Mmapper with Neko's review fixes (#2503)

DeS

.clang-format

@@ -0,0 +1,33 @@
+Standard: Cpp11
+UseTab: ForIndentation
+TabWidth: 1
+IndentWidth: 1
+AccessModifierOffset: -1
+PointerAlignment: Left
+NamespaceIndentation: All
+ColumnLimit: 200
+BreakBeforeBraces: Allman
+BreakConstructorInitializersBeforeComma: true
+BreakBeforeBinaryOperators: false
+BreakBeforeTernaryOperators: false
+AlwaysBreakTemplateDeclarations: true
+AllowShortIfStatementsOnASingleLine: true
+AllowShortBlocksOnASingleLine: true
+AllowShortCaseLabelsOnASingleLine: true
+AllowShortFunctionsOnASingleLine: false
+AllowShortLoopsOnASingleLine: true
+Cpp11BracedListStyle: true
+IndentCaseLabels: false
+SortIncludes: false
+ReflowComments: true
+AlignConsecutiveAssignments: true
+AlignTrailingComments: true
+AlignAfterOpenBracket: DontAlign
+ConstructorInitializerAllOnOneLineOrOnePerLine: false
+BinPackArguments: true
+BinPackParameters: true
+AlwaysBreakAfterReturnType: None
+KeepEmptyLinesAtTheStartOfBlocks: true
+IndentWrappedFunctionNames: false
+
+#SpaceAfterTemplateKeyword: false

.editorconfig

@@ -0,0 +1,7 @@
+root = true
+
+[*.{h,cpp,hpp}]
+charset = utf-8
+indent_style = tab
+indent_size = 4
+trim_trailing_whitespace = true

.gitignore

@@ -38,6 +38,8 @@
 /rpcs3/Debug
 /rpcs3/Release
 /llvm_build
+/Vulkan/Vulkan-build
+/Vulkan/glslang-build
 
 /wxWidgets/lib
 /bin/rpcs3.ini

.gitmodules

@@ -3,7 +3,7 @@
 	url = https://github.com/wxWidgets/wxWidgets
 	ignore = dirty
 [submodule "rpcs3-ffmpeg"]
-	path = ffmpeg
+	path = 3rdparty/ffmpeg
 	url = https://github.com/hrydgard/ppsspp-ffmpeg
 [submodule "asmjit"]
 	path = asmjit
@@ -12,16 +12,32 @@
 [submodule "llvm"]
 	path = llvm
 	url = https://github.com/llvm-mirror/llvm
-	branch = release_36
-[submodule "minidx9"]
-	path = minidx9
-	url = https://github.com/hrydgard/minidx9.git
+	branch = release_38
 [submodule "rsx_program_decompiler"]
 	path = rsx_program_decompiler
 	url = https://github.com/RPCS3/rsx_program_decompiler
 [submodule "GSL"]
-	path = GSL
+	path = 3rdparty/GSL
 	url = https://github.com/Microsoft/GSL.git
 [submodule "libpng"]
-	path = libpng
+	path = 3rdparty/libpng
 	url = https://github.com/RPCS3/libpng
+	ignore = dirty
+[submodule "Vulkan/glslang"]
+	path = Vulkan/glslang
+	url = https://github.com/KhronosGroup/glslang.git
+[submodule "Vulkan/Vulkan-LoaderAndValidationLayers"]
+	path = Vulkan/Vulkan-LoaderAndValidationLayers
+	url = https://github.com/KhronosGroup/Vulkan-LoaderAndValidationLayers
+[submodule "Utilities/yaml-cpp"]
+	path = Utilities/yaml-cpp
+	url = https://github.com/RPCS3/yaml-cpp
+[submodule "3rdparty/pugixml"]
+	path = 3rdparty/pugixml
+	url = https://github.com/RPCS3/pugixml
+[submodule "3rdparty/cereal"]
+	path = 3rdparty/cereal
+	url = https://github.com/USCiLab/cereal.git
+[submodule "rsx-debugger"]
+	path = rsx-debugger
+	url = https://github.com/RPCS3/rsx-debugger.git

.mention-bot

@@ -1,3 +1,3 @@
 {
-  "userBlacklist": ["AlexAltea"]
-}
+  "userBlacklist": ["AlexAltea", "tambry", "DHrpcs3"]
+}

.travis.yml

@@ -1,98 +1,96 @@
 language: cpp
+sudo: false
+dist: trusty
 
 os:
- - linux
- - osx
+  - linux
+  #- osx
 
 osx_image: xcode6.4
 
 compiler:
- - clang
- - gcc
+  - clang
+  - gcc
+
+cache:
+  ccache: true
+  directories:
+    - $HOME/hombebrew_cache
 
 env:
   global:
-   - secure: "Vf+FY48nip9JppMnq11105NealdErSWsoUhHo63/V3V+LKfA9guenxCp93/qoSIdSGC/sJwb0yIIMGvkTT/rxDJNh6Z+BWUTb2E0WEIIQbvTJNOSUzoq7dfF1LT61XjVjByFzcbC2xjtaBowmcAYEs1jGUUuEjYVCMmD5lY8hUg="
-   # Which Travis environment to run Coverity on
-   - coverity_scan_run_condition='"$TRAVIS_OS_NAME" = linux -a "$CC" = gcc'
-   # Test mode is for testing if it's working with Coverity. Change to true if testing, to avoid reaching the quota.
-   - coverity_scan_script_test_mode=false
-
-branches:
-  except:
-    - ppu_recompiler
+    - secure: "Vf+FY48nip9JppMnq11105NealdErSWsoUhHo63/V3V+LKfA9guenxCp93/qoSIdSGC/sJwb0yIIMGvkTT/rxDJNh6Z+BWUTb2E0WEIIQbvTJNOSUzoq7dfF1LT61XjVjByFzcbC2xjtaBowmcAYEs1jGUUuEjYVCMmD5lY8hUg="
+    # Which Travis environment to run Coverity on
+    - coverity_scan_run_condition='"$TRAVIS_OS_NAME" = linux -a "$CC" = gcc'
+    # Test mode is for testing if it's working with Coverity. Change to true if testing, to avoid reaching the quota.
+    - coverity_scan_script_test_mode=false
 
 matrix:
- exclude:
-  - os: osx
-    compiler: gcc
+  exclude:
+    - os: osx
+      compiler: gcc
 
 git:
   submodules: false
 
 before_install:
-# shutdown services on Travis, which may have a memory impact
-  - if [ "$TRAVIS_OS_NAME" = "linux" ]; then
-      sudo apt-get install libwxgtk3.0-dev;
-      sudo apt-add-repository -y ppa:libreoffice/ppa;
-      sudo apt-get update;
-      sudo apt-get install libglew-dev;
-    fi;
   - if [ "$TRAVIS_OS_NAME" = "linux" ] && [ "$CXX" = "g++" ]; then
       export CXX="g++-5" CC="gcc-5" CXXFLAGS="-Wno-format-security";
       export GCC_COLORS='error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01';
-    elif [ "$TRAVIS_OS_NAME" = "linux" ]; then
-      export CXX="clang++-3.6" CC="clang-3.6";
     fi;
-# Add coverall for C++ so coverall.io could be triggered. Even it should be --coverage and gcov.
+  # Add coverall for C++ so coverall.io could be triggered. Even it should be --coverage and gcov.
   - if [ "$TRAVIS_OS_NAME" = "linux" ]; then
-      sudo pip install cpp-coveralls requests[security];
+      pip install --user cpp-coveralls requests[security];
     else
       brew update; brew update;
-      brew install glew wxwidgets llvm36;
+      brew install ccache glew wxwidgets llvm38;
     fi;
 
 before_script:
- - git submodule update --init asmjit ffmpeg rsx_program_decompiler GSL libpng
- - mkdir build
- - cd build
- - if [ "$TRAVIS_OS_NAME" = "linux" ]; then cmake ..; else cmake .. -DLLVM_DIR=/usr/local/opt/llvm36/lib/llvm-3.6/share/llvm/cmake; fi
+  - git submodule update --init rsx_program_decompiler asmjit 3rdparty/ffmpeg 3rdparty/pugixml 3rdparty/GSL 3rdparty/libpng Utilities/yaml-cpp 3rdparty/cereal
+  - mkdir build
+  - cd build
+  - if [ "$TRAVIS_OS_NAME" = "linux" ]; then
+      cmake ..;
+    else
+      cmake .. -DLLVM_DIR=/usr/local/opt/llvm38/lib/llvm-3.8/share/llvm/cmake;
+    fi;
 
 script:
-# Add a command to show all the variables. May be useful for debugging Travis.
-# - echo "--Shell Export Lists START--" ; export -p; echo "--Shell Export Lists STOP--";
-# And to ensure the versions of toolchain
-- echo "--CXX version?"; "$CXX" --version; echo "--CXX version confirmed";
-- if [ "$COVERITY_SCAN_BRANCH" != 1 ]; then make -j 4; fi
+  # Add a command to show all the variables. May be useful for debugging Travis.
+  #- echo "--Shell Export Lists START--" ; export -p; echo "--Shell Export Lists STOP--";
+  # And to ensure the versions of toolchain
+  - echo "--CXX version?"; "$CXX" --version; echo "--CXX version confirmed";
+  - if [ "$COVERITY_SCAN_BRANCH" != 1 ]; then make -j 3; fi
 
 addons:
   apt:
     sources:
-    - ubuntu-toolchain-r-test
-    - llvm-toolchain-precise-3.6
-    - kubuntu-backports
+      - ubuntu-toolchain-r-test
+      #- llvm-toolchain-trusty-3.8 temporarily disabled
     packages:
-    - cmake
-    - libopenal-dev
-    - freeglut3-dev
-#    - libglew-dev apt version is too old
-    - libc6-dev
-    - llvm-3.6
-    - llvm-3.6-dev
-    - libedit-dev
-    - g++-5
-    - gcc-5
-    - clang-3.6
-    - libstdc++-4.8-dev
-    - lib32stdc++6
-    - zlib1g-dev
+      - cmake
+      - libopenal-dev
+      - freeglut3-dev
+      - libglew-dev
+      - libc6-dev
+      #- llvm-3.8
+      #- llvm-3.8-dev
+      - libedit-dev
+      - g++-5
+      - gcc-5
+      #- clang-3.6
+      - libstdc++-5-dev
+      - lib32stdc++6
+      - zlib1g-dev
+      - libwxgtk3.0-dev
   coverity_scan:
     project:
       name: $TRAVIS_REPO_SLUG
       description: "PS3 emulator/debugger"
-    notification_email: raul.tambre@gmail.com
-    build_command: "make -j 4"
+    notification_email: nekotekina@gmail.com
+    build_command: "make -j 3"
     branch_pattern: coverity_scan
 
 after_success:
-- if [ "$COVERITY_SCAN_BRANCH" != 1 ] && [ "$TRAVIS_OS_NAME" = linux ]; then coveralls --extension .c --extension .cpp --extension .h; fi
+  - if [ "$COVERITY_SCAN_BRANCH" != 1 ] && [ "$TRAVIS_OS_NAME" = linux ]; then coveralls --extension .c --extension .cpp --extension .h; fi

3rdparty/XAudio2_7/audiodefs.h

@@ -0,0 +1,263 @@
+/***************************************************************************
+ *
+ *  Copyright (c) Microsoft Corporation.  All rights reserved.
+ *
+ *  File:     audiodefs.h
+ *  Content:  Basic constants and data types for audio work.
+ *
+ *  Remarks:  This header file defines all of the audio format constants and
+ *            structures required for XAudio2 and XACT work.  Providing these
+ *            in a single location avoids certain dependency problems in the
+ *            legacy audio headers (mmreg.h, mmsystem.h, ksmedia.h).
+ *
+ *            NOTE: Including the legacy headers after this one may cause a
+ *            compilation error, because they define some of the same types
+ *            defined here without preprocessor guards to avoid multiple
+ *            definitions.  If a source file needs one of the old headers,
+ *            it must include it before including audiodefs.h.
+ *
+ ***************************************************************************/
+
+#ifndef __AUDIODEFS_INCLUDED__
+#define __AUDIODEFS_INCLUDED__
+
+#include <windef.h>  // For WORD, DWORD, etc.
+
+#pragma pack(push, 1)  // Pack structures to 1-byte boundaries
+
+
+/**************************************************************************
+ *
+ *  WAVEFORMATEX: Base structure for many audio formats.  Format-specific
+ *  extensions can be defined for particular formats by using a non-zero
+ *  cbSize value and adding extra fields to the end of this structure.
+ *
+ ***************************************************************************/
+
+#ifndef _WAVEFORMATEX_
+
+    #define _WAVEFORMATEX_
+    typedef struct tWAVEFORMATEX
+    {
+        WORD wFormatTag;        // Integer identifier of the format
+        WORD nChannels;         // Number of audio channels
+        DWORD nSamplesPerSec;   // Audio sample rate
+        DWORD nAvgBytesPerSec;  // Bytes per second (possibly approximate)
+        WORD nBlockAlign;       // Size in bytes of a sample block (all channels)
+        WORD wBitsPerSample;    // Size in bits of a single per-channel sample
+        WORD cbSize;            // Bytes of extra data appended to this struct
+    } WAVEFORMATEX;
+
+#endif
+
+// Defining pointer types outside of the #if block to make sure they are
+// defined even if mmreg.h or mmsystem.h is #included before this file
+
+typedef WAVEFORMATEX *PWAVEFORMATEX, *NPWAVEFORMATEX, *LPWAVEFORMATEX;
+typedef const WAVEFORMATEX *PCWAVEFORMATEX, *LPCWAVEFORMATEX;
+
+
+/**************************************************************************
+ *
+ *  WAVEFORMATEXTENSIBLE: Extended version of WAVEFORMATEX that should be
+ *  used as a basis for all new audio formats.  The format tag is replaced
+ *  with a GUID, allowing new formats to be defined without registering a
+ *  format tag with Microsoft.  There are also new fields that can be used
+ *  to specify the spatial positions for each channel and the bit packing
+ *  used for wide samples (e.g. 24-bit PCM samples in 32-bit containers).
+ *
+ ***************************************************************************/
+
+#ifndef _WAVEFORMATEXTENSIBLE_
+
+    #define _WAVEFORMATEXTENSIBLE_
+    typedef struct
+    {
+        WAVEFORMATEX Format;          // Base WAVEFORMATEX data
+        union
+        {
+            WORD wValidBitsPerSample; // Valid bits in each sample container
+            WORD wSamplesPerBlock;    // Samples per block of audio data; valid
+                                      // if wBitsPerSample=0 (but rarely used).
+            WORD wReserved;           // Zero if neither case above applies.
+        } Samples;
+        DWORD dwChannelMask;          // Positions of the audio channels
+        GUID SubFormat;               // Format identifier GUID
+    } WAVEFORMATEXTENSIBLE;
+
+#endif
+
+typedef WAVEFORMATEXTENSIBLE *PWAVEFORMATEXTENSIBLE, *LPWAVEFORMATEXTENSIBLE;
+typedef const WAVEFORMATEXTENSIBLE *PCWAVEFORMATEXTENSIBLE, *LPCWAVEFORMATEXTENSIBLE;
+
+
+
+/**************************************************************************
+ *
+ *  Define the most common wave format tags used in WAVEFORMATEX formats.
+ *
+ ***************************************************************************/
+
+#ifndef WAVE_FORMAT_PCM  // Pulse Code Modulation
+
+    // If WAVE_FORMAT_PCM is not defined, we need to define some legacy types
+    // for compatibility with the Windows mmreg.h / mmsystem.h header files.
+
+    // Old general format structure (information common to all formats)
+    typedef struct waveformat_tag
+    {
+        WORD wFormatTag;
+        WORD nChannels;
+        DWORD nSamplesPerSec;
+        DWORD nAvgBytesPerSec;
+        WORD nBlockAlign;
+    } WAVEFORMAT, *PWAVEFORMAT, NEAR *NPWAVEFORMAT, FAR *LPWAVEFORMAT;
+
+    // Specific format structure for PCM data
+    typedef struct pcmwaveformat_tag
+    {
+        WAVEFORMAT wf;
+        WORD wBitsPerSample;
+    } PCMWAVEFORMAT, *PPCMWAVEFORMAT, NEAR *NPPCMWAVEFORMAT, FAR *LPPCMWAVEFORMAT;
+
+    #define WAVE_FORMAT_PCM 0x0001
+
+#endif
+
+#ifndef WAVE_FORMAT_ADPCM  // Microsoft Adaptive Differental PCM
+
+    // Replicate the Microsoft ADPCM type definitions from mmreg.h.
+
+    typedef struct adpcmcoef_tag
+    {
+        short iCoef1;
+        short iCoef2;
+    } ADPCMCOEFSET;
+
+    #pragma warning(push)
+    #pragma warning(disable:4200)  // Disable zero-sized array warnings
+
+    typedef struct adpcmwaveformat_tag {
+        WAVEFORMATEX wfx;
+        WORD wSamplesPerBlock;
+        WORD wNumCoef;
+        ADPCMCOEFSET aCoef[];  // Always 7 coefficient pairs for MS ADPCM
+    } ADPCMWAVEFORMAT;
+
+    #pragma warning(pop)
+
+    #define WAVE_FORMAT_ADPCM 0x0002
+
+#endif
+
+// Other frequently used format tags
+
+#ifndef WAVE_FORMAT_UNKNOWN
+    #define WAVE_FORMAT_UNKNOWN         0x0000 // Unknown or invalid format tag
+#endif
+
+#ifndef WAVE_FORMAT_IEEE_FLOAT
+    #define WAVE_FORMAT_IEEE_FLOAT      0x0003 // 32-bit floating-point
+#endif
+
+#ifndef WAVE_FORMAT_MPEGLAYER3
+    #define WAVE_FORMAT_MPEGLAYER3      0x0055 // ISO/MPEG Layer3
+#endif
+
+#ifndef WAVE_FORMAT_DOLBY_AC3_SPDIF
+    #define WAVE_FORMAT_DOLBY_AC3_SPDIF 0x0092 // Dolby Audio Codec 3 over S/PDIF
+#endif
+
+#ifndef WAVE_FORMAT_WMAUDIO2
+    #define WAVE_FORMAT_WMAUDIO2        0x0161 // Windows Media Audio
+#endif
+
+#ifndef WAVE_FORMAT_WMAUDIO3
+    #define WAVE_FORMAT_WMAUDIO3        0x0162 // Windows Media Audio Pro
+#endif
+
+#ifndef WAVE_FORMAT_WMASPDIF
+    #define WAVE_FORMAT_WMASPDIF        0x0164 // Windows Media Audio over S/PDIF
+#endif
+
+#ifndef WAVE_FORMAT_EXTENSIBLE
+    #define WAVE_FORMAT_EXTENSIBLE      0xFFFE // All WAVEFORMATEXTENSIBLE formats
+#endif
+
+
+/**************************************************************************
+ *
+ *  Define the most common wave format GUIDs used in WAVEFORMATEXTENSIBLE
+ *  formats.  Note that including the Windows ksmedia.h header after this
+ *  one will cause build problems; this cannot be avoided, since ksmedia.h
+ *  defines these macros without preprocessor guards.
+ *
+ ***************************************************************************/
+
+#ifdef __cplusplus // uuid() and __uuidof() are only available in C++
+
+    #ifndef KSDATAFORMAT_SUBTYPE_PCM
+        struct __declspec(uuid("00000001-0000-0010-8000-00aa00389b71")) KSDATAFORMAT_SUBTYPE_PCM_STRUCT;
+        #define KSDATAFORMAT_SUBTYPE_PCM __uuidof(KSDATAFORMAT_SUBTYPE_PCM_STRUCT)
+    #endif
+
+    #ifndef KSDATAFORMAT_SUBTYPE_ADPCM
+        struct __declspec(uuid("00000002-0000-0010-8000-00aa00389b71")) KSDATAFORMAT_SUBTYPE_ADPCM_STRUCT;
+        #define KSDATAFORMAT_SUBTYPE_ADPCM __uuidof(KSDATAFORMAT_SUBTYPE_ADPCM_STRUCT)
+    #endif
+
+    #ifndef KSDATAFORMAT_SUBTYPE_IEEE_FLOAT
+        struct __declspec(uuid("00000003-0000-0010-8000-00aa00389b71")) KSDATAFORMAT_SUBTYPE_IEEE_FLOAT_STRUCT;
+        #define KSDATAFORMAT_SUBTYPE_IEEE_FLOAT __uuidof(KSDATAFORMAT_SUBTYPE_IEEE_FLOAT_STRUCT)
+    #endif
+
+#endif
+
+
+/**************************************************************************
+ *
+ *  Speaker positions used in the WAVEFORMATEXTENSIBLE dwChannelMask field.
+ *
+ ***************************************************************************/
+
+#ifndef SPEAKER_FRONT_LEFT
+    #define SPEAKER_FRONT_LEFT            0x00000001
+    #define SPEAKER_FRONT_RIGHT           0x00000002
+    #define SPEAKER_FRONT_CENTER          0x00000004
+    #define SPEAKER_LOW_FREQUENCY         0x00000008
+    #define SPEAKER_BACK_LEFT             0x00000010
+    #define SPEAKER_BACK_RIGHT            0x00000020
+    #define SPEAKER_FRONT_LEFT_OF_CENTER  0x00000040
+    #define SPEAKER_FRONT_RIGHT_OF_CENTER 0x00000080
+    #define SPEAKER_BACK_CENTER           0x00000100
+    #define SPEAKER_SIDE_LEFT             0x00000200
+    #define SPEAKER_SIDE_RIGHT            0x00000400
+    #define SPEAKER_TOP_CENTER            0x00000800
+    #define SPEAKER_TOP_FRONT_LEFT        0x00001000
+    #define SPEAKER_TOP_FRONT_CENTER      0x00002000
+    #define SPEAKER_TOP_FRONT_RIGHT       0x00004000
+    #define SPEAKER_TOP_BACK_LEFT         0x00008000
+    #define SPEAKER_TOP_BACK_CENTER       0x00010000
+    #define SPEAKER_TOP_BACK_RIGHT        0x00020000
+    #define SPEAKER_RESERVED              0x7FFC0000
+    #define SPEAKER_ALL                   0x80000000
+    #define _SPEAKER_POSITIONS_
+#endif
+
+#ifndef SPEAKER_STEREO
+    #define SPEAKER_MONO             (SPEAKER_FRONT_CENTER)
+    #define SPEAKER_STEREO           (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT)
+    #define SPEAKER_2POINT1          (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_LOW_FREQUENCY)
+    #define SPEAKER_SURROUND         (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_BACK_CENTER)
+    #define SPEAKER_QUAD             (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT)
+    #define SPEAKER_4POINT1          (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT)
+    #define SPEAKER_5POINT1          (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT)
+    #define SPEAKER_7POINT1          (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT | SPEAKER_FRONT_LEFT_OF_CENTER | SPEAKER_FRONT_RIGHT_OF_CENTER)
+    #define SPEAKER_5POINT1_SURROUND (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_SIDE_LEFT | SPEAKER_SIDE_RIGHT)
+    #define SPEAKER_7POINT1_SURROUND (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT | SPEAKER_SIDE_LEFT  | SPEAKER_SIDE_RIGHT)
+#endif
+
+
+#pragma pack(pop)
+
+#endif // #ifndef __AUDIODEFS_INCLUDED__

3rdparty/XAudio2_7/comdecl.h

@@ -0,0 +1,65 @@
+// comdecl.h: Macros to facilitate COM interface and GUID declarations.
+// Copyright (c) Microsoft Corporation.  All rights reserved.
+
+#ifndef _COMDECL_H_
+#define _COMDECL_H_
+
+#ifndef _XBOX
+    #include <basetyps.h>   // For standard COM interface macros
+#else
+    #pragma warning(push)
+    #pragma warning(disable:4061)
+    #include <xtl.h>        // Required by xobjbase.h
+    #include <xobjbase.h>   // Special definitions for Xbox build
+    #pragma warning(pop)
+#endif
+
+// The DEFINE_CLSID() and DEFINE_IID() macros defined below allow COM GUIDs to
+// be declared and defined in such a way that clients can obtain the GUIDs using
+// either the __uuidof() extension or the old-style CLSID_Foo / IID_IFoo names.
+// If using the latter approach, the client can also choose whether to get the
+// GUID definitions by defining the INITGUID preprocessor constant or by linking
+// to a GUID library.  This works in either C or C++.
+
+#ifndef _MSC_VER
+
+#define DEFINE_UUID(name, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) __CRT_UUID_DECL(name, 0x##l, 0x##w1, 0x##w2, 0x##b1, 0x##b2, 0x##b3, 0x##b4, 0x##b5, 0x##b6, 0x##b7, 0x##b8)
+#define DEFINE_CLSID(className, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) class className; DEFINE_UUID(className, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8)
+#define DEFINE_IID(interfaceName, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) struct interfaceName; DEFINE_UUID(interfaceName, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8)
+
+#elif __cplusplus
+
+    #define DECLSPEC_UUID_WRAPPER(x) __declspec(uuid(#x))
+    #ifdef INITGUID
+
+        #define DEFINE_CLSID(className, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) \
+            class DECLSPEC_UUID_WRAPPER(l##-##w1##-##w2##-##b1##b2##-##b3##b4##b5##b6##b7##b8) className; \
+            EXTERN_C const GUID DECLSPEC_SELECTANY CLSID_##className = __uuidof(className)
+
+        #define DEFINE_IID(interfaceName, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) \
+            interface DECLSPEC_UUID_WRAPPER(l##-##w1##-##w2##-##b1##b2##-##b3##b4##b5##b6##b7##b8) interfaceName; \
+            EXTERN_C const GUID DECLSPEC_SELECTANY IID_##interfaceName = __uuidof(interfaceName)
+
+    #else // INITGUID
+
+        #define DEFINE_CLSID(className, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) \
+            class DECLSPEC_UUID_WRAPPER(l##-##w1##-##w2##-##b1##b2##-##b3##b4##b5##b6##b7##b8) className; \
+            EXTERN_C const GUID CLSID_##className
+
+        #define DEFINE_IID(interfaceName, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) \
+            interface DECLSPEC_UUID_WRAPPER(l##-##w1##-##w2##-##b1##b2##-##b3##b4##b5##b6##b7##b8) interfaceName; \
+            EXTERN_C const GUID IID_##interfaceName
+
+    #endif // INITGUID
+
+#else // __cplusplus
+
+    #define DEFINE_CLSID(className, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) \
+        DEFINE_GUID(CLSID_##className, 0x##l, 0x##w1, 0x##w2, 0x##b1, 0x##b2, 0x##b3, 0x##b4, 0x##b5, 0x##b6, 0x##b7, 0x##b8)
+
+    #define DEFINE_IID(interfaceName, l, w1, w2, b1, b2, b3, b4, b5, b6, b7, b8) \
+        DEFINE_GUID(IID_##interfaceName, 0x##l, 0x##w1, 0x##w2, 0x##b1, 0x##b2, 0x##b3, 0x##b4, 0x##b5, 0x##b6, 0x##b7, 0x##b8)
+
+#endif // __cplusplus
+
+#endif // #ifndef _COMDECL_H_

3rdparty/XAudio2_7/dxsdkver.h

@@ -0,0 +1,18 @@
+/*==========================================================================;
+ *
+ *
+ *  File:   dxsdkver.h
+ *  Content:    DirectX SDK Version Include File
+ *
+ ****************************************************************************/
+
+#ifndef _DXSDKVER_H_
+#define _DXSDKVER_H_
+
+#define _DXSDK_PRODUCT_MAJOR  9     
+#define _DXSDK_PRODUCT_MINOR  29        
+#define _DXSDK_BUILD_MAJOR    1962        
+#define _DXSDK_BUILD_MINOR    0        
+
+#endif // _DXSDKVER_H_
+

