Vertex buffers.

CMakeLists.txt

@@ -13,7 +13,7 @@ if (MSVC)
     add_compile_definitions(_HAS_EXCEPTIONS=1)
     add_compile_definitions(_CRT_SECURE_NO_WARNINGS)
 else ()
-    set(CMAKE_CXX_FLAGS "-Wall -fno-rtti -fno-exceptions")
+    set(CMAKE_CXX_FLAGS "-Wall -g -fno-rtti -fno-exceptions")
 endif ()
 
 include(add_shader.cmake)

aster/buffer.cpp

@@ -5,6 +5,8 @@
 
 #include "buffer.h"
 
+#include <ranges>
+
 void
 Buffer::Destroy(const Device *device)
 {
@@ -16,28 +18,91 @@ Buffer::Destroy(const Device *device)
 }
 
 void
-UniformBuffer::Init(const Device *device, usize size, cstr name)
+Buffer::Allocate(const Device *device, usize size, vk::BufferUsageFlags bufferUsage,
+                 VmaAllocationCreateFlags allocationFlags, VmaMemoryUsage memoryUsage, cstr name)
 {
+    assert(size <= SIZE_MASK);
+
     vk::BufferCreateInfo bufferCreateInfo = {
         .size = size,
-        .usage = vk::BufferUsageFlagBits::eUniformBuffer,
+        .usage = bufferUsage,
         .sharingMode = vk::SharingMode::eExclusive,
     };
-    constexpr VmaAllocationCreateInfo allocationCreateInfo = {
-        .flags = VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT |
-                 VMA_ALLOCATION_CREATE_HOST_ACCESS_ALLOW_TRANSFER_INSTEAD_BIT | VMA_ALLOCATION_CREATE_MAPPED_BIT,
-        .usage = VMA_MEMORY_USAGE_AUTO,
+    const VmaAllocationCreateInfo allocationCreateInfo = {
+        .flags = allocationFlags,
+        .usage = memoryUsage,
     };
 
     VkBuffer buffer;
     VmaAllocation allocation;
+    VmaAllocationInfo allocationInfo;
     auto result = Cast<vk::Result>(vmaCreateBuffer(device->m_Allocator, Recast<VkBufferCreateInfo *>(&bufferCreateInfo),
-                                                   &allocationCreateInfo, &buffer, &allocation, nullptr));
+                                                   &allocationCreateInfo, &buffer, &allocation, &allocationInfo));
     ERROR_IF(Failed(result), "Could not allocate buffer. Cause: {}", result) THEN_ABORT(result);
 
     m_Buffer = buffer;
-    m_Size = size;
+    m_Size = size & SIZE_MASK;
     m_Allocation = allocation;
+    m_Mapped = allocationInfo.pMappedData;
+
+    vk::MemoryPropertyFlags memoryPropertyFlags;
+    vmaGetAllocationMemoryProperties(device->m_Allocator, allocation,
+                                     Recast<VkMemoryPropertyFlags *>(&memoryPropertyFlags));
+    if (memoryPropertyFlags & vk::MemoryPropertyFlagBits::eHostVisible)
+    {
+        m_Size |= HOST_ACCESSIBLE_BIT;
+    }
 
     device->SetName(m_Buffer, name);
 }
+
+void
+Buffer::Write(const Device *device, usize offset, usize size, void *data)
+{
+    assert(IsHostVisible());
+
+    if (!IsMapped())
+    {
+        auto result = Cast<vk::Result>(vmaMapMemory(device->m_Allocator, m_Allocation, &m_Mapped));
+        ERROR_IF(Failed(result), "Memory mapping failed. Cause: {}", result);
+        if (!Failed(result))
+        {
+            memcpy(m_Mapped, data, size);
+
+            vmaUnmapMemory(device->m_Allocator, m_Allocation);
+            m_Mapped = nullptr;
+        }
+    }
+    else
+    {
+        memcpy(m_Mapped, data, size);
+    }
+
+    // TODO: Debug this.
+    // auto result = Cast<vk::Result>(vmaCopyMemoryToAllocation(device->m_Allocator, &data, m_Allocation, 0, size));
+    // ERROR_IF(Failed(result), "Writing to buffer failed. Cause: {}", result) THEN_ABORT(result);
+}
+
+void
+UniformBuffer::Init(const Device *device, const usize size, const cstr name)
+{
+    Allocate(device, size, vk::BufferUsageFlagBits::eUniformBuffer,
+             VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT |
+                 VMA_ALLOCATION_CREATE_HOST_ACCESS_ALLOW_TRANSFER_INSTEAD_BIT | VMA_ALLOCATION_CREATE_MAPPED_BIT,
+             VMA_MEMORY_USAGE_AUTO, name);
+}
+
+void
+VertexBuffer::Init(const Device *device, usize size, cstr name)
+{
+    Allocate(device, size, vk::BufferUsageFlagBits::eVertexBuffer | vk::BufferUsageFlagBits::eTransferDst,
+             VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT, VMA_MEMORY_USAGE_AUTO, name);
+}
+
+void
+StagingBuffer::Init(const Device *device, usize size, cstr name)
+{
+    Allocate(device, size, vk::BufferUsageFlagBits::eTransferSrc,
+             VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT | VMA_ALLOCATION_CREATE_MAPPED_BIT,
+             VMA_MEMORY_USAGE_AUTO, name);
+}

