Merge pull request #5279 from stenzek/compressed-custom-textures

Native compressed custom texture support

Source/Core/Common/Common.vcxproj

@@ -93,6 +93,7 @@
     <ClInclude Include="GL\GLExtensions\ARB_uniform_buffer_object.h" />
     <ClInclude Include="GL\GLExtensions\ARB_vertex_array_object.h" />
     <ClInclude Include="GL\GLExtensions\ARB_viewport_array.h" />
+    <ClInclude Include="GL\GLExtensions\EXT_texture_compression_s3tc.h" />
     <ClInclude Include="GL\GLExtensions\EXT_texture_filter_anisotropic.h" />
     <ClInclude Include="GL\GLExtensions\GLExtensions.h" />
     <ClInclude Include="GL\GLExtensions\gl_1_1.h" />

Source/Core/Common/Common.vcxproj.filters

@@ -248,6 +248,9 @@
     <ClInclude Include="GL\GLExtensions\ARB_compute_shader.h">
       <Filter>GL\GLExtensions</Filter>
     </ClInclude>
+    <ClInclude Include="GL\GLExtensions\EXT_texture_compression_s3tc.h">
+      <Filter>GL\GLExtensions</Filter>
+    </ClInclude>
   </ItemGroup>
   <ItemGroup>
     <ClCompile Include="CDUtils.cpp" />

Source/Core/Common/GL/GLExtensions/EXT_texture_compression_s3tc.h

@@ -0,0 +1,29 @@
+/*
+** Copyright (c) 2013-2015 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a
+** copy of this software and/or associated documentation files (the
+** "Materials"), to deal in the Materials without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Materials, and to
+** permit persons to whom the Materials are furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be included
+** in all copies or substantial portions of the Materials.
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+*/
+
+#include "Common/GL/GLExtensions/gl_common.h"
+
+#define GL_COMPRESSED_RGB_S3TC_DXT1_EXT 0x83F0
+#define GL_COMPRESSED_RGBA_S3TC_DXT1_EXT 0x83F1
+#define GL_COMPRESSED_RGBA_S3TC_DXT3_EXT 0x83F2
+#define GL_COMPRESSED_RGBA_S3TC_DXT5_EXT 0x83F3

Source/Core/Common/GL/GLExtensions/GLExtensions.h

@@ -31,6 +31,7 @@
 #include "Common/GL/GLExtensions/ARB_uniform_buffer_object.h"
 #include "Common/GL/GLExtensions/ARB_vertex_array_object.h"
 #include "Common/GL/GLExtensions/ARB_viewport_array.h"
+#include "Common/GL/GLExtensions/EXT_texture_compression_s3tc.h"
 #include "Common/GL/GLExtensions/EXT_texture_filter_anisotropic.h"
 #include "Common/GL/GLExtensions/HP_occlusion_test.h"
 #include "Common/GL/GLExtensions/KHR_debug.h"

Source/Core/VideoBackends/D3D/D3DBase.cpp

@@ -239,6 +239,19 @@ D3D_FEATURE_LEVEL GetFeatureLevel(IDXGIAdapter* adapter)
   return feat_level;
 }
 
+static bool SupportsS3TCTextures(ID3D11Device* device)
+{
+  UINT bc1_support, bc2_support, bc3_support;
+  if (FAILED(device->CheckFormatSupport(DXGI_FORMAT_BC1_UNORM, &bc1_support)) ||
+      FAILED(device->CheckFormatSupport(DXGI_FORMAT_BC2_UNORM, &bc2_support)) ||
+      FAILED(device->CheckFormatSupport(DXGI_FORMAT_BC3_UNORM, &bc3_support)))
+  {
+    return false;
+  }
+
+  return ((bc1_support & bc2_support & bc3_support) & D3D11_FORMAT_SUPPORT_TEXTURE2D) != 0;
+}
+
 HRESULT Create(HWND wnd)
 {
   hWnd = wnd;
@@ -427,6 +440,7 @@ HRESULT Create(HWND wnd)
   UINT format_support;
   device->CheckFormatSupport(DXGI_FORMAT_B8G8R8A8_UNORM, &format_support);
   bgra_textures_supported = (format_support & D3D11_FORMAT_SUPPORT_TEXTURE2D) != 0;
+  g_Config.backend_info.bSupportsST3CTextures = SupportsS3TCTextures(device);
 
   stateman = new StateManager;
   return S_OK;

Source/Core/VideoBackends/D3D/TextureCache.cpp

@@ -7,6 +7,7 @@
 #include <algorithm>
 #include <memory>
 
+#include "Common/Assert.h"
 #include "Common/CommonTypes.h"
 #include "Common/Logging/Log.h"
 
@@ -29,6 +30,25 @@ static std::unique_ptr<PSTextureEncoder> g_encoder;
 const size_t MAX_COPY_BUFFERS = 32;
 ID3D11Buffer* efbcopycbuf[MAX_COPY_BUFFERS] = {0};
 
+static DXGI_FORMAT GetDXGIFormatForHostFormat(HostTextureFormat format)
+{
+  switch (format)
+  {
+  case HostTextureFormat::DXT1:
+    return DXGI_FORMAT_BC1_UNORM;
+
+  case HostTextureFormat::DXT3:
+    return DXGI_FORMAT_BC2_UNORM;
+
+  case HostTextureFormat::DXT5:
+    return DXGI_FORMAT_BC3_UNORM;
+
+  case HostTextureFormat::RGBA8:
+  default:
+    return DXGI_FORMAT_R8G8B8A8_UNORM;
+  }
+}
+
 TextureCache::TCacheEntry::~TCacheEntry()
 {
   texture->Release();
@@ -41,6 +61,11 @@ void TextureCache::TCacheEntry::Bind(unsigned int stage)
 
 bool TextureCache::TCacheEntry::Save(const std::string& filename, unsigned int level)
 {
+  // We can't dump compressed textures currently (it would mean drawing them to a RGBA8
+  // framebuffer, and saving that). TextureCache does not call Save for custom textures
+  // anyway, so this is fine for now.
+  _assert_(config.format == HostTextureFormat::RGBA8);
+
   // Create a staging/readback texture with the dimensions of the specified mip level.
   u32 mip_width = std::max(config.width >> level, 1u);
   u32 mip_height = std::max(config.height >> level, 1u);
@@ -129,28 +154,30 @@ void TextureCache::TCacheEntry::CopyRectangleFromTexture(const TCacheEntryBase*
   g_renderer->RestoreAPIState();
 }
 
-void TextureCache::TCacheEntry::Load(const u8* buffer, u32 width, u32 height, u32 expanded_width,
+void TextureCache::TCacheEntry::Load(u32 level, u32 width, u32 height, u32 row_length,
-                                     u32 level)
+                                     const u8* buffer, size_t buffer_size)
 {
-  unsigned int src_pitch = 4 * expanded_width;
+  size_t src_pitch = CalculateHostTextureLevelPitch(config.format, row_length);
-  D3D::context->UpdateSubresource(texture->GetTex(), level, nullptr, buffer, src_pitch, 0);
+  D3D::context->UpdateSubresource(texture->GetTex(), level, nullptr, buffer,
+                                  static_cast<UINT>(src_pitch), 0);
 }
 
 TextureCacheBase::TCacheEntryBase* TextureCache::CreateTexture(const TCacheEntryConfig& config)
 {
+  DXGI_FORMAT dxgi_format = GetDXGIFormatForHostFormat(config.format);
   if (config.rendertarget)
   {
     return new TCacheEntry(
-        config, D3DTexture2D::Create(
+        config, D3DTexture2D::Create(config.width, config.height,
-                    config.width, config.height, (D3D11_BIND_FLAG)((int)D3D11_BIND_RENDER_TARGET |
+                                     (D3D11_BIND_FLAG)((int)D3D11_BIND_RENDER_TARGET |
                                                        (int)D3D11_BIND_SHADER_RESOURCE),
-                    D3D11_USAGE_DEFAULT, DXGI_FORMAT_R8G8B8A8_UNORM, 1, config.layers));
+                                     D3D11_USAGE_DEFAULT, dxgi_format, 1, config.layers));
   }
   else
   {
     const D3D11_TEXTURE2D_DESC texdesc =
-        CD3D11_TEXTURE2D_DESC(DXGI_FORMAT_R8G8B8A8_UNORM, config.width, config.height, 1,
+        CD3D11_TEXTURE2D_DESC(dxgi_format, config.width, config.height, 1, config.levels,
-                              config.levels, D3D11_BIND_SHADER_RESOURCE, D3D11_USAGE_DEFAULT, 0);
+                              D3D11_BIND_SHADER_RESOURCE, D3D11_USAGE_DEFAULT, 0);
 
     ID3D11Texture2D* pTexture;
     const HRESULT hr = D3D::device->CreateTexture2D(&texdesc, nullptr, &pTexture);

Source/Core/VideoBackends/D3D/TextureCache.h

@@ -30,7 +30,8 @@ private:
                                   const MathUtil::Rectangle<int>& srcrect,
                                   const MathUtil::Rectangle<int>& dstrect) override;
 
-    void Load(const u8* buffer, u32 width, u32 height, u32 expanded_width, u32 levels) override;
+    void Load(u32 level, u32 width, u32 height, u32 row_length, const u8* buffer,
+              size_t buffer_size) override;
 
     void FromRenderTarget(bool is_depth_copy, const EFBRectangle& srcRect, bool scaleByHalf,
                           unsigned int cbufid, const float* colmat) override;

Source/Core/VideoBackends/D3D/main.cpp

@@ -77,6 +77,7 @@ void VideoBackend::InitBackendInfo()
   g_Config.backend_info.bSupportsMultithreading = false;
   g_Config.backend_info.bSupportsInternalResolutionFrameDumps = false;
   g_Config.backend_info.bSupportsGPUTextureDecoding = false;
+  g_Config.backend_info.bSupportsST3CTextures = false;
 
   IDXGIFactory* factory;
   IDXGIAdapter* ad;

Source/Core/VideoBackends/D3D12/D3DBase.cpp

@@ -262,6 +262,21 @@ std::vector<DXGI_SAMPLE_DESC> EnumAAModes(ID3D12Device* device)
   return aa_modes;
 }
 
+static bool SupportsS3TCTextures(ID3D12Device* device)
+{
+  auto CheckForFormat = [](ID3D12Device* device, DXGI_FORMAT format) {
+    D3D12_FEATURE_DATA_FORMAT_SUPPORT data = {format};
+    if (FAILED(device->CheckFeatureSupport(D3D12_FEATURE_FORMAT_SUPPORT, &data, sizeof(data))))
+      return false;
+
+    return (data.Support1 & D3D12_FORMAT_SUPPORT1_TEXTURE2D) != 0;
+  };
+
+  return CheckForFormat(device, DXGI_FORMAT_BC1_UNORM) &&
+         CheckForFormat(device, DXGI_FORMAT_BC2_UNORM) &&
+         CheckForFormat(device, DXGI_FORMAT_BC3_UNORM);
+}
+
 HRESULT Create(HWND wnd)
 {
   hWnd = wnd;
@@ -478,6 +493,8 @@ HRESULT Create(HWND wnd)
   SAFE_RELEASE(factory);
   SAFE_RELEASE(adapter);
 