3rdparty/XAudio2_7/xma2defs.h

@@ -0,0 +1,718 @@
+/***************************************************************************
+ *
+ * Copyright (c) Microsoft Corporation.  All rights reserved.
+ *
+ * File:     xma2defs.h
+ * Content:  Constants, data types and functions for XMA2 compressed audio.
+ *
+ ***************************************************************************/
+
+#ifndef __XMA2DEFS_INCLUDED__
+#define __XMA2DEFS_INCLUDED__
+
+#include <sal.h>        // Markers for documenting API semantics
+#include <winerror.h>   // For S_OK, E_FAIL
+#include <audiodefs.h>  // Basic data types and constants for audio work
+
+
+/***************************************************************************
+ *  Overview
+ ***************************************************************************/
+
+// A typical XMA2 file contains these RIFF chunks:
+//
+// 'fmt' or 'XMA2' chunk (or both): A description of the XMA data's structure
+// and characteristics (length, channels, sample rate, loops, block size, etc).
+//
+// 'seek' chunk: A seek table to help navigate the XMA data.
+//
+// 'data' chunk: The encoded XMA2 data.
+//
+// The encoded XMA2 data is structured as a set of BLOCKS, which contain PACKETS,
+// which contain FRAMES, which contain SUBFRAMES (roughly speaking).  The frames
+// in a file may also be divided into several subsets, called STREAMS.
+//
+// FRAME: A variable-sized segment of XMA data that decodes to exactly 512 mono
+//      or stereo PCM samples.  This is the smallest unit of XMA data that can
+//      be decoded in isolation.  Frames are an arbitrary number of bits in
+//      length, and need not be byte-aligned.  See "XMA frame structure" below.
+//
+// SUBFRAME: A region of bits in an XMA frame that decodes to 128 mono or stereo
+//      samples.  The XMA decoder cannot decode a subframe in isolation; it needs
+//      a whole frame to work with.  However, it can begin emitting the frame's
+//      decoded samples at any one of the four subframe boundaries.  Subframes
+//      can be addressed for seeking and looping purposes.
+//
+// PACKET: A 2Kb region containing a 32-bit header and some XMA frames.  Frames
+//      can (and usually do) span packets.  A packet's header includes the offset
+//      in bits of the first frame that begins within that packet.  All of the
+//      frames that begin in a given packet belong to the same "stream" (see the
+//      Multichannel Audio section below).
+//
+// STREAM: A set of packets within an XMA file that all contain data for the
+//      same mono or stereo component of a PCM file with more than two channels.
+//      The packets comprising a given stream may be interleaved with each other
+//      more or less arbitrarily; see Multichannel Audio.
+//
+// BLOCK: An array of XMA packets; or, to break it down differently, a series of
+//      consecutive XMA frames, padded at the end with reserved data.  A block
+//      must contain at least one 2Kb packet per stream, and it can hold up to
+//      4095 packets (8190Kb), but its size is typically in the 32Kb-128Kb range.
+//      (The size chosen involves a trade-off between memory use and efficiency
+//      of reading from permanent storage.)
+//
+//      XMA frames do not span blocks, so a block is guaranteed to begin with a
+//      set of complete frames, one per stream.  Also, a block in a multi-stream
+//      XMA2 file always contains the same number of samples for each stream;
+//      see Multichannel Audio.
+//
+// The 'data' chunk in an XMA2 file is an array of XMA2WAVEFORMAT.BlockCount XMA
+// blocks, all the same size (as specified in XMA2WAVEFORMAT.BlockSizeInBytes)
+// except for the last one, which may be shorter.
+
+
+// MULTICHANNEL AUDIO: the XMA decoder can only decode raw XMA data into either
+// mono or stereo PCM data.  In order to encode a 6-channel file (say), the file
+// must be deinterleaved into 3 stereo streams that are encoded independently,
+// producing 3 encoded XMA data streams.  Then the packets in these 3 streams
+// are interleaved to produce a single XMA2 file, and some information is added
+// to the file so that the original 6-channel audio can be reconstructed at
+// decode time.  This works using the concept of an XMA stream (see above).
+//
+// The frames for all the streams in an XMA file are interleaved in an arbitrary
+// order.  To locate a frame that belongs to a given stream in a given XMA block,
+// you must examine the first few packets in the block.  Here (and only here) the
+// packets are guaranteed to be presented in stream order, so that all frames
+// beginning in packet 0 belong to stream 0 (the first stereo pair), etc.
+//
+// (This means that when decoding multi-stream XMA files, only entire XMA blocks
+// should be submitted to the decoder; otherwise it cannot know which frames
+// belong to which stream.)
+//
+// Once you have one frame that belongs to a given stream, you can find the next
+// one by looking at the frame's 'NextFrameOffsetBits' value (which is stored in
+// its first 15 bits; see XMAFRAME below).  The GetXmaFrameBitPosition function
+// uses this technique.
+
+
+// SEEKING IN XMA2 FILES: Here is some pseudocode to find the byte position and
+// subframe in an XMA2 file which will contain sample S when decoded.
+//
+// 1. Traverse the seek table to find the XMA2 block containing sample S. The
+//    seek table is an array of big-endian DWORDs, one per block in the file.
+//    The Nth DWORD is the total number of PCM samples that would be obtained
+//    by decoding the entire XMA file up to the end of block N.  Hence, the
+//    block we want is the first one whose seek table entry is greater than S.
+//    (See the GetXmaBlockContainingSample helper function.)
+//
+// 2. Calculate which frame F within the block found above contains sample S.
+//    Since each frame decodes to 512 samples, this is straightforward.  The
+//    first frame in the block produces samples X to X + 512, where X is the
+//    seek table entry for the prior block.  So F is (S - X) / 512.
+//
+// 3. Find the bit offset within the block where frame F starts.  Since frames
+//    are variable-sized, this can only be done by traversing all the frames in
+//    the block until we reach frame F.  (See GetXmaFrameBitPosition.)
+//
+// 4. Frame F has four 128-sample subframes.  To find the subframe containing S,
+//    we can use the formula (S % 512) / 128.
+//
+// In the case of multi-stream XMA files, sample S is a multichannel sample with
+// parts coming from several frames, one per stream.  To find all these frames,
+// steps 2-4 need to be repeated for each stream N, using the knowledge that the
+// first packets in a block are presented in stream order.  The frame traversal
+// in step 3 must be started at the first frame in the Nth packet of the block,
+// which will be the first frame for stream N.  (And the packet header will tell
+// you the first frame's start position within the packet.)
+//
+// Step 1 can be performed using the GetXmaBlockContainingSample function below,
+// and steps 2-4 by calling GetXmaDecodePositionForSample once for each stream.
+
+
+
+/***************************************************************************
+ *  XMA constants
+ ***************************************************************************/
+
+// Size of the PCM samples produced by the XMA decoder
+#define XMA_OUTPUT_SAMPLE_BYTES         2u
+#define XMA_OUTPUT_SAMPLE_BITS          (XMA_OUTPUT_SAMPLE_BYTES * 8u)
+
+// Size of an XMA packet
+#define XMA_BYTES_PER_PACKET            2048u
+#define XMA_BITS_PER_PACKET             (XMA_BYTES_PER_PACKET * 8u)
+
+// Size of an XMA packet header
+#define XMA_PACKET_HEADER_BYTES         4u
+#define XMA_PACKET_HEADER_BITS          (XMA_PACKET_HEADER_BYTES * 8u)
+
+// Sample blocks in a decoded XMA frame
+#define XMA_SAMPLES_PER_FRAME           512u
+
+// Sample blocks in a decoded XMA subframe
+#define XMA_SAMPLES_PER_SUBFRAME        128u
+
+// Maximum encoded data that can be submitted to the XMA decoder at a time
+#define XMA_READBUFFER_MAX_PACKETS      4095u
+#define XMA_READBUFFER_MAX_BYTES        (XMA_READBUFFER_MAX_PACKETS * XMA_BYTES_PER_PACKET)
+
+// Maximum size allowed for the XMA decoder's output buffers
+#define XMA_WRITEBUFFER_MAX_BYTES       (31u * 256u)
+
+// Required byte alignment of the XMA decoder's output buffers
+#define XMA_WRITEBUFFER_BYTE_ALIGNMENT  256u
+
+// Decode chunk sizes for the XMA_PLAYBACK_INIT.subframesToDecode field
+#define XMA_MIN_SUBFRAMES_TO_DECODE     1u
+#define XMA_MAX_SUBFRAMES_TO_DECODE     8u
+#define XMA_OPTIMAL_SUBFRAMES_TO_DECODE 4u
+
+// LoopCount<255 means finite repetitions; LoopCount=255 means infinite looping
+#define XMA_MAX_LOOPCOUNT               254u
+#define XMA_INFINITE_LOOP               255u
+
+
+
+/***************************************************************************
+ *  XMA format structures
+ ***************************************************************************/
+
+// The currently recommended way to express format information for XMA2 files
+// is the XMA2WAVEFORMATEX structure.  This structure is fully compliant with
+// the WAVEFORMATEX standard and contains all the information needed to parse
+// and manage XMA2 files in a compact way.
+
+#define WAVE_FORMAT_XMA2 0x166
+
+typedef struct XMA2WAVEFORMATEX
+{
+    WAVEFORMATEX wfx;
+    // Meaning of the WAVEFORMATEX fields here:
+    //    wFormatTag;        // Audio format type; always WAVE_FORMAT_XMA2
+    //    nChannels;         // Channel count of the decoded audio
+    //    nSamplesPerSec;    // Sample rate of the decoded audio
+    //    nAvgBytesPerSec;   // Used internally by the XMA encoder
+    //    nBlockAlign;       // Decoded sample size; channels * wBitsPerSample / 8
+    //    wBitsPerSample;    // Bits per decoded mono sample; always 16 for XMA
+    //    cbSize;            // Size in bytes of the rest of this structure (34)
+
+    WORD  NumStreams;        // Number of audio streams (1 or 2 channels each)
+    DWORD ChannelMask;       // Spatial positions of the channels in this file,
+                             // stored as SPEAKER_xxx values (see audiodefs.h)
+    DWORD SamplesEncoded;    // Total number of PCM samples the file decodes to
+    DWORD BytesPerBlock;     // XMA block size (but the last one may be shorter)
+    DWORD PlayBegin;         // First valid sample in the decoded audio
+    DWORD PlayLength;        // Length of the valid part of the decoded audio
+    DWORD LoopBegin;         // Beginning of the loop region in decoded sample terms
+    DWORD LoopLength;        // Length of the loop region in decoded sample terms
+    BYTE  LoopCount;         // Number of loop repetitions; 255 = infinite
+    BYTE  EncoderVersion;    // Version of XMA encoder that generated the file
+    WORD  BlockCount;        // XMA blocks in file (and entries in its seek table)
+} XMA2WAVEFORMATEX, *PXMA2WAVEFORMATEX;
+
+
+// The legacy XMA format structures are described here for reference, but they
+// should not be used in new content.  XMAWAVEFORMAT was the structure used in
+// XMA version 1 files.  XMA2WAVEFORMAT was used in early XMA2 files; it is not
+// placed in the usual 'fmt' RIFF chunk but in its own 'XMA2' chunk.
+
+#ifndef WAVE_FORMAT_XMA
+#define WAVE_FORMAT_XMA 0x0165
+
+// Values used in the ChannelMask fields below.  Similar to the SPEAKER_xxx
+// values defined in audiodefs.h, but modified to fit in a single byte.
+#ifndef XMA_SPEAKER_LEFT
+    #define XMA_SPEAKER_LEFT            0x01
+    #define XMA_SPEAKER_RIGHT           0x02
+    #define XMA_SPEAKER_CENTER          0x04
+    #define XMA_SPEAKER_LFE             0x08
+    #define XMA_SPEAKER_LEFT_SURROUND   0x10
+    #define XMA_SPEAKER_RIGHT_SURROUND  0x20
+    #define XMA_SPEAKER_LEFT_BACK       0x40
+    #define XMA_SPEAKER_RIGHT_BACK      0x80
+#endif
+
+
+// Used in XMAWAVEFORMAT for per-stream data
+typedef struct XMASTREAMFORMAT
+{
+    DWORD PsuedoBytesPerSec; // Used by the XMA encoder (typo preserved for legacy reasons)
+    DWORD SampleRate;        // The stream's decoded sample rate (in XMA2 files,
+                             // this is the same for all streams in the file).
+    DWORD LoopStart;         // Bit offset of the frame containing the loop start
+                             // point, relative to the beginning of the stream.
+    DWORD LoopEnd;           // Bit offset of the frame containing the loop end.
+    BYTE  SubframeData;      // Two 4-bit numbers specifying the exact location of
+                             // the loop points within the frames that contain them.
+                             //   SubframeEnd: Subframe of the loop end frame where
+                             //                the loop ends.  Ranges from 0 to 3.
+                             //   SubframeSkip: Subframes to skip in the start frame to
+                             //                 reach the loop.  Ranges from 0 to 4.
+    BYTE  Channels;          // Number of channels in the stream (1 or 2)
+    WORD  ChannelMask;       // Spatial positions of the channels in the stream
+} XMASTREAMFORMAT;
+
+// Legacy XMA1 format structure
+typedef struct XMAWAVEFORMAT
+{
+    WORD FormatTag;          // Audio format type (always WAVE_FORMAT_XMA)
+    WORD BitsPerSample;      // Bit depth (currently required to be 16)
+    WORD EncodeOptions;      // Options for XMA encoder/decoder
+    WORD LargestSkip;        // Largest skip used in interleaving streams
+    WORD NumStreams;         // Number of interleaved audio streams
+    BYTE LoopCount;          // Number of loop repetitions; 255 = infinite
+    BYTE Version;            // XMA encoder version that generated the file.
+                             // Always 3 or higher for XMA2 files.
+    XMASTREAMFORMAT XmaStreams[1]; // Per-stream format information; the actual
+                                   // array length is in the NumStreams field.
+} XMAWAVEFORMAT;
+
+
+// Used in XMA2WAVEFORMAT for per-stream data
+typedef struct XMA2STREAMFORMAT
+{
+    BYTE Channels;           // Number of channels in the stream (1 or 2)
+    BYTE RESERVED;           // Reserved for future use
+    WORD ChannelMask;        // Spatial positions of the channels in the stream
+} XMA2STREAMFORMAT;
+
+// Legacy XMA2 format structure (big-endian byte ordering)
+typedef struct XMA2WAVEFORMAT
+{
+    BYTE  Version;           // XMA encoder version that generated the file.
+                             // Always 3 or higher for XMA2 files.
+    BYTE  NumStreams;        // Number of interleaved audio streams
+    BYTE  RESERVED;          // Reserved for future use
+    BYTE  LoopCount;         // Number of loop repetitions; 255 = infinite
+    DWORD LoopBegin;         // Loop begin point, in samples
+    DWORD LoopEnd;           // Loop end point, in samples
+    DWORD SampleRate;        // The file's decoded sample rate
+    DWORD EncodeOptions;     // Options for the XMA encoder/decoder
+    DWORD PsuedoBytesPerSec; // Used internally by the XMA encoder
+    DWORD BlockSizeInBytes;  // Size in bytes of this file's XMA blocks (except
+                             // possibly the last one).  Always a multiple of
+                             // 2Kb, since XMA blocks are arrays of 2Kb packets.
+    DWORD SamplesEncoded;    // Total number of PCM samples encoded in this file
+    DWORD SamplesInSource;   // Actual number of PCM samples in the source
+                             // material used to generate this file
+    DWORD BlockCount;        // Number of XMA blocks in this file (and hence
+                             // also the number of entries in its seek table)
+    XMA2STREAMFORMAT Streams[1]; // Per-stream format information; the actual
+                                 // array length is in the NumStreams field.
+} XMA2WAVEFORMAT;
+
+#endif // #ifndef WAVE_FORMAT_XMA
+
+
+
+/***************************************************************************
+ *  XMA packet structure (in big-endian form)
+ ***************************************************************************/
+
+typedef struct XMA2PACKET
+{
+    int FrameCount        :  6;  // Number of XMA frames that begin in this packet
+    int FrameOffsetInBits : 15;  // Bit of XmaData where the first complete frame begins
+    int PacketMetaData    :  3;  // Metadata stored in the packet (always 1 for XMA2)
+    int PacketSkipCount   :  8;  // How many packets belonging to other streams must be
+                                 // skipped to find the next packet belonging to this one
+    BYTE XmaData[XMA_BYTES_PER_PACKET - sizeof(DWORD)];  // XMA encoded data
+} XMA2PACKET;
+
+// E.g. if the first DWORD of a packet is 0x30107902:
+//
+// 001100 000001000001111 001 00000010
+//    |          |         |      |____ Skip 2 packets to find the next one for this stream
+//    |          |         |___________ XMA2 signature (always 001)
+//    |          |_____________________ First frame starts 527 bits into packet
+//    |________________________________ Packet contains 12 frames
+
+
+// Helper functions to extract the fields above from an XMA packet.  (Note that
+// the bitfields cannot be read directly on little-endian architectures such as
+// the Intel x86, as they are laid out in big-endian form.)
+
+__inline DWORD GetXmaPacketFrameCount(__in_bcount(1) const BYTE* pPacket)
+{
+    return (DWORD)(pPacket[0] >> 2);
+}
+
+__inline DWORD GetXmaPacketFirstFrameOffsetInBits(__in_bcount(3) const BYTE* pPacket)
+{
+    return ((DWORD)(pPacket[0] & 0x3) << 13) |
+           ((DWORD)(pPacket[1]) << 5) |
+           ((DWORD)(pPacket[2]) >> 3);
+}
+
+__inline DWORD GetXmaPacketMetadata(__in_bcount(3) const BYTE* pPacket)
+{
+    return (DWORD)(pPacket[2] & 0x7);
+}
+
+__inline DWORD GetXmaPacketSkipCount(__in_bcount(4) const BYTE* pPacket)
+{
+    return (DWORD)(pPacket[3]);
+}
+
+
+
+/***************************************************************************
+ *  XMA frame structure
+ ***************************************************************************/
+
+// There is no way to represent the XMA frame as a C struct, since it is a
+// variable-sized string of bits that need not be stored at a byte-aligned
+// position in memory.  This is the layout:
+//
+// XMAFRAME
+// {
+//    LengthInBits: A 15-bit number representing the length of this frame.
+//    XmaData: Encoded XMA data; its size in bits is (LengthInBits - 15).
+// }
+
+// Size in bits of the frame's initial LengthInBits field
+#define XMA_BITS_IN_FRAME_LENGTH_FIELD 15
+
+// Special LengthInBits value that marks an invalid final frame
+#define XMA_FINAL_FRAME_MARKER 0x7FFF
+
+
+
+/***************************************************************************
+ *  XMA helper functions
+ ***************************************************************************/
+
+// We define a local ASSERT macro to equal the global one if it exists.
+// You can define XMA2DEFS_ASSERT in advance to override this default.
+#ifndef XMA2DEFS_ASSERT
+    #ifdef ASSERT
+        #define XMA2DEFS_ASSERT ASSERT
+    #else
+        #define XMA2DEFS_ASSERT(a) /* No-op by default */
+    #endif
+#endif
+
+
+// GetXmaBlockContainingSample: Use a given seek table to find the XMA block
+// containing a given decoded sample.  Note that the seek table entries in an
+// XMA file are stored in big-endian form and may need to be converted prior
+// to calling this function.
+
+__inline HRESULT GetXmaBlockContainingSample
+(
+    DWORD nBlockCount,                      // Blocks in the file (= seek table entries)
+    __in_ecount(nBlockCount) const DWORD* pSeekTable,  // Pointer to the seek table data
+    DWORD nDesiredSample,                   // Decoded sample to locate
+    __out DWORD* pnBlockContainingSample,   // Index of the block containing the sample
+    __out DWORD* pnSampleOffsetWithinBlock  // Position of the sample in this block
+)
+{
+    DWORD nPreviousTotalSamples = 0;
+    DWORD nBlock;
+    DWORD nTotalSamplesSoFar;
+
+    XMA2DEFS_ASSERT(pSeekTable);
+    XMA2DEFS_ASSERT(pnBlockContainingSample);
+    XMA2DEFS_ASSERT(pnSampleOffsetWithinBlock);
+
+    for (nBlock = 0; nBlock < nBlockCount; ++nBlock)
+    {
+        nTotalSamplesSoFar = pSeekTable[nBlock];
+        if (nTotalSamplesSoFar > nDesiredSample)
+        {
+            *pnBlockContainingSample = nBlock;
+            *pnSampleOffsetWithinBlock = nDesiredSample - nPreviousTotalSamples;
+            return S_OK;
+        }
+        nPreviousTotalSamples = nTotalSamplesSoFar;
+    }
+
+    return E_FAIL;
+}
+
+
+// GetXmaFrameLengthInBits: Reads a given frame's LengthInBits field.
+
+__inline DWORD GetXmaFrameLengthInBits
+(
+    __in_bcount(nBitPosition / 8 + 3)
+    __in const BYTE* pPacket,  // Pointer to XMA packet[s] containing the frame
+    DWORD nBitPosition         // Bit offset of the frame within this packet
+)
+{
+    DWORD nRegion;
+    DWORD nBytePosition = nBitPosition / 8;
+    DWORD nBitOffset = nBitPosition % 8;
+
+    if (nBitOffset < 2) // Only need to read 2 bytes (and might not be safe to read more)
+    {
+        nRegion = (DWORD)(pPacket[nBytePosition+0]) << 8 |
+                  (DWORD)(pPacket[nBytePosition+1]);
+        return (nRegion >> (1 - nBitOffset)) & 0x7FFF;  // Last 15 bits
+    }
+    else // Need to read 3 bytes
+    {
+        nRegion = (DWORD)(pPacket[nBytePosition+0]) << 16 |
+                  (DWORD)(pPacket[nBytePosition+1]) << 8 |
+                  (DWORD)(pPacket[nBytePosition+2]);
+        return (nRegion >> (9 - nBitOffset)) & 0x7FFF;  // Last 15 bits
+    }
+}
+
+
+// GetXmaFrameBitPosition: Calculates the bit offset of a given frame within
+// an XMA block or set of blocks.  Returns 0 on failure.
+
+__inline DWORD GetXmaFrameBitPosition
+(
+    __in_bcount(nXmaDataBytes) const BYTE* pXmaData,  // Pointer to XMA block[s]
+    DWORD nXmaDataBytes,                              // Size of pXmaData in bytes
+    DWORD nStreamIndex,                               // Stream within which to seek
+    DWORD nDesiredFrame                               // Frame sought
+)
+{
+    const BYTE* pCurrentPacket;
+    DWORD nPacketsExamined = 0;
+    DWORD nFrameCountSoFar = 0;
+    DWORD nFramesToSkip;
+    DWORD nFrameBitOffset;
+
+    XMA2DEFS_ASSERT(pXmaData);
+    XMA2DEFS_ASSERT(nXmaDataBytes % XMA_BYTES_PER_PACKET == 0);
+
+    // Get the first XMA packet belonging to the desired stream, relying on the
+    // fact that the first packets for each stream are in consecutive order at
+    // the beginning of an XMA block.
+
+    pCurrentPacket = pXmaData + nStreamIndex * XMA_BYTES_PER_PACKET;
+    for (;;)
+    {
+        // If we have exceeded the size of the XMA data, return failure
+        if (pCurrentPacket + XMA_BYTES_PER_PACKET > pXmaData + nXmaDataBytes)
+        {
+            return 0;
+        }
+
+        // If the current packet contains the frame we are looking for...
+        if (nFrameCountSoFar + GetXmaPacketFrameCount(pCurrentPacket) > nDesiredFrame)
+        {
+            // See how many frames in this packet we need to skip to get to it
+            XMA2DEFS_ASSERT(nDesiredFrame >= nFrameCountSoFar);
+            nFramesToSkip = nDesiredFrame - nFrameCountSoFar;
+
+            // Get the bit offset of the first frame in this packet
+            nFrameBitOffset = XMA_PACKET_HEADER_BITS + GetXmaPacketFirstFrameOffsetInBits(pCurrentPacket);
+
+            // Advance nFrameBitOffset to the frame of interest
+            while (nFramesToSkip--)
+            {
+                nFrameBitOffset += GetXmaFrameLengthInBits(pCurrentPacket, nFrameBitOffset);
+            }
+
+            // The bit offset to return is the number of bits from pXmaData to
+            // pCurrentPacket plus the bit offset of the frame of interest
+            return (DWORD)(pCurrentPacket - pXmaData) * 8 + nFrameBitOffset;
+        }
+
+        // If we haven't found the right packet yet, advance our counters
+        ++nPacketsExamined;
+        nFrameCountSoFar += GetXmaPacketFrameCount(pCurrentPacket);
+
+        // And skip to the next packet belonging to the same stream
+        pCurrentPacket += XMA_BYTES_PER_PACKET * (GetXmaPacketSkipCount(pCurrentPacket) + 1);
+    }
+}
+
+
+// GetLastXmaFrameBitPosition: Calculates the bit offset of the last complete
+// frame in an XMA block or set of blocks.
+
+__inline DWORD GetLastXmaFrameBitPosition
+(
+    __in_bcount(nXmaDataBytes) const BYTE* pXmaData,  // Pointer to XMA block[s]
+    DWORD nXmaDataBytes,                              // Size of pXmaData in bytes
+    DWORD nStreamIndex                                // Stream within which to seek
+)
+{
+    const BYTE* pLastPacket;
+    DWORD nBytesToNextPacket;
+    DWORD nFrameBitOffset;
+    DWORD nFramesInLastPacket;
+
+    XMA2DEFS_ASSERT(pXmaData);
+    XMA2DEFS_ASSERT(nXmaDataBytes % XMA_BYTES_PER_PACKET == 0);
+    XMA2DEFS_ASSERT(nXmaDataBytes >= XMA_BYTES_PER_PACKET * (nStreamIndex + 1));
+
+    // Get the first XMA packet belonging to the desired stream, relying on the
+    // fact that the first packets for each stream are in consecutive order at
+    // the beginning of an XMA block.
+    pLastPacket = pXmaData + nStreamIndex * XMA_BYTES_PER_PACKET;
+
+    // Search for the last packet belonging to the desired stream
+    for (;;)
+    {
+        nBytesToNextPacket = XMA_BYTES_PER_PACKET * (GetXmaPacketSkipCount(pLastPacket) + 1);
+        XMA2DEFS_ASSERT(nBytesToNextPacket);
+        if (pLastPacket + nBytesToNextPacket + XMA_BYTES_PER_PACKET > pXmaData + nXmaDataBytes)
+        {
+            break;  // The next packet would extend beyond the end of pXmaData
+        }
+        pLastPacket += nBytesToNextPacket;
+    }
+
+    // The last packet can sometimes have no seekable frames, in which case we
+    // have to use the previous one
+    if (GetXmaPacketFrameCount(pLastPacket) == 0)
+    {
+        pLastPacket -= nBytesToNextPacket;
+    }
+
+    // Found the last packet.  Get the bit offset of its first frame.
+    nFrameBitOffset = XMA_PACKET_HEADER_BITS + GetXmaPacketFirstFrameOffsetInBits(pLastPacket);
+
+    // Traverse frames until we reach the last one
+    nFramesInLastPacket = GetXmaPacketFrameCount(pLastPacket);
+    while (--nFramesInLastPacket)
+    {
+        nFrameBitOffset += GetXmaFrameLengthInBits(pLastPacket, nFrameBitOffset);
+    }
+
+    // The bit offset to return is the number of bits from pXmaData to
+    // pLastPacket plus the offset of the last frame in this packet.
+    return (DWORD)(pLastPacket - pXmaData) * 8 + nFrameBitOffset;
+}
+
+
+// GetXmaDecodePositionForSample: Obtains the information needed to make the
+// decoder generate audio starting at a given sample position relative to the
+// beginning of the given XMA block: the bit offset of the appropriate frame,
+// and the right subframe within that frame.  This data can be passed directly
+// to the XMAPlaybackSetDecodePosition function.
+
+__inline HRESULT GetXmaDecodePositionForSample
+(
+    __in_bcount(nXmaDataBytes) const BYTE* pXmaData,  // Pointer to XMA block[s]
+    DWORD nXmaDataBytes,                              // Size of pXmaData in bytes
+    DWORD nStreamIndex,                               // Stream within which to seek
+    DWORD nDesiredSample,                             // Sample sought
+    __out DWORD* pnBitOffset,                         // Returns the bit offset within pXmaData of
+                                                      // the frame containing the sample sought
+    __out DWORD* pnSubFrame                           // Returns the subframe containing the sample
+)
+{
+    DWORD nDesiredFrame = nDesiredSample / XMA_SAMPLES_PER_FRAME;
+    DWORD nSubFrame = (nDesiredSample % XMA_SAMPLES_PER_FRAME) / XMA_SAMPLES_PER_SUBFRAME;
+    DWORD nBitOffset = GetXmaFrameBitPosition(pXmaData, nXmaDataBytes, nStreamIndex, nDesiredFrame);
+
+    XMA2DEFS_ASSERT(pnBitOffset);
+    XMA2DEFS_ASSERT(pnSubFrame);
+
+    if (nBitOffset)
+    {
+        *pnBitOffset = nBitOffset;
+        *pnSubFrame = nSubFrame;
+        return S_OK;
+    }
+    else
+    {
+        return E_FAIL;
+    }
+}
+
+
+// GetXmaSampleRate: Obtains the legal XMA sample rate (24, 32, 44.1 or 48Khz)
+// corresponding to a generic sample rate.
+
+__inline DWORD GetXmaSampleRate(DWORD dwGeneralRate)
+{
+    DWORD dwXmaRate = 48000; // Default XMA rate for all rates above 44100Hz
+
+    if (dwGeneralRate <= 24000)      dwXmaRate = 24000;
+    else if (dwGeneralRate <= 32000) dwXmaRate = 32000;
+    else if (dwGeneralRate <= 44100) dwXmaRate = 44100;
+
+    return dwXmaRate;
+}
+
+
+// Functions to convert between WAVEFORMATEXTENSIBLE channel masks (combinations
+// of the SPEAKER_xxx flags defined in audiodefs.h) and XMA channel masks (which
+// are limited to eight possible speaker positions: left, right, center, low
+// frequency, side left, side right, back left and back right).
+
+__inline DWORD GetStandardChannelMaskFromXmaMask(BYTE bXmaMask)
+{
+    DWORD dwStandardMask = 0;
+
+    if (bXmaMask & XMA_SPEAKER_LEFT)           dwStandardMask |= SPEAKER_FRONT_LEFT;
+    if (bXmaMask & XMA_SPEAKER_RIGHT)          dwStandardMask |= SPEAKER_FRONT_RIGHT;
+    if (bXmaMask & XMA_SPEAKER_CENTER)         dwStandardMask |= SPEAKER_FRONT_CENTER;
+    if (bXmaMask & XMA_SPEAKER_LFE)            dwStandardMask |= SPEAKER_LOW_FREQUENCY;
+    if (bXmaMask & XMA_SPEAKER_LEFT_SURROUND)  dwStandardMask |= SPEAKER_SIDE_LEFT;
+    if (bXmaMask & XMA_SPEAKER_RIGHT_SURROUND) dwStandardMask |= SPEAKER_SIDE_RIGHT;
+    if (bXmaMask & XMA_SPEAKER_LEFT_BACK)      dwStandardMask |= SPEAKER_BACK_LEFT;
+    if (bXmaMask & XMA_SPEAKER_RIGHT_BACK)     dwStandardMask |= SPEAKER_BACK_RIGHT;
+
+    return dwStandardMask;
+}
+
+__inline BYTE GetXmaChannelMaskFromStandardMask(DWORD dwStandardMask)
+{
+    BYTE bXmaMask = 0;
+
+    if (dwStandardMask & SPEAKER_FRONT_LEFT)    bXmaMask |= XMA_SPEAKER_LEFT;
+    if (dwStandardMask & SPEAKER_FRONT_RIGHT)   bXmaMask |= XMA_SPEAKER_RIGHT;
+    if (dwStandardMask & SPEAKER_FRONT_CENTER)  bXmaMask |= XMA_SPEAKER_CENTER;
+    if (dwStandardMask & SPEAKER_LOW_FREQUENCY) bXmaMask |= XMA_SPEAKER_LFE;
+    if (dwStandardMask & SPEAKER_SIDE_LEFT)     bXmaMask |= XMA_SPEAKER_LEFT_SURROUND;
+    if (dwStandardMask & SPEAKER_SIDE_RIGHT)    bXmaMask |= XMA_SPEAKER_RIGHT_SURROUND;
+    if (dwStandardMask & SPEAKER_BACK_LEFT)     bXmaMask |= XMA_SPEAKER_LEFT_BACK;
+    if (dwStandardMask & SPEAKER_BACK_RIGHT)    bXmaMask |= XMA_SPEAKER_RIGHT_BACK;
+
+    return bXmaMask;
+}
+
+
+// LocalizeXma2Format: Modifies a XMA2WAVEFORMATEX structure in place to comply
+// with the current platform's byte-ordering rules (little- or big-endian).
+
+__inline HRESULT LocalizeXma2Format(__inout XMA2WAVEFORMATEX* pXma2Format)
+{
+    #define XMASWAP2BYTES(n) ((WORD)(((n) >> 8) | (((n) & 0xff) << 8)))
+    #define XMASWAP4BYTES(n) ((DWORD)((n) >> 24 | (n) << 24 | ((n) & 0xff00) << 8 | ((n) & 0xff0000) >> 8))
+
+    if (pXma2Format->wfx.wFormatTag == WAVE_FORMAT_XMA2)
+    {
+        return S_OK;
+    }
+    else if (XMASWAP2BYTES(pXma2Format->wfx.wFormatTag) == WAVE_FORMAT_XMA2)
+    {
+        pXma2Format->wfx.wFormatTag      = XMASWAP2BYTES(pXma2Format->wfx.wFormatTag);
+        pXma2Format->wfx.nChannels       = XMASWAP2BYTES(pXma2Format->wfx.nChannels);
+        pXma2Format->wfx.nSamplesPerSec  = XMASWAP4BYTES(pXma2Format->wfx.nSamplesPerSec);
+        pXma2Format->wfx.nAvgBytesPerSec = XMASWAP4BYTES(pXma2Format->wfx.nAvgBytesPerSec);
+        pXma2Format->wfx.nBlockAlign     = XMASWAP2BYTES(pXma2Format->wfx.nBlockAlign);
+        pXma2Format->wfx.wBitsPerSample  = XMASWAP2BYTES(pXma2Format->wfx.wBitsPerSample);
+        pXma2Format->wfx.cbSize          = XMASWAP2BYTES(pXma2Format->wfx.cbSize);
+        pXma2Format->NumStreams          = XMASWAP2BYTES(pXma2Format->NumStreams);
+        pXma2Format->ChannelMask         = XMASWAP4BYTES(pXma2Format->ChannelMask);
+        pXma2Format->SamplesEncoded      = XMASWAP4BYTES(pXma2Format->SamplesEncoded);
+        pXma2Format->BytesPerBlock       = XMASWAP4BYTES(pXma2Format->BytesPerBlock);
+        pXma2Format->PlayBegin           = XMASWAP4BYTES(pXma2Format->PlayBegin);
+        pXma2Format->PlayLength          = XMASWAP4BYTES(pXma2Format->PlayLength);
+        pXma2Format->LoopBegin           = XMASWAP4BYTES(pXma2Format->LoopBegin);
+        pXma2Format->LoopLength          = XMASWAP4BYTES(pXma2Format->LoopLength);
+        pXma2Format->BlockCount          = XMASWAP2BYTES(pXma2Format->BlockCount);
+        return S_OK;
+    }
+    else
+    {
+        return E_FAIL; // Not a recognizable XMA2 format
+    }
+
+    #undef XMASWAP2BYTES
+    #undef XMASWAP4BYTES
+}
+
+
+#endif // #ifndef __XMA2DEFS_INCLUDED__

CMakeLists.txt

@@ -11,13 +11,36 @@ if(NOT CMAKE_SIZEOF_VOID_P EQUAL 8)
 	message( FATAL_ERROR "RPCS3 can only be compiled on 64-bit platforms." )
 endif()
 
+if (CMAKE_COMPILER_IS_GNUCC AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS 5.1)
+	message( FATAL_ERROR "RPCS3 requires at least gcc-5.1." )
+endif()
+
+if (CMAKE_COMPILER_IS_GNUCC AND CMAKE_CXX_COMPILER_VERSION VERSION_EQUAL 6.1)
+	message( FATAL_ERROR "RPCS3 can't be compiled with gcc-6.1, see #1691." )
+endif()
+
+find_program(CCACHE_FOUND ccache)
+if (CCACHE_FOUND)
+	set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ccache)
+	set_property(GLOBAL PROPERTY RULE_LAUNCH_LINK ccache)
+endif()
+
+add_definitions(-DCMAKE_BUILD)
+
 # We use libpng's static library and don't need to build the shared library and run the tests
 set(PNG_SHARED OFF CACHE BOOL "Build shared lib." FORCE)
 set(PNG_TESTS  OFF CACHE BOOL "Build tests." FORCE)
 
-add_subdirectory( asmjit )
-add_subdirectory( libpng )
+# Select the version of libpng to use, default is builtin
+if (NOT USE_SYSTEM_LIBPNG)
+	add_subdirectory( 3rdparty/libpng )
+endif()
+
+option(VULKAN_PREBUILT "" OFF)
+
 # TODO: do real installation, including copying directory structure
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY_RELEASE "${PROJECT_BINARY_DIR}/bin")
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY_DEBUG "${PROJECT_BINARY_DIR}/bin")
+add_subdirectory( Vulkan )
 add_subdirectory( rpcs3 )
+add_subdirectory( rsx_program_decompiler )

README.md

@@ -4,16 +4,18 @@ RPCS3
 [![Build Status](https://travis-ci.org/RPCS3/rpcs3.svg?branch=master)](https://travis-ci.org/RPCS3/rpcs3)
 [![Build status](https://ci.appveyor.com/api/projects/status/411c4clmiohtx7eo/branch/master?svg=true)](https://ci.appveyor.com/project/rpcs3/rpcs3/branch/master)
 [![Coverity Status](https://img.shields.io/coverity/scan/3960.svg)](https://scan.coverity.com/projects/3960)
-[![Coverage Status](https://coveralls.io/repos/RPCS3/rpcs3/badge.svg)](https://coveralls.io/r/RPCS3/rpcs3)
 
 An open-source PlayStation 3 emulator/debugger written in C++.
 
-You can find some basic information in the [FAQ](https://github.com/RPCS3/rpcs3/wiki/FAQ). For discussion about this emulator and PS3 emulation please visit the [official forums](http://www.emunewz.net/forum/forumdisplay.php?fid=162).
+You can find some basic information in the [FAQ](https://github.com/RPCS3/rpcs3/wiki/FAQ). For discussion about this emulator and PS3 emulation please visit the [official forums](http://www.emunewz.net/forum/forumdisplay.php?fid=172).
 
+Official Discord server: https://discord.me/RPCS3
+
+Support Nekotekina on Patreon: https://www.patreon.com/Nekotekina
 
 ### Development
 
-If you want to contribute please take a took at the [Coding Style](https://github.com/RPCS3/rpcs3/wiki/Coding-Style), [Roadmap](https://github.com/RPCS3/rpcs3/wiki/Roadmap) and [Developer Information](https://github.com/RPCS3/rpcs3/wiki/Developer-Information) pages. You should as well contact any of the developers in the forum in order to know about the current situation of the emulator.
+If you want to contribute please take a look at the [Coding Style](https://github.com/RPCS3/rpcs3/wiki/Coding-Style), [Roadmap](https://github.com/RPCS3/rpcs3/wiki/Roadmap) and [Developer Information](https://github.com/RPCS3/rpcs3/wiki/Developer-Information) pages. You should as well contact any of the developers in the forum in order to know about the current situation of the emulator.
 
 
 ### Dependencies
@@ -21,30 +23,47 @@ If you want to contribute please take a took at the [Coding Style](https://githu
 __Windows__
 * [Visual Studio 2015](https://www.visualstudio.com/en-us/downloads/download-visual-studio-vs.aspx)
 * [Visual C++ Redistributable Packages for Visual Studio 2015](http://www.microsoft.com/en-us/download/details.aspx?id=48145)
-* [Cmake 3.1.0+](http://www.cmake.org/download/) (optional, required only for LLVM build; add to PATH)
-* [Python](https://www.python.org/downloads/) (optional, required only for LLVM build; add to PATH)
+* [Cmake 3.1.0+](http://www.cmake.org/download/) (required; add to PATH)
+* [Python 3.3+](https://www.python.org/downloads/) (required; add to PATH)
 
 __Linux__
-* GCC 5.1+ or Clang 3.5.0+
-* Debian & Ubuntu: `sudo apt-get install libopenal-dev libwxgtk3.0-dev build-essential libglew-dev`
+* GCC 5.1+ or Clang 3.5.0+ ([not GCC 6.1](https://github.com/RPCS3/rpcs3/issues/1691))
+* Debian & Ubuntu: `sudo apt-get install cmake build-essential libopenal-dev libwxgtk3.0-dev libglew-dev zlib1g-dev libedit-dev libvulkan-dev git`
 * Arch: `sudo pacman -S glew openal wxgtk cmake llvm`
 
 __Mac OSX__
+Mac OSX is not supported at this moment because it doesn't meet system requirements (OpenGL 4.3)
 * Xcode 6+ (tested with Xcode 6.4)
-* Install with Homebrew: `brew install glew wxwidgets` (add `llvm36` to that list if you want to build with ppu jit)
+* Install with Homebrew: `brew install glew wxwidgets`
 * Remove '-framework QuickTime' from '_ldflags' in /usr/local/bin/wx-config
 
-
 ### Building
 
-To initialize the repository don't forget to execute `git submodule update --init` to pull the wxWidgets source.
-* __Windows__:
-Open the *.SLN* file, and press *Build* > *Clean Solution*, then *Build Solution*. *Rebuild* may not work correctly.
-* __Linux & Mac OSX__:
-If you want to build with LLVM, then LLVM 3.6.2 is required.
-`cd rpcs3 && cmake CMakeLists.txt && make && cd ../` then run with `cd bin && ./rpcs3`.
-If you are on OSX and want to build with llvm don't forget to add `-DLLVM_DIR=/usr/local/opt/llvm36/lib/llvm-3.6/share/llvm/cmake` (or wherever llvm brew was installed) to cmake invocation.
+- __Windows__: </br>
+1) To initialize the repository don't forget to execute `git submodule update --init` to pull the submodules. </br>
+2) Open the *.SLN* file. </br>
+3) Build the projects in *__BUILD_BEFORE* folder: right-click on every project > *Build*. </br>
+4) Press *BUILD* > *Build Solution* or *Rebuild Solution*. </br>
+
+- __Linux & Mac OSX__: </br>
+1) `git clone https://github.com/RPCS3/rpcs3.git` </br>
+2) `cd rpcs3/` </br>
+3) `git submodule update --init` </br>
+4) `cmake CMakeLists.txt && make GitVersion && make` </br>
+5) Run RPCS3 with `./bin/rpcs3` </br>
+
+If you are on OSX and want to build with llvm don't forget to add `-DLLVM_DIR=...` (or wherever llvm brew was installed) to cmake invocation.
+
 When using GDB, configure it to ignore SIGSEGV signal (`handle SIGSEGV nostop noprint`).
 
-### Support
-* [Donate by PayPal](https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick&hosted_button_id=MPJ3S9XQXCE3G)
+##### CMake Build Options (Linux & Mac OSX)
+
+- ```-DUSE_SYSTEM_LIBPNG=ON/OFF``` (default = *OFF*) </br>
+Build against the shared libpng instead of using the builtin one. libpng 1.6+ highly recommended. Try this option if you get version conflict errors or only see black game icons.
+
+- ```-DUSE_SYSTEM_FFMPEG=ON/OFF``` (default = *OFF*) </br>
+Build against the shared ffmpeg libraries instead of using the builtin patched version. Try this if the builtin version breaks the OpenGL renderer for you.
+
+### License
+
+Most files are licensed under the terms of GNU GPLv2 License, see LICENSE file for details. Some files may be licensed differently, check appropriate file headers for details.

