        std::vector<s16> new_samples(samples.size() / system_channels * device_channels, 0);
-        new_samples.resize_destructive(samples.size() / system_channels * device_channels);
-        std::memset(new_samples.data(), 0, new_samples.size() * sizeof(s16));
 
         for (u32 read_index = 0, write_index = 0; read_index < samples.size();
              read_index += system_channels, write_index += device_channels) {
@@ -102,7 +100,7 @@ void SinkStream::AppendBuffer(SinkBuffer& buffer, Common::ScratchBuffer<s16>& sa
         }
     }
 
-    samples_buffer.Push(samples.data(), samples.size());
+    samples_buffer.Push(samples);
     queue.enqueue(buffer);
     queued_buffers++;
 }

src/audio_core/sink/sink_stream.h

@@ -14,7 +14,6 @@
 #include "common/common_types.h"
 #include "common/reader_writer_queue.h"
 #include "common/ring_buffer.h"
-#include "common/scratch_buffer.h"
 
 namespace Core {
 class System;
@@ -170,7 +169,7 @@ public:
      * @param buffer  - Audio buffer information to be queued.
      * @param samples - The s16 samples to be queue for playback.
      */
-    virtual void AppendBuffer(SinkBuffer& buffer, Common::ScratchBuffer<s16>& samples);
+    virtual void AppendBuffer(SinkBuffer& buffer, std::vector<s16>& samples);
 
     /**
      * Release a buffer. Audio In only, will fill a buffer with recorded samples.

src/common/scratch_buffer.h

@@ -24,24 +24,6 @@ public:
 
     ~ScratchBuffer() = default;
 
-    ScratchBuffer(ScratchBuffer&& rhs) {
-        last_requested_size = rhs.last_requested_size;
-        buffer_capacity = rhs.buffer_capacity;
-        buffer = std::move(rhs.buffer);
-
-        rhs.last_requested_size = 0;
-        rhs.buffer_capacity = 0;
-    }
-
-    void operator=(ScratchBuffer&& rhs) {
-        last_requested_size = rhs.last_requested_size;
-        buffer_capacity = rhs.buffer_capacity;
-        buffer = std::move(rhs.buffer);
-
-        rhs.last_requested_size = 0;
-        rhs.buffer_capacity = 0;
-    }
-
     /// This will only grow the buffer's capacity if size is greater than the current capacity.
     /// The previously held data will remain intact.
     void resize(size_t size) {

src/video_core/buffer_cache/buffer_base.h

@@ -289,6 +289,11 @@ public:
         return cpu_addr;
     }
 
+    /// Update the base CPU address of the buffer
+    void UpdateCpuAddr(VAddr cpu_addr_) noexcept {
+        cpu_addr = cpu_addr_;
+    }
+
     /// Returns the offset relative to the given CPU address
     /// @pre IsInBounds returns true
     [[nodiscard]] u32 Offset(VAddr other_cpu_addr) const noexcept {

src/video_core/buffer_cache/buffer_cache.h

@@ -205,9 +205,9 @@ public:
         current_draw_indirect = current_draw_indirect_;
     }
 
-    [[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectCount();
+    [[nodiscard]] typename BufferCache<P>::Buffer* GetDrawIndirectCount();
 
-    [[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectBuffer();
+    [[nodiscard]] typename BufferCache<P>::Buffer* GetDrawIndirectBuffer();
 
     std::recursive_mutex mutex;
     Runtime& runtime;
@@ -385,6 +385,9 @@ private:
     SlotVector<Buffer> slot_buffers;
     DelayedDestructionRing<Buffer, 8> delayed_destruction_ring;
 
+    Buffer indirect_buffer;
+    Buffer indirect_count_buffer;
+
     const Tegra::Engines::DrawManager::IndirectParams* current_draw_indirect{};
 
     u32 last_index_count = 0;
@@ -458,7 +461,9 @@ private:
 template <class P>
 BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
                             Core::Memory::Memory& cpu_memory_, Runtime& runtime_)
-    : runtime{runtime_}, rasterizer{rasterizer_}, cpu_memory{cpu_memory_} {
+    : runtime{runtime_}, rasterizer{rasterizer_}, cpu_memory{cpu_memory_},
+      indirect_buffer(runtime, rasterizer, 0, 0x1000),
+      indirect_count_buffer(runtime, rasterizer, 0, 0x1000) {
     // Ensure the first slot is used for the null buffer
     void(slot_buffers.insert(runtime, NullBufferParams{}));
     common_ranges.clear();
@@ -1064,15 +1069,13 @@ void BufferCache<P>::BindHostVertexBuffers() {
 