+  g_Config.backend_info.bSupportsST3CTextures = SupportsS3TCTextures(device12);
+
   return S_OK;
 }
 

Source/Core/VideoBackends/D3D12/TextureCache.cpp

@@ -33,6 +33,25 @@ static u32 s_efb_copy_last_cbuf_id = UINT_MAX;
 static ID3D12Resource* s_texture_cache_entry_readback_buffer = nullptr;
 static size_t s_texture_cache_entry_readback_buffer_size = 0;
 
+static DXGI_FORMAT GetDXGIFormatForHostFormat(HostTextureFormat format)
+{
+  switch (format)
+  {
+  case HostTextureFormat::DXT1:
+    return DXGI_FORMAT_BC1_UNORM;
+
+  case HostTextureFormat::DXT3:
+    return DXGI_FORMAT_BC2_UNORM;
+
+  case HostTextureFormat::DXT5:
+    return DXGI_FORMAT_BC3_UNORM;
+
+  case HostTextureFormat::RGBA8:
+  default:
+    return DXGI_FORMAT_R8G8B8A8_UNORM;
+  }
+}
+
 TextureCache::TCacheEntry::~TCacheEntry()
 {
   m_texture->Release();
@@ -51,6 +70,11 @@ bool TextureCache::TCacheEntry::Save(const std::string& filename, unsigned int l
       Common::AlignUp(level_width * sizeof(u32), D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
   size_t required_readback_buffer_size = level_pitch * level_height;
 
+  // We can't dump compressed textures currently (it would mean drawing them to a RGBA8
+  // framebuffer, and saving that). TextureCache does not call Save for custom textures
+  // anyway, so this is fine for now.
+  _assert_(config.format == HostTextureFormat::RGBA8);
+
   // Check if the current readback buffer is large enough
   if (required_readback_buffer_size > s_texture_cache_entry_readback_buffer_size)
   {
@@ -175,22 +199,24 @@ void TextureCache::TCacheEntry::CopyRectangleFromTexture(const TCacheEntryBase*
   g_renderer->RestoreAPIState();
 }
 
-void TextureCache::TCacheEntry::Load(const u8* buffer, u32 width, u32 height, u32 expanded_width,
+void TextureCache::TCacheEntry::Load(u32 level, u32 width, u32 height, u32 row_length,
-                                     u32 level)
+                                     const u8* buffer, size_t buffer_size)
 {
-  unsigned int src_pitch = 4 * expanded_width;
+  size_t src_pitch = CalculateHostTextureLevelPitch(config.format, row_length);
-  D3D::ReplaceRGBATexture2D(m_texture->GetTex12(), buffer, width, height, src_pitch, level,
+  D3D::ReplaceRGBATexture2D(m_texture->GetTex12(), buffer, width, height,
+                            static_cast<unsigned int>(src_pitch), level,
                             m_texture->GetResourceUsageState());
 }
 
 TextureCacheBase::TCacheEntryBase* TextureCache::CreateTexture(const TCacheEntryConfig& config)
 {
+  DXGI_FORMAT dxgi_format = GetDXGIFormatForHostFormat(config.format);
   if (config.rendertarget)
   {
     D3DTexture2D* texture =
         D3DTexture2D::Create(config.width, config.height,
                              TEXTURE_BIND_FLAG_SHADER_RESOURCE | TEXTURE_BIND_FLAG_RENDER_TARGET,
-                             DXGI_FORMAT_R8G8B8A8_UNORM, 1, config.layers);
+                             dxgi_format, 1, config.layers);
 
     TCacheEntry* entry = new TCacheEntry(config, texture);
 
@@ -204,8 +230,8 @@ TextureCacheBase::TCacheEntryBase* TextureCache::CreateTexture(const TCacheEntry
   {
     ID3D12Resource* texture_resource = nullptr;
 
-    D3D12_RESOURCE_DESC texture_resource_desc = CD3DX12_RESOURCE_DESC::Tex2D(
+    D3D12_RESOURCE_DESC texture_resource_desc =
-        DXGI_FORMAT_R8G8B8A8_UNORM, config.width, config.height, 1, config.levels);
+        CD3DX12_RESOURCE_DESC::Tex2D(dxgi_format, config.width, config.height, 1, config.levels);
 
     CheckHR(D3D::device12->CreateCommittedResource(
         &CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_DEFAULT), D3D12_HEAP_FLAG_NONE,

Source/Core/VideoBackends/D3D12/TextureCache.h

@@ -39,7 +39,8 @@ private:
                                   const MathUtil::Rectangle<int>& src_rect,
                                   const MathUtil::Rectangle<int>& dst_rect) override;
 
-    void Load(const u8* buffer, u32 width, u32 height, u32 expanded_width, u32 levels) override;
+    void Load(u32 level, u32 width, u32 height, u32 row_length, const u8* buffer,
+              size_t buffer_size) override;
 
     void FromRenderTarget(bool is_depth_copy, const EFBRectangle& src_rect, bool scale_by_half,
                           unsigned int cbuf_id, const float* colmat) override;

Source/Core/VideoBackends/D3D12/main.cpp

@@ -80,6 +80,7 @@ void VideoBackend::InitBackendInfo()
   g_Config.backend_info.bSupportsMultithreading = false;
   g_Config.backend_info.bSupportsInternalResolutionFrameDumps = false;
   g_Config.backend_info.bSupportsGPUTextureDecoding = false;
+  g_Config.backend_info.bSupportsST3CTextures = false;
 
   IDXGIFactory* factory;
   IDXGIAdapter* ad;

Source/Core/VideoBackends/Null/NullBackend.cpp

@@ -45,6 +45,7 @@ void VideoBackend::InitBackendInfo()
   g_Config.backend_info.bSupportsMultithreading = false;
   g_Config.backend_info.bSupportsInternalResolutionFrameDumps = false;
   g_Config.backend_info.bSupportsGPUTextureDecoding = false;
+  g_Config.backend_info.bSupportsST3CTextures = false;
 
   // aamodes: We only support 1 sample, so no MSAA
   g_Config.backend_info.Adapters.clear();

Source/Core/VideoBackends/Null/TextureCache.h

@@ -31,7 +31,10 @@ private:
   {
     TCacheEntry(const TCacheEntryConfig& _config) : TCacheEntryBase(_config) {}
     ~TCacheEntry() {}
-    void Load(const u8* buffer, u32 width, u32 height, u32 expanded_width, u32 level) override {}
+    void Load(u32 level, u32 width, u32 height, u32 row_length, const u8* buffer,
+              size_t buffer_size) override
+    {
+    }
     void FromRenderTarget(bool is_depth_copy, const EFBRectangle& src_rect, bool scale_by_half,
                           unsigned int cbufid, const float* colmat) override
     {

Source/Core/VideoBackends/OGL/Render.cpp

@@ -462,6 +462,8 @@ Renderer::Renderer()
   g_ogl_config.bSupportsConservativeDepth = GLExtensions::Supports("GL_ARB_conservative_depth");
   g_ogl_config.bSupportsAniso = GLExtensions::Supports("GL_EXT_texture_filter_anisotropic");
   g_Config.backend_info.bSupportsComputeShaders = GLExtensions::Supports("GL_ARB_compute_shader");
+  g_Config.backend_info.bSupportsST3CTextures =
+      GLExtensions::Supports("GL_EXT_texture_compression_s3tc");
 
   if (GLInterface->GetMode() == GLInterfaceMode::MODE_OPENGLES3)
   {

Source/Core/VideoBackends/OGL/TextureCache.cpp

@@ -11,6 +11,7 @@
 #include <memory>
 #include <vector>
 
+#include "Common/Assert.h"
 #include "Common/GL/GLInterfaceBase.h"
 #include "Common/MsgHandler.h"
 #include "Common/StringUtil.h"
@@ -87,6 +88,48 @@ bool SaveTexture(const std::string& filename, u32 textarget, u32 tex, int virtua
   return TextureToPng(data.data(), width * 4, filename, width, height, true);
 }
 
+static GLenum GetGLInternalFormatForTextureFormat(HostTextureFormat format, bool storage)
+{
+  switch (format)
+  {
+  case HostTextureFormat::DXT1:
+    return GL_COMPRESSED_RGBA_S3TC_DXT1_EXT;
+  case HostTextureFormat::DXT3:
+    return GL_COMPRESSED_RGBA_S3TC_DXT3_EXT;
+  case HostTextureFormat::DXT5:
+    return GL_COMPRESSED_RGBA_S3TC_DXT5_EXT;
+  case HostTextureFormat::RGBA8:
+  default:
+    return storage ? GL_RGBA8 : GL_RGBA;
+  }
+}
+
+static GLenum GetGLFormatForTextureFormat(HostTextureFormat format)
+{
+  switch (format)
+  {
+  case HostTextureFormat::RGBA8:
+    return GL_RGBA;
+
+  // Compressed texture formats don't use this parameter.
+  default:
+    return GL_UNSIGNED_BYTE;
+  }
+}
+
+static GLenum GetGLTypeForTextureFormat(HostTextureFormat format)
+{
+  switch (format)
+  {
+  case HostTextureFormat::RGBA8:
+    return GL_UNSIGNED_BYTE;
+
+  // Compressed texture formats don't use this parameter.
+  default:
+    return GL_UNSIGNED_BYTE;
+  }
+}
+
 TextureCache::TCacheEntry::~TCacheEntry()
 {
   if (texture)
@@ -129,6 +172,11 @@ void TextureCache::TCacheEntry::Bind(unsigned int stage)
 
 bool TextureCache::TCacheEntry::Save(const std::string& filename, unsigned int level)
 {
+  // We can't dump compressed textures currently (it would mean drawing them to a RGBA8
+  // framebuffer, and saving that). TextureCache does not call Save for custom textures
+  // anyway, so this is fine for now.
+  _assert_(config.format == HostTextureFormat::RGBA8);
+
   return SaveTexture(filename, GL_TEXTURE_2D_ARRAY, texture, config.width, config.height, level);
 }
 
@@ -143,12 +191,15 @@ TextureCache::TCacheEntryBase* TextureCache::CreateTexture(const TCacheEntryConf
 
   if (g_ogl_config.bSupportsTextureStorage)
   {
-    glTexStorage3D(GL_TEXTURE_2D_ARRAY, config.levels, GL_RGBA8, config.width, config.height,
+    GLenum gl_internal_format = GetGLInternalFormatForTextureFormat(config.format, true);
-                   config.layers);
+    glTexStorage3D(GL_TEXTURE_2D_ARRAY, config.levels, gl_internal_format, config.width,
+                   config.height, config.layers);
   }
 
   if (config.rendertarget)
   {
+    // We can't render to compressed formats.
+    _assert_(!IsCompressedHostTextureFormat(config.format));
     if (!g_ogl_config.bSupportsTextureStorage)
     {
       for (u32 level = 0; level < config.levels; level++)
@@ -202,8 +253,8 @@ void TextureCache::TCacheEntry::CopyRectangleFromTexture(const TCacheEntryBase*
   g_renderer->RestoreAPIState();
 }
 