Utilities/AtomicPtr.h

@@ -0,0 +1,150 @@
+#pragma once
+
+#include "Atomic.h"
+#include <memory>
+
+// Unfinished. Only std::default_delete will work as expected.
+template<typename T, typename D>
+class atomic_ptr_base : D
+{
+protected:
+	atomic_t<T*> m_ptr;
+
+	constexpr atomic_ptr_base(T* ptr)
+		: m_ptr(ptr)
+	{
+	}
+
+public:
+	~atomic_ptr_base()
+	{
+		if (m_ptr)
+		{
+			(*this)(m_ptr.load());
+		}
+	}
+
+	D& get_deleter()
+	{
+		return *this;
+	}
+
+	const D& get_deleter() const
+	{
+		return *this;
+	}
+};
+
+// Simple atomic pointer with unique ownership. Draft, unfinished.
+template<typename T, typename D = std::default_delete<T>>
+class atomic_ptr final : atomic_ptr_base<T, D>
+{
+	using base = atomic_ptr_base<T, D>;
+
+	static_assert(sizeof(T*) == sizeof(base), "atomic_ptr<> error: invalid deleter (empty class expected)");
+
+public:
+	constexpr atomic_ptr()
+		: base(nullptr)
+	{
+	}
+
+	constexpr atomic_ptr(std::nullptr_t)
+		: base(nullptr)
+	{
+	}
+
+	explicit atomic_ptr(T* ptr)
+		: base(ptr)
+	{
+	}
+
+	template<typename T2, typename = std::enable_if_t<std::is_convertible<T2, T>::value>>
+	atomic_ptr(std::unique_ptr<T2, D>&& ptr)
+		: base(ptr.release())
+	{
+	}
+
+	atomic_ptr& operator =(std::nullptr_t)
+	{
+		if (T* old = base::m_ptr.exchange(nullptr))
+		{
+			this->get_deleter()(old);
+		}
+
+		return *this;
+	}
+
+	template<typename T2, typename = std::enable_if_t<std::is_convertible<T2, T>::value>>
+	atomic_ptr& operator =(std::unique_ptr<T2, D>&& ptr)
+	{
+		if (T* old = base::m_ptr.exchange(ptr.release()))
+		{
+			this->get_deleter()(old);
+		}
+
+		return *this;
+	}
+
+	void swap(std::unique_ptr<T, D>& ptr)
+	{
+		ptr.reset(base::m_ptr.exchange(ptr.release()));
+	}
+
+	std::add_lvalue_reference_t<T> operator *() const
+	{
+		return *base::m_ptr;
+	}
+
+	T* operator ->() const
+	{
+		return base::m_ptr;
+	}
+
+	T* get() const
+	{
+		return base::m_ptr;
+	}
+
+	explicit operator bool() const
+	{
+		return base::m_ptr != nullptr;
+	}
+
+	T* release() const
+	{
+		return base::m_ptr.exchange(0);
+	}
+
+	void reset(T* ptr = nullptr)
+	{
+		if (T* old = base::m_ptr.exchange(ptr))
+		{
+			this->get_deleter()(old);
+		}
+	}
+
+	// Steal the pointer from `ptr`, convert old value to unique_ptr
+	std::unique_ptr<T, D> exchange(std::unique_ptr<T, D>&& ptr)
+	{
+		return std::unique_ptr<T, D>(base::m_ptr.exchange(ptr.release()));
+	}
+
+	// If pointer is null, steal it from `ptr`
+	bool test_and_swap(std::unique_ptr<T, D>&& ptr)
+	{
+		if (base::m_ptr.compare_and_swap_test(nullptr, ptr.get()))
+		{
+			ptr.release();
+			return true;
+		}
+
+		return false;
+	}
+};
+
+template<typename T, typename D>
+class atomic_ptr<T[], D> final : atomic_ptr_base<T[], D>
+{
+	// TODO
+};

Utilities/AutoPause.cpp

@@ -1,57 +1,29 @@
 #include "stdafx.h"
-#include "AutoPause.h"
-#include "Utilities/Log.h"
-#include "Utilities/File.h"
+#include "Config.h"
 #include "Emu/System.h"
-#include "Emu/state.h"
+#include "AutoPause.h"
 
-using namespace Debug;
+cfg::bool_entry g_cfg_debug_autopause_syscall(cfg::root.misc, "Auto Pause at System Call");
+cfg::bool_entry g_cfg_debug_autopause_func_call(cfg::root.misc, "Auto Pause at Function Call");
 
-std::unique_ptr<AutoPause> g_autopause;
-
-AutoPause& AutoPause::getInstance(void)
+debug::autopause& debug::autopause::get_instance()
 {
-	if (!g_autopause)
+	// Use magic static
+	static autopause instance;
+	return instance;
+}
+
+// Load Auto Pause Configuration from file "pause.bin"
+void debug::autopause::reload(void)
+{
+	auto& instance = get_instance();
+
+	instance.m_pause_function.clear();
+	instance.m_pause_syscall.clear();
+
+	// TODO: better format, possibly a config entry
+	if (fs::file list{ fs::get_config_dir() + "pause.bin" })
 	{
-		g_autopause.reset(new AutoPause);
-	}
-
-	return *g_autopause;
-}
-
-//Still use binary format. Default Setting should be "disable all auto pause".
-AutoPause::AutoPause(void)
-{
-	m_pause_function.reserve(16);
-	m_pause_syscall.reserve(16);
-	initialized = false;
-	//Reload(false, false);
-	Reload();
-}
-
-//Notice: I would not allow to write the binary to file in this command.
-AutoPause::~AutoPause(void)
-{
-	initialized = false;
-	m_pause_function.clear();
-	m_pause_syscall.clear();
-	m_pause_function_enable = false;
-	m_pause_syscall_enable = false;
-}
-
-//Load Auto Pause Configuration from file "pause.bin"
-//This would be able to create in a GUI window.
-void AutoPause::Reload(void)
-{
-	if (fs::is_file(fs::get_config_dir() + "pause.bin"))
-	{
-		m_pause_function.clear();
-		m_pause_function.reserve(16);
-		m_pause_syscall.clear();
-		m_pause_syscall.reserve(16);
-
-		fs::file list(fs::get_config_dir() + "pause.bin");
-		//System calls ID and Function calls ID are all u32 iirc.
 		u32 num;
 		size_t fmax = list.size();
 		size_t fcur = 0;
@@ -64,60 +36,38 @@ void AutoPause::Reload(void)
 			
 			if (num < 1024)
 			{
-				//Less than 1024 - be regarded as a system call.
-				//emplace_back may not cause reductant move/copy operation.
-				m_pause_syscall.emplace_back(num);
-				LOG_WARNING(HLE, "Auto Pause: Find System Call ID 0x%x", num);
+				instance.m_pause_syscall.emplace(num);
+				LOG_WARNING(HLE, "Set autopause at syscall %lld", num);
 			}
 			else
 			{
-				m_pause_function.emplace_back(num);
-				LOG_WARNING(HLE, "Auto Pause: Find Function Call ID 0x%x", num);
+				instance.m_pause_function.emplace(num);
+				LOG_WARNING(HLE, "Set autopause at function 0x%08x", num);
 			}
 		}
 	}
-
-	m_pause_syscall_enable = rpcs3::config.misc.debug.auto_pause_syscall.value();
-	m_pause_function_enable = rpcs3::config.misc.debug.auto_pause_func_call.value();
-	initialized = true;
 }
 
-void AutoPause::TryPause(u32 code)
+bool debug::autopause::pause_syscall(u64 code)
 {
-	if (code < 1024)
+	if (g_cfg_debug_autopause_syscall && get_instance().m_pause_syscall.count(code) != 0)
 	{
-		//Would first check Enable setting. Then the list length.
-		if ((!m_pause_syscall_enable)
-			|| (m_pause_syscall.size() <= 0))
-		{
-			return;
-		}
-
-		for (u32 i = 0; i < m_pause_syscall.size(); ++i)
-		{
-			if (code == m_pause_syscall[i])
-			{
-				Emu.Pause();
-				LOG_ERROR(HLE, "Auto Pause Triggered: System call 0x%x", code); // Used Error
-			}
-		}
+		Emu.Pause();
+		LOG_SUCCESS(HLE, "Autopause triggered at syscall %lld", code);
+		return true;
 	}
-	else
-	{
-		//Well similiar.. Seperate the list caused by possible setting difference.
-		if ((!m_pause_function_enable)
-			|| (m_pause_function.size() <= 0))
-		{
-			return;
-		}
 
-		for (u32 i = 0; i < m_pause_function.size(); ++i)
-		{
-			if (code == m_pause_function[i])
-			{
-				Emu.Pause();
-				LOG_ERROR(HLE, "Auto Pause Triggered: Function call 0x%x", code); // Used Error
-			}
-		}
-	}
+	return false;
+}
+
+bool debug::autopause::pause_function(u32 code)
+{
+	if (g_cfg_debug_autopause_func_call && get_instance().m_pause_function.count(code) != 0)
+	{
+		Emu.Pause();
+		LOG_SUCCESS(HLE, "Autopause triggered at function 0x%08x", code);
+		return true;
+	}
+
+	return false;
 }

Utilities/AutoPause.h

@@ -1,24 +1,22 @@
 #pragma once
 
-//Regarded as a Debugger Enchantment
-namespace Debug {
-	//To store the pause function/call id, and let those pause there.
-	//Would be with a GUI to configure those.
-	struct AutoPause
+#include <unordered_set>
+
+// Regarded as a Debugger Enchantment
+namespace debug
+{
+	// To store the pause function/call id, and let those pause there.
+	// Would be with a GUI to configure those.
+	class autopause
 	{
-		std::vector<u32> m_pause_syscall;
-		std::vector<u32> m_pause_function;
-		bool initialized;
-		bool m_pause_syscall_enable;
-		bool m_pause_function_enable;
+		std::unordered_set<u64> m_pause_syscall;
+		std::unordered_set<u32> m_pause_function;
 
-		AutoPause();
-		~AutoPause();
+		static autopause& get_instance();
 	public:
-		static AutoPause& getInstance(void);
 
-		void Reload(void);
-
-		void TryPause(u32 code);
+		static void reload();
+		static bool pause_syscall(u64 code);
+		static bool pause_function(u32 code);
 	};
-}
+}

Utilities/BEType.h

@@ -1,71 +1,58 @@
 #pragma once
 
-#ifdef _MSC_VER
-#include <intrin.h>
-#else
-#include <x86intrin.h>
-#endif
+#include "types.h"
 
-#define IS_LE_MACHINE // only draft
-
-union v128
+// 128-bit vector type and also se_storage<> storage type
+union alignas(16) v128
 {
-	template<typename T, std::size_t N, std::size_t M> class masked_array_t // array type accessed as (index ^ M)
+	char _bytes[16];
+
+	template <typename T, std::size_t N, std::size_t M>
+	struct masked_array_t // array type accessed as (index ^ M)
 	{
 		T m_data[N];
 
 	public:
-		T& operator [](std::size_t index)
+		T& operator[](std::size_t index)
 		{
 			return m_data[index ^ M];
 		}
 
-		const T& operator [](std::size_t index) const
+		const T& operator[](std::size_t index) const
 		{
 			return m_data[index ^ M];
 		}
-
-		T& at(std::size_t index)
-		{
-			return (index ^ M) < N ? m_data[index ^ M] : throw std::out_of_range(__FUNCTION__);
-		}
-
-		const T& at(std::size_t index) const
-		{
-			return (index ^ M) < N ? m_data[index ^ M] : throw std::out_of_range(__FUNCTION__);
-		}
 	};
 
-#ifdef IS_LE_MACHINE
-	template<typename T, std::size_t N = 16 / sizeof(T)> using normal_array_t = masked_array_t<T, N, 0>;
-	template<typename T, std::size_t N = 16 / sizeof(T)> using reversed_array_t = masked_array_t<T, N, N - 1>;
-#else
-	template<typename T, std::size_t N = 16 / sizeof(T)> using normal_array_t = masked_array_t<T, N, N - 1>;
-	template<typename T, std::size_t N = 16 / sizeof(T)> using reversed_array_t = masked_array_t<T, N, 0>;
+#if IS_LE_MACHINE == 1
+	template <typename T, std::size_t N = 16 / sizeof(T)>
+	using normal_array_t = masked_array_t<T, N, 0>;
+	template <typename T, std::size_t N = 16 / sizeof(T)>
+	using reversed_array_t = masked_array_t<T, N, N - 1>;
 #endif
 
-	normal_array_t<u64>   _u64;
-	normal_array_t<s64>   _s64;
+	normal_array_t<u64> _u64;
+	normal_array_t<s64> _s64;
 	reversed_array_t<u64> u64r;
 	reversed_array_t<s64> s64r;
 
-	normal_array_t<u32>   _u32;
-	normal_array_t<s32>   _s32;
+	normal_array_t<u32> _u32;
+	normal_array_t<s32> _s32;
 	reversed_array_t<u32> u32r;
 	reversed_array_t<s32> s32r;
 
-	normal_array_t<u16>   _u16;
-	normal_array_t<s16>   _s16;
+	normal_array_t<u16> _u16;
+	normal_array_t<s16> _s16;
 	reversed_array_t<u16> u16r;
 	reversed_array_t<s16> s16r;
 
-	normal_array_t<u8>    _u8;
-	normal_array_t<s8>    _s8;
-	reversed_array_t<u8>  u8r;
-	reversed_array_t<s8>  s8r;
+	normal_array_t<u8> _u8;
+	normal_array_t<s8> _s8;
+	reversed_array_t<u8> u8r;
+	reversed_array_t<s8> s8r;
 
-	normal_array_t<f32>   _f;
-	normal_array_t<f64>   _d;
+	normal_array_t<f32> _f;
+	normal_array_t<f64> _d;
 	reversed_array_t<f32> fr;
 	reversed_array_t<f64> dr;
 
@@ -73,7 +60,7 @@ union v128
 	__m128i vi;
 	__m128d vd;
 
-	class bit_array_128
+	struct bit_array_128
 	{
 		u64 m_data[2];
 
@@ -95,7 +82,7 @@ union v128
 				return (data & mask) != 0;
 			}
 
-			bit_element& operator =(const bool right)
+			bit_element& operator=(const bool right)
 			{
 				if (right)
 				{
@@ -108,7 +95,7 @@ union v128
 				return *this;
 			}
 
-			bit_element& operator =(const bit_element& right)
+			bit_element& operator=(const bit_element& right)
 			{
 				if (right)
 				{
@@ -123,40 +110,21 @@ union v128
 		};
 
 		// Index 0 returns the MSB and index 127 returns the LSB
-		bit_element operator [](u32 index)
+		bit_element operator[](u32 index)
 		{
-#ifdef IS_LE_MACHINE
+#if IS_LE_MACHINE == 1
 			return bit_element(m_data[1 - (index >> 6)], 0x8000000000000000ull >> (index & 0x3F));
-#else
-			return bit_element(m_data[index >> 6], 0x8000000000000000ull >> (index & 0x3F));
 #endif
 		}
 
 		// Index 0 returns the MSB and index 127 returns the LSB
-		bool operator [](u32 index) const
+		bool operator[](u32 index) const
 		{
-#ifdef IS_LE_MACHINE
+#if IS_LE_MACHINE == 1
 			return (m_data[1 - (index >> 6)] & (0x8000000000000000ull >> (index & 0x3F))) != 0;
-#else
-			return (m_data[index >> 6] & (0x8000000000000000ull >> (index & 0x3F))) != 0;
 #endif
 		}
-
-		bit_element at(u32 index)
-		{
-			if (index >= 128) throw std::out_of_range(__FUNCTION__);
-
-			return operator[](index);
-		}
-
-		bool at(u32 index) const
-		{
-			if (index >= 128) throw std::out_of_range(__FUNCTION__);
-
-			return operator[](index);
-		}
-	}
-	_bit;
+	} _bit;
 
 	static v128 from64(u64 _0, u64 _1 = 0)
 	{
@@ -310,26 +278,16 @@ union v128
 		return fromV(_mm_cmpeq_epi32(left.vi, right.vi));
 	}
 
-	bool operator ==(const v128& right) const
+	bool operator==(const v128& right) const
 	{
 		return _u64[0] == right._u64[0] && _u64[1] == right._u64[1];
 	}
 
-	bool operator !=(const v128& right) const
+	bool operator!=(const v128& right) const
 	{
 		return _u64[0] != right._u64[0] || _u64[1] != right._u64[1];
 	}
 
-	bool is_any_1() const // check if any bit is 1
-	{
-		return _u64[0] || _u64[1];
-	}
-
-	bool is_any_0() const // check if any bit is 0
-	{
-		return ~_u64[0] || ~_u64[1];
-	}
-
 	// result = (~left) & (right)
 	static inline v128 andnot(const v128& left, const v128& right)
 	{
@@ -341,54 +299,73 @@ union v128
 		_u64[0] = 0;
 		_u64[1] = 0;
 	}
-
-	std::string to_hex() const;
-
-	std::string to_xyzw() const;
-
-	static inline v128 byteswap(const v128 val)
-	{
-		return fromV(_mm_shuffle_epi8(val.vi, _mm_set_epi8(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)));
-	}
 };
 
-CHECK_SIZE_ALIGN(v128, 16, 16);
-
-inline v128 operator |(const v128& left, const v128& right)
+inline v128 operator|(const v128& left, const v128& right)
 {
 	return v128::fromV(_mm_or_si128(left.vi, right.vi));
 }
 
-inline v128 operator &(const v128& left, const v128& right)
+inline v128 operator&(const v128& left, const v128& right)
 {
 	return v128::fromV(_mm_and_si128(left.vi, right.vi));
 }
 
-inline v128 operator ^(const v128& left, const v128& right)
+inline v128 operator^(const v128& left, const v128& right)
 {
 	return v128::fromV(_mm_xor_si128(left.vi, right.vi));
 }
 
-inline v128 operator ~(const v128& other)
+inline v128 operator~(const v128& other)
 {
 	return v128::from64(~other._u64[0], ~other._u64[1]);
 }
 
-template<typename T, std::size_t Size = sizeof(T)> struct se_storage
+template <typename T, std::size_t Align, std::size_t Size>
+struct se_storage
 {
-	static_assert(!Size, "Bad se_storage<> type");
+	using type = std::aligned_storage_t<Size, Align>;
+
+	// Unoptimized generic byteswap for unaligned data
+	static void reverse(u8* dst, const u8* src)
+	{
+		for (std::size_t i = 0; i < Size; i++)
+		{
+			dst[i] = src[Size - 1 - i];
+		}
+	}
+
+	static type to(const T& src)
+	{
+		type result;
+		reverse(reinterpret_cast<u8*>(&result), reinterpret_cast<const u8*>(&src));
+		return result;
+	}
+
+	static T from(const type& src)
+	{
+		T result;
+		reverse(reinterpret_cast<u8*>(&result), reinterpret_cast<const u8*>(&src));
+		return result;
+	}
+
+	static type copy(const type& src)
+	{
+		type result;
+		for (std::size_t i = 0; i < Size; i++)
+		{
+			reinterpret_cast<u8*>(&result)[i] = reinterpret_cast<const u8*>(&src)[i];
+		}
+		return result;
+	}
 };
 
-template<typename T> struct se_storage<T, 2>
+template <typename T>
+struct se_storage<T, 2, 2>
 {
 	using type = u16;
 
-	[[deprecated]] static constexpr u16 _swap(u16 src) // for reference
-	{
-		return (src >> 8) | (src << 8);
-	}
-
-	static inline u16 swap(u16 src)
+	static constexpr u16 swap(u16 src)
 	{
 #if defined(__GNUG__)
 		return __builtin_bswap16(src);
@@ -407,18 +384,19 @@ template<typename T> struct se_storage<T, 2>
 		const u16 result = swap(src);
 		return reinterpret_cast<const T&>(result);
 	}
+
+	static inline T copy(const T& src)
+	{
+		return src;
+	}
 };
 
-template<typename T> struct se_storage<T, 4>
+template <typename T>
+struct se_storage<T, 4, 4>
 {
 	using type = u32;
 
-	[[deprecated]] static constexpr u32 _swap(u32 src) // for reference
-	{
-		return (src >> 24) | (src << 24) | ((src >> 8) & 0x0000ff00) | ((src << 8) & 0x00ff0000);
-	}
-
-	static inline u32 swap(u32 src)
+	static constexpr u32 swap(u32 src)
 	{
 #if defined(__GNUG__)
 		return __builtin_bswap32(src);
@@ -437,24 +415,19 @@ template<typename T> struct se_storage<T, 4>
 		const u32 result = swap(src);
 		return reinterpret_cast<const T&>(result);
 	}
+
+	static inline T copy(const T& src)
+	{
+		return src;
+	}
 };
 
-template<typename T> struct se_storage<T, 8>
+template <typename T>
+struct se_storage<T, 8, 8>
 {
 	using type = u64;
 
-	[[deprecated]] static constexpr u64 _swap(u64 src) // for reference
-	{
-		return (src >> 56) | (src << 56) |
-			((src >> 40) & 0x000000000000ff00) |
-			((src >> 24) & 0x0000000000ff0000) |
-			((src >> 8)  & 0x00000000ff000000) |
-			((src << 8)  & 0x000000ff00000000) |
-			((src << 24) & 0x0000ff0000000000) |
-			((src << 40) & 0x00ff000000000000);
-	}
-
-	static inline u64 swap(u64 src)
+	static constexpr u64 swap(u64 src)
 	{
 #if defined(__GNUG__)
 		return __builtin_bswap64(src);
@@ -473,81 +446,54 @@ template<typename T> struct se_storage<T, 8>
 		const u64 result = swap(src);
 		return reinterpret_cast<const T&>(result);
 	}
+
+	static inline T copy(const T& src)
+	{
+		return src;
+	}
 };
 
-template<typename T> struct se_storage<T, 16>
+template <typename T>
+struct se_storage<T, 16, 16>
 {
 	using type = v128;
 
+	static inline v128 swap(const v128& src)
+	{
+		return v128::fromV(_mm_shuffle_epi8(src.vi, _mm_set_epi8(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)));
+	}
+
 	static inline v128 to(const T& src)
 	{
-		return v128::byteswap(reinterpret_cast<const v128&>(src));
+		return swap(reinterpret_cast<const v128&>(src));
 	}
 
 	static inline T from(const v128& src)
 	{
-		const v128 result = v128::byteswap(src);
+		const v128 result = swap(src);
 		return reinterpret_cast<const T&>(result);
 	}
-};
 
-template<typename T> using se_storage_t = typename se_storage<T>::type;
-
-template<typename T1, typename T2> struct se_convert
-{
-	using type_from = std::remove_cv_t<T1>;
-	using type_to = std::remove_cv_t<T2>;
-	using stype_from = se_storage_t<std::remove_cv_t<T1>>;
-	using stype_to = se_storage_t<std::remove_cv_t<T2>>;
-	using storage_from = se_storage<std::remove_cv_t<T1>>;
-	using storage_to = se_storage<std::remove_cv_t<T2>>;
-
-	static inline std::enable_if_t<std::is_same<type_from, type_to>::value, stype_to> convert(const stype_from& data)
+	static inline T copy(const T& src)
 	{
-		return data;
-	}
-
-	static inline stype_to convert(const stype_from& data, ...)
-	{
-		return storage_to::to(storage_from::from(data));
+		return src;
 	}
 };
 
-static struct se_raw_tag_t {} constexpr se_raw{};
-
-template<typename T, bool Se = true> class se_t;
-
-// se_t with switched endianness
-template<typename T> class se_t<T, true>
+// Switched endianness
+template <typename T, std::size_t Align>
+class se_t<T, true, Align>
 {
 	using type = typename std::remove_cv<T>::type;
-	using stype = se_storage_t<type>;
-	using storage = se_storage<type>;
+	using stype = typename se_storage<type, Align>::type;
+	using storage = se_storage<type, Align>;
 
 	stype m_data;
 
-	static_assert(!std::is_union<type>::value && !std::is_class<type>::value || std::is_same<type, v128>::value || std::is_same<type, u128>::value, "se_t<> error: invalid type (struct or union)");
 	static_assert(!std::is_pointer<type>::value, "se_t<> error: invalid type (pointer)");
 	static_assert(!std::is_reference<type>::value, "se_t<> error: invalid type (reference)");
 	static_assert(!std::is_array<type>::value, "se_t<> error: invalid type (array)");
-	//static_assert(!std::is_enum<type>::value, "se_t<> error: invalid type (enumeration), use integral type instead");
-	static_assert(alignof(type) == alignof(stype), "se_t<> error: unexpected alignment");
-
-	template<typename T2, typename = void> struct bool_converter
-	{
-		static inline bool to_bool(const se_t<T2>& value)
-		{
-			return static_cast<bool>(value.value());
-		}
-	};
-
-	template<typename T2> struct bool_converter<T2, std::enable_if_t<std::is_integral<T2>::value>>
-	{
-		static inline bool to_bool(const se_t<T2>& value)
-		{
-			return value.m_data != 0;
-		}
-	};
+	static_assert(sizeof(type) == alignof(type), "se_t<> error: unexpected alignment");
 
 public:
 	se_t() = default;
@@ -558,181 +504,146 @@ public:
 		: m_data(storage::to(value))
 	{
 	}
-	
-	// construct directly from raw data (don't use)
-	constexpr se_t(const stype& raw_value, const se_raw_tag_t&)
-		: m_data(raw_value)
-	{
-	}
 
 	type value() const
 	{
 		return storage::from(m_data);
 	}
 
-	// access underlying raw data (don't use)
-	constexpr const stype& raw_data() const noexcept
-	{
-		return m_data;
-	}
-	
-	se_t& operator =(const se_t&) = default;
+	se_t& operator=(const se_t&) = default;
 
-	se_t& operator =(type value)
+	se_t& operator=(type value)
 	{
 		return m_data = storage::to(value), *this;
 	}
 
+	using simple_type = simple_t<T>;
+
 	operator type() const
 	{
 		return storage::from(m_data);
 	}
-
-	// optimization
-	explicit operator bool() const
-	{
-		return bool_converter<type>::to_bool(*this);
-	}
-
-	// optimization
-	template<typename T2> std::enable_if_t<IS_BINARY_COMPARABLE(T, T2), se_t&> operator &=(const se_t<T2>& right)
-	{
-		return m_data &= right.raw_data(), *this;
-	}
-
-	// optimization
-	template<typename CT> std::enable_if_t<IS_INTEGRAL(T) && std::is_convertible<CT, T>::value, se_t&> operator &=(CT right)
-	{
-		return m_data &= storage::to(right), *this;
-	}
-
-	// optimization
-	template<typename T2> std::enable_if_t<IS_BINARY_COMPARABLE(T, T2), se_t&> operator |=(const se_t<T2>& right)
-	{
-		return m_data |= right.raw_data(), *this;
-	}
-
-	// optimization
-	template<typename CT> std::enable_if_t<IS_INTEGRAL(T) && std::is_convertible<CT, T>::value, se_t&> operator |=(CT right)
-	{
-		return m_data |= storage::to(right), *this;
-	}
-
-	// optimization
-	template<typename T2> std::enable_if_t<IS_BINARY_COMPARABLE(T, T2), se_t&> operator ^=(const se_t<T2>& right)
-	{
-		return m_data ^= right.raw_data(), *this;
-	}
-
-	// optimization
-	template<typename CT> std::enable_if_t<IS_INTEGRAL(T) && std::is_convertible<CT, T>::value, se_t&> operator ^=(CT right)
-	{
-		return m_data ^= storage::to(right), *this;
-	}
 };
 
-// se_t with native endianness
-template<typename T> class se_t<T, false>
+// Native endianness
+template <typename T, std::size_t Align>
+class se_t<T, false, Align>
 {
 	using type = typename std::remove_cv<T>::type;
+	using stype = typename se_storage<type, Align>::type;
+	using storage = se_storage<type, Align>;
 
-	type m_data;
-
-	static_assert(!std::is_union<type>::value && !std::is_class<type>::value || std::is_same<type, v128>::value || std::is_same<type, u128>::value, "se_t<> error: invalid type (struct or union)");
 	static_assert(!std::is_pointer<type>::value, "se_t<> error: invalid type (pointer)");
 	static_assert(!std::is_reference<type>::value, "se_t<> error: invalid type (reference)");
 	static_assert(!std::is_array<type>::value, "se_t<> error: invalid type (array)");
-	//static_assert(!std::is_enum<type>::value, "se_t<> error: invalid type (enumeration), use integral type instead");
+	static_assert(sizeof(type) == alignof(type), "se_t<> error: unexpected alignment");
+
+	stype m_data;
 
 public:
 	se_t() = default;
 
-	se_t(const se_t&) = default;
-
-	constexpr se_t(type value)
-		: m_data(value)
+	se_t(type value)
+		: m_data(reinterpret_cast<const stype&>(value))
 	{
 	}
 
 	type value() const
 	{
-		return m_data;
+		return storage::copy(reinterpret_cast<const type&>(m_data));
 	}
 
-	se_t& operator =(const se_t& value) = default;
+	se_t& operator=(const se_t& value) = default;
 
-	se_t& operator =(type value)
+	se_t& operator=(type value)
 	{
-		return m_data = value, *this;
+		return m_data = reinterpret_cast<const stype&>(value), *this;
 	}
 
+	using simple_type = simple_t<T>;
+
 	operator type() const
 	{
-		return m_data;
-	}
-
-	template<typename CT> std::enable_if_t<IS_INTEGRAL(T) && std::is_convertible<CT, T>::value, se_t&> operator &=(const CT& right)
-	{
-		return m_data &= right, *this;
-	}
-
-	template<typename CT> std::enable_if_t<IS_INTEGRAL(T) && std::is_convertible<CT, T>::value, se_t&> operator |=(const CT& right)
-	{
-		return m_data |= right, *this;
-	}
-
-	template<typename CT> std::enable_if_t<IS_INTEGRAL(T) && std::is_convertible<CT, T>::value, se_t&> operator ^=(const CT& right)
-	{
-		return m_data ^= right, *this;
+		return storage::copy(reinterpret_cast<const type&>(m_data));
 	}
 };
 
-// se_t with native endianness (alias)
-template<typename T> using nse_t = se_t<T, false>;
+// se_t<> with native endianness
+template <typename T, std::size_t Align = alignof(T)>
+using nse_t = se_t<T, false, Align>;
 
-template<typename T, bool Se, typename T1> inline se_t<T, Se>& operator +=(se_t<T, Se>& left, const T1& right)
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator+=(se_t<T, Se, Align>& left, const T1& right)
 {
 	auto value = left.value();
 	return left = (value += right);
 }
 
-template<typename T, bool Se, typename T1> inline se_t<T, Se>& operator -=(se_t<T, Se>& left, const T1& right)
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator-=(se_t<T, Se, Align>& left, const T1& right)
 {
 	auto value = left.value();
 	return left = (value -= right);
 }
 
-template<typename T, bool Se, typename T1> inline se_t<T, Se>& operator *=(se_t<T, Se>& left, const T1& right)
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator*=(se_t<T, Se, Align>& left, const T1& right)
 {
 	auto value = left.value();
 	return left = (value *= right);
 }
 
-template<typename T, bool Se, typename T1> inline se_t<T, Se>& operator /=(se_t<T, Se>& left, const T1& right)
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator/=(se_t<T, Se, Align>& left, const T1& right)
 {
 	auto value = left.value();
 	return left = (value /= right);
 }
 
-template<typename T, bool Se, typename T1> inline se_t<T, Se>& operator %=(se_t<T, Se>& left, const T1& right)
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator%=(se_t<T, Se, Align>& left, const T1& right)
 {
 	auto value = left.value();
 	return left = (value %= right);
 }
 
-template<typename T, bool Se, typename T1> inline se_t<T, Se>& operator <<=(se_t<T, Se>& left, const T1& right)
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator&=(se_t<T, Se, Align>& left, const T1& right)
+{
+	auto value = left.value();
+	return left = (value &= right);
+}
+
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator|=(se_t<T, Se, Align>& left, const T1& right)
+{
+	auto value = left.value();
+	return left = (value |= right);
+}
+
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator^=(se_t<T, Se, Align>& left, const T1& right)
+{
+	auto value = left.value();
+	return left = (value ^= right);
+}
+
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator<<=(se_t<T, Se, Align>& left, const T1& right)
 {
 	auto value = left.value();
 	return left = (value <<= right);
 }
 
-template<typename T, bool Se, typename T1> inline se_t<T, Se>& operator >>=(se_t<T, Se>& left, const T1& right)
+template <typename T, bool Se, std::size_t Align, typename T1>
+inline se_t<T, Se, Align>& operator>>=(se_t<T, Se, Align>& left, const T1& right)
 {
 	auto value = left.value();
 	return left = (value >>= right);
 }
 
-template<typename T, bool Se> inline se_t<T, Se> operator ++(se_t<T, Se>& left, int)
+template <typename T, bool Se, std::size_t Align>
+inline se_t<T, Se, Align> operator++(se_t<T, Se, Align>& left, int)
 {
 	auto value = left.value();
 	auto result = value++;
@@ -740,7 +651,8 @@ template<typename T, bool Se> inline se_t<T, Se> operator ++(se_t<T, Se>& left,
 	return result;
 }
 
-template<typename T, bool Se> inline se_t<T, Se> operator --(se_t<T, Se>& left, int)
+template <typename T, bool Se, std::size_t Align>
+inline se_t<T, Se, Align> operator--(se_t<T, Se, Align>& left, int)
 {
 	auto value = left.value();
 	auto result = value--;
@@ -748,193 +660,140 @@ template<typename T, bool Se> inline se_t<T, Se> operator --(se_t<T, Se>& left,
 	return result;
 }
 
-template<typename T, bool Se> inline se_t<T, Se>& operator ++(se_t<T, Se>& right)
+template <typename T, bool Se, std::size_t Align>
+inline se_t<T, Se, Align>& operator++(se_t<T, Se, Align>& right)
 {
 	auto value = right.value();
 	return right = ++value;
 }
 
-template<typename T, bool Se> inline se_t<T, Se>& operator --(se_t<T, Se>& right)
+template <typename T, bool Se, std::size_t Align>
+inline se_t<T, Se, Align>& operator--(se_t<T, Se, Align>& right)
 {
 	auto value = right.value();
 	return right = --value;
 }
 
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_BINARY_COMPARABLE(T1, T2), bool> operator ==(const se_t<T1>& left, const se_t<T2>& right)
-{
-	return left.raw_data() == right.raw_data();
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGRAL(T1) && IS_INTEGER(T2) && sizeof(T1) >= sizeof(T2), bool> operator ==(const se_t<T1>& left, T2 right)
-{
-	return left.raw_data() == se_storage<T1>::to(right);
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGER(T1) && IS_INTEGRAL(T2) && sizeof(T1) <= sizeof(T2), bool> operator ==(T1 left, const se_t<T2>& right)
-{
-	return se_storage<T2>::to(left) == right.raw_data();
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_BINARY_COMPARABLE(T1, T2), bool> operator !=(const se_t<T1>& left, const se_t<T2>& right)
-{
-	return left.raw_data() != right.raw_data();
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGRAL(T1) && IS_INTEGER(T2) && sizeof(T1) >= sizeof(T2), bool> operator !=(const se_t<T1>& left, T2 right)
-{
-	return left.raw_data() != se_storage<T1>::to(right);
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGER(T1) && IS_INTEGRAL(T2) && sizeof(T1) <= sizeof(T2), bool> operator !=(T1 left, const se_t<T2>& right)
-{
-	return se_storage<T2>::to(left) != right.raw_data();
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_BINARY_COMPARABLE(T1, T2) && sizeof(T1) >= 4, se_t<decltype(T1() & T2())>> operator &(const se_t<T1>& left, const se_t<T2>& right)
-{
-	return{ left.raw_data() & right.raw_data(), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGRAL(T1) && IS_INTEGER(T2) && sizeof(T1) >= sizeof(T2) && sizeof(T1) >= 4, se_t<decltype(T1() & T2())>> operator &(const se_t<T1>& left, T2 right)
-{
-	return{ left.raw_data() & se_storage<T1>::to(right), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGER(T1) && IS_INTEGRAL(T2) && sizeof(T1) <= sizeof(T2) && sizeof(T2) >= 4, se_t<decltype(T1() & T2())>> operator &(T1 left, const se_t<T2>& right)
-{
-	return{ se_storage<T2>::to(left) & right.raw_data(), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_BINARY_COMPARABLE(T1, T2) && sizeof(T1) >= 4, se_t<decltype(T1() | T2())>> operator |(const se_t<T1>& left, const se_t<T2>& right)
-{
-	return{ left.raw_data() | right.raw_data(), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGRAL(T1) && IS_INTEGER(T2) && sizeof(T1) >= sizeof(T2) && sizeof(T1) >= 4, se_t<decltype(T1() | T2())>> operator |(const se_t<T1>& left, T2 right)
-{
-	return{ left.raw_data() | se_storage<T1>::to(right), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGER(T1) && IS_INTEGRAL(T2) && sizeof(T1) <= sizeof(T2) && sizeof(T2) >= 4, se_t<decltype(T1() | T2())>> operator |(T1 left, const se_t<T2>& right)
-{
-	return{ se_storage<T2>::to(left) | right.raw_data(), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_BINARY_COMPARABLE(T1, T2) && sizeof(T1) >= 4, se_t<decltype(T1() ^ T2())>> operator ^(const se_t<T1>& left, const se_t<T2>& right)
-{
-	return{ left.raw_data() ^ right.raw_data(), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGRAL(T1) && IS_INTEGER(T2) && sizeof(T1) >= sizeof(T2) && sizeof(T1) >= 4, se_t<decltype(T1() ^ T2())>> operator ^(const se_t<T1>& left, T2 right)
-{
-	return{ left.raw_data() ^ se_storage<T1>::to(right), se_raw };
-}
-
-// optimization
-template<typename T1, typename T2> inline std::enable_if_t<IS_INTEGER(T1) && IS_INTEGRAL(T2) && sizeof(T1) <= sizeof(T2) && sizeof(T2) >= 4, se_t<decltype(T1() ^ T2())>> operator ^(T1 left, const se_t<T2>& right)
-{
-	return{ se_storage<T2>::to(left) ^ right.raw_data(), se_raw };
-}
-
-// optimization
-template<typename T> inline std::enable_if_t<IS_INTEGRAL(T) && sizeof(T) >= 4, se_t<decltype(~T())>> operator ~(const se_t<T>& right)
-{
-	return{ ~right.raw_data(), se_raw };
-}
-
-#ifdef IS_LE_MACHINE
-template<typename T> using be_t = se_t<T, true>;
-template<typename T> using le_t = se_t<T, false>;
-#else
-template<typename T> using be_t = se_t<T, false>;
-template<typename T> using le_t = se_t<T, true>;
+#if IS_LE_MACHINE == 1
+template <typename T, std::size_t Align = alignof(T)>
+using be_t = se_t<T, true, Align>;
+template <typename T, std::size_t Align = alignof(T)>
+using le_t = se_t<T, false, Align>;
 #endif
 
-
-template<typename T, bool Se, typename = void> struct to_se
+// Type converter: converts native endianness arithmetic/enum types to appropriate se_t<> type
+template <typename T, bool Se, typename = void>
+struct to_se
 {
-	using type = typename std::conditional<std::is_arithmetic<T>::value || std::is_enum<T>::value, se_t<T, Se>, T>::type;
+	template <typename T2, typename = void>
+	struct to_se_
+	{
+		using type = T2;
+	};
+
+	template <typename T2>
+	struct to_se_<T2, std::enable_if_t<std::is_arithmetic<T2>::value || std::is_enum<T2>::value>>
+	{
+		using type = se_t<T2, Se>;
+	};
+
+	// Convert arithmetic and enum types
+	using type = typename to_se_<T>::type;
 };
 
