vk_image.c

Go to the documentation of this file.

/*
Copyright (C) 1997-2001 Id Software, Inc.
Copyright (C) 2018-2019 Krzysztof Kondrak
 
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
 
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  
 
See the GNU General Public License for more details.
 
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
 
*/
 
#include "vk_local.h"
 
image_t     vktextures[MAX_VKTEXTURES];
int         numvktextures;
int         base_textureid;     // gltextures[i] = base_textureid+i
// texture for storing raw image data (cinematics, endscreens, etc.)
qvktexture_t vk_rawTexture = QVVKTEXTURE_INIT;
 
static byte          intensitytable[256];
static unsigned char gammatable[256];
 
cvar_t      *intensity;
extern cvar_t   *vk_mip_nearfilter;
 
unsigned    d_8to24table[256];
 
uint32_t Vk_Upload8 (byte *data, int width, int height,  qboolean mipmap, qboolean is_sky );
uint32_t Vk_Upload32 (unsigned *data, int width, int height,  qboolean mipmap);
 
// default global texture and lightmap samplers
qvksampler_t vk_current_sampler = S_MIPMAP_LINEAR;
qvksampler_t vk_current_lmap_sampler = S_MIPMAP_LINEAR;
 
// internal helper
static VkImageAspectFlags getDepthStencilAspect(VkFormat depthFormat)
{
    switch (depthFormat)
    {
    case VK_FORMAT_D32_SFLOAT_S8_UINT:
    case VK_FORMAT_D24_UNORM_S8_UINT:
    case VK_FORMAT_D16_UNORM_S8_UINT:
        return VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT;
    default:
        return VK_IMAGE_ASPECT_DEPTH_BIT;
    }
}
 
// internal helper
static void transitionImageLayout(const VkCommandBuffer *cmdBuffer, const VkQueue *queue, const qvktexture_t *texture, const VkImageLayout oldLayout, const VkImageLayout newLayout)
{
    VkPipelineStageFlags srcStage = 0;
    VkPipelineStageFlags dstStage = 0;
 
    VkImageMemoryBarrier imgBarrier = {
        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
        .pNext = NULL,
        .oldLayout = oldLayout,
        .newLayout = newLayout,
        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
        .image = texture->image,
        .subresourceRange.baseMipLevel = 0, // no mip mapping levels
        .subresourceRange.baseArrayLayer = 0,
        .subresourceRange.layerCount = 1,
        .subresourceRange.levelCount = texture->mipLevels
    };
 
    if (newLayout == VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL)
    {
        imgBarrier.subresourceRange.aspectMask = getDepthStencilAspect(texture->format);
    }
    else
    {
        imgBarrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
    }
 
    if (oldLayout == VK_IMAGE_LAYOUT_UNDEFINED && newLayout == VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL)
    {
        imgBarrier.srcAccessMask = 0;
        imgBarrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
        srcStage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
        dstStage = VK_PIPELINE_STAGE_TRANSFER_BIT;
    }
    // transiton that may occur when updating existing image
    else if (oldLayout == VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL && newLayout == VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL)
    {
        imgBarrier.srcAccessMask = 0;
        imgBarrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
        srcStage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
        dstStage = VK_PIPELINE_STAGE_TRANSFER_BIT;
    }
    else if (oldLayout == VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL && newLayout == VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL)
    {
        if (vk_device.transferQueue == vk_device.gfxQueue)
        {
            imgBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
            imgBarrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
            srcStage = VK_PIPELINE_STAGE_TRANSFER_BIT;
            dstStage = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
        }
        else
        {
            if (vk_device.transferQueue == *queue)
            {
                // if the image is exclusively shared, start queue ownership transfer process (release) - only for VK_SHARING_MODE_EXCLUSIVE
                imgBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
                imgBarrier.dstAccessMask = 0;
                imgBarrier.srcQueueFamilyIndex = vk_device.transferFamilyIndex;
                imgBarrier.dstQueueFamilyIndex = vk_device.gfxFamilyIndex;
                srcStage = VK_PIPELINE_STAGE_TRANSFER_BIT;
                dstStage = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
            }
            else
            {
                // continuing queue transfer (acquisition) - this will only happen for VK_SHARING_MODE_EXCLUSIVE images
                if (texture->sharingMode == VK_SHARING_MODE_EXCLUSIVE)
                {
                    imgBarrier.srcAccessMask = 0;
                    imgBarrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
                    imgBarrier.srcQueueFamilyIndex = vk_device.transferFamilyIndex;
                    imgBarrier.dstQueueFamilyIndex = vk_device.gfxFamilyIndex;
                    srcStage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
                    dstStage = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
                }
                else
                {
                    imgBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
                    imgBarrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
                    srcStage = VK_PIPELINE_STAGE_TRANSFER_BIT;
                    dstStage = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
                }
            }
        }
    }
    else if (oldLayout == VK_IMAGE_LAYOUT_UNDEFINED && newLayout == VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL)
    {
        imgBarrier.srcAccessMask = 0;
        imgBarrier.dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
        srcStage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
        dstStage = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT;
    }
    else if (oldLayout == VK_IMAGE_LAYOUT_UNDEFINED && newLayout == VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL)
    {
        imgBarrier.srcAccessMask = 0;
        imgBarrier.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
        srcStage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
        dstStage = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
    }
    else
    {
        assert(0 && !"Invalid image stage!");
    }
 
    vkCmdPipelineBarrier(*cmdBuffer, srcStage, dstStage, 0, 0, NULL, 0, NULL, 1, &imgBarrier);
}
 
// internal helper
static void generateMipmaps(const VkCommandBuffer *cmdBuffer, const qvktexture_t *texture, uint32_t width, uint32_t height)
{
    int32_t mipWidth = width;
    int32_t mipHeight = height;
    VkFilter mipFilter = vk_mip_nearfilter->value > 0 ? VK_FILTER_NEAREST : VK_FILTER_LINEAR;
 
    VkImageMemoryBarrier imgBarrier = {
        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
        .pNext = NULL,
        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
        .image = texture->image,
        .subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
        .subresourceRange.levelCount = 1,
        .subresourceRange.baseArrayLayer = 0,
        .subresourceRange.layerCount = 1
    };
 
    // copy rescaled mip data between consecutive levels (each higher level is half the size of the previous level)
    for (uint32_t i = 1; i < texture->mipLevels; ++i)
    {
        imgBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
        imgBarrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
        imgBarrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
        imgBarrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
        imgBarrier.subresourceRange.baseMipLevel = i - 1;
 
        vkCmdPipelineBarrier(*cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, NULL, 0, NULL, 1, &imgBarrier);
 
        VkImageBlit blit = {
            .srcSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
            .srcSubresource.mipLevel = i - 1,
            .srcSubresource.baseArrayLayer = 0,
            .srcSubresource.layerCount = 1,
            .srcOffsets[0] = { 0, 0, 0 },
            .srcOffsets[1] = { mipWidth, mipHeight, 1 },
            .dstSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
            .dstSubresource.mipLevel = i,
            .dstSubresource.baseArrayLayer = 0,
            .dstSubresource.layerCount = 1,
            .dstOffsets[0] = { 0, 0, 0 },
            .dstOffsets[1] = { mipWidth > 1 ? mipWidth >> 1 : 1,
                              mipHeight > 1 ? mipHeight >> 1 : 1, 1 } // each mip level is half the size of the previous level
        };
 