aster/buffer.h

@@ -14,12 +14,55 @@ struct Buffer
 {
     vk::Buffer m_Buffer = nullptr;
     VmaAllocation m_Allocation = nullptr;
-    usize m_Size = 0;
+    void *m_Mapped = nullptr;
+
+    [[nodiscard]] usize GetSize() const;
+    [[nodiscard]] bool IsHostVisible() const;
+    [[nodiscard]] bool IsMapped() const;
 
     void Destroy(const Device *device);
+    void Write(const Device *device, usize offset, usize size, void *data);
+
+  protected:
+    void Allocate(const Device *device, usize size, vk::BufferUsageFlags bufferUsage,
+                  VmaAllocationCreateFlags allocationFlags, VmaMemoryUsage memoryUsage, cstr name);
+
+    usize m_Size = 0;
+
+    constexpr static usize HOST_ACCESSIBLE_BIT = 1llu << 63;
+    constexpr static usize SIZE_MASK = ~HOST_ACCESSIBLE_BIT;
 };
 
 struct UniformBuffer : Buffer
 {
     void Init(const Device *device, usize size, cstr name = nullptr);
 };
+
+struct VertexBuffer : Buffer
+{
+    void Init(const Device *device, usize size, cstr name = nullptr);
+    void Write(const Device *device, void *data, usize size, usize offset) const = delete;
+};
+
+struct StagingBuffer : Buffer
+{
+    void Init(const Device *device, usize size, cstr name = nullptr);
+};
+
+inline usize
+Buffer::GetSize() const
+{
+    return m_Size & SIZE_MASK;
+}
+
+inline bool
+Buffer::IsHostVisible() const
+{
+    return m_Size & HOST_ACCESSIBLE_BIT;
+}
+
+inline bool
+Buffer::IsMapped() const
+{
+    return m_Mapped;
+}

aster/device.h

@@ -33,14 +33,12 @@ struct Device final
     ~Device();
 
     template <typename T>
-        requires vk::isVulkanHandleType<T>::value
-    void SetName(const T &object, cstr name) const;
+    requires vk::isVulkanHandleType<T>::value void SetName(const T &object, cstr name) const;
     [[nodiscard]] vk::Queue GetQueue(u32 familyIndex, u32 queueIndex) const;
 };
 
 template <typename T>
-    requires vk::isVulkanHandleType<T>::value
-void
+requires vk::isVulkanHandleType<T>::value void
 Device::SetName(const T &object, cstr name) const
 {
     if (!name)

aster/global.h

@@ -30,7 +30,7 @@
 
 #include <vulkan/vulkan.hpp>
 
-constexpr u32 ASTER_API_VERSION = vk::ApiVersion13;
+constexpr u32 ASTER_API_VERSION = VK_API_VERSION_1_3;
 
