#include #include #include #include #include "driver/CustomAssert.h" #include #include "driver/vkExt.h" #include "QPUassembler/qpu_assembler.h" //#define GLFW_INCLUDE_VULKAN //#define VK_USE_PLATFORM_WIN32_KHR //#include //#define GLFW_EXPOSE_NATIVE_WIN32 //#include //GLFWwindow * window; //#define WINDOW_WIDTH 640 //#define WINDOW_HEIGHT 480 // Note: support swap chain recreation (not only required for resized windows!) // Note: window resize may not result in Vulkan telling that the swap chain should be recreated, should be handled explicitly! void run(); void setupVulkan(); void mainLoop(); void cleanup(); void createInstance(); void createWindowSurface(); void findPhysicalDevice(); void checkSwapChainSupport(); void findQueueFamilies(); void createLogicalDevice(); void createSemaphores(); void createSwapChain(); void createCommandQueues(); void draw(); void CreateRenderPass(); void CreateFramebuffer(); void CreateShaders(); void CreatePipeline(); void CreateUniformBuffer(); void CreateDescriptorSet(); void CreateVertexBuffer(); void CreateTexture(); void recordCommandBuffers(); VkSurfaceFormatKHR chooseSurfaceFormat(const std::vector& availableFormats); VkExtent2D chooseSwapExtent(const VkSurfaceCapabilitiesKHR& surfaceCapabilities); VkPresentModeKHR choosePresentMode(const std::vector presentModes); VkInstance instance; // VkSurfaceKHR windowSurface; // VkPhysicalDevice physicalDevice; VkDevice device; // VkSemaphore imageAvailableSemaphore; // VkSemaphore renderingFinishedSemaphore; // VkSwapchainKHR swapChain; // VkCommandPool commandPool; // std::vector presentCommandBuffers; // std::vector swapChainImages; // VkRenderPass renderPass; // std::vector fbs; // VkShaderModule shaderModule; // VkPipeline pipeline; // VkQueue graphicsQueue; VkQueue presentQueue; VkBuffer vertexBuffer; VkDeviceMemory vertexBufferMemory; VkPhysicalDeviceMemoryProperties pdmp; std::vector views; //? VkSurfaceFormatKHR swapchainFormat; VkExtent2D swapChainExtent; VkPipelineLayout pipelineLayout; uint32_t graphicsQueueFamily; uint32_t presentQueueFamily; void cleanup() { vkDeviceWaitIdle(device); // Note: this is done implicitly when the command pool is freed, but nice to know about vkFreeCommandBuffers(device, commandPool, presentCommandBuffers.size(), presentCommandBuffers.data()); vkDestroyCommandPool(device, commandPool, nullptr); vkDestroySemaphore(device, imageAvailableSemaphore, nullptr); vkDestroySemaphore(device, renderingFinishedSemaphore, nullptr); for(int c = 0; c < views.size(); ++c) vkDestroyImageView(device, views[c], 0); for (int c = 0; c < fbs.size(); ++c) vkDestroyFramebuffer(device, fbs[c], 0); vkDestroyRenderPass(device, renderPass, 0); vkDestroyShaderModule(device, shaderModule, 0); vkDestroyPipeline(device, pipeline, 0); // Note: implicitly destroys images (in fact, we're not allowed to do that explicitly) vkDestroySwapchainKHR(device, swapChain, nullptr); vkDestroyDevice(device, nullptr); vkDestroySurfaceKHR(instance, windowSurface, nullptr); vkDestroyInstance(instance, nullptr); } void run() { // Note: dynamically loading loader may be a better idea to fail gracefully when Vulkan is not supported // Create window for Vulkan //glfwInit(); //glfwWindowHint(GLFW_CLIENT_API, GLFW_NO_API); //glfwWindowHint(GLFW_RESIZABLE, GLFW_FALSE); //window = glfwCreateWindow(WINDOW_WIDTH, WINDOW_HEIGHT, "The 630 line cornflower blue window", nullptr, nullptr); // Use Vulkan setupVulkan(); mainLoop(); cleanup(); } void setupVulkan() { createInstance(); findPhysicalDevice(); createWindowSurface(); checkSwapChainSupport(); findQueueFamilies(); createLogicalDevice(); createSemaphores(); createSwapChain(); createCommandQueues(); CreateRenderPass(); CreateFramebuffer(); CreateVertexBuffer(); //CreateUniformBuffer(); CreateShaders(); CreatePipeline(); recordCommandBuffers(); } void mainLoop() { //while (!glfwWindowShouldClose(window)) { for(int c = 0; c < 300; ++c){ draw(); //glfwPollEvents(); } } void createInstance() { VkApplicationInfo appInfo = {}; appInfo.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO; appInfo.pApplicationName = "VulkanTriangle"; appInfo.applicationVersion = VK_MAKE_VERSION(1, 0, 0); appInfo.pEngineName = "TriangleEngine"; appInfo.engineVersion = VK_MAKE_VERSION(1, 0, 0); appInfo.apiVersion = VK_API_VERSION_1_0; // Get instance extensions required by GLFW to draw to window //unsigned int glfwExtensionCount; //const char** glfwExtensions; //glfwExtensions = glfwGetRequiredInstanceExtensions(&glfwExtensionCount); // Check for extensions uint32_t extensionCount = 0; vkEnumerateInstanceExtensionProperties(nullptr, &extensionCount, nullptr); if (extensionCount == 0) { std::cerr << "no extensions supported!" << std::endl; assert(0); } std::vector availableExtensions(extensionCount); vkEnumerateInstanceExtensionProperties(nullptr, &extensionCount, availableExtensions.data()); std::cout << "supported extensions:" << std::endl; for (const auto& extension : availableExtensions) { std::cout << "\t" << extension.extensionName << std::endl; } const char* enabledExtensions[] = { "VK_KHR_surface", "VK_KHR_display" }; VkInstanceCreateInfo createInfo = {}; createInfo.