-void TextureCache::TCacheEntry::Load(const u8* buffer, u32 width, u32 height, u32 expanded_width,
+void TextureCache::TCacheEntry::Load(u32 level, u32 width, u32 height, u32 row_length,
-                                     u32 level)
+                                     const u8* buffer, size_t buffer_size)
 {
   if (level >= config.levels)
     PanicAlert("Texture only has %d levels, can't update level %d", config.levels, level);
@@ -216,21 +267,40 @@ void TextureCache::TCacheEntry::Load(const u8* buffer, u32 width, u32 height, u3
   glActiveTexture(GL_TEXTURE9);
   glBindTexture(GL_TEXTURE_2D_ARRAY, texture);
 
-  if (expanded_width != width)
+  if (row_length != width)
-    glPixelStorei(GL_UNPACK_ROW_LENGTH, expanded_width);
+    glPixelStorei(GL_UNPACK_ROW_LENGTH, row_length);
 
+  GLenum gl_internal_format = GetGLInternalFormatForTextureFormat(config.format, false);
+  if (IsCompressedHostTextureFormat(config.format))
+  {
     if (g_ogl_config.bSupportsTextureStorage)
     {
-    glTexSubImage3D(GL_TEXTURE_2D_ARRAY, level, 0, 0, 0, width, height, 1, GL_RGBA,
+      glCompressedTexSubImage3D(GL_TEXTURE_2D_ARRAY, level, 0, 0, 0, width, height, 1,
-                    GL_UNSIGNED_BYTE, buffer);
+                                gl_internal_format, static_cast<GLsizei>(buffer_size), buffer);
     }
     else
     {
-    glTexImage3D(GL_TEXTURE_2D_ARRAY, level, GL_RGBA, width, height, 1, 0, GL_RGBA,
+      glCompressedTexImage3D(GL_TEXTURE_2D_ARRAY, level, gl_internal_format, width, height, 1, 0,
-                 GL_UNSIGNED_BYTE, buffer);
+                             static_cast<GLsizei>(buffer_size), buffer);
+    }
+  }
+  else
+  {
+    GLenum gl_format = GetGLFormatForTextureFormat(config.format);
+    GLenum gl_type = GetGLTypeForTextureFormat(config.format);
+    if (g_ogl_config.bSupportsTextureStorage)
+    {
+      glTexSubImage3D(GL_TEXTURE_2D_ARRAY, level, 0, 0, 0, width, height, 1, gl_format, gl_type,
+                      buffer);
+    }
+    else
+    {
+      glTexImage3D(GL_TEXTURE_2D_ARRAY, level, gl_internal_format, width, height, 1, 0, gl_format,
+                   gl_type, buffer);
+    }
   }
 
-  if (expanded_width != width)
+  if (row_length != width)
     glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
 
   TextureCache::SetStage();

Source/Core/VideoBackends/OGL/TextureCache.h

@@ -45,7 +45,8 @@ private:
                                   const MathUtil::Rectangle<int>& srcrect,
                                   const MathUtil::Rectangle<int>& dstrect) override;
 
-    void Load(const u8* buffer, u32 width, u32 height, u32 expanded_width, u32 level) override;
+    void Load(u32 level, u32 width, u32 height, u32 row_length, const u8* buffer,
+              size_t buffer_size) override;
 
     void FromRenderTarget(bool is_depth_copy, const EFBRectangle& srcRect, bool scaleByHalf,
                           unsigned int cbufid, const float* colmat) override;

Source/Core/VideoBackends/OGL/main.cpp

@@ -103,6 +103,7 @@ void VideoBackend::InitBackendInfo()
   g_Config.backend_info.bSupportsPaletteConversion = true;
   g_Config.backend_info.bSupportsClipControl = true;
   g_Config.backend_info.bSupportsDepthClamp = true;
+  g_Config.backend_info.bSupportsST3CTextures = false;
 
   g_Config.backend_info.Adapters.clear();
 

Source/Core/VideoBackends/Software/SWmain.cpp

@@ -65,7 +65,10 @@ private:
   {
     TCacheEntry(const TCacheEntryConfig& _config) : TCacheEntryBase(_config) {}
     ~TCacheEntry() {}
-    void Load(const u8* buffer, u32 width, u32 height, u32 expanded_width, u32 level) override {}
+    void Load(u32 level, u32 width, u32 height, u32 row_length, const u8* buffer,
+              size_t buffer_size) override
+    {
+    }
     void FromRenderTarget(bool is_depth_copy, const EFBRectangle& srcRect, bool scaleByHalf,
                           unsigned int cbufid, const float* colmat) override
     {
@@ -134,6 +137,7 @@ void VideoSoftware::InitBackendInfo()
   g_Config.backend_info.bSupportsComputeShaders = false;
   g_Config.backend_info.bSupportsInternalResolutionFrameDumps = false;
   g_Config.backend_info.bSupportsGPUTextureDecoding = false;
+  g_Config.backend_info.bSupportsST3CTextures = false;
 
   // aamodes
   g_Config.backend_info.AAModes = {1};

Source/Core/VideoBackends/Vulkan/Constants.h

@@ -109,7 +109,7 @@ constexpr size_t MAXIMUM_TEXTURE_UPLOAD_BUFFER_SIZE = 64 * 1024 * 1024;
 // streaming buffer and be blocking frequently. Games are unlikely to have textures this
 // large anyway, so it's only really an issue for HD texture packs, and memory is not
 // a limiting factor in these scenarios anyway.
-constexpr size_t STAGING_TEXTURE_UPLOAD_THRESHOLD = 1024 * 1024 * 4;
+constexpr size_t STAGING_TEXTURE_UPLOAD_THRESHOLD = 1024 * 1024 * 8;
 
 // Streaming uniform buffer size
 constexpr size_t INITIAL_UNIFORM_STREAM_BUFFER_SIZE = 16 * 1024 * 1024;

Source/Core/VideoBackends/Vulkan/StagingBuffer.cpp

@@ -127,7 +127,7 @@ void StagingBuffer::InvalidateCPUCache(VkDeviceSize offset, VkDeviceSize size)
 void StagingBuffer::Read(VkDeviceSize offset, void* data, size_t size, bool invalidate_caches)
 {
   _assert_((offset + size) <= m_size);
-  _assert_(offset >= m_map_offset && size < (m_map_size + (offset - m_map_offset)));
+  _assert_(offset >= m_map_offset && size <= (m_map_size + (offset - m_map_offset)));
   if (invalidate_caches)
     InvalidateCPUCache(offset, size);
 
@@ -138,7 +138,7 @@ void StagingBuffer::Write(VkDeviceSize offset, const void* data, size_t size,
                           bool invalidate_caches)
 {
   _assert_((offset + size) <= m_size);
-  _assert_(offset >= m_map_offset && size < (m_map_size + (offset - m_map_offset)));
+  _assert_(offset >= m_map_offset && size <= (m_map_size + (offset - m_map_offset)));
 
   memcpy(m_map_pointer + (offset - m_map_offset), data, size);
   if (invalidate_caches)

Source/Core/VideoBackends/Vulkan/TextureCache.cpp

@@ -9,6 +9,7 @@
 #include <string>
 #include <vector>
 
+#include "Common/Align.h"
 #include "Common/Assert.h"
 #include "Common/CommonFuncs.h"
 #include "Common/Logging/Log.h"
@@ -238,9 +239,10 @@ TextureCacheBase::TCacheEntryBase* TextureCache::CreateTexture(const TCacheEntry
     usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
 
   // Allocate texture object
+  VkFormat vk_format = Util::GetVkFormatForHostTextureFormat(config.format);
   std::unique_ptr<Texture2D> texture = Texture2D::Create(
-      config.width, config.height, config.levels, config.layers, TEXTURECACHE_TEXTURE_FORMAT,
+      config.width, config.height, config.levels, config.layers, vk_format, VK_SAMPLE_COUNT_1_BIT,
-      VK_SAMPLE_COUNT_1_BIT, VK_IMAGE_VIEW_TYPE_2D_ARRAY, VK_IMAGE_TILING_OPTIMAL, usage);
+      VK_IMAGE_VIEW_TYPE_2D_ARRAY, VK_IMAGE_TILING_OPTIMAL, usage);
 
   if (!texture)
     return nullptr;
@@ -341,8 +343,8 @@ TextureCache::TCacheEntry::~TCacheEntry()
     g_command_buffer_mgr->DeferFramebufferDestruction(m_framebuffer);
 }
 
-void TextureCache::TCacheEntry::Load(const u8* buffer, unsigned int width, unsigned int height,
+void TextureCache::TCacheEntry::Load(u32 level, u32 width, u32 height, u32 row_length,
-                                     unsigned int expanded_width, unsigned int level)
+                                     const u8* buffer, size_t buffer_size)
 {
   // Can't copy data larger than the texture extents.
   width = std::max(1u, std::min(width, m_texture->GetWidth() >> level));
@@ -366,87 +368,68 @@ void TextureCache::TCacheEntry::Load(const u8* buffer, unsigned int width, unsig
   m_texture->TransitionToLayout(g_command_buffer_mgr->GetCurrentInitCommandBuffer(),
                                 VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
 
-  // Does this texture data fit within the streaming buffer?
+  // For unaligned textures, we can save some memory in the transfer buffer by skipping the rows
-  u32 upload_width = width;
+  // that lie outside of the texture's dimensions.
-  u32 upload_pitch = upload_width * sizeof(u32);
-  u32 upload_size = upload_pitch * height;
   u32 upload_alignment = static_cast<u32>(g_vulkan_context->GetBufferImageGranularity());
-  u32 source_pitch = expanded_width * 4;
+  u32 block_size = Util::GetBlockSize(m_texture->GetFormat());
-  if ((upload_size + upload_alignment) <= STAGING_TEXTURE_UPLOAD_THRESHOLD &&
+  u32 num_rows = Common::AlignUp(height, block_size) / block_size;
-      (upload_size + upload_alignment) <= MAXIMUM_TEXTURE_UPLOAD_BUFFER_SIZE)
+  size_t source_pitch = CalculateHostTextureLevelPitch(config.format, row_length);
-  {
+  size_t upload_size = source_pitch * num_rows;
-    // Assume tightly packed rows, with no padding as the buffer source.
+  std::unique_ptr<StagingBuffer> temp_buffer;
-    StreamBuffer* upload_buffer = TextureCache::GetInstance()->m_texture_upload_buffer.get();
+  VkBuffer upload_buffer;
+  VkDeviceSize upload_buffer_offset;
 
