Merge branch 'shadps4-emu:main' into main

.github/workflows/build.yml

@@ -14,14 +14,14 @@ env:
 
 jobs:
   reuse:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-24.04
     continue-on-error: true
     steps:
     - uses: actions/checkout@v4
     - uses: fsfe/reuse-action@v5
 
   clang-format:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-24.04
     continue-on-error: true
     steps:
     - uses: actions/checkout@v4
@@ -39,7 +39,7 @@ jobs:
       run: ./.ci/clang-format.sh
       
   get-info:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-24.04
     outputs:
       date: ${{ steps.vars.outputs.date }}
       shorthash: ${{ steps.vars.outputs.shorthash }}
@@ -57,7 +57,7 @@ jobs:
         echo "fullhash=$(git rev-parse HEAD)" >> $GITHUB_OUTPUT
 
   windows-sdl:
-    runs-on: windows-latest
+    runs-on: windows-2025
     needs: get-info
     steps:
     - uses: actions/checkout@v4
@@ -101,7 +101,7 @@ jobs:
         path: ${{github.workspace}}/build/shadPS4.exe
 
   windows-qt:
-    runs-on: windows-latest
+    runs-on: windows-2025
     needs: get-info
     steps:
     - uses: actions/checkout@v4

src/video_core/buffer_cache/buffer.h

@@ -119,19 +119,23 @@ public:
         return buffer;
     }
 
-    std::optional<vk::BufferMemoryBarrier2> GetBarrier(vk::AccessFlagBits2 dst_acess_mask,
-                                                       vk::PipelineStageFlagBits2 dst_stage) {
+    std::optional<vk::BufferMemoryBarrier2> GetBarrier(
+        vk::Flags<vk::AccessFlagBits2> dst_acess_mask, vk::PipelineStageFlagBits2 dst_stage,
+        u32 offset = 0) {
         if (dst_acess_mask == access_mask && stage == dst_stage) {
             return {};
         }
 
+        DEBUG_ASSERT(offset < size_bytes);
+
         auto barrier = vk::BufferMemoryBarrier2{
             .srcStageMask = stage,
             .srcAccessMask = access_mask,
             .dstStageMask = dst_stage,
             .dstAccessMask = dst_acess_mask,
             .buffer = buffer.buffer,
-            .size = size_bytes,
+            .offset = offset,
+            .size = size_bytes - offset,
         };
         access_mask = dst_acess_mask;
         stage = dst_stage;
@@ -150,8 +154,10 @@ public:
     Vulkan::Scheduler* scheduler;
     MemoryUsage usage;
     UniqueBuffer buffer;
-    vk::AccessFlagBits2 access_mask{vk::AccessFlagBits2::eNone};
-    vk::PipelineStageFlagBits2 stage{vk::PipelineStageFlagBits2::eNone};
+    vk::Flags<vk::AccessFlagBits2> access_mask{
+        vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite |
+        vk::AccessFlagBits2::eTransferRead | vk::AccessFlagBits2::eTransferWrite};
+    vk::PipelineStageFlagBits2 stage{vk::PipelineStageFlagBits2::eAllCommands};
 };
 
 class StreamBuffer : public Buffer {

src/video_core/buffer_cache/buffer_cache.cpp

@@ -34,21 +34,10 @@ BufferCache::BufferCache(const Vulkan::Instance& instance_, Vulkan::Scheduler& s
 
     // Ensure the first slot is used for the null buffer
     const auto null_id =
-        slot_buffers.insert(instance, scheduler, MemoryUsage::DeviceLocal, 0, ReadFlags, 1);
+        slot_buffers.insert(instance, scheduler, MemoryUsage::DeviceLocal, 0, ReadFlags, 16);
     ASSERT(null_id.index == 0);
     const vk::Buffer& null_buffer = slot_buffers[null_id].buffer;
     Vulkan::SetObjectName(instance.GetDevice(), null_buffer, "Null Buffer");
-
-    const vk::BufferViewCreateInfo null_view_ci = {
-        .buffer = null_buffer,
-        .format = vk::Format::eR8Unorm,
-        .offset = 0,
-        .range = VK_WHOLE_SIZE,
-    };
-    const auto [null_view_result, null_view] = instance.GetDevice().createBufferView(null_view_ci);
-    ASSERT_MSG(null_view_result == vk::Result::eSuccess, "Failed to create null buffer view.");
-    null_buffer_view = null_view;
-    Vulkan::SetObjectName(instance.GetDevice(), null_buffer_view, "Null Buffer View");
 }
 