        // src image == dst image, because we're blitting between different mip levels of the same image
        vkCmdBlitImage(*cmdBuffer, texture->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
            texture->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &blit, mipFilter);
 
        imgBarrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
        imgBarrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
        imgBarrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
        imgBarrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
 
        vkCmdPipelineBarrier(*cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, 0, 0, NULL, 0, NULL, 1, &imgBarrier);
 
        // avoid zero-sized mip levels
        if (mipWidth > 1)  mipWidth >>= 1;
        if (mipHeight > 1) mipHeight >>= 1;
    }
 
    imgBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
    imgBarrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
    imgBarrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
    imgBarrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
    imgBarrier.subresourceRange.baseMipLevel = texture->mipLevels - 1;
 
    vkCmdPipelineBarrier(*cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, 0, 0, NULL, 0, NULL, 1, &imgBarrier);
}
 
// internal helper
static void createTextureImage(qvktexture_t *dstTex, const unsigned char *data, uint32_t width, uint32_t height)
{
    int unifiedTransferAndGfx = vk_device.transferQueue == vk_device.gfxQueue ? 1 : 0;
    // assuming 32bit images
    uint32_t imageSize = width * height * 4;
 
    VkBuffer staging_buffer;
    VkCommandBuffer command_buffer;
    uint32_t staging_offset;
    void *imgData = QVk_GetStagingBuffer(imageSize, 4, &command_buffer, &staging_buffer, &staging_offset);
    memcpy(imgData, data, (size_t)imageSize);
 
    VkImageUsageFlags imageUsage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT;
    // set extra image usage flag if we're dealing with mipmapped image - will need it for copying data between mip levels
    if (dstTex->mipLevels > 1)
        imageUsage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
 
    VK_VERIFY(QVk_CreateImage(width, height, dstTex->format, VK_IMAGE_TILING_OPTIMAL, imageUsage, VMA_MEMORY_USAGE_GPU_ONLY, dstTex));
 
    transitionImageLayout(&command_buffer, &vk_device.transferQueue, dstTex, VK_IMAGE_LAYOUT_UNDEFINED, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
    // copy buffer to image
    VkBufferImageCopy region = {
        .bufferOffset = staging_offset,
        .bufferRowLength = 0,
        .bufferImageHeight = 0,
        .imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
        .imageSubresource.mipLevel = 0,
        .imageSubresource.baseArrayLayer = 0,
        .imageSubresource.layerCount = 1,
        .imageOffset = { 0, 0, 0 },
        .imageExtent = { width, height, 1 }
    };
 
    vkCmdCopyBufferToImage(command_buffer, staging_buffer, dstTex->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &region);
 
    if (dstTex->mipLevels > 1)
    {
        // vkCmdBlitImage requires a queue with GRAPHICS_BIT present
        generateMipmaps(&command_buffer, dstTex, width, height);
    }
    else
    {
        // for non-unified transfer and graphics, this step begins queue ownership transfer to graphics queue (for exclusive sharing only)
        if (unifiedTransferAndGfx || dstTex->sharingMode == VK_SHARING_MODE_EXCLUSIVE)
            transitionImageLayout(&command_buffer, &vk_device.transferQueue, dstTex, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
 
        if (!unifiedTransferAndGfx)
        {
            transitionImageLayout(&command_buffer, &vk_device.gfxQueue, dstTex, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
        }
    }
}
 
VkResult QVk_CreateImageView(const VkImage *image, VkImageAspectFlags aspectFlags, VkImageView *imageView, VkFormat format, uint32_t mipLevels)
{
    VkImageViewCreateInfo ivCreateInfo = {
        .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
        .pNext = NULL,
        .flags = 0,
        .image = *image,
        .viewType = VK_IMAGE_VIEW_TYPE_2D,
        .format = format,
        .components.r = VK_COMPONENT_SWIZZLE_IDENTITY,
        .components.g = VK_COMPONENT_SWIZZLE_IDENTITY,
        .components.b = VK_COMPONENT_SWIZZLE_IDENTITY,
        .components.a = VK_COMPONENT_SWIZZLE_IDENTITY,
        .subresourceRange.aspectMask = aspectFlags,
        .subresourceRange.baseArrayLayer = 0,
        .subresourceRange.baseMipLevel = 0,
        .subresourceRange.layerCount = 1,
        .subresourceRange.levelCount = mipLevels
    };
 
    return vkCreateImageView(vk_device.logical, &ivCreateInfo, NULL, imageView);
}
 
VkResult QVk_CreateImage(uint32_t width, uint32_t height, VkFormat format, VkImageTiling tiling, VkImageUsageFlags usage, VmaMemoryUsage memUsage, qvktexture_t *texture)
{
    VkImageCreateInfo imageInfo = {
        .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
        .imageType = VK_IMAGE_TYPE_2D,
        .extent.width = width,
        .extent.height = height,
        .extent.depth = 1,
        .mipLevels = texture->mipLevels,
        .arrayLayers = 1,
        .format = format,
        .tiling = tiling,
        .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
        .usage = usage,
        .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
        .samples = texture->sampleCount,
        .flags = 0
    };
 
    uint32_t queueFamilies[] = { (uint32_t)vk_device.gfxFamilyIndex, (uint32_t)vk_device.transferFamilyIndex };
    if (vk_device.gfxFamilyIndex != vk_device.transferFamilyIndex)
    {
        imageInfo.sharingMode = VK_SHARING_MODE_CONCURRENT;
        imageInfo.queueFamilyIndexCount = 2;
        imageInfo.pQueueFamilyIndices = queueFamilies;
    }
 
    VmaAllocationCreateInfo vmallocInfo = {
        .flags = texture->vmaFlags,
        .usage = memUsage
    };
 
    texture->sharingMode = imageInfo.sharingMode;
    return vmaCreateImage(vk_malloc, &imageInfo, &vmallocInfo, &texture->image, &texture->allocation, &texture->allocInfo);
}
 
void QVk_CreateDepthBuffer(VkSampleCountFlagBits sampleCount, qvktexture_t *depthBuffer)
{
    depthBuffer->format = QVk_FindDepthFormat();
    depthBuffer->sampleCount = sampleCount;
    // On 64-bit builds, Intel drivers throw a warning:
    // "Mapping an image with layout VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used."
    // Minor annoyance but we don't want any validation warnings, so we create dedicated allocation for depth buffer.
    // more details: https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator/issues/34
    // Note that this is a false positive which in other cases could be ignored: https://gpuopen-librariesandsdks.github.io/VulkanMemoryAllocator/html/general_considerations.html#general_considerations_validation_layer_warnings
    depthBuffer->vmaFlags = VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT;
 
    VK_VERIFY(QVk_CreateImage(vk_swapchain.extent.width, vk_swapchain.extent.height, depthBuffer->format, VK_IMAGE_TILING_OPTIMAL, VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT, VMA_MEMORY_USAGE_GPU_ONLY, depthBuffer));
    VK_VERIFY(QVk_CreateImageView(&depthBuffer->image, getDepthStencilAspect(depthBuffer->format), &depthBuffer->imageView, depthBuffer->format, depthBuffer->mipLevels));
}
 
void QVk_CreateColorBuffer(VkSampleCountFlagBits sampleCount, qvktexture_t *colorBuffer, int extraFlags)
{
    colorBuffer->format = vk_swapchain.format;
    colorBuffer->sampleCount = sampleCount;
    // On 64-bit builds, Intel drivers throw a warning:
    // "Mapping an image with layout VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used."
    // Minor annoyance but we don't want any validation warnings, so we create dedicated allocation for color buffer.
    // more details: https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator/issues/34
    // Note that this is a false positive which in other cases could be ignored: https://gpuopen-librariesandsdks.github.io/VulkanMemoryAllocator/html/general_considerations.html#general_considerations_validation_layer_warnings
    colorBuffer->vmaFlags = VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT;
 