-template<typename T, bool Se> struct to_se<const T, Se, std::enable_if_t<!std::is_array<T>::value>> // move const qualifier
+template <bool Se>
+struct to_se<v128, Se>
 {
+	using type = se_t<v128, Se>;
+};
+
+template <bool Se>
+struct to_se<u128, Se>
+{
+	using type = se_t<u128, Se>;
+};
+
+template <bool Se>
+struct to_se<s128, Se>
+{
+	using type = se_t<s128, Se>;
+};
+
+template <bool Se>
+struct to_se<bool, Se>
+{
+	using type = bool;
+};
+
+template <bool Se>
+struct to_se<char, Se>
+{
+	using type = char;
+};
+
+template <bool Se>
+struct to_se<u8, Se>
+{
+	using type = u8;
+};
+
+template <bool Se>
+struct to_se<s8, Se>
+{
+	using type = s8;
+};
+
+template <typename T, bool Se>
+struct to_se<const T, Se, std::enable_if_t<!std::is_array<T>::value>>
+{
+	// Move const qualifier
 	using type = const typename to_se<T, Se>::type;
 };
 
-template<typename T, bool Se> struct to_se<volatile T, Se, std::enable_if_t<!std::is_array<T>::value && !std::is_const<T>::value>> // move volatile qualifier
+template <typename T, bool Se>
+struct to_se<volatile T, Se, std::enable_if_t<!std::is_array<T>::value && !std::is_const<T>::value>>
 {
+	// Move volatile qualifier
 	using type = volatile typename to_se<T, Se>::type;
 };
 
-template<typename T, bool Se> struct to_se<T[], Se>
+template <typename T, bool Se>
+struct to_se<T[], Se>
 {
+	// Move array qualifier
 	using type = typename to_se<T, Se>::type[];
 };
 
-template<typename T, bool Se, std::size_t N> struct to_se<T[N], Se>
+template <typename T, bool Se, std::size_t N>
+struct to_se<T[N], Se>
 {
+	// Move array qualifier
 	using type = typename to_se<T, Se>::type[N];
 };
 
-template<bool Se> struct to_se<u128, Se> { using type = se_t<u128, Se>; };
-template<bool Se> struct to_se<v128, Se> { using type = se_t<v128, Se>; };
-template<bool Se> struct to_se<bool, Se> { using type = bool; };
-template<bool Se> struct to_se<char, Se> { using type = char; };
-template<bool Se> struct to_se<u8, Se>   { using type = u8; };
-template<bool Se> struct to_se<s8, Se>   { using type = s8; };
-
-#ifdef IS_LE_MACHINE
-template<typename T> using to_be_t = typename to_se<T, true>::type;
-template<typename T> using to_le_t = typename to_se<T, false>::type;
-#else
-template<typename T> using to_be_t = typename to_se<T, false>::type;
-template<typename T> using to_le_t = typename to_se<T, true>::type;
+// BE/LE aliases for to_se<>
+#if IS_LE_MACHINE == 1
+template <typename T>
+using to_be_t = typename to_se<T, true>::type;
+template <typename T>
+using to_le_t = typename to_se<T, false>::type;
 #endif
 
+// BE/LE aliases for atomic_t
+#if IS_LE_MACHINE == 1
+template <typename T>
+using atomic_be_t = atomic_t<be_t<T>>;
+template <typename T>
+using atomic_le_t = atomic_t<le_t<T>>;
+#endif
 
-template<typename T, typename = void> struct to_ne
+template <typename T, bool Se, std::size_t Align>
+struct fmt_unveil<se_t<T, Se, Align>, void>
 {
-	using type = T;
-};
+	using type = typename fmt_unveil<T>::type;
 
-template<typename T, bool Se> struct to_ne<se_t<T, Se>>
-{
-	using type = typename std::remove_cv<T>::type;
+	static inline auto get(const se_t<T, Se, Align>& arg)
+	{
+		return fmt_unveil<T>::get(arg);
+	}
 };
-
-template<typename T> struct to_ne<const T, std::enable_if_t<!std::is_array<T>::value>> // move const qualifier
-{
-	using type = const typename to_ne<T>::type;
-};
-
-template<typename T> struct to_ne<volatile T, std::enable_if_t<!std::is_array<T>::value && !std::is_const<T>::value>> // move volatile qualifier
-{
-	using type = volatile typename to_ne<T>::type;
-};
-
-template<typename T> struct to_ne<T[]>
-{
-	using type = typename to_ne<T>::type[];
-};
-
-template<typename T, std::size_t N> struct to_ne<T[N]>
-{
-	using type = typename to_ne<T>::type[N];
-};
-
-// restore native endianness for T: returns T for be_t<T> or le_t<T>, T otherwise
-template<typename T> using to_ne_t = typename to_ne<T>::type;

Utilities/BitField.h

@@ -1,140 +1,280 @@
 #pragma once
 
-// BitField access helper class (N bits from I position), intended to be put in union
-template<typename T, u32 I, u32 N> class bf_t
+#include "types.h"
+
+template<typename T, uint N>
+struct bf_base
 {
-	// Checks
-	static_assert(I < sizeof(T) * 8, "bf_t<> error: I out of bounds");
-	static_assert(N < sizeof(T) * 8, "bf_t<> error: N out of bounds");
-	static_assert(I + N <= sizeof(T) * 8, "bf_t<> error: values out of bounds");
+	using type = T;
+	using vtype = simple_t<type>;
+	using utype = typename std::make_unsigned<vtype>::type;
 
-	// Underlying data type
-	using type = typename std::remove_cv<T>::type;
+	// Datatype bitsize
+	static constexpr uint bitmax = sizeof(T) * 8; static_assert(N - 1 < bitmax, "bf_base<> error: N out of bounds");
+	
+	// Field bitsize
+	static constexpr uint bitsize = N;
 
-	// Underlying value type (native endianness)
-	using vtype = typename to_ne<type>::type;
+	// Value mask
+	static constexpr utype vmask = static_cast<utype>(~utype{} >> (bitmax - bitsize));
 
-	// Mask of size N
-	constexpr static vtype s_mask = (static_cast<vtype>(1) << N) - 1;
-
-	// Underlying data member
+protected:
 	type m_data;
+};
 
-	// Conversion operator helper (uses SFINAE)
-	template<typename T2, typename = void> struct converter {};
+// Bitfield accessor (N bits from I position, 0 is LSB)
+template<typename T, uint I, uint N>
+struct bf_t : bf_base<T, N>
+{
+	using type = typename bf_t::type;
+	using vtype = typename bf_t::vtype;
+	using utype = typename bf_t::utype;
 
-	template<typename T2> struct converter<T2, std::enable_if_t<std::is_unsigned<T2>::value>>
+	// Field offset
+	static constexpr uint bitpos = I; static_assert(bitpos + N <= bf_t::bitmax, "bf_t<> error: I out of bounds");
+
+	// Get bitmask of size N, at I pos
+	static constexpr utype data_mask()
+	{
+		return static_cast<utype>(~utype{} >> (bf_t::bitmax - bf_t::bitsize) << bitpos);
+	}
+
+	// Bitfield extraction helper
+	template<typename T2, typename = void>
+	struct extract_impl
+	{
+		static_assert(!sizeof(T2), "bf_t<> error: Invalid type");
+	};
+
+	template<typename T2>
+	struct extract_impl<T2, std::enable_if_t<std::is_unsigned<T2>::value>>
 	{
 		// Load unsigned value
-		static inline T2 convert(const type& data)
+		static constexpr T2 extract(const T& data)
 		{
-			return (data >> I) & s_mask;
+			return static_cast<T2>((static_cast<utype>(data) >> bitpos) & bf_t::vmask);
 		}
 	};
 
-	template<typename T2> struct converter<T2, std::enable_if_t<std::is_signed<T2>::value>>
+	template<typename T2>
+	struct extract_impl<T2, std::enable_if_t<std::is_signed<T2>::value>>
 	{
 		// Load signed value (sign-extended)
-		static inline T2 convert(const type& data)
+		static constexpr T2 extract(const T& data)
 		{
-			return data << (sizeof(T) * 8 - I - N) >> (sizeof(T) * 8 - N);
+			return static_cast<T2>(static_cast<vtype>(static_cast<utype>(data) << (bf_t::bitmax - bitpos - N)) >> (bf_t::bitmax - N));
 		}
 	};
 
-public:
-	// Assignment operator (store bitfield value)
-	bf_t& operator =(vtype value)
+	// Bitfield extraction
+	static constexpr vtype extract(const T& data)
 	{
-		m_data = (m_data & ~(s_mask << I)) | (value & s_mask) << I;
-		return *this;
+		return extract_impl<vtype>::extract(data);
 	}
 
-	// Conversion operator (load bitfield value)
-	operator vtype() const
+	// Bitfield insertion
+	static constexpr vtype insert(vtype value)
 	{
-		return converter<vtype>::convert(m_data);
+		return static_cast<vtype>((value & bf_t::vmask) << bitpos);
 	}
 
-	// Get raw data with mask applied
-	type unshifted() const
+	// Load bitfield value
+	constexpr operator vtype() const
 	{
-		return (m_data & (s_mask << I));
+		return extract(this->m_data);
 	}
 
-	// Optimized bool conversion
-	explicit operator bool() const
+	// Load raw data with mask applied
+	constexpr T unshifted() const
+	{
+		return static_cast<T>(this->m_data & data_mask());
+	}
+
+	// Optimized bool conversion (must be removed if inappropriate)
+	explicit constexpr operator bool() const
 	{
 		return unshifted() != 0;
 	}
 
-	// Postfix increment operator
+	// Store bitfield value
+	bf_t& operator =(vtype value)
+	{
+		this->m_data = static_cast<vtype>((this->m_data & ~data_mask()) | insert(value));
+		return *this;
+	}
+
 	vtype operator ++(int)
 	{
-		vtype result = *this;
-		*this = result + 1;
+		utype result = *this;
+		*this = static_cast<vtype>(result + 1);
 		return result;
 	}
 
-	// Prefix increment operator
 	bf_t& operator ++()
 	{
 		return *this = *this + 1;
 	}
 
-	// Postfix decrement operator
 	vtype operator --(int)
 	{
-		vtype result = *this;
-		*this = result - 1;
+		utype result = *this;
+		*this = static_cast<vtype>(result - 1);
 		return result;
 	}
 
-	// Prefix decrement operator
 	bf_t& operator --()
 	{
 		return *this = *this - 1;
 	}
 
-	// Addition assignment operator
 	bf_t& operator +=(vtype right)
 	{
 		return *this = *this + right;
 	}
 
-	// Subtraction assignment operator
 	bf_t& operator -=(vtype right)
 	{
 		return *this = *this - right;
 	}
 
-	// Multiplication assignment operator
 	bf_t& operator *=(vtype right)
 	{
 		return *this = *this * right;
 	}
 
-	// Bitwise AND assignment operator
 	bf_t& operator &=(vtype right)
 	{
-		m_data &= (right & s_mask) << I;
+		this->m_data &= static_cast<vtype>((static_cast<utype>(right) & bf_t::vmask) << bitpos);
 		return *this;
 	}
 
-	// Bitwise OR assignment operator
 	bf_t& operator |=(vtype right)
 	{
-		m_data |= (right & s_mask) << I;
+		this->m_data |= static_cast<vtype>((static_cast<utype>(right) & bf_t::vmask) << bitpos);
 		return *this;
 	}
 
-	// Bitwise XOR assignment operator
 	bf_t& operator ^=(vtype right)
 	{
-		m_data ^= (right & s_mask) << I;
+		this->m_data ^= static_cast<vtype>((static_cast<utype>(right) & bf_t::vmask) << bitpos);
 		return *this;
 	}
 };
 
-template<typename T, u32 I, u32 N> using bf_be_t = bf_t<be_t<T>, I, N>;
+// Field pack (concatenated from left to right)
+template<typename F = void, typename... Fields>
+struct cf_t : bf_base<typename F::type, F::bitsize + cf_t<Fields...>::bitsize>
+{
+	using type = typename cf_t::type;
+	using vtype = typename cf_t::vtype;
+	using utype = typename cf_t::utype;
 
-template<typename T, u32 I, u32 N> using bf_le_t = bf_t<le_t<T>, I, N>;
+	// Get disjunction of all "data" masks of concatenated values
+	static constexpr vtype data_mask()
+	{
+		return static_cast<vtype>(F::data_mask() | cf_t<Fields...>::data_mask());
+	}
+
+	// Extract all bitfields and concatenate
+	static constexpr vtype extract(const type& data)
+	{
+		return static_cast<vtype>(static_cast<utype>(F::extract(data)) << cf_t<Fields...>::bitsize | cf_t<Fields...>::extract(data));
+	}
+
+	// Split bitfields and insert them
+	static constexpr vtype insert(vtype value)
+	{
+		return static_cast<vtype>(F::insert(value >> cf_t<Fields...>::bitsize) | cf_t<Fields...>::insert(value));
+	}
+
+	// Load value
+	constexpr operator vtype() const
+	{
+		return extract(this->m_data);
+	}
+
+	// Store value
+	cf_t& operator =(vtype value)
+	{
+		this->m_data = (this->m_data & ~data_mask()) | insert(value);
+		return *this;
+	}
+};
+
+// Empty field pack (recursion terminator)
+template<>
+struct cf_t<void>
+{
+	static constexpr uint bitsize = 0;
+
+	static constexpr uint data_mask()
+	{
+		return 0;
+	}
+
+	template<typename T>
+	static constexpr auto extract(const T& data) -> decltype(+T())
+	{
+		return 0;
+	}
+
+	template<typename T>
+	static constexpr T insert(T value)
+	{
+		return 0;
+	}
+};
+
+// Fixed field (provides constant values in field pack)
+template<typename T, T V, uint N>
+struct ff_t : bf_base<T, N>
+{
+	using type = typename ff_t::type;
+	using vtype = typename ff_t::vtype;
+
+	// Return constant value
+	static constexpr vtype extract(const type& data)
+	{
+		static_assert((V & ff_t::vmask) == V, "ff_t<> error: V out of bounds");
+		return V;
+	}
+
+	// Get value
+	operator vtype() const
+	{
+		return V;
+	}
+};
+
+template<typename T, uint I, uint N>
+struct fmt_unveil<bf_t<T, I, N>, void>
+{
+	using type = typename fmt_unveil<simple_t<T>>::type;
+
+	static inline auto get(const bf_t<T, I, N>& bf)
+	{
+		return fmt_unveil<type>::get(bf);
+	}
+};
+
+template<typename F, typename... Fields>
+struct fmt_unveil<cf_t<F, Fields...>, void>
+{
+	using type = typename fmt_unveil<simple_t<typename F::type>>::type;
+
+	static inline auto get(const cf_t<F, Fields...>& cf)
+	{
+		return fmt_unveil<type>::get(cf);
+	}
+};
+
+template<typename T, T V, uint N>
+struct fmt_unveil<ff_t<T, V, N>, void>
+{
+	using type = typename fmt_unveil<simple_t<T>>::type;
+
+	static inline auto get(const ff_t<T, V, N>& ff)
+	{
+		return fmt_unveil<type>::get(ff);
+	}
+};

Utilities/Config.cpp

@@ -0,0 +1,285 @@
+#include "stdafx.h"
+#include "Config.h"
+
+#include "yaml-cpp/yaml.h"
+
+namespace cfg
+{
+	logs::channel cfg("CFG", logs::level::notice);
+
+	entry_base::entry_base(type _type)
+		: m_type(_type)
+	{
+		if (_type != type::node)
+		{
+			fmt::throw_exception<std::logic_error>("Invalid root node" HERE);
+		}
+	}
+
+	entry_base::entry_base(type _type, node& owner, const std::string& name)
+		: m_type(_type)
+	{
+		if (!owner.m_nodes.emplace(name, this).second)
+		{
+			fmt::throw_exception<std::logic_error>("Node already exists: %s" HERE, name);
+		}
+	}
+
+	entry_base& entry_base::operator[](const std::string& name) const
+	{
+		if (m_type == type::node)
+		{
+			return *static_cast<const node&>(*this).m_nodes.at(name);
+		}
+
+		fmt::throw_exception<std::logic_error>("Invalid node type" HERE);
+	}
+
+	entry_base& entry_base::operator[](const char* name) const
+	{
+		if (m_type == type::node)
+		{
+			return *static_cast<const node&>(*this).m_nodes.at(name);
+		}
+
+		fmt::throw_exception<std::logic_error>("Invalid node type" HERE);
+	}
+
+	bool entry_base::from_string(const std::string&)
+	{
+		fmt::throw_exception<std::logic_error>("from_string() purecall" HERE);
+	}
+
+	bool entry_base::from_list(std::vector<std::string>&&)
+	{
+		fmt::throw_exception<std::logic_error>("from_list() purecall" HERE);
+	}
+
+	// Emit YAML
+	static void encode(YAML::Emitter& out, const class entry_base& rhs);
+
+	// Incrementally load config entries from YAML::Node.
+	// The config value is preserved if the corresponding YAML node doesn't exist.
+	static void decode(const YAML::Node& data, class entry_base& rhs);
+}
+
+bool cfg::try_to_int64(s64* out, const std::string& value, s64 min, s64 max)
+{
+	// TODO: this could be rewritten without exceptions (but it should be as safe as possible and provide logs)
+	s64 result;
+	std::size_t pos;
+
+	try
+	{
+		result = std::stoll(value, &pos, 0 /* Auto-detect numeric base */);
+	}
+	catch (const std::exception& e)
+	{
+		if (out) cfg.error("cfg::try_to_int('%s'): exception: %s", value, e.what());
+		return false;
+	}
+
+	if (pos != value.size())
+	{
+		if (out) cfg.error("cfg::try_to_int('%s'): unexpected characters (pos=%zu)", value, pos);
+		return false;
+	}
+
+	if (result < min || result > max)
+	{
+		if (out) cfg.error("cfg::try_to_int('%s'): out of bounds (%lld..%lld)", value, min, max);
+		return false;
+	}
+
+	if (out) *out = result;
+	return true;
+}
+
+bool cfg::try_to_enum_value(u64* out, decltype(&fmt_class_string<int>::format) func, const std::string& value)
+{
+	for (u64 i = 0;; i++)
+	{
+		std::string var;
+		func(var, i);
+
+		if (var == value)
+		{
+			if (out) *out = i;
+			return true;
+		}
+
+		std::string hex;
+		fmt_class_string<u64>::format(hex, i);
+		if (var == hex)
+		{
+			break;
+		}
+	}
+
+	try
+	{
+		const auto val = std::stoull(value, nullptr, 0);
+
+		if (out) *out = val;
+		return true;
+	}
+	catch (...)
+	{
+		return false;
+	}
+}
+
+std::vector<std::string> cfg::try_to_enum_list(decltype(&fmt_class_string<int>::format) func)
+{
+	std::vector<std::string> result;
+
+	for (u64 i = 0;; i++)
+	{
+		std::string var;
+		func(var, i);
+
+		std::string hex;
+		fmt_class_string<u64>::format(hex, i);
+		if (var == hex)
+		{
+			break;
+		}
+
+		result.emplace_back(std::move(var));
+	}
+
+	return result;
+}
+
+void cfg::encode(YAML::Emitter& out, const cfg::entry_base& rhs)
+{
+	switch (rhs.get_type())
+	{
+	case type::node:
+	{
+		out << YAML::BeginMap;
+		for (const auto& np : static_cast<const node&>(rhs).get_nodes())
+		{
+			out << YAML::Key << np.first;
+			out << YAML::Value; encode(out, *np.second);
+		}
+
+		out << YAML::EndMap;
+		return;
+	}
+	case type::set:
+	{
+		out << YAML::BeginSeq;
+		for (const auto& str : static_cast<const set_entry&>(rhs).get_set())
+		{
+			out << str;
+		}
+
+		out << YAML::EndSeq;
+		return;
+	}
+	}
+
+	out << rhs.to_string();
+}
+
+void cfg::decode(const YAML::Node& data, cfg::entry_base& rhs)
+{
+	switch (rhs.get_type())
+	{
+	case type::node:
+	{
+		if (data.IsScalar() || data.IsSequence())
+		{
+			return; // ???
+		}
+
+		for (const auto& pair : data)
+		{
+			if (!pair.first.IsScalar()) continue;
+
+			// Find the key among existing nodes
+			const auto name = pair.first.Scalar();
+			const auto found = static_cast<node&>(rhs).get_nodes().find(name);
+
+			if (found != static_cast<node&>(rhs).get_nodes().cend())
+			{
+				decode(pair.second, *found->second);
+			}
+			else
+			{
+				// ???
+			}
+		}
+
+		break;
+	}
+	case type::set:
+	{
+		std::vector<std::string> values;
+
+		if (YAML::convert<decltype(values)>::decode(data, values))
+		{
+			rhs.from_list(std::move(values));
+		}
+
+		break;
+	}
+	default:
+	{
+		std::string value;
+
+		if (YAML::convert<std::string>::decode(data, value))
+		{
+			rhs.from_string(value);
+		}
+
+		break; // ???
+	}
+	}
+}
+
+std::string cfg::node::to_string() const
+{
+	YAML::Emitter out;
+	cfg::encode(out, *this);
+
+	return{ out.c_str(), out.size() };
+}
+
+bool cfg::node::from_string(const std::string& value)
+{
+	cfg::decode(YAML::Load(value), *this);
+	return true;
+}
+
+void cfg::node::from_default()
+{
+	for (auto& node : m_nodes)
+	{
+		node.second->from_default();
+	}
+}
+
+void cfg::bool_entry::from_default()
+{
+	value = def;
+}
+
+void cfg::string_entry::from_default()
+{
+	*this = def;
+}
+
+void cfg::set_entry::from_default()
+{
+	std::lock_guard<std::mutex> lock(m_mutex);
+	m_set = {};
+}
+
+cfg::root_node& cfg::get_root()
+{
+	// Magic static
+	static root_node root;
+	return root;
+}

Utilities/Config.h

@@ -0,0 +1,510 @@
+#pragma once
+
+#include "Utilities/types.h"
+#include "Utilities/Atomic.h"
+#include "Utilities/StrFmt.h"
+
+#include <initializer_list>
+#include <exception>
+#include <utility>
+#include <string>
+#include <vector>
+#include <set>
+#include <unordered_map>
+#include <map>
+#include <mutex>
+
+namespace cfg
+{
+	// Convert string to signed integer
+	bool try_to_int64(s64* out, const std::string& value, s64 min, s64 max);
+
+	// Internal hack
+	bool try_to_enum_value(u64* out, decltype(&fmt_class_string<int>::format) func, const std::string&);
+
+	// Internal hack
+	std::vector<std::string> try_to_enum_list(decltype(&fmt_class_string<int>::format) func);
+
+	// Config tree entry type.
+	enum class type : uint
+	{
+		node = 0, // cfg::node type
+		boolean, // cfg::bool_entry type
+		fixed_map, // cfg::map_entry type
+		enumeration, // cfg::enum_entry type
+		integer, // cfg::int_entry type
+		string, // cfg::string_entry type
+		set, // cfg::set_entry type
+	};
+
+	// Config tree entry abstract base class
+	class entry_base
+	{
+		const type m_type;
+
+	protected:
+		// Ownerless entry constructor
+		entry_base(type _type);
+
+		// Owned entry constructor
+		entry_base(type _type, class node& owner, const std::string& name);
+
+	public:
+		// Disallow copy/move constructors and assignments
+		entry_base(const entry_base&) = delete;
+
+		// Get type
+		type get_type() const { return m_type; }
+
+		// Access child node (must exist)
+		entry_base& operator [](const std::string& name) const; entry_base& operator [](const char* name) const;
+
+		// Reset defaults
+		virtual void from_default() = 0;
+
+		// Convert to string (optional)
+		virtual std::string to_string() const
+		{
+			return{};
+		}
+
+		// Try to convert from string (optional)
+		virtual bool from_string(const std::string&);
+
+		// Get string list (optional)
+		virtual std::vector<std::string> to_list() const
+		{
+			return{};
+		}
+
+		// Set multiple values. Implementation-specific, optional.
+		virtual bool from_list(std::vector<std::string>&&);
+	};
+
+	// Config tree node which contains another nodes
+	class node : public entry_base
+	{
+		std::map<std::string, entry_base*> m_nodes;
+
+		friend class entry_base;
+
+	public:
+		// Root node constructor
+		node()
+			: entry_base(type::node)
+		{
+		}
+
+		// Registered node constructor
+		node(node& owner, const std::string& name)
+			: entry_base(type::node, owner, name)
+		{
+		}
+
+		// Get child nodes
+		const std::map<std::string, entry_base*>& get_nodes() const
+		{
+			return m_nodes;
+		}
+
+		// Serialize node
+		std::string to_string() const override;
+
+		// Deserialize node
+		bool from_string(const std::string& value) override;
+
+		// Set default values
+		void from_default() override;
+	};
+
+	struct bool_entry final : public entry_base
+	{
+		atomic_t<bool> value;
+
+		const bool def;
+
+		bool_entry(node& owner, const std::string& name, bool def = false)
+			: entry_base(type::boolean, owner, name)
+			, value(def)
+			, def(def)
+		{
+		}
+
+		explicit operator bool() const
+		{
+			return value.load();
+		}
+
+		bool_entry& operator =(bool value)
+		{
+			value = value;
+			return *this;
+		}
+
+		void from_default() override;
+
+		std::string to_string() const override
+		{
+			return value.load() ? "true" : "false";
+		}
+
+		bool from_string(const std::string& value) override
+		{
+			if (value == "false")
+				this->value = false;
+			else if (value == "true")
+				this->value = true;
+			else
+				return false;
+
+			return true;
+		}
+	};
+
+	// Value node with fixed set of possible values, each maps to a value of type T.
+	template<typename T>
+	struct map_entry final : public entry_base
+	{
+		using init_type  = std::initializer_list<std::pair<std::string, T>>;
+		using map_type   = std::unordered_map<std::string, T>;
+		using list_type  = std::vector<std::string>;
+		using value_type = typename map_type::value_type;
+
+		static map_type make_map(init_type init)
+		{
+			map_type map(init.size());
+
+			for (const auto& v : init)
+			{
+				// Ensure elements are unique
+				verify(HERE), map.emplace(v.first, v.second).second;
+			}
+
+			return map;
+		}
+
+		static list_type make_list(init_type init)
+		{
+			list_type list; list.reserve(init.size());
+
+			for (const auto& v : init)
+			{
+				list.emplace_back(v.first);
+			}
+
+			return list;
+		}
+
+	public:
+		const map_type map;
+		const list_type list; // Element list sorted in original order
+		const value_type& def; // Pointer to the default value
+
+	private:
+		atomic_t<const value_type*> m_value;
+
+	public:
+		map_entry(node& owner, const std::string& name, const std::string& def, init_type init)
+			: entry_base(type::fixed_map, owner, name)
+			, map(make_map(init))
+			, list(make_list(init))
+			, def(*map.find(def))
+			, m_value(&this->def)
+		{
+		}
+
+		map_entry(node& owner, const std::string& name, std::size_t def_index, init_type init)
+			: map_entry(owner, name, (init.begin() + (def_index < init.size() ? def_index : 0))->first, init)
+		{
+		}
+
+		map_entry(node& owner, const std::string& name, init_type init)
+			: map_entry(owner, name, 0, init)
+		{
+		}
+
+		const T& get() const
+		{
+			return m_value.load()->second;
+		}
+
+		void from_default() override
+		{
+			m_value = &def;
+		}
+
+		std::string to_string() const override
+		{
+			return m_value.load()->first;
+		}
+
+		bool from_string(const std::string& value) override
+		{
+			const auto found = map.find(value);
+
+			if (found == map.end())
+			{
+				return false;
+			}
+			else
+			{
+				m_value = &*found;
+				return true;
+			}
+		}
+
+		std::vector<std::string> to_list() const override
+		{
+			return list;
+		}
+	};
+
+	// Value node with fixed set of possible values, each maps to an enum value of type T.
+	template<typename T, bool External = false>
+	class enum_entry final : public entry_base
+	{
+		// Value or reference
+		std::conditional_t<External, atomic_t<T>&, atomic_t<T>> m_value;
+
+	public:
+		const T def;
+
+		enum_entry(node& owner, const std::string& name, std::conditional_t<External, atomic_t<T>&, T> value)
+			: entry_base(type::enumeration, owner, name)
+			, m_value(value)
+			, def(value)
+		{
+		}
+
+		operator T() const
+		{
+			return m_value.load();
+		}
+
+		enum_entry& operator =(T value)
+		{
+			m_value = value;
+			return *this;
+		}
+
+		void from_default() override
+		{
+			m_value = def;
+		}
+
+		std::string to_string() const override
+		{
+			std::string result;
+			fmt_class_string<T>::format(result, fmt_unveil<T>::get(m_value));
+			return result; // TODO: ???
+		}
+
+		bool from_string(const std::string& value) override
+		{
+			u64 result;
+
+			if (try_to_enum_value(&result, &fmt_class_string<T>::format, value))
+			{
+				// No narrowing check, it's hard to do right there
+				m_value = static_cast<T>(static_cast<std::underlying_type_t<T>>(result));
+				return true;
+			}
+
+			return false;
+		}
+
+		std::vector<std::string> to_list() const override
+		{
+			return try_to_enum_list(&fmt_class_string<T>::format);
+		}
+	};
+
+	// Signed 32/64-bit integer entry with custom Min/Max range.
+	template<s64 Min, s64 Max>
+	class int_entry final : public entry_base
+	{
+		static_assert(Min < Max, "Invalid cfg::int_entry range");
+
+		// Prefer 32 bit type if possible
+		using int_type = std::conditional_t<Min >= INT32_MIN && Max <= INT32_MAX, s32, s64>;
+
+		atomic_t<int_type> m_value;
+
+	public:
+		const int_type def;
+
+		int_entry(node& owner, const std::string& name, int_type def = std::min<int_type>(Max, std::max<int_type>(Min, 0)))
+			: entry_base(type::integer, owner, name)
+			, m_value(def)
+			, def(def)
+		{
+		}
+
+		operator int_type() const
+		{
+			return m_value.load();
+		}
+
+		int_entry& operator =(int_type value)
+		{
+			if (value < Min || value > Max)
+			{
+				fmt::throw_exception("Value out of the valid range: %lld" HERE, s64{ value });
+			}
+
+			m_value = value;
+			return *this;
+		}
+
+		void from_default() override
+		{
+			m_value = def;
+		}
+
+		std::string to_string() const override
+		{
+			return std::to_string(m_value.load());
+		}
+
+		bool from_string(const std::string& value) override
+		{
+			s64 result;
+			if (try_to_int64(&result, value, Min, Max))
+			{
+				m_value = static_cast<int_type>(result);
+				return true;
+			}
+
+			return false;
+		}
+	};
+
+	// Alias for 32 bit int
+	using int32_entry = int_entry<INT32_MIN, INT32_MAX>;
+
+	// Alias for 64 bit int
+	using int64_entry = int_entry<INT64_MIN, INT64_MAX>;
+
+	// Simple string entry with mutex
+	class string_entry final : public entry_base
+	{
+		mutable std::mutex m_mutex;
+		std::string m_value;
+
+	public:
+		const std::string def;
+
+		string_entry(node& owner, const std::string& name, const std::string& def = {})
+			: entry_base(type::string, owner, name)
+			, m_value(def)
+			, def(def)
+		{
+		}
+
+		operator std::string() const
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+			return m_value;
+		}
+
+		std::string get() const
+		{
+			return *this;
+		}
+
+		string_entry& operator =(const std::string& value)
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+			m_value = value;
+			return *this;
+		}
+
+		std::size_t size() const
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+			return m_value.size();
+		}
+
+		void from_default() override;
+
+		std::string to_string() const override
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+			return m_value;
+		}
+
+		bool from_string(const std::string& value) override
+		{
+			*this = value;
+			return true;
+		}
+	};
+
+	// Simple set entry with mutex (TODO: template for various types)
+	class set_entry final : public entry_base
+	{
+		mutable std::mutex m_mutex;
+
+		std::set<std::string> m_set;
+
+	public:
+		// Default value is empty list in current implementation
+		set_entry(node& owner, const std::string& name)
+			: entry_base(type::set, owner, name)
+		{
+		}
+
+		std::set<std::string> get_set() const
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+
+			return m_set;
+		}
+
+		void set_set(std::set<std::string>&& set)
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+			m_set = std::move(set);
+		}
+
+		void from_default() override;
+
+		std::vector<std::string> to_list() const override
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+
+			return{ m_set.begin(), m_set.end() };
+		}
+
+		bool from_list(std::vector<std::string>&& list) override
+		{
+			std::lock_guard<std::mutex> lock(m_mutex);
+			m_set = { std::make_move_iterator(list.begin()), std::make_move_iterator(list.end()) };
+
+			return true;
+		}
+	};
+
+	// Root type with some predefined nodes. Don't change it, this is not mandatory for adding nodes.
+	struct root_node : node
+	{
+		node core  { *this, "Core" };
+		node vfs   { *this, "VFS" };
+		node log   { *this, "Log" };
+		node video { *this, "Video" };
+		node audio { *this, "Audio" };
+		node io    { *this, "Input/Output" };
+		node sys   { *this, "System" };
+		node net   { *this, "Net" };
+		node misc  { *this, "Miscellaneous" };
+	};
+
+	// Get global configuration root instance
+	extern root_node& get_root();
+
+	// Global configuration root instance (cached reference)
+	static root_node& root = get_root();
+}
+
+// Registered log channel
+#define LOG_CHANNEL(name) extern logs::channel name; namespace logs { static cfg::enum_entry<logs::level, true> name(cfg::root.log, #name, ::name.enabled); }

Utilities/File.h

@@ -1,29 +1,50 @@
 #pragma once
 
-namespace fom // file open mode
-{
-	enum open_mode : u32
-	{
-		read = 1 << 0, // enable reading
-		write = 1 << 1, // enable writing
-		append = 1 << 2, // enable appending (always write to the end of file)
-		create = 1 << 3, // create file if it doesn't exist
-		trunc = 1 << 4, // clear opened file if it's not empty
-		excl = 1 << 5, // failure if the file already exists (used with `create`)
+#include "types.h"
+#include "bit_set.h"
 
-		rewrite = write | create | trunc,
-	};
-};
+#include <memory>
+#include <string>
+#include <vector>
+#include <algorithm>
 
 namespace fs
 {
-	enum seek_mode : u32 // file seek mode
+	// File open mode flags
+	enum class open_mode : u32
+	{
+		read,
+		write,
+		append,
+		create,
+		trunc,
+		excl,
+
+		__bitset_enum_max
+	};
+
+	constexpr auto read    = +open_mode::read; // Enable reading
+	constexpr auto write   = +open_mode::write; // Enable writing
+	constexpr auto append  = +open_mode::append; // Always append to the end of the file
+	constexpr auto create  = +open_mode::create; // Create file if it doesn't exist
+	constexpr auto trunc   = +open_mode::trunc; // Clear opened file if it's not empty
+	constexpr auto excl    = +open_mode::excl; // Failure if the file already exists (used with `create`)
+
+	constexpr auto rewrite = open_mode::write + open_mode::create + open_mode::trunc;
+
+	// File seek mode
+	enum class seek_mode : u32
 	{
 		seek_set,
 		seek_cur,
 		seek_end,
 	};
 
+	constexpr auto seek_set = seek_mode::seek_set; // From beginning
+	constexpr auto seek_cur = seek_mode::seek_cur; // From current position
+	constexpr auto seek_end = seek_mode::seek_end; // From end
+
+	// File attributes (TODO)
 	struct stat_t
 	{
 		bool is_directory;
@@ -34,7 +55,58 @@ namespace fs
 		s64 ctime;
 	};
 
-	// Get parent directory for the path (returns empty string on failure)
+	// File handle base
+	struct file_base
+	{
+		virtual ~file_base();
+
+		virtual stat_t stat() = 0;
+		virtual bool trunc(u64 length) = 0;
+		virtual u64 read(void* buffer, u64 size) = 0;
+		virtual u64 write(const void* buffer, u64 size) = 0;
+		virtual u64 seek(s64 offset, seek_mode whence) = 0;
+		virtual u64 size() = 0;
+	};
+
+	// Directory entry (TODO)
+	struct dir_entry : stat_t
+	{
+		std::string name;
+	};
+
+	// Directory handle base
+	struct dir_base
+	{
+		virtual ~dir_base();
+
+		virtual bool read(dir_entry&) = 0;
+		virtual void rewind() = 0;
+	};
+
+	// Virtual device
+	struct device_base
+	{
+		virtual ~device_base();
+
+		virtual bool stat(const std::string& path, stat_t& info) = 0;
+		virtual bool remove_dir(const std::string& path) = 0;
+		virtual bool create_dir(const std::string& path) = 0;
+		virtual bool rename(const std::string& from, const std::string& to) = 0;
+		virtual bool remove(const std::string& path) = 0;
+		virtual bool trunc(const std::string& path, u64 length) = 0;
+		virtual bool utime(const std::string& path, s64 atime, s64 mtime) = 0;
+
+		virtual std::unique_ptr<file_base> open(const std::string& path, bs_t<open_mode> mode) = 0;
+		virtual std::unique_ptr<dir_base> open_dir(const std::string& path) = 0;
+	};
+
+	// Get virtual device for specified path (nullptr for real path)
+	std::shared_ptr<device_base> get_virtual_device(const std::string& path);
+
+	// Set virtual device with specified name (nullptr for deletion)
+	std::shared_ptr<device_base> set_virtual_device(const std::string& root_name, const std::shared_ptr<device_base>&);
+
+	// Try to get parent directory (returns empty string on failure)
 	std::string get_parent_dir(const std::string& path);
 