-    // Allocate memory from the streaming buffer for the texture data.
+  // Does this texture data fit within the streaming buffer?
-    if (!upload_buffer->ReserveMemory(upload_size, g_vulkan_context->GetBufferImageGranularity()))
+  if (upload_size <= STAGING_TEXTURE_UPLOAD_THRESHOLD &&
+      upload_size <= MAXIMUM_TEXTURE_UPLOAD_BUFFER_SIZE)
+  {
+    StreamBuffer* stream_buffer = TextureCache::GetInstance()->m_texture_upload_buffer.get();
+    if (!stream_buffer->ReserveMemory(upload_size, upload_alignment))
     {
       // Execute the command buffer first.
       WARN_LOG(VIDEO, "Executing command list while waiting for space in texture upload buffer");
       Util::ExecuteCurrentCommandsAndRestoreState(false);
 
       // Try allocating again. This may cause a fence wait.
-      if (!upload_buffer->ReserveMemory(upload_size, g_vulkan_context->GetBufferImageGranularity()))
+      if (!stream_buffer->ReserveMemory(upload_size, upload_alignment))
         PanicAlert("Failed to allocate space in texture upload buffer");
     }
 
-    // Grab buffer pointers
+    // Copy to the streaming buffer.
-    VkBuffer image_upload_buffer = upload_buffer->GetBuffer();
+    upload_buffer = stream_buffer->GetBuffer();
-    VkDeviceSize image_upload_buffer_offset = upload_buffer->GetCurrentOffset();
+    upload_buffer_offset = stream_buffer->GetCurrentOffset();
-    u8* image_upload_buffer_pointer = upload_buffer->GetCurrentHostPointer();
+    std::memcpy(stream_buffer->GetCurrentHostPointer(), buffer, upload_size);
-
+    stream_buffer->CommitMemory(upload_size);
-    // Copy to the buffer using the stride from the subresource layout
-    const u8* source_ptr = buffer;
-    if (upload_pitch != source_pitch)
-    {
-      VkDeviceSize copy_pitch = std::min(source_pitch, upload_pitch);
-      for (unsigned int row = 0; row < height; row++)
-      {
-        memcpy(image_upload_buffer_pointer + row * upload_pitch, source_ptr + row * source_pitch,
-               copy_pitch);
-      }
   }
   else
   {
-      // Can copy the whole thing in one block, the pitch matches
+    // Create a temporary staging buffer that is destroyed after the image is copied.
-      memcpy(image_upload_buffer_pointer, source_ptr, upload_size);
+    temp_buffer = StagingBuffer::Create(STAGING_BUFFER_TYPE_UPLOAD, upload_size,
-    }
+                                        VK_BUFFER_USAGE_TRANSFER_SRC_BIT);
-
+    if (!temp_buffer || !temp_buffer->Map())
-    // Flush buffer memory if necessary
-    upload_buffer->CommitMemory(upload_size);
-
-    // Copy from the streaming buffer to the actual image.
-    VkBufferImageCopy image_copy = {
-        image_upload_buffer_offset,                // VkDeviceSize                bufferOffset
-        0,                                         // uint32_t                    bufferRowLength
-        0,                                         // uint32_t                    bufferImageHeight
-        {VK_IMAGE_ASPECT_COLOR_BIT, level, 0, 1},  // VkImageSubresourceLayers    imageSubresource
-        {0, 0, 0},                                 // VkOffset3D                  imageOffset
-        {width, height, 1}                         // VkExtent3D                  imageExtent
-    };
-    vkCmdCopyBufferToImage(g_command_buffer_mgr->GetCurrentInitCommandBuffer(), image_upload_buffer,
-                           m_texture->GetImage(), VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1,
-                           &image_copy);
-  }
-  else
-  {
-    // Slow path. The data for the image is too large to fit in the streaming buffer, so we need
-    // to allocate a temporary texture to store the data in, then copy to the real texture.
-    std::unique_ptr<StagingTexture2D> staging_texture = StagingTexture2D::Create(
-        STAGING_BUFFER_TYPE_UPLOAD, width, height, TEXTURECACHE_TEXTURE_FORMAT);
-
-    if (!staging_texture || !staging_texture->Map())
     {
       PanicAlert("Failed to allocate staging texture for large texture upload.");
       return;
     }
 
-    // Copy data to staging texture first, then to the "real" texture.
+    upload_buffer = temp_buffer->GetBuffer();
-    staging_texture->WriteTexels(0, 0, width, height, buffer, source_pitch);
+    upload_buffer_offset = 0;
-    staging_texture->CopyToImage(g_command_buffer_mgr->GetCurrentInitCommandBuffer(),
+    temp_buffer->Write(0, buffer, upload_size, true);
-                                 m_texture->GetImage(), VK_IMAGE_ASPECT_COLOR_BIT, 0, 0, width,
+    temp_buffer->Unmap();
-                                 height, level, 0);
   }
+
+  // Copy from the streaming buffer to the actual image.
+  VkBufferImageCopy image_copy = {
+      upload_buffer_offset,                      // VkDeviceSize                bufferOffset
+      row_length,                                // uint32_t                    bufferRowLength
+      0,                                         // uint32_t                    bufferImageHeight
+      {VK_IMAGE_ASPECT_COLOR_BIT, level, 0, 1},  // VkImageSubresourceLayers    imageSubresource
+      {0, 0, 0},                                 // VkOffset3D                  imageOffset
+      {width, height, 1}                         // VkExtent3D                  imageExtent
+  };
+  vkCmdCopyBufferToImage(g_command_buffer_mgr->GetCurrentInitCommandBuffer(), upload_buffer,
+                         m_texture->GetImage(), VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1,
+                         &image_copy);
 }
 
 void TextureCache::TCacheEntry::FromRenderTarget(bool is_depth_copy, const EFBRectangle& src_rect,
@@ -544,6 +527,11 @@ bool TextureCache::TCacheEntry::Save(const std::string& filename, unsigned int l
 {
   _assert_(level < config.levels);
 
+  // We can't dump compressed textures currently (it would mean drawing them to a RGBA8
+  // framebuffer, and saving that). TextureCache does not call Save for custom textures
+  // anyway, so this is fine for now.
+  _assert_(config.format == HostTextureFormat::RGBA8);
+
   // Determine dimensions of image we want to save.
   u32 level_width = std::max(1u, config.width >> level);
   u32 level_height = std::max(1u, config.height >> level);
@@ -582,7 +570,8 @@ bool TextureCache::TCacheEntry::Save(const std::string& filename, unsigned int l
   // It's okay to throw this texture away immediately, since we're done with it, and
   // we blocked until the copy completed on the GPU anyway.
   bool result = TextureToPng(reinterpret_cast<u8*>(staging_texture->GetMapPointer()),
-                             staging_texture->GetRowStride(), filename, level_width, level_height);
+                             static_cast<u32>(staging_texture->GetRowStride()), filename,
+                             level_width, level_height);
 
   staging_texture->Unmap();
   return result;

Source/Core/VideoBackends/Vulkan/TextureCache.h

@@ -27,8 +27,8 @@ public:
 
     Texture2D* GetTexture() const { return m_texture.get(); }
     VkFramebuffer GetFramebuffer() const { return m_framebuffer; }
-    void Load(const u8* buffer, unsigned int width, unsigned int height,
+    void Load(u32 level, u32 width, u32 height, u32 row_length, const u8* buffer,
-              unsigned int expanded_width, unsigned int level) override;
+              size_t buffer_size) override;
     void FromRenderTarget(bool is_depth_copy, const EFBRectangle& src_rect, bool scale_by_half,
                           unsigned int cbufid, const float* colmat) override;
     void CopyRectangleFromTexture(const TCacheEntryBase* source,

Source/Core/VideoBackends/Vulkan/Util.cpp

@@ -53,6 +53,20 @@ bool IsDepthFormat(VkFormat format)
   }
 }
 
+bool IsCompressedFormat(VkFormat format)
+{
+  switch (format)
+  {
+  case VK_FORMAT_BC1_RGBA_UNORM_BLOCK:
+  case VK_FORMAT_BC2_UNORM_BLOCK:
+  case VK_FORMAT_BC3_UNORM_BLOCK:
+    return true;
+
+  default:
+    return false;
+  }
+}
+
 VkFormat GetLinearFormat(VkFormat format)
 {
   switch (format)
@@ -74,6 +88,25 @@ VkFormat GetLinearFormat(VkFormat format)
   }
 }
 
+VkFormat GetVkFormatForHostTextureFormat(HostTextureFormat format)
+{
+  switch (format)
+  {
+  case HostTextureFormat::DXT1:
+    return VK_FORMAT_BC1_RGBA_UNORM_BLOCK;
+
+  case HostTextureFormat::DXT3:
+    return VK_FORMAT_BC2_UNORM_BLOCK;
+
+  case HostTextureFormat::DXT5:
+    return VK_FORMAT_BC3_UNORM_BLOCK;
+
+  case HostTextureFormat::RGBA8:
+  default:
+    return VK_FORMAT_R8G8B8A8_UNORM;
+  }
+}
+
 u32 GetTexelSize(VkFormat format)
 {
   // Only contains pixel formats we use.
@@ -91,12 +124,33 @@ u32 GetTexelSize(VkFormat format)
   case VK_FORMAT_B8G8R8A8_UNORM:
     return 4;
 
+  case VK_FORMAT_BC1_RGBA_UNORM_BLOCK:
+    return 8;
+
+  case VK_FORMAT_BC2_UNORM_BLOCK:
+  case VK_FORMAT_BC3_UNORM_BLOCK:
+    return 16;
+
   default:
     PanicAlert("Unhandled pixel format");
     return 1;
   }
 }
 
+u32 GetBlockSize(VkFormat format)
+{
+  switch (format)
+  {
+  case VK_FORMAT_BC1_RGBA_UNORM_BLOCK:
+  case VK_FORMAT_BC2_UNORM_BLOCK:
+  case VK_FORMAT_BC3_UNORM_BLOCK:
+    return 4;
+
+  default:
+    return 1;
+  }
+}
+
 VkRect2D ClampRect2D(const VkRect2D& rect, u32 width, u32 height)
 {
   VkRect2D out;

Source/Core/VideoBackends/Vulkan/Util.h

@@ -25,8 +25,11 @@ size_t AlignBufferOffset(size_t offset, size_t alignment);
 u32 MakeRGBA8Color(float r, float g, float b, float a);
 
 bool IsDepthFormat(VkFormat format);
+bool IsCompressedFormat(VkFormat format);
 VkFormat GetLinearFormat(VkFormat format);
+VkFormat GetVkFormatForHostTextureFormat(HostTextureFormat format);
 u32 GetTexelSize(VkFormat format);
+u32 GetBlockSize(VkFormat format);
 
 // Clamps a VkRect2D to the specified dimensions.
 VkRect2D ClampRect2D(const VkRect2D& rect, u32 width, u32 height);