    VK_VERIFY(QVk_CreateImage(vk_swapchain.extent.width, vk_swapchain.extent.height, colorBuffer->format, VK_IMAGE_TILING_OPTIMAL, extraFlags | VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT, VMA_MEMORY_USAGE_GPU_ONLY, colorBuffer));
    VK_VERIFY(QVk_CreateImageView(&colorBuffer->image, VK_IMAGE_ASPECT_COLOR_BIT, &colorBuffer->imageView, colorBuffer->format, colorBuffer->mipLevels));
}
 
void QVk_CreateTexture(qvktexture_t *texture, const unsigned char *data, uint32_t width, uint32_t height, qvksampler_t samplerType)
{
    createTextureImage(texture, data, width, height);
    VK_VERIFY(QVk_CreateImageView(&texture->image, VK_IMAGE_ASPECT_COLOR_BIT, &texture->imageView, texture->format, texture->mipLevels));
 
    // create descriptor set for the texture
    VkDescriptorSetAllocateInfo dsAllocInfo = {
        .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO,
        .pNext = NULL,
        .descriptorPool = vk_descriptorPool,
        .descriptorSetCount = 1,
        .pSetLayouts = &vk_samplerDescSetLayout
    };
 
    VK_VERIFY(vkAllocateDescriptorSets(vk_device.logical, &dsAllocInfo, &texture->descriptorSet));
 
    // attach sampler
    QVk_UpdateTextureSampler(texture, samplerType);
}
 
void QVk_UpdateTextureData(qvktexture_t *texture, const unsigned char *data, uint32_t offset_x, uint32_t offset_y, uint32_t width, uint32_t height)
{
    int unifiedTransferAndGfx = vk_device.transferQueue == vk_device.gfxQueue ? 1 : 0;
    // assuming 32bit images
    uint32_t imageSize = width * height * 4;
 
    VkBuffer staging_buffer;
    VkCommandBuffer command_buffer;
    uint32_t staging_offset;
    void *imgData = QVk_GetStagingBuffer(imageSize, 4, &command_buffer, &staging_buffer, &staging_offset);
    memcpy(imgData, data, (size_t)imageSize);
 
    transitionImageLayout(&command_buffer, &vk_device.transferQueue, texture, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
    // copy buffer to image
    VkBufferImageCopy region = {
        .bufferOffset = staging_offset,
        .bufferRowLength = 0,
        .bufferImageHeight = 0,
        .imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
        .imageSubresource.mipLevel = 0,
        .imageSubresource.baseArrayLayer = 0,
        .imageSubresource.layerCount = 1,
        .imageOffset = { offset_x, offset_y, 0 },
        .imageExtent = { width, height, 1 }
    };
 
    vkCmdCopyBufferToImage(command_buffer, staging_buffer, texture->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &region);
 
    if (texture->mipLevels > 1)
    {
        // vkCmdBlitImage requires a queue with GRAPHICS_BIT present
        generateMipmaps(&command_buffer, texture, width, height);
    }
    else
    {
        // for non-unified transfer and graphics, this step begins queue ownership transfer to graphics queue (for exclusive sharing only)
        if (unifiedTransferAndGfx || texture->sharingMode == VK_SHARING_MODE_EXCLUSIVE)
            transitionImageLayout(&command_buffer, &vk_device.transferQueue, texture, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
 
        if (!unifiedTransferAndGfx)
        {
            transitionImageLayout(&command_buffer, &vk_device.gfxQueue, texture, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
        }
    }
}
 
void QVk_ReleaseTexture(qvktexture_t *texture)
{
    QVk_SubmitStagingBuffers();
    vkDeviceWaitIdle(vk_device.logical);
 
    if (texture->image != VK_NULL_HANDLE)
        vmaDestroyImage(vk_malloc, texture->image, texture->allocation);
    if (texture->imageView != VK_NULL_HANDLE)
        vkDestroyImageView(vk_device.logical, texture->imageView, NULL);
    if (texture->descriptorSet != VK_NULL_HANDLE)
        vkFreeDescriptorSets(vk_device.logical, vk_descriptorPool, 1, &texture->descriptorSet);
 
    texture->image = VK_NULL_HANDLE;
    texture->imageView = VK_NULL_HANDLE;
    texture->descriptorSet = VK_NULL_HANDLE;
}
 
void QVk_ReadPixels(uint8_t *dstBuffer, uint32_t width, uint32_t height)
{
    qvkbuffer_t buff;
    VkCommandBuffer cmdBuffer;
    qvkbufferopts_t buffOpts = {
        .usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT,
        .reqMemFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
        .prefMemFlags = VK_MEMORY_PROPERTY_HOST_CACHED_BIT,
        .vmaUsage = VMA_MEMORY_USAGE_CPU_ONLY,
        // When taking a screenshot on Intel, the Linux driver may throw a warning:
        // "Mapping an image with layout VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used."
        // Minor annoyance but we don't want any validation warnings, so we create dedicated allocation for the image buffer.
        // more details: https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator/issues/34
        // Note that this is a false positive which in other cases could be ignored: https://gpuopen-librariesandsdks.github.io/VulkanMemoryAllocator/html/general_considerations.html#general_considerations_validation_layer_warnings
        .vmaFlags = VMA_ALLOCATION_CREATE_MAPPED_BIT | VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT
    };
 
    VK_VERIFY(QVk_CreateBuffer(width * height * 4, &buff, buffOpts));
    cmdBuffer = QVk_CreateCommandBuffer(&vk_commandPool, VK_COMMAND_BUFFER_LEVEL_PRIMARY);
    VK_VERIFY(QVk_BeginCommand(&cmdBuffer));
    
    // transition the current swapchain image to be a source of data transfer to our buffer
    extern int vk_activeBufferIdx;
    VkImageMemoryBarrier imgBarrier = {
        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
        .pNext = NULL,
        .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
        .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
        .oldLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR,
        .newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
        .image = vk_swapchain.images[vk_activeBufferIdx],
        .subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
        .subresourceRange.baseMipLevel = 0,
        .subresourceRange.baseArrayLayer = 0,
        .subresourceRange.layerCount = 1,
        .subresourceRange.levelCount = 1
    };
 
    vkCmdPipelineBarrier(cmdBuffer, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, NULL, 0, NULL, 1, &imgBarrier);
 
    VkBufferImageCopy region = {
        .bufferOffset = 0,
        .bufferRowLength = width,
        .bufferImageHeight = height,
        .imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
        .imageSubresource.mipLevel = 0,
        .imageSubresource.baseArrayLayer = 0,
        .imageSubresource.layerCount = 1,
        .imageOffset = { 0, 0, 0 },
        .imageExtent = { width, height, 1 }
    };
 
    // copy the swapchain image
    vkCmdCopyImageToBuffer(cmdBuffer, vk_swapchain.images[vk_activeBufferIdx], VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, buff.buffer, 1, &region);
    VK_VERIFY(vkDeviceWaitIdle(vk_device.logical));
    QVk_SubmitCommand(&cmdBuffer, &vk_device.gfxQueue);
 