 	// Get file information
@@ -70,79 +142,109 @@ namespace fs
 	// Change file size (possibly appending zeros)
 	bool truncate_file(const std::string& path, u64 length);
 
+	// Set file access/modification time
+	bool utime(const std::string& path, s64 atime, s64 mtime);
+
 	class file final
 	{
-		using handle_type = std::intptr_t;
+		std::unique_ptr<file_base> m_file;
 
-		constexpr static handle_type null = -1;
-
-		handle_type m_fd = null;
-
-		friend class file_read_map;
-		friend class file_write_map;
+		[[noreturn]] void xnull() const;
+		[[noreturn]] void xfail() const;
 
 	public:
+		// Default constructor
 		file() = default;
 
-		explicit file(const std::string& path, u32 mode = fom::read)
-		{
-			open(path, mode);
-		}
+		// Open file with specified mode
+		explicit file(const std::string& path, bs_t<open_mode> mode = ::fs::read);		
 
-		file(file&& other)
-			: m_fd(other.m_fd)
-		{
-			other.m_fd = null;
-		}
+		// Open memory for read
+		explicit file(const void* ptr, std::size_t size);
 
-		file& operator =(file&& right)
+		// Open file with specified args (forward to constructor)
+		template <typename... Args>
+		bool open(Args&&... args)
 		{
-			std::swap(m_fd, right.m_fd);
-			return *this;
-		}
-
-		~file();
-
-		// Check whether the handle is valid (opened file)
-		bool is_opened() const
-		{
-			return m_fd != null;
+			*this = fs::file(std::forward<Args>(args)...);
+			return m_file.operator bool();
 		}
 
 		// Check whether the handle is valid (opened file)
 		explicit operator bool() const
 		{
-			return is_opened();
+			return m_file.operator bool();
 		}
 
-		// Open specified file with specified mode
-		bool open(const std::string& path, u32 mode = fom::read);
+		// Close the file explicitly
+		void close()
+		{
+			m_file.reset();
+		}
+
+		void reset(std::unique_ptr<file_base>&& ptr)
+		{
+			m_file = std::move(ptr);
+		}
+
+		std::unique_ptr<file_base> release()
+		{
+			return std::move(m_file);
+		}
 
 		// Change file size (possibly appending zero bytes)
-		bool trunc(u64 size) const;
+		bool trunc(u64 length) const
+		{
+			if (!m_file) xnull();
+			return m_file->trunc(length);
+		}
 
 		// Get file information
-		bool stat(stat_t& info) const;
-
-		// Close the file explicitly (destructor automatically closes the file)
-		void close();
+		stat_t stat() const
+		{
+			if (!m_file) xnull();
+			return m_file->stat();
+		}
 
 		// Read the data from the file and return the amount of data written in buffer
-		u64 read(void* buffer, u64 count) const;
+		u64 read(void* buffer, u64 count) const
+		{
+			if (!m_file) xnull();
+			return m_file->read(buffer, count);
+		}
 
 		// Write the data to the file and return the amount of data actually written
-		u64 write(const void* buffer, u64 count) const;
+		u64 write(const void* buffer, u64 count) const
+		{
+			if (!m_file) xnull();
+			return m_file->write(buffer, count);
+		}
 
-		// Move file pointer
-		u64 seek(s64 offset, seek_mode whence = seek_set) const;
+		// Change current position, returns resulting position
+		u64 seek(s64 offset, seek_mode whence = seek_set) const
+		{
+			if (!m_file) xnull();
+			return m_file->seek(offset, whence);
+		}
 
 		// Get file size
-		u64 size() const;
+		u64 size() const
+		{
+			if (!m_file) xnull();
+			return m_file->size();
+		}
+
+		// Get current position
+		u64 pos() const
+		{
+			if (!m_file) xnull();
+			return m_file->seek(0, seek_cur);
+		}
 
 		// Write std::string unconditionally
 		const file& write(const std::string& str) const
 		{
-			CHECK_ASSERTION(write(str.data(), str.size()) == str.size());
+			if (write(str.data(), str.size()) != str.size()) xfail();
 			return *this;
 		}
 
@@ -150,7 +252,7 @@ namespace fs
 		template<typename T>
 		std::enable_if_t<std::is_pod<T>::value && !std::is_pointer<T>::value, const file&> write(const T& data) const
 		{
-			CHECK_ASSERTION(write(std::addressof(data), sizeof(T)) == sizeof(T));
+			if (write(std::addressof(data), sizeof(T)) != sizeof(T)) xfail();
 			return *this;
 		}
 
@@ -158,7 +260,7 @@ namespace fs
 		template<typename T>
 		std::enable_if_t<std::is_pod<T>::value && !std::is_pointer<T>::value, const file&> write(const std::vector<T>& vec) const
 		{
-			CHECK_ASSERTION(write(vec.data(), vec.size() * sizeof(T)) == vec.size() * sizeof(T));
+			if (write(vec.data(), vec.size() * sizeof(T)) != vec.size() * sizeof(T)) xfail();
 			return *this;
 		}
 
@@ -168,6 +270,13 @@ namespace fs
 			return read(&str[0], str.size()) == str.size();
 		}
 
+		// Read std::string
+		bool read(std::string& str, std::size_t size) const
+		{
+			str.resize(size);
+			return read(&str[0], size) == size;
+		}
+
 		// Read POD, sizeof(T) is used
 		template<typename T>
 		std::enable_if_t<std::is_pod<T>::value && !std::is_pointer<T>::value, bool> read(T& data) const
@@ -182,12 +291,20 @@ namespace fs
 			return read(vec.data(), sizeof(T) * vec.size()) == sizeof(T) * vec.size();
 		}
 
+		// Read POD std::vector
+		template<typename T>
+		std::enable_if_t<std::is_pod<T>::value && !std::is_pointer<T>::value, bool> read(std::vector<T>& vec, std::size_t size) const
+		{
+			vec.resize(size);
+			return read(vec.data(), sizeof(T) * size) == sizeof(T) * size;
+		}
+
 		// Read POD (experimental)
 		template<typename T>
 		std::enable_if_t<std::is_pod<T>::value && !std::is_pointer<T>::value, T> read() const
 		{
 			T result;
-			CHECK_ASSERTION(read(result));
+			if (!read(result)) xfail();
 			return result;
 		}
 
@@ -196,7 +313,7 @@ namespace fs
 		{
 			std::string result;
 			result.resize(size());
-			CHECK_ASSERTION(seek(0) != -1 && read(result));
+			if (seek(0), !read(result)) xfail();
 			return result;
 		}
 
@@ -206,164 +323,69 @@ namespace fs
 		{
 			std::vector<T> result;
 			result.resize(size() / sizeof(T));
-			CHECK_ASSERTION(seek(0) != -1 && read(result));
+			if (seek(0), !read(result)) xfail();
 			return result;
 		}
 	};
 
-	// TODO
-	class file_read_map final
-	{
-		char* m_ptr = nullptr;
-		u64 m_size;
-
-	public:
-		file_read_map() = default;
-
-		file_read_map(file_read_map&& right)
-			: m_ptr(right.m_ptr)
-			, m_size(right.m_size)
-		{
-			right.m_ptr = 0;
-		}
-
-		file_read_map& operator =(file_read_map&& right)
-		{
-			std::swap(m_ptr, right.m_ptr);
-			std::swap(m_size, right.m_size);
-			return *this;
-		}
-
-		file_read_map(const file& f)
-		{
-			reset(f);
-		}
-
-		~file_read_map()
-		{
-			reset();
-		}
-
-		// Open file mapping
-		void reset(const file& f);
-		
-		// Close file mapping
-		void reset();
-
-		// Get pointer
-		operator const char*() const
-		{
-			return m_ptr;
-		}
-	};
-
-	// TODO
-	class file_write_map final
-	{
-		char* m_ptr = nullptr;
-		u64 m_size;
-
-	public:
-		file_write_map() = default;
-
-		file_write_map(file_write_map&& right)
-			: m_ptr(right.m_ptr)
-			, m_size(right.m_size)
-		{
-			right.m_ptr = 0;
-		}
-
-		file_write_map& operator =(file_write_map&& right)
-		{
-			std::swap(m_ptr, right.m_ptr);
-			std::swap(m_size, right.m_size);
-			return *this;
-		}
-
-		file_write_map(const file& f)
-		{
-			reset(f);
-		}
-
-		~file_write_map()
-		{
-			reset();
-		}
-
-		// Open file mapping
-		void reset(const file& f);
-
-		// Close file mapping
-		void reset();
-
-		// Get pointer
-		operator char*() const
-		{
-			return m_ptr;
-		}
-	};
-
 	class dir final
 	{
-		std::unique_ptr<char[]> m_path;
-		std::intptr_t m_dd; // handle (aux)
+		std::unique_ptr<dir_base> m_dir;
+
+		[[noreturn]] void xnull() const;
 
 	public:
 		dir() = default;
 
-		explicit dir(const std::string& dirname)
+		// Open dir handle
+		explicit dir(const std::string& path)
 		{
-			open(dirname);
+			open(path);
 		}
 
-		dir(dir&& other)
-			: m_dd(other.m_dd)
-			, m_path(std::move(other.m_path))
-		{
-		}
-
-		dir& operator =(dir&& right)
-		{
-			std::swap(m_dd, right.m_dd);
-			std::swap(m_path, right.m_path);
-			return *this;
-		}
-
-		~dir();
-
-		// Check whether the handle is valid (opened directory)
-		bool is_opened() const
-		{
-			return m_path.operator bool();
-		}
+		// Open specified directory
+		bool open(const std::string& path);
 
 		// Check whether the handle is valid (opened directory)
 		explicit operator bool() const
 		{
-			return is_opened();
+			return m_dir.operator bool();
 		}
 
-		// Open specified directory
-		bool open(const std::string& dirname);
-		
-		// Close the directory explicitly (destructor automatically closes the directory)
-		void close();
-
-		// Get next directory entry (UTF-8 name and file stat)
-		bool read(std::string& name, stat_t& info);
-
-		bool first(std::string& name, stat_t& info);
-
-		struct entry
+		// Close the directory explicitly
+		void close()
 		{
-			std::string name;
-			stat_t info;
-		};
+			m_dir.reset();
+		}
+
+		void reset(std::unique_ptr<dir_base>&& ptr)
+		{
+			m_dir = std::move(ptr);
+		}
+		
+		std::unique_ptr<dir_base> release()
+		{
+			return std::move(m_dir);
+		}
+
+		// Get next directory entry
+		bool read(dir_entry& out) const
+		{
+			if (!m_dir) xnull();
+			return m_dir->read(out);
+		}
+
+		// Reset to the beginning
+		void rewind() const
+		{
+			if (!m_dir) xnull();
+			return m_dir->rewind();
+		}
 
 		class iterator
 		{
-			entry m_entry;
 			dir* m_parent;
+			dir_entry m_entry;
 
 		public:
 			enum class mode
@@ -382,20 +404,16 @@ namespace fs
 
 				if (mode_ == mode::from_first)
 				{
-					m_parent->first(m_entry.name, m_entry.info);
-				}
-				else
-				{
-					m_parent->read(m_entry.name, m_entry.info);
+					m_parent->rewind();
 				}
 
-				if (m_entry.name.empty())
+				if (!m_parent->read(m_entry))
 				{
 					m_parent = nullptr;
 				}
 			}
 
-			entry& operator *()
+			dir_entry& operator *()
 			{
 				return m_entry;
 			}
@@ -414,7 +432,7 @@ namespace fs
 
 		iterator begin()
 		{
-			return{ this };
+			return{ m_dir ? this : nullptr };
 		}
 
 		iterator end()
@@ -426,6 +444,127 @@ namespace fs
 	// Get configuration directory
 	const std::string& get_config_dir();
 
-	// Get executable directory
-	const std::string& get_executable_dir();
+	// Get data/cache directory for specified prefix and suffix
+	std::string get_data_dir(const std::string& prefix, const std::string& location, const std::string& suffix);
+
+	// Get data/cache directory for specified prefix and path (suffix will be filename)
+	std::string get_data_dir(const std::string& prefix, const std::string& path);
+
+	// Delete directory and all its contents recursively
+	void remove_all(const std::string& path, bool remove_root = true);
+
+	// Get size of all files recursively
+	u64 get_dir_size(const std::string& path);
+
+	enum class error : uint
+	{
+		ok = 0,
+
+		inval,
+		noent,
+		exist,
+		acces,
+	};
+
+	// Error code returned
+	extern thread_local error g_tls_error;
+
+	template <typename T>
+	struct container_stream final : file_base
+	{
+		// T can be a reference, but this is not recommended
+		using value_type = typename std::remove_reference_t<T>::value_type;
+
+		T obj;
+		u64 pos;
+
+		container_stream(T&& obj)
+			: obj(std::forward<T>(obj))
+			, pos(0)
+		{
+		}
+
+		~container_stream() override
+		{
+		}
+
+		stat_t stat() override
+		{
+			fmt::raw_error("fs::container_stream<>::stat(): not supported");
+		}
+
+		bool trunc(u64 length) override
+		{
+			obj.resize(length);
+			return true;
+		}
+
+		u64 read(void* buffer, u64 size) override
+		{
+			const u64 end = obj.size();
+
+			if (pos < end)
+			{
+				// Get readable size
+				if (const u64 max = std::min<u64>(size, end - pos))
+				{
+					std::copy(obj.cbegin() + pos, obj.cbegin() + pos + max, static_cast<value_type*>(buffer));
+					pos = pos + max;
+					return max;
+				}
+			}
+
+			return 0;
+		}
+
+		u64 write(const void* buffer, u64 size) override
+		{
+			const u64 old_size = obj.size();
+
+			if (old_size + size < old_size)
+			{
+				fmt::raw_error("fs::container_stream<>::write(): overflow");
+			}
+
+			if (pos > old_size)
+			{
+				// Fill gap if necessary (default-initialized)
+				obj.resize(pos);
+			}
+
+			const auto src = static_cast<const value_type*>(buffer);
+
+			// Overwrite existing part
+			const u64 overlap = std::min<u64>(obj.size() - pos, size);
+			std::copy(src, src + overlap, obj.begin() + pos);
+
+			// Append new data
+			obj.insert(obj.end(), src + overlap, src + size);
+			pos += size;
+
+			return size;
+		}
+
+		u64 seek(s64 offset, seek_mode whence) override
+		{
+			return
+				whence == fs::seek_set ? pos = offset :
+				whence == fs::seek_cur ? pos = offset + pos :
+				whence == fs::seek_end ? pos = offset + size() :
+				(fmt::raw_error("fs::container_stream<>::seek(): invalid whence"), 0);
+		}
+
+		u64 size() override
+		{
+			return obj.size();
+		}
+	};
+
+	template <typename T>
+	file make_stream(T&& container = T{})
+	{
+		file result;
+		result.reset(std::make_unique<container_stream<T>>(std::forward<T>(container)));
+		return result;
+	}
 }

Utilities/GNU.cpp

@@ -1,51 +0,0 @@
-#include "GNU.h"
-
-#ifdef __APPLE__
-#include <sys/types.h>
-#include <sys/_types/_timespec.h>
-#include <mach/mach.h>
-#include <mach/clock.h>
-#include <mach/mach_time.h>
-#undef CPU_STATE_MAX
-
-#define MT_NANO (+1.0E-9)
-#define MT_GIGA UINT64_C(1000000000)
-
-// TODO create a list of timers,
-static double mt_timebase = 0.0;
-static uint64_t mt_timestart = 0;
-
-// TODO be more careful in a multithreaded environement
-int clock_gettime(clockid_t clk_id, struct timespec *tp)
-{
-    kern_return_t retval = KERN_SUCCESS;
-    if( clk_id == TIMER_ABSTIME)
-    {
-        if (!mt_timestart) { // only one timer, initilized on the first call to the TIMER
-            mach_timebase_info_data_t tb = { 0 };
-            mach_timebase_info(&tb);
-            mt_timebase = tb.numer;
-            mt_timebase /= tb.denom;
-            mt_timestart = mach_absolute_time();
-        }
-
-        double diff = (mach_absolute_time() - mt_timestart) * mt_timebase;
-        tp->tv_sec = diff * MT_NANO;
-        tp->tv_nsec = diff - (tp->tv_sec * MT_GIGA);
-    }
-    else // other clk_ids are mapped to the coresponding mach clock_service
-    {
-        clock_serv_t cclock;
-        mach_timespec_t mts;
-
-        host_get_clock_service(mach_host_self(), clk_id, &cclock);
-        retval = clock_get_time(cclock, &mts);
-        mach_port_deallocate(mach_task_self(), cclock);
-
-        tp->tv_sec = mts.tv_sec;
-        tp->tv_nsec = mts.tv_nsec;
-    }
-
-    return retval;
-}
-#endif /* __APPLE__ */

Utilities/GNU.h

@@ -1,292 +0,0 @@
-#pragma once
-
-#include <emmintrin.h>
-
-#if defined(_MSC_VER) && _MSC_VER <= 1800
-#define thread_local __declspec(thread)
-#elif __APPLE__
-#define thread_local __thread
-#endif
-
-#if defined(_MSC_VER)
-#define never_inline __declspec(noinline)
-#else
-#define never_inline __attribute__((noinline))
-#endif
-
-#if defined(_MSC_VER)
-#define safe_buffers __declspec(safebuffers)
-#else
-#define safe_buffers
-#endif
-
-#if defined(_MSC_VER)
-#define force_inline __forceinline
-#else
-#define force_inline __attribute__((always_inline))
-#endif
-
-#if defined(_MSC_VER) && _MSC_VER <= 1800
-#define alignas(x) _CRT_ALIGN(x)
-#endif
-
-#if defined(__GNUG__)
-
-#include <stdlib.h>
-#include <cstdint>
-
-#define _fpclass(x) std::fpclassify(x)
-#define INFINITE 0xFFFFFFFF
-
-#ifdef __APPLE__
-
-// XXX only supports a single timer
-#define TIMER_ABSTIME -1
-/* The opengroup spec isn't clear on the mapping from REALTIME to CALENDAR
- being appropriate or not.
- http://pubs.opengroup.org/onlinepubs/009695299/basedefs/time.h.html */
-#define CLOCK_REALTIME  1 // #define CALENDAR_CLOCK 1 from mach/clock_types.h
-#define CLOCK_MONOTONIC 0 // #define SYSTEM_CLOCK 0
-
-typedef int clockid_t;
-
-/* the mach kernel uses struct mach_timespec, so struct timespec
-    is loaded from <sys/_types/_timespec.h> for compatability */
-// struct timespec { time_t tv_sec; long tv_nsec; };
-
-int clock_gettime(clockid_t clk_id, struct timespec *tp);
-
-#endif /* __APPLE__ */
-#endif /* __GNUG__ */
-
-#if defined(_MSC_VER)
-
-// Unsigned 128-bit integer implementation
-struct alignas(16) u128
-{
-	std::uint64_t lo, hi;
-
-	u128() = default;
-
-	u128(const u128&) = default;
-
-	u128(std::uint64_t l)
-		: lo(l)
-		, hi(0)
-	{
-	}
-
-	u128 operator +(const u128& r) const
-	{
-		u128 value;
-		_addcarry_u64(_addcarry_u64(0, r.lo, lo, &value.lo), r.hi, hi, &value.hi);
-		return value;
-	}
-
-	friend u128 operator +(const u128& l, std::uint64_t r)
-	{
-		u128 value;
-		_addcarry_u64(_addcarry_u64(0, r, l.lo, &value.lo), l.hi, 0, &value.hi);
-		return value;
-	}
-
-	friend u128 operator +(std::uint64_t l, const u128& r)
-	{
-		u128 value;
-		_addcarry_u64(_addcarry_u64(0, r.lo, l, &value.lo), 0, r.hi, &value.hi);
-		return value;
-	}
-
-	u128 operator -(const u128& r) const
-	{
-		u128 value;
-		_subborrow_u64(_subborrow_u64(0, r.lo, lo, &value.lo), r.hi, hi, &value.hi);
-		return value;
-	}
-
-	friend u128 operator -(const u128& l, std::uint64_t r)
-	{
-		u128 value;
-		_subborrow_u64(_subborrow_u64(0, r, l.lo, &value.lo), 0, l.hi, &value.hi);
-		return value;
-	}
-
-	friend u128 operator -(std::uint64_t l, const u128& r)
-	{
-		u128 value;
-		_subborrow_u64(_subborrow_u64(0, r.lo, l, &value.lo), r.hi, 0, &value.hi);
-		return value;
-	}
-
-	u128 operator +() const
-	{
-		return *this;
-	}
-
-	u128 operator -() const
-	{
-		u128 value;
-		_subborrow_u64(_subborrow_u64(0, lo, 0, &value.lo), hi, 0, &value.hi);
-		return value;
-	}
-
-	u128& operator ++()
-	{
-		_addcarry_u64(_addcarry_u64(0, 1, lo, &lo), 0, hi, &hi);
-		return *this;
-	}
-
-	u128 operator ++(int)
-	{
-		u128 value = *this;
-		_addcarry_u64(_addcarry_u64(0, 1, lo, &lo), 0, hi, &hi);
-		return value;
-	}
-
-	u128& operator --()
-	{
-		_subborrow_u64(_subborrow_u64(0, 1, lo, &lo), 0, hi, &hi);
-		return *this;
-	}
-
-	u128 operator --(int)
-	{
-		u128 value = *this;
-		_subborrow_u64(_subborrow_u64(0, 1, lo, &lo), 0, hi, &hi);
-		return value;
-	}
-
-	u128 operator ~() const
-	{
-		u128 value;
-		value.lo = ~lo;
-		value.hi = ~hi;
-		return value;
-	}
-
-	u128 operator &(const u128& r) const
-	{
-		u128 value;
-		value.lo = lo & r.lo;
-		value.hi = hi & r.hi;
-		return value;
-	}
-
-	u128 operator |(const u128& r) const
-	{
-		u128 value;
-		value.lo = lo | r.lo;
-		value.hi = hi | r.hi;
-		return value;
-	}
-
-	u128 operator ^(const u128& r) const
-	{
-		u128 value;
-		value.lo = lo ^ r.lo;
-		value.hi = hi ^ r.hi;
-		return value;
-	}
-
-	u128& operator +=(const u128& r)
-	{
-		_addcarry_u64(_addcarry_u64(0, r.lo, lo, &lo), r.hi, hi, &hi);
-		return *this;
-	}
-
-	u128& operator +=(uint64_t r)
-	{
-		_addcarry_u64(_addcarry_u64(0, r, lo, &lo), 0, hi, &hi);
-		return *this;
-	}
-
-	u128& operator &=(const u128& r)
-	{
-		lo &= r.lo;
-		hi &= r.hi;
-		return *this;
-	}
-
-	u128& operator |=(const u128& r)
-	{
-		lo |= r.lo;
-		hi |= r.hi;
-		return *this;
-	}
-
-	u128& operator ^=(const u128& r)
-	{
-		lo ^= r.lo;
-		hi ^= r.hi;
-		return *this;
-	}
-};
-#endif
-
-inline std::uint32_t cntlz32(std::uint32_t arg)
-{
-#if defined(_MSC_VER)
-	unsigned long res;
-	return _BitScanReverse(&res, arg) ? res ^ 31 : 32;
-#else
-	return arg ? __builtin_clzll(arg) - 32 : 32;
-#endif
-}
-
-inline std::uint64_t cntlz64(std::uint64_t arg)
-{
-#if defined(_MSC_VER)
-	unsigned long res;
-	return _BitScanReverse64(&res, arg) ? res ^ 63 : 64;
-#else
-	return arg ? __builtin_clzll(arg) : 64;
-#endif
-}
-
-// compare 16 packed unsigned bytes (greater than)
-inline __m128i sse_cmpgt_epu8(__m128i A, __m128i B)
-{
-	// (A xor 0x80) > (B xor 0x80)
-	const auto sign = _mm_set1_epi32(0x80808080);
-	return _mm_cmpgt_epi8(_mm_xor_si128(A, sign), _mm_xor_si128(B, sign));
-}
-
-inline __m128i sse_cmpgt_epu16(__m128i A, __m128i B)
-{
-	const auto sign = _mm_set1_epi32(0x80008000);
-	return _mm_cmpgt_epi16(_mm_xor_si128(A, sign), _mm_xor_si128(B, sign));
-}
-
-inline __m128i sse_cmpgt_epu32(__m128i A, __m128i B)
-{
-	const auto sign = _mm_set1_epi32(0x80000000);
-	return _mm_cmpgt_epi32(_mm_xor_si128(A, sign), _mm_xor_si128(B, sign));
-}
-
-inline __m128 sse_exp2_ps(__m128 A)
-{
-	const auto x0 = _mm_max_ps(_mm_min_ps(A, _mm_set1_ps(127.4999961f)), _mm_set1_ps(-127.4999961f));
-	const auto x1 = _mm_add_ps(x0, _mm_set1_ps(0.5f));
-	const auto x2 = _mm_sub_epi32(_mm_cvtps_epi32(x1), _mm_and_si128(_mm_castps_si128(_mm_cmpnlt_ps(_mm_setzero_ps(), x1)), _mm_set1_epi32(1)));
-	const auto x3 = _mm_sub_ps(x0, _mm_cvtepi32_ps(x2));
-	const auto x4 = _mm_mul_ps(x3, x3);
-	const auto x5 = _mm_mul_ps(x3, _mm_add_ps(_mm_mul_ps(_mm_add_ps(_mm_mul_ps(x4, _mm_set1_ps(0.023093347705f)), _mm_set1_ps(20.20206567f)), x4), _mm_set1_ps(1513.906801f)));
-	const auto x6 = _mm_mul_ps(x5, _mm_rcp_ps(_mm_sub_ps(_mm_add_ps(_mm_mul_ps(_mm_set1_ps(233.1842117f), x4), _mm_set1_ps(4368.211667f)), x5)));
-	return _mm_mul_ps(_mm_add_ps(_mm_add_ps(x6, x6), _mm_set1_ps(1.0f)), _mm_castsi128_ps(_mm_slli_epi32(_mm_add_epi32(x2, _mm_set1_epi32(127)), 23)));
-}
-
-inline __m128 sse_log2_ps(__m128 A)
-{
-	const auto _1 = _mm_set1_ps(1.0f);
-	const auto _c = _mm_set1_ps(1.442695040f);
-	const auto x0 = _mm_max_ps(A, _mm_castsi128_ps(_mm_set1_epi32(0x00800000)));
-	const auto x1 = _mm_or_ps(_mm_and_ps(x0, _mm_castsi128_ps(_mm_set1_epi32(0x807fffff))), _1);
-	const auto x2 = _mm_rcp_ps(_mm_add_ps(x1, _1));
-	const auto x3 = _mm_mul_ps(_mm_sub_ps(x1, _1), x2);
-	const auto x4 = _mm_add_ps(x3, x3);
-	const auto x5 = _mm_mul_ps(x4, x4);
-	const auto x6 = _mm_add_ps(_mm_mul_ps(_mm_add_ps(_mm_mul_ps(_mm_set1_ps(-0.7895802789f), x5), _mm_set1_ps(16.38666457f)), x5), _mm_set1_ps(-64.1409953f));
-	const auto x7 = _mm_rcp_ps(_mm_add_ps(_mm_mul_ps(_mm_add_ps(_mm_mul_ps(_mm_set1_ps(-35.67227983f), x5), _mm_set1_ps(312.0937664f)), x5), _mm_set1_ps(-769.6919436f)));
-	const auto x8 = _mm_cvtepi32_ps(_mm_sub_epi32(_mm_srli_epi32(_mm_castps_si128(x0), 23), _mm_set1_epi32(127)));
-	return _mm_add_ps(_mm_mul_ps(_mm_mul_ps(_mm_mul_ps(_mm_mul_ps(x5, x6), x7), x4), _c), _mm_add_ps(_mm_mul_ps(x4, _c), x8));
-}

Utilities/GSL.h

@@ -0,0 +1,10 @@
+#pragma once
+
+#define GSL_THROW_ON_CONTRACT_VIOLATION
+
+#pragma push_macro("new")
+#undef new
+#include <gsl.h>
+#pragma pop_macro("new")
+#undef Expects
+#undef Ensures

Utilities/JIT.cpp

@@ -0,0 +1,410 @@
+#ifdef LLVM_AVAILABLE
+
+#include <unordered_map>
+#include <map>
+#include <unordered_set>
+#include <set>
+#include <array>
+
+#include "types.h"
+#include "StrFmt.h"
+#include "File.h"
+#include "Log.h"
+
+#ifdef _MSC_VER
+#pragma warning(push, 0)
+#endif
+#include "llvm/Support/TargetSelect.h"
+#include "llvm/Support/FormattedStream.h"
+#include "llvm/ExecutionEngine/ExecutionEngine.h"
+#include "llvm/ExecutionEngine/RTDyldMemoryManager.h"
+#include "llvm/ExecutionEngine/JITEventListener.h"
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif
+
+#ifdef _WIN32
+#include <Windows.h>
+#else
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <sys/types.h>
+#endif
+
+#include "JIT.h"
+
+// Global LLVM context (thread-unsafe)
+llvm::LLVMContext g_llvm_ctx;
+
+// Size of virtual memory area reserved: 512 MB
+static const u64 s_memory_size = 0x20000000;
+
+// Try to reserve a portion of virtual memory in the first 2 GB address space beforehand, if possible.
+static void* const s_memory = []() -> void*
+{
+#ifdef _WIN32
+	for (u64 addr = 0x10000000; addr <= 0x60000000; addr += 0x1000000)
+	{
+		if (VirtualAlloc((void*)addr, s_memory_size, MEM_RESERVE, PAGE_NOACCESS))
+		{
+			return (void*)addr;
+		}
+	}
+
+	return VirtualAlloc(NULL, s_memory_size, MEM_RESERVE, PAGE_NOACCESS);
+#else
+	return ::mmap((void*)0x10000000, s_memory_size, PROT_NONE, MAP_ANON | MAP_PRIVATE, -1, 0);
+#endif
+}();
+
+// Code section
+static u8* s_code_addr;
+static u64 s_code_size;
+
+// EH frames
+static u8* s_unwind_info;
+static u64 s_unwind_size;
+
+#ifdef _WIN32
+static std::vector<std::vector<RUNTIME_FUNCTION>> s_unwind; // .pdata
+#endif
+
+// Helper class
+struct MemoryManager final : llvm::RTDyldMemoryManager
+{
+	std::unordered_map<std::string, std::uintptr_t>& m_link;
+
+	MemoryManager(std::unordered_map<std::string, std::uintptr_t>& table)
+		: m_link(table)
+	{
+	}
+
+	[[noreturn]] static void null()
+	{
+		fmt::throw_exception("Null function" HERE);
+	}
+
+	virtual u64 getSymbolAddress(const std::string& name) override
+	{
+		const auto found = m_link.find(name);
+
+		if (found != m_link.end())
+		{
+			return found->second;
+		}
+
+		if (u64 addr = RTDyldMemoryManager::getSymbolAddress(name))
+		{
+			// This may be bad if LLVM requests some built-in functions like fma.
+			LOG_ERROR(GENERAL, "LLVM: Symbol requested: %s -> 0x%016llx", name, addr);
+			return addr;
+		}
+
+		// It's fine if some function is never called, for example.
+		LOG_ERROR(GENERAL, "LLVM: Symbol not found: %s", name);
+		return (u64)null;
+	}
+
+	virtual u8* allocateCodeSection(std::uintptr_t size, uint align, uint sec_id, llvm::StringRef sec_name) override
+	{
+		// Simple allocation
+		const u64 next = ::align((u64)m_next + size, 4096);
+
+		if (next > (u64)s_memory + s_memory_size)
+		{
+			LOG_FATAL(GENERAL, "LLVM: Out of memory (size=0x%llx, aligned 0x%x)", size, align);
+			return nullptr;
+		}
+
+#ifdef _WIN32
+		if (!VirtualAlloc(m_next, size, MEM_COMMIT, PAGE_EXECUTE_READWRITE))
+#else
+		if (::mprotect(m_next, size, PROT_READ | PROT_WRITE | PROT_EXEC))
+#endif
+		{
+			LOG_FATAL(GENERAL, "LLVM: Failed to allocate memory at %p", m_next);
+			return nullptr;
+		}
+
+		s_code_addr = (u8*)m_next;
+		s_code_size = size;
+
+		LOG_NOTICE(GENERAL, "LLVM: Code section %u '%s' allocated -> %p (size=0x%llx, aligned 0x%x)", sec_id, sec_name.data(), m_next, size, align);
+		return (u8*)std::exchange(m_next, (void*)next);
+	}
+
+	virtual u8* allocateDataSection(std::uintptr_t size, uint align, uint sec_id, llvm::StringRef sec_name, bool is_ro) override
+	{
+		// Simple allocation
+		const u64 next = ::align((u64)m_next + size, 4096);
+
+		if (next > (u64)s_memory + s_memory_size)
+		{
+			LOG_FATAL(GENERAL, "LLVM: Out of memory (size=0x%llx, aligned 0x%x)", size, align);
+			return nullptr;
+		}
+
+		if (!is_ro)
+		{
+			LOG_ERROR(GENERAL, "LLVM: Writeable data section not supported!");
+		}
+
+#ifdef _WIN32
+		if (!VirtualAlloc(m_next, size, MEM_COMMIT, PAGE_READWRITE))
+#else
+		if (::mprotect(m_next, size, PROT_READ | PROT_WRITE))
+#endif
+		{
+			LOG_FATAL(GENERAL, "LLVM: Failed to allocate memory at %p", m_next);
+			return nullptr;
+		}
+
+		LOG_NOTICE(GENERAL, "LLVM: Data section %u '%s' allocated -> %p (size=0x%llx, aligned 0x%x, %s)", sec_id, sec_name.data(), m_next, size, align, is_ro ? "ro" : "rw");
+		return (u8*)std::exchange(m_next, (void*)next);
+	}
+
+	virtual bool finalizeMemory(std::string* = nullptr) override
+	{
+		// TODO: make only read-only sections read-only
+//#ifdef _WIN32
+//		DWORD op;
+//		VirtualProtect(s_memory, (u64)m_next - (u64)s_memory, PAGE_READONLY, &op);
+//		VirtualProtect(s_code_addr, s_code_size, PAGE_EXECUTE_READ, &op);
+//#else
+//		::mprotect(s_memory, (u64)m_next - (u64)s_memory, PROT_READ);
+//		::mprotect(s_code_addr, s_code_size, PROT_READ | PROT_EXEC);
+//#endif
+		return false;
+	}
+
+	virtual void registerEHFrames(u8* addr, u64 load_addr, std::size_t size) override
+	{
+		s_unwind_info = addr;
+		s_unwind_size = size;
+
+		return RTDyldMemoryManager::registerEHFrames(addr, load_addr, size);
+	}
+
+	virtual void deregisterEHFrames(u8* addr, u64 load_addr, std::size_t size) override
+	{
+		LOG_ERROR(GENERAL, "deregisterEHFrames() called"); // Not expected
+
+		return RTDyldMemoryManager::deregisterEHFrames(addr, load_addr, size);
+	}
+
+	~MemoryManager()
+	{
+#ifdef _WIN32
+		for (auto&& unwind : s_unwind)
+		{
+			if (!RtlDeleteFunctionTable(unwind.data()))
+			{
+				LOG_FATAL(GENERAL, "RtlDeleteFunctionTable() failed! Error %u", GetLastError());
+			}
+		}
+
+		s_unwind.clear();
+
+		if (!VirtualFree(s_memory, 0, MEM_DECOMMIT))
+		{
+			LOG_FATAL(GENERAL, "VirtualFree(%p) failed! Error %u", s_memory, GetLastError());
+		}
+#else
+		if (!::mmap(s_memory, s_memory_size, PROT_NONE, MAP_FIXED | MAP_ANON | MAP_PRIVATE, -1, 0))
+		{
+			LOG_FATAL(GENERAL, "mmap(%p) failed! Error %d", s_memory, errno);
+		}
+
+		// TODO: unregister EH frames if necessary
+#endif
+	}
+
+private:
+	void* m_next = s_memory;
+};
+
+// Helper class
+struct EventListener final : llvm::JITEventListener
+{
+	std::string path;
+
+	virtual void NotifyObjectEmitted(const llvm::object::ObjectFile& obj, const llvm::RuntimeDyld::LoadedObjectInfo& inf) override
+	{
+		if (!path.empty())
+		{
+			const llvm::StringRef elf = obj.getData();
+			fs::file(path, fs::rewrite).write(elf.data(), elf.size());
+		}
+	}
+};
+
+static EventListener s_listener;
+
+static void dummy()
+{
+}
+
+jit_compiler::jit_compiler(std::unordered_map<std::string, std::uintptr_t> init_linkage_info)
+	: m_link(std::move(init_linkage_info))
+{
+#ifdef _MSC_VER
+	m_link.emplace("__chkstk", (u64)&dummy);
+#endif
+
+	verify(HERE), s_memory;
+
+	// Initialization
+	llvm::InitializeNativeTarget();
+	llvm::InitializeNativeTargetAsmPrinter();
+	LLVMLinkInMCJIT();
+	m_cpu = llvm::sys::getHostCPUName();
+
+	if (m_cpu == "skylake")
+	{
+		m_cpu = "haswell";
+	}
+
+	std::string result;
+
+	m_engine.reset(llvm::EngineBuilder(std::make_unique<llvm::Module>("", g_llvm_ctx))
+		.setErrorStr(&result)
+		.setMCJITMemoryManager(std::make_unique<MemoryManager>(m_link))
+		.setOptLevel(llvm::CodeGenOpt::Aggressive)
+		.setCodeModel((u64)s_memory <= 0x60000000 ? llvm::CodeModel::Small : llvm::CodeModel::Large) // TODO
+		.setMCPU(m_cpu)
+		.create());
+
+	if (!m_engine)
+	{
+		fmt::throw_exception("LLVM: Failed to create ExecutionEngine: %s", result);
+	}
+
+	m_engine->setProcessAllSections(true); // ???
+	m_engine->RegisterJITEventListener(&s_listener);
+}
+
+void jit_compiler::load(std::unique_ptr<llvm::Module> module, std::unique_ptr<llvm::object::ObjectFile> object)
+{
+	s_listener.path.clear();
+
+	auto* module_ptr = module.get();
+
+	m_engine->addModule(std::move(module));
+	m_engine->addObjectFile(std::move(object));
+	m_engine->finalizeObject();
+
+	m_map.clear();
+
+	for (auto& func : module_ptr->functions())
+	{
+		const std::string& name = func.getName();
+
+		if (!m_link.count(name))
+		{
+			// Register compiled function
+			m_map[name] = m_engine->getFunctionAddress(name);
+		}
+	}
+
+	init();
+}
+
+void jit_compiler::make(std::unique_ptr<llvm::Module> module, std::string path)
+{
+	s_listener.path = std::move(path);
+
+	auto* module_ptr = module.get();
+
+	m_engine->addModule(std::move(module));
+	m_engine->finalizeObject();
+
+	m_map.clear();
+
+	for (auto& func : module_ptr->functions())
+	{
+		if (!func.empty())
+		{
+			const std::string& name = func.getName();
+
+			// Register compiled function
+			m_map[name] = m_engine->getFunctionAddress(name);
+		}
+
+		// Delete IR to lower memory consumption
+		func.deleteBody();
+	}
+
+	init();
+}
+
+void jit_compiler::init()
+{
+#ifdef _WIN32
+	// Register .xdata UNWIND_INFO (.pdata section is empty for some reason)
+	std::set<u64> func_set;
+
+	for (const auto& pair : m_map)
+	{
+		func_set.emplace(pair.second);
+	}
+
+	const u64 base = (u64)s_memory;
+	const u8* bits = s_unwind_info;
+
+	std::vector<RUNTIME_FUNCTION> unwind;
+	unwind.reserve(m_map.size());
+
+	for (const u64 addr : func_set)
+	{
+		// Find next function address
+		const auto _next = func_set.upper_bound(addr);
+		const u64 next = _next != func_set.end() ? *_next : (u64)s_code_addr + s_code_size;
+
+		// Generate RUNTIME_FUNCTION record
+		RUNTIME_FUNCTION uw;
+		uw.BeginAddress = static_cast<u32>(addr - base);
+		uw.EndAddress   = static_cast<u32>(next - base);
+		uw.UnwindData   = static_cast<u32>((u64)bits - base);
+		unwind.emplace_back(uw);
+
+		// Parse .xdata UNWIND_INFO record
+		const u8 flags = *bits++; // Version and flags
+		const u8 prolog = *bits++; // Size of prolog
+		const u8 count = *bits++; // Count of unwind codes
+		const u8 frame = *bits++; // Frame Reg + Off
+		bits += ::align(std::max<u8>(1, count), 2) * sizeof(u16); // UNWIND_CODE array
+
+		if (flags != 1) 
+		{
+			// Can't happen for trivial code
+			LOG_ERROR(GENERAL, "LLVM: unsupported UNWIND_INFO version/flags (0x%02x)", flags);
+			break;
+		}
+
+		LOG_TRACE(GENERAL, "LLVM: .xdata at 0x%llx: function 0x%x..0x%x: p0x%02x, c0x%02x, f0x%02x", uw.UnwindData + base, uw.BeginAddress + base, uw.EndAddress + base, prolog, count, frame);
+	}
+
+	if (s_unwind_info + s_unwind_size != bits)
+	{
+		LOG_ERROR(GENERAL, "LLVM: .xdata analysis failed! (%p != %p)", s_unwind_info + s_unwind_size, bits);
+	}
+	else if (!RtlAddFunctionTable(unwind.data(), (DWORD)unwind.size(), base))
+	{
+		LOG_ERROR(GENERAL, "RtlAddFunctionTable(%p) failed! Error %u", s_unwind_info, GetLastError());
+	}
+	else
+	{
+		LOG_NOTICE(GENERAL, "LLVM: UNWIND_INFO registered (%p, size=0x%llx)", s_unwind_info, s_unwind_size);
+	}
+
+	s_unwind.emplace_back(std::move(unwind));
+#endif
+}
+
+jit_compiler::~jit_compiler()
+{
+}
+
+#endif

