/*
 * Copyright (c) 2015-2022 The Khronos Group Inc.
 * Copyright (c) 2015-2022 Valve Corporation
 * Copyright (c) 2015-2022 LunarG, Inc.
 * Copyright (c) 2022 NVIDIA CORPORATION.  All rights reserved.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Author: Rodrigo Locatti <rlocatti@nvidia.com>
 */

#include <chrono>
#include <thread>

#include "cast_utils.h"
#include "layer_validation_tests.h"

// Tests for NVIDIA-specific best practices
const char *kEnableNVIDIAValidation = "VALIDATION_CHECK_ENABLE_VENDOR_SPECIFIC_NVIDIA";

static constexpr float defaultQueuePriority = 0.0f;

TEST_F(VkNvidiaBestPracticesLayerTest, PageableDeviceLocalMemory) {
    AddRequiredExtensions(VK_EXT_PAGEABLE_DEVICE_LOCAL_MEMORY_EXTENSION_NAME);
    InitBestPracticesFramework(kEnableNVIDIAValidation);

    if (!AreRequiredExtensionsEnabled()) {
        GTEST_SKIP() << RequiredExtensionsNotSupported() << " not supported";
    }

    VkDeviceQueueCreateInfo queue_ci = LvlInitStruct<VkDeviceQueueCreateInfo>();
    queue_ci.queueCount = 1;
    queue_ci.pQueuePriorities = &defaultQueuePriority;

    VkDeviceCreateInfo device_ci = LvlInitStruct<VkDeviceCreateInfo>();
    device_ci.queueCreateInfoCount = 1;
    device_ci.pQueueCreateInfos = &queue_ci;

    VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT pageable = LvlInitStruct<VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT>();
    pageable.pageableDeviceLocalMemory = VK_TRUE;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreateDevice-PageableDeviceLocalMemory");
        VkDevice test_device = VK_NULL_HANDLE;
        VkResult err = vk::CreateDevice(gpu(), &device_ci, nullptr, &test_device);
        m_errorMonitor->VerifyFound();
        if (err == VK_SUCCESS) {
            vk::DestroyDevice(test_device, nullptr);
        }
    }

    // Now enable the expected features
    device_ci.enabledExtensionCount = m_device_extension_names.size();
    device_ci.ppEnabledExtensionNames = m_device_extension_names.data();
    device_ci.pNext = &pageable;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreateDevice-PageableDeviceLocalMemory");
        vk_testing::Device test_device(gpu());
        test_device.init(device_ci);
        m_errorMonitor->Finish();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, TilingLinear) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    VkImageCreateInfo image_ci = LvlInitStruct<VkImageCreateInfo>();
    image_ci.imageType = VK_IMAGE_TYPE_2D;
    image_ci.format = VK_FORMAT_A8B8G8R8_UNORM_PACK32;
    image_ci.extent = { 512, 512, 1 };
    image_ci.mipLevels = 1;
    image_ci.arrayLayers = 1;
    image_ci.samples = VK_SAMPLE_COUNT_1_BIT;
    image_ci.tiling = VK_IMAGE_TILING_OPTIMAL;
    image_ci.usage = VK_IMAGE_USAGE_SAMPLED_BIT;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreateImage-TilingLinear");
        vk_testing::Image image(*m_device, image_ci, vk_testing::no_mem);
        m_errorMonitor->Finish();
    }

    image_ci.tiling = VK_IMAGE_TILING_LINEAR;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreateImage-TilingLinear");
        vk_testing::Image image(*m_device, image_ci, vk_testing::no_mem);
        m_errorMonitor->VerifyFound();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, Depth32Format) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    VkImageCreateInfo image_ci = LvlInitStruct<VkImageCreateInfo>();
    image_ci.imageType = VK_IMAGE_TYPE_2D;
    // This should be VK_FORMAT_D24_UNORM_S8_UINT, but that's not a required format.
    image_ci.format = VK_FORMAT_A8B8G8R8_UNORM_PACK32;
    image_ci.extent = { 512, 512, 1 };
    image_ci.mipLevels = 1;
    image_ci.arrayLayers = 1;
    image_ci.samples = VK_SAMPLE_COUNT_1_BIT;
    image_ci.tiling = VK_IMAGE_TILING_OPTIMAL;
    image_ci.usage = VK_IMAGE_USAGE_SAMPLED_BIT;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreateImage-Depth32Format");
        vk_testing::Image image(*m_device, image_ci, vk_testing::no_mem);
        m_errorMonitor->Finish();
    }

    VkFormat formats[] = { VK_FORMAT_D32_SFLOAT, VK_FORMAT_D32_SFLOAT_S8_UINT };

    for (VkFormat format : formats) {
        image_ci.format = format;

        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreateImage-Depth32Format");
        vk_testing::Image image(*m_device, image_ci, vk_testing::no_mem);
        m_errorMonitor->VerifyFound();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, QueueBindSparse_NotAsync) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    if (!m_device->phy().features().sparseBinding) {
        GTEST_SKIP() << "Test requires sparseBinding";
    }

    VkDeviceQueueCreateInfo general_queue_ci = LvlInitStruct<VkDeviceQueueCreateInfo>();
    general_queue_ci.queueFamilyIndex = m_device->QueueFamilyMatching(VK_QUEUE_GRAPHICS_BIT | VK_QUEUE_SPARSE_BINDING_BIT, 0);
    general_queue_ci.queueCount = 1;
    general_queue_ci.pQueuePriorities = &defaultQueuePriority;

    VkDeviceQueueCreateInfo transfer_queue_ci = LvlInitStruct<VkDeviceQueueCreateInfo>();
    transfer_queue_ci.queueFamilyIndex = m_device->QueueFamilyMatching(VK_QUEUE_TRANSFER_BIT | VK_QUEUE_SPARSE_BINDING_BIT,
                                                                       VK_QUEUE_COMPUTE_BIT | VK_QUEUE_GRAPHICS_BIT);
    transfer_queue_ci.queueCount = 1;
    transfer_queue_ci.pQueuePriorities = &defaultQueuePriority;

    if (general_queue_ci.queueFamilyIndex == UINT32_MAX || transfer_queue_ci.queueFamilyIndex == UINT32_MAX) {
        GTEST_SKIP() << "Test requires a general and a transfer queue";
    }

    VkDeviceQueueCreateInfo queue_cis[2] = {
        general_queue_ci,
        transfer_queue_ci,
    };
    uint32_t queue_family_indices[] = {
        general_queue_ci.queueFamilyIndex,
        transfer_queue_ci.queueFamilyIndex,
    };

    VkPhysicalDeviceFeatures features = {};
    features.sparseBinding = VK_TRUE;

