"Merge Tagged PR 9588"

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()) {
+                    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/effect/aux_.cpp

@@ -4,6 +4,7 @@
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "audio_core/renderer/command/effect/aux_.h"
 #include "audio_core/renderer/effect/aux_.h"
+#include "core/core.h"
 #include "core/memory.h"
 
 namespace AudioCore::AudioRenderer {
@@ -19,10 +20,24 @@ static void ResetAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr aux_in
         return;
     }
 
-    auto info{reinterpret_cast<AuxInfo::AuxInfoDsp*>(memory.GetPointer(aux_info))};
-    info->read_offset = 0;
-    info->write_offset = 0;
-    info->total_sample_count = 0;
+    AuxInfo::AuxInfoDsp info{};
+    auto info_ptr{&info};
+    bool host_safe{(aux_info & Core::Memory::YUZU_PAGEMASK) <=
+                   (Core::Memory::YUZU_PAGESIZE - 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_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));
+    }
 }
 
 /**
@@ -40,11 +55,10 @@ 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_,
-                             [[maybe_unused]] u32 sample_count, const CpuAddr send_buffer,
-                             const u32 count_max, std::span<const s32> input,
-                             const u32 write_count_, const u32 write_offset,
-                             const u32 update_count) {
+static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr send_info_,
+                             [[maybe_unused]] u32 sample_count, CpuAddr send_buffer, u32 count_max,
+                             std::span<const s32> input, u32 write_count_, u32 write_offset,
+                             u32 update_count) {
     if (write_count_ > count_max) {
         LOG_ERROR(Service_Audio,
                   "write_count must be smaller than count_max! write_count {}, count_max {}",
@@ -52,6 +66,11 @@ 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;
@@ -67,33 +86,47 @@ static u32 WriteAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr send_in
     }
 
     AuxInfo::AuxInfoDsp send_info{};
-    memory.ReadBlockUnsafe(send_info_, &send_info, sizeof(AuxInfo::AuxInfoDsp));
+    auto send_ptr = &send_info;
+    bool host_safe = (send_info_ & Core::Memory::YUZU_PAGEMASK) <=
+                     (Core::Memory::YUZU_PAGESIZE - sizeof(AuxInfo::AuxInfoDsp));
 
-    u32 target_write_offset{send_info.write_offset + write_offset};
-    if (target_write_offset > count_max || write_count_ == 0) {
+    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};
+    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 > 0) {
+        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[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_ptr->write_offset = (send_ptr->write_offset + update_count) % count_max;
     }
 
-    memory.WriteBlockUnsafe(send_info_, &send_info, sizeof(AuxInfo::AuxInfoDsp));
+    if (!host_safe) [[unlikely]] {
+        memory.WriteBlockUnsafe(send_info_, send_ptr, sizeof(AuxInfo::AuxInfoDsp));
+    }
 
     return write_count_;
 }
@@ -102,7 +135,7 @@ 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.
@@ -112,16 +145,21 @@ 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_,
-                            const CpuAddr return_buffer, const u32 count_max, std::span<s32> output,
-                            const u32 count_, const u32 read_offset, const u32 update_count) {
+static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, CpuAddr return_info_,
+                            CpuAddr return_buffer, u32 count_max, std::span<s32> output,
+                            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;
     }
 
@@ -136,35 +174,49 @@ static u32 ReadAuxBufferDsp(Core::Memory::Memory& memory, const CpuAddr return_i
     }
 
     AuxInfo::AuxInfoDsp return_info{};
-    memory.ReadBlockUnsafe(return_info_, &return_info, sizeof(AuxInfo::AuxInfoDsp));
+    auto return_ptr = &return_info;
+    bool host_safe = (return_info_ & Core::Memory::YUZU_PAGEMASK) <=
+                     (Core::Memory::YUZU_PAGESIZE - sizeof(AuxInfo::AuxInfoDsp));
 
-    u32 target_read_offset{return_info.read_offset + read_offset};
+    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};
     if (target_read_offset > count_max) {
         return 0;
     }
 
-    u32 read_count{count_};
-    u32 read_pos{0};
+    u32 read_count{read_count_};
+    u32 write_pos{0};
     while (read_count > 0) {
         u32 to_read{std::min(count_max - target_read_offset, read_count)};
-
-        if (to_read > 0) {
+        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[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_ptr->read_offset = (return_ptr->read_offset + update_count) % count_max;
     }
 