 #define CODE_LOC " @ " __FILE__ ":" VULKAN_HPP_STRINGIFY(__LINE__)
 

samples/01_triangle/CMakeLists.txt

@@ -3,7 +3,7 @@
 cmake_minimum_required(VERSION 3.13)
 
 add_executable(triangle "triangle.cpp")
-add_shader(triangle shader/triangle.vs.hlsl)
+add_shader(triangle shader/triangle.vert.glsl)
 add_shader(triangle shader/triangle.frag.glsl)
 
 target_link_libraries(triangle PRIVATE aster_core)

samples/01_triangle/shader/triangle.vert.glsl

@@ -0,0 +1,27 @@
+#version 450
+#pragma shader_stage(vertex)
+
+layout(location=0) in vec4 position;
+layout(location=1) in vec4 color;
+
+layout(location=0) out vec3 outColor;
+
+void main() {
+    /*
+    vec3 points[] = {
+        vec3(-0.5f, -0.5f, 0.0f),
+        vec3(0.5f, -0.5f, 0.0f),
+        vec3(0.0f,  0.5f, 0.0f)
+    };
+    vec3 colors[] = {
+        vec3( 1.0f, 0.0f, 0.0f ),
+        vec3( 0.0f, 1.0f, 0.0f ),
+        vec3( 0.0f, 0.0f, 1.0f ),
+    };
+
+    gl_Position = vec4(points[gl_VertexIndex], 1.0f);
+    outColor = vec3(colors[gl_VertexIndex]); //*/
+    //*
+    gl_Position = vec4(position.xyz, 1.0f);
+    outColor = vec3(color.rgb); //*/
+}

samples/01_triangle/triangle.cpp

@@ -3,6 +3,7 @@
 //  Copyright (c) 2020-2024 Anish Bhobe
 // =============================================
 
+#include "buffer.h"
 #include "constants.h"
 #include "context.h"
 #include "device.h"
@@ -18,12 +19,43 @@
 #include <EASTL/array.h>
 
 constexpr u32 MAX_FRAMES_IN_FLIGHT = 3;
-constexpr auto VERTEX_SHADER_FILE = "shader/triangle.vs.hlsl.spv";
+constexpr auto VERTEX_SHADER_FILE = "shader/triangle.vert.glsl.spv";
 constexpr auto FRAGMENT_SHADER_FILE = "shader/triangle.frag.glsl.spv";
 
 vk::ShaderModule CreateShader(const Device *device, cstr shaderFile);
 Pipeline CreatePipeline(const Device *device, const Swapchain *swapchain);
 
+struct Vertex
+{
+    float m_Position[4];
+    float m_Color[4];
+
+    constexpr static vk::VertexInputBindingDescription
+    GetBinding(const u32 binding)
+    {
+        return {.binding = binding, .stride = sizeof(Vertex), .inputRate = vk::VertexInputRate::eVertex};
+    }
+
+    constexpr static eastl::array<vk::VertexInputAttributeDescription, 2>
+    GetAttributes(const u32 binding)
+    {
+        return {
+            vk::VertexInputAttributeDescription{
+                .location = 0,
+                .binding = binding,
+                .format = vk::Format::eR32G32B32A32Sfloat,
+                .offset = offsetof(Vertex, m_Position),
+            },
+            vk::VertexInputAttributeDescription{
+                .location = 1,
+                .binding = binding,
+                .format = vk::Format::eR32G32B32A32Sfloat,
+                .offset = offsetof(Vertex, m_Color),
+            },
+        };
+    }
+};
+
 struct Frame
 {
     const Device *m_Device;
@@ -51,16 +83,80 @@ main(int, char **)
 
     INFO("Using {} as the primary device.", deviceToUse.m_DeviceProperties.deviceName.data());
 
-    Features enabledDeviceFeatures = {.m_Vulkan13Features = {.dynamicRendering = vk::True}};
+    Features enabledDeviceFeatures = {.m_Vulkan13Features = {.dynamicRendering = true}};
     QueueAllocation queueAllocation = FindAppropriateQueueAllocation(&deviceToUse);
     Device device = {&context, &deviceToUse, &enabledDeviceFeatures, {queueAllocation}, "Primary Device"};
 
-    vk::Queue commandQueue = device.GetQueue(queueAllocation.m_Family, 1);
+    vk::Queue commandQueue = device.GetQueue(queueAllocation.m_Family, 0);
 
     Swapchain swapchain = {&window, &device, "Primary Chain"};
 
     Pipeline pipeline = CreatePipeline(&device, &swapchain);
 