Source/Core/VideoBackends/Vulkan/VulkanContext.cpp

@@ -245,6 +245,7 @@ void VulkanContext::PopulateBackendInfo(VideoConfig* config)
   config->backend_info.bSupportsFragmentStoresAndAtomics = false;     // Dependent on features.
   config->backend_info.bSupportsSSAA = false;                         // Dependent on features.
   config->backend_info.bSupportsDepthClamp = false;                   // Dependent on features.
+  config->backend_info.bSupportsST3CTextures = false;                 // Dependent on features.
   config->backend_info.bSupportsReversedDepthRange = false;  // No support yet due to driver bugs.
 }
 
@@ -283,6 +284,9 @@ void VulkanContext::PopulateBackendInfoFeatures(VideoConfig* config, VkPhysicalD
   config->backend_info.bSupportsDepthClamp =
       (features.depthClamp == VK_TRUE && features.shaderClipDistance == VK_TRUE);
 
+  // textureCompressionBC implies BC1 through BC7, which is a superset of DXT1/3/5, which we need.
+  config->backend_info.bSupportsST3CTextures = features.textureCompressionBC == VK_TRUE;
+
   // Our usage of primitive restart appears to be broken on AMD's binary drivers.
   // Seems to be fine on GCN Gen 1-2, unconfirmed on GCN Gen 3, causes driver resets on GCN Gen 4.
   if (DriverDetails::HasBug(DriverDetails::BUG_PRIMITIVE_RESTART))
@@ -459,6 +463,7 @@ bool VulkanContext::SelectDeviceFeatures()
   m_device_features.occlusionQueryPrecise = available_features.occlusionQueryPrecise;
   m_device_features.shaderClipDistance = available_features.shaderClipDistance;
   m_device_features.depthClamp = available_features.depthClamp;
+  m_device_features.textureCompressionBC = available_features.textureCompressionBC;
   return true;
 }
 

Source/Core/VideoCommon/CMakeLists.txt

@@ -14,6 +14,7 @@ set(SRCS
   GeometryShaderGen.cpp
   GeometryShaderManager.cpp
   HiresTextures.cpp
+  HiresTextures_DDSLoader.cpp
   ImageWrite.cpp
   IndexGenerator.cpp
   LightingShaderGen.cpp

Source/Core/VideoCommon/HiresTextures.cpp

@@ -366,6 +366,21 @@ std::string HiresTexture::GenBaseName(const u8* texture, size_t texture_size, co
   return name;
 }
 
+u32 HiresTexture::CalculateMipCount(u32 width, u32 height)
+{
+  u32 mip_width = width;
+  u32 mip_height = height;
+  u32 mip_count = 1;
+  while (mip_width > 1 || mip_height > 1)
+  {
+    mip_width = std::max(mip_width / 2, 1u);
+    mip_height = std::max(mip_height / 2, 1u);
+    mip_count++;
+  }
+
+  return mip_count;
+}
+
 std::shared_ptr<HiresTexture> HiresTexture::Search(const u8* texture, size_t texture_size,
                                                    const u8* tlut, size_t tlut_size, u32 width,
                                                    u32 height, int format, bool has_mipmaps)
@@ -395,81 +410,134 @@ std::shared_ptr<HiresTexture> HiresTexture::Search(const u8* texture, size_t tex
 std::unique_ptr<HiresTexture> HiresTexture::Load(const std::string& base_filename, u32 width,
                                                  u32 height)
 {
-  std::unique_ptr<HiresTexture> ret;
+  // We need to have a level 0 custom texture to even consider loading.
-  for (int level = 0;; level++)
+  auto filename_iter = s_textureMap.find(base_filename);
+  if (filename_iter == s_textureMap.end())
+    return nullptr;
+
+  // Try to load level 0 (and any mipmaps) from a DDS file.
+  // If this fails, it's fine, we'll just load level0 again using SOIL.
+  // Can't use make_unique due to private constructor.
+  std::unique_ptr<HiresTexture> ret = std::unique_ptr<HiresTexture>(new HiresTexture());
+  const std::string& first_mip_filename = filename_iter->second;
+  LoadDDSTexture(ret.get(), first_mip_filename);
+
+  // Load remaining mip levels, or from the start if it's not a DDS texture.
+  for (u32 mip_level = static_cast<u32>(ret->m_levels.size());; mip_level++)
   {
     std::string filename = base_filename;
-    if (level)
+    if (mip_level != 0)
-    {
+      filename += StringFromFormat("_mip%u", mip_level);
-      filename += StringFromFormat("_mip%u", level);
-    }
 
-    if (s_textureMap.find(filename) != s_textureMap.end())
+    filename_iter = s_textureMap.find(filename);
-    {
+    if (filename_iter == s_textureMap.end())
-      Level l;
+      break;
 
+    // Try loading DDS textures first, that way we maintain compression of DXT formats.
+    // TODO: Reduce the number of open() calls here. We could use one fd.
+    Level level;
+    if (!LoadDDSTexture(level, filename_iter->second))
+    {
       File::IOFile file;
-      file.Open(s_textureMap[filename], "rb");
+      file.Open(filename_iter->second, "rb");
       std::vector<u8> buffer(file.GetSize());
       file.ReadBytes(buffer.data(), file.GetSize());
-
+      if (!LoadTexture(level, buffer))
-      int channels;
-      l.data =
-          SOILPointer(SOIL_load_image_from_memory(buffer.data(), (int)buffer.size(), (int*)&l.width,
-                                                  (int*)&l.height, &channels, SOIL_LOAD_RGBA),
-                      SOIL_free_image_data);
-      l.data_size = (size_t)l.width * l.height * 4;
-
-      if (l.data == nullptr)
       {
         ERROR_LOG(VIDEO, "Custom texture %s failed to load", filename.c_str());
         break;
       }
-
-      if (!level)
-      {
-        if (l.width * height != l.height * width)
-          ERROR_LOG(VIDEO, "Invalid custom texture size %dx%d for texture %s. The aspect differs "
-                           "from the native size %dx%d.",
-                    l.width, l.height, filename.c_str(), width, height);
-        if (width && height && (l.width % width || l.height % height))
-          WARN_LOG(VIDEO, "Invalid custom texture size %dx%d for texture %s. Please use an integer "
-                          "upscaling factor based on the native size %dx%d.",
-                   l.width, l.height, filename.c_str(), width, height);
-        width = l.width;
-        height = l.height;
-      }
-      else if (width != l.width || height != l.height)
-      {
-        ERROR_LOG(
-            VIDEO,
-            "Invalid custom texture size %dx%d for texture %s. This mipmap layer _must_ be %dx%d.",
-            l.width, l.height, filename.c_str(), width, height);
-        l.data.reset();
-        break;
     }
 
-      if (!ret)
+    ret->m_levels.push_back(std::move(level));
-        ret = std::unique_ptr<HiresTexture>(new HiresTexture);
+  }
-      ret->m_levels.push_back(std::move(l));
 
-      // no more mipmaps available
+  // If we failed to load any mip levels, we can't use this texture at all.
-      if (width == 1 && height == 1)
+  if (ret->m_levels.empty())
-        break;
+    return nullptr;
 
-      // calculate the size of the next mipmap
+  // Verify that the aspect ratio of the texture hasn't changed, as this could have side-effects.
-      width = std::max(1u, width >> 1);
+  const Level& first_mip = ret->m_levels[0];
-      height = std::max(1u, height >> 1);
+  if (first_mip.width * height != first_mip.height * width)
+  {
+    ERROR_LOG(VIDEO, "Invalid custom texture size %ux%u for texture %s. The aspect differs "
+                     "from the native size %ux%u.",
+              first_mip.width, first_mip.height, first_mip_filename.c_str(), width, height);
+  }
+
+  // Same deal if the custom texture isn't a multiple of the native size.
+  if (width != 0 && height != 0 && (first_mip.width % width || first_mip.height % height))
+  {
+    ERROR_LOG(VIDEO, "Invalid custom texture size %ux%u for texture %s. Please use an integer "
+                     "upscaling factor based on the native size %ux%u.",
+              first_mip.width, first_mip.height, first_mip_filename.c_str(), width, height);
+  }
+
+  // Verify that each mip level is the correct size (divide by 2 each time).
+  u32 current_mip_width = first_mip.width;
+  u32 current_mip_height = first_mip.height;
+  for (u32 mip_level = 1; mip_level < static_cast<u32>(ret->m_levels.size()); mip_level++)
+  {
+    if (current_mip_width != 1 || current_mip_height != 1)
+    {
+      current_mip_width = std::max(current_mip_width / 2, 1u);
+      current_mip_height = std::max(current_mip_height / 2, 1u);
+
+      const Level& level = ret->m_levels[mip_level];
+      if (current_mip_width == level.width && current_mip_height == level.height)
+        continue;
+
+      ERROR_LOG(VIDEO,
+                "Invalid custom texture size %dx%d for texture %s. Mipmap level %u must be %dx%d.",
+                level.width, level.height, first_mip_filename.c_str(), mip_level, current_mip_width,
+                current_mip_height);
     }
     else
     {
-      break;
+      // It is invalid to have more than a single 1x1 mipmap.
+      ERROR_LOG(VIDEO, "Custom texture %s has too many 1x1 mipmaps. Skipping extra levels.",
+                first_mip_filename.c_str());
     }
+
+    // Drop this mip level and any others after it.
+    while (ret->m_levels.size() > mip_level)
+      ret->m_levels.pop_back();
+  }
+
+  // All levels have to have the same format.
+  if (std::any_of(ret->m_levels.begin(), ret->m_levels.end(),
+                  [&ret](const Level& l) { return l.format != ret->m_levels[0].format; }))
+  {
+    ERROR_LOG(VIDEO, "Custom texture %s has inconsistent formats across mip levels.",
+              first_mip_filename.c_str());
+
+    return nullptr;
   }
 
   return ret;
 }
 
+bool HiresTexture::LoadTexture(Level& level, const std::vector<u8>& buffer)
+{
+  int channels;
+  int width;
+  int height;
+
+  u8* data = SOIL_load_image_from_memory(buffer.data(), static_cast<int>(buffer.size()), &width,
+                                         &height, &channels, SOIL_LOAD_RGBA);
+  if (!data)
+    return false;
+
+  // Images loaded by SOIL are converted to RGBA.
+  level.width = static_cast<u32>(width);
+  level.height = static_cast<u32>(height);
+  level.format = HostTextureFormat::RGBA8;
+  level.data = ImageDataPointer(data, SOIL_free_image_data);
+  level.row_length = level.width;
+  level.data_size = static_cast<size_t>(level.row_length) * 4 * level.height;
+  return true;
+}
+
 std::string HiresTexture::GetTextureDirectory(const std::string& game_id)
 {
   const std::string texture_directory = File::GetUserPath(D_HIRESTEXTURES_IDX) + game_id;
@@ -484,3 +552,8 @@ std::string HiresTexture::GetTextureDirectory(const std::string& game_id)
 HiresTexture::~HiresTexture()
 {
 }
+
+HostTextureFormat HiresTexture::GetFormat() const
+{
+  return m_levels.at(0).format;
+}