    VkDeviceCreateInfo device_ci = LvlInitStruct<VkDeviceCreateInfo>();
    device_ci.queueCreateInfoCount = 2;
    device_ci.pQueueCreateInfos = queue_cis;
    device_ci.pEnabledFeatures = &features;

    vk_testing::Device test_device(gpu());
    test_device.init(device_ci);

    VkQueue graphics_queue = VK_NULL_HANDLE;
    VkQueue transfer_queue = VK_NULL_HANDLE;
    vk::GetDeviceQueue(test_device.handle(), general_queue_ci.queueFamilyIndex, 0, &graphics_queue);
    vk::GetDeviceQueue(test_device.handle(), transfer_queue_ci.queueFamilyIndex, 0, &transfer_queue);

    VkBufferCreateInfo sparse_buffer_ci = LvlInitStruct<VkBufferCreateInfo>();
    sparse_buffer_ci.flags = VK_BUFFER_CREATE_SPARSE_BINDING_BIT;
    sparse_buffer_ci.size = 0x10000;
    sparse_buffer_ci.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT;
    sparse_buffer_ci.sharingMode = VK_SHARING_MODE_CONCURRENT;
    sparse_buffer_ci.queueFamilyIndexCount = 2;
    sparse_buffer_ci.pQueueFamilyIndices = queue_family_indices;

    vk_testing::Buffer sparse_buffer(test_device, sparse_buffer_ci, vk_testing::no_mem);

    const VkMemoryRequirements memory_requirements = sparse_buffer.memory_requirements();
    ASSERT_NE(memory_requirements.memoryTypeBits, 0);

    // Find first valid bit, whatever it is
    uint32_t memory_type_index = 0;
    while (((memory_requirements.memoryTypeBits >> memory_type_index) & 1) == 0) {
        ++memory_type_index;
    }

    VkMemoryAllocateInfo memory_ai = LvlInitStruct<VkMemoryAllocateInfo>();
    memory_ai.allocationSize = memory_requirements.size;
    memory_ai.memoryTypeIndex = memory_type_index;

    vk_testing::DeviceMemory sparse_memory(test_device, memory_ai);

    VkSparseMemoryBind bind = {};
    bind.resourceOffset = 0;
    bind.size = sparse_buffer_ci.size;
    bind.memory = sparse_memory.handle();
    bind.memoryOffset = 0;
    bind.flags = 0;

    VkSparseBufferMemoryBindInfo sparse_buffer_bind = {};
    sparse_buffer_bind.buffer = sparse_buffer.handle();
    sparse_buffer_bind.bindCount = 1;
    sparse_buffer_bind.pBinds = &bind;

    VkBindSparseInfo bind_info = LvlInitStruct<VkBindSparseInfo>();
    bind_info.bufferBindCount = 1;
    bind_info.pBufferBinds = &sparse_buffer_bind;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-QueueBindSparse-NotAsync");
        vk::QueueBindSparse(transfer_queue, 1, &bind_info, VK_NULL_HANDLE);
        m_errorMonitor->Finish();
    }

    test_device.wait();

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-QueueBindSparse-NotAsync");
        vk::QueueBindSparse(graphics_queue, 1, &bind_info, VK_NULL_HANDLE);
        m_errorMonitor->VerifyFound();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, AccelerationStructure_NotAsync) {
    SetTargetApiVersion(VK_API_VERSION_1_1);
    AddRequiredExtensions(VK_KHR_ACCELERATION_STRUCTURE_EXTENSION_NAME);
    AddRequiredExtensions(VK_KHR_BUFFER_DEVICE_ADDRESS_EXTENSION_NAME);
    InitBestPracticesFramework(kEnableNVIDIAValidation);

    if (IsDriver(VK_DRIVER_ID_AMD_PROPRIETARY)) {
        GTEST_SKIP() << "Test is crashing on AMD hardware for unknown reasons.";
    }
    if (DeviceValidationVersion() < VK_API_VERSION_1_1) {
        GTEST_SKIP() << "Vulkan >= 1.1 required";
    }
    if (!AreRequiredExtensionsEnabled()) {
        GTEST_SKIP() << RequiredExtensionsNotSupported() << " not supported";
    }
    InitState();

    VkDeviceQueueCreateInfo general_queue_ci = LvlInitStruct<VkDeviceQueueCreateInfo>();
    general_queue_ci.queueFamilyIndex = m_device->QueueFamilyMatching(VK_QUEUE_GRAPHICS_BIT, 0);
    general_queue_ci.queueCount = 1;
    general_queue_ci.pQueuePriorities = &defaultQueuePriority;

    VkDeviceQueueCreateInfo compute_queue_ci = LvlInitStruct<VkDeviceQueueCreateInfo>();
    compute_queue_ci.queueFamilyIndex = m_device->QueueFamilyMatching(VK_QUEUE_COMPUTE_BIT, VK_QUEUE_GRAPHICS_BIT);
    compute_queue_ci.queueCount = 1;
    compute_queue_ci.pQueuePriorities = &defaultQueuePriority;

    if (general_queue_ci.queueFamilyIndex == UINT32_MAX || compute_queue_ci.queueFamilyIndex == UINT32_MAX) {
        // There's no asynchronous compute queue, skip.
        return;
    }

    VkDeviceQueueCreateInfo queue_cis[2] = {
        general_queue_ci,
        compute_queue_ci,
    };
    uint32_t queue_family_indices[] = {
        general_queue_ci.queueFamilyIndex,
        compute_queue_ci.queueFamilyIndex,
    };

    VkPhysicalDeviceAccelerationStructureFeaturesKHR acceleration_structures_features = LvlInitStruct<VkPhysicalDeviceAccelerationStructureFeaturesKHR>();
    acceleration_structures_features.accelerationStructure = VK_TRUE;

    VkPhysicalDeviceBufferDeviceAddressFeaturesEXT buffer_device_address_features = LvlInitStruct<VkPhysicalDeviceBufferDeviceAddressFeaturesEXT>();
    buffer_device_address_features.pNext = &acceleration_structures_features;
    buffer_device_address_features.bufferDeviceAddress = VK_TRUE;

    VkDeviceCreateInfo device_ci = LvlInitStruct<VkDeviceCreateInfo>();
    device_ci.pNext = &buffer_device_address_features;
    device_ci.queueCreateInfoCount = 2;
    device_ci.pQueueCreateInfos = queue_cis;
    device_ci.enabledExtensionCount = m_device_extension_names.size();
    device_ci.ppEnabledExtensionNames = m_device_extension_names.data();

    vk_testing::Device test_device(gpu());
    test_device.init(device_ci);