 template <class P>
 void BufferCache<P>::BindHostDrawIndirectBuffers() {
-    const auto bind_buffer = [this](const Binding& binding) {
+    const auto bind_buffer = [this](const Binding& binding, Buffer& buffer) {
-        Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer, binding.buffer_id);
         SynchronizeBuffer(buffer, binding.cpu_addr, binding.size);
     };
     if (current_draw_indirect->include_count) {
-        bind_buffer(count_buffer_binding);
+        bind_buffer(count_buffer_binding, indirect_count_buffer);
     }
-    bind_buffer(indirect_buffer_binding);
+    bind_buffer(indirect_buffer_binding, indirect_buffer);
 }
 
 template <class P>
@@ -1409,14 +1412,16 @@ void BufferCache<P>::UpdateDrawIndirect() {
         binding = Binding{
             .cpu_addr = *cpu_addr,
             .size = static_cast<u32>(size),
-            .buffer_id = FindBuffer(*cpu_addr, static_cast<u32>(size)),
+            .buffer_id = NULL_BUFFER_ID,
         };
     };
     if (current_draw_indirect->include_count) {
         update(current_draw_indirect->count_start_address, sizeof(u32), count_buffer_binding);
+        indirect_count_buffer.UpdateCpuAddr(count_buffer_binding.cpu_addr);
     }
     update(current_draw_indirect->indirect_start_address, current_draw_indirect->buffer_size,
            indirect_buffer_binding);
+    indirect_buffer.UpdateCpuAddr(indirect_buffer_binding.cpu_addr);
 }
 
 template <class P>
@@ -2000,15 +2005,13 @@ bool BufferCache<P>::HasFastUniformBufferBound(size_t stage, u32 binding_index)
 }
 
 template <class P>
-std::pair<typename BufferCache<P>::Buffer*, u32> BufferCache<P>::GetDrawIndirectCount() {
+typename BufferCache<P>::Buffer* BufferCache<P>::GetDrawIndirectCount() {
-    auto& buffer = slot_buffers[count_buffer_binding.buffer_id];
+    return &indirect_count_buffer;
-    return std::make_pair(&buffer, buffer.Offset(count_buffer_binding.cpu_addr));
 }
 
 template <class P>
-std::pair<typename BufferCache<P>::Buffer*, u32> BufferCache<P>::GetDrawIndirectBuffer() {
+typename BufferCache<P>::Buffer* BufferCache<P>::GetDrawIndirectBuffer() {
-    auto& buffer = slot_buffers[indirect_buffer_binding.buffer_id];
+    return &indirect_buffer;
-    return std::make_pair(&buffer, buffer.Offset(indirect_buffer_binding.cpu_addr));
 }
 
 } // namespace VideoCommon

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -285,23 +285,19 @@ void RasterizerOpenGL::DrawIndirect() {
     const auto& params = maxwell3d->draw_manager->GetIndirectParams();
     buffer_cache.SetDrawIndirect(&params);
     PrepareDraw(params.is_indexed, [this, &params](GLenum primitive_mode) {
-        const auto [buffer, offset] = buffer_cache.GetDrawIndirectBuffer();
+        const auto buffer = buffer_cache.GetDrawIndirectBuffer();
-        const GLvoid* const gl_offset =
-            reinterpret_cast<const GLvoid*>(static_cast<uintptr_t>(offset));
         glBindBuffer(GL_DRAW_INDIRECT_BUFFER, buffer->Handle());
         if (params.include_count) {
-            const auto [draw_buffer, offset_base] = buffer_cache.GetDrawIndirectCount();
+            const auto draw_buffer = buffer_cache.GetDrawIndirectCount();
             glBindBuffer(GL_PARAMETER_BUFFER, draw_buffer->Handle());
 