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO; createInfo.pNext = 0; createInfo.pApplicationInfo = &appInfo; createInfo.enabledExtensionCount = sizeof(enabledExtensions) / sizeof(const char*); createInfo.ppEnabledExtensionNames = enabledExtensions; createInfo.enabledLayerCount = 0; createInfo.ppEnabledLayerNames = 0; // Initialize Vulkan instance if (vkCreateInstance(&createInfo, nullptr, &instance) != VK_SUCCESS) { std::cerr << "failed to create instance!" << std::endl; assert(0); } else { std::cout << "created vulkan instance" << std::endl; } } void createWindowSurface() { windowSurface = 0; uint32_t displayCount; vkGetPhysicalDeviceDisplayPropertiesKHR(physicalDevice, &displayCount, 0); VkDisplayPropertiesKHR* displayProperties = (VkDisplayPropertiesKHR*)malloc(sizeof(VkDisplayPropertiesKHR)*displayCount); vkGetPhysicalDeviceDisplayPropertiesKHR(physicalDevice, &displayCount, displayProperties); printf("Enumerated displays\n"); for(uint32_t c = 0; c < displayCount; ++c) { printf("Display ID %i\n", displayProperties[c].display); printf("Display name %s\n", displayProperties[c].displayName); printf("Display width %i\n", displayProperties[c].physicalDimensions.width); printf("Display height %i\n", displayProperties[c].physicalDimensions.height); printf("Display horizontal resolution %i\n", displayProperties[c].physicalResolution.width); printf("Display vertical resolution %i\n", displayProperties[c].physicalResolution.height); } uint32_t modeCount; vkGetDisplayModePropertiesKHR(physicalDevice, displayProperties[0].display, &modeCount, 0); VkDisplayModePropertiesKHR* displayModeProperties = (VkDisplayModePropertiesKHR*)malloc(sizeof(VkDisplayModePropertiesKHR)*modeCount); vkGetDisplayModePropertiesKHR(physicalDevice, displayProperties[0].display, &modeCount, displayModeProperties); // printf("\nEnumerated modes\n"); // for(uint32_t c = 0; c < modeCount; ++c) // { // printf("Mode refresh rate %i\n", displayModeProperties[c].parameters.refreshRate); // printf("Mode width %i\n", displayModeProperties[c].parameters.visibleRegion.width); // printf("Mode height %i\n\n", displayModeProperties[c].parameters.visibleRegion.height); // } VkDisplaySurfaceCreateInfoKHR dsci = {}; dsci.sType = VK_STRUCTURE_TYPE_DISPLAY_SURFACE_CREATE_INFO_KHR; dsci.displayMode = displayModeProperties[0].displayMode; dsci.transform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR; dsci.alphaMode = VK_DISPLAY_PLANE_ALPHA_OPAQUE_BIT_KHR; dsci.imageExtent = displayModeProperties[0].parameters.visibleRegion; vkCreateDisplayPlaneSurfaceKHR(instance, &dsci, 0, &windowSurface); std::cout << "created window surface" << std::endl; } void findPhysicalDevice() { // Try to find 1 Vulkan supported device // Note: perhaps refactor to loop through devices and find first one that supports all required features and extensions uint32_t deviceCount = 1; VkResult res = vkEnumeratePhysicalDevices(instance, &deviceCount, &physicalDevice); if (res != VK_SUCCESS && res != VK_INCOMPLETE) { std::cerr << "enumerating physical devices failed!" << std::endl; assert(0); } if (deviceCount == 0) { std::cerr << "no physical devices that support vulkan!" << std::endl; assert(0); } std::cout << "physical device with vulkan support found" << std::endl; vkGetPhysicalDeviceMemoryProperties(physicalDevice, &pdmp); // Check device features // Note: will apiVersion >= appInfo.apiVersion? Probably yes, but spec is unclear. VkPhysicalDeviceProperties deviceProperties; VkPhysicalDeviceFeatures deviceFeatures; vkGetPhysicalDeviceProperties(physicalDevice, &deviceProperties); vkGetPhysicalDeviceFeatures(physicalDevice, &deviceFeatures); uint32_t supportedVersion[] = { VK_VERSION_MAJOR(deviceProperties.apiVersion), VK_VERSION_MINOR(deviceProperties.apiVersion), VK_VERSION_PATCH(deviceProperties.apiVersion) }; std::cout << "physical device supports version " << supportedVersion[0] << "." << supportedVersion[1] << "." << supportedVersion[2] << std::endl; } void checkSwapChainSupport() { uint32_t extensionCount = 0; vkEnumerateDeviceExtensionProperties(physicalDevice, nullptr, &extensionCount, nullptr); if (extensionCount == 0) { std::cerr << "physical device doesn't support any extensions" << std::endl; assert(0); } std::vector deviceExtensions(extensionCount); vkEnumerateDeviceExtensionProperties(physicalDevice, nullptr, &extensionCount, deviceExtensions.data()); for (const auto& extension : deviceExtensions) { if (strcmp(extension.extensionName, VK_KHR_SWAPCHAIN_EXTENSION_NAME) == 0) { std::cout << "physical device supports swap chains" << std::endl; return; } } std::cerr << "physical device doesn't support swap chains" << std::endl; assert(0); } void findQueueFamilies() { // Check queue families uint32_t queueFamilyCount = 0; vkGetPhysicalDeviceQueueFamilyProperties(physicalDevice, &queueFamilyCount, nullptr); if (queueFamilyCount == 0) { std::cout << "physical device has no queue families!" << std::endl; assert(0); } // Find queue family with graphics support // Note: is a transfer queue necessary to copy vertices to the gpu or can a graphics queue handle that? std::vector queueFamilies(queueFamilyCount); vkGetPhysicalDeviceQueueFamilyProperties(physicalDevice, &queueFamilyCount, queueFamilies.data()); std::cout << "physical device has " << queueFamilyCount << " queue families" << std::endl; bool foundGraphicsQueueFamily = false; bool foundPresentQueueFamily = false; for (uint32_t i = 0; i < queueFamilyCount; i++) { VkBool32 presentSupport = false; vkGetPhysicalDeviceSurfaceSupportKHR(physicalDevice, i, windowSurface, &presentSupport); if (queueFamilies[i].queueCount > 0 && queueFamilies[i].queueFlags & VK_QUEUE_GRAPHICS_BIT) { graphicsQueueFamily = i; foundGraphicsQueueFamily = true; if (presentSupport) { presentQueueFamily = i; foundPresentQueueFamily = true; break; } } if (!foundPresentQueueFamily && presentSupport) { presentQueueFamily = i; foundPresentQueueFamily = true; } } if (foundGraphicsQueueFamily) { std::cout << "queue family #" << graphicsQueueFamily << " supports graphics" << std::endl; if (foundPresentQueueFamily) { std::cout << "queue family #" << presentQueueFamily << " supports presentation" << std::endl; } else { std::cerr << "could