-    memory.WriteBlockUnsafe(return_info_, &return_info, sizeof(AuxInfo::AuxInfoDsp));
+    if (!host_safe) [[unlikely]] {
+        memory.WriteBlockUnsafe(return_info_, return_ptr, sizeof(AuxInfo::AuxInfoDsp));
+    }
 
-    return count_;
+    return read_count_;
 }
 
 void AuxCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor& processor,
@@ -189,7 +241,7 @@ 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/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -292,7 +292,7 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
 
     Optimization::PositionPass(env, program);
 
-    Optimization::GlobalMemoryToStorageBufferPass(program, host_info);
+    Optimization::GlobalMemoryToStorageBufferPass(program);
     Optimization::TexturePass(env, program, host_info);
 
     if (Settings::values.resolution_info.active) {

src/shader_recompiler/host_translate_info.h

@@ -15,7 +15,6 @@ struct HostTranslateInfo {
     bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
     bool support_snorm_render_buffer{};  ///< True when the device supports SNORM render buffers
     bool support_viewport_index_layer{}; ///< True when the device supports gl_Layer in VS
-    u32 min_ssbo_alignment{};            ///< Minimum alignment supported by the device for SSBOs
     bool support_geometry_shader_passthrough{}; ///< True when the device supports geometry
                                                 ///< passthrough shaders
 };

src/shader_recompiler/ir_opt/global_memory_to_storage_buffer_pass.cpp

@@ -11,7 +11,6 @@
 #include "shader_recompiler/frontend/ir/breadth_first_search.h"
 #include "shader_recompiler/frontend/ir/ir_emitter.h"
 #include "shader_recompiler/frontend/ir/value.h"
-#include "shader_recompiler/host_translate_info.h"
 #include "shader_recompiler/ir_opt/passes.h"
 
 namespace Shader::Optimization {
@@ -403,7 +402,7 @@ void CollectStorageBuffers(IR::Block& block, IR::Inst& inst, StorageInfo& info)
 }
 
 /// Returns the offset in indices (not bytes) for an equivalent storage instruction
-IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer, u32 alignment) {
+IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer) {
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::U32 offset;
     if (const std::optional<LowAddrInfo> low_addr{TrackLowAddress(&inst)}) {
@@ -416,10 +415,7 @@ IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer
     }
     // Subtract the least significant 32 bits from the guest offset. The result is the storage
     // buffer offset in bytes.
-    IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
-
-    // Align the offset base to match the host alignment requirements
-    low_cbuf = ir.BitwiseAnd(low_cbuf, ir.Imm32(~(alignment - 1U)));
+    const IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
     return ir.ISub(offset, low_cbuf);
 }
 
@@ -514,7 +510,7 @@ void Replace(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
 }
 } // Anonymous namespace
 
-void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info) {
+void GlobalMemoryToStorageBufferPass(IR::Program& program) {
     StorageInfo info;
     for (IR::Block* const block : program.post_order_blocks) {
         for (IR::Inst& inst : block->Instructions()) {
@@ -538,8 +534,7 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateIn
         const IR::U32 index{IR::Value{static_cast<u32>(info.set.index_of(it))}};
         IR::Block* const block{storage_inst.block};
         IR::Inst* const inst{storage_inst.inst};
-        const IR::U32 offset{
-            StorageOffset(*block, *inst, storage_buffer, host_info.min_ssbo_alignment)};
+        const IR::U32 offset{StorageOffset(*block, *inst, storage_buffer)};
         Replace(*block, *inst, index, offset);
     }
 }

src/shader_recompiler/ir_opt/passes.h

@@ -15,7 +15,7 @@ namespace Shader::Optimization {
 void CollectShaderInfoPass(Environment& env, IR::Program& program);
 void ConstantPropagationPass(Environment& env, IR::Program& program);
 void DeadCodeEliminationPass(IR::Program& program);
-void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info);
+void GlobalMemoryToStorageBufferPass(IR::Program& program);
 void IdentityRemovalPass(IR::Program& program);
 void LowerFp16ToFp32(IR::Program& program);
 void LowerInt64ToInt32(IR::Program& program);

