#include "vk_textures.h" #include "vk_common.h" #include "vk_core.h" #include "vk_const.h" #include "vk_descriptor.h" #include "xash3d_mathlib.h" #include "crtlib.h" #include "crclib.h" #include "com_strings.h" #include "eiface.h" #include #include #define TEXTURES_HASH_SIZE (MAX_TEXTURES >> 2) static vk_texture_t vk_textures[MAX_TEXTURES]; static vk_texture_t* vk_texturesHashTable[TEXTURES_HASH_SIZE]; static uint vk_numTextures; vk_textures_global_t tglob; static void VK_CreateInternalTextures(void); void initTextures( void ) { memset( vk_textures, 0, sizeof( vk_textures )); memset( vk_texturesHashTable, 0, sizeof( vk_texturesHashTable )); vk_numTextures = 0; // create unused 0-entry Q_strncpy( vk_textures->name, "*unused*", sizeof( vk_textures->name )); vk_textures->hashValue = COM_HashKey( vk_textures->name, TEXTURES_HASH_SIZE ); vk_textures->nextHash = vk_texturesHashTable[vk_textures->hashValue]; vk_texturesHashTable[vk_textures->hashValue] = vk_textures; vk_numTextures = 1; /* FIXME // validate cvars R_SetTextureParameters(); */ VK_CreateInternalTextures(); /* FIXME gEngine.Cmd_AddCommand( "texturelist", R_TextureList_f, "display loaded textures list" ); */ } void destroyTextures( void ) { for( unsigned int i = 0; i < vk_numTextures; i++ ) VK_FreeTexture( i ); //memset( tr.lightmapTextures, 0, sizeof( tr.lightmapTextures )); memset( vk_texturesHashTable, 0, sizeof( vk_texturesHashTable )); memset( vk_textures, 0, sizeof( vk_textures )); vk_numTextures = 0; } vk_texture_t *findTexture(int index) { ASSERT(index >= 0); ASSERT(index < MAX_TEXTURES); return vk_textures + index; } static vk_texture_t *Common_AllocTexture( const char *name, texFlags_t flags ) { vk_texture_t *tex; uint i; // find a free texture_t slot for( i = 0, tex = vk_textures; i < vk_numTextures; i++, tex++ ) if( !tex->name[0] ) break; if( i == vk_numTextures ) { if( vk_numTextures == MAX_TEXTURES ) gEngine.Host_Error( "VK_AllocTexture: MAX_TEXTURES limit exceeds\n" ); vk_numTextures++; } tex = &vk_textures[i]; // copy initial params Q_strncpy( tex->name, name, sizeof( tex->name )); if( FBitSet( flags, TF_SKYSIDE )) tex->texnum = tglob.skyboxbasenum++; else tex->texnum = i; // texnum is used for fast acess into vk_textures array too tex->flags = flags; // add to hash table tex->hashValue = COM_HashKey( name, TEXTURES_HASH_SIZE ); tex->nextHash = vk_texturesHashTable[tex->hashValue]; vk_texturesHashTable[tex->hashValue] = tex; return tex; } static qboolean Common_CheckTexName( const char *name ) { int len; if( !COM_CheckString( name )) return false; len = Q_strlen( name ); // because multi-layered textures can exceed name string if( len >= sizeof( vk_textures->name )) { gEngine.Con_Printf( S_ERROR "LoadTexture: too long name %s (%d)\n", name, len ); return false; } return true; } static vk_texture_t *Common_TextureForName( const char *name ) { vk_texture_t *tex; uint hash; // find the texture in array hash = COM_HashKey( name, TEXTURES_HASH_SIZE ); for( tex = vk_texturesHashTable[hash]; tex != NULL; tex = tex->nextHash ) { if( !Q_stricmp( tex->name, name )) return tex; } return NULL; } static rgbdata_t *Common_FakeImage( int width, int height, int depth, int flags ) { static byte data2D[1024]; // 16x16x4 static rgbdata_t r_image; // also use this for bad textures, but without alpha r_image.width = Q_max( 1, width ); r_image.height = Q_max( 1, height ); r_image.depth = Q_max( 1, depth ); r_image.flags = flags; r_image.type = PF_RGBA_32; r_image.size = r_image.width * r_image.height * r_image.depth * 4; r_image.buffer = (r_image.size > sizeof( data2D )) ? NULL : data2D; r_image.palette = NULL; r_image.numMips = 1; r_image.encode = 0; if( FBitSet( r_image.flags, IMAGE_CUBEMAP )) r_image.size *= 6; memset( data2D, 0xFF, sizeof( data2D )); return &r_image; } /* =============== GL_ProcessImage