+    vk::CommandPool copyPool;
+    vk::CommandBuffer copyBuffer;
+    {
+        vk::CommandPoolCreateInfo poolCreateInfo = {
+            .flags = vk::CommandPoolCreateFlagBits::eTransient,
+            .queueFamilyIndex = queueAllocation.m_Family,
+        };
+        auto result = device.m_Device.createCommandPool(&poolCreateInfo, nullptr, &copyPool);
+        ERROR_IF(Failed(result), "Copy command pool creation failed. Cause: {}", result) THEN_ABORT(result);
+        vk::CommandBufferAllocateInfo bufferAllocateInfo = {
+            .commandPool = copyPool,
+            .level = vk::CommandBufferLevel::ePrimary,
+            .commandBufferCount = 1,
+        };
+        result = device.m_Device.allocateCommandBuffers(&bufferAllocateInfo, &copyBuffer);
+        ERROR_IF(Failed(result), "Copy command buffer allocation failed. Cause: {}", result) THEN_ABORT(result);
+    }
+
+    // eastl::array<Vertex, 3> vertices{};
+    eastl::array vertices = {
+        Vertex{.m_Position = {-0.5f, -0.5f, 0.0f, 1.0f}, .m_Color = {1.0f, 0.0f, 0.0f, 1.0f}},
+        Vertex{.m_Position = {0.5f, -0.5f, 0.0f, 1.0f}, .m_Color = {0.0f, 1.0f, 0.0f, 1.0f}},
+        Vertex{.m_Position = {0.0f, 0.5f, 0.0f, 1.0f}, .m_Color = {0.0f, 0.0f, 1.0f, 1.0f}},
+    };
+    VertexBuffer vbo;
+    vbo.Init(&device, vertices.size() * sizeof vertices[0], "VBO");
+    {
+        StagingBuffer staging;
+        staging.Init(&device, vertices.size() * sizeof vertices[0], "Staging");
+        staging.Write(&device, 0, vertices.size() * sizeof vertices[0], vertices.data());
+
+        vk::Fence fence;
+        vk::FenceCreateInfo fenceCreateInfo = {};
+        auto result = device.m_Device.createFence(&fenceCreateInfo, nullptr, &fence);
+        ERROR_IF(Failed(result), "Fence creation failed. Cause: {}", result) THEN_ABORT(result);
+
+        vk::CommandBufferBeginInfo beginInfo = {.flags = vk::CommandBufferUsageFlagBits::eOneTimeSubmit};
+        result = copyBuffer.begin(&beginInfo);
+        ERROR_IF(Failed(result), "Copy begin failed. Cause: {}", result) THEN_ABORT(result);
+
+        vk::BufferCopy bufferCopy = {.srcOffset = 0, .dstOffset = 0, .size = staging.GetSize()};
+        copyBuffer.copyBuffer(staging.m_Buffer, vbo.m_Buffer, 1, &bufferCopy);
+
+        result = copyBuffer.end();
+        ERROR_IF(Failed(result), "Copy end failed. Cause: {}", result) THEN_ABORT(result);
+
+        vk::SubmitInfo submitInfo = {
+            .commandBufferCount = 1,
+            .pCommandBuffers = &copyBuffer,
+        };
+
+        result = commandQueue.submit(1, &submitInfo, fence);
+        ERROR_IF(Failed(result), "Submit failed. Cause: {}", result) THEN_ABORT(result) ELSE_INFO("Submit copy");
+
+        result = device.m_Device.waitForFences(1, &fence, true, MaxValue<u64>);
+        ERROR_IF(Failed(result), "Fence wait failed. Cause: {}", result) THEN_ABORT(result) ELSE_INFO("Fence wait");
+
+        result = device.m_Device.resetCommandPool(copyPool, {});
+        ERROR_IF(Failed(result), "Couldn't reset command pool. Cause: {}", result) THEN_ABORT(result);
+
+        device.m_Device.destroy(fence, nullptr);
+        staging.Destroy(&device);
+    }
+
     // Persistent variables
     vk::Viewport viewport = {
         .x = 0,
@@ -105,12 +201,13 @@ main(int, char **)
         frames.emplace_back(&device, queueAllocation.m_Family, i);
     }
 
+    INFO("Starting loop");
     u32 frameIndex = 0;
     while (window.Poll())
     {
         Frame *currentFrame = &frames[frameIndex];
 
-        auto result = device.m_Device.waitForFences(1, &currentFrame->m_FrameAvailableFence, vk::True, MaxValue<u64>);
+        auto result = device.m_Device.waitForFences(1, &currentFrame->m_FrameAvailableFence, true, MaxValue<u64>);
         ERROR_IF(Failed(result), "Waiting for fence {} failed. Cause: {}", frameIndex, result)
         THEN_ABORT(result);
 
