[vulkan] Adjusted VK_EXT_vertex_input_dynamic_state

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -10,6 +10,7 @@
 #include <span>
 #include <vector>
 
+#include "common/settings.h"
 #include "video_core/buffer_cache/buffer_cache_base.h"
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
 
@@ -334,6 +335,9 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m
       staging_pool{staging_pool_}, guest_descriptor_queue{guest_descriptor_queue_},
       quad_index_pass(device, scheduler, descriptor_pool, staging_pool,
                       compute_pass_descriptor_queue) {
+    use_vertex_input_dynamic_state = device.IsExtVertexInputDynamicStateSupported() &&
+                                     Settings::values.vertex_input_dynamic_state.GetValue();
+
     const VkDriverIdKHR driver_id = device.GetDriverID();
     limit_dynamic_storage_buffers = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||
                                     driver_id == VK_DRIVER_ID_ARM_PROPRIETARY;
@@ -556,7 +560,10 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset
     if (index >= device.GetMaxVertexInputBindings()) {
         return;
     }
-    if (device.IsExtExtendedDynamicStateSupported()) {
+    const bool use_bind_vertex_buffers2_stride =
+        device.IsExtExtendedDynamicStateSupported() && !use_vertex_input_dynamic_state;
+
+    if (use_bind_vertex_buffers2_stride) {
         scheduler.Record([index, buffer, offset, size, stride](vk::CommandBuffer cmdbuf) {
             const VkDeviceSize vk_offset = buffer != VK_NULL_HANDLE ? offset : 0;
             const VkDeviceSize vk_size = buffer != VK_NULL_HANDLE ? size : VK_WHOLE_SIZE;
@@ -596,7 +603,10 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
     if (binding_count == 0) {
         return;
     }
-    if (device.IsExtExtendedDynamicStateSupported()) {
+    const bool use_bind_vertex_buffers2_stride =
+        device.IsExtExtendedDynamicStateSupported() && !use_vertex_input_dynamic_state;
+
+    if (use_bind_vertex_buffers2_stride) {
         scheduler.Record([bindings_ = std::move(bindings), buffer_handles_ = std::move(buffer_handles), binding_count](vk::CommandBuffer cmdbuf) {
             cmdbuf.BindVertexBuffers2EXT(bindings_.min_index, binding_count, buffer_handles_.data(), bindings_.offsets.data(), bindings_.sizes.data(), bindings_.strides.data());
         });

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -183,6 +183,7 @@ private:
     std::unique_ptr<Uint8Pass> uint8_pass;
     QuadIndexedPass quad_index_pass;
 
+    bool use_vertex_input_dynamic_state = false;
     bool limit_dynamic_storage_buffers = false;
     u32 max_dynamic_storage_buffers = (std::numeric_limits<u32>::max)();
 };

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -1732,39 +1732,54 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
     }
     dirty[Dirty::VertexInput] = false;
 
+    const size_t max_vertex_attributes =
+        (std::min)(Maxwell::NumVertexAttributes,
+                   static_cast<size_t>(device.GetMaxVertexInputAttributes()));
+    const size_t max_vertex_bindings =
+        (std::min)(Maxwell::NumVertexArrays,
+                   static_cast<size_t>(device.GetMaxVertexInputBindings()));
+
+    // Dynamic state path must respect device limits, same as static pipeline creation.
+    for (size_t index = max_vertex_attributes; index < Maxwell::NumVertexAttributes; ++index) {
+        dirty[Dirty::VertexAttribute0 + index] = false;
+    }
+    for (size_t index = max_vertex_bindings; index < Maxwell::NumVertexArrays; ++index) {
+        dirty[Dirty::VertexBinding0 + index] = false;
+    }
+
     boost::container::static_vector<VkVertexInputBindingDescription2EXT, 32> bindings;
     boost::container::static_vector<VkVertexInputAttributeDescription2EXT, 32> attributes;
 