src/video_core/buffer_cache/buffer_cache.h

@@ -1938,21 +1938,14 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
                                                                       bool is_written) const {
     const GPUVAddr gpu_addr = gpu_memory->Read<u64>(ssbo_addr);
     const u32 size = gpu_memory->Read<u32>(ssbo_addr + 8);
-    const u32 alignment = runtime.GetStorageBufferAlignment();
-
-    const GPUVAddr aligned_gpu_addr = Common::AlignDown(gpu_addr, alignment);
-    const u32 aligned_size =
-        Common::AlignUp(static_cast<u32>(gpu_addr - aligned_gpu_addr) + size, alignment);
-
-    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(aligned_gpu_addr);
+    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
     if (!cpu_addr || size == 0) {
         return NULL_BINDING;
     }
-
-    const VAddr cpu_end = Common::AlignUp(*cpu_addr + aligned_size, Core::Memory::YUZU_PAGESIZE);
+    const VAddr cpu_end = Common::AlignUp(*cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
     const Binding binding{
         .cpu_addr = *cpu_addr,
-        .size = is_written ? aligned_size : static_cast<u32>(cpu_end - *cpu_addr),
+        .size = is_written ? size : static_cast<u32>(cpu_end - *cpu_addr),
         .buffer_id = BufferId{},
     };
     return binding;

src/video_core/gpu_thread.cpp

@@ -32,7 +32,8 @@ static void RunThread(std::stop_token stop_token, Core::System& system,
     VideoCore::RasterizerInterface* const rasterizer = renderer.ReadRasterizer();
 
     while (!stop_token.stop_requested()) {
-        CommandDataContainer next = state.queue.PopWait(stop_token);
+        CommandDataContainer next;
+        state.queue.Pop(next, stop_token);
         if (stop_token.stop_requested()) {
             break;
         }

src/video_core/gpu_thread.h

@@ -10,8 +10,8 @@
 #include <thread>
 #include <variant>
 
+#include "common/bounded_threadsafe_queue.h"
 #include "common/polyfill_thread.h"
-#include "common/threadsafe_queue.h"
 #include "video_core/framebuffer_config.h"
 
 namespace Tegra {
@@ -97,7 +97,7 @@ struct CommandDataContainer {
 
 /// Struct used to synchronize the GPU thread
 struct SynchState final {
-    using CommandQueue = Common::MPSCQueue<CommandDataContainer, true>;
+    using CommandQueue = Common::MPSCQueue<CommandDataContainer>;
     std::mutex write_lock;
     CommandQueue queue;
     u64 last_fence{};

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -160,10 +160,6 @@ public:
         return device.CanReportMemoryUsage();
     }
 
-    u32 GetStorageBufferAlignment() const {
-        return static_cast<u32>(device.GetShaderStorageBufferAlignment());
-    }
-
 private:
     static constexpr std::array PABO_LUT{
         GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV,          GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -236,7 +236,6 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .needs_demote_reorder = device.IsAmd(),
           .support_snorm_render_buffer = false,
           .support_viewport_index_layer = device.HasVertexViewportLayer(),
-          .min_ssbo_alignment = static_cast<u32>(device.GetShaderStorageBufferAlignment()),
           .support_geometry_shader_passthrough = device.HasGeometryShaderPassthrough(),
       } {
     if (use_asynchronous_shaders) {

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -330,10 +330,6 @@ bool BufferCacheRuntime::CanReportMemoryUsage() const {
     return device.CanReportMemoryUsage();
 }
 
-u32 BufferCacheRuntime::GetStorageBufferAlignment() const {
-    return static_cast<u32>(device.GetStorageBufferAlignment());
-}
-
 void BufferCacheRuntime::Finish() {
     scheduler.Finish();
 }

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -73,8 +73,6 @@ public:
 
     bool CanReportMemoryUsage() const;
 
-    u32 GetStorageBufferAlignment() const;
-
     [[nodiscard]] StagingBufferRef UploadStagingBuffer(size_t size);
 
     [[nodiscard]] StagingBufferRef DownloadStagingBuffer(size_t size);

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -344,7 +344,6 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
             driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE,
         .support_snorm_render_buffer = true,
         .support_viewport_index_layer = device.IsExtShaderViewportIndexLayerSupported(),
-        .min_ssbo_alignment = static_cast<u32>(device.GetStorageBufferAlignment()),
         .support_geometry_shader_passthrough = device.IsNvGeometryShaderPassthroughSupported(),
     };