do specified actions on pixels =============== */ static void VK_ProcessImage( vk_texture_t *tex, rgbdata_t *pic ) { float emboss_scale = 0.0f; uint img_flags = 0; // force upload texture as RGB or RGBA (detail textures requires this) if( tex->flags & TF_FORCE_COLOR ) pic->flags |= IMAGE_HAS_COLOR; if( pic->flags & IMAGE_HAS_ALPHA ) tex->flags |= TF_HAS_ALPHA; //FIXME provod: ??? tex->encode = pic->encode; // share encode method if( ImageDXT( pic->type )) { if( !pic->numMips ) tex->flags |= TF_NOMIPMAP; // disable mipmapping by user request // clear all the unsupported flags tex->flags &= ~TF_KEEP_SOURCE; } else { // copy flag about luma pixels if( pic->flags & IMAGE_HAS_LUMA ) tex->flags |= TF_HAS_LUMA; if( pic->flags & IMAGE_QUAKEPAL ) tex->flags |= TF_QUAKEPAL; // create luma texture from quake texture if( tex->flags & TF_MAKELUMA ) { img_flags |= IMAGE_MAKE_LUMA; tex->flags &= ~TF_MAKELUMA; } if( tex->flags & TF_ALLOW_EMBOSS ) { img_flags |= IMAGE_EMBOSS; tex->flags &= ~TF_ALLOW_EMBOSS; } /* FIXME provod: ??? if( !FBitSet( tex->flags, TF_IMG_UPLOADED ) && FBitSet( tex->flags, TF_KEEP_SOURCE )) tex->original = gEngfuncs.FS_CopyImage( pic ); // because current pic will be expanded to rgba */ // we need to expand image into RGBA buffer if( pic->type == PF_INDEXED_24 || pic->type == PF_INDEXED_32 ) img_flags |= IMAGE_FORCE_RGBA; /* FIXME provod: ??? // dedicated server doesn't register this variable if( gl_emboss_scale != NULL ) emboss_scale = gl_emboss_scale->value; */ // processing image before uploading (force to rgba, make luma etc) if( pic->buffer ) gEngine.Image_Process( &pic, 0, 0, img_flags, emboss_scale ); if( FBitSet( tex->flags, TF_LUMINANCE )) ClearBits( pic->flags, IMAGE_HAS_COLOR ); } } static void VK_CreateInternalTextures( void ) { int dx2, dy, d; int x, y; rgbdata_t *pic; // emo-texture from quake1 pic = Common_FakeImage( 16, 16, 1, IMAGE_HAS_COLOR ); for( y = 0; y < 16; y++ ) { for( x = 0; x < 16; x++ ) { if(( y < 8 ) ^ ( x < 8 )) ((uint *)pic->buffer)[y*16+x] = 0xFFFF00FF; else ((uint *)pic->buffer)[y*16+x] = 0xFF000000; } } tglob.defaultTexture = VK_LoadTextureInternal( REF_DEFAULT_TEXTURE, pic, TF_COLORMAP ); // particle texture from quake1 pic = Common_FakeImage( 16, 16, 1, IMAGE_HAS_COLOR|IMAGE_HAS_ALPHA ); for( x = 0; x < 16; x++ ) { dx2 = x - 8; dx2 = dx2 * dx2; for( y = 0; y < 16; y++ ) { dy = y - 8; d = 255 - 35 * sqrt( dx2 + dy * dy ); pic->buffer[( y * 16 + x ) * 4 + 3] = bound( 0, d, 255 ); } } tglob.particleTexture = VK_LoadTextureInternal( REF_PARTICLE_TEXTURE, pic, TF_CLAMP ); // white texture pic = Common_FakeImage( 4, 4, 1, IMAGE_HAS_COLOR ); for( x = 0; x < 16; x++ ) ((uint *)pic->buffer)[x] = 0xFFFFFFFF; tglob.whiteTexture = VK_LoadTextureInternal( REF_WHITE_TEXTURE, pic, TF_COLORMAP ); // gray texture pic = Common_FakeImage( 4, 4, 1, IMAGE_HAS_COLOR ); for( x = 0; x < 16; x++ ) ((uint *)pic->buffer)[x] = 0xFF7F7F7F; tglob.grayTexture = VK_LoadTextureInternal( REF_GRAY_TEXTURE, pic, TF_COLORMAP ); // black texture pic = Common_FakeImage( 4, 4, 1, IMAGE_HAS_COLOR ); for( x = 0; x < 16; x++ ) ((uint *)pic->buffer)[x] = 0xFF000000; tglob.blackTexture = VK_LoadTextureInternal( REF_BLACK_TEXTURE, pic, TF_COLORMAP ); // cinematic dummy pic = Common_FakeImage( 640, 100, 1, IMAGE_HAS_COLOR ); tglob.cinTexture = VK_LoadTextureInternal( "*cintexture", pic, TF_NOMIPMAP|TF_CLAMP ); } static VkFormat VK_GetFormat(pixformat_t format) { switch(format) { case PF_RGBA_32: return VK_FORMAT_R8G8B8A8_UNORM; default: gEngine.Con_Printf(S_WARN "FIXME unsupported pixformat_t %d\n", format); return VK_FORMAT_UNDEFINED; } } static qboolean VK_UploadTexture(vk_texture_t *tex, rgbdata_t *pic) { const VkFormat format = VK_GetFormat(pic->type); const VkImageTiling tiling = VK_IMAGE_TILING_OPTIMAL; const VkImageUsageFlags usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT; gEngine.Con_Reportf("Uploading texture %s\n", tex->name); if (!pic->buffer) return false; // 1. Create VkImage w/ usage = DST|SAMPLED, layout=UNDEFINED { VkImageCreateInfo image_create_info = { .