-    // There seems to be a bug on Nvidia's driver where updating only higher attributes ends up
+    std::array<bool, Maxwell::NumVertexArrays> used_bindings{};
-    // generating dirty state. Track the highest dirty attribute and update all attributes until
+    for (size_t index = 0; index < max_vertex_attributes; ++index) {
-    // that one.
-    size_t highest_dirty_attr{};
-    for (size_t index = 0; index < Maxwell::NumVertexAttributes; ++index) {
-        if (dirty[Dirty::VertexAttribute0 + index]) {
-            highest_dirty_attr = index;
-        }
-    }
-    for (size_t index = 0; index < highest_dirty_attr; ++index) {
-        const Maxwell::VertexAttribute attribute{regs.vertex_attrib_format[index]};
-        const u32 binding{attribute.buffer};
         dirty[Dirty::VertexAttribute0 + index] = false;
-        dirty[Dirty::VertexBinding0 + static_cast<size_t>(binding)] = true;
+
-        if (!attribute.constant) {
+        const Maxwell::VertexAttribute attribute{regs.vertex_attrib_format[index]};
-            attributes.push_back({
+        if (attribute.constant) {
-                .sType = VK_STRUCTURE_TYPE_VERTEX_INPUT_ATTRIBUTE_DESCRIPTION_2_EXT,
-                .pNext = nullptr,
-                .location = static_cast<u32>(index),
-                .binding = binding,
-                .format = MaxwellToVK::VertexFormat(device, attribute.type, attribute.size),
-                .offset = attribute.offset,
-            });
-        }
-    }
-    for (size_t index = 0; index < Maxwell::NumVertexAttributes; ++index) {
-        if (!dirty[Dirty::VertexBinding0 + index]) {
             continue;
         }
+
+        const size_t binding{attribute.buffer};
+        if (binding >= max_vertex_bindings) {
+            continue;
+        }
+
+        used_bindings[binding] = true;
+        attributes.push_back({
+            .sType = VK_STRUCTURE_TYPE_VERTEX_INPUT_ATTRIBUTE_DESCRIPTION_2_EXT,
+            .pNext = nullptr,
+            .location = static_cast<u32>(index),
+            .binding = static_cast<u32>(binding),
+            .format = MaxwellToVK::VertexFormat(device, attribute.type, attribute.size),
+            .offset = attribute.offset,
+        });
+    }
+
+    for (size_t index = 0; index < max_vertex_bindings; ++index) {
         dirty[Dirty::VertexBinding0 + index] = false;
+        if (!used_bindings[index]) {
+            continue;
+        }
 
         const u32 binding{static_cast<u32>(index)};
         const auto& input_binding{regs.vertex_streams[binding]};
@@ -1778,6 +1793,10 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
             .divisor = is_instanced ? input_binding.frequency : 1,
         });
     }
+
+    ASSERT(attributes.size() <= max_vertex_attributes);
+    ASSERT(bindings.size() <= max_vertex_bindings);
+
     scheduler.Record([bindings, attributes](vk::CommandBuffer cmdbuf) {
         cmdbuf.SetVertexInputEXT(bindings, attributes);
     });

src/video_core/vulkan_common/vulkan_device.cpp

@@ -1112,34 +1112,11 @@ bool Device::GetSuitability(bool requires_swapchain) {
     // VK_DYNAMIC_STATE
 
     // Driver detection variables for workarounds in GetSuitability
-    const VkDriverId driver_id = properties.driver.driverID;
 
     // VK_EXT_extended_dynamic_state2 below this will appear drivers that need workarounds.
 
     // VK_EXT_extended_dynamic_state3 below this will appear drivers that need workarounds.
 
-    // Samsung: Broken extendedDynamicState3ColorBlendEquation
-    // Disable blend equation dynamic state, force static pipeline state
-    if (extensions.extended_dynamic_state3 &&
-        (driver_id == VK_DRIVER_ID_SAMSUNG_PROPRIETARY)) {
-        LOG_WARNING(Render_Vulkan,
-                    "Samsung: Disabling broken extendedDynamicState3ColorBlendEquation");
-        features.extended_dynamic_state3.extendedDynamicState3ColorBlendEnable = false;
-        features.extended_dynamic_state3.extendedDynamicState3ColorBlendEquation = false;
-    }
-
-    // Intel Windows < 27.20.100.0: Broken VertexInputDynamicState
-    // Same for NVIDIA Proprietary < 580.119.02, unknown when VIDS was first NOT broken
-    // Disable VertexInputDynamicState on old Intel Windows drivers
-    if (extensions.vertex_input_dynamic_state) {
-        const u32 version = (properties.properties.driverVersion << 3) >> 3;
-        if ((driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS && version < VK_MAKE_API_VERSION(27, 20, 100, 0))
-        || (driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY && version < VK_MAKE_API_VERSION(580, 119, 02, 0))) {
-            LOG_WARNING(Render_Vulkan, "Disabling broken VK_EXT_vertex_input_dynamic_state");
-            RemoveExtensionFeature(extensions.vertex_input_dynamic_state, features.vertex_input_dynamic_state, VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME);
-        }
-    }
-
     if (u32(Settings::values.dyna_state.GetValue()) == 0) {
         LOG_INFO(Render_Vulkan, "Extended Dynamic State disabled by user setting, clearing all EDS features");
         features.extended_dynamic_state.extendedDynamicState = false;