    VkQueue graphics_queue = VK_NULL_HANDLE;
    VkQueue transfer_queue = VK_NULL_HANDLE;
    vk::GetDeviceQueue(test_device.handle(), general_queue_ci.queueFamilyIndex, 0, &graphics_queue);
    vk::GetDeviceQueue(test_device.handle(), compute_queue_ci.queueFamilyIndex, 0, &transfer_queue);

    VkBufferCreateInfo acceleration_structure_buffer_ci = LvlInitStruct<VkBufferCreateInfo>();
    acceleration_structure_buffer_ci.size = 0x10000;
    acceleration_structure_buffer_ci.usage = VK_BUFFER_USAGE_ACCELERATION_STRUCTURE_STORAGE_BIT_KHR;

    VkBufferCreateInfo scratch_buffer_ci = LvlInitStruct<VkBufferCreateInfo>();
    scratch_buffer_ci.size = 1ULL << 20;
    scratch_buffer_ci.usage = VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;

    vk_testing::Buffer acceleration_structure_buffer(test_device, acceleration_structure_buffer_ci);
    vk_testing::Buffer scratch_buffer(test_device, scratch_buffer_ci);

    auto vkGetBufferDeviceAddressKHR =
        (PFN_vkGetBufferDeviceAddress)vk::GetDeviceProcAddr(m_device->device(), "vkGetBufferDeviceAddressKHR");
    ASSERT_TRUE(vkGetBufferDeviceAddressKHR != nullptr);

    VkBufferDeviceAddressInfo scratch_buffer_device_address_info = LvlInitStruct<VkBufferDeviceAddressInfo>();
    scratch_buffer_device_address_info.buffer = scratch_buffer.handle();
    VkDeviceAddress scratch_address = vkGetBufferDeviceAddressKHR(test_device.handle(), &scratch_buffer_device_address_info);

    VkAccelerationStructureCreateInfoKHR acceleration_structure_ci = LvlInitStruct<VkAccelerationStructureCreateInfoKHR>();
    acceleration_structure_ci.buffer = acceleration_structure_buffer.handle();
    acceleration_structure_ci.offset = 0;
    acceleration_structure_ci.size = acceleration_structure_buffer_ci.size;
    acceleration_structure_ci.type = VK_ACCELERATION_STRUCTURE_TYPE_BOTTOM_LEVEL_KHR;
    acceleration_structure_ci.deviceAddress = 0;

    vk_testing::AccelerationStructureKHR acceleration_structure(test_device, acceleration_structure_ci);

    VkAccelerationStructureGeometryTrianglesDataKHR triangle_data = LvlInitStruct<VkAccelerationStructureGeometryTrianglesDataKHR>();
    triangle_data.vertexFormat = VK_FORMAT_A8B8G8R8_UNORM_PACK32;
    triangle_data.vertexData.deviceAddress = 0;
    triangle_data.vertexStride = 0;
    triangle_data.maxVertex = 3;
    triangle_data.indexType = VK_INDEX_TYPE_UINT16;
    triangle_data.indexData.deviceAddress = 0;
    triangle_data.transformData.deviceAddress = 0;

    VkAccelerationStructureGeometryKHR geometry = LvlInitStruct<VkAccelerationStructureGeometryKHR>();
    geometry.geometryType = VK_GEOMETRY_TYPE_TRIANGLES_KHR;
    geometry.geometry.triangles = triangle_data;
    geometry.flags = 0;

    VkAccelerationStructureBuildGeometryInfoKHR geometry_info = LvlInitStruct<VkAccelerationStructureBuildGeometryInfoKHR>();
    geometry_info.type = VK_ACCELERATION_STRUCTURE_TYPE_BOTTOM_LEVEL_KHR;
    geometry_info.flags = 0;
    geometry_info.mode = VK_BUILD_ACCELERATION_STRUCTURE_MODE_BUILD_KHR;
    geometry_info.srcAccelerationStructure = VK_NULL_HANDLE;
    geometry_info.dstAccelerationStructure = acceleration_structure.handle();
    geometry_info.geometryCount = 1;
    geometry_info.pGeometries = &geometry;
    geometry_info.ppGeometries = nullptr;
    geometry_info.scratchData.deviceAddress = scratch_address;

    VkAccelerationStructureBuildRangeInfoKHR build_range_info = {};
    build_range_info.primitiveCount = 1;
    build_range_info.primitiveOffset = 0;
    build_range_info.firstVertex = 0;
    build_range_info.transformOffset = 0;

    VkAccelerationStructureBuildRangeInfoKHR *p_build_range_info = &build_range_info;

    auto vkCmdBuildAccelerationStructuresKHR = reinterpret_cast<PFN_vkCmdBuildAccelerationStructuresKHR>(
        vk::GetDeviceProcAddr(test_device.handle(), "vkCmdBuildAccelerationStructuresKHR"));
    ASSERT_NE(vkCmdBuildAccelerationStructuresKHR, nullptr);

    for (uint32_t queue_family_index : queue_family_indices) {
        VkCommandPoolCreateInfo command_pool_ci = LvlInitStruct<VkCommandPoolCreateInfo>();
        command_pool_ci.queueFamilyIndex = queue_family_index;
        vk_testing::CommandPool command_pool(test_device, command_pool_ci);

        VkCommandBufferAllocateInfo command_buffer_ai = LvlInitStruct<VkCommandBufferAllocateInfo>();
        command_buffer_ai.commandPool = command_pool.handle();
        command_buffer_ai.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
        command_buffer_ai.commandBufferCount = 1;

        vk_testing::CommandBuffer command_buffer(test_device, command_buffer_ai);
        VkCommandBufferBeginInfo command_buffer_bi = LvlInitStruct<VkCommandBufferBeginInfo>();
        command_buffer_bi.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
        command_buffer.begin(&command_buffer_bi);

        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-AccelerationStructure-NotAsync");
        vkCmdBuildAccelerationStructuresKHR(command_buffer.handle(), 1, &geometry_info, &p_build_range_info);

        if (queue_family_index == compute_queue_ci.queueFamilyIndex) {
            m_errorMonitor->Finish();
        } else {
            m_errorMonitor->VerifyFound();
        }

        command_buffer.end();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, AllocateMemory_SetPriority) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    VkMemoryAllocateInfo memory_ai = LvlInitStruct<VkMemoryAllocateInfo>();
    memory_ai.allocationSize = 0x100000;
    memory_ai.memoryTypeIndex = 0;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-AllocateMemory-SetPriority");
        vk_testing::DeviceMemory memory(*m_device, memory_ai);
        m_errorMonitor->VerifyFound();
    }