             if (params.is_indexed) {
                 const GLenum format = MaxwellToGL::IndexFormat(maxwell3d->regs.index_buffer.format);
-                glMultiDrawElementsIndirectCount(primitive_mode, format, gl_offset,
+                glMultiDrawElementsIndirectCount(primitive_mode, format, nullptr, 0,
-                                                 static_cast<GLintptr>(offset_base),
                                                  static_cast<GLsizei>(params.max_draw_counts),
                                                  static_cast<GLsizei>(params.stride));
             } else {
-                glMultiDrawArraysIndirectCount(primitive_mode, gl_offset,
+                glMultiDrawArraysIndirectCount(primitive_mode, nullptr, 0,
-                                               static_cast<GLintptr>(offset_base),
                                                static_cast<GLsizei>(params.max_draw_counts),
                                                static_cast<GLsizei>(params.stride));
             }
@@ -309,11 +305,11 @@ void RasterizerOpenGL::DrawIndirect() {
         }
         if (params.is_indexed) {
             const GLenum format = MaxwellToGL::IndexFormat(maxwell3d->regs.index_buffer.format);
-            glMultiDrawElementsIndirect(primitive_mode, format, gl_offset,
+            glMultiDrawElementsIndirect(primitive_mode, format, nullptr,
                                         static_cast<GLsizei>(params.max_draw_counts),
                                         static_cast<GLsizei>(params.stride));
         } else {
-            glMultiDrawArraysIndirect(primitive_mode, gl_offset,
+            glMultiDrawArraysIndirect(primitive_mode, nullptr,
                                       static_cast<GLsizei>(params.max_draw_counts),
                                       static_cast<GLsizei>(params.stride));
         }

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -230,35 +230,29 @@ void RasterizerVulkan::DrawIndirect() {
     const auto& params = maxwell3d->draw_manager->GetIndirectParams();
     buffer_cache.SetDrawIndirect(&params);
     PrepareDraw(params.is_indexed, [this, &params] {
-        const auto indirect_buffer = buffer_cache.GetDrawIndirectBuffer();
+        const auto buffer = buffer_cache.GetDrawIndirectBuffer();
-        const auto& buffer = indirect_buffer.first;
-        const auto& offset = indirect_buffer.second;
         if (params.include_count) {
-            const auto count = buffer_cache.GetDrawIndirectCount();
+            const auto draw_buffer = buffer_cache.GetDrawIndirectCount();
-            const auto& draw_buffer = count.first;
-            const auto& offset_base = count.second;
             scheduler.Record([draw_buffer_obj = draw_buffer->Handle(),
-                              buffer_obj = buffer->Handle(), offset_base, offset,
+                              buffer_obj = buffer->Handle(), params](vk::CommandBuffer cmdbuf) {
-                              params](vk::CommandBuffer cmdbuf) {
                 if (params.is_indexed) {
-                    cmdbuf.DrawIndexedIndirectCount(
+                    cmdbuf.DrawIndexedIndirectCount(buffer_obj, 0, draw_buffer_obj, 0,
-                        buffer_obj, offset, draw_buffer_obj, offset_base,
+                                                    static_cast<u32>(params.max_draw_counts),
-                        static_cast<u32>(params.max_draw_counts), static_cast<u32>(params.stride));
+                                                    static_cast<u32>(params.stride));
                 } else {
-                    cmdbuf.DrawIndirectCount(buffer_obj, offset, draw_buffer_obj, offset_base,
+                    cmdbuf.DrawIndirectCount(buffer_obj, 0, draw_buffer_obj, 0,
                                              static_cast<u32>(params.max_draw_counts),
                                              static_cast<u32>(params.stride));
                 }
             });
             return;
         }
-        scheduler.Record([buffer_obj = buffer->Handle(), offset, params](vk::CommandBuffer cmdbuf) {
+        scheduler.Record([buffer_obj = buffer->Handle(), params](vk::CommandBuffer cmdbuf) {
             if (params.is_indexed) {
-                cmdbuf.DrawIndexedIndirect(buffer_obj, offset,
+                cmdbuf.DrawIndexedIndirect(buffer_obj, 0, static_cast<u32>(params.max_draw_counts),
-                                           static_cast<u32>(params.max_draw_counts),
                                            static_cast<u32>(params.stride));
             } else {
-                cmdbuf.DrawIndirect(buffer_obj, offset, static_cast<u32>(params.max_draw_counts),
+                cmdbuf.DrawIndirect(buffer_obj, 0, static_cast<u32>(params.max_draw_counts),
                                     static_cast<u32>(params.stride));
             }
         });