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO, .imageType = VK_IMAGE_TYPE_2D, .extent.width = pic->width, .extent.height = pic->height, .extent.depth = 1, .format = format, .mipLevels = 1, // TODO pic->numMips, .arrayLayers = 1, .tiling = tiling, .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED, .usage = usage, .samples = VK_SAMPLE_COUNT_1_BIT, .sharingMode = VK_SHARING_MODE_EXCLUSIVE, }; XVK_CHECK(vkCreateImage(vk_core.device, &image_create_info, NULL, &tex->vk.image)); if (vk_core.debug) { VkDebugUtilsObjectNameInfoEXT debug_name = { .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_OBJECT_NAME_INFO_EXT, .objectHandle = (uint64_t)tex->vk.image, .objectType = VK_OBJECT_TYPE_IMAGE, .pObjectName = tex->name, }; XVK_CHECK(vkSetDebugUtilsObjectNameEXT(vk_core.device, &debug_name)); } } // 2. Alloc mem for VkImage and bind it (DEV_LOCAL) { VkMemoryRequirements memreq; vkGetImageMemoryRequirements(vk_core.device, tex->vk.image, &memreq); tex->vk.device_memory = allocateDeviceMemory(memreq, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT, 0); XVK_CHECK(vkBindImageMemory(vk_core.device, tex->vk.image, tex->vk.device_memory.device_memory, tex->vk.device_memory.offset)); } { const size_t staging_offset = 0; // TODO multiple staging buffer users params.staging->ptr // 5. Create/get cmdbuf for transitions VkCommandBufferBeginInfo beginfo = { .sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO, .flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT, }; // 5.1 upload buf -> image:layout:DST // 5.1.1 transitionToLayout(UNDEFINED -> DST) VkImageMemoryBarrier image_barrier = { .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, .image = tex->vk.image, .srcAccessMask = 0, .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT, .oldLayout = VK_IMAGE_LAYOUT_UNDEFINED, .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, .subresourceRange = (VkImageSubresourceRange) { .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, .baseMipLevel = 0, .levelCount = 1, // TODO .baseArrayLayer = 0, .layerCount = 1, }}; XVK_CHECK(vkBeginCommandBuffer(vk_core.cb_tex, &beginfo)); vkCmdPipelineBarrier(vk_core.cb_tex, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, NULL, 0, NULL, 1, &image_barrier); // 5.1.2 copyBufferToImage for all mip levels //for (int i = 0; i < 1/* TODO params.mipmaps_count*/; ++i) { VkBufferImageCopy region = {0}; region.bufferOffset = /*TODO mip->offset +*/ staging_offset; region.bufferRowLength = 0; region.bufferImageHeight = 0; region.imageSubresource = (VkImageSubresourceLayers){ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, .mipLevel = 0, // TODO mip->mip_level, .baseArrayLayer = 0, .layerCount = 1, }; region.imageExtent = (VkExtent3D){ .width = pic->width, // TODO mip->width, .height = pic->height, // TODO mip->height, .depth = 1, }; memcpy(((uint8_t*)vk_core.staging.mapped) + staging_offset, pic->buffer, pic->size); vkCmdCopyBufferToImage(vk_core.cb_tex, vk_core.staging.buffer, tex->vk.image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, ®ion); } // 5.2 image:layout:DST -> image:layout:SAMPLED // 5.2.1 transitionToLayout(DST -> SHADER_READ_ONLY) image_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT; image_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT; image_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL; image_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; image_barrier.subresourceRange = (VkImageSubresourceRange){ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, .baseMipLevel = 0, .levelCount = 1, // FIXME params.mipmaps_count, .baseArrayLayer = 0, .layerCount = 1, }; vkCmdPipelineBarrier(vk_core.cb_tex, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, 0, 0, NULL, 0, NULL, 1, &image_barrier); XVK_CHECK(vkEndCommandBuffer(vk_core.cb_tex)); } { VkSubmitInfo subinfo = {.