    // store image in destination buffer
    memcpy(dstBuffer, (uint8_t *)buff.allocInfo.pMappedData, width * height * 4);
 
    QVk_FreeBuffer(&buff);
}
 
/*
===============
Vk_ImageList_f
===============
*/
void    Vk_ImageList_f (void)
{
    int     i;
    image_t *image;
    int     texels;
 
    ri.Con_Printf(PRINT_ALL, "------------------\n");
    texels = 0;
 
    for (i = 0, image = vktextures; i < numvktextures; i++, image++)
    {
        if (image->vk_texture.image == VK_NULL_HANDLE)
            continue;
        texels += image->upload_width*image->upload_height;
        switch (image->type)
        {
        case it_skin:
            ri.Con_Printf(PRINT_ALL, "M");
            break;
        case it_sprite:
            ri.Con_Printf(PRINT_ALL, "S");
            break;
        case it_wall:
            ri.Con_Printf(PRINT_ALL, "W");
            break;
        case it_pic:
            ri.Con_Printf(PRINT_ALL, "P");
            break;
        default:
            ri.Con_Printf(PRINT_ALL, " ");
            break;
        }
 
        ri.Con_Printf(PRINT_ALL, " %3i %3i RGB: %s\n",
            image->upload_width, image->upload_height, image->name);
    }
    ri.Con_Printf(PRINT_ALL, "Total texel count (not counting mipmaps): %i\n", texels);
}
 
 
/*
=============================================================================
 
  scrap allocation
 
  Allocate all the little status bar obejcts into a single texture
  to crutch up inefficient hardware / drivers
 
=============================================================================
*/
 
#define MAX_SCRAPS      1
#define BLOCK_WIDTH     256
#define BLOCK_HEIGHT    256
 
int         scrap_allocated[MAX_SCRAPS][BLOCK_WIDTH];
byte        scrap_texels[MAX_SCRAPS][BLOCK_WIDTH*BLOCK_HEIGHT];
// textures for storing scrap image data (tiny image atlas)
qvktexture_t vk_scrapTextures[MAX_SCRAPS] = { QVVKTEXTURE_INIT };
 
// returns a texture number and the position inside it
int Scrap_AllocBlock (int w, int h, int *x, int *y)
{
    int     i, j;
    int     best, best2;
    int     texnum;
 
    for (texnum=0 ; texnum<MAX_SCRAPS ; texnum++)
    {
        best = BLOCK_HEIGHT;
 
        for (i=0 ; i<BLOCK_WIDTH-w ; i++)
        {
            best2 = 0;
 
            for (j=0 ; j<w ; j++)
            {
                if (scrap_allocated[texnum][i+j] >= best)
                    break;
                if (scrap_allocated[texnum][i+j] > best2)
                    best2 = scrap_allocated[texnum][i+j];
            }
            if (j == w)
            {   // this is a valid spot
                *x = i;
                *y = best = best2;
            }
        }
 
        if (best + h > BLOCK_HEIGHT)
            continue;
 
        for (i=0 ; i<w ; i++)
            scrap_allocated[texnum][*x + i] = best + h;
 
        return texnum;
    }
 
    return -1;
//  Sys_Error ("Scrap_AllocBlock: full");
}
 
typedef struct
{
    char *name;
    qvksampler_t samplerType;
} vkmode_t;
 
vkmode_t modes[] = {
    {"VK_NEAREST", S_NEAREST },
    {"VK_LINEAR", S_LINEAR },
    {"VK_MIPMAP_NEAREST", S_MIPMAP_NEAREST },
    {"VK_MIPMAP_LINEAR", S_MIPMAP_LINEAR }
};
 
#define NUM_VK_MODES (sizeof(modes) / sizeof (vkmode_t))
 
/*
 ===============
 Vk_TextureMode
 ===============
 */
void Vk_TextureMode( char *string )
{
    int     i,j;
    image_t *image;
    static char prev_mode[32] = { "VK_MIPMAP_LINEAR" };
 
    for (i = 0; i < NUM_VK_MODES; i++)
    {
        if (!Q_stricmp(modes[i].name, string))
            break;
    }
 
    if (i == NUM_VK_MODES)
    {
        ri.Con_Printf(PRINT_ALL, "bad filter name (valid values: VK_NEAREST, VK_LINEAR, VK_MIPMAP_NEAREST, VK_MIPMAP_LINEAR)\n");
        ri.Cvar_Set("vk_texturemode", prev_mode);
        return;
    }
 
    memcpy(prev_mode, string, strlen(string));
    prev_mode[strlen(string)] = '\0';
 
    i += vk_aniso->value > 0 ? NUM_VK_MODES : 0;
    vk_current_sampler = i;
 
    vkDeviceWaitIdle(vk_device.logical);
    for (j = 0, image = vktextures; j < numvktextures; j++, image++)
    {
        // skip console characters - we want them unfiltered at all times
        if (image->vk_texture.image != VK_NULL_HANDLE && Q_stricmp(image->name, "pics/conchars.pcx"))
            QVk_UpdateTextureSampler(&image->vk_texture, i);
    }
 
    for (j = 0; j < MAX_SCRAPS; j++)
    {
        if (vk_scrapTextures[j].image != VK_NULL_HANDLE)
            QVk_UpdateTextureSampler(&vk_scrapTextures[j], i);
    }
 
    if (vk_rawTexture.image != VK_NULL_HANDLE)
        QVk_UpdateTextureSampler(&vk_rawTexture, i);
}
 
/*
 ===============
 Vk_LmapTextureMode
 ===============
 */
void Vk_LmapTextureMode( char *string )
{
    int     i,j;
    static char prev_mode[32] = { "VK_MIPMAP_LINEAR" };
 
    for (i = 0; i < NUM_VK_MODES; i++)
    {
        if (!Q_stricmp(modes[i].name, string))
            break;
    }
 
    if (i == NUM_VK_MODES)
    {
        ri.Con_Printf(PRINT_ALL, "bad filter name (valid values: VK_NEAREST, VK_LINEAR, VK_MIPMAP_NEAREST, VK_MIPMAP_LINEAR)\n");
        ri.Cvar_Set("vk_lmaptexturemode", prev_mode);
        return;
    }
 
    memcpy(prev_mode, string, strlen(string));
    prev_mode[strlen(string)] = '\0';
 
    i += vk_aniso->value > 0 ? NUM_VK_MODES : 0;
    vk_current_lmap_sampler = i;
 
    vkDeviceWaitIdle(vk_device.logical);
    for (j = 0; j < MAX_LIGHTMAPS*2; j++)
    {
        if (vk_state.lightmap_textures[j].image != VK_NULL_HANDLE)
            QVk_UpdateTextureSampler(&vk_state.lightmap_textures[j], i);
    }
}
 
/*
=================================================================
 
PCX LOADING
 
=================================================================
*/
 
 
/*
==============
LoadPCX
==============
*/
void LoadPCX (char *filename, byte **pic, byte **palette, int *width, int *height)
{
    byte    *raw;
    pcx_t   *pcx;
    int     x, y;
    int     len;
    int     dataByte, runLength;
    byte    *out, *pix;
 
    *pic = NULL;
    *palette = NULL;
 