    VkMemoryPriorityAllocateInfoEXT priority = LvlInitStruct<VkMemoryPriorityAllocateInfoEXT>();
    priority.priority = 0.5f;
    memory_ai.pNext = &priority;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-AllocateMemory-SetPriority");
        vk_testing::DeviceMemory memory(*m_device, memory_ai);
        m_errorMonitor->Finish();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, AllocateMemory_ReuseAllocations) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    VkMemoryAllocateInfo memory_ai = LvlInitStruct<VkMemoryAllocateInfo>();
    memory_ai.allocationSize = 0x100000;
    memory_ai.memoryTypeIndex = 0;

    VkMemoryPriorityAllocateInfoEXT priority = LvlInitStruct<VkMemoryPriorityAllocateInfoEXT>();
    priority.priority = 0.5f;
    memory_ai.pNext = &priority;

    m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                         "UNASSIGNED-BestPractices-AllocateMemory-ReuseAllocations");
    {
        vk_testing::DeviceMemory memory(*m_device, memory_ai);
    }

    std::this_thread::sleep_for(std::chrono::seconds{6});

    {
        vk_testing::DeviceMemory memory(*m_device, memory_ai);
    }

    m_errorMonitor->Finish();

    m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                         "UNASSIGNED-BestPractices-AllocateMemory-ReuseAllocations");

    {
        vk_testing::DeviceMemory memory(*m_device, memory_ai);
    }

    m_errorMonitor->VerifyFound();
}

TEST_F(VkNvidiaBestPracticesLayerTest, BindMemory_NoPriority) {
    SetTargetApiVersion(VK_API_VERSION_1_1);
    AddRequiredExtensions(VK_EXT_PAGEABLE_DEVICE_LOCAL_MEMORY_EXTENSION_NAME);
    AddRequiredExtensions(VK_KHR_MAINTENANCE_4_EXTENSION_NAME);
    InitBestPracticesFramework(kEnableNVIDIAValidation);

    if (DeviceValidationVersion() < VK_API_VERSION_1_1) {
        GTEST_SKIP() << "Vulkan >= 1.1 required";
    }
    if (!AreRequiredExtensionsEnabled()) {
        GTEST_SKIP() << RequiredExtensionsNotSupported() << " not supported";
    }
    InitState();

    VkDeviceQueueCreateInfo queue_ci = LvlInitStruct<VkDeviceQueueCreateInfo>();
    queue_ci.queueFamilyIndex = 0;
    queue_ci.queueCount = 1;
    queue_ci.pQueuePriorities = &defaultQueuePriority;

    VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT pageable_features = LvlInitStruct<VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT>();
    pageable_features.pNext = nullptr;
    pageable_features.pageableDeviceLocalMemory = VK_TRUE;

    VkPhysicalDeviceMaintenance4Features maintenance4_features = LvlInitStruct<VkPhysicalDeviceMaintenance4Features>();
    maintenance4_features.pNext = &pageable_features;
    maintenance4_features.maintenance4 = VK_TRUE;

    VkDeviceCreateInfo device_ci = LvlInitStruct<VkDeviceCreateInfo>();
    device_ci.pNext = &maintenance4_features;
    device_ci.queueCreateInfoCount = 1;
    device_ci.pQueueCreateInfos = &queue_ci;
    device_ci.enabledExtensionCount = m_device_extension_names.size();
    device_ci.ppEnabledExtensionNames = m_device_extension_names.data();

    vk_testing::Device test_device(gpu());
    test_device.init(device_ci);

    VkBufferCreateInfo buffer_ci = LvlInitStruct<VkBufferCreateInfo>();
    buffer_ci.flags = 0;
    buffer_ci.size = 0x100000;
    buffer_ci.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT;
    buffer_ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
    buffer_ci.queueFamilyIndexCount = 0;
    buffer_ci.pQueueFamilyIndices = nullptr;

    vk_testing::Buffer buffer_a(test_device, buffer_ci, vk_testing::no_mem);
    vk_testing::Buffer buffer_b(test_device, buffer_ci, vk_testing::no_mem);

    const VkMemoryRequirements memory_requirements = buffer_a.memory_requirements();
    ASSERT_NE(memory_requirements.memoryTypeBits, 0);

    // Find first valid bit, whatever it is
    uint32_t memory_type_index = 0;
    while (((memory_requirements.memoryTypeBits >> memory_type_index) & 1) == 0) {
        ++memory_type_index;
    }

    VkMemoryAllocateInfo memory_ai = LvlInitStruct<VkMemoryAllocateInfo>();
    memory_ai.allocationSize = memory_requirements.size;
    memory_ai.memoryTypeIndex = memory_type_index;

    vk_testing::DeviceMemory memory(test_device, memory_ai);

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-BindMemory-NoPriority");
        vk::BindBufferMemory(test_device.handle(), buffer_a.handle(), memory.handle(), 0);
        m_errorMonitor->VerifyFound();
    }

    auto vkSetDeviceMemoryPriorityEXT = reinterpret_cast<PFN_vkSetDeviceMemoryPriorityEXT>(
        vk::GetDeviceProcAddr(test_device.handle(), "vkSetDeviceMemoryPriorityEXT"));
    vkSetDeviceMemoryPriorityEXT(test_device.handle(), memory.handle(), 0.5f);

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-BindMemory-NoPriority");
        vk::BindBufferMemory(test_device.handle(), buffer_b.handle(), memory.handle(), 0);
        m_errorMonitor->Finish();
    }
}

static VkDescriptorSetLayoutBinding CreateSingleDescriptorBinding(VkDescriptorType type, uint32_t binding) {
    VkDescriptorSetLayoutBinding layout_binding = {};
    layout_binding.binding = binding;
    layout_binding.descriptorType = type;
    layout_binding.descriptorCount = 1;
    layout_binding.stageFlags = VK_SHADER_STAGE_VERTEX_BIT;
    layout_binding.pImmutableSamplers = nullptr;
    return layout_binding;
}

TEST_F(VkNvidiaBestPracticesLayerTest, CreatePipelineLayout_SeparateSampler) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    VkDescriptorSetLayoutBinding separate_bindings[] = {
        CreateSingleDescriptorBinding(VK_DESCRIPTOR_TYPE_SAMPLER, 0),
        CreateSingleDescriptorBinding(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, 1),
    };
    VkDescriptorSetLayoutBinding combined_bindings[] = {
        CreateSingleDescriptorBinding(VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, 0),
    };