not find a valid queue family with present support" << std::endl; assert(0); } } else { std::cerr << "could not find a valid queue family with graphics support" << std::endl; assert(0); } } void createLogicalDevice() { // Greate one graphics queue and optionally a separate presentation queue float queuePriority = 1.0f; VkDeviceQueueCreateInfo queueCreateInfo[2] = {}; queueCreateInfo[0].sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO; queueCreateInfo[0].queueFamilyIndex = graphicsQueueFamily; queueCreateInfo[0].queueCount = 1; queueCreateInfo[0].pQueuePriorities = &queuePriority; queueCreateInfo[0].sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO; queueCreateInfo[0].queueFamilyIndex = presentQueueFamily; queueCreateInfo[0].queueCount = 1; queueCreateInfo[0].pQueuePriorities = &queuePriority; // Create logical device from physical device // Note: there are separate instance and device extensions! VkDeviceCreateInfo deviceCreateInfo = {}; deviceCreateInfo.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO; deviceCreateInfo.pQueueCreateInfos = queueCreateInfo; if (graphicsQueueFamily == presentQueueFamily) { deviceCreateInfo.queueCreateInfoCount = 1; } else { deviceCreateInfo.queueCreateInfoCount = 2; } const char* deviceExtensions = VK_KHR_SWAPCHAIN_EXTENSION_NAME; deviceCreateInfo.enabledExtensionCount = 1; deviceCreateInfo.ppEnabledExtensionNames = &deviceExtensions; if (vkCreateDevice(physicalDevice, &deviceCreateInfo, nullptr, &device) != VK_SUCCESS) { std::cerr << "failed to create logical device" << std::endl; assert(0); } std::cout << "created logical device" << std::endl; // Get graphics and presentation queues (which may be the same) vkGetDeviceQueue(device, graphicsQueueFamily, 0, &graphicsQueue); vkGetDeviceQueue(device, presentQueueFamily, 0, &presentQueue); std::cout << "acquired graphics and presentation queues" << std::endl; } void createSemaphores() { VkSemaphoreCreateInfo createInfo = {}; createInfo.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO; if (vkCreateSemaphore(device, &createInfo, nullptr, &imageAvailableSemaphore) != VK_SUCCESS || vkCreateSemaphore(device, &createInfo, nullptr, &renderingFinishedSemaphore) != VK_SUCCESS) { std::cerr << "failed to create semaphores" << std::endl; assert(0); } else { std::cout << "created semaphores" << std::endl; } } void createSwapChain() { // Find surface capabilities VkSurfaceCapabilitiesKHR surfaceCapabilities; if (vkGetPhysicalDeviceSurfaceCapabilitiesKHR(physicalDevice, windowSurface, &surfaceCapabilities) != VK_SUCCESS) { std::cerr << "failed to acquire presentation surface capabilities" << std::endl; assert(0); } // Find supported surface formats uint32_t formatCount; if (vkGetPhysicalDeviceSurfaceFormatsKHR(physicalDevice, windowSurface, &formatCount, nullptr) != VK_SUCCESS || formatCount == 0) { std::cerr << "failed to get number of supported surface formats" << std::endl; assert(0); } std::vector surfaceFormats(formatCount); if (vkGetPhysicalDeviceSurfaceFormatsKHR(physicalDevice, windowSurface, &formatCount, surfaceFormats.data()) != VK_SUCCESS) { std::cerr << "failed to get supported surface formats" << std::endl; assert(0); } // Find supported present modes uint32_t presentModeCount; if (vkGetPhysicalDeviceSurfacePresentModesKHR(physicalDevice, windowSurface, &presentModeCount, nullptr) != VK_SUCCESS || presentModeCount == 0) { std::cerr << "failed to get number of supported presentation modes" << std::endl; assert(0); } std::vector presentModes(presentModeCount); if (vkGetPhysicalDeviceSurfacePresentModesKHR(physicalDevice, windowSurface, &presentModeCount, presentModes.data()) != VK_SUCCESS) { std::cerr << "failed to get supported presentation modes" << std::endl; assert(0); } // Determine number of images for swap chain uint32_t imageCount = surfaceCapabilities.minImageCount + 1; if (surfaceCapabilities.maxImageCount != 0 && imageCount > surfaceCapabilities.maxImageCount) { imageCount = surfaceCapabilities.maxImageCount; } std::cout << "using " << imageCount << " images for swap chain" << std::endl; // Select a surface format swapchainFormat = chooseSurfaceFormat(surfaceFormats); // Select swap chain size swapChainExtent = chooseSwapExtent(surfaceCapabilities); // Check if swap chain supports being the destination of an image transfer // Note: AMD driver bug, though it would be nice to implement a workaround that doesn't use transfering //if (!(surfaceCapabilities.supportedUsageFlags & VK_IMAGE_USAGE_TRANSFER_DST_BIT)) { // std::cerr << "swap chain image does not support VK_IMAGE_TRANSFER_DST usage" << std::endl; //assert(0); //} // Determine transformation to use (preferring no transform) VkSurfaceTransformFlagBitsKHR surfaceTransform; if (surfaceCapabilities.supportedTransforms & VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR) { surfaceTransform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR; } else { surfaceTransform = surfaceCapabilities.currentTransform; } // Choose presentation mode (preferring MAILBOX ~= triple buffering) VkPresentModeKHR presentMode = choosePresentMode(presentModes); // Finally, create the swap chain VkSwapchainCreateInfoKHR createInfo = {}; createInfo.sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR; createInfo.surface = windowSurface; createInfo.minImageCount = imageCount; createInfo.imageFormat = swapchainFormat.format; createInfo.imageColorSpace = swapchainFormat.colorSpace; createInfo.imageExtent = swapChainExtent; createInfo.imageArrayLayers = 1; createInfo.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; createInfo.