    //
    // load the file
    //
    len = ri.FS_LoadFile (filename, (void **)&raw);
    if (!raw)
    {
        ri.Con_Printf (PRINT_DEVELOPER, "Bad pcx file %s\n", filename);
        return;
    }
 
    //
    // parse the PCX file
    //
    pcx = (pcx_t *)raw;
 
    pcx->xmin = LittleShort(pcx->xmin);
    pcx->ymin = LittleShort(pcx->ymin);
    pcx->xmax = LittleShort(pcx->xmax);
    pcx->ymax = LittleShort(pcx->ymax);
    pcx->hres = LittleShort(pcx->hres);
    pcx->vres = LittleShort(pcx->vres);
    pcx->bytes_per_line = LittleShort(pcx->bytes_per_line);
    pcx->palette_type = LittleShort(pcx->palette_type);
 
    raw = &pcx->data;
 
    if (pcx->manufacturer != 0x0a
        || pcx->version != 5
        || pcx->encoding != 1
        || pcx->bits_per_pixel != 8
        || pcx->xmax >= 640
        || pcx->ymax >= 480)
    {
        ri.Con_Printf (PRINT_ALL, "Bad pcx file %s\n", filename);
        return;
    }
 
    out = malloc ( (pcx->ymax+1) * (pcx->xmax+1) );
 
    *pic = out;
 
    pix = out;
 
    if (palette)
    {
        *palette = malloc(768);
        memcpy (*palette, (byte *)pcx + len - 768, 768);
    }
 
    if (width)
        *width = pcx->xmax+1;
    if (height)
        *height = pcx->ymax+1;
 
    for (y=0 ; y<=pcx->ymax ; y++, pix += pcx->xmax+1)
    {
        for (x=0 ; x<=pcx->xmax ; )
        {
            dataByte = *raw++;
 
            if((dataByte & 0xC0) == 0xC0)
            {
                runLength = dataByte & 0x3F;
                dataByte = *raw++;
            }
            else
                runLength = 1;
 
            while(runLength-- > 0)
                pix[x++] = dataByte;
        }
 
    }
 
    if ( raw - (byte *)pcx > len)
    {
        ri.Con_Printf (PRINT_DEVELOPER, "PCX file %s was malformed", filename);
        free (*pic);
        *pic = NULL;
    }
 
    ri.FS_FreeFile (pcx);
}
 
/*
=========================================================
 
TARGA LOADING
 
=========================================================
*/
 
typedef struct _TargaHeader {
    unsigned char   id_length, colormap_type, image_type;
    unsigned short  colormap_index, colormap_length;
    unsigned char   colormap_size;
    unsigned short  x_origin, y_origin, width, height;
    unsigned char   pixel_size, attributes;
} TargaHeader;
 
 
/*
=============
LoadTGA
=============
*/
void LoadTGA (char *name, byte **pic, int *width, int *height)
{
    int     columns, rows, numPixels;
    byte    *pixbuf;
    int     row, column;
    byte    *buf_p;
    byte    *buffer;
    int     length;
    TargaHeader     targa_header;
    byte            *targa_rgba;
    byte tmp[2];
 
    *pic = NULL;
 
    //
    // load the file
    //
    length = ri.FS_LoadFile (name, (void **)&buffer);
    if (!buffer)
    {
        ri.Con_Printf (PRINT_DEVELOPER, "Bad tga file %s\n", name);
        return;
    }
 
    buf_p = buffer;
 
    targa_header.id_length = *buf_p++;
    targa_header.colormap_type = *buf_p++;
    targa_header.image_type = *buf_p++;
    
    tmp[0] = buf_p[0];
    tmp[1] = buf_p[1];
    targa_header.colormap_index = LittleShort ( *((short *)tmp) );
    buf_p+=2;
    tmp[0] = buf_p[0];
    tmp[1] = buf_p[1];
    targa_header.colormap_length = LittleShort ( *((short *)tmp) );
    buf_p+=2;
    targa_header.colormap_size = *buf_p++;
    targa_header.x_origin = LittleShort ( *((short *)buf_p) );
    buf_p+=2;
    targa_header.y_origin = LittleShort ( *((short *)buf_p) );
    buf_p+=2;
    targa_header.width = LittleShort ( *((short *)buf_p) );
    buf_p+=2;
    targa_header.height = LittleShort ( *((short *)buf_p) );
    buf_p+=2;
    targa_header.pixel_size = *buf_p++;
    targa_header.attributes = *buf_p++;
 
    if (targa_header.image_type!=2 
        && targa_header.image_type!=10) 
        ri.Sys_Error (ERR_DROP, "LoadTGA: Only type 2 and 10 targa RGB images supported\n");
 
    if (targa_header.colormap_type !=0 
        || (targa_header.pixel_size!=32 && targa_header.pixel_size!=24))
        ri.Sys_Error (ERR_DROP, "LoadTGA: Only 32 or 24 bit images supported (no colormaps)\n");
 
    columns = targa_header.width;
    rows = targa_header.height;
    numPixels = columns * rows;
 
    if (width)
        *width = columns;
    if (height)
        *height = rows;
 
    targa_rgba = malloc (numPixels*4);
    *pic = targa_rgba;
 
    if (targa_header.id_length != 0)
        buf_p += targa_header.id_length;  // skip TARGA image comment
    
    if (targa_header.image_type==2) {  // Uncompressed, RGB images
        for(row=rows-1; row>=0; row--) {
            pixbuf = targa_rgba + row*columns*4;
            for(column=0; column<columns; column++) {
                unsigned char red,green,blue,alphabyte;
                switch (targa_header.pixel_size) {
                    case 24:
                            
                            blue = *buf_p++;
                            green = *buf_p++;
                            red = *buf_p++;
                            *pixbuf++ = red;
                            *pixbuf++ = green;
                            *pixbuf++ = blue;
                            *pixbuf++ = 255;
                            break;
                    case 32:
                            blue = *buf_p++;
                            green = *buf_p++;
                            red = *buf_p++;
                            alphabyte = *buf_p++;
                            *pixbuf++ = red;
                            *pixbuf++ = green;
                            *pixbuf++ = blue;
                            *pixbuf++ = alphabyte;
                            break;
                }
            }
        }
    }
    else if (targa_header.image_type==10) {   // Runlength encoded RGB images
        unsigned char red = 0,green = 0,blue = 0,alphabyte = 0,packetHeader,packetSize,j;
        for(row=rows-1; row>=0; row--) {
            pixbuf = targa_rgba + row*columns*4;
            for(column=0; column<columns; ) {
                packetHeader= *buf_p++;
                packetSize = 1 + (packetHeader & 0x7f);
                if (packetHeader & 0x80) {        // run-length packet
                    switch (targa_header.pixel_size) {
                        case 24:
                                blue = *buf_p++;
                                green = *buf_p++;
                                red = *buf_p++;
                                alphabyte = 255;
                                break;
                        case 32:
                                blue = *buf_p++;
                                green = *buf_p++;
                                red = *buf_p++;
                                alphabyte = *buf_p++;
                                break;
                    }
    