@@ -181,6 +278,8 @@ main(int, char **)
         cmd.setViewport(0, 1, &viewport);
         cmd.setScissor(0, 1, &scissor);
         cmd.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline.m_Pipeline);
+        usize offsets = 0;
+        cmd.bindVertexBuffers(0, 1, &vbo.m_Buffer, &offsets);
         cmd.draw(3, 1, 0, 0);
 
         cmd.endRendering();
@@ -239,6 +338,9 @@ main(int, char **)
     auto result = device.m_Device.waitIdle();
     ERROR_IF(Failed(result), "Wait idle failed. Cause: {}", result);
 
+    device.m_Device.destroy(copyPool, nullptr);
+    vbo.Destroy(&device);
+
     return 0;
 }
 
@@ -283,6 +385,7 @@ Frame::AllocateCommandBuffer() const
 
     return commandBuffer;
 }
+
 Pipeline
 CreatePipeline(const Device *device, const Swapchain *swapchain)
 {
@@ -314,15 +417,18 @@ CreatePipeline(const Device *device, const Swapchain *swapchain)
     ERROR_IF(Failed(result), "Could not create a pipeline layout. Cause: {}", result) THEN_ABORT(result);
     device->SetName(pipelineLayout, "Triangle Layout");
 
+    vk::VertexInputBindingDescription inputBindingDescription = Vertex::GetBinding(0);
+    auto inputAttributeDescription = Vertex::GetAttributes(0);
+
     vk::PipelineVertexInputStateCreateInfo vertexInputStateCreateInfo = {
-        .vertexBindingDescriptionCount = 0,
-        .pVertexBindingDescriptions = nullptr,
-        .vertexAttributeDescriptionCount = 0,
-        .pVertexAttributeDescriptions = nullptr,
+        .vertexBindingDescriptionCount = 1,
+        .pVertexBindingDescriptions = &inputBindingDescription,
+        .vertexAttributeDescriptionCount = Cast<u32>(inputAttributeDescription.size()),
+        .pVertexAttributeDescriptions = inputAttributeDescription.data(),
     };
     vk::PipelineInputAssemblyStateCreateInfo inputAssemblyStateCreateInfo = {
         .topology = vk::PrimitiveTopology::eTriangleList,
-        .primitiveRestartEnable = vk::False,
+        .primitiveRestartEnable = false,
     };
 
     vk::PipelineViewportStateCreateInfo viewportStateCreateInfo = {
@@ -331,24 +437,24 @@ CreatePipeline(const Device *device, const Swapchain *swapchain)
     };
 
     vk::PipelineRasterizationStateCreateInfo rasterizationStateCreateInfo = {
-        .depthClampEnable = vk::False,
-        .rasterizerDiscardEnable = vk::False,
+        .depthClampEnable = false,
+        .rasterizerDiscardEnable = false,
         .polygonMode = vk::PolygonMode::eFill,
         .cullMode = vk::CullModeFlagBits::eNone,
         .frontFace = vk::FrontFace::eCounterClockwise,
-        .depthBiasEnable = vk::False,
+        .depthBiasEnable = false,
         .lineWidth = 1.0,
     };
     vk::PipelineMultisampleStateCreateInfo multisampleStateCreateInfo = {
         .rasterizationSamples = vk::SampleCountFlagBits::e1,
-        .sampleShadingEnable = vk::False,
+        .sampleShadingEnable = false,
     };
     vk::PipelineDepthStencilStateCreateInfo depthStencilStateCreateInfo = {
-        .depthTestEnable = vk::False,
-        .depthWriteEnable = vk::False,
+        .depthTestEnable = false,
+        .depthWriteEnable = false,
     };
     vk::PipelineColorBlendAttachmentState colorBlendAttachmentState = {
-        .blendEnable = vk::False,
+        .blendEnable = false,
         .srcColorBlendFactor = vk::BlendFactor::eSrcColor,
         .dstColorBlendFactor = vk::BlendFactor::eOneMinusSrcColor,
         .colorBlendOp = vk::BlendOp::eAdd,
@@ -359,7 +465,7 @@ CreatePipeline(const Device *device, const Swapchain *swapchain)
                           vk::ColorComponentFlagBits::eB | vk::ColorComponentFlagBits::eA,
     };
     vk::PipelineColorBlendStateCreateInfo colorBlendStateCreateInfo = {
-        .logicOpEnable = vk::False,
+        .logicOpEnable = false,
         .attachmentCount = 1,
         .pAttachments = &colorBlendAttachmentState,
     };