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO}; subinfo.commandBufferCount = 1; subinfo.pCommandBuffers = &vk_core.cb_tex; XVK_CHECK(vkQueueSubmit(vk_core.queue, 1, &subinfo, VK_NULL_HANDLE)); XVK_CHECK(vkQueueWaitIdle(vk_core.queue)); } { VkImageViewCreateInfo ivci = {.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO}; ivci.viewType = VK_IMAGE_VIEW_TYPE_2D; ivci.format = format; ivci.image = tex->vk.image; ivci.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; ivci.subresourceRange.baseMipLevel = 0; ivci.subresourceRange.levelCount = 1; // FIXME params.mipmaps_count; ivci.subresourceRange.baseArrayLayer = 0; ivci.subresourceRange.layerCount = 1; ivci.components = (VkComponentMapping){0, 0, 0, (pic->flags & IMAGE_HAS_ALPHA) ? 0 : VK_COMPONENT_SWIZZLE_ONE}; XVK_CHECK(vkCreateImageView(vk_core.device, &ivci, NULL, &tex->vk.image_view)); } if (vk_core.debug) { VkDebugUtilsObjectNameInfoEXT debug_name = { .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_OBJECT_NAME_INFO_EXT, .objectHandle = (uint64_t)tex->vk.image_view, .objectType = VK_OBJECT_TYPE_IMAGE_VIEW, .pObjectName = tex->name, }; XVK_CHECK(vkSetDebugUtilsObjectNameEXT(vk_core.device, &debug_name)); } // TODO how should we approach this: // - per-texture desc sets can be inconvenient if texture is used in different incompatible contexts // - update descriptor sets in batch? if (vk_desc.next_free != MAX_TEXTURES) { VkDescriptorImageInfo dii_tex = { .imageView = tex->vk.image_view, .imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL, }; VkWriteDescriptorSet wds[] = { { .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, .dstBinding = 0, .dstArrayElement = 0, .descriptorCount = 1, .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, .pImageInfo = &dii_tex, }}; wds[0].dstSet = tex->vk.descriptor = vk_desc.sets[vk_desc.next_free++]; vkUpdateDescriptorSets(vk_core.device, ARRAYSIZE(wds), wds, 0, NULL); } else { tex->vk.descriptor = VK_NULL_HANDLE; } return true; } ///////////// Render API funcs ///////////// // Texture tools int VK_FindTexture( const char *name ) { vk_texture_t *tex; if( !Common_CheckTexName( name )) return 0; // see if already loaded if(( tex = Common_TextureForName( name ))) return (tex - vk_textures); return 0; } const char* VK_TextureName( unsigned int texnum ) { ASSERT( texnum >= 0 && texnum < MAX_TEXTURES ); return vk_textures[texnum].name; } const byte* VK_TextureData( unsigned int texnum ) { gEngine.Con_Printf(S_WARN "VK FIXME: %s\n", __FUNCTION__); // We don't store original texture data // TODO do we need to? return NULL; } int VK_LoadTexture( const char *name, const byte *buf, size_t size, int flags ) { vk_texture_t *tex; rgbdata_t *pic; uint picFlags = 0; if( !Common_CheckTexName( name )) return 0; // see if already loaded if(( tex = Common_TextureForName( name ))) return (tex - vk_textures); if( FBitSet( flags, TF_NOFLIP_TGA )) SetBits( picFlags, IL_DONTFLIP_TGA ); if( FBitSet( flags, TF_KEEP_SOURCE ) && !FBitSet( flags, TF_EXPAND_SOURCE )) SetBits( picFlags, IL_KEEP_8BIT ); // set some image flags gEngine.Image_SetForceFlags( picFlags ); pic = gEngine.FS_LoadImage( name, buf, size ); if( !pic ) return 0; // couldn't loading image // allocate the new one tex = Common_AllocTexture( name, flags ); // upload texture VK_ProcessImage( tex, pic ); if( !VK_UploadTexture( tex, pic )) { memset( tex, 0, sizeof( vk_texture_t )); gEngine.FS_FreeImage( pic ); // release source texture return 0; } /* FIXME VK_ApplyTextureParams( tex ); // update texture filter, wrap etc */ tex->width = pic->width; tex->height = pic->height; gEngine.FS_FreeImage( pic ); // release source texture // NOTE: always return texnum as index in array or engine will stop work !!! return tex - vk_textures; } int VK_CreateTexture( const char *name, int width, int height, const void *buffer, texFlags_t flags ) { gEngine.Con_Printf("VK FIXME: %s\n", __FUNCTION__); return 0; } int VK_LoadTextureArray( const char **names, int flags ) { gEngine.Con_Printf("VK FIXME: %s\n", __FUNCTION__); return 0; } int VK_CreateTextureArray( const char *name, int width, int height, int depth, const void *buffer, texFlags_t flags ) { gEngine.Con_Printf("VK FIXME: %s\n", __FUNCTION__); return 0; } void VK_FreeTexture( unsigned int texnum ) { vk_texture_t *tex; vk_texture_t **prev; vk_texture_t *cur; if( texnum <= 0 ) return; tex = vk_textures + texnum; ASSERT( tex != NULL ); // already freed? if( !tex->vk.image ) return; // debug if( !tex->name[0] ) { gEngine.Con_Printf( S_ERROR "GL_DeleteTexture: trying to free unnamed texture with index %u\n", texnum ); return; } // remove from hash table prev = &vk_texturesHashTable[tex->hashValue]; while( 1 ) { cur = *prev; if( !cur ) break; if( cur == tex ) { *prev = cur->nextHash; break; } prev = &cur->nextHash; } /* // release source if( tex->original ) gEngfuncs.FS_FreeImage( tex->original ); */ vkDestroyImageView(vk_core.device, tex->vk.image_view, NULL); vkDestroyImage(vk_core.device, tex->vk.image, NULL); freeDeviceMemory(&tex->vk.device_memory); memset(tex, 0, sizeof(*tex)); } int VK_LoadTextureFromBuffer( const char *name, rgbdata_t *pic, texFlags_t flags, qboolean update ) { vk_texture_t *tex; if( !Common_CheckTexName( name )) return 0; // see if already loaded if(( tex = Common_TextureForName( name )) && !update ) return (tex - vk_textures); // couldn't loading image if( !pic ) return 0; if( update ) { if( tex == NULL ) gEngine.Host_Error( "VK_LoadTextureFromBuffer: couldn't find texture %s for update\n", name ); SetBits( tex->flags, flags ); } else { // allocate the new one tex = Common_AllocTexture( name, flags ); } VK_ProcessImage( tex, pic ); if( !VK_UploadTexture( tex, pic )) { memset( tex, 0, sizeof( vk_texture_t )); return 0; } /* FIXME VK_ApplyTextureParams( tex ); // update texture filter, wrap etc */ return (tex - vk_textures); } vk_image_t VK_ImageCreate(uint32_t width, uint32_t height, VkFormat format, VkImageTiling tiling, VkImageUsageFlags usage) { vk_image_t image; VkMemoryRequirements memreq; VkImageViewCreateInfo ivci = {.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO }; VkImageCreateInfo ici = { .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO, .imageType = VK_IMAGE_TYPE_2D, .extent.width = width, .extent.height = height, .extent.depth = 1, .mipLevels = 1, .arrayLayers = 1, .format = format, .tiling = tiling, .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED, .usage = usage, .samples = VK_SAMPLE_COUNT_1_BIT, .sharingMode = VK_SHARING_MODE_EXCLUSIVE, }; XVK_CHECK(vkCreateImage(vk_core.device, &ici, NULL, &image.image)); vkGetImageMemoryRequirements(vk_core.device, image.image, &memreq); image.devmem = allocateDeviceMemory(memreq, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT, 0); XVK_CHECK(vkBindImageMemory(vk_core.device, image.image, image.devmem.device_memory, 0)); ivci.viewType = VK_IMAGE_VIEW_TYPE_2D; ivci.format = ici.format; ivci.image = image.image; ivci.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; ivci.subresourceRange.levelCount = 1; ivci.subresourceRange.layerCount = 1; XVK_CHECK(vkCreateImageView(vk_core.device, &ivci, NULL, &image.view)); return image; } void VK_ImageDestroy(vk_image_t *img) { vkDestroyImageView(vk_core.device, img->view, NULL); vkDestroyImage(vk_core.device, img->image, NULL); freeDeviceMemory(&img->devmem); *img = (vk_image_t){0}; }