Utilities/JIT.h

@@ -0,0 +1,73 @@
+#pragma once
+
+#ifdef LLVM_AVAILABLE
+
+#include <memory>
+#include <string>
+#include <unordered_map>
+
+#include "types.h"
+
+#include "restore_new.h"
+#ifdef _MSC_VER
+#pragma warning(push, 0)
+#endif
+#include "llvm/IR/LLVMContext.h"
+#include "llvm/IR/Module.h"
+#include "llvm/ExecutionEngine/ExecutionEngine.h"
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif
+#include "define_new_memleakdetect.h"
+
+extern llvm::LLVMContext g_llvm_ctx;
+
+// Temporary compiler interface
+class jit_compiler final
+{
+	// Execution instance
+	std::unique_ptr<llvm::ExecutionEngine> m_engine;
+
+	// Compiled functions
+	std::unordered_map<std::string, std::uintptr_t> m_map;
+
+	// Linkage cache
+	std::unordered_map<std::string, std::uintptr_t> m_link;
+
+	// Arch
+	std::string m_cpu;
+
+	// Internal
+	void init();
+
+public:
+	jit_compiler(std::unordered_map<std::string, std::uintptr_t>);
+	~jit_compiler();
+
+	// Compile module
+	void make(std::unique_ptr<llvm::Module>, std::string);
+
+	// Load object
+	void load(std::unique_ptr<llvm::Module>, std::unique_ptr<llvm::object::ObjectFile>);
+
+	// Get compiled function address
+	std::uintptr_t get(const std::string& name) const
+	{
+		const auto found = m_map.find(name);
+		
+		if (found != m_map.end())
+		{
+			return found->second;
+		}
+
+		return 0;
+	}
+
+	// Get CPU info
+	const std::string& cpu() const
+	{
+		return m_cpu;
+	}
+};
+
+#endif

Utilities/LICENSE_1_0.txt

@@ -0,0 +1,23 @@
+Boost Software License - Version 1.0 - August 17th, 2003
+
+Permission is hereby granted, free of charge, to any person or organization
+obtaining a copy of the software and accompanying documentation covered by
+this license (the "Software") to use, reproduce, display, distribute,
+execute, and transmit the Software, and to prepare derivative works of the
+Software, and to permit third-parties to whom the Software is furnished to
+do so, all subject to the following:
+
+The copyright notices in the Software and this entire statement, including
+the above license grant, this restriction and the following disclaimer,
+must be included in all copies of the Software, in whole or in part, and
+all derivative works of the Software, unless such copies or derivative
+works are solely in the form of machine-executable object code generated by
+a source language processor.
+
+THE SOFTWARE IS 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, TITLE AND NON-INFRINGEMENT. IN NO EVENT
+SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE
+FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE,
+ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+DEALINGS IN THE SOFTWARE.

Utilities/Log.cpp

@@ -1,26 +1,111 @@
-#include "stdafx.h"
-#include "Thread.h"
+#include "Log.h"
 #include "File.h"
-#include "Log.h"
+#include "StrFmt.h"
+
+#include "rpcs3_version.h"
+#include <string>
 
 #ifdef _WIN32
 #include <Windows.h>
+#else
+#include <chrono>
 #endif
 
-namespace _log
+// Thread-specific log prefix provider
+thread_local std::string(*g_tls_log_prefix)() = nullptr;
+
+template<>
+void fmt_class_string<logs::level>::format(std::string& out, u64 arg)
 {
-	logger& get_logger()
+	format_enum(out, arg, [](auto lev)
 	{
-		// Use magic static for global logger instance
-		static logger instance;
-		return instance;
+		switch (lev)
+		{
+		case logs::level::always: return "Nothing";
+		case logs::level::fatal: return "Fatal";
+		case logs::level::error: return "Error";
+		case logs::level::todo: return "TODO";
+		case logs::level::success: return "Success";
+		case logs::level::warning: return "Warning";
+		case logs::level::notice: return "Notice";
+		case logs::level::trace: return "Trace";
+		}
+
+		return unknown;
+	});
+}
+
+namespace logs
+{
+	class file_writer
+	{
+		// Could be memory-mapped file
+		fs::file m_file;
+
+	public:
+		file_writer(const std::string& name);
+
+		virtual ~file_writer() = default;
+
+		// Append raw data
+		void log(const char* text, std::size_t size);
+	};
+
+	struct file_listener : public file_writer, public listener
+	{
+		file_listener(const std::string& name)
+			: file_writer(name)
+			, listener()
+		{
+			const std::string& start = fmt::format("\xEF\xBB\xBF" "RPCS3 v%s\n", rpcs3::version.to_string());
+			file_writer::log(start.data(), start.size());
+		}
+
+		// Encode level, current thread name, channel name and write log message
+		virtual void log(u64 stamp, const message& msg, const std::string& prefix, const std::string& text) override;
+	};
+
+	static file_listener* get_logger()
+	{
+		// Use magic static
+		static file_listener logger("RPCS3.log");
+		return &logger;
 	}
 
-	file_listener g_log_file(_PRGNAME_ ".log");
+	static u64 get_stamp()
+	{
+		static struct time_initializer
+		{
+#ifdef _WIN32
+			LARGE_INTEGER freq;
+			LARGE_INTEGER start;
 
-	file_writer g_tty_file("TTY.log");
+			time_initializer()
+			{
+				QueryPerformanceFrequency(&freq);
+				QueryPerformanceCounter(&start);
+			}
+#else
+			std::chrono::steady_clock::time_point start = std::chrono::steady_clock::now();
+#endif
 
-	channel GENERAL("", level::notice);
+			u64 get() const
+			{
+#ifdef _WIN32
+				LARGE_INTEGER now;
+				QueryPerformanceCounter(&now);
+				const LONGLONG diff = now.QuadPart - start.QuadPart;
+				return diff / freq.QuadPart * 1'000'000 + diff % freq.QuadPart * 1'000'000 / freq.QuadPart;
+#else
+				return std::chrono::duration_cast<std::chrono::microseconds>(std::chrono::steady_clock::now() - start).count();
+#endif
+			}
+		} timebase{};
+
+		return timebase.get();
+	}
+
+	channel GENERAL(nullptr, level::notice);
 	channel LOADER("LDR", level::notice);
 	channel MEMORY("MEM", level::notice);
 	channel RSX("RSX", level::notice);
@@ -30,121 +115,106 @@ namespace _log
 	channel ARMv7("ARMv7");
 }
 
-_log::listener::listener()
+logs::listener::~listener()
 {
-	// Register self
-	get_logger().add_listener(this);
 }
 
-_log::listener::~listener()
+void logs::listener::add(logs::listener* _new)
 {
-	// Unregister self
-	get_logger().remove_listener(this);
-}
+	// Get first (main) listener
+	listener* lis = get_logger();
 
-_log::channel::channel(const std::string& name, _log::level init_level)
-	: name{ name }
-	, enabled{ init_level }
-{
-	// TODO: register config property "name" associated with "enabled" member
-}
-
-void _log::logger::add_listener(_log::listener* listener)
-{
-	std::lock_guard<shared_mutex> lock(m_mutex);
-
-	m_listeners.emplace(listener);
-}
-
-void _log::logger::remove_listener(_log::listener* listener)
-{
-	std::lock_guard<shared_mutex> lock(m_mutex);
-
-	m_listeners.erase(listener);
-}
-
-void _log::logger::broadcast(const _log::channel& ch, _log::level sev, const std::string& text) const
-{
-	reader_lock lock(m_mutex);
-
-	for (auto listener : m_listeners)
+	// Install new listener at the end of linked list
+	while (lis->m_next || !lis->m_next.compare_and_swap_test(nullptr, _new))
 	{
-		listener->log(ch, sev, text);
+		lis = lis->m_next;
 	}
 }
 
-void _log::broadcast(const _log::channel& ch, _log::level sev, const std::string& text)
+void logs::message::broadcast(const char* fmt, const fmt_type_info* sup, const u64* args)
 {
-	get_logger().broadcast(ch, sev, text);
+	// Get timestamp
+	const u64 stamp = get_stamp();
+
+	std::string text; fmt::raw_append(text, fmt, sup, args);
+	std::string prefix(g_tls_log_prefix ? g_tls_log_prefix() : "");
+
+	// Get first (main) listener
+	listener* lis = get_logger();
+	
+	// Send message to all listeners
+	while (lis)
+	{
+		lis->log(stamp, *this, prefix, text);
+		lis = lis->m_next;
+	}
 }
 
-_log::file_writer::file_writer(const std::string& name)
+[[noreturn]] extern void catch_all_exceptions();
+
+logs::file_writer::file_writer(const std::string& name)
 {
 	try
 	{
-		if (!m_file.open(fs::get_config_dir() + name, fom::rewrite | fom::append))
+		if (!m_file.open(fs::get_config_dir() + name, fs::rewrite + fs::append))
 		{
-			throw EXCEPTION("Can't create log file %s (error %d)", name, errno);
+			fmt::throw_exception("Can't create log file %s (error %s)", name, fs::g_tls_error);
 		}
 	}
-	catch (const fmt::exception& e)
+	catch (...)
 	{
-#ifdef _WIN32
-		MessageBoxA(0, e.what(), "_log::file_writer() failed", MB_ICONERROR);
-#else
-		std::printf("_log::file_writer() failed: %s\n", e.what());
-#endif
+		catch_all_exceptions();
 	}
 }
 
-void _log::file_writer::log(const std::string& text)
+void logs::file_writer::log(const char* text, std::size_t size)
 {
-	m_file.write(text);
+	m_file.write(text, size);
 }
 
-std::size_t _log::file_writer::size() const
+void logs::file_listener::log(u64 stamp, const logs::message& msg, const std::string& prefix, const std::string& _text)
 {
-	return m_file.seek(0, fs::seek_cur);
-}
-
-void _log::file_listener::log(const _log::channel& ch, _log::level sev, const std::string& text)
-{
-	std::string msg; msg.reserve(text.size() + 200);
+	std::string text; text.reserve(prefix.size() + _text.size() + 200);
 
 	// Used character: U+00B7 (Middle Dot)
-	switch (sev)
+	switch (msg.sev)
 	{
-	case level::always:  msg = u8"·A "; break;
-	case level::fatal:   msg = u8"·F "; break;
-	case level::error:   msg = u8"·E "; break;
-	case level::todo:    msg = u8"·U "; break;
-	case level::success: msg = u8"·S "; break;
-	case level::warning: msg = u8"·W "; break;
-	case level::notice:  msg = u8"·! "; break;
-	case level::trace:   msg = u8"·T "; break;
+	case level::always:  text = u8"·A "; break;
+	case level::fatal:   text = u8"·F "; break;
+	case level::error:   text = u8"·E "; break;
+	case level::todo:    text = u8"·U "; break;
+	case level::success: text = u8"·S "; break;
+	case level::warning: text = u8"·W "; break;
+	case level::notice:  text = u8"·! "; break;
+	case level::trace:   text = u8"·T "; break;
 	}
 
-	// TODO: print time?
+	// Print miscosecond timestamp
+	const u64 hours = stamp / 3600'000'000;
+	const u64 mins = (stamp % 3600'000'000) / 60'000'000;
+	const u64 secs = (stamp % 60'000'000) / 1'000'000;
+	const u64 frac = (stamp % 1'000'000);
+	fmt::append(text, "%u:%02u:%02u.%06u ", hours, mins, secs, frac);
 
-	if (auto t = thread_ctrl::get_current())
+	if (prefix.size() > 0)
 	{
-		msg += '{';
-		msg += t->get_name();
-		msg += "} ";
-	}
-
-	if (ch.name.size())
-	{
-		msg += ch.name;
-		msg += sev == level::todo ? " TODO: " : ": ";
-	}
-	else if (sev == level::todo)
-	{
-		msg += "TODO: ";
+		text += "{";
+		text += prefix;
+		text += "} ";
 	}
 	
-	msg += text;
-	msg += '\n';
+	if (msg.ch->name)
+	{
+		text += msg.ch->name;
+		text += msg.sev == level::todo ? " TODO: " : ": ";
+	}
+	else if (msg.sev == level::todo)
+	{
+		text += "TODO: ";
+	}
+	
+	text += _text;
+	text += '\n';
 
-	file_writer::log(msg);
+	file_writer::log(text.data(), text.size());
 }

Utilities/Log.h

@@ -1,8 +1,10 @@
 #pragma once
 
-#include "SharedMutex.h"
+#include "types.h"
+#include "Atomic.h"
+#include "StrFmt.h"
 
-namespace _log
+namespace logs
 {
 	enum class level : uint
 	{
@@ -17,61 +19,64 @@ namespace _log
 	};
 
 	struct channel;
-	struct listener;
 
-	// Log manager
-	class logger final
+	// Message information (temporary data)
+	struct message
 	{
-		mutable shared_mutex m_mutex;
+		const channel* ch;
+		level sev;
 
-		std::set<listener*> m_listeners;
-
-	public:
-		// Register listener
-		void add_listener(listener* listener);
-
-		// Unregister listener
-		void remove_listener(listener* listener);
-
-		// Send log message to all listeners
-		void broadcast(const channel& ch, level sev, const std::string& text) const;
+		// Send log message to global logger instance
+		void broadcast(const char*, const fmt_type_info*, const u64*);
 	};
 
-	// Send log message to global logger instance
-	void broadcast(const channel& ch, level sev, const std::string& text);
+	class listener
+	{
+		// Next listener (linked list)
+		atomic_t<listener*> m_next{};
+
+		friend struct message;
+
+	public:
+		constexpr listener() = default;
+
+		virtual ~listener();
+
+		// Process log message
+		virtual void log(u64 stamp, const message& msg, const std::string& prefix, const std::string& text) = 0;
+
+		// Add new listener
+		static void add(listener*);
+	};
 
-	// Log channel (source)
 	struct channel
 	{
-		// Channel prefix (also used for identification)
-		const std::string name;
+		// Channel prefix (added to every log message)
+		const char* const name;
 
 		// The lowest logging level enabled for this channel (used for early filtering)
-		std::atomic<level> enabled;
+		atomic_t<level> enabled;
 
-		// Initialization (max level enabled by default)
-		channel(const std::string& name, level = level::trace);
-
-		virtual ~channel() = default;
-
-		// Log without formatting
-		force_inline void log(level sev, const std::string& text) const
+		// Constant initialization: name and initial log level
+		constexpr channel(const char* name, level enabled = level::trace)
+			: name(name)
+			, enabled(enabled)
 		{
-			if (sev <= enabled)
-				broadcast(*this, sev, text);
 		}
 
-		// Log with formatting
+		// Formatting function
 		template<typename... Args>
-		force_inline safe_buffers void format(level sev, const char* fmt, const Args&... args) const
+		SAFE_BUFFERS FORCE_INLINE void format(level sev, const char* fmt, const Args&... args) const
 		{
-			if (sev <= enabled)
-				broadcast(*this, sev, fmt::format(fmt, fmt::do_unveil(args)...));
+			if (UNLIKELY(sev <= enabled))
+			{
+				message{this, sev}.broadcast(fmt, fmt::get_type_info<fmt_unveil_t<Args>...>(), fmt_args_t<Args...>{fmt_unveil<Args>::get(args)...});
+			}
 		}
 
 #define GEN_LOG_METHOD(_sev)\
 		template<typename... Args>\
-		force_inline void _sev(const char* fmt, const Args&... args)\
+		SAFE_BUFFERS void _sev(const char* fmt, const Args&... args) const\
 		{\
 			return format<Args...>(level::_sev, fmt, args...);\
 		}
@@ -87,52 +92,7 @@ namespace _log
 #undef GEN_LOG_METHOD
 	};
 
-	// Log listener (destination)
-	struct listener
-	{
-		listener();
-		
-		virtual ~listener();
-
-		virtual void log(const channel& ch, level sev, const std::string& text) = 0;
-	};
-
-	class file_writer
-	{
-		// Could be memory-mapped file
-		fs::file m_file;
-
-	public:
-		file_writer(const std::string& name);
-
-		virtual ~file_writer() = default;
-
-		// Append raw data
-		void log(const std::string& text);
-
-		// Get current file size (may be used by secondary readers)
-		std::size_t size() const;
-	};
-
-	struct file_listener : public file_writer, public listener
-	{
-		file_listener(const std::string& name)
-			: file_writer(name)
-			, listener()
-		{
-		}
-
-		// Encode level, current thread name, channel name and write log message
-		virtual void log(const channel& ch, level sev, const std::string& text) override;
-	};
-
-	// Global variable for RPCS3.log
-	extern file_listener g_log_file;
-
-	// Global variable for TTY.log
-	extern file_writer g_tty_file;
-
-	// Small set of predefined channels:
+	/* Small set of predefined channels */
 
 	extern channel GENERAL;
 	extern channel LOADER;
@@ -146,10 +106,10 @@ namespace _log
 
 // Legacy:
 
-#define LOG_SUCCESS(ch, fmt, ...) _log::ch.success(fmt, ##__VA_ARGS__)
-#define LOG_NOTICE(ch, fmt, ...)  _log::ch.notice (fmt, ##__VA_ARGS__)
-#define LOG_WARNING(ch, fmt, ...) _log::ch.warning(fmt, ##__VA_ARGS__)
-#define LOG_ERROR(ch, fmt, ...)   _log::ch.error  (fmt, ##__VA_ARGS__)
-#define LOG_TODO(ch, fmt, ...)    _log::ch.todo   (fmt, ##__VA_ARGS__)
-#define LOG_TRACE(ch, fmt, ...)   _log::ch.trace  (fmt, ##__VA_ARGS__)
-#define LOG_FATAL(ch, fmt, ...)   _log::ch.fatal  (fmt, ##__VA_ARGS__)
+#define LOG_SUCCESS(ch, fmt, ...) logs::ch.success("" fmt, ##__VA_ARGS__)
+#define LOG_NOTICE(ch, fmt, ...)  logs::ch.notice ("" fmt, ##__VA_ARGS__)
+#define LOG_WARNING(ch, fmt, ...) logs::ch.warning("" fmt, ##__VA_ARGS__)
+#define LOG_ERROR(ch, fmt, ...)   logs::ch.error  ("" fmt, ##__VA_ARGS__)
+#define LOG_TODO(ch, fmt, ...)    logs::ch.todo   ("" fmt, ##__VA_ARGS__)
+#define LOG_TRACE(ch, fmt, ...)   logs::ch.trace  ("" fmt, ##__VA_ARGS__)
+#define LOG_FATAL(ch, fmt, ...)   logs::ch.fatal  ("" fmt, ##__VA_ARGS__)

Utilities/MTProgressDialog.h

@@ -61,7 +61,7 @@ public:
 		Show();
 	}
 
-	force_inline void Update(const u8 thread_id, const u64 value, const wxString& msg)
+	void Update(const u8 thread_id, const u64 value, const wxString& msg)
 	{
 		if(thread_id > m_cores) return;
 

Utilities/Semaphore.cpp

@@ -1,120 +0,0 @@
-#include "stdafx.h"
-#include "Utilities/Semaphore.h"
-
-bool semaphore_t::try_wait()
-{
-	// check m_value without interlocked op
-	if (m_var.load().value == 0)
-	{
-		return false;
-	}
-
-	// try to decrement m_value atomically
-	const auto old = m_var.atomic_op([](sync_var_t& var)
-	{
-		if (var.value)
-		{
-			var.value--;
-		}
-	});
-
-	// recheck atomic result
-	if (old.value == 0)
-	{
-		return false;
-	}
-
-	return true;
-}
-
-bool semaphore_t::try_post()
-{
-	// check m_value without interlocked op
-	if (m_var.load().value >= max_value)
-	{
-		return false;
-	}
-
-	// try to increment m_value atomically
-	const auto old = m_var.atomic_op([&](sync_var_t& var)
-	{
-		if (var.value < max_value)
-		{
-			var.value++;
-		}
-	});
-
-	// recheck atomic result
-	if (old.value >= max_value)
-	{
-		return false;
-	}
-
-	if (old.waiters)
-	{
-		// notify waiting thread
-		std::lock_guard<std::mutex> lock(m_mutex);
-
-		m_cv.notify_one();
-	}
-
-	return true;
-}
-
-void semaphore_t::wait()
-{
-	if (m_var.atomic_op([](sync_var_t& var) -> bool
-	{
-		if (var.value)
-		{
-			var.value--;
-
-			return true;
-		}
-		else
-		{
-			//var.waiters++;
-
-			return false;
-		}
-	}))
-	{
-		return;
-	}
-
-	std::unique_lock<std::mutex> lock(m_mutex);
-
-	m_var.atomic_op([](sync_var_t& var)
-	{
-		var.waiters++;
-	});
-
-	while (!m_var.atomic_op([](sync_var_t& var) -> bool
-	{
-		if (var.value)
-		{
-			var.value--;
-			var.waiters--;
-
-			return true;
-		}
-		else
-		{
-			return false;
-		}
-	}))
-	{
-		m_cv.wait(lock);
-	}
-}
-
-bool semaphore_t::post_and_wait()
-{
-	// TODO: merge these functions? Probably has a race condition.
-	if (try_wait()) return false;
-
-	try_post();
-	wait();
-
-	return true;
-}

Utilities/Semaphore.h

@@ -1,37 +0,0 @@
-#pragma once
-
-class semaphore_t
-{
-	// semaphore mutex
-	std::mutex m_mutex;
-
-	// semaphore condition variable
-	std::condition_variable m_cv;
-
-	struct sync_var_t
-	{
-		u32 value; // current semaphore value
-		u32 waiters; // current amount of waiters
-	};
-
-	// current semaphore value
-	atomic_t<sync_var_t> m_var;
-
-public:
-	// max semaphore value
-	const u32 max_value;
-
-	semaphore_t(u32 max_value = 1, u32 value = 0)
-		: m_var(sync_var_t{ value, 0 })
-		, max_value(max_value)
-	{
-	}
-
-	bool try_wait();
-
-	bool try_post();
-
-	void wait();
-
-	bool post_and_wait();
-};

Utilities/SharedMutex.cpp

@@ -1,103 +0,0 @@
-#include "stdafx.h"
-#include "SharedMutex.h"
-
-void shared_mutex::impl_lock_shared(u32 old_value)
-{
-	// Throw if reader count breaks the "second" limit (it should be impossible)
-	CHECK_ASSERTION((old_value & SM_READER_COUNT) != SM_READER_COUNT);
-
-	std::unique_lock<std::mutex> lock(m_mutex);
-
-	// Notify non-zero reader queue size
-	m_ctrl |= SM_READER_QUEUE;
-
-	// Compensate incorrectly increased reader count
-	if ((--m_ctrl & SM_READER_COUNT) == 0 && m_wq_size)
-	{
-		// Notify current exclusive owner (condition passed)
-		m_ocv.notify_one();
-	}
-
-	CHECK_ASSERTION(++m_rq_size);
-
-	// Obtain the reader lock
-	while (!atomic_op(m_ctrl, op_lock_shared))
-	{
-		m_rcv.wait(lock);
-	}
-
-	CHECK_ASSERTION(m_rq_size--);
-
-	if (m_rq_size == 0)
-	{
-		m_ctrl &= ~SM_READER_QUEUE;
-	}
-}
-
-void shared_mutex::impl_unlock_shared(u32 new_value)
-{
-	// Throw if reader count was zero
-	CHECK_ASSERTION((new_value & SM_READER_COUNT) != SM_READER_COUNT);
-
-	// Mutex cannot be unlocked before notification because m_ctrl has been changed outside
-	std::lock_guard<std::mutex> lock(m_mutex);
-
-	if (m_wq_size && (new_value & SM_READER_COUNT) == 0)
-	{
-		// Notify current exclusive owner that the latest reader is gone
-		m_ocv.notify_one();
-	}
-	else if (m_rq_size)
-	{
-		m_rcv.notify_one();
-	}
-}
-
-void shared_mutex::impl_lock_excl(u32 value)
-{
-	std::unique_lock<std::mutex> lock(m_mutex);
-
-	// Notify non-zero writer queue size
-	m_ctrl |= SM_WRITER_QUEUE;
-
-	CHECK_ASSERTION(++m_wq_size);
-
-	// Obtain the writer lock
-	while (!atomic_op(m_ctrl, op_lock_excl))
-	{
-		m_wcv.wait(lock);
-	}
-
-	// Wait for remaining readers
-	while ((m_ctrl & SM_READER_COUNT) != 0)
-	{
-		m_ocv.wait(lock);
-	}
-
-	CHECK_ASSERTION(m_wq_size--);
-
-	if (m_wq_size == 0)
-	{
-		m_ctrl &= ~SM_WRITER_QUEUE;
-	}
-}
-
-void shared_mutex::impl_unlock_excl(u32 value)
-{
-	// Throw if was not locked exclusively
-	CHECK_ASSERTION(value & SM_WRITER_LOCK);
-
-	// Mutex cannot be unlocked before notification because m_ctrl has been changed outside
-	std::lock_guard<std::mutex> lock(m_mutex);
-	
-	if (m_wq_size)
-	{
-		// Notify next exclusive owner
-		m_wcv.notify_one();
-	}
-	else if (m_rq_size)
-	{
-		// Notify all readers
-		m_rcv.notify_all();
-	}
-}

Utilities/SharedMutex.h

@@ -1,136 +0,0 @@
-#pragma once
-
-//! An attempt to create effective implementation of "shared mutex", lock-free in optimistic case.
-//! All locking and unlocking may be done by single LOCK XADD or LOCK CMPXCHG instructions.
-//! MSVC implementation of std::shared_timed_mutex seems suboptimal.
-//! std::shared_mutex is not available until C++17.
-class shared_mutex final
-{
-	enum : u32
-	{
-		SM_WRITER_LOCK  = 1u << 31, // Exclusive lock flag, must be MSB
-		SM_WRITER_QUEUE = 1u << 30, // Flag set if m_wq_size != 0
-		SM_READER_QUEUE = 1u << 29, // Flag set if m_rq_size != 0
-
-		SM_READER_COUNT = SM_READER_QUEUE - 1, // Valid reader count bit mask
-		SM_READER_MAX   = 1u << 24, // Max reader count
-	};
-
-	std::atomic<u32> m_ctrl{}; // Control atomic variable: reader count | SM_* flags
-	std::thread::id m_owner{}; // Current exclusive owner (TODO: implement only for debug mode?)
-
-	std::mutex m_mutex;
-
-	u32 m_rq_size{}; // Reader queue size (threads waiting on m_rcv)
-	u32 m_wq_size{}; // Writer queue size (threads waiting on m_wcv+m_ocv)
-	
-	std::condition_variable m_rcv; // Reader queue
-	std::condition_variable m_wcv; // Writer queue
-	std::condition_variable m_ocv; // For current exclusive owner
-
-	static bool op_lock_shared(u32& ctrl)
-	{
-		// Check writer flags and reader limit
-		return (ctrl & ~SM_READER_QUEUE) < SM_READER_MAX ? ctrl++, true : false;
-	}
-
-	static bool op_lock_excl(u32& ctrl)
-	{
-		// Test and set writer lock
-		return (ctrl & SM_WRITER_LOCK) == 0 ? ctrl |= SM_WRITER_LOCK, true : false;
-	}
-
-	void impl_lock_shared(u32 old_ctrl);
-	void impl_unlock_shared(u32 new_ctrl);
-	void impl_lock_excl(u32 ctrl);
-	void impl_unlock_excl(u32 ctrl);
-
-public:
-	shared_mutex() = default;
-
-	// Lock in shared mode
-	void lock_shared()
-	{
-		const u32 old_ctrl = m_ctrl++;
-
-		// Check flags and reader limit
-		if (old_ctrl >= SM_READER_MAX)
-		{
-			impl_lock_shared(old_ctrl);
-		}
-	}
-
-	// Try to lock in shared mode
-	bool try_lock_shared()
-	{
-		return atomic_op(m_ctrl, [](u32& ctrl)
-		{
-			// Check flags and reader limit
-			return ctrl < SM_READER_MAX ? ctrl++, true : false;
-		});
-	}
-
-	// Unlock in shared mode
-	void unlock_shared()
-	{
-		const u32 new_ctrl = --m_ctrl;
-
-		// Check if notification required
-		if (new_ctrl >= SM_READER_MAX)
-		{
-			impl_unlock_shared(new_ctrl);
-		}
-	}
-
-	// Lock exclusively
-	void lock()
-	{
-		u32 value = 0;
-
-		if (!m_ctrl.compare_exchange_strong(value, SM_WRITER_LOCK))
-		{
-			impl_lock_excl(value);
-		}
-	}
-
-	// Try to lock exclusively
-	bool try_lock()
-	{
-		u32 value = 0;
-
-		return m_ctrl.compare_exchange_strong(value, SM_WRITER_LOCK);
-	}
-
-	// Unlock exclusively
-	void unlock()
-	{
-		const u32 value = m_ctrl.fetch_add(SM_WRITER_LOCK);
-
-		// Check if notification required
-		if (value != SM_WRITER_LOCK)
-		{
-			impl_unlock_excl(value);
-		}
-	}
-};
-
-//! Simplified shared (reader) lock implementation, similar to std::lock_guard.
-//! std::shared_lock may be used instead if necessary.
-class reader_lock final
-{
-	shared_mutex& m_mutex;
-
-public:
-	reader_lock(const reader_lock&) = delete;
-
-	reader_lock(shared_mutex& mutex)
-		: m_mutex(mutex)
-	{
-		m_mutex.lock_shared();
-	}
-
-	~reader_lock()
-	{
-		m_mutex.unlock_shared();
-	}
-};

Utilities/SleepQueue.cpp

@@ -1,55 +0,0 @@
-#include "stdafx.h"
-#include "Emu/CPU/CPUThread.h"
-
-#include "SleepQueue.h"
-
-void sleep_queue_entry_t::add_entry()
-{
-	m_queue.emplace_back(std::static_pointer_cast<CPUThread>(m_thread.shared_from_this()));
-}
-
-void sleep_queue_entry_t::remove_entry()
-{
-	for (auto it = m_queue.begin(); it != m_queue.end(); it++)
-	{
-		if (it->get() == &m_thread)
-		{
-			m_queue.erase(it);
-			return;
-		}
-	}
-}
-
-bool sleep_queue_entry_t::find() const
-{
-	for (auto it = m_queue.begin(); it != m_queue.end(); it++)
-	{
-		if (it->get() == &m_thread)
-		{
-			return true;
-		}
-	}
-
-	return false;
-}
-
-sleep_queue_entry_t::sleep_queue_entry_t(sleep_entry_t& cpu, sleep_queue_t& queue)
-	: m_thread(cpu)
-	, m_queue(queue)
-{
-	add_entry();
-	cpu.sleep();
-}
-
-sleep_queue_entry_t::sleep_queue_entry_t(sleep_entry_t& cpu, sleep_queue_t& queue, const defer_sleep_t&)
-	: m_thread(cpu)
-	, m_queue(queue)
-{
-	cpu.sleep();
-}
-
-sleep_queue_entry_t::~sleep_queue_entry_t()
-{
-	remove_entry();
-	m_thread.awake();
-}

Utilities/SleepQueue.h

@@ -1,45 +0,0 @@
-#pragma once
-
-using sleep_entry_t = class CPUThread;
-using sleep_queue_t = std::deque<std::shared_ptr<sleep_entry_t>>;
-
-static struct defer_sleep_t {} const defer_sleep{};
-
-// automatic object handling a thread entry in the sleep queue
-class sleep_queue_entry_t final
-{
-	sleep_entry_t& m_thread;
-	sleep_queue_t& m_queue;
-
-	void add_entry();
-	void remove_entry();
-	bool find() const;
-
-public:
-	// add specified thread to the sleep queue
-	sleep_queue_entry_t(sleep_entry_t& entry, sleep_queue_t& queue);
-
-	// don't add specified thread to the sleep queue
-	sleep_queue_entry_t(sleep_entry_t& entry, sleep_queue_t& queue, const defer_sleep_t&);
-
-	// removes specified thread from the sleep queue if added
-	~sleep_queue_entry_t();
-
-	// add thread to the sleep queue
-	void enter()
-	{
-		add_entry();
-	}
-
-	// remove thread from the sleep queue
-	void leave()
-	{
-		remove_entry();
-	}
-
-	// check whether the thread exists in the sleep queue
-	explicit operator bool() const
-	{
-		return find();
-	}
-};