Source/Core/VideoCommon/HiresTextures.h

@@ -9,11 +9,12 @@
 #include <vector>
 
 #include "Common/CommonTypes.h"
+#include "VideoCommon/VideoCommon.h"
 
 class HiresTexture
 {
 public:
-  using SOILPointer = std::unique_ptr<u8, void (*)(unsigned char*)>;
+  using ImageDataPointer = std::unique_ptr<u8, void (*)(unsigned char*)>;
 
   static void Init();
   static void Update();
@@ -27,22 +28,30 @@ public:
                                  size_t tlut_size, u32 width, u32 height, int format,
                                  bool has_mipmaps, bool dump = false);
 
+  static u32 CalculateMipCount(u32 width, u32 height);
+
   ~HiresTexture();
 
+  HostTextureFormat GetFormat() const;
   struct Level
   {
     Level();
 
-    SOILPointer data;
+    ImageDataPointer data;
-    size_t data_size = 0;
+    HostTextureFormat format = HostTextureFormat::RGBA8;
     u32 width = 0;
     u32 height = 0;
+    u32 row_length = 0;
+    size_t data_size = 0;
   };
   std::vector<Level> m_levels;
 
 private:
   static std::unique_ptr<HiresTexture> Load(const std::string& base_filename, u32 width,
                                             u32 height);
+  static bool LoadDDSTexture(HiresTexture* tex, const std::string& filename);
+  static bool LoadDDSTexture(Level& level, const std::string& filename);
+  static bool LoadTexture(Level& level, const std::vector<u8>& buffer);
   static void Prefetch();
 
   static std::string GetTextureDirectory(const std::string& game_id);