                    for(j=0;j<packetSize;j++) {
                        *pixbuf++=red;
                        *pixbuf++=green;
                        *pixbuf++=blue;
                        *pixbuf++=alphabyte;
                        column++;
                        if (column==columns) { // run spans across rows
                            column=0;
                            if (row>0)
                                row--;
                            else
                                goto breakOut;
                            pixbuf = targa_rgba + row*columns*4;
                        }
                    }
                }
                else {                            // non run-length packet
                    for(j=0;j<packetSize;j++) {
                        switch (targa_header.pixel_size) {
                            case 24:
                                    blue = *buf_p++;
                                    green = *buf_p++;
                                    red = *buf_p++;
                                    *pixbuf++ = red;
                                    *pixbuf++ = green;
                                    *pixbuf++ = blue;
                                    *pixbuf++ = 255;
                                    break;
                            case 32:
                                    blue = *buf_p++;
                                    green = *buf_p++;
                                    red = *buf_p++;
                                    alphabyte = *buf_p++;
                                    *pixbuf++ = red;
                                    *pixbuf++ = green;
                                    *pixbuf++ = blue;
                                    *pixbuf++ = alphabyte;
                                    break;
                        }
                        column++;
                        if (column==columns) { // pixel packet run spans across rows
                            column=0;
                            if (row>0)
                                row--;
                            else
                                goto breakOut;
                            pixbuf = targa_rgba + row*columns*4;
                        }                       
                    }
                }
            }
            breakOut:;
        }
    }
 
    ri.FS_FreeFile (buffer);
}
 
 
/*
====================================================================
 
IMAGE FLOOD FILLING
 
====================================================================
*/
 
 
/*
=================
Mod_FloodFillSkin
 
Fill background pixels so mipmapping doesn't have haloes
=================
*/
 
typedef struct
{
    short       x, y;
} floodfill_t;
 
// must be a power of 2
#define FLOODFILL_FIFO_SIZE 0x1000
#define FLOODFILL_FIFO_MASK (FLOODFILL_FIFO_SIZE - 1)
 
#define FLOODFILL_STEP( off, dx, dy ) \
{ \
    if (pos[off] == fillcolor) \
    { \
        pos[off] = 255; \
        fifo[inpt].x = x + (dx), fifo[inpt].y = y + (dy); \
        inpt = (inpt + 1) & FLOODFILL_FIFO_MASK; \
    } \
    else if (pos[off] != 255) fdc = pos[off]; \
}
 
void R_FloodFillSkin( byte *skin, int skinwidth, int skinheight )
{
    byte                fillcolor = *skin; // assume this is the pixel to fill
    floodfill_t         fifo[FLOODFILL_FIFO_SIZE];
    int                 inpt = 0, outpt = 0;
    int                 filledcolor = -1;
    int                 i;
 
    if (filledcolor == -1)
    {
        filledcolor = 0;
        // attempt to find opaque black
        for (i = 0; i < 256; ++i)
            if (d_8to24table[i] == (255 << 0)) // alpha 1.0
            {
                filledcolor = i;
                break;
            }
    }
 
    // can't fill to filled color or to transparent color (used as visited marker)
    if ((fillcolor == filledcolor) || (fillcolor == 255))
    {
        //printf( "not filling skin from %d to %d\n", fillcolor, filledcolor );
        return;
    }
 
    fifo[inpt].x = 0, fifo[inpt].y = 0;
    inpt = (inpt + 1) & FLOODFILL_FIFO_MASK;
 
    while (outpt != inpt)
    {
        int         x = fifo[outpt].x, y = fifo[outpt].y;
        int         fdc = filledcolor;
        byte        *pos = &skin[x + skinwidth * y];
 
        outpt = (outpt + 1) & FLOODFILL_FIFO_MASK;
 
        if (x > 0)              FLOODFILL_STEP( -1, -1, 0 );
        if (x < skinwidth - 1)  FLOODFILL_STEP( 1, 1, 0 );
        if (y > 0)              FLOODFILL_STEP( -skinwidth, 0, -1 );
        if (y < skinheight - 1) FLOODFILL_STEP( skinwidth, 0, 1 );
        skin[x + skinwidth * y] = fdc;
    }
}
 
//=======================================================
 
 
/*
================
Vk_ResampleTexture
================
*/
void Vk_ResampleTexture (unsigned *in, int inwidth, int inheight, unsigned *out,  int outwidth, int outheight)
{
    int     i, j;
    unsigned    *inrow, *inrow2;
    unsigned    frac, fracstep;
    unsigned    p1[1024], p2[1024];
    byte        *pix1, *pix2, *pix3, *pix4;
 
    fracstep = inwidth*0x10000/outwidth;
 
    frac = fracstep>>2;
    for (i=0 ; i<outwidth ; i++)
    {
        p1[i] = 4*(frac>>16);
        frac += fracstep;
    }
    frac = 3*(fracstep>>2);
    for (i=0 ; i<outwidth ; i++)
    {
        p2[i] = 4*(frac>>16);
        frac += fracstep;
    }
 
    for (i=0 ; i<outheight ; i++, out += outwidth)
    {
        inrow = in + inwidth*(int)((i+0.25)*inheight/outheight);
        inrow2 = in + inwidth*(int)((i+0.75)*inheight/outheight);
 
        for (j=0 ; j<outwidth ; j++)
        {
            pix1 = (byte *)inrow + p1[j];
            pix2 = (byte *)inrow + p2[j];
            pix3 = (byte *)inrow2 + p1[j];
            pix4 = (byte *)inrow2 + p2[j];
            ((byte *)(out+j))[0] = (pix1[0] + pix2[0] + pix3[0] + pix4[0])>>2;
            ((byte *)(out+j))[1] = (pix1[1] + pix2[1] + pix3[1] + pix4[1])>>2;
            ((byte *)(out+j))[2] = (pix1[2] + pix2[2] + pix3[2] + pix4[2])>>2;
            ((byte *)(out+j))[3] = (pix1[3] + pix2[3] + pix3[3] + pix4[3])>>2;
        }
    }
}
 
/*
================
Vk_LightScaleTexture
 
Scale up the pixel values in a texture to increase the
lighting range
================
*/
void Vk_LightScaleTexture (unsigned *in, int inwidth, int inheight, qboolean only_gamma )
{
    if ( only_gamma )
    {
        int     i, c;
        byte    *p;
 
        p = (byte *)in;
 
        c = inwidth*inheight;
        for (i=0 ; i<c ; i++, p+=4)
        {
            p[0] = gammatable[p[0]];
            p[1] = gammatable[p[1]];
            p[2] = gammatable[p[2]];
        }
    }
    else
    {
        int     i, c;
        byte    *p;
 
        p = (byte *)in;
 
        c = inwidth*inheight;
        for (i=0 ; i<c ; i++, p+=4)
        {
            p[0] = gammatable[intensitytable[p[0]]];
            p[1] = gammatable[intensitytable[p[1]]];
            p[2] = gammatable[intensitytable[p[2]]];
        }
    }
}
 
/*
===============
Vk_Upload32
 
Returns number of mip levels
===============
*/
int     upload_width, upload_height;
unsigned int texBuffer[256 * 256];
 
uint32_t Vk_Upload32 (unsigned *data, int width, int height,  qboolean mipmap)
{
    unsigned    scaled[256 * 256];
    int         scaled_width, scaled_height;
 
    for (scaled_width = 1; scaled_width < width; scaled_width <<= 1)
        ;
    if (vk_round_down->value && scaled_width > width && mipmap)
        scaled_width >>= 1;
    for (scaled_height = 1; scaled_height < height; scaled_height <<= 1)
        ;
    if (vk_round_down->value && scaled_height > height && mipmap)
        scaled_height >>= 1;
 