Utilities/StrFmt.cpp

@@ -1,84 +1,297 @@
-#include "stdafx.h"
-#pragma warning(push)
-#pragma message("TODO: remove wx dependency: <wx/string.h>")
-#pragma warning(disable : 4996)
-#include <wx/string.h>
-#pragma warning(pop)
+#include "StrFmt.h"
+#include "BEType.h"
+#include "StrUtil.h"
+#include "cfmt.h"
 
-std::string v128::to_hex() const
+#include <algorithm>
+
+#ifdef _WIN32
+#include <Windows.h>
+#else
+#include <errno.h>
+#endif
+
+template <>
+void fmt_class_string<std::pair<const fmt_type_info*, u64>>::format(std::string& out, u64 arg)
 {
-	return fmt::format("%016llx%016llx", _u64[1], _u64[0]);
+	// Dynamic format arg
+	const auto& pair = get_object(arg);
+
+	if (pair.first)
+	{
+		pair.first->fmt_string(out, pair.second);
+	}
 }
 
-std::string v128::to_xyzw() const
+void fmt_class_string<const void*>::format(std::string& out, u64 arg)
 {
-	return fmt::format("x: %g y: %g z: %g w: %g", _f[3], _f[2], _f[1], _f[0]);
+	if (arg)
+	{
+		fmt::append(out, "%p", reinterpret_cast<const void*>(static_cast<std::uintptr_t>(arg)));
+	}
+	else
+	{
+		out += "(NULL)";		
+	}
 }
 
-std::string fmt::to_hex(u64 value, u64 count)
+void fmt_class_string<const char*>::format(std::string& out, u64 arg)
 {
-	if (count - 1 >= 16)
+	if (arg)
 	{
-		throw EXCEPTION("Invalid count: 0x%llx", count);
+		out += reinterpret_cast<const char*>(static_cast<std::uintptr_t>(arg));
 	}
-
-	count = std::max<u64>(count, 16 - cntlz64(value) / 4);
-
-	char res[16] = {};
-
-	for (size_t i = count - 1; ~i; i--, value /= 16)
+	else
 	{
-		res[i] = "0123456789abcdef"[value % 16];
+		out += "(NULL)";
 	}
-
-	return std::string(res, count);
 }
 
-std::string fmt::to_udec(u64 value)
+template <>
+void fmt_class_string<std::string>::format(std::string& out, u64 arg)
 {
-	char res[20] = {};
-	size_t first = sizeof(res);
-
-	if (!value)
-	{
-		res[--first] = '0';
-	}
-
-	for (; value; value /= 10)
-	{
-		res[--first] = '0' + (value % 10);
-	}
-
-	return std::string(&res[first], sizeof(res) - first);
+	out += get_object(arg).c_str(); // TODO?
 }
 
-std::string fmt::to_sdec(s64 svalue)
+template <>
+void fmt_class_string<std::vector<char>>::format(std::string& out, u64 arg)
 {
-	const bool sign = svalue < 0;
-	u64 value = sign ? -svalue : svalue;
-
-	char res[20] = {};
-	size_t first = sizeof(res);
-
-	if (!value)
-	{
-		res[--first] = '0';
-	}
-
-	for (; value; value /= 10)
-	{
-		res[--first] = '0' + (value % 10);
-	}
-
-	if (sign)
-	{
-		res[--first] = '-';
-	}
-
-	return std::string(&res[first], sizeof(res) - first);
+	const std::vector<char>& obj = get_object(arg);
+	out.append(obj.cbegin(), obj.cend());
 }
 
-//extern const std::string fmt::placeholder = "???";
+template <>
+void fmt_class_string<char>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#hhx", static_cast<char>(arg));
+}
+
+template <>
+void fmt_class_string<uchar>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#hhx", static_cast<uchar>(arg));
+}
+
+template <>
+void fmt_class_string<schar>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#hhx", static_cast<schar>(arg));
+}
+
+template <>
+void fmt_class_string<short>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#hx", static_cast<short>(arg));
+}
+
+template <>
+void fmt_class_string<ushort>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#hx", static_cast<ushort>(arg));
+}
+
+template <>
+void fmt_class_string<int>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#x", static_cast<int>(arg));
+}
+
+template <>
+void fmt_class_string<uint>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#x", static_cast<uint>(arg));
+}
+
+template <>
+void fmt_class_string<long>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#lx", static_cast<long>(arg));
+}
+
+template <>
+void fmt_class_string<ulong>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#lx", static_cast<ulong>(arg));
+}
+
+template <>
+void fmt_class_string<llong>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#llx", static_cast<llong>(arg));
+}
+
+template <>
+void fmt_class_string<ullong>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%#llx", static_cast<ullong>(arg));
+}
+
+template <>
+void fmt_class_string<float>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%gf", static_cast<float>(reinterpret_cast<f64&>(arg)));
+}
+
+template <>
+void fmt_class_string<double>::format(std::string& out, u64 arg)
+{
+	fmt::append(out, "%g", reinterpret_cast<f64&>(arg));
+}
+
+template <>
+void fmt_class_string<bool>::format(std::string& out, u64 arg)
+{
+	out += arg ? "true" : "false";
+}
+
+template <>
+void fmt_class_string<v128>::format(std::string& out, u64 arg)
+{
+	const v128& vec = get_object(arg);
+	fmt::append(out, "0x%016llx%016llx", vec._u64[1], vec._u64[0]);
+}
+
+namespace fmt
+{
+	void raw_error(const char* msg)
+	{
+		throw std::runtime_error{msg};
+	}
+
+	void raw_verify_error(const char* msg, const fmt_type_info* sup, u64 arg)
+	{
+		std::string out{"Verification failed"};
+
+		// Print error code (may be irrelevant)
+#ifdef _WIN32
+		if (DWORD error = GetLastError())
+		{
+			fmt::append(out, " (e=%#x)", error);
+		}
+#else
+		if (int error = errno)
+		{
+			fmt::append(out, " (e=%d)", error);
+		}
+#endif
+
+		if (sup)
+		{
+			out += " (";
+			sup->fmt_string(out, arg); // Print value
+			out += ")";
+		}
+
+		if (msg)
+		{
+			out += ": ";
+			out += msg;
+		}
+
+		throw std::runtime_error{out};
+	}
+
+	void raw_narrow_error(const char* msg, const fmt_type_info* sup, u64 arg)
+	{
+		std::string out{"Narrow error"};
+
+		if (sup)
+		{
+			out += " (";
+			sup->fmt_string(out, arg); // Print value
+			out += ")";
+		}
+
+		if (msg)
+		{
+			out += ": ";
+			out += msg;
+		}
+
+		throw std::range_error{out};
+	}
+
+	// Hidden template
+	template <typename T>
+	void raw_throw_exception(const char* fmt, const fmt_type_info* sup, const u64* args)
+	{
+		std::string out;
+		raw_append(out, fmt, sup, args);
+		throw T{out};
+	}
+
+	// Explicit instantiations (not exhaustive)
+	template void raw_throw_exception<std::runtime_error>(const char*, const fmt_type_info*, const u64*);
+	template void raw_throw_exception<std::logic_error>(const char*, const fmt_type_info*, const u64*);
+	template void raw_throw_exception<std::domain_error>(const char*, const fmt_type_info*, const u64*);
+	template void raw_throw_exception<std::invalid_argument>(const char*, const fmt_type_info*, const u64*);
+	template void raw_throw_exception<std::out_of_range>(const char*, const fmt_type_info*, const u64*);
+	template void raw_throw_exception<std::range_error>(const char*, const fmt_type_info*, const u64*);
+	template void raw_throw_exception<std::overflow_error>(const char*, const fmt_type_info*, const u64*);
+	template void raw_throw_exception<std::underflow_error>(const char*, const fmt_type_info*, const u64*);
+
+	struct cfmt_src;
+}
+
+// Temporary implementation
+struct fmt::cfmt_src
+{
+	const fmt_type_info* sup;
+	const u64* args;
+
+	bool test(std::size_t index) const
+	{
+		if (!sup[index].fmt_string)
+		{
+			return false;
+		}
+
+		return true;
+	}
+
+	template <typename T>
+	T get(std::size_t index) const
+	{
+		return *reinterpret_cast<const T*>(reinterpret_cast<const u8*>(args + index));
+	}
+
+	void skip(std::size_t extra)
+	{
+		++sup += extra;
+		++args += extra;
+	}
+
+	std::size_t fmt_string(std::string& out, std::size_t extra) const
+	{
+		const std::size_t start = out.size();
+		sup[extra].fmt_string(out, args[extra]);
+		return out.size() - start;
+	}
+
+	// Returns type size (0 if unknown, pointer, unsigned, assumed max)
+	std::size_t type(std::size_t extra) const
+	{
+// Hack: use known function pointers to determine type
+#define TYPE(type) \
+	if (sup[extra].fmt_string == &fmt_class_string<type>::format) return sizeof(type);
+
+		TYPE(int);
+		TYPE(llong);
+		TYPE(schar);
+		TYPE(short);
+		if (std::is_signed<char>::value) TYPE(char);
+		TYPE(long);
+
+#undef TYPE
+
+		return 0;
+	}
+};
+
+void fmt::raw_append(std::string& out, const char* fmt, const fmt_type_info* sup, const u64* args) noexcept
+{
+	cfmt_append(out, fmt, cfmt_src{sup, args});
+}
 
 std::string fmt::replace_first(const std::string& src, const std::string& from, const std::string& to)
 {
@@ -92,7 +305,7 @@ std::string fmt::replace_first(const std::string& src, const std::string& from,
 	return (pos ? src.substr(0, pos) + to : to) + std::string(src.c_str() + pos + from.length());
 }
 
-std::string fmt::replace_all(const std::string &src, const std::string& from, const std::string& to)
+std::string fmt::replace_all(const std::string& src, const std::string& from, const std::string& to)
 {
 	std::string target = src;
 	for (auto pos = target.find(from); pos != std::string::npos; pos = target.find(from, pos + 1))
@@ -104,83 +317,6 @@ std::string fmt::replace_all(const std::string &src, const std::string& from, co
 	return target;
 }
 
-//TODO: move this wx Stuff somewhere else
-//convert a wxString to a std::string encoded in utf8
-//CAUTION, only use this to interface with wxWidgets classes
-std::string fmt::ToUTF8(const wxString& right)
-{
-	auto ret = std::string(((const char *)right.utf8_str()));
-	return ret;
-}
-
-//convert a std::string encoded in utf8 to a wxString
-//CAUTION, only use this to interface with wxWidgets classes
-wxString fmt::FromUTF8(const std::string& right)
-{
-	auto ret = wxString::FromUTF8(right.c_str());
-	return ret;
-}
-
-//TODO: remove this after every snippet that uses it is gone
-//WARNING: not fully compatible with CmpNoCase from wxString
-int fmt::CmpNoCase(const std::string& a, const std::string& b)
-{
-	if (a.length() != b.length())
-	{
-		return -1;
-	}
-	else
-	{
-		return std::equal(a.begin(),
-			a.end(),
-			b.begin(),
-			[](const char& a, const char& b){return ::tolower(a) == ::tolower(b); })
-			? 0 : -1;
-	}
-}
-
-//TODO: remove this after every snippet that uses it is gone
-//WARNING: not fully compatible with CmpNoCase from wxString
-void fmt::Replace(std::string &str, const std::string &searchterm, const std::string& replaceterm)
-{
-	size_t cursor = 0;
-	do
-	{
-		cursor = str.find(searchterm, cursor);
-		if (cursor != std::string::npos)
-		{
-			str.replace(cursor, searchterm.size(), replaceterm);
-			cursor += replaceterm.size();
-		}
-		else
-		{
-			break;
-		}
-	} while (true);
-}
-
-std::vector<std::string> fmt::rSplit(const std::string& source, const std::string& delim)
-{
-	std::vector<std::string> ret;
-	size_t cursor = 0;
-	do
-	{
-		size_t prevcurs = cursor;
-		cursor = source.find(delim, cursor);
-		if (cursor != std::string::npos)
-		{
-			ret.push_back(source.substr(prevcurs,cursor-prevcurs));
-			cursor += delim.size();
-		}
-		else
-		{
-			ret.push_back(source.substr(prevcurs));
-			break;
-		}
-	} while (true);
-	return ret;
-}
-
 std::vector<std::string> fmt::split(const std::string& source, std::initializer_list<std::string> separators, bool is_skip_empty)
 {
 	std::vector<std::string> result;
@@ -189,7 +325,7 @@ std::vector<std::string> fmt::split(const std::string& source, std::initializer_
 
 	for (size_t cursor_end = 0; cursor_end < source.length(); ++cursor_end)
 	{
-		for (auto &separator : separators)
+		for (auto& separator : separators)
 		{
 			if (strncmp(source.c_str() + cursor_end, separator.c_str(), separator.length()) == 0)
 			{
@@ -198,7 +334,7 @@ std::vector<std::string> fmt::split(const std::string& source, std::initializer_
 					result.push_back(candidate);
 
 				cursor_begin = cursor_end + separator.length();
-				cursor_end = cursor_begin - 1;
+				cursor_end   = cursor_begin - 1;
 				break;
 			}
 		}
@@ -217,50 +353,20 @@ std::string fmt::trim(const std::string& source, const std::string& values)
 	std::size_t begin = source.find_first_not_of(values);
 
 	if (begin == source.npos)
-		return{};
+		return {};
 
 	return source.substr(begin, source.find_last_not_of(values) + 1);
 }
 
-std::string fmt::tolower(std::string source)
+std::string fmt::to_upper(const std::string& string)
 {
-	std::transform(source.begin(), source.end(), source.begin(), ::tolower);
-
-	return source;
+	std::string result;
+	result.resize(string.size());
+	std::transform(string.begin(), string.end(), result.begin(), ::toupper);
+	return result;
 }
 
-std::string fmt::toupper(std::string source)
-{
-	std::transform(source.begin(), source.end(), source.begin(), ::toupper);
-
-	return source;
-}
-
-std::string fmt::escape(std::string source)
-{
-	const std::pair<std::string, std::string> escape_list[] =
-	{
-		{ "\\", "\\\\" },
-		{ "\a", "\\a" },
-		{ "\b", "\\b" },
-		{ "\f", "\\f" },
-		{ "\n", "\\n\n" },
-		{ "\r", "\\r" },
-		{ "\t", "\\t" },
-		{ "\v", "\\v" },
-	};
-
-	source = fmt::replace_all(source, escape_list);
-
-	for (char c = 0; c < 32; c++)
-	{
-		if (c != '\n') source = fmt::replace_all(source, std::string(1, c), fmt::format("\\x%02X", c));
-	}
-
-	return source;
-}
-
-bool fmt::match(const std::string &source, const std::string &mask)
+bool fmt::match(const std::string& source, const std::string& mask)
 {
 	std::size_t source_position = 0, mask_position = 0;
 

Utilities/StrFmt.h

@@ -1,359 +1,265 @@
 #pragma once
 
-class wxString;
+#include "types.h"
 
-#if defined(_MSC_VER) && _MSC_VER <= 1800
-#define snprintf _snprintf
-#endif
+#include <exception>
+#include <stdexcept>
+#include <string>
 
 namespace fmt
 {
-	//struct empty_t{};
+	template <typename... Args>
+	static std::string format(const char*, const Args&...);
+}
 
-	//extern const std::string placeholder;
+template <typename T, typename>
+struct fmt_unveil
+{
+	static_assert(sizeof(T) > 0, "fmt_unveil<>: cannot pass forward-declared object");
 
-	template <typename T>
-	std::string AfterLast(const std::string& source, T searchstr)
+	using type = T;
+
+	static inline u64 get(const T& arg)
 	{
-		size_t search_pos = source.rfind(searchstr);
-		search_pos = search_pos == std::string::npos ? 0 : search_pos;
-		return source.substr(search_pos);
+		return reinterpret_cast<std::uintptr_t>(&arg);
 	}
 
-	template <typename T>
-	std::string BeforeLast(const std::string& source, T searchstr)
+	// Temporary value container (can possibly be created by other fmt_unveil<> specializations)
+	struct u64_wrapper
 	{
-		size_t search_pos = source.rfind(searchstr);
-		search_pos = search_pos == std::string::npos ? 0 : search_pos;
-		return source.substr(0, search_pos);
-	}
+		T arg;
 
-	template <typename T>
-	std::string AfterFirst(const std::string& source, T searchstr)
-	{
-		size_t search_pos = source.find(searchstr);
-		search_pos = search_pos == std::string::npos ? 0 : search_pos;
-		return source.substr(search_pos);
-	}
-
-	template <typename T>
-	std::string BeforeFirst(const std::string& source, T searchstr)
-	{
-		size_t search_pos = source.find(searchstr);
-		search_pos = search_pos == std::string::npos ? 0 : search_pos;
-		return source.substr(0, search_pos);
-	}
-
-	// write `fmt` from `pos` to the first occurence of `fmt::placeholder` to
-	// the stream `os`. Then write `arg` to to the stream. If there's no
-	// `fmt::placeholder` after `pos` everything in `fmt` after pos is written
-	// to `os`. Then `arg` is written to `os` after appending a space character
-	//template<typename T>
-	//empty_t write(const std::string &fmt, std::ostream &os, std::string::size_type &pos, T &&arg)
-	//{
-	//	std::string::size_type ins = fmt.find(placeholder, pos);
-
-	//	if (ins == std::string::npos)
-	//	{
-	//		os.write(fmt.data() + pos, fmt.size() - pos);
-	//		os << ' ' << arg;
-
-	//		pos = fmt.size();
-	//	}
-	//	else
-	//	{
-	//		os.write(fmt.data() + pos, ins - pos);
-	//		os << arg;
-
-	//		pos = ins + placeholder.size();
-	//	}
-	//	return{};
-	//}
-
-	// typesafe version of a sprintf-like function. Returns the printed to
-	// string. To mark positions where the arguments are supposed to be
-	// inserted use `fmt::placeholder`. If there's not enough placeholders
-	// the rest of the arguments are appended at the end, seperated by spaces
-	//template<typename  ... Args>
-	//std::string SFormat(const std::string &fmt, Args&& ... parameters)
-	//{
-	//	std::ostringstream os;
-	//	std::string::size_type pos = 0;
-	//	std::initializer_list<empty_t> { write(fmt, os, pos, parameters)... };
-
-	//	if (!fmt.empty())
-	//	{
-	//		os.write(fmt.data() + pos, fmt.size() - pos);
-	//	}
-
-	//	std::string result = os.str();
-	//	return result;
-	//}
-
-	std::string replace_first(const std::string& src, const std::string& from, const std::string& to);
-	std::string replace_all(const std::string &src, const std::string& from, const std::string& to);
-
-	template<size_t list_size>
-	std::string replace_all(std::string src, const std::pair<std::string, std::string>(&list)[list_size])
-	{
-		for (size_t pos = 0; pos < src.length(); ++pos)
+		// Allow implicit conversion
+		operator u64() const
 		{
-			for (size_t i = 0; i < list_size; ++i)
+			return reinterpret_cast<std::uintptr_t>(&arg);
+		}
+	};
+
+	// This overload resolution takes the precedence
+	static inline u64_wrapper get(T&& arg)
+	{
+		return u64_wrapper{std::move(arg)};
+	}
+};
+
+template <typename T>
+struct fmt_unveil<T, std::enable_if_t<std::is_integral<T>::value && sizeof(T) <= 8 && alignof(T) <= 8>>
+{
+	using type = T;
+
+	static inline u64 get(T arg)
+	{
+		return static_cast<T>(arg);
+	}
+};
+
+template <typename T>
+struct fmt_unveil<T, std::enable_if_t<std::is_floating_point<T>::value && sizeof(T) <= 8 && alignof(T) <= 8>>
+{
+	using type = T;
+
+	// Convert FP to f64 and reinterpret (TODO?)
+	static inline u64 get(const f64 arg)
+	{
+		return *reinterpret_cast<const u64*>(reinterpret_cast<const u8*>(&arg));
+	}
+};
+
+template <typename T>
+struct fmt_unveil<T, std::enable_if_t<std::is_enum<T>::value>>
+{
+	using type = T;
+
+	static inline u64 get(T arg)
+	{
+		return static_cast<std::underlying_type_t<T>>(arg);
+	}
+};
+
+template <typename T>
+struct fmt_unveil<T*, void>
+{
+	using type = const T*;
+
+	static inline u64 get(const T* arg)
+	{
+		return reinterpret_cast<std::uintptr_t>(arg);
+	}
+};
+
+template <typename T, std::size_t N>
+struct fmt_unveil<T[N], void>
+{
+	using type = const T*;
+
+	static inline u64 get(const T* arg)
+	{
+		return reinterpret_cast<std::uintptr_t>(arg);
+	}
+};
+
+template <>
+struct fmt_unveil<b8, void>
+{
+	using type = bool;
+
+	static inline u64 get(const b8& value)
+	{
+		return fmt_unveil<bool>::get(value);
+	}
+};
+
+// String type format provider, also type classifier (format() called if an argument is formatted as "%s")
+template <typename T, typename = void>
+struct fmt_class_string
+{
+	// Formatting function (must be explicitly specialized)
+	static void format(std::string& out, u64 arg);
+
+	// Helper typedef (visible in format())
+	using type = T;
+
+	// Helper function (converts arg to object reference)
+	static SAFE_BUFFERS FORCE_INLINE const T& get_object(u64 arg)
+	{
+		return *reinterpret_cast<const T*>(static_cast<std::uintptr_t>(arg));
+	}
+
+	// Helper function (safely converts arg to enum value)
+	static SAFE_BUFFERS FORCE_INLINE void format_enum(std::string& out, u64 arg, const char* (*get)(T value))
+	{
+		const auto value = static_cast<std::underlying_type_t<T>>(arg);
+
+		// Check narrowing
+		if (static_cast<u64>(value) == arg)
+		{
+			if (const char* str = get(static_cast<T>(value)))
 			{
-				const size_t comp_length = list[i].first.length();
+				out += str;
+				return;
+			}
+		}
 
-				if (src.length() - pos < comp_length)
-					continue;
+		// Fallback to underlying type formatting
+		fmt_class_string<std::underlying_type_t<T>>::format(out, static_cast<u64>(value));
+	}
 
-				if (src.substr(pos, comp_length) == list[i].first)
+	// Helper function (bitset formatting)
+	static SAFE_BUFFERS FORCE_INLINE void format_bitset(std::string& out, u64 arg, const char* prefix, const char* delim, const char* suffix, void (*fmt)(std::string&, u64))
+	{
+		// Start from raw value
+		fmt_class_string<u64>::format(out, arg);
+
+		out += prefix;
+
+		for (u64 i = 0; i < 63; i++)
+		{
+			const u64 mask = 1ull << i;
+
+			if (arg & mask)
+			{
+				fmt(out, i);
+
+				if (arg >> (i + 1))
 				{
-					src = (pos ? src.substr(0, pos) + list[i].second : list[i].second) + src.substr(pos + comp_length);
-					pos += list[i].second.length() - 1;
-					break;
+					out += delim;
 				}
 			}
 		}
 
-		return src;
+		if (arg & (1ull << 63))
+		{
+			fmt(out, 63);
+		}
+
+		out += suffix;
 	}
 
-	template<size_t list_size>
-	std::string replace_all(std::string src, const std::pair<std::string, std::function<std::string()>>(&list)[list_size])
+	// Helper constant (may be used in format_enum as lambda return value)
+	static constexpr const char* unknown = nullptr;
+};
+
+template <>
+struct fmt_class_string<const void*, void>
+{
+	static void format(std::string& out, u64 arg);
+};
+
+template <typename T>
+struct fmt_class_string<T*, void> : fmt_class_string<const void*, void>
+{
+	// Classify all pointers as const void*
+};
+
+template <>
+struct fmt_class_string<const char*, void>
+{
+	static void format(std::string& out, u64 arg);
+};
+
+template <>
+struct fmt_class_string<char*, void> : fmt_class_string<const char*>
+{
+	// Classify char* as const char*
+};
+
+struct fmt_type_info
+{
+	decltype(&fmt_class_string<int>::format) fmt_string;
+
+	template <typename T>
+	static constexpr fmt_type_info make()
 	{
-		for (size_t pos = 0; pos < src.length(); ++pos)
+		return fmt_type_info
 		{
-			for (size_t i = 0; i < list_size; ++i)
-			{
-				const size_t comp_length = list[i].first.length();
-
-				if (src.length() - pos < comp_length)
-					continue;
-
-				if (src.substr(pos, comp_length) == list[i].first)
-				{
-					src = (pos ? src.substr(0, pos) + list[i].second() : list[i].second()) + src.substr(pos + comp_length);
-					pos += list[i].second().length() - 1;
-					break;
-				}
-			}
-		}
-
-		return src;
+			&fmt_class_string<T>::format,
+		};
 	}
+};
 
-	std::string to_hex(u64 value, u64 count = 1);
-	std::string to_udec(u64 value);
-	std::string to_sdec(s64 value);
+// Argument array type (each element generated via fmt_unveil<>)
+template <typename... Args>
+using fmt_args_t = const u64(&&)[sizeof...(Args) + 1];
 
-	template<typename T, bool is_enum = std::is_enum<T>::value> struct unveil
+namespace fmt
+{
+	template <typename... Args>
+	SAFE_BUFFERS FORCE_INLINE const fmt_type_info* get_type_info()
 	{
-		using result_type = T;
-
-		force_inline static result_type get_value(const T& arg)
-		{
-			return arg;
-		}
-	};
-
-	template<> struct unveil<const char*, false>
-	{
-		using result_type = const char* const;
-
-		force_inline static result_type get_value(const char* const& arg)
-		{
-			return arg;
-		}
-	};
-
-	template<std::size_t N> struct unveil<char[N], false>
-	{
-		using result_type = const char* const;
-
-		force_inline static result_type get_value(const char(&arg)[N])
-		{
-			return arg;
-		}
-	};
-
-	template<> struct unveil<std::string, false>
-	{
-		using result_type = const char*;
-
-		force_inline static result_type get_value(const std::string& arg)
-		{
-			return arg.c_str();
-		}
-	};
-
-	template<typename T> struct unveil<T, true>
-	{
-		using result_type = std::underlying_type_t<T>;
-
-		force_inline static result_type get_value(const T& arg)
-		{
-			return static_cast<result_type>(arg);
-		}
-	};
-
-	template<typename T, bool Se> struct unveil<se_t<T, Se>, false>
-	{
-		using result_type = typename unveil<T>::result_type;
-
-		force_inline static result_type get_value(const se_t<T, Se>& arg)
-		{
-			return unveil<T>::get_value(arg);
-		}
-	};
-
-	template<typename T>
-	force_inline typename unveil<T>::result_type do_unveil(const T& arg)
-	{
-		return unveil<T>::get_value(arg);
-	}
-
-	// Formatting function with special functionality:
-	//
-	// std::string is forced to .c_str()
-	// be_t<> is forced to .value() (fmt::do_unveil reverts byte order automatically)
-	//
-	// External specializations for fmt::do_unveil (can be found in another headers):
-	// vm::ptr, vm::bptr, ... (fmt::do_unveil) (vm_ptr.h) (with appropriate address type, using .addr() can be avoided)
-	// vm::ref, vm::bref, ... (fmt::do_unveil) (vm_ref.h)
-	//
-	template<typename... Args>
-	safe_buffers std::string format(const char* fmt, const Args&... args)
-	{
-		// fixed stack buffer for the first attempt
-		std::array<char, 4096> fixed_buf;
-
-		// possibly dynamically allocated buffer for the second attempt
-		std::unique_ptr<char[]> buf;
-
-		// pointer to the current buffer
-		char* buf_addr = fixed_buf.data();
-
-		for (std::size_t buf_size = fixed_buf.size();;)
-		{
-#ifndef _MSC_VER
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wformat-security"
-#endif
-			const std::size_t len = std::snprintf(buf_addr, buf_size, fmt, do_unveil(args)...);
-#ifndef _MSC_VER
-#pragma GCC diagnostic pop
-#endif
-			if (len > INT_MAX)
-			{
-				throw std::runtime_error("std::snprintf() failed");
-			}
-
-			if (len < buf_size)
-			{
-				return{ buf_addr, len };
-			}
-
-			buf.reset(buf_addr = new char[buf_size = len + 1]);
-		}
-	}
-
-	struct exception : public std::exception
-	{
-		std::unique_ptr<char[]> message;
-
-		template<typename... Args> never_inline safe_buffers exception(const char* file, int line, const char* func, const char* text, Args... args) noexcept
-		{
-			const std::string data = format(text, args...) + format("\n(in file %s:%d, in function %s)", file, line, func);
-
-			message.reset(new char[data.size() + 1]);
-
-			std::memcpy(message.get(), data.c_str(), data.size() + 1);
-		}
-
-		exception(const exception& other) noexcept
-		{
-			const std::size_t size = std::strlen(other.message.get());
-
-			message.reset(new char[size + 1]);
-
-			std::memcpy(message.get(), other.message.get(), size + 1);
-		}
-
-		virtual const char* what() const noexcept override
-		{
-			return message.get();
-		}
-	};
-
-	//convert a wxString to a std::string encoded in utf8
-	//CAUTION, only use this to interface with wxWidgets classes
-	std::string ToUTF8(const wxString& right);
-
-	//convert a std::string encoded in utf8 to a wxString
-	//CAUTION, only use this to interface with wxWidgets classes
-	wxString FromUTF8(const std::string& right);
-
-	//TODO: remove this after every snippet that uses it is gone
-	//WARNING: not fully compatible with CmpNoCase from wxString
-	int CmpNoCase(const std::string& a, const std::string& b);
-
-	//TODO: remove this after every snippet that uses it is gone
-	//WARNING: not fully compatible with Replace from wxString
-	void Replace(std::string &str, const std::string &searchterm, const std::string& replaceterm);
-
-	std::vector<std::string> rSplit(const std::string& source, const std::string& delim);
-
-	std::vector<std::string> split(const std::string& source, std::initializer_list<std::string> separators, bool is_skip_empty = true);
-	std::string trim(const std::string& source, const std::string& values = " \t");
-
-	template<typename T>
-	std::string merge(const T& source, const std::string& separator)
-	{
-		if (!source.size())
-		{
-			return{};
-		}
-
-		std::string result;
-
-		auto it = source.begin();
-		auto end = source.end();
-		for (--end; it != end; ++it)
-		{
-			result += *it + separator;
-		}
-
-		return result + source.back();
-	}
-
-	template<typename T>
-	std::string merge(std::initializer_list<T> sources, const std::string& separator)
-	{
-		if (!sources.size())
-		{
-			return{};
-		}
-
-		std::string result;
-		bool first = true;
-
-		for (auto &v : sources)
-		{
-			if (first)
-			{
-				result = fmt::merge(v, separator);
-				first = false;
-			}
-			else
-			{
-				result += separator + fmt::merge(v, separator);
-			}
-		}
+		// Constantly initialized null-terminated list of type-specific information
+		static constexpr fmt_type_info result[sizeof...(Args) + 1]{fmt_type_info::make<Args>()...};
 
 		return result;
 	}
 
-	std::string tolower(std::string source);
-	std::string toupper(std::string source);
-	std::string escape(std::string source);
-	bool match(const std::string &source, const std::string &mask);
+	// Internal formatting function
+	void raw_append(std::string& out, const char*, const fmt_type_info*, const u64*) noexcept;
+
+	// Formatting function
+	template <typename... Args>
+	SAFE_BUFFERS FORCE_INLINE void append(std::string& out, const char* fmt, const Args&... args)
+	{
+		raw_append(out, fmt, fmt::get_type_info<fmt_unveil_t<Args>...>(), fmt_args_t<Args...>{fmt_unveil<Args>::get(args)...});
+	}
+
+	// Formatting function
+	template <typename... Args>
+	SAFE_BUFFERS FORCE_INLINE std::string format(const char* fmt, const Args&... args)
+	{
+		std::string result;
+		append<Args...>(result, fmt, args...);
+		return result;
+	}
+
+	// Internal exception message formatting template, must be explicitly specialized or instantiated in cpp to minimize code bloat
+	template <typename T>
+	[[noreturn]] void raw_throw_exception(const char*, const fmt_type_info*, const u64*);
+
+	// Throw exception with formatting
+	template <typename T = std::runtime_error, typename... Args>
+	[[noreturn]] SAFE_BUFFERS FORCE_INLINE void throw_exception(const char* fmt, const Args&... args)
+	{
+		raw_throw_exception<T>(fmt, fmt::get_type_info<fmt_unveil_t<Args>...>(), fmt_args_t<Args...>{fmt_unveil<Args>::get(args)...});
+	}
 }

Utilities/StrUtil.h

@@ -0,0 +1,133 @@
+#pragma once
+
+#include <cstdlib>
+#include <cstring>
+#include <string>
+#include <vector>
+#include <functional>
+
+// Copy null-terminated string from std::string to char array with truncation
+template <std::size_t N>
+inline void strcpy_trunc(char (&dst)[N], const std::string& src)
+{
+	const std::size_t count = src.size() >= N ? N - 1 : src.size();
+	std::memcpy(dst, src.c_str(), count);
+	dst[count] = '\0';
+}
+
+// Copy null-terminated string from char array to another char array with truncation
+template <std::size_t N, std::size_t N2>
+inline void strcpy_trunc(char (&dst)[N], const char (&src)[N2])
+{
+	const std::size_t count = N2 >= N ? N - 1 : N2;
+	std::memcpy(dst, src, count);
+	dst[count] = '\0';
+}
+
+namespace fmt
+{
+	std::string replace_first(const std::string& src, const std::string& from, const std::string& to);
+	std::string replace_all(const std::string& src, const std::string& from, const std::string& to);
+
+	template <size_t list_size>
+	std::string replace_all(std::string src, const std::pair<std::string, std::string> (&list)[list_size])
+	{
+		for (size_t pos = 0; pos < src.length(); ++pos)
+		{
+			for (size_t i = 0; i < list_size; ++i)
+			{
+				const size_t comp_length = list[i].first.length();
+
+				if (src.length() - pos < comp_length)
+					continue;
+
+				if (src.substr(pos, comp_length) == list[i].first)
+				{
+					src = (pos ? src.substr(0, pos) + list[i].second : list[i].second) + src.substr(pos + comp_length);
+					pos += list[i].second.length() - 1;
+					break;
+				}
+			}
+		}
+
+		return src;
+	}
+
+	template <size_t list_size>
+	std::string replace_all(std::string src, const std::pair<std::string, std::function<std::string()>> (&list)[list_size])
+	{
+		for (size_t pos = 0; pos < src.length(); ++pos)
+		{
+			for (size_t i = 0; i < list_size; ++i)
+			{
+				const size_t comp_length = list[i].first.length();
+
+				if (src.length() - pos < comp_length)
+					continue;
+
+				if (src.substr(pos, comp_length) == list[i].first)
+				{
+					src = (pos ? src.substr(0, pos) + list[i].second() : list[i].second()) + src.substr(pos + comp_length);
+					pos += list[i].second().length() - 1;
+					break;
+				}
+			}
+		}
+
+		return src;
+	}
+
+	std::vector<std::string> split(const std::string& source, std::initializer_list<std::string> separators, bool is_skip_empty = true);
+	std::string trim(const std::string& source, const std::string& values = " \t");
+
+	template <typename T>
+	std::string merge(const T& source, const std::string& separator)
+	{
+		if (!source.size())
+		{
+			return {};
+		}
+
+		std::string result;
+
+		auto it  = source.begin();
+		auto end = source.end();
+		for (--end; it != end; ++it)
+		{
+			result += *it + separator;
+		}
+
+		return result + source.back();
+	}
+
+	template <typename T>
+	std::string merge(std::initializer_list<T> sources, const std::string& separator)
+	{
+		if (!sources.size())
+		{
+			return {};
+		}
+
+		std::string result;
+		bool first = true;
+
+		for (auto& v : sources)
+		{
+			if (first)
+			{
+				result = fmt::merge(v, separator);
+				first  = false;
+			}
+			else
+			{
+				result += separator + fmt::merge(v, separator);
+			}
+		}
+
+		return result;
+	}
+
+	std::string to_upper(const std::string& string);
+
+	bool match(const std::string& source, const std::string& mask);
+}

Utilities/Thread.h

@@ -1,510 +1,317 @@
 #pragma once
 
+#include "types.h"
+#include "Atomic.h"
+
+#include <exception>
+#include <string>
+#include <memory>
+
+#include "sema.h"
+#include "cond.h"
+
 // Will report exception and call std::abort() if put in catch(...)
 [[noreturn]] void catch_all_exceptions();
 
-// Thread control class
-class thread_ctrl final
+// Simple list of void() functors
+class task_stack
 {
-	static thread_local thread_ctrl* g_tls_this_thread;
+	struct task_base
+	{
+		std::unique_ptr<task_base> next;
 