Source/Core/VideoCommon/HiresTextures_DDSLoader.cpp

@@ -0,0 +1,486 @@
+// Copyright 2017 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#include "VideoCommon/HiresTextures.h"
+#include <algorithm>
+#include <cstddef>
+#include <cstdint>
+#include <cstring>
+#include <functional>
+#include "Common/Align.h"
+#include "Common/FileUtil.h"
+#include "Common/Swap.h"
+#include "VideoCommon/VideoConfig.h"
+
+namespace
+{
+// From https://raw.githubusercontent.com/Microsoft/DirectXTex/master/DirectXTex/DDS.h
+//
+// This header defines constants and structures that are useful when parsing
+// DDS files.  DDS files were originally designed to use several structures
+// and constants that are native to DirectDraw and are defined in ddraw.h,
+// such as DDSURFACEDESC2 and DDSCAPS2.  This file defines similar
+// (compatible) constants and structures so that one can use DDS files
+// without needing to include ddraw.h.
+//
+// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
+// PARTICULAR PURPOSE.
+//
+// Copyright (c) Microsoft Corporation. All rights reserved.
+//
+// http://go.microsoft.com/fwlink/?LinkId=248926
+
+#pragma pack(push, 1)
+
+const uint32_t DDS_MAGIC = 0x20534444;  // "DDS "
+
+struct DDS_PIXELFORMAT
+{
+  uint32_t dwSize;
+  uint32_t dwFlags;
+  uint32_t dwFourCC;
+  uint32_t dwRGBBitCount;
+  uint32_t dwRBitMask;
+  uint32_t dwGBitMask;
+  uint32_t dwBBitMask;
+  uint32_t dwABitMask;
+};
+
+#define DDS_FOURCC 0x00000004      // DDPF_FOURCC
+#define DDS_RGB 0x00000040         // DDPF_RGB
+#define DDS_RGBA 0x00000041        // DDPF_RGB | DDPF_ALPHAPIXELS
+#define DDS_LUMINANCE 0x00020000   // DDPF_LUMINANCE
+#define DDS_LUMINANCEA 0x00020001  // DDPF_LUMINANCE | DDPF_ALPHAPIXELS
+#define DDS_ALPHA 0x00000002       // DDPF_ALPHA
+#define DDS_PAL8 0x00000020        // DDPF_PALETTEINDEXED8
+#define DDS_PAL8A 0x00000021       // DDPF_PALETTEINDEXED8 | DDPF_ALPHAPIXELS
+#define DDS_BUMPDUDV 0x00080000    // DDPF_BUMPDUDV
+
+#ifndef MAKEFOURCC
+#define MAKEFOURCC(ch0, ch1, ch2, ch3)                                                             \
+  ((uint32_t)(uint8_t)(ch0) | ((uint32_t)(uint8_t)(ch1) << 8) | ((uint32_t)(uint8_t)(ch2) << 16) | \
+   ((uint32_t)(uint8_t)(ch3) << 24))
+#endif /* defined(MAKEFOURCC) */
+
+#define DDS_HEADER_FLAGS_TEXTURE                                                                   \
+  0x00001007  // DDSD_CAPS | DDSD_HEIGHT | DDSD_WIDTH | DDSD_PIXELFORMAT
+#define DDS_HEADER_FLAGS_MIPMAP 0x00020000      // DDSD_MIPMAPCOUNT
+#define DDS_HEADER_FLAGS_VOLUME 0x00800000      // DDSD_DEPTH
+#define DDS_HEADER_FLAGS_PITCH 0x00000008       // DDSD_PITCH
+#define DDS_HEADER_FLAGS_LINEARSIZE 0x00080000  // DDSD_LINEARSIZE
+
+// Subset here matches D3D10_RESOURCE_DIMENSION and D3D11_RESOURCE_DIMENSION
+enum DDS_RESOURCE_DIMENSION
+{
+  DDS_DIMENSION_TEXTURE1D = 2,
+  DDS_DIMENSION_TEXTURE2D = 3,
+  DDS_DIMENSION_TEXTURE3D = 4,
+};
+
+struct DDS_HEADER
+{
+  uint32_t dwSize;
+  uint32_t dwFlags;
+  uint32_t dwHeight;
+  uint32_t dwWidth;
+  uint32_t dwPitchOrLinearSize;
+  uint32_t dwDepth;  // only if DDS_HEADER_FLAGS_VOLUME is set in dwFlags
+  uint32_t dwMipMapCount;
+  uint32_t dwReserved1[11];
+  DDS_PIXELFORMAT ddspf;
+  uint32_t dwCaps;
+  uint32_t dwCaps2;
+  uint32_t dwCaps3;
+  uint32_t dwCaps4;
+  uint32_t dwReserved2;
+};
+
+struct DDS_HEADER_DXT10
+{
+  uint32_t dxgiFormat;
+  uint32_t resourceDimension;
+  uint32_t miscFlag;  // see DDS_RESOURCE_MISC_FLAG
+  uint32_t arraySize;
+  uint32_t miscFlags2;  // see DDS_MISC_FLAGS2
+};
+
+#pragma pack(pop)
+
+static_assert(sizeof(DDS_HEADER) == 124, "DDS Header size mismatch");
+static_assert(sizeof(DDS_HEADER_DXT10) == 20, "DDS DX10 Extended Header size mismatch");
+
+constexpr DDS_PIXELFORMAT DDSPF_A8R8G8B8 = {
+    sizeof(DDS_PIXELFORMAT), DDS_RGBA, 0, 32, 0x00ff0000, 0x0000ff00, 0x000000ff, 0xff000000};
+constexpr DDS_PIXELFORMAT DDSPF_X8R8G8B8 = {
+    sizeof(DDS_PIXELFORMAT), DDS_RGB, 0, 32, 0x00ff0000, 0x0000ff00, 0x000000ff, 0x00000000};
+constexpr DDS_PIXELFORMAT DDSPF_A8B8G8R8 = {
+    sizeof(DDS_PIXELFORMAT), DDS_RGBA, 0, 32, 0x000000ff, 0x0000ff00, 0x00ff0000, 0xff000000};
+constexpr DDS_PIXELFORMAT DDSPF_X8B8G8R8 = {
+    sizeof(DDS_PIXELFORMAT), DDS_RGB, 0, 32, 0x000000ff, 0x0000ff00, 0x00ff0000, 0x00000000};
+constexpr DDS_PIXELFORMAT DDSPF_R8G8B8 = {
+    sizeof(DDS_PIXELFORMAT), DDS_RGB, 0, 24, 0x00ff0000, 0x0000ff00, 0x000000ff, 0x00000000};
+
+// End of Microsoft code from DDS.h.
+
+bool DDSPixelFormatMatches(const DDS_PIXELFORMAT& pf1, const DDS_PIXELFORMAT& pf2)
+{
+  return std::tie(pf1.dwSize, pf1.dwFlags, pf1.dwFourCC, pf1.dwRGBBitCount, pf1.dwRBitMask,
+                  pf1.dwGBitMask, pf1.dwGBitMask, pf1.dwBBitMask, pf1.dwABitMask) ==
+         std::tie(pf2.dwSize, pf2.dwFlags, pf2.dwFourCC, pf2.dwRGBBitCount, pf2.dwRBitMask,
+                  pf2.dwGBitMask, pf2.dwGBitMask, pf2.dwBBitMask, pf2.dwABitMask);
+}
+
+struct DDSLoadInfo
+{
+  u32 block_size = 1;
+  u32 bytes_per_block = 4;
+  u32 width = 0;
+  u32 height = 0;
+  u32 mip_count = 0;
+  HostTextureFormat format = HostTextureFormat::RGBA8;
+  size_t first_mip_offset = 0;
+  size_t first_mip_size = 0;
+  u32 first_mip_row_length = 0;
+
+  std::function<void(HiresTexture::Level*)> conversion_function;
+};
+
+u32 GetBlockCount(u32 extent, u32 block_size)
+{
+  return std::max(Common::AlignUp(extent, block_size) / block_size, 1u);
+}
+
+HiresTexture::ImageDataPointer AllocateLevelData(size_t size)
+{
+  return HiresTexture::ImageDataPointer(new u8[size], [](u8* data) { delete[] data; });
+}
+
+void ConvertTexture_X8B8G8R8(HiresTexture::Level* level)
+{
+  u8* data_ptr = level->data.get();
+  for (u32 row = 0; row < level->height; row++)
+  {
+    for (u32 x = 0; x < level->row_length; x++)
+    {
+      // Set alpha channel to full intensity.
+      data_ptr[3] = 0xFF;
+      data_ptr += sizeof(u32);
+    }
+  }
+}
+
+void ConvertTexture_A8R8G8B8(HiresTexture::Level* level)
+{
+  u8* data_ptr = level->data.get();
+  for (u32 row = 0; row < level->height; row++)
+  {
+    for (u32 x = 0; x < level->row_length; x++)
+    {
+      // Byte swap ABGR -> RGBA
+      u32 val;
+      std::memcpy(&val, data_ptr, sizeof(val));
+      val = ((val & 0xFF00FF00) | ((val >> 16) & 0xFF) | ((val << 16) & 0xFF0000));
+      std::memcpy(data_ptr, &val, sizeof(u32));
+      data_ptr += sizeof(u32);
+    }
+  }
+}
+
+void ConvertTexture_X8R8G8B8(HiresTexture::Level* level)
+{
+  u8* data_ptr = level->data.get();
+  for (u32 row = 0; row < level->height; row++)
+  {
+    for (u32 x = 0; x < level->row_length; x++)
+    {
+      // Byte swap XBGR -> RGBX, and set alpha to full intensity.
+      u32 val;
+      std::memcpy(&val, data_ptr, sizeof(val));
+      val = ((val & 0x0000FF00) | ((val >> 16) & 0xFF) | ((val << 16) & 0xFF0000)) | 0xFF000000;
+      std::memcpy(data_ptr, &val, sizeof(u32));
+      data_ptr += sizeof(u32);
+    }
+  }
+}
+
+void ConvertTexture_R8G8B8(HiresTexture::Level* level)
+{
+  // Have to reallocate the buffer for this one, since the data in the file
+  // does not have an alpha byte.
+  level->data_size = level->row_length * level->height * sizeof(u32);
+  HiresTexture::ImageDataPointer rgb_data = AllocateLevelData(level->data_size);
+  std::swap(level->data, rgb_data);
+
+  const u8* rgb_data_ptr = rgb_data.get();
+  u8* data_ptr = level->data.get();
+
+  for (u32 row = 0; row < level->height; row++)
+  {
+    for (u32 x = 0; x < level->row_length; x++)
+    {
+      // This is BGR in memory.
+      u32 val;
+      std::memcpy(&val, rgb_data_ptr, sizeof(val));
+      val = ((val & 0x0000FF00) | ((val >> 16) & 0xFF) | ((val << 16) & 0xFF0000)) | 0xFF000000;
+      std::memcpy(data_ptr, &val, sizeof(u32));
+      data_ptr += sizeof(u32);
+      rgb_data_ptr += 3;
+    }
+  }
+}
+
+bool ParseDDSHeader(File::IOFile& file, DDSLoadInfo* info)
+{
+  // Exit as early as possible for non-DDS textures, since all extensions are currently
+  // passed through this function.
+  u32 magic;
+  if (!file.ReadBytes(&magic, sizeof(magic)) || magic != DDS_MAGIC)
+    return false;
+
+  DDS_HEADER header;
+  size_t header_size = sizeof(header);
+  if (!file.ReadBytes(&header, header_size) || header.dwSize < header_size)
+    return false;
+
+  // Required fields.
+  if ((header.dwFlags & DDS_HEADER_FLAGS_TEXTURE) != DDS_HEADER_FLAGS_TEXTURE)
+    return false;
+
+  // Image should be 2D.
+  if (header.dwFlags & DDS_HEADER_FLAGS_VOLUME)
+    return false;
+
+  // Presence of width/height fields is already tested by DDS_HEADER_FLAGS_TEXTURE.
+  info->width = header.dwWidth;
+  info->height = header.dwHeight;
+  if (info->width == 0 || info->height == 0)
+    return false;
+
+  // Check for mip levels.
+  if (header.dwFlags & DDS_HEADER_FLAGS_MIPMAP)
+  {
+    info->mip_count = header.dwMipMapCount;
+    if (header.dwMipMapCount != 0)
+      info->mip_count = header.dwMipMapCount;
+    else
+      info->mip_count = HiresTexture::CalculateMipCount(info->width, info->height);
+  }
+  else
+  {
+    info->mip_count = 1;
+  }
+
+  // Handle fourcc formats vs uncompressed formats.
+  bool has_fourcc = (header.ddspf.dwFlags & DDS_FOURCC) != 0;
+  bool needs_s3tc = false;
+  if (has_fourcc)
+  {
+    // Handle DX10 extension header.
+    u32 dxt10_format = 0;
+    if (header.ddspf.dwFourCC == MAKEFOURCC('D', 'X', '1', '0'))
+    {
+      DDS_HEADER_DXT10 dxt10_header;
+      if (!file.ReadBytes(&dxt10_header, sizeof(dxt10_header)))
+        return false;
+
+      // Can't handle array textures here. Doesn't make sense to use them, anyway.
+      if (dxt10_header.resourceDimension != DDS_DIMENSION_TEXTURE2D || dxt10_header.arraySize != 1)
+        return false;
+
+      header_size += sizeof(dxt10_header);
+      dxt10_format = dxt10_header.dxgiFormat;
+    }
+
+    // Currently, we only handle compressed textures here, and leave the rest to the SOIL loader.
+    // In the future, this could be extended, but these isn't much benefit in doing so currently.
+    if (header.ddspf.dwFourCC == MAKEFOURCC('D', 'X', 'T', '1') || dxt10_format == 71)
+    {
+      info->format = HostTextureFormat::DXT1;
+      info->block_size = 4;
+      info->bytes_per_block = 8;
+      needs_s3tc = true;
+    }
+    else if (header.ddspf.dwFourCC == MAKEFOURCC('D', 'X', 'T', '3') || dxt10_format == 74)
+    {
+      info->format = HostTextureFormat::DXT3;
+      info->block_size = 4;
+      info->bytes_per_block = 16;
+      needs_s3tc = true;
+    }
+    else if (header.ddspf.dwFourCC == MAKEFOURCC('D', 'X', 'T', '5') || dxt10_format == 77)
+    {
+      info->format = HostTextureFormat::DXT5;
+      info->block_size = 4;
+      info->bytes_per_block = 16;
+      needs_s3tc = true;
+    }
+    else
+    {
+      // Leave all remaining formats to SOIL.
+      return false;
+    }
+  }
+  else
+  {
+    if (DDSPixelFormatMatches(header.ddspf, DDSPF_A8R8G8B8))
+    {
+      info->conversion_function = ConvertTexture_A8R8G8B8;
+    }
+    else if (DDSPixelFormatMatches(header.ddspf, DDSPF_X8R8G8B8))
+    {
+      info->conversion_function = ConvertTexture_X8R8G8B8;
+    }
+    else if (DDSPixelFormatMatches(header.ddspf, DDSPF_X8B8G8R8))
+    {
+      info->conversion_function = ConvertTexture_X8B8G8R8;
+    }
+    else if (DDSPixelFormatMatches(header.ddspf, DDSPF_R8G8B8))
+    {
+      info->conversion_function = ConvertTexture_R8G8B8;
+    }
+    else if (DDSPixelFormatMatches(header.ddspf, DDSPF_A8B8G8R8))
+    {
+      // This format is already in RGBA order, so no conversion necessary.
+    }
+    else
+    {
+      return false;
+    }
+
+    // All these formats are RGBA, just with byte swapping.
+    info->format = HostTextureFormat::RGBA8;
+    info->block_size = 1;
+    info->bytes_per_block = header.ddspf.dwRGBBitCount / 8;
+  }
+
+  // We also need to ensure the backend supports these formats natively before loading them,
+  // otherwise, fallback to SOIL, which will decompress them to RGBA.
+  if (needs_s3tc && !g_ActiveConfig.backend_info.bSupportsST3CTextures)
+    return false;
+
+  // Mip levels smaller than the block size are padded to multiples of the block size.
+  u32 blocks_wide = GetBlockCount(info->width, info->block_size);
+  u32 blocks_high = GetBlockCount(info->height, info->block_size);
+
+  // Pitch can be specified in the header, otherwise we can derive it from the dimensions. For
+  // compressed formats, both DDS_HEADER_FLAGS_LINEARSIZE and DDS_HEADER_FLAGS_PITCH should be
+  // set. See https://msdn.microsoft.com/en-us/library/windows/desktop/bb943982(v=vs.85).aspx
+  if (header.dwFlags & DDS_HEADER_FLAGS_PITCH && header.dwFlags & DDS_HEADER_FLAGS_LINEARSIZE)
+  {
+    // Convert pitch (in bytes) to texels/row length.
+    if (header.dwPitchOrLinearSize < info->bytes_per_block)
+    {
+      // Likely a corrupted or invalid file.
+      return false;
+    }
+
+    info->first_mip_row_length =
+        std::max(header.dwPitchOrLinearSize / info->bytes_per_block, 1u) * info->block_size;
+    info->first_mip_size = static_cast<size_t>(info->first_mip_row_length / info->block_size) *
+                           info->block_size * blocks_high;
+  }
+  else
+  {
+    // Assume no padding between rows of blocks.
+    info->first_mip_row_length = blocks_wide * info->block_size;
+    info->first_mip_size = blocks_wide * static_cast<size_t>(info->bytes_per_block) * blocks_high;
+  }
+
+  // Check for truncated or corrupted files.
+  info->first_mip_offset = sizeof(magic) + header_size;
+  if (info->first_mip_offset >= file.GetSize())
+    return false;
+
+  return true;
+}
+
+bool ReadMipLevel(HiresTexture::Level* level, File::IOFile& file, const DDSLoadInfo& info,
+                  u32 width, u32 height, u32 row_length, size_t size)
+{
+  // Copy to the final storage location. The deallocator here is simple, nothing extra is
+  // needed, compared to the SOIL-based loader.
+  level->width = width;
+  level->height = height;
+  level->format = info.format;
+  level->row_length = row_length;
+  level->data_size = size;
+  level->data = AllocateLevelData(level->data_size);
+  if (!file.ReadBytes(level->data.get(), level->data_size))
+  {
+    level->data.reset();
+    return false;
+  }
+
+  // Apply conversion function for uncompressed textures.
+  if (info.conversion_function)
+    info.conversion_function(level);
+
+  return true;
+}
+
+}  // namespace
+
+bool HiresTexture::LoadDDSTexture(HiresTexture* tex, const std::string& filename)
+{
+  File::IOFile file;
+  file.Open(filename, "rb");
+  if (!file.IsOpen())
+    return false;
+
+  DDSLoadInfo info;
+  if (!ParseDDSHeader(file, &info))
+    return false;
+
+  // Read first mip level, as it may have a custom pitch.
+  Level first_level;
+  if (!file.Seek(info.first_mip_offset, SEEK_SET) ||
+      !ReadMipLevel(&first_level, file, info, info.width, info.height, info.first_mip_row_length,
+                    info.first_mip_size))
+  {
+    return false;
+  }
+
+  tex->m_levels.push_back(std::move(first_level));
+
+  // Read in any remaining mip levels in the file.
+  // If the .dds file does not contain a full mip chain, we'll fall back to the old path.
+  u32 mip_width = info.width;
+  u32 mip_height = info.height;
+  for (u32 i = 1; i < info.mip_count; i++)
+  {
+    mip_width = std::max(mip_width / 2, 1u);
+    mip_height = std::max(mip_height / 2, 1u);
+
+    // Pitch can't be specified with each mip level, so we have to calculate it ourselves.
+    u32 blocks_wide = GetBlockCount(mip_width, info.block_size);
+    u32 blocks_high = GetBlockCount(mip_height, info.block_size);
+    u32 mip_row_length = blocks_wide * info.block_size;
+    size_t mip_size = blocks_wide * static_cast<size_t>(info.bytes_per_block) * blocks_high;
+    Level level;
+    if (!ReadMipLevel(&level, file, info, mip_width, mip_height, mip_row_length, mip_size))
+      break;
+
+    tex->m_levels.push_back(std::move(level));
+  }
+
+  return true;
+}
+
+bool HiresTexture::LoadDDSTexture(Level& level, const std::string& filename)
+{
+  // Only loading a single mip level.
+  File::IOFile file;
+  file.Open(filename, "rb");
+  if (!file.IsOpen())
+    return false;
+
+  DDSLoadInfo info;
+  if (!ParseDDSHeader(file, &info))
+    return false;
+
+  return ReadMipLevel(&level, file, info, info.width, info.height, info.first_mip_row_length,
+                      info.first_mip_size);
+}