    VkDescriptorSetLayoutCreateInfo separate_set_layout_ci = LvlInitStruct<VkDescriptorSetLayoutCreateInfo>();
    separate_set_layout_ci.flags = 0;
    separate_set_layout_ci.bindingCount = sizeof(separate_bindings) / sizeof(separate_bindings[0]);
    separate_set_layout_ci.pBindings = separate_bindings;

    VkDescriptorSetLayoutCreateInfo combined_set_layout_ci = LvlInitStruct<VkDescriptorSetLayoutCreateInfo>();
    combined_set_layout_ci.flags = 0;
    combined_set_layout_ci.bindingCount = sizeof(combined_bindings) / sizeof(combined_bindings[0]);
    combined_set_layout_ci.pBindings = combined_bindings;

    vk_testing::DescriptorSetLayout separate_set_layout(*m_device, separate_set_layout_ci);
    vk_testing::DescriptorSetLayout combined_set_layout(*m_device, combined_set_layout_ci);

    VkPipelineLayoutCreateInfo pipeline_layout_ci = LvlInitStruct<VkPipelineLayoutCreateInfo>();
    pipeline_layout_ci.flags = 0;
    pipeline_layout_ci.pushConstantRangeCount = 0;
    pipeline_layout_ci.pPushConstantRanges = nullptr;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreatePipelineLayout-SeparateSampler");
        vk_testing::PipelineLayout pipeline_layout(*m_device, pipeline_layout_ci, {&separate_set_layout});
        m_errorMonitor->VerifyFound();
    }

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreatePipelineLayout-SeparateSampler");
        vk_testing::PipelineLayout pipeline_layout(*m_device, pipeline_layout_ci, {&combined_set_layout});
        m_errorMonitor->Finish();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, CreatePipelineLayout_LargePipelineLayout) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    VkDescriptorSetLayoutBinding large_bindings[] = {
        { 0, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, 16, VK_SHADER_STAGE_VERTEX_BIT, nullptr },
        { 1, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, 1, VK_SHADER_STAGE_VERTEX_BIT, nullptr },
    };
    VkDescriptorSetLayoutBinding small_bindings[] = {
        { 0, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, 16, VK_SHADER_STAGE_VERTEX_BIT, nullptr },
    };

    VkDescriptorSetLayoutCreateInfo large_set_layout_ci = LvlInitStruct<VkDescriptorSetLayoutCreateInfo>();
    large_set_layout_ci.flags = 0;
    large_set_layout_ci.bindingCount = sizeof(large_bindings) / sizeof(large_bindings[0]);
    large_set_layout_ci.pBindings = large_bindings;

    VkDescriptorSetLayoutCreateInfo small_set_layout_ci = LvlInitStruct<VkDescriptorSetLayoutCreateInfo>();
    small_set_layout_ci.flags = 0;
    small_set_layout_ci.bindingCount = sizeof(small_bindings) / sizeof(small_bindings[0]);
    small_set_layout_ci.pBindings = small_bindings;

    vk_testing::DescriptorSetLayout large_set_layout(*m_device, large_set_layout_ci);
    vk_testing::DescriptorSetLayout small_set_layout(*m_device, small_set_layout_ci);

    VkPipelineLayoutCreateInfo pipeline_layout_ci = LvlInitStruct<VkPipelineLayoutCreateInfo>();
    pipeline_layout_ci.flags = 0;
    pipeline_layout_ci.pushConstantRangeCount = 0;
    pipeline_layout_ci.pPushConstantRanges = nullptr;

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreatePipelineLayout-LargePipelineLayout");
        vk_testing::PipelineLayout pipeline_layout(*m_device, pipeline_layout_ci, {&large_set_layout});
        m_errorMonitor->VerifyFound();
    }

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-CreatePipelineLayout-LargePipelineLayout");
        vk_testing::PipelineLayout pipeline_layout(*m_device, pipeline_layout_ci, {&small_set_layout});
        m_errorMonitor->Finish();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, BindPipeline_SwitchTessGeometryMesh)
{
    SetTargetApiVersion(VK_API_VERSION_1_1);

    AddRequiredExtensions(VK_KHR_DYNAMIC_RENDERING_EXTENSION_NAME);

    ASSERT_NO_FATAL_FAILURE(InitBestPracticesFramework(kEnableNVIDIAValidation));

    if (DeviceValidationVersion() < VK_API_VERSION_1_1) {
        GTEST_SKIP() << "At least Vulkan version 1.1 is required";
    }

    auto dynamic_rendering_features = LvlInitStruct<VkPhysicalDeviceDynamicRenderingFeaturesKHR>();
    auto features2 = GetPhysicalDeviceFeatures2(dynamic_rendering_features);
    if (!dynamic_rendering_features.dynamicRendering) {
        GTEST_SKIP() << "This test requires dynamicRendering";
    }

    ASSERT_NO_FATAL_FAILURE(InitState(nullptr, &features2));

    char const *vsSource = R"glsl(
        #version 450
        void main() {}
    )glsl";

    char const *gsSource = R"glsl(
        #version 450
        layout(triangles) in;
        layout(triangle_strip, max_vertices = 3) out;
        void main() {}
    )glsl";

    VkShaderObj vs(this, vsSource, VK_SHADER_STAGE_VERTEX_BIT);
    VkShaderObj gs(this, gsSource, VK_SHADER_STAGE_GEOMETRY_BIT);

    VkPipelineRenderingCreateInfo pipeline_rendering_info = LvlInitStruct<VkPipelineRenderingCreateInfo>();

    CreatePipelineHelper vsPipe(*this);
    vsPipe.InitInfo();
    vsPipe.shader_stages_ = {vs.GetStageCreateInfo()};
    vsPipe.InitState();
    vsPipe.gp_ci_.pNext = &pipeline_rendering_info;
    vsPipe.CreateGraphicsPipeline();

    CreatePipelineHelper vgsPipe(*this);
    vgsPipe.InitInfo();
    vgsPipe.shader_stages_ = {vs.GetStageCreateInfo(), gs.GetStageCreateInfo()};
    vgsPipe.InitState();
    vgsPipe.gp_ci_.pNext = &pipeline_rendering_info;
    vgsPipe.CreateGraphicsPipeline();

    m_commandBuffer->begin();

    {
        m_errorMonitor->SetAllowedFailureMsg("UNASSIGNED-BestPractices-Pipeline-SortAndBind");
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-BindPipeline-SwitchTessGeometryMesh");
        vk::CmdBindPipeline(m_commandBuffer->handle(), VK_PIPELINE_BIND_POINT_GRAPHICS, vsPipe.pipeline_);
        m_errorMonitor->Finish();
    }
    {
        m_errorMonitor->SetAllowedFailureMsg("UNASSIGNED-BestPractices-Pipeline-SortAndBind");
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-BindPipeline-SwitchTessGeometryMesh");
        for (int i = 0; i < 10; ++i) {
            vk::CmdBindPipeline(m_commandBuffer->handle(), VK_PIPELINE_BIND_POINT_GRAPHICS, vgsPipe.pipeline_);
            vk::CmdBindPipeline(m_commandBuffer->handle(), VK_PIPELINE_BIND_POINT_GRAPHICS, vsPipe.pipeline_);
        }
        m_errorMonitor->VerifyFound();
    }
}