 BufferCache::~BufferCache() = default;
@@ -479,43 +468,36 @@ void BufferCache::JoinOverlap(BufferId new_buffer_id, BufferId overlap_id,
     };
     scheduler.EndRendering();
     const auto cmdbuf = scheduler.CommandBuffer();
-    const std::array pre_barriers = {
-        vk::BufferMemoryBarrier2{
-            .srcStageMask = vk::PipelineStageFlagBits2::eAllCommands,
-            .srcAccessMask = vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite,
-            .dstStageMask = vk::PipelineStageFlagBits2::eTransfer,
-            .dstAccessMask = vk::AccessFlagBits2::eTransferRead,
-            .buffer = overlap.Handle(),
-            .offset = 0,
-            .size = overlap.SizeBytes(),
-        },
-    };
-    const std::array post_barriers = {
-        vk::BufferMemoryBarrier2{
-            .srcStageMask = vk::PipelineStageFlagBits2::eTransfer,
-            .srcAccessMask = vk::AccessFlagBits2::eTransferRead,
-            .dstStageMask = vk::PipelineStageFlagBits2::eAllCommands,
-            .dstAccessMask = vk::AccessFlagBits2::eMemoryWrite,
-            .buffer = overlap.Handle(),
-            .offset = 0,
-            .size = overlap.SizeBytes(),
-        },
-        vk::BufferMemoryBarrier2{
-            .srcStageMask = vk::PipelineStageFlagBits2::eTransfer,
-            .srcAccessMask = vk::AccessFlagBits2::eTransferWrite,
-            .dstStageMask = vk::PipelineStageFlagBits2::eAllCommands,
-            .dstAccessMask = vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite,
-            .buffer = new_buffer.Handle(),
-            .offset = dst_base_offset,
-            .size = overlap.SizeBytes(),
-        },
-    };
+
+    boost::container::static_vector<vk::BufferMemoryBarrier2, 2> pre_barriers{};
+    if (auto src_barrier = overlap.GetBarrier(vk::AccessFlagBits2::eTransferRead,
+                                              vk::PipelineStageFlagBits2::eTransfer)) {
+        pre_barriers.push_back(*src_barrier);
+    }
+    if (auto dst_barrier =
+            new_buffer.GetBarrier(vk::AccessFlagBits2::eTransferWrite,
+                                  vk::PipelineStageFlagBits2::eTransfer, dst_base_offset)) {
+        pre_barriers.push_back(*dst_barrier);
+    }
     cmdbuf.pipelineBarrier2(vk::DependencyInfo{
         .dependencyFlags = vk::DependencyFlagBits::eByRegion,
-        .bufferMemoryBarrierCount = 1,
+        .bufferMemoryBarrierCount = static_cast<u32>(pre_barriers.size()),
         .pBufferMemoryBarriers = pre_barriers.data(),
     });
+
     cmdbuf.copyBuffer(overlap.Handle(), new_buffer.Handle(), copy);
+
+    boost::container::static_vector<vk::BufferMemoryBarrier2, 2> post_barriers{};
+    if (auto src_barrier =
+            overlap.GetBarrier(vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite,
+                               vk::PipelineStageFlagBits2::eAllCommands)) {
+        post_barriers.push_back(*src_barrier);
+    }
+    if (auto dst_barrier = new_buffer.GetBarrier(
+            vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite,
+            vk::PipelineStageFlagBits2::eAllCommands, dst_base_offset)) {
+        post_barriers.push_back(*dst_barrier);
+    }
     cmdbuf.pipelineBarrier2(vk::DependencyInfo{
         .dependencyFlags = vk::DependencyFlagBits::eByRegion,
         .bufferMemoryBarrierCount = static_cast<u32>(post_barriers.size()),
@@ -626,7 +608,8 @@ void BufferCache::SynchronizeBuffer(Buffer& buffer, VAddr device_addr, u32 size,
     const auto cmdbuf = scheduler.CommandBuffer();
     const vk::BufferMemoryBarrier2 pre_barrier = {
         .srcStageMask = vk::PipelineStageFlagBits2::eAllCommands,
-        .srcAccessMask = vk::AccessFlagBits2::eMemoryRead,
+        .srcAccessMask = vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite |
+                         vk::AccessFlagBits2::eTransferRead | vk::AccessFlagBits2::eTransferWrite,
         .dstStageMask = vk::PipelineStageFlagBits2::eTransfer,
         .dstAccessMask = vk::AccessFlagBits2::eTransferWrite,
         .buffer = buffer.Handle(),

src/video_core/buffer_cache/buffer_cache.h

@@ -71,10 +71,6 @@ public:
         return slot_buffers[id];
     }
 
-    [[nodiscard]] vk::BufferView& NullBufferView() {
-        return null_buffer_view;
-    }
-
     /// Invalidates any buffer in the logical page range.
     void InvalidateMemory(VAddr device_addr, u64 size);
 
@@ -160,7 +156,6 @@ private:
     std::shared_mutex mutex;
     Common::SlotVector<Buffer> slot_buffers;
     RangeSet gpu_modified_ranges;
-    vk::BufferView null_buffer_view;
     MemoryTracker memory_tracker;
     PageTable page_table;
 };