    // let people sample down the world textures for speed
    if (mipmap)
    {
        scaled_width >>= (int)vk_picmip->value;
        scaled_height >>= (int)vk_picmip->value;
    }
 
    // don't ever bother with >256 textures
    if (scaled_width > 256)
        scaled_width = 256;
    if (scaled_height > 256)
        scaled_height = 256;
 
    if (scaled_width < 1)
        scaled_width = 1;
    if (scaled_height < 1)
        scaled_height = 1;
 
    upload_width = scaled_width;
    upload_height = scaled_height;
 
    if (scaled_width * scaled_height > sizeof(scaled) / 4)
        ri.Sys_Error(ERR_DROP, "Vk_Upload32: too big");
 
    if (scaled_width == width && scaled_height == height)
    {
        if (!mipmap)
        {
            memcpy(texBuffer, data, scaled_width * scaled_height * 4);
            return 1;
        }
        memcpy(scaled, data, width*height * 4);
    }
    else
        Vk_ResampleTexture(data, width, height, scaled, scaled_width, scaled_height);
 
    Vk_LightScaleTexture(scaled, scaled_width, scaled_height, !mipmap);
 
    memcpy(texBuffer, scaled, sizeof(scaled));
    if (mipmap)
    {
        int     miplevel = 1;
 
        while (scaled_width > 1 || scaled_height > 1)
        {
            scaled_width >>= 1;
            scaled_height >>= 1;
            if (scaled_width < 1)
                scaled_width = 1;
            if (scaled_height < 1)
                scaled_height = 1;
            miplevel++;
        }
 
        return miplevel;
    }
 
    return 1;
}
 
/*
===============
Vk_Upload8
 
Returns number of mip levels
===============
*/
 
uint32_t Vk_Upload8 (byte *data, int width, int height,  qboolean mipmap, qboolean is_sky )
{
    static unsigned trans[512 * 256];
    int         i, s;
    int         p;
 
    s = width * height;
 
    if (s > sizeof(trans) / 4)
        ri.Sys_Error(ERR_DROP, "Vk_Upload8: too large");
 
    for (i = 0; i < s; i++)
    {
        p = data[i];
        trans[i] = d_8to24table[p];
 
        if (p == 255)
        {   // transparent, so scan around for another color
            // to avoid alpha fringes
            // FIXME: do a full flood fill so mips work...
            if (i > width && data[i - width] != 255)
                p = data[i - width];
            else if (i < s - width && data[i + width] != 255)
                p = data[i + width];
            else if (i > 0 && data[i - 1] != 255)
                p = data[i - 1];
            else if (i < s - 1 && data[i + 1] != 255)
                p = data[i + 1];
            else
                p = 0;
            // copy rgb components
            ((byte *)&trans[i])[0] = ((byte *)&d_8to24table[p])[0];
            ((byte *)&trans[i])[1] = ((byte *)&d_8to24table[p])[1];
            ((byte *)&trans[i])[2] = ((byte *)&d_8to24table[p])[2];
        }
    }
 
    return Vk_Upload32(trans, width, height, mipmap);
}
 
 
/*
================
Vk_LoadPic
 
This is also used as an entry point for the generated r_notexture
================
*/
image_t *Vk_LoadPic (char *name, byte *pic, int width, int height, imagetype_t type, int bits, qvksampler_t *samplerType)
{
    image_t     *image;
    int         i;
 
    // find a free image_t
    for (i = 0, image = vktextures; i<numvktextures; i++, image++)
    {
        if (image->vk_texture.image == VK_NULL_HANDLE && !image->scrap)
            break;
    }
    if (i == numvktextures)
    {
        if (numvktextures == MAX_VKTEXTURES)
            ri.Sys_Error(ERR_DROP, "MAX_VKTEXTURES");
        numvktextures++;
    }
    image = &vktextures[i];
 
    if (strlen(name) >= sizeof(image->name))
        ri.Sys_Error(ERR_DROP, "Draw_LoadPic: \"%s\" is too long", name);
    strcpy(image->name, name);
    image->registration_sequence = registration_sequence;
    // zero-clear Vulkan texture handle
    QVVKTEXTURE_CLEAR(image->vk_texture);
    image->width = width;
    image->height = height;
    image->type = type;
 
    if (type == it_skin && bits == 8)
        R_FloodFillSkin(pic, width, height);
 
    // load little pics into the scrap
    if (image->type == it_pic && bits == 8
        && image->width < 64 && image->height < 64)
    {
        int     x, y;
        int     i, j, k;
        int     texnum;
 
        texnum = Scrap_AllocBlock(image->width, image->height, &x, &y);
        if (texnum == -1)
            goto nonscrap;
 
        // copy the texels into the scrap block
        k = 0;
        for (i = 0; i<image->height; i++)
            for (j = 0; j<image->width; j++, k++)
                scrap_texels[texnum][(y + i)*BLOCK_WIDTH + x + j] = pic[k];
        image->scrap = true;
        image->sl = (x + 0.01) / (float)BLOCK_WIDTH;
        image->sh = (x + image->width - 0.01) / (float)BLOCK_WIDTH;
        image->tl = (y + 0.01) / (float)BLOCK_WIDTH;
        image->th = (y + image->height - 0.01) / (float)BLOCK_WIDTH;
        image->upload_width = BLOCK_WIDTH;
        image->upload_height = BLOCK_HEIGHT;
 
        // update scrap data
        Vk_Upload8(scrap_texels[texnum], BLOCK_WIDTH, BLOCK_HEIGHT, false, false);
 
        if (vk_scrapTextures[texnum].image != VK_NULL_HANDLE)
        {
            QVk_UpdateTextureData(&vk_scrapTextures[texnum], (unsigned char*)texBuffer, 0, 0, image->upload_width, image->upload_height);
        }
        else
        {
            QVVKTEXTURE_CLEAR(vk_scrapTextures[texnum]);
            QVk_CreateTexture(&vk_scrapTextures[texnum], (unsigned char*)texBuffer, image->upload_width, image->upload_height, samplerType ? *samplerType : vk_current_sampler);
            QVk_DebugSetObjectName((uint64_t)vk_scrapTextures[texnum].image, VK_OBJECT_TYPE_IMAGE, va("Image: %s", name));
            QVk_DebugSetObjectName((uint64_t)vk_scrapTextures[texnum].imageView, VK_OBJECT_TYPE_IMAGE_VIEW, va("Image View: %s", name));
            QVk_DebugSetObjectName((uint64_t)vk_scrapTextures[texnum].descriptorSet, VK_OBJECT_TYPE_DESCRIPTOR_SET, va("Descriptor Set: %s", name));
            QVk_DebugSetObjectName((uint64_t)vk_scrapTextures[texnum].allocInfo.deviceMemory, VK_OBJECT_TYPE_DEVICE_MEMORY, "Memory: scrap texture");
        }
 
        image->vk_texture = vk_scrapTextures[texnum];
    }
    else
    {
    nonscrap:
        image->scrap = false;
        if (bits == 8)
            image->vk_texture.mipLevels = Vk_Upload8(pic, width, height, (image->type != it_pic && image->type != it_sky), image->type == it_sky);
        else
            image->vk_texture.mipLevels = Vk_Upload32((unsigned *)pic, width, height, (image->type != it_pic && image->type != it_sky));
 
        image->upload_width = upload_width;     // after power of 2 and scales
        image->upload_height = upload_height;
        image->sl = 0;
        image->sh = 1;
        image->tl = 0;
        image->th = 1;
 