TEST_F(VkNvidiaBestPracticesLayerTest, BindPipeline_ZcullDirection)
{
    SetTargetApiVersion(VK_API_VERSION_1_3);

    ASSERT_NO_FATAL_FAILURE(InitBestPracticesFramework(kEnableNVIDIAValidation));

    if (DeviceValidationVersion() < VK_API_VERSION_1_3) {
        GTEST_SKIP() << "At least Vulkan version 1.3 is required";
    }

    VkPhysicalDeviceDynamicRenderingFeaturesKHR dynamic_rendering_features = LvlInitStruct<VkPhysicalDeviceDynamicRenderingFeaturesKHR>();
    VkPhysicalDeviceSynchronization2Features synchronization2_features = LvlInitStruct<VkPhysicalDeviceSynchronization2Features>(&dynamic_rendering_features);
    VkPhysicalDeviceFeatures2 features2 = GetPhysicalDeviceFeatures2(synchronization2_features);
    if (!dynamic_rendering_features.dynamicRendering) {
        GTEST_SKIP() << "This test requires dynamicRendering";
    }
    if (!synchronization2_features.synchronization2) {
        GTEST_SKIP() << "This test requires synchronization2";
    }

    ASSERT_NO_FATAL_FAILURE(InitState(nullptr, &features2));

    VkFormat depth_format = VK_FORMAT_D32_SFLOAT_S8_UINT;
    VkPipelineRenderingCreateInfo pipeline_rendering_info = LvlInitStruct<VkPipelineRenderingCreateInfo>();
    pipeline_rendering_info.depthAttachmentFormat = depth_format;
    pipeline_rendering_info.stencilAttachmentFormat = depth_format;

    VkImageObj image(m_device);
    // 3 array layers
    image.Init(image.ImageCreateInfo2D(32, 32, 1, 3, depth_format,
                                       VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT,
                                       VK_IMAGE_TILING_OPTIMAL, 0));
    ASSERT_TRUE(image.initialized());

    VkImageViewCreateInfo image_view_ci = LvlInitStruct<VkImageViewCreateInfo>();
    image_view_ci.image = image.handle();
    image_view_ci.viewType = VK_IMAGE_VIEW_TYPE_2D;
    image_view_ci.format = depth_format;
    // rendering to layer index 1
    image_view_ci.subresourceRange = {VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT, 0, 1, 1, 1};

    vk_testing::ImageView depth_image_view(*m_device, image_view_ci);

    VkRenderingAttachmentInfo depth_attachment = LvlInitStruct<VkRenderingAttachmentInfo>();
    depth_attachment.imageLayout = VK_IMAGE_LAYOUT_GENERAL;
    depth_attachment.imageView = depth_image_view.handle();
    depth_attachment.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;

    VkRenderingInfo begin_rendering_info = LvlInitStruct<VkRenderingInfo>();
    begin_rendering_info.renderArea.extent = {32, 32};
    begin_rendering_info.layerCount = 1;
    begin_rendering_info.pDepthAttachment = &depth_attachment;
    begin_rendering_info.pStencilAttachment = &depth_attachment;

    VkClearRect clear_rect{};
    clear_rect.baseArrayLayer = 0;
    clear_rect.layerCount = 1;
    clear_rect.rect.extent.width = 32;
    clear_rect.rect.extent.height = 32;

    VkClearAttachment attachment{};
    attachment.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;

    VkImageMemoryBarrier discard_barrier = LvlInitStruct<VkImageMemoryBarrier>();
    discard_barrier.srcAccessMask = VK_ACCESS_MEMORY_READ_BIT;
    discard_barrier.dstAccessMask = VK_ACCESS_MEMORY_WRITE_BIT | VK_ACCESS_MEMORY_READ_BIT;
    discard_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
    discard_barrier.newLayout = VK_IMAGE_LAYOUT_GENERAL;
    discard_barrier.image = image.handle();
    discard_barrier.subresourceRange = {VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT, 0, 1, 1,
                                        VK_REMAINING_ARRAY_LAYERS};

    VkImageMemoryBarrier2 discard_barrier2 = LvlInitStruct<VkImageMemoryBarrier2>();
    discard_barrier2.srcAccessMask = VK_ACCESS_2_MEMORY_READ_BIT;
    discard_barrier2.dstAccessMask = VK_ACCESS_2_MEMORY_WRITE_BIT | VK_ACCESS_MEMORY_READ_BIT;
    discard_barrier2.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
    discard_barrier2.newLayout = VK_IMAGE_LAYOUT_GENERAL;
    discard_barrier2.image = image.handle();
    discard_barrier2.subresourceRange = {VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT, 0, 1, 1,
                                         VK_REMAINING_ARRAY_LAYERS};

    VkDependencyInfo discard_dependency_info = LvlInitStruct<VkDependencyInfo>();
    discard_dependency_info.imageMemoryBarrierCount = 1;
    discard_dependency_info.pImageMemoryBarriers = &discard_barrier2;

    auto set_desired_failure_msg = [this] {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-Zcull-LessGreaterRatio");
    };

    const char *vsSource = R"glsl(
        #version 450
        void main() {}
    )glsl";

    VkShaderObj vs(this, vsSource, VK_SHADER_STAGE_VERTEX_BIT);

    VkPipelineObj pipe(m_device);
    pipe.AddShader(&vs);
    pipe.AddDefaultColorAttachment();
    pipe.MakeDynamic(VK_DYNAMIC_STATE_DEPTH_TEST_ENABLE);
    pipe.MakeDynamic(VK_DYNAMIC_STATE_DEPTH_COMPARE_OP);

    VkPipelineDepthStencilStateCreateInfo depth_stencil_state_ci = LvlInitStruct<VkPipelineDepthStencilStateCreateInfo>();

    VkPipelineLayoutCreateInfo pipeline_layout_ci = LvlInitStruct<VkPipelineLayoutCreateInfo>();
    vk_testing::PipelineLayout pipeline_layout(*m_device, pipeline_layout_ci);