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -537,6 +537,7 @@ void Rasterizer::BindBuffers(const Shader::Info& stage, Shader::Backend::Binding
     }
 
     // Second pass to re-bind buffers that were updated after binding
+    auto& null_buffer = buffer_cache.GetBuffer(VideoCore::NULL_BUFFER_ID);
     for (u32 i = 0; i < buffer_bindings.size(); i++) {
         const auto& [buffer_id, vsharp] = buffer_bindings[i];
         const auto& desc = stage.buffers[i];
@@ -548,7 +549,6 @@ void Rasterizer::BindBuffers(const Shader::Info& stage, Shader::Backend::Binding
             } else if (instance.IsNullDescriptorSupported()) {
                 buffer_infos.emplace_back(VK_NULL_HANDLE, 0, VK_WHOLE_SIZE);
             } else {
-                auto& null_buffer = buffer_cache.GetBuffer(VideoCore::NULL_BUFFER_ID);
                 buffer_infos.emplace_back(null_buffer.Handle(), 0, VK_WHOLE_SIZE);
             }
         } else {
@@ -582,17 +582,19 @@ void Rasterizer::BindBuffers(const Shader::Info& stage, Shader::Backend::Binding
         ++binding.buffer;
     }
 
-    const auto null_buffer_view =
-        instance.IsNullDescriptorSupported() ? VK_NULL_HANDLE : buffer_cache.NullBufferView();
     for (u32 i = 0; i < texbuffer_bindings.size(); i++) {
         const auto& [buffer_id, vsharp] = texbuffer_bindings[i];
         const auto& desc = stage.texture_buffers[i];
-        vk::BufferView& buffer_view = buffer_views.emplace_back(null_buffer_view);
+        // Fallback format for null buffer view; never used in valid buffer case.
+        const auto data_fmt = vsharp.GetDataFmt() != AmdGpu::DataFormat::FormatInvalid
+                                  ? vsharp.GetDataFmt()
+                                  : AmdGpu::DataFormat::Format8;
+        const u32 fmt_stride = AmdGpu::NumBits(data_fmt) >> 3;
+        vk::BufferView buffer_view;
         if (buffer_id) {
             const u32 alignment = instance.TexelBufferMinAlignment();
             const auto [vk_buffer, offset] = buffer_cache.ObtainBuffer(
                 vsharp.base_address, vsharp.GetSize(), desc.is_written, true, buffer_id);
-            const u32 fmt_stride = AmdGpu::NumBits(vsharp.GetDataFmt()) >> 3;
             const u32 buf_stride = vsharp.GetStride();
             ASSERT_MSG(buf_stride % fmt_stride == 0,
                        "Texel buffer stride must match format stride");
@@ -600,9 +602,8 @@ void Rasterizer::BindBuffers(const Shader::Info& stage, Shader::Backend::Binding
             const u32 adjust = offset - offset_aligned;
             ASSERT(adjust % fmt_stride == 0);
             push_data.AddTexelOffset(binding.buffer, buf_stride / fmt_stride, adjust / fmt_stride);
-            buffer_view =
-                vk_buffer->View(offset_aligned, vsharp.GetSize() + adjust, desc.is_written,
-                                vsharp.GetDataFmt(), vsharp.GetNumberFmt());
+            buffer_view = vk_buffer->View(offset_aligned, vsharp.GetSize() + adjust,
+                                          desc.is_written, data_fmt, vsharp.GetNumberFmt());
             if (auto barrier =
                     vk_buffer->GetBarrier(desc.is_written ? vk::AccessFlagBits2::eShaderWrite
                                                           : vk::AccessFlagBits2::eShaderRead,
@@ -612,6 +613,11 @@ void Rasterizer::BindBuffers(const Shader::Info& stage, Shader::Backend::Binding
             if (desc.is_written) {
                 texture_cache.InvalidateMemoryFromGPU(vsharp.base_address, vsharp.GetSize());
             }
+        } else if (instance.IsNullDescriptorSupported()) {
+            buffer_view = VK_NULL_HANDLE;
+        } else {
+            buffer_view =
+                null_buffer.View(0, fmt_stride, desc.is_written, data_fmt, vsharp.GetNumberFmt());
         }
 
         set_writes.push_back({
@@ -621,7 +627,7 @@ void Rasterizer::BindBuffers(const Shader::Info& stage, Shader::Backend::Binding
             .descriptorCount = 1,
             .descriptorType = desc.is_written ? vk::DescriptorType::eStorageTexelBuffer
                                               : vk::DescriptorType::eUniformTexelBuffer,
-            .pTexelBufferView = &buffer_view,
+            .pTexelBufferView = &buffer_views.emplace_back(buffer_view),
         });
         ++binding.buffer;
     }