        QVk_CreateTexture(&image->vk_texture, (unsigned char*)texBuffer, image->upload_width, image->upload_height, samplerType ? *samplerType : vk_current_sampler);
        QVk_DebugSetObjectName((uint64_t)image->vk_texture.image, VK_OBJECT_TYPE_IMAGE, va("Image: %s", name));
        QVk_DebugSetObjectName((uint64_t)image->vk_texture.imageView, VK_OBJECT_TYPE_IMAGE_VIEW, va("Image View: %s", name));
        QVk_DebugSetObjectName((uint64_t)image->vk_texture.descriptorSet, VK_OBJECT_TYPE_DESCRIPTOR_SET, va("Descriptor Set: %s", name));
        QVk_DebugSetObjectName((uint64_t)image->vk_texture.allocInfo.deviceMemory, VK_OBJECT_TYPE_DEVICE_MEMORY, "Memory: game textures");
    }
 
    return image;
}
 
 
/*
================
Vk_LoadWal
================
*/
image_t *Vk_LoadWal (char *name)
{
    miptex_t    *mt;
    int         width, height, ofs;
    image_t     *image;
 
    ri.FS_LoadFile (name, (void **)&mt);
    if (!mt)
    {
        ri.Con_Printf (PRINT_ALL, "Vk_FindImage: can't load %s\n", name);
        return r_notexture;
    }
 
    width = LittleLong (mt->width);
    height = LittleLong (mt->height);
    ofs = LittleLong (mt->offsets[0]);
 
    image = Vk_LoadPic (name, (byte *)mt + ofs, width, height, it_wall, 8, NULL);
 
    ri.FS_FreeFile ((void *)mt);
 
    return image;
}
 
/*
===============
Vk_FindImage
 
Finds or loads the given image
===============
*/
image_t *Vk_FindImage (char *name, imagetype_t type, qvksampler_t *samplerType)
{
    image_t *image;
    int     i, len;
    byte    *pic, *palette;
    int     width, height;
 
    if (!name)
        return NULL;    //  ri.Sys_Error (ERR_DROP, "Vk_FindImage: NULL name");
    len = (int)strlen(name);
    if (len<5)
        return NULL;    //  ri.Sys_Error (ERR_DROP, "Vk_FindImage: bad name: %s", name);
 
    // look for it
    for (i=0, image=vktextures ; i<numvktextures ; i++,image++)
    {
        if (!strcmp(name, image->name))
        {
            image->registration_sequence = registration_sequence;
            return image;
        }
    }
 
    //
    // load the pic from disk
    //
    pic = NULL;
    palette = NULL;
    if (!strcmp(name+len-4, ".pcx"))
    {
        LoadPCX (name, &pic, &palette, &width, &height);
        if (!pic)
            return NULL; // ri.Sys_Error (ERR_DROP, "Vk_FindImage: can't load %s", name);
        image = Vk_LoadPic (name, pic, width, height, type, 8, samplerType);
    }
    else if (!strcmp(name+len-4, ".wal"))
    {
        image = Vk_LoadWal (name);
    }
    else if (!strcmp(name+len-4, ".tga"))
    {
        LoadTGA (name, &pic, &width, &height);
        if (!pic)
            return NULL; // ri.Sys_Error (ERR_DROP, "Vk_FindImage: can't load %s", name);
        image = Vk_LoadPic (name, pic, width, height, type, 32, samplerType);
    }
    else
        return NULL;    //  ri.Sys_Error (ERR_DROP, "Vk_FindImage: bad extension on: %s", name);
 
 
    if (pic)
        free(pic);
    if (palette)
        free(palette);
 
    return image;
}
 
 
 
/*
===============
R_RegisterSkin
===============
*/
struct image_s *R_RegisterSkin (char *name)
{
    return Vk_FindImage (name, it_skin, NULL);
}
 
 
/*
================
Vk_FreeUnusedImages
 
Any image that was not touched on this registration sequence
will be freed.
================
*/
void Vk_FreeUnusedImages (void)
{
    int     i;
    image_t *image;
 
    // never free r_notexture or particle texture
    r_notexture->registration_sequence = registration_sequence;
    r_particletexture->registration_sequence = registration_sequence;
 
    for (i = 0, image = vktextures; i < numvktextures; i++, image++)
    {
        if (image->registration_sequence == registration_sequence)
            continue;       // used this sequence
        if (!image->registration_sequence)
            continue;       // free image_t slot
        if (image->type == it_pic)
            continue;       // don't free pics
        // free it
        QVk_ReleaseTexture(&image->vk_texture);
        memset(image, 0, sizeof(*image));
    }
}
 
 
/*
===============
Draw_GetPalette
===============
*/
int Draw_GetPalette (void)
{
    int     i;
    int     r, g, b;
    unsigned    v;
    byte    *pic, *pal;
    int     width, height;
 
    // get the palette
 
    LoadPCX ("pics/colormap.pcx", &pic, &pal, &width, &height);
    if (!pal)
        ri.Sys_Error (ERR_FATAL, "Couldn't load pics/colormap.pcx");
 
    for (i=0 ; i<256 ; i++)
    {
        r = pal[i*3+0];
        g = pal[i*3+1];
        b = pal[i*3+2];
        
        v = (255<<24) + (r<<0) + (g<<8) + (b<<16);
        d_8to24table[i] = LittleLong(v);
    }
 
    d_8to24table[255] &= LittleLong(0xffffff);  // 255 is transparent
 
    free (pic);
    free (pal);
 
    return 0;
}
 
 
/*
===============
Vk_InitImages
===============
*/
void    Vk_InitImages (void)
{
    int     i, j;
    float   g = vid_gamma->value;
 
    registration_sequence = 1;
 
    // init intensity conversions
    intensity = ri.Cvar_Get("intensity", "2", 0);
 
    if (intensity->value <= 1)
        ri.Cvar_Set("intensity", "1");
 
    vk_state.inverse_intensity = 1 / intensity->value;
 
    Draw_GetPalette();
 
    for (i = 0; i < 256; i++)
    {
        if (g == 1)
        {
            gammatable[i] = i;
        }
        else
        {
            float inf;
 
            inf = 255 * pow((i + 0.5) / 255.5, g) + 0.5;
            if (inf < 0)
                inf = 0;
            if (inf > 255)
                inf = 255;
            gammatable[i] = inf;
        }
    }
 
    for (i = 0; i<256; i++)
    {
        j = i * intensity->value;
        if (j > 255)
            j = 255;
        intensitytable[i] = j;
    }
}
 
/*
===============
Vk_ShutdownImages
===============
*/
void    Vk_ShutdownImages (void)
{
    int     i;
    image_t *image;
 
    for (i = 0, image = vktextures; i<numvktextures; i++, image++)
    {
        if (!image->registration_sequence)
            continue;       // free image_t slot
        // free it (scraps are stored in a separate Vulkan buffer)
        if (!image->scrap)
            QVk_ReleaseTexture(&image->vk_texture);
        memset(image, 0, sizeof(*image));
    }
 
    QVk_ReleaseTexture(&vk_rawTexture);
 
    for(i = 0; i < MAX_SCRAPS; i++)
        QVk_ReleaseTexture(&vk_scrapTextures[i]);
 
    for(i = 0; i < MAX_LIGHTMAPS*2; i++)
        QVk_ReleaseTexture(&vk_state.lightmap_textures[i]);
}