    VkGraphicsPipelineCreateInfo create_info = LvlInitStruct<VkGraphicsPipelineCreateInfo>();
    pipe.InitGraphicsPipelineCreateInfo(&create_info);
    create_info.pNext = &pipeline_rendering_info;
    create_info.pDepthStencilState = &depth_stencil_state_ci;
    pipe.CreateVKPipeline(pipeline_layout.handle(), VK_NULL_HANDLE, &create_info);

    VkViewport viewport = {};
    viewport.width = 32;
    viewport.height = 32;
    viewport.minDepth = 1.0f;
    viewport.maxDepth = 1.0f;

    VkRect2D scissor = {};
    scissor.extent.width = 32;
    scissor.extent.height = 32;

    VkCommandBufferBeginInfo begin_info = LvlInitStruct<VkCommandBufferBeginInfo>();
    m_commandBuffer->begin(&begin_info);

    auto cmd = m_commandBuffer->handle();

    vk::CmdBindPipeline(cmd, VK_PIPELINE_BIND_POINT_GRAPHICS, pipe.handle());
    vk::CmdSetDepthTestEnable(cmd, VK_TRUE);

    vk::CmdSetViewport(cmd, 0, 1, &viewport);
    vk::CmdSetScissor(cmd, 0, 1, &scissor);

    {
        SCOPED_TRACE("Unbalance");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 90; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 10; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    {
        SCOPED_TRACE("Balance");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->VerifyFound();
    }

    {
        // This should miss because BeginRendering uses LOAD_OP_CLEAR
        SCOPED_TRACE("Balance with end rendering");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    {
        SCOPED_TRACE("Clear before balance");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdClearAttachments(cmd, 1, &attachment, 1, &clear_rect);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    {
        SCOPED_TRACE("Balance before clear");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdClearAttachments(cmd, 1, &attachment, 1, &clear_rect);
        m_errorMonitor->VerifyFound();

        vk::CmdEndRendering(cmd);
    }

    // The tests below use LOAD_OP for depth
    depth_attachment.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;

    {
        SCOPED_TRACE("Load previous scope");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->VerifyFound();
    }

    {
        SCOPED_TRACE("Transition to discard");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        vk::CmdPipelineBarrier(cmd, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, 0, nullptr, 0,
                               nullptr, 1, &discard_barrier);

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    {
        SCOPED_TRACE("Transition to discard 2");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        vk::CmdPipelineBarrier2(cmd, &discard_dependency_info);

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    {
        SCOPED_TRACE("Balance before discard");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        set_desired_failure_msg();
        vk::CmdPipelineBarrier2(cmd, &discard_dependency_info);
        m_errorMonitor->VerifyFound();
    }

    {
        SCOPED_TRACE("Transfer clear to discard");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        VkClearDepthStencilValue ds_value{};
        vk::CmdClearDepthStencilImage(cmd, image.handle(), VK_IMAGE_LAYOUT_GENERAL, &ds_value, 1, &discard_barrier.subresourceRange);

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    {
        SCOPED_TRACE("Transfer clear to validate");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        set_desired_failure_msg();
        VkClearDepthStencilValue ds_value{};
        vk::CmdClearDepthStencilImage(cmd, image.handle(), VK_IMAGE_LAYOUT_GENERAL, &ds_value, 1, &discard_barrier.subresourceRange);
        m_errorMonitor->VerifyFound();
    }

    discard_barrier.subresourceRange.layerCount = VK_REMAINING_ARRAY_LAYERS;

    {
        SCOPED_TRACE("Transfer clear with VK_REMAINING_ARRAY_LAYERS");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        VkClearDepthStencilValue ds_value{};
        vk::CmdClearDepthStencilImage(cmd, image.handle(), VK_IMAGE_LAYOUT_GENERAL, &ds_value, 1,
                                      &discard_barrier.subresourceRange);

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    discard_barrier.subresourceRange.levelCount = VK_REMAINING_MIP_LEVELS;

    {
        SCOPED_TRACE("Transfer clear with VK_REMAINING_MIP_LEVELS");

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_LESS);
        for (int i = 0; i < 60; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        vk::CmdEndRendering(cmd);

        VkClearDepthStencilValue ds_value{};
        vk::CmdClearDepthStencilImage(cmd, image.handle(), VK_IMAGE_LAYOUT_GENERAL, &ds_value, 1,
                                      &discard_barrier.subresourceRange);

        vk::CmdBeginRendering(cmd, &begin_rendering_info);

        vk::CmdSetDepthCompareOp(cmd, VK_COMPARE_OP_GREATER);
        for (int i = 0; i < 40; ++i) m_commandBuffer->Draw(0, 0, 0, 0);

        set_desired_failure_msg();
        vk::CmdEndRendering(cmd);
        m_errorMonitor->Finish();
    }

    m_commandBuffer->end();
}

TEST_F(VkNvidiaBestPracticesLayerTest, ClearColor_NotCompressed)
{
    SetTargetApiVersion(VK_API_VERSION_1_3);

    ASSERT_NO_FATAL_FAILURE(InitBestPracticesFramework(kEnableNVIDIAValidation));

    if (DeviceValidationVersion() < VK_API_VERSION_1_3) {
        GTEST_SKIP() << "At least Vulkan version 1.3 is required";
    }

    VkPhysicalDeviceDynamicRenderingFeaturesKHR dynamic_rendering_features = LvlInitStruct<VkPhysicalDeviceDynamicRenderingFeaturesKHR>();
    VkPhysicalDeviceFeatures2 features2 = GetPhysicalDeviceFeatures2(dynamic_rendering_features);
    if (!dynamic_rendering_features.dynamicRendering) {
        GTEST_SKIP() << "This test requires dynamicRendering";
    }

    ASSERT_NO_FATAL_FAILURE(InitState(nullptr, &features2));

    auto set_desired = [this] {
        m_errorMonitor->Finish();
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, "UNASSIGNED-BestPractices-ClearColor-NotCompressed");
        m_errorMonitor->SetAllowedFailureMsg("UNASSIGNED-BestPractices-DrawState-ClearCmdBeforeDraw");
    };

    VkImageObj image(m_device);
    image.Init((uint32_t)m_width, (uint32_t)m_height, 1, VK_FORMAT_B8G8R8A8_UNORM,
               VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT, VK_IMAGE_TILING_OPTIMAL, 0);
    ASSERT_TRUE(image.initialized());

    VkImageViewCreateInfo image_view_ci = LvlInitStruct<VkImageViewCreateInfo>();
    image_view_ci.image = image.handle();
    image_view_ci.viewType = VK_IMAGE_VIEW_TYPE_2D;
    image_view_ci.format = VK_FORMAT_B8G8R8A8_UNORM;
    image_view_ci.subresourceRange = {VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1};

    vk_testing::ImageView image_view(*m_device, image_view_ci);