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; createInfo.queueFamilyIndexCount = 0; createInfo.pQueueFamilyIndices = nullptr; createInfo.preTransform = surfaceTransform; createInfo.compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR; createInfo.presentMode = presentMode; createInfo.clipped = VK_TRUE; createInfo.oldSwapchain = VK_NULL_HANDLE; if (vkCreateSwapchainKHR(device, &createInfo, nullptr, &swapChain) != VK_SUCCESS) { std::cerr << "failed to create swap chain" << std::endl; assert(0); } else { std::cout << "created swap chain" << std::endl; } // Store the images used by the swap chain // Note: these are the images that swap chain image indices refer to // Note: actual number of images may differ from requested number, since it's a lower bound uint32_t actualImageCount = 0; if (vkGetSwapchainImagesKHR(device, swapChain, &actualImageCount, nullptr) != VK_SUCCESS || actualImageCount == 0) { std::cerr << "failed to acquire number of swap chain images" << std::endl; assert(0); } swapChainImages.resize(actualImageCount); views.resize(actualImageCount); if (vkGetSwapchainImagesKHR(device, swapChain, &actualImageCount, swapChainImages.data()) != VK_SUCCESS) { std::cerr << "failed to acquire swap chain images" << std::endl; assert(0); } std::cout << "acquired swap chain images" << std::endl; } VkSurfaceFormatKHR chooseSurfaceFormat(const std::vector& availableFormats) { // We can either choose any format if (availableFormats.size() == 1 && availableFormats[0].format == VK_FORMAT_UNDEFINED) { return { VK_FORMAT_R8G8B8A8_UNORM, VK_COLORSPACE_SRGB_NONLINEAR_KHR }; } // Or go with the standard format - if available for (const auto& availableSurfaceFormat : availableFormats) { if (availableSurfaceFormat.format == VK_FORMAT_R8G8B8A8_UNORM) { return availableSurfaceFormat; } } // Or fall back to the first available one return availableFormats[0]; } VkExtent2D chooseSwapExtent(const VkSurfaceCapabilitiesKHR& surfaceCapabilities) { if (surfaceCapabilities.currentExtent.width == -1) { VkExtent2D swapChainExtent = {}; #define min(a, b) (a < b ? a : b) #define max(a, b) (a > b ? a : b) swapChainExtent.width = min(max(640, surfaceCapabilities.minImageExtent.width), surfaceCapabilities.maxImageExtent.width); swapChainExtent.height = min(max(480, surfaceCapabilities.minImageExtent.height), surfaceCapabilities.maxImageExtent.height); return swapChainExtent; } else { return surfaceCapabilities.currentExtent; } } VkPresentModeKHR choosePresentMode(const std::vector presentModes) { for (const auto& presentMode : presentModes) { if (presentMode == VK_PRESENT_MODE_MAILBOX_KHR) { return presentMode; } } // If mailbox is unavailable, fall back to FIFO (guaranteed to be available) return VK_PRESENT_MODE_FIFO_KHR; } void createCommandQueues() { // Create presentation command pool // Note: only command buffers for a single queue family can be created from this pool VkCommandPoolCreateInfo poolCreateInfo = {}; poolCreateInfo.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO; poolCreateInfo.queueFamilyIndex = presentQueueFamily; if (vkCreateCommandPool(device, &poolCreateInfo, nullptr, &commandPool) != VK_SUCCESS) { std::cerr << "failed to create command queue for presentation queue family" << std::endl; assert(0); } else { std::cout << "created command pool for presentation queue family" << std::endl; } // Get number of swap chain images and create vector to hold command queue for each one presentCommandBuffers.resize(swapChainImages.size()); // Allocate presentation command buffers // Note: secondary command buffers are only for nesting in primary command buffers VkCommandBufferAllocateInfo allocInfo = {}; allocInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO; allocInfo.commandPool = commandPool; allocInfo.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY; allocInfo.commandBufferCount = (uint32_t)swapChainImages.size(); if (vkAllocateCommandBuffers(device, &allocInfo, presentCommandBuffers.data()) != VK_SUCCESS) { std::cerr << "failed to allocate presentation command buffers" << std::endl; assert(0); } else { std::cout << "allocated presentation command buffers" << std::endl; } } void recordCommandBuffers() { // Prepare data for recording command buffers VkCommandBufferBeginInfo beginInfo = {}; beginInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; beginInfo.flags = VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT; // Note: contains value for each subresource range VkClearColorValue clearColor = { { 0.4f, 0.6f, 0.9f, 1.0f } // R, G, B, A }; VkClearValue clearValue = {}; clearValue.color = clearColor; VkImageSubresourceRange subResourceRange = {}; subResourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; subResourceRange.baseMipLevel = 0; subResourceRange.levelCount = 1; subResourceRange.baseArrayLayer = 0; subResourceRange.layerCount = 1; VkRenderPassBeginInfo renderPassInfo = {}; renderPassInfo.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; renderPassInfo.renderPass = renderPass; renderPassInfo.renderArea.offset.x = 0; renderPassInfo.renderArea.offset.y = 0; renderPassInfo.renderArea.extent.width = swapChainExtent.width; renderPassInfo.renderArea.extent.height = swapChainExtent.height; renderPassInfo.clearValueCount = 1; renderPassInfo.pClearValues = &clearValue; VkViewport viewport = { 0 }; viewport.height = (float)swapChainExtent.width; viewport.width = (float)swapChainExtent.height; viewport.minDepth = (float)0.0f; viewport.maxDepth = (float)1.0f; VkRect2D scissor = { 0 }; scissor.extent.width = swapChainExtent.width; scissor.extent.height = swapChainExtent.height; scissor.offset.x = 0; scissor.offset.y = 0; // Record the command buffer for every swap chain image for (uint32_t i = 0; i < swapChainImages.size(); i++) { // Record command buffer vkBeginCommandBuffer(presentCommandBuffers[i], &beginInfo); renderPassInfo.framebuffer = fbs[i]; vkCmdBeginRenderPass(presentCommandBuffers[i], &renderPassInfo, VK_SUBPASS_CONTENTS_INLINE); vkCmdBindPipeline(presentCommandBuffers[i], VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); //vkCmdSetViewport(presentCommandBuffers[i], 0, 1, &viewport); //vkCmdSetScissor(presentCommandBuffers[i], 0, 1, &scissor); VkDeviceSize offsets[2] = {0, sizeof(float)*2*3}; VkBuffer buffers[2] = {vertexBuffer, vertexBuffer}; vkCmdBindVertexBuffers(presentCommandBuffers[i], 0, 2, buffers, offsets); float Wcoeff = 1.0f; //1.0f / Wc = 2.0 - Wcoeff float viewportScaleX = (float)(swapChainExtent.width) * 0.5f * 16.0f; float viewportScaleY = 1.0f * (float)(swapChainExtent.height) * 0.5f * 16.0f; float Zs = 1.0f; uint32_t pushConstants[4]; pushConstants[0] = *(uint32_t*)&Wcoeff; pushConstants[1] = *(uint32_t*)&viewportScaleX; pushConstants[2] = *(uint32_t*)&viewportScaleY; pushConstants[3] = *(uint32_t*)&Zs; vkCmdPushConstants(presentCommandBuffers[i], pipelineLayout, VK_SHADER_STAGE_VERTEX_BIT, 0, sizeof(pushConstants), &pushConstants); vkCmdDraw(presentCommandBuffers[i], 3, 1, 0, 0); vkCmdEndRenderPass(presentCommandBuffers[i]); if (vkEndCommandBuffer(presentCommandBuffers[i]) != VK_SUCCESS) { std::cerr << "failed to record command buffer" << std::endl; assert(0); } else { std::cout << "recorded command buffer for image " << i << std::endl; } } } void draw() { // Acquire image uint32_t imageIndex; VkResult res = vkAcquireNextImageKHR(device, swapChain, UINT64_MAX, imageAvailableSemaphore, VK_NULL_HANDLE, &imageIndex); if (res != VK_SUCCESS && res != VK_SUBOPTIMAL_KHR) { std::cerr << "failed to acquire image" << std::endl; assert(0); } std::cout << "acquired image" << std::endl; // Wait for image to be available and draw VkSubmitInfo submitInfo = {}; submitInfo.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; submitInfo.waitSemaphoreCount = 1; submitInfo.pWaitSemaphores = &imageAvailableSemaphore; submitInfo.signalSemaphoreCount = 1; submitInfo.pSignalSemaphores = &renderingFinishedSemaphore; submitInfo.commandBufferCount = 1; submitInfo.pCommandBuffers = &presentCommandBuffers[imageIndex]; if (vkQueueSubmit(presentQueue, 1, &submitInfo, VK_NULL_HANDLE) != VK_SUCCESS) { std::cerr << "failed to submit draw command buffer" << std::endl; assert(0); } std::cout << "submitted draw command buffer" << std::endl; // Present drawn image // Note: semaphore here is not strictly necessary, because commands are processed in submission order within a single queue VkPresentInfoKHR presentInfo = {}; presentInfo.sType = VK_STRUCTURE_TYPE_PRESENT_INFO_KHR; presentInfo.waitSemaphoreCount = 1; presentInfo.pWaitSemaphores = &renderingFinishedSemaphore; presentInfo.swapchainCount = 1; presentInfo.pSwapchains = &swapChain; presentInfo.pImageIndices = &imageIndex; res = vkQueuePresentKHR(presentQueue, &presentInfo); if (res != VK_SUCCESS) { std::cerr << "failed to submit present command buffer" << std::endl; assert(0); } std::cout << "submitted presentation command buffer" << std::endl; } void CreateRenderPass() { VkAttachmentReference attachRef = {}; attachRef.attachment = 0; attachRef.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; VkSubpassDescription subpassDesc = {}; subpassDesc.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; subpassDesc.colorAttachmentCount = 1; subpassDesc.pColorAttachments = &attachRef; VkAttachmentDescription attachDesc = {}; attachDesc.format = swapchainFormat.format; // attachDesc.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR; attachDesc.storeOp = VK_ATTACHMENT_STORE_OP_STORE; attachDesc.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; attachDesc.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; attachDesc.initialLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR; attachDesc.finalLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR; attachDesc.samples = VK_SAMPLE_COUNT_1_BIT; VkRenderPassCreateInfo renderPassCreateInfo = {}; renderPassCreateInfo.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; renderPassCreateInfo.attachmentCount = 1; renderPassCreateInfo.pAttachments = &attachDesc; renderPassCreateInfo.subpassCount = 1; renderPassCreateInfo.pSubpasses = &subpassDesc; VkResult res = vkCreateRenderPass(device, &renderPassCreateInfo, NULL, &renderPass); printf("Created a render pass\n"); } void CreateFramebuffer() { fbs.resize(swapChainImages.size()); VkResult res; for (uint32_t i = 0; i < swapChainImages.size(); i++) { VkImageViewCreateInfo ViewCreateInfo = {}; ViewCreateInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; ViewCreateInfo.image = swapChainImages[i]; ViewCreateInfo.format = swapchainFormat.format; // ViewCreateInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; ViewCreateInfo.components.r = VK_COMPONENT_SWIZZLE_IDENTITY; ViewCreateInfo.components.g = VK_COMPONENT_SWIZZLE_IDENTITY; ViewCreateInfo.components.b = VK_COMPONENT_SWIZZLE_IDENTITY; ViewCreateInfo.components.a = VK_COMPONENT_SWIZZLE_IDENTITY; ViewCreateInfo.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; ViewCreateInfo.subresourceRange.baseMipLevel = 0; ViewCreateInfo.subresourceRange.levelCount = 1; ViewCreateInfo.subresourceRange.baseArrayLayer = 0; ViewCreateInfo.subresourceRange.layerCount = 1; res = vkCreateImageView(device, &ViewCreateInfo, NULL, &views[i]); VkFramebufferCreateInfo fbCreateInfo = {}; fbCreateInfo.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; fbCreateInfo.renderPass = renderPass; fbCreateInfo.attachmentCount = 1; fbCreateInfo.pAttachments = &views[i]; fbCreateInfo.width = swapChainExtent.width; fbCreateInfo.height = swapChainExtent.height; fbCreateInfo.layers = 1; res = vkCreateFramebuffer(device, &fbCreateInfo, NULL, &fbs[i]); } printf("Frame buffers created\n"); } void CreateShaders() { /** //FS prog reading tex with varyings 0x100049e0203e303e nop nop, r0, r0 ; fmul r0, vary, ra15 //mul by w 0x100248e1213e317e fadd r3, r0, r5 ; fmul r1, vary, ra15 //add interpolation coefficient C, mul by w 0x600208a7019e7340 sig_thread_switch fadd r2, r1, r5 ; nop nop, r0, r0 0x10021e67159e7480 mov tmu0_t, r2 ; nop nop, r0, r0 0x10021e27159e76c0 mov tmu0_s, r3 ; nop nop, r0, r0 0xa00009e7009e7000 load_tmu0 nop nop, r0, r0 ; nop nop, r0, r0 0x190208e7049e7900 fmax r3, r4.8a, r4.8a ; nop nop, r0, r0 0x1b424823849e791b fmax r0, r4.8b, r4.8b ; mov r3.8a, r3 0x1d524863849e7900 fmax r1, r4.8c, r4.8c ; mov r3.8b, r0 0x1f6248a3849e7909 fmax r2, r4.8d, r4.8d ; mov r3.8c, r1 0x117049e3809e7012 nop nop, r0, r0 ; mov r3.8d, r2 0x10020ba7159e76c0 mov tlb_color_all, r3 ; nop nop, r0, r0 0x300009e7009e7000 sig_end nop nop, r0, r0 ; nop nop, r0, r0 0x100009e7009e7000 nop nop, r0, r0 ; nop nop, r0, r0 0x500009e7009e7000 sig_unlock_score nop nop, r0, r0 ; nop nop, r0, r0 /**/ char vs_asm_code[] = ///0x40000000 = 2.0 ///uni = 1.0 ///rb0 = 2 - 1 = 1 "sig_small_imm ; rx0 = fsub.ws.always(b, a, uni, 0x40000000) ; nop = nop(r0, r0) ;\n" ///set up VPM read for subsequent reads ///0x00201a00: 0000 0000 0010 0000 0001 1010 0000 0000 ///addr: 0 ///size: 32bit ///packed ///horizontal ///stride=1 ///vectors to read = 5 "sig_load_imm ; vr_setup = load32.always(0x00301a00) ; nop = load32.always() ;\n" ///uni = viewportXScale ///r0 = vpm * uni "sig_none ; nop = nop(r0, r0, vpm_read, uni) ; r0 = fmul.always(a, b) ;\n" ///r1 = r0 * rb0 (1) "sig_none ; nop = nop(r0, r0, nop, rb0) ; r1 = fmul.always(r0, b) ;\n" ///uni = viewportYScale ///ra0.16a = int(r1), r2 = vpm * uni "sig_none ; rx0.16a = ftoi.always(r1, r1, vpm_read, uni) ; r2 = fmul.always(a, b) ;\n" ///r3 = r2 * rb0 ///r0 = vpm "sig_none ; rx1 = or.always(a, a, vpm_read, rb0) ; r3 = fmul.always(r2, b) ;\n" ///ra0.16b = int(r3) ///r1 = vpm "sig_none ; rx0.16b = ftoi.always(r3, r3) ; nop = nop(r0, r0) ;\n" ///set up VPM write for subsequent writes ///0x00001a00: 0000 0000 0000 0000 0001 1010 0000 0000 ///addr: 0 ///size: 32bit ///horizontal ///stride = 1 "sig_load_imm ; vw_setup = load32.always.ws(0x00001a00) ; nop = load32.always() ;\n" ///shaded vertex format for PSE /// Ys and Xs ///vpm = ra0 "sig_none ; vpm = or.always(a, a, ra0, nop) ; nop = nop(r0, r0);\n" /// Zs ///uni = 0.5 ///vpm = uni "sig_none ; vpm = or.always(a, a, uni, nop) ; nop = nop(r0, r0);\n" /// 1.0 / Wc ///vpm = rb0 (1) "sig_none ; vpm = or.always(b, b, nop, rb0) ; nop = nop(r0, r0);\n" ///vpm = r0 "sig_none ; vpm.nop = fmax.always.8a(a, a, ra1) ; nop = nop(r0, r0) ; " "sig_none ; vpm.nop = fmax.always.8b(a, a, ra1) ; nop = nop(r0, r0) ; " "sig_none ; vpm.nop = fmax.always.8c(a, a, ra1) ; nop = nop(r0, r0) ; " "sig_none ; vpm.nop = fmax.always.8d(a, a, ra1) ; nop = nop(r0, r0) ; " ///END "sig_end ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;\n" "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;\n" "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;\n" "\0"; char cs_asm_code[] = ///uni = 1.0 ///r3 = 2.0 - uni "sig_small_imm ; r3 = fsub.always(b, a, uni, 0x40000000) ; nop = nop(r0, r0);\n" "sig_load_imm ; vr_setup = load32.always(0x00201a00) ; nop = load32.always() ;\n" ///r2 = vpm "sig_none ; r2 = or.always(a, a, vpm_read, nop) ; nop = nop(r0, r0);\n" "sig_load_imm ; vw_setup = load32.always.ws(0x00001a00) ; nop = load32.always() ;\n" ///shaded coordinates format for PTB /// write Xc ///r1 = vpm, vpm = r2 "sig_none ; r1 = or.always(a, a, vpm_read, nop) ; vpm = v8min.always(r2, r2);\n" /// write Yc ///uni = viewportXscale ///vpm = r1, r2 = r2 * uni "sig_none ; vpm = or.always(r1, r1, uni, nop) ; r2 = fmul.always(r2, a);\n" ///uni = viewportYscale ///r1 = r1 * uni "sig_none ; nop = nop(r0, r0, uni, nop) ; r1 = fmul.always(r1, a);\n" ///r0 = r2 * r3 "sig_none ; nop = nop(r0, r0) ; r0 = fmul.always(r2, r3);\n" ///ra0.16a = r0, r1 = r1 * r3 "sig_none ; rx0.16a = ftoi.always(r0, r0) ; r1 = fmul.always(r1, r3) ;\n" ///ra0.16b = r1 "sig_none ; rx0.16b = ftoi.always(r1, r1) ; nop = nop(r0, r0) ;\n" ///write Zc ///vpm = 0 "sig_small_imm ; vpm = or.always(b, b, nop, 0) ; nop = nop(r0, r0) ;\n" ///write Wc ///vpm = 1.0 "sig_small_imm ; vpm = or.always(b, b, nop, 0x3f800000) ; nop = nop(r0, r0) ;\n" ///write Ys and Xs ///vpm = ra0 "sig_none ; vpm = or.always(a, a, ra0, nop) ; nop = nop(r0, r0) ;\n" ///write Zs ///uni = 0.5 ///vpm = uni "sig_none ; vpm = or.always(a, a, uni, nop) ; nop = nop(r0, r0) ;\n" ///write 1/Wc ///vpm = r3 "sig_none ; vpm = or.always(r3, r3) ; nop = nop(r0, r0) ;\n" ///END "sig_end ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;\n" "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;\n" "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;\n" "\0"; //clever: use small immedate -1 interpreted as 0xffffffff (white) to set color to white //"sig_small_imm ; tlb_color_all = or.always(b, b, nop, -1) ; nop = nop(r0, r0) ;" //8bit access //abcd //BGRA /** //rainbow colors char fs_asm_code[] = "sig_none ; r1 = itof.always(a, a, x_pix, uni) ; r3 = v8min.always(b, b) ;" //can't use mul pipeline for conversion :( "sig_load_imm ; r2 = load32.always(0x3a088888) ; nop = load32() ;" //1/1920 "sig_none ; nop = nop(r0, r0) ; r2 = fmul.always(r2, r3);\n" "sig_none ; r1 = itof.pm.always(b, b, x_pix, y_pix) ; r0.8c = fmul.always(r1, r2) ;" "sig_load_imm ; r2 = load32.always(0x3a72b9d6) ; nop = load32() ;" //1/1080 "sig_none ; nop = nop(r0, r0) ; r2 = fmul.always(r2, r3);\n" "sig_none ; nop = nop.pm(r0, r0) ; r0.8b = fmul.always(r1, r2) ;" "sig_small_imm ; nop = nop.pm(r0, r0, nop, 0) ; r0.8a = v8min.always(b, b) ;" "sig_small_imm ; nop = nop.pm(r0, r0, nop, 0x3f800000) ; r0.8d = v8min.always(b, b) ;" "sig_none ; tlb_color_all = or.always(r0, r0) ; nop = nop(r0, r0) ;" "sig_end ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "sig_unlock_score ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "\0"; /**/ /** //display a color char fs_asm_code[] = "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "sig_load_imm ; r0 = load32.always(0xffa14ccc) ; nop = load32() ;" "sig_none ; tlb_color_all = or.always(r0, r0) ; nop = nop(r0, r0) ;" "sig_end ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "sig_unlock_score ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "\0"; /**/ /**/ //Varyings are interpolated using (A * (x - x0) + B * (y - y0)) * W + C //Hardware calculates (A * (x - x0) + B * (y - y0)) for us for each varying component //but we still need to mul by W and add C //C is loaded into R5 when we read a varying (to be used in next instruction) //display a varying char fs_asm_code[] = ///r0 = varyingX * W "sig_none ; nop = nop(r0, r0, pay_zw, vary) ; r0 = fmul.always(a, b) ;" ///r2 = r0 + r5 (C) "sig_none ; rx0 = fadd.always(r0, r5, pay_zw, vary) ; r1 = fmul.always(a, b) ;" "sig_none ; rx1 = fadd.always(r1, r5, pay_zw, vary) ; r2 = fmul.always(a, b) ;" "sig_none ; rx2 = fadd.always(r2, r5, pay_zw, vary) ; r3 = fmul.always(a, b) ;" "sig_none ; rx3 = fadd.always(r3, r5) ; nop = nop(r0, r0) ;" "sig_none ; nop = nop.pm(r0, r0, ra0, nop) ; r0.8a = v8min.always(a, a) ;" "sig_none ; nop = nop.pm(r0, r0, ra1, nop) ; r0.8b = v8min.always(a, a) ;" "sig_none ; nop = nop.pm(r0, r0, ra2, nop) ; r0.8c = v8min.always(a, a) ;" "sig_none ; nop = nop.pm(r0, r0, ra3, nop) ; r0.8d = v8min.always(a, a) ;" "sig_none ; tlb_color_all = or.always(r0, r0) ; nop = nop(r0, r0) ;" "sig_end ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "sig_none ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "sig_unlock_score ; nop = nop(r0, r0) ; nop = nop(r0, r0) ;" "\0"; /**/ char* asm_strings[] = { (char*)cs_asm_code, (char*)vs_asm_code, (char*)fs_asm_code, 0 }; VkRpiAssemblyMappingEXT vertexMappings[] = { //vertex shader uniforms { VK_RPI_ASSEMBLY_MAPPING_TYPE_PUSH_CONSTANT, VK_DESCRIPTOR_TYPE_MAX_ENUM, //descriptor type 0, //descriptor set # 0, //descriptor binding # 0, //descriptor array element # 0, //resource offset }, { VK_RPI_ASSEMBLY_MAPPING_TYPE_PUSH_CONSTANT, VK_DESCRIPTOR_TYPE_MAX_ENUM, //descriptor type 0, //descriptor set # 0, //descriptor binding # 0, //descriptor array element # 4, //resource offset }, { VK_RPI_ASSEMBLY_MAPPING_TYPE_PUSH_CONSTANT, VK_DESCRIPTOR_TYPE_MAX_ENUM, //descriptor type 0, //descriptor set # 0, //descriptor binding # 0, //descriptor array element # 8, //resource offset }, { VK_RPI_ASSEMBLY_MAPPING_TYPE_PUSH_CONSTANT, VK_DESCRIPTOR_TYPE_MAX_ENUM, //descriptor type 0, //descriptor set # 0, //descriptor binding # 0, //descriptor array element # 12, //resource offset } }; uint32_t spirv[6]; uint64_t* asm_ptrs[4] = {}; uint32_t asm_sizes[4] = {}; VkRpiAssemblyMappingEXT* asm_mappings[4] = {}; uint32_t asm_mappings_sizes[4] = {}; VkRpiShaderModuleAssemblyCreateInfoEXT shaderModuleCreateInfo = {}; shaderModuleCreateInfo.instructions = asm_ptrs; shaderModuleCreateInfo.numInstructions = asm_sizes; shaderModuleCreateInfo.mappings = asm_mappings; shaderModuleCreateInfo.numMappings = asm_mappings_sizes; asm_mappings[VK_RPI_ASSEMBLY_TYPE_VERTEX] = vertexMappings; asm_mappings_sizes[VK_RPI_ASSEMBLY_TYPE_VERTEX] = sizeof(vertexMappings) / sizeof(VkRpiAssemblyMappingEXT); { //assemble cs code asm_sizes[0] = get_num_instructions(cs_asm_code); uint32_t size = sizeof(uint64_t)*asm_sizes[0]; asm_ptrs[0] = (uint64_t*)malloc(size); assemble_qpu_asm(cs_asm_code, asm_ptrs[0]); } { //assemble vs code asm_sizes[1] = get_num_instructions(vs_asm_code); uint32_t size = sizeof(uint64_t)*asm_sizes[1]; asm_ptrs[1] = (uint64_t*)malloc(size); assemble_qpu_asm(vs_asm_code, asm_ptrs[1]); } { //assemble fs code asm_sizes[2] = get_num_instructions(fs_asm_code); uint32_t size = sizeof(uint64_t)*asm_sizes[2]; asm_ptrs[2] = (uint64_t*)malloc(size); assemble_qpu_asm(fs_asm_code, asm_ptrs[2]); } asm_sizes[3] = 0; asm_ptrs[3] = 0; spirv[0] = 0x07230203; spirv[1] = 0x00010000; spirv[2] = 0x14E45250; spirv[3] = 1; spirv[4] = (uint32_t)&shaderModuleCreateInfo; //words start here spirv[5] = 1 << 16; VkShaderModuleCreateInfo smci = {}; smci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; smci.codeSize = sizeof(uint32_t)*6; smci.pCode = spirv; vkCreateShaderModule(device, &smci, 0, &shaderModule); for(uint32_t c = 0; c < 4; ++c) { free(asm_ptrs[c]); } } #define VERTEX_BUFFER_BIND_ID 0 void CreatePipeline() { VkPushConstantRange pushConstantRanges[1]; pushConstantRanges[0].offset = 0; pushConstantRanges[0].size = 4 * 4; //4 * 32bits pushConstantRanges[0].stageFlags = VK_SHADER_STAGE_VERTEX_BIT; VkPipelineLayoutCreateInfo pipelineLayoutCI = {}; pipelineLayoutCI.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; pipelineLayoutCI.setLayoutCount = 0; pipelineLayoutCI.pushConstantRangeCount = 1; pipelineLayoutCI.pPushConstantRanges = &pushConstantRanges[0]; vkCreatePipelineLayout(device, &pipelineLayoutCI, 0, &pipelineLayout); VkPipelineShaderStageCreateInfo shaderStageCreateInfo[2] = {}; shaderStageCreateInfo[0].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; shaderStageCreateInfo[0].stage = VK_SHADER_STAGE_VERTEX_BIT; shaderStageCreateInfo[0].module = shaderModule; shaderStageCreateInfo[0].pName = "main"; shaderStageCreateInfo[1].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; shaderStageCreateInfo[1].stage = VK_SHADER_STAGE_FRAGMENT_BIT; shaderStageCreateInfo[1].module = shaderModule; shaderStageCreateInfo[1].pName = "main"; VkVertexInputBindingDescription vertexInputBindingDescription[2] = { { 0, sizeof(float) * 2, VK_VERTEX_INPUT_RATE_VERTEX }, { 1, sizeof(uint8_t) * 4, VK_VERTEX_INPUT_RATE_VERTEX }, }; VkVertexInputAttributeDescription vertexInputAttributeDescriptions[2] = { { 0, 0, VK_FORMAT_R32G32_SFLOAT, 0 }, { 1, 1, VK_FORMAT_R8G8B8A8_UNORM, 0 } }; VkPipelineVertexInputStateCreateInfo vertexInputInfo = {}; vertexInputInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO; vertexInputInfo.vertexAttributeDescriptionCount = 2; vertexInputInfo.pVertexAttributeDescriptions = vertexInputAttributeDescriptions; vertexInputInfo.vertexBindingDescriptionCount = 2; vertexInputInfo.pVertexBindingDescriptions = vertexInputBindingDescription; VkPipelineInputAssemblyStateCreateInfo pipelineIACreateInfo = {}; pipelineIACreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO; pipelineIACreateInfo.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST; VkViewport vp = {}; vp.x = 0.0f; vp.y = 0.0f; vp.width = (float)swapChainExtent.width; vp.height = (float)swapChainExtent.height; vp.minDepth = 0.0f; vp.maxDepth = 1.0f; VkPipelineViewportStateCreateInfo vpCreateInfo = {}; vpCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO; vpCreateInfo.viewportCount = 1; vpCreateInfo.pViewports = &vp; VkPipelineRasterizationStateCreateInfo rastCreateInfo = {}; rastCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; rastCreateInfo.polygonMode = VK_POLYGON_MODE_FILL; rastCreateInfo.cullMode = VK_CULL_MODE_NONE; rastCreateInfo.frontFace = VK_FRONT_FACE_COUNTER_CLOCKWISE; rastCreateInfo.lineWidth = 1.0f; VkPipelineMultisampleStateCreateInfo pipelineMSCreateInfo = {}; pipelineMSCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; VkPipelineColorBlendAttachmentState blendAttachState = {}; blendAttachState.colorWriteMask = 0xf; blendAttachState.blendEnable = false; VkPipelineColorBlendStateCreateInfo blendCreateInfo = {}; blendCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO; blendCreateInfo.attachmentCount = 1; blendCreateInfo.pAttachments = &blendAttachState; VkPipelineDepthStencilStateCreateInfo depthStencilState = {}; depthStencilState.depthTestEnable = false; depthStencilState.stencilTestEnable = false; VkGraphicsPipelineCreateInfo pipelineInfo = {}; pipelineInfo.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO; pipelineInfo.stageCount = 2; pipelineInfo.pStages = &shaderStageCreateInfo[0]; pipelineInfo.pVertexInputState = &vertexInputInfo; pipelineInfo.pInputAssemblyState = &pipelineIACreateInfo; pipelineInfo.pViewportState = &vpCreateInfo; pipelineInfo.pRasterizationState = &rastCreateInfo; pipelineInfo.pMultisampleState = &pipelineMSCreateInfo; pipelineInfo.pColorBlendState = &blendCreateInfo; pipelineInfo.renderPass = renderPass; pipelineInfo.basePipelineIndex = -1; pipelineInfo.pDepthStencilState = &depthStencilState; pipelineInfo.layout = pipelineLayout; VkResult res = vkCreateGraphicsPipelines(device, VK_NULL_HANDLE, 1, &pipelineInfo, NULL, &pipeline); printf("Graphics pipeline created\n"); } uint32_t getMemoryTypeIndex(VkPhysicalDeviceMemoryProperties deviceMemoryProperties, uint32_t typeBits, VkMemoryPropertyFlags properties) { // Iterate over all memory types available for the device used in this example for (uint32_t i = 0; i < deviceMemoryProperties.memoryTypeCount; i++) { if ((typeBits & 1) == 1) { if ((deviceMemoryProperties.memoryTypes[i].propertyFlags & properties) == properties) { return i; } } typeBits >>= 1; } assert(0); } void CreateVertexBuffer() { unsigned vboSize = sizeof(float) * (2 * 3 + 3); //3 x vec2 + 3 VkMemoryRequirements mr; { //create staging buffer VkBufferCreateInfo ci = {}; ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; ci.size = vboSize; ci.usage = VK_BUFFER_USAGE_VERTEX_BUFFER_BIT; VkResult res = vkCreateBuffer(device, &ci, 0, &vertexBuffer); vkGetBufferMemoryRequirements(device, vertexBuffer, &mr); VkMemoryAllocateInfo mai = {}; mai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO; mai.allocationSize = mr.size; mai.memoryTypeIndex = getMemoryTypeIndex(pdmp, mr.memoryTypeBits, VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT); res = vkAllocateMemory(device, &mai, 0, &vertexBufferMemory); uint32_t vertices[9]; ((float*)vertices)[0] = -1; ((float*)vertices)[1] = 1; ((float*)vertices)[2] = 1; ((float*)vertices)[3] = 1; ((float*)vertices)[4] = 0; ((float*)vertices)[5] = -1; ((uint8_t*)vertices)[6*4 + 0] = 255; ((uint8_t*)vertices)[6*4 + 1] = 0; ((uint8_t*)vertices)[6*4 + 2] = 0; ((uint8_t*)vertices)[6*4 + 3] = 0; ((uint8_t*)vertices)[7*4 + 0] = 0; ((uint8_t*)vertices)[7*4 + 1] = 255; ((uint8_t*)vertices)[7*4 + 2] = 0; ((uint8_t*)vertices)[7*4 + 3] = 0; ((uint8_t*)vertices)[8*4 + 0] = 0; ((uint8_t*)vertices)[8*4 + 1] = 0; ((uint8_t*)vertices)[8*4 + 2] = 255; ((uint8_t*)vertices)[8*4 + 3] = 0; void* data; res = vkMapMemory(device, vertexBufferMemory, 0, mr.size, 0, &data); memcpy(data, vertices, vboSize); vkUnmapMemory(device, vertexBufferMemory); res = vkBindBufferMemory(device, vertexBuffer, vertexBufferMemory, 0); } printf("Vertex buffer created\n"); } int main() { // Note: dynamically loading loader may be a better idea to fail gracefully when Vulkan is not supported // Create window for Vulkan //glfwInit(); //glfwWindowHint(GLFW_CLIENT_API, GLFW_NO_API); //glfwWindowHint(GLFW_RESIZABLE, GLFW_FALSE); //window = glfwCreateWindow(WINDOW_WIDTH, WINDOW_HEIGHT, "The 630 line cornflower blue window", nullptr, nullptr); // Use Vulkan setupVulkan(); mainLoop(); cleanup(); return 0; }