Source/Core/VideoCommon/TextureCacheBase.cpp

@@ -47,6 +47,29 @@ TextureCacheBase::TCacheEntryBase::~TCacheEntryBase()
 {
 }
 
+bool TextureCacheBase::IsCompressedHostTextureFormat(HostTextureFormat format)
+{
+  // This will need to be changed if we add any other uncompressed formats.
+  return format != HostTextureFormat::RGBA8;
+}
+
+size_t TextureCacheBase::CalculateHostTextureLevelPitch(HostTextureFormat format, u32 row_length)
+{
+  switch (format)
+  {
+  case HostTextureFormat::DXT1:
+    return static_cast<size_t>(std::max(1u, row_length / 4)) * 8;
+
+  case HostTextureFormat::DXT3:
+  case HostTextureFormat::DXT5:
+    return static_cast<size_t>(std::max(1u, row_length / 4)) * 16;
+
+  case HostTextureFormat::RGBA8:
+  default:
+    return static_cast<size_t>(row_length) * 4;
+  }
+}
+
 void TextureCacheBase::CheckTempSize(size_t required_size)
 {
   if (required_size <= temp_size)
@@ -750,8 +773,6 @@ TextureCacheBase::TCacheEntryBase* TextureCacheBase::Load(const u32 stage)
       }
       expandedWidth = level.width;
       expandedHeight = level.height;
-      CheckTempSize(level.data_size);
-      memcpy(temp, level.data.get(), level.data_size);
     }
   }
 
@@ -774,6 +795,7 @@ TextureCacheBase::TCacheEntryBase* TextureCacheBase::Load(const u32 stage)
   config.width = width;
   config.height = height;
   config.levels = texLevels;
+  config.format = hires_tex ? hires_tex->GetFormat() : HostTextureFormat::RGBA8;
 
   TCacheEntryBase* entry = AllocateTexture(config);
   GFX_DEBUGGER_PAUSE_AT(NEXT_NEW_TEXTURE, true);
@@ -784,17 +806,20 @@ TextureCacheBase::TCacheEntryBase* TextureCacheBase::Load(const u32 stage)
   const u8* tlut = &texMem[tlutaddr];
   if (hires_tex)
   {
-    entry->Load(temp, width, height, expandedWidth, 0);
+    const auto& level = hires_tex->m_levels[0];
+    entry->Load(0, level.width, level.height, level.row_length, level.data.get(), level.data_size);
   }
-  else if (decode_on_gpu)
+  if (!hires_tex && decode_on_gpu)
   {
     u32 row_stride = bytes_per_block * (expandedWidth / bsw);
     g_texture_cache->DecodeTextureOnGPU(
         entry, 0, src_data, texture_size, static_cast<TextureFormat>(texformat), width, height,
         expandedWidth, expandedHeight, row_stride, tlut, static_cast<TlutFormat>(tlutfmt));
   }
-  else
+  else if (!hires_tex)
   {
+    size_t decoded_texture_size = expandedWidth * sizeof(u32) * expandedHeight;
+    CheckTempSize(decoded_texture_size);
     if (!(texformat == GX_TF_RGBA8 && from_tmem))
     {
       TexDecoder_Decode(temp, src_data, expandedWidth, expandedHeight, texformat, tlut,
@@ -807,7 +832,7 @@ TextureCacheBase::TCacheEntryBase* TextureCacheBase::Load(const u32 stage)
       TexDecoder_DecodeRGBA8FromTmem(temp, src_data, src_data_gb, expandedWidth, expandedHeight);
     }
 
-    entry->Load(temp, width, height, expandedWidth, 0);
+    entry->Load(0, width, height, expandedWidth, temp, decoded_texture_size);
   }
 
   iter = textures_by_address.emplace(address, entry);
@@ -837,9 +862,8 @@ TextureCacheBase::TCacheEntryBase* TextureCacheBase::Load(const u32 stage)
     for (u32 level_index = 1; level_index != texLevels; ++level_index)
     {
       const auto& level = hires_tex->m_levels[level_index];
-      CheckTempSize(level.data_size);
+      entry->Load(level_index, level.width, level.height, level.row_length, level.data.get(),
-      memcpy(temp, level.data.get(), level.data_size);
+                  level.data_size);
-      entry->Load(temp, level.width, level.height, level.width, level_index);
     }
   }
   else
@@ -877,9 +901,11 @@ TextureCacheBase::TCacheEntryBase* TextureCacheBase::Load(const u32 stage)
       }
       else
       {
+        // No need to call CheckTempSize here, as mips will always be smaller than the base level.
+        size_t decoded_mip_size = expanded_mip_width * sizeof(u32) * expanded_mip_height;
         TexDecoder_Decode(temp, mip_src_data, expanded_mip_width, expanded_mip_height, texformat,
                           tlut, (TlutFormat)tlutfmt);
-        entry->Load(temp, mip_width, mip_height, expanded_mip_width, level);
+        entry->Load(level, mip_width, mip_height, expanded_mip_width, temp, decoded_mip_size);
       }
 
       mip_src_data += mip_size;

Source/Core/VideoCommon/TextureCacheBase.h

@@ -27,16 +27,16 @@ public:
 
     bool operator==(const TCacheEntryConfig& o) const
     {
-      return std::tie(width, height, levels, layers, rendertarget) ==
+      return std::tie(width, height, levels, layers, format, rendertarget) ==
-             std::tie(o.width, o.height, o.levels, o.layers, o.rendertarget);
+             std::tie(o.width, o.height, o.levels, o.layers, o.format, o.rendertarget);
     }
 
     struct Hasher : std::hash<u64>
     {
       size_t operator()(const TCacheEntryConfig& c) const
       {
-        u64 id = (u64)c.rendertarget << 63 | (u64)c.layers << 48 | (u64)c.levels << 32 |
+        u64 id = (u64)c.rendertarget << 63 | (u64)c.format << 50 | (u64)c.layers << 48 |
-                 (u64)c.height << 16 | (u64)c.width;
+                 (u64)c.levels << 32 | (u64)c.height << 16 | (u64)c.width;
         return std::hash<u64>::operator()(id);
       }
     };
@@ -45,6 +45,7 @@ public:
     u32 height = 0;
     u32 levels = 1;
     u32 layers = 1;
+    HostTextureFormat format = HostTextureFormat::RGBA8;
     bool rendertarget = false;
   };
 
@@ -129,7 +130,8 @@ public:
                                           const MathUtil::Rectangle<int>& srcrect,
                                           const MathUtil::Rectangle<int>& dstrect) = 0;
 
-    virtual void Load(const u8* buffer, u32 width, u32 height, u32 expanded_width, u32 level) = 0;
+    virtual void Load(u32 level, u32 width, u32 height, u32 row_length, const u8* buffer,
+                      size_t buffer_size) = 0;
     virtual void FromRenderTarget(bool is_depth_copy, const EFBRectangle& srcRect, bool scaleByHalf,
                                   unsigned int cbufid, const float* colmat) = 0;
 
@@ -144,6 +146,10 @@ public:
 
   virtual ~TextureCacheBase();  // needs virtual for DX11 dtor
 
+  // TODO: Move these to AbstractTexture once it is finished.
+  static bool IsCompressedHostTextureFormat(HostTextureFormat format);
+  static size_t CalculateHostTextureLevelPitch(HostTextureFormat format, u32 row_length);
+
   void OnConfigChanged(VideoConfig& config);
 
   // Removes textures which aren't used for more than TEXTURE_KILL_THRESHOLD frames,

Source/Core/VideoCommon/VideoCommon.h

@@ -77,6 +77,15 @@ enum class APIType
   Nothing
 };
 
+// Texture formats that videocommon can upload/use.
+enum class HostTextureFormat : u32
+{
+  RGBA8,
+  DXT1,
+  DXT3,
+  DXT5
+};
+
 inline u32 RGBA8ToRGBA6ToRGBA8(u32 src)
 {
   u32 color = src;

Source/Core/VideoCommon/VideoCommon.vcxproj

@@ -69,6 +69,7 @@
     <ClCompile Include="FPSCounter.cpp" />
     <ClCompile Include="FramebufferManagerBase.cpp" />
     <ClCompile Include="HiresTextures.cpp" />
+    <ClCompile Include="HiresTextures_DDSLoader.cpp" />
     <ClCompile Include="ImageWrite.cpp" />
     <ClCompile Include="IndexGenerator.cpp" />
     <ClCompile Include="MainBase.cpp" />

Source/Core/VideoCommon/VideoCommon.vcxproj.filters

@@ -164,6 +164,9 @@
     <ClCompile Include="LightingShaderGen.cpp">
       <Filter>Shader Generators</Filter>
     </ClCompile>
+    <ClCompile Include="HiresTextures_DDSLoader.cpp">
+      <Filter>Util</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="CommandProcessor.h" />
@@ -203,9 +206,6 @@
     <ClInclude Include="TextureDecoder.h">
       <Filter>Decoding</Filter>
     </ClInclude>
-    <ClInclude Include="TextureDecoder_Util.h">
-      <Filter>Decoding</Filter>
-    </ClInclude>
     <ClInclude Include="BPFunctions.h">
       <Filter>Register Sections</Filter>
     </ClInclude>

Source/Core/VideoCommon/VideoConfig.cpp

@@ -41,6 +41,7 @@ VideoConfig::VideoConfig()
   backend_info.bSupportsExclusiveFullscreen = false;
   backend_info.bSupportsMultithreading = false;
   backend_info.bSupportsInternalResolutionFrameDumps = false;
+  backend_info.bSupportsST3CTextures = false;
 
   bEnableValidationLayer = false;
   bBackendMultithreading = true;

Source/Core/VideoCommon/VideoConfig.h

@@ -197,6 +197,7 @@ struct VideoConfig final
     bool bSupportsMultithreading;
     bool bSupportsInternalResolutionFrameDumps;
     bool bSupportsGPUTextureDecoding;
+    bool bSupportsST3CTextures;
   } backend_info;
 
   // Utility