-	// Name getter
-	std::function<std::string()> m_name;
+		virtual ~task_base();
 
-	// Thread handle (be careful)
-	std::thread m_thread;
+		virtual void invoke()
+		{
+			if (next)
+			{
+				next->invoke();
+			}
+		}
+	};
 
-	// Thread result
-	std::future<void> m_future;
+	template <typename F>
+	struct task_type final : task_base
+	{
+		std::remove_reference_t<F> func;
 
-	// Functions scheduled at thread exit
-	std::deque<std::function<void()>> m_atexit;
+		task_type(F&& func)
+			: func(std::forward<F>(func))
+		{
+		}
 
-	// Called at the thread start
-	static void initialize();
+		void invoke() final override
+		{
+			func();
+			task_base::invoke();
+		}
+	};
 
-	// Called at the thread end
-	static void finalize() noexcept;
+	std::unique_ptr<task_base> m_stack;
 
 public:
-	template<typename T>
-	thread_ctrl(T&& name)
-		: m_name(std::forward<T>(name))
+	task_stack() = default;
+
+	template <typename F>
+	task_stack(F&& func)
+		: m_stack(new task_type<F>(std::forward<F>(func)))
 	{
 	}
 
-	// Disable copy/move constructors and operators
+	void push(task_stack stack)
+	{
+		auto _top = stack.m_stack.release();
+		auto _next = m_stack.release();
+		m_stack.reset(_top);
+		while (UNLIKELY(_top->next)) _top = _top->next.get();
+		_top->next.reset(_next);
+	}
+
+	void reset()
+	{
+		m_stack.reset();
+	}
+
+	void invoke() const
+	{
+		if (m_stack)
+		{
+			m_stack->invoke();
+		}
+	}
+};
+
+// Thread control class
+class thread_ctrl final
+{
+	// Current thread
+	static thread_local thread_ctrl* g_tls_this_thread;
+
+	// Self pointer
+	std::shared_ptr<thread_ctrl> m_self;
+
+	// Thread handle (platform-specific)
+	atomic_t<std::uintptr_t> m_thread{0};
+
+	// Thread mutex
+	mutable semaphore<> m_mutex;
+
+	// Thread condition variable
+	cond_variable m_cond;
+
+	// Thread flags
+	atomic_t<u32> m_signal{0};
+
+	// Thread joining condition variable
+	cond_variable m_jcv;
+
+	// Remotely set or caught exception
+	std::exception_ptr m_exception;
+
+	// Thread initial task or atexit task
+	task_stack m_task;
+
+	// Fixed name
+	std::string m_name;
+
+	// Start thread
+	static void start(const std::shared_ptr<thread_ctrl>&, task_stack);
+
+	// Called at the thread start
+	void initialize();
+
+	// Called at the thread end
+	void finalize(std::exception_ptr) noexcept;
+
+	// Add task (atexit)
+	static void _push(task_stack);
+
+	// Internal waiting function, may throw. Infinite value is -1.
+	static bool _wait_for(u64 usec);
+
+	// Internal throwing function. Mutex must be locked and will be unlocked.
+	[[noreturn]] void _throw();
+
+	// Internal notification function
+	void _notify(cond_variable thread_ctrl::*);
+
+public:
+	thread_ctrl(std::string&& name);
+
 	thread_ctrl(const thread_ctrl&) = delete;
 
 	~thread_ctrl();
 
 	// Get thread name
-	std::string get_name() const;
-
-	// Get future result (may throw)
-	void join()
+	const std::string& get_name() const
 	{
-		return m_future.get();
+		return m_name;
 	}
 
+	// Get exception
+	std::exception_ptr get_exception() const;
+
+	// Set exception
+	void set_exception(std::exception_ptr ptr);
+
+	// Get thread result (may throw, simultaneous joining allowed)
+	void join();
+
+	// Notify the thread
+	void notify();
+
+	// Wait once with timeout. Abortable, may throw. May spuriously return false.
+	static inline bool wait_for(u64 usec)
+	{
+		return _wait_for(usec);
+	}
+
+	// Wait. Abortable, may throw.
+	static inline void wait()
+	{
+		_wait_for(-1);
+	}
+
+	// Wait until pred(). Abortable, may throw.
+	template<typename F, typename RT = std::result_of_t<F()>>
+	static inline RT wait(F&& pred)
+	{
+		while (true)
+		{
+			if (RT result = pred())
+			{
+				return result;
+			}
+
+			_wait_for(-1);
+		}
+	}
+
+	// Wait eternally until aborted.
+	[[noreturn]] static inline void eternalize()
+	{
+		while (true)
+		{
+			_wait_for(-1);
+		}
+	}
+
+	// Test exception (may throw).
+	static void test();
+
 	// Get current thread (may be nullptr)
-	static const thread_ctrl* get_current()
+	static thread_ctrl* get_current()
 	{
 		return g_tls_this_thread;
 	}
 
 	// Register function at thread exit (for the current thread)
-	template<typename T>
-	static inline void at_exit(T&& func)
+	template<typename F>
+	static inline void atexit(F&& func)
 	{
-		CHECK_ASSERTION(g_tls_this_thread);
+		_push(std::forward<F>(func));
+	}
 
-		g_tls_this_thread->m_atexit.emplace_front(std::forward<T>(func));
+	// Create detached named thread
+	template<typename N, typename F>
+	static inline void spawn(N&& name, F&& func)
+	{
+		auto out = std::make_shared<thread_ctrl>(std::forward<N>(name));
+
+		thread_ctrl::start(out, std::forward<F>(func));
 	}
 
 	// Named thread factory
 	template<typename N, typename F>
-	static inline std::shared_ptr<thread_ctrl> spawn(N&& name, F&& func)
+	static inline void spawn(std::shared_ptr<thread_ctrl>& out, N&& name, F&& func)
 	{
-		auto ctrl = std::make_shared<thread_ctrl>(std::forward<N>(name));
+		out = std::make_shared<thread_ctrl>(std::forward<N>(name));
 
-		std::promise<void> promise;
-
-		ctrl->m_future = promise.get_future();
-
-		ctrl->m_thread = std::thread([ctrl, task = std::forward<F>(func)](std::promise<void> promise)
-		{
-			g_tls_this_thread = ctrl.get();
-
-			try
-			{
-				initialize();
-				task();
-				finalize();
-				promise.set_value();
-			}
-			catch (...)
-			{
-				finalize();
-				promise.set_exception(std::current_exception());
-			}
-
-		}, std::move(promise));
-
-		return ctrl;
+		thread_ctrl::start(out, std::forward<F>(func));
 	}
 };
 
-class named_thread_t : public std::enable_shared_from_this<named_thread_t>
+class named_thread
 {
 	// Pointer to managed resource (shared with actual thread)
 	std::shared_ptr<thread_ctrl> m_thread;
 
 public:
-	// Thread condition variable for external use (this thread waits on it, other threads may notify)
-	std::condition_variable cv;
+	named_thread();
 
-	// Thread mutex for external use (can be used with `cv`)
-	std::mutex mutex;
+	virtual ~named_thread();
+
+	// Deleted copy/move constructors + copy/move operators
+	named_thread(const named_thread&) = delete;
+
+	// Get thread name
+	virtual std::string get_name() const;
 
 protected:
+	// Start thread (cannot be called from the constructor: should throw in such case)
+	void start_thread(const std::shared_ptr<void>& _this);
+
 	// Thread task (called in the thread)
 	virtual void on_task() = 0;
 
 	// Thread finalization (called after on_task)
 	virtual void on_exit() {}
 
-	// ID initialization (called through id_aux_initialize)
-	virtual void on_id_aux_initialize() { start(); }
-
-	// ID finalization (called through id_aux_finalize)
-	virtual void on_id_aux_finalize() { join(); }
-
 public:
-	named_thread_t() = default;
+	// ID initialization
+	virtual void on_init(const std::shared_ptr<void>& _this)
+	{
+		return start_thread(_this);
+	}
 
-	virtual ~named_thread_t() = default;
+	// ID finalization
+	virtual void on_stop()
+	{
+		m_thread->join();
+	}
 
-	// Deleted copy/move constructors + copy/move operators
-	named_thread_t(const named_thread_t&) = delete;
+	// Access thread_ctrl
+	thread_ctrl* get() const
+	{
+		return m_thread.get();
+	}
 
-	// Get thread name
-	virtual std::string get_name() const;
+	void join() const
+	{
+		return m_thread->join();
+	}
 
-	// Start thread (cannot be called from the constructor: should throw bad_weak_ptr in such case)
-	void start();
-
-	// Join thread (get future result)
-	void join();
-
-	// Check whether the thread is not in "empty state"
-	bool is_started() const { return m_thread.operator bool(); }
-
-	// Compare with the current thread
-	bool is_current() const { CHECK_ASSERTION(m_thread); return thread_ctrl::get_current() == m_thread.get(); }
-
-	// Get thread_ctrl
-	const thread_ctrl* get_thread_ctrl() const { return m_thread.get(); }
-
-	friend void id_aux_initialize(named_thread_t* ptr) { ptr->on_id_aux_initialize(); }
-	friend void id_aux_finalize(named_thread_t* ptr) { ptr->on_id_aux_finalize(); }
+	void notify() const
+	{
+		return m_thread->notify();
+	}
 };
 
 // Wrapper for named thread, joins automatically in the destructor, can only be used in function scope
-class scope_thread_t final
+class scope_thread final
 {
 	std::shared_ptr<thread_ctrl> m_thread;
 
 public:
 	template<typename N, typename F>
-	scope_thread_t(N&& name, F&& func)
-		: m_thread(thread_ctrl::spawn(std::forward<N>(name), std::forward<F>(func)))
+	scope_thread(N&& name, F&& func)
 	{
+		thread_ctrl::spawn(m_thread, std::forward<N>(name), std::forward<F>(func));
 	}
 
 	// Deleted copy/move constructors + copy/move operators
-	scope_thread_t(const scope_thread_t&) = delete;
+	scope_thread(const scope_thread&) = delete;
 
 	// Destructor with exceptions allowed
-	~scope_thread_t() noexcept(false)
+	~scope_thread() noexcept(false)
 	{
 		m_thread->join();
 	}
-};
 
-extern const std::function<bool()> SQUEUE_ALWAYS_EXIT;
-extern const std::function<bool()> SQUEUE_NEVER_EXIT;
-
-bool squeue_test_exit();
-
-template<typename T, u32 sq_size = 256>
-class squeue_t
-{
-	struct squeue_sync_var_t
+	// Access thread_ctrl
+	thread_ctrl* get() const
 	{
-		struct
-		{
-			u32 position : 31;
-			u32 pop_lock : 1;
-		};
-		struct
-		{
-			u32 count : 31;
-			u32 push_lock : 1;
-		};
-	};
-
-	atomic_t<squeue_sync_var_t> m_sync;
-
-	mutable std::mutex m_rcv_mutex;
-	mutable std::mutex m_wcv_mutex;
-	mutable std::condition_variable m_rcv;
-	mutable std::condition_variable m_wcv;
-
-	T m_data[sq_size];
-
-	enum squeue_sync_var_result : u32
-	{
-		SQSVR_OK = 0,
-		SQSVR_LOCKED = 1,
-		SQSVR_FAILED = 2,
-	};
-
-public:
-	squeue_t()
-		: m_sync(squeue_sync_var_t{})
-	{
-	}
-
-	u32 get_max_size() const
-	{
-		return sq_size;
-	}
-
-	bool is_full() const
-	{
-		return m_sync.load().count == sq_size;
-	}
-
-	bool push(const T& data, const std::function<bool()>& test_exit)
-	{
-		u32 pos = 0;
-
-		while (u32 res = m_sync.atomic_op([&pos](squeue_sync_var_t& sync) -> u32
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-
-			if (sync.push_lock)
-			{
-				return SQSVR_LOCKED;
-			}
-			if (sync.count == sq_size)
-			{
-				return SQSVR_FAILED;
-			}
-
-			sync.push_lock = 1;
-			pos = sync.position + sync.count;
-			return SQSVR_OK;
-		}))
-		{
-			if (res == SQSVR_FAILED && (test_exit() || squeue_test_exit()))
-			{
-				return false;
-			}
-
-			std::unique_lock<std::mutex> wcv_lock(m_wcv_mutex);
-			m_wcv.wait_for(wcv_lock, std::chrono::milliseconds(1));
-		}
-
-		m_data[pos >= sq_size ? pos - sq_size : pos] = data;
-
-		m_sync.atomic_op([](squeue_sync_var_t& sync)
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-			assert(sync.push_lock);
-			sync.push_lock = 0;
-			sync.count++;
-		});
-
-		m_rcv.notify_one();
-		m_wcv.notify_one();
-		return true;
-	}
-
-	bool push(const T& data, const volatile bool* do_exit)
-	{
-		return push(data, [do_exit](){ return do_exit && *do_exit; });
-	}
-
-	force_inline bool push(const T& data)
-	{
-		return push(data, SQUEUE_NEVER_EXIT);
-	}
-
-	force_inline bool try_push(const T& data)
-	{
-		return push(data, SQUEUE_ALWAYS_EXIT);
-	}
-
-	bool pop(T& data, const std::function<bool()>& test_exit)
-	{
-		u32 pos = 0;
-
-		while (u32 res = m_sync.atomic_op([&pos](squeue_sync_var_t& sync) -> u32
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-
-			if (!sync.count)
-			{
-				return SQSVR_FAILED;
-			}
-			if (sync.pop_lock)
-			{
-				return SQSVR_LOCKED;
-			}
-
-			sync.pop_lock = 1;
-			pos = sync.position;
-			return SQSVR_OK;
-		}))
-		{
-			if (res == SQSVR_FAILED && (test_exit() || squeue_test_exit()))
-			{
-				return false;
-			}
-
-			std::unique_lock<std::mutex> rcv_lock(m_rcv_mutex);
-			m_rcv.wait_for(rcv_lock, std::chrono::milliseconds(1));
-		}
-
-		data = m_data[pos];
-
-		m_sync.atomic_op([](squeue_sync_var_t& sync)
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-			assert(sync.pop_lock);
-			sync.pop_lock = 0;
-			sync.position++;
-			sync.count--;
-			if (sync.position == sq_size)
-			{
-				sync.position = 0;
-			}
-		});
-
-		m_rcv.notify_one();
-		m_wcv.notify_one();
-		return true;
-	}
-
-	bool pop(T& data, const volatile bool* do_exit)
-	{
-		return pop(data, [do_exit](){ return do_exit && *do_exit; });
-	}
-
-	force_inline bool pop(T& data)
-	{
-		return pop(data, SQUEUE_NEVER_EXIT);
-	}
-
-	force_inline bool try_pop(T& data)
-	{
-		return pop(data, SQUEUE_ALWAYS_EXIT);
-	}
-
-	bool peek(T& data, u32 start_pos, const std::function<bool()>& test_exit)
-	{
-		assert(start_pos < sq_size);
-		u32 pos = 0;
-
-		while (u32 res = m_sync.atomic_op([&pos, start_pos](squeue_sync_var_t& sync) -> u32
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-
-			if (sync.count <= start_pos)
-			{
-				return SQSVR_FAILED;
-			}
-			if (sync.pop_lock)
-			{
-				return SQSVR_LOCKED;
-			}
-
-			sync.pop_lock = 1;
-			pos = sync.position + start_pos;
-			return SQSVR_OK;
-		}))
-		{
-			if (res == SQSVR_FAILED && (test_exit() || squeue_test_exit()))
-			{
-				return false;
-			}
-
-			std::unique_lock<std::mutex> rcv_lock(m_rcv_mutex);
-			m_rcv.wait_for(rcv_lock, std::chrono::milliseconds(1));
-		}
-
-		data = m_data[pos >= sq_size ? pos - sq_size : pos];
-
-		m_sync.atomic_op([](squeue_sync_var_t& sync)
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-			assert(sync.pop_lock);
-			sync.pop_lock = 0;
-		});
-
-		m_rcv.notify_one();
-		return true;
-	}
-
-	bool peek(T& data, u32 start_pos, const volatile bool* do_exit)
-	{
-		return peek(data, start_pos, [do_exit](){ return do_exit && *do_exit; });
-	}
-
-	force_inline bool peek(T& data, u32 start_pos = 0)
-	{
-		return peek(data, start_pos, SQUEUE_NEVER_EXIT);
-	}
-
-	force_inline bool try_peek(T& data, u32 start_pos = 0)
-	{
-		return peek(data, start_pos, SQUEUE_ALWAYS_EXIT);
-	}
-
-	class squeue_data_t
-	{
-		T* const m_data;
-		const u32 m_pos;
-		const u32 m_count;
-
-		squeue_data_t(T* data, u32 pos, u32 count)
-			: m_data(data)
-			, m_pos(pos)
-			, m_count(count)
-		{
-		}
-
-	public:
-		T& operator [] (u32 index)
-		{
-			assert(index < m_count);
-			index += m_pos;
-			index = index < sq_size ? index : index - sq_size;
-			return m_data[index];
-		}
-	};
-
-	void process(void(*proc)(squeue_data_t data))
-	{
-		u32 pos, count;
-
-		while (m_sync.atomic_op([&pos, &count](squeue_sync_var_t& sync) -> u32
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-
-			if (sync.pop_lock || sync.push_lock)
-			{
-				return SQSVR_LOCKED;
-			}
-
-			pos = sync.position;
-			count = sync.count;
-			sync.pop_lock = 1;
-			sync.push_lock = 1;
-			return SQSVR_OK;
-		}))
-		{
-			std::unique_lock<std::mutex> rcv_lock(m_rcv_mutex);
-			m_rcv.wait_for(rcv_lock, std::chrono::milliseconds(1));
-		}
-
-		proc(squeue_data_t(m_data, pos, count));
-
-		m_sync.atomic_op([](squeue_sync_var_t& sync)
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-			assert(sync.pop_lock && sync.push_lock);
-			sync.pop_lock = 0;
-			sync.push_lock = 0;
-		});
-
-		m_wcv.notify_one();
-		m_rcv.notify_one();
-	}
-
-	void clear()
-	{
-		while (m_sync.atomic_op([](squeue_sync_var_t& sync) -> u32
-		{
-			assert(sync.count <= sq_size);
-			assert(sync.position < sq_size);
-
-			if (sync.pop_lock || sync.push_lock)
-			{
-				return SQSVR_LOCKED;
-			}
-
-			sync.pop_lock = 1;
-			sync.push_lock = 1;
-			return SQSVR_OK;
-		}))
-		{
-			std::unique_lock<std::mutex> rcv_lock(m_rcv_mutex);
-			m_rcv.wait_for(rcv_lock, std::chrono::milliseconds(1));
-		}
-
-		m_sync.exchange({});
-		m_wcv.notify_one();
-		m_rcv.notify_one();
+		return m_thread.get();
 	}
 };

Utilities/Timer.h

@@ -6,8 +6,8 @@ class Timer
 {
 private:
 	bool m_stopped;
-	std::chrono::high_resolution_clock::time_point m_start;
-	std::chrono::high_resolution_clock::time_point m_end;
+	std::chrono::steady_clock::time_point m_start;
+	std::chrono::steady_clock::time_point m_end;
 
 public:
 	Timer() : m_stopped(false)
@@ -17,13 +17,13 @@ public:
 	void Start()
 	{
 		m_stopped = false;
-		m_start = std::chrono::high_resolution_clock::now();
+		m_start = std::chrono::steady_clock::now();
 	}
 
 	void Stop()
 	{
 		m_stopped = true;
-		m_end = std::chrono::high_resolution_clock::now();
+		m_end = std::chrono::steady_clock::now();
 	}
 
 	double GetElapsedTimeInSec() const
@@ -38,14 +38,14 @@ public:
 
 	u64 GetElapsedTimeInMicroSec() const
 	{
-		std::chrono::high_resolution_clock::time_point now = m_stopped ? m_end : std::chrono::high_resolution_clock::now();
+		std::chrono::steady_clock::time_point now = m_stopped ? m_end : std::chrono::steady_clock::now();
 
 		return std::chrono::duration_cast<std::chrono::microseconds>(now - m_start).count();
 	}
 
 	u64 GetElapsedTimeInNanoSec() const
 	{
-		std::chrono::high_resolution_clock::time_point now = m_stopped ? m_end : std::chrono::high_resolution_clock::now();
+		std::chrono::steady_clock::time_point now = m_stopped ? m_end : std::chrono::steady_clock::now();
 
 		return std::chrono::duration_cast<std::chrono::nanoseconds>(now - m_start).count();
 	}

Utilities/VirtualMemory.cpp

@@ -13,33 +13,30 @@
 
 namespace memory_helper
 {
-	void* reserve_memory(size_t size)
+	void* reserve_memory(std::size_t size)
 	{
 #ifdef _WIN32
-		void* ret = VirtualAlloc(NULL, size, MEM_RESERVE, PAGE_NOACCESS);
-		CHECK_ASSERTION(ret != NULL);
+		return verify("reserve_memory" HERE, VirtualAlloc(NULL, size, MEM_RESERVE, PAGE_NOACCESS));
 #else
-		void* ret = mmap(nullptr, size, PROT_NONE, MAP_ANON | MAP_PRIVATE, -1, 0);
-		CHECK_ASSERTION(ret != 0);
-#endif
-		return ret;
-	}
-
-	void commit_page_memory(void* pointer, size_t page_size)
-	{
-#ifdef _WIN32
-		CHECK_ASSERTION(VirtualAlloc((u8*)pointer, page_size, MEM_COMMIT, PAGE_READWRITE) != NULL);
-#else
-		CHECK_ASSERTION(mprotect((u8*)pointer, page_size, PROT_READ | PROT_WRITE) != -1);
+		return verify("reserve_memory" HERE, ::mmap(nullptr, size, PROT_NONE, MAP_ANON | MAP_PRIVATE, -1, 0));
 #endif
 	}
 
-	void free_reserved_memory(void* pointer, size_t size)
+	void commit_page_memory(void* pointer, std::size_t size)
 	{
 #ifdef _WIN32
-		CHECK_ASSERTION(VirtualFree(pointer, 0, MEM_RELEASE) != 0);
+		verify(HERE), VirtualAlloc(pointer, size, MEM_COMMIT, PAGE_READWRITE);
 #else
-		CHECK_ASSERTION(munmap(pointer, size) == 0);
+		verify(HERE), ::mprotect((void*)((u64)pointer & -4096), ::align(size, 4096), PROT_READ | PROT_WRITE) != -1;
+#endif
+	}
+
+	void free_reserved_memory(void* pointer, std::size_t size)
+	{
+#ifdef _WIN32
+		verify(HERE), VirtualFree(pointer, 0, MEM_DECOMMIT);
+#else
+		verify(HERE), ::mmap(pointer, size, PROT_NONE, MAP_FIXED | MAP_ANON | MAP_PRIVATE, -1, 0);
 #endif
 	}
 }

Utilities/VirtualMemory.h

@@ -3,20 +3,20 @@
 namespace memory_helper
 {
 	/**
-	* Reserve size bytes of virtual memory and returns it.
+	* Reserve `size` bytes of virtual memory and returns it.
 	* The memory should be commited before usage.
 	*/
-	void* reserve_memory(size_t size);
+	void* reserve_memory(std::size_t size);
 
 	/**
-	* Commit page_size bytes of virtual memory starting at pointer.
+	* Commit `size` bytes of virtual memory starting at pointer.
 	* That is, bake reserved memory with physical memory.
 	* pointer should belong to a range of reserved memory.
 	*/
-	void commit_page_memory(void* pointer, size_t page_size);
+	void commit_page_memory(void* pointer, std::size_t size);
 
 	/**
-	* Free memory alloced via reserve_memory.
+	* Decommit all memory committed via commit_page_memory.
 	*/
-	void free_reserved_memory(void* pointer, size_t size);
-}
+	void free_reserved_memory(void* pointer, std::size_t size);
+}

Utilities/bit_set.h

@@ -0,0 +1,732 @@
+#pragma once
+
+/*
+This header helps to extend scoped enum types (enum class) in two possible ways:
+1) Enabling bitwise operators for enums
+2) Advanced bs_t<> template (this converts enum type to another "bitset" enum type)
+
+To enable bitwise operators, enum scope must contain `__bitwise_ops` entry.
+
+enum class flags
+{
+	__bitwise_ops, // Not essential, but recommended to put it first
+
+	flag1 = 1 << 0,
+	flag2 = 1 << 1,
+};
+
+Examples:
+`flags::flag1 | flags::flag2` - bitwise OR
+`flags::flag1 & flags::flag2` - bitwise AND
+`flags::flag1 ^ flags::flag2` - bitwise XOR
+`~flags::flag1` - bitwise NEG
+
+To enable bs_t<> template, enum scope must contain `__bitset_enum_max` entry.
+
+enum class flagzz : u32
+{
+	flag1, // Bit indices start from zero
+	flag2,
+
+	__bitset_enum_max // It must be the last value
+};
+
+Now some operators are enabled for two enum types: `flagzz` and `bs_t<flagzz>`.
+These are very different from previously described bitwise operators.
+
+Examples:
+`+flagzz::flag1` - unary `+` operator convert flagzz value to bs_t<flagzz>
+`flagzz::flag1 + flagzz::flag2` - bitset union
+`flagzz::flag1 - flagzz::flag2` - bitset difference
+Intersection (&) and symmetric difference (^) is also available.
+
+*/
+
+#include "types.h"
+
+// Helper template
+template<typename T>
+struct bs_base
+{
+	// Underlying type
+	using under = std::underlying_type_t<T>;
+
+	// Actual bitset type
+	enum class type : under
+	{
+		null = 0, // Empty bitset
+
+		__bitset_set_type = 0 // SFINAE marker
+	};
+
+	static constexpr std::size_t bitmax = sizeof(T) * 8;
+	static constexpr std::size_t bitsize = static_cast<under>(T::__bitset_enum_max);
+
+	static_assert(std::is_enum<T>::value, "bs_t<> error: invalid type (must be enum)");
+	static_assert(!bitsize || bitsize <= bitmax, "bs_t<> error: invalid __bitset_enum_max");
+
+	// Helper function
+	static constexpr under shift(T value)
+	{
+		return static_cast<under>(1) << static_cast<under>(value);
+	}
+
+	friend type& operator +=(type& lhs, type rhs)
+	{
+		reinterpret_cast<under&>(lhs) |= static_cast<under>(rhs);
+		return lhs;
+	}
+
+	friend type& operator -=(type& lhs, type rhs)
+	{
+		reinterpret_cast<under&>(lhs) &= ~static_cast<under>(rhs);
+		return lhs;
+	}
+
+	friend type& operator &=(type& lhs, type rhs)
+	{
+		reinterpret_cast<under&>(lhs) &= static_cast<under>(rhs);
+		return lhs;
+	}
+
+	friend type& operator ^=(type& lhs, type rhs)
+	{
+		reinterpret_cast<under&>(lhs) ^= static_cast<under>(rhs);
+		return lhs;
+	}
+
+	friend type& operator +=(type& lhs, T rhs)
+	{
+		reinterpret_cast<under&>(lhs) |= shift(rhs);
+		return lhs;
+	}
+
+	friend type& operator -=(type& lhs, T rhs)
+	{
+		reinterpret_cast<under&>(lhs) &= ~shift(rhs);
+		return lhs;
+	}
+
+	friend type& operator &=(type& lhs, T rhs)
+	{
+		reinterpret_cast<under&>(lhs) &= shift(rhs);
+		return lhs;
+	}
+
+	friend type& operator ^=(type& lhs, T rhs)
+	{
+		reinterpret_cast<under&>(lhs) ^= shift(rhs);
+		return lhs;
+	}
+
+	friend constexpr type operator +(type lhs, type rhs)
+	{
+		return static_cast<type>(static_cast<under>(lhs) | static_cast<under>(rhs));
+	}
+
+	friend constexpr type operator -(type lhs, type rhs)
+	{
+		return static_cast<type>(static_cast<under>(lhs) & ~static_cast<under>(rhs));
+	}
+
+	friend constexpr type operator &(type lhs, type rhs)
+	{
+		return static_cast<type>(static_cast<under>(lhs) & static_cast<under>(rhs));
+	}
+
+	friend constexpr type operator ^(type lhs, type rhs)
+	{
+		return static_cast<type>(static_cast<under>(lhs) ^ static_cast<under>(rhs));
+	}
+
+	friend constexpr type operator &(type lhs, T rhs)
+	{
+		return static_cast<type>(static_cast<under>(lhs) & shift(rhs));
+	}
+
+	friend constexpr type operator ^(type lhs, T rhs)
+	{
+		return static_cast<type>(static_cast<under>(lhs) ^ shift(rhs));
+	}
+
+	friend constexpr type operator &(T lhs, type rhs)
+	{
+		return static_cast<type>(shift(lhs) & static_cast<under>(rhs));
+	}
+
+	friend constexpr type operator ^(T lhs, type rhs)
+	{
+		return static_cast<type>(shift(lhs) ^ static_cast<under>(rhs));
+	}
+
+	friend constexpr bool operator ==(T lhs, type rhs)
+	{
+		return shift(lhs) == rhs;
+	}
+
+	friend constexpr bool operator ==(type lhs, T rhs)
+	{
+		return lhs == shift(rhs);
+	}
+
+	friend constexpr bool operator !=(T lhs, type rhs)
+	{
+		return shift(lhs) != rhs;
+	}
+
+	friend constexpr bool operator !=(type lhs, T rhs)
+	{
+		return lhs != shift(rhs);
+	}
+
+	friend constexpr bool test(type value)
+	{
+		return static_cast<under>(value) != 0;
+	}
+
+	friend constexpr bool test(type lhs, type rhs)
+	{
+		return (static_cast<under>(lhs) & static_cast<under>(rhs)) != 0;
+	}
+
+	friend constexpr bool test(type lhs, T rhs)
+	{
+		return (static_cast<under>(lhs) & shift(rhs)) != 0;
+	}
+
+	friend constexpr bool test(T lhs, type rhs)
+	{
+		return (shift(lhs) & static_cast<under>(rhs)) != 0;
+	}
+
+	friend bool test_and_set(type& lhs, type rhs)
+	{
+		return test_and_set(reinterpret_cast<under&>(lhs), static_cast<under>(rhs));
+	}
+
+	friend bool test_and_set(type& lhs, T rhs)
+	{
+		return test_and_set(reinterpret_cast<under&>(lhs), shift(rhs));
+	}
+
+	friend bool test_and_reset(type& lhs, type rhs)
+	{
+		return test_and_reset(reinterpret_cast<under&>(lhs), static_cast<under>(rhs));
+	}
+
+	friend bool test_and_reset(type& lhs, T rhs)
+	{
+		return test_and_reset(reinterpret_cast<under&>(lhs), shift(rhs));
+	}
+
+	friend bool test_and_complement(type& lhs, type rhs)
+	{
+		return test_and_complement(reinterpret_cast<under&>(lhs), static_cast<under>(rhs));
+	}
+
+	friend bool test_and_complement(type& lhs, T rhs)
+	{
+		return test_and_complement(reinterpret_cast<under&>(lhs), shift(rhs));
+	}
+};
+
+// Bitset type for enum class with available bits [0, T::__bitset_enum_max)
+template<typename T>
+using bs_t = typename bs_base<T>::type;
+
+// Unary '+' operator: promote plain enum value to bitset value
+template<typename T, typename = decltype(T::__bitset_enum_max)>
+constexpr bs_t<T> operator +(T value)
+{
+	return static_cast<bs_t<T>>(bs_base<T>::shift(value));
+}
+
+// Binary '+' operator: bitset union
+template<typename T, typename = decltype(T::__bitset_enum_max)>
+constexpr bs_t<T> operator +(T lhs, T rhs)
+{
+	return static_cast<bs_t<T>>(bs_base<T>::shift(lhs) | bs_base<T>::shift(rhs));
+}
+
+// Binary '+' operator: bitset union
+template<typename T, typename = decltype(T::__bitset_enum_max)>
+constexpr bs_t<T> operator +(typename bs_base<T>::type lhs, T rhs)
+{
+	return static_cast<bs_t<T>>(static_cast<typename bs_base<T>::under>(lhs) | bs_base<T>::shift(rhs));
+}
+
+// Binary '+' operator: bitset union
+template<typename T, typename = decltype(T::__bitset_enum_max)>
+constexpr bs_t<T> operator +(T lhs, typename bs_base<T>::type rhs)
+{
+	return static_cast<bs_t<T>>(bs_base<T>::shift(lhs) | static_cast<typename bs_base<T>::under>(rhs));
+}
+
+// Binary '-' operator: bitset difference
+template<typename T, typename = decltype(T::__bitset_enum_max)>
+constexpr bs_t<T> operator -(T lhs, T rhs)
+{
+	return static_cast<bs_t<T>>(bs_base<T>::shift(lhs) & ~bs_base<T>::shift(rhs));
+}
+
+// Binary '-' operator: bitset difference
+template<typename T, typename = decltype(T::__bitset_enum_max)>
+constexpr bs_t<T> operator -(typename bs_base<T>::type lhs, T rhs)
+{
+	return static_cast<bs_t<T>>(static_cast<typename bs_base<T>::under>(lhs) & ~bs_base<T>::shift(rhs));
+}
+
+// Binary '-' operator: bitset difference
+template<typename T, typename = decltype(T::__bitset_enum_max)>
+constexpr bs_t<T> operator -(T lhs, typename bs_base<T>::type rhs)
+{
+	return static_cast<bs_t<T>>(bs_base<T>::shift(lhs) & ~static_cast<typename bs_base<T>::under>(rhs));
+}
+
+template<typename BS, typename T>
+struct atomic_add<BS, T, void_t<decltype(T::__bitset_enum_max), std::enable_if_t<std::is_same<BS, bs_t<T>>::value>>>
+{
+	using under = typename bs_base<T>::under;
+
+	static inline bs_t<T> op1(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::fetch_or(reinterpret_cast<under&>(left), bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline bs_t<T> op2(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::or_fetch(reinterpret_cast<under&>(left), bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename BS, typename T>
+struct atomic_sub<BS, T, void_t<decltype(T::__bitset_enum_max), std::enable_if_t<std::is_same<BS, bs_t<T>>::value>>>
+{
+	using under = typename bs_base<T>::under;
+
+	static inline bs_t<T> op1(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::fetch_and(reinterpret_cast<under&>(left), ~bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline bs_t<T> op2(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::and_fetch(reinterpret_cast<under&>(left), ~bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename BS, typename T>
+struct atomic_and<BS, T, void_t<decltype(T::__bitset_enum_max), std::enable_if_t<std::is_same<BS, bs_t<T>>::value>>>
+{
+	using under = typename bs_base<T>::under;
+
+	static inline bs_t<T> op1(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::fetch_and(reinterpret_cast<under&>(left), bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline bs_t<T> op2(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::and_fetch(reinterpret_cast<under&>(left), bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename BS, typename T>
+struct atomic_xor<BS, T, void_t<decltype(T::__bitset_enum_max), std::enable_if_t<std::is_same<BS, bs_t<T>>::value>>>
+{
+	using under = typename bs_base<T>::under;
+
+	static inline bs_t<T> op1(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::fetch_xor(reinterpret_cast<under&>(left), bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline bs_t<T> op2(bs_t<T>& left, T right)
+	{
+		return static_cast<bs_t<T>>(atomic_storage<under>::xor_fetch(reinterpret_cast<under&>(left), bs_base<T>::shift(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename T>
+struct atomic_add<T, T, void_t<decltype(T::__bitset_set_type)>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline T op1(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::fetch_or(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline T op2(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::or_fetch(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename T>
+struct atomic_sub<T, T, void_t<decltype(T::__bitset_set_type)>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline T op1(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::fetch_and(reinterpret_cast<under&>(left), ~static_cast<under>(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline T op2(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::and_fetch(reinterpret_cast<under&>(left), ~static_cast<under>(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename T>
+struct atomic_and<T, T, void_t<decltype(T::__bitset_set_type)>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline T op1(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::fetch_and(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline T op2(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::and_fetch(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename T>
+struct atomic_xor<T, T, void_t<decltype(T::__bitset_set_type)>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline T op1(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::fetch_xor(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline T op2(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::xor_fetch(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename BS, typename T>
+struct atomic_test_and_set<BS, T, void_t<decltype(T::__bitset_enum_max), std::enable_if_t<std::is_same<BS, bs_t<T>>::value>>>
+{
+	using under = typename bs_base<T>::under;
+
+	static inline bool _op(bs_t<T>& left, T value)
+	{
+		return atomic_storage<under>::bts(reinterpret_cast<under&>(left), static_cast<uint>(static_cast<under>(value)));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+template<typename BS, typename T>
+struct atomic_test_and_reset<BS, T, void_t<decltype(T::__bitset_enum_max), std::enable_if_t<std::is_same<BS, bs_t<T>>::value>>>
+{
+	using under = typename bs_base<T>::under;
+
+	static inline bool _op(bs_t<T>& left, T value)
+	{
+		return atomic_storage<under>::btr(reinterpret_cast<under&>(left), static_cast<uint>(static_cast<under>(value)));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+template<typename BS, typename T>
+struct atomic_test_and_complement<