    VkRenderingAttachmentInfo color_attachment = LvlInitStruct<VkRenderingAttachmentInfo>();
    color_attachment.imageView = image_view.handle();
    color_attachment.imageLayout = VK_IMAGE_LAYOUT_GENERAL;
    color_attachment.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
    color_attachment.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
    color_attachment.clearValue = {};

    VkRenderingInfo begin_rendering_info = LvlInitStruct<VkRenderingInfo>();
    begin_rendering_info.renderArea.extent = {(uint32_t)m_width, (uint32_t)m_height};
    begin_rendering_info.layerCount = 1;
    begin_rendering_info.colorAttachmentCount = 1;
    begin_rendering_info.pColorAttachments = &color_attachment;

    VkClearAttachment clear{};
    clear.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
    clear.clearValue.color.float32[0] = 0.0f;
    clear.clearValue.color.float32[1] = 0.0f;
    clear.clearValue.color.float32[2] = 0.0f;
    clear.clearValue.color.float32[3] = 0.0f;
    clear.colorAttachment = 0;

    auto set_clear_color = [&clear](const std::array<float, 4>& color) {
        for (size_t i = 0; i < 4; ++i) {
            clear.clearValue.color.float32[i] = color[i];
        }
    };

    VkClearRect clear_rect = {{{0, 0}, {(uint32_t)m_width, (uint32_t)m_height}}, 0, 1};

    m_commandBuffer->begin();
    vk::CmdBeginRendering(m_commandBuffer->handle(), &begin_rendering_info);

    {
        set_desired();
        set_clear_color({1.0f, 0.5f, 0.25f, 0.0f});

        for (int i = 0; i < 16 + 1; ++i) {
            clear.clearValue.color.float32[3] += 0.05f;
            vk::CmdClearAttachments(m_commandBuffer->handle(), 1, &clear, 1, &clear_rect);
        }
        m_errorMonitor->VerifyFound();
    }
    {
        set_desired();
        set_clear_color({1.0f, 1.0f, 1.0f, 1.0f});
        vk::CmdClearAttachments(m_commandBuffer->handle(), 1, &clear, 1, &clear_rect);
        m_errorMonitor->Finish();
    }
    {
        set_desired();
        set_clear_color({0.0f, 0.0f, 0.0f, 0.0f});
        vk::CmdClearAttachments(m_commandBuffer->handle(), 1, &clear, 1, &clear_rect);
        m_errorMonitor->Finish();
    }
    {
        set_desired();
        set_clear_color({0.9f, 1.0f, 1.0f, 1.0f});
        vk::CmdClearAttachments(m_commandBuffer->handle(), 1, &clear, 1, &clear_rect);
        m_errorMonitor->VerifyFound();
    }

    vk::CmdEndRendering(m_commandBuffer->handle());

    {
        set_desired();
        vk::CmdBeginRendering(m_commandBuffer->handle(), &begin_rendering_info);
        m_errorMonitor->Finish();
        vk::CmdEndRendering(m_commandBuffer->handle());
    }
    {
        color_attachment.clearValue.color.float32[0] = 0.55f;

        set_desired();
        vk::CmdBeginRendering(m_commandBuffer->handle(), &begin_rendering_info);
        m_errorMonitor->VerifyFound();
    }

    m_commandBuffer->end();
}

TEST_F(VkNvidiaBestPracticesLayerTest, BeginCommandBuffer_OneTimeSubmit) {
    InitBestPracticesFramework(kEnableNVIDIAValidation);
    InitState();

    VkResult err = VK_SUCCESS;

    VkCommandPoolCreateInfo command_pool_ci = LvlInitStruct<VkCommandPoolCreateInfo>();
    command_pool_ci.flags = VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT;
    command_pool_ci.queueFamilyIndex = m_device->graphics_queue_node_index_;

    vk_testing::CommandPool command_pool(*m_device, command_pool_ci);

    VkCommandBufferAllocateInfo allocate_info = LvlInitStruct<VkCommandBufferAllocateInfo>();
    allocate_info.commandPool = command_pool.handle();
    allocate_info.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
    allocate_info.commandBufferCount = 1;
    vk_testing::CommandBuffer command_buffer0(*m_device, allocate_info);
    vk_testing::CommandBuffer command_buffer1(*m_device, allocate_info);
    vk_testing::CommandBuffer command_buffer2(*m_device, allocate_info);

    VkSubmitInfo submit_info = LvlInitStruct<VkSubmitInfo>();
    submit_info.commandBufferCount = 1;

    VkCommandBufferBeginInfo begin_info = LvlInitStruct<VkCommandBufferBeginInfo>();

    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-vkBeginCommandBuffer-one-time-submit");

        submit_info.pCommandBuffers = &command_buffer0.handle();

        command_buffer0.begin(&begin_info);
        command_buffer0.end();

        err = vk::QueueSubmit(m_device->m_queue, 1, &submit_info, VK_NULL_HANDLE);
        ASSERT_VK_SUCCESS(err);

        m_device->wait();

        err = vk::BeginCommandBuffer(command_buffer0.handle(), &begin_info);
        m_errorMonitor->VerifyFound();

        if (err == VK_SUCCESS) {
            command_buffer0.end();
        }
    }
    {
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-vkBeginCommandBuffer-one-time-submit");

        submit_info.pCommandBuffers = &command_buffer1.handle();

        command_buffer1.begin(&begin_info);
        command_buffer1.end();

        for (int i = 0; i < 2; ++i) {
            err = vk::QueueSubmit(m_device->m_queue, 1, &submit_info, VK_NULL_HANDLE);
            ASSERT_VK_SUCCESS(err);

            m_device->wait();
        }

        err = vk::BeginCommandBuffer(command_buffer1.handle(), &begin_info);
        m_errorMonitor->Finish();

        if (err == VK_SUCCESS) {
            command_buffer1.end();
        }
    }
    {
        begin_info.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
        m_errorMonitor->SetDesiredFailureMsg(VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
                                             "UNASSIGNED-BestPractices-vkBeginCommandBuffer-one-time-submit");

        submit_info.pCommandBuffers = &command_buffer2.handle();

        command_buffer2.begin(&begin_info);
        command_buffer2.end();

        err = vk::QueueSubmit(m_device->m_queue, 1, &submit_info, VK_NULL_HANDLE);
        ASSERT_VK_SUCCESS(err);

        m_device->wait();

        err = vk::BeginCommandBuffer(command_buffer2.handle(), &begin_info);
        m_errorMonitor->Finish();

        if (err == VK_SUCCESS) {
            command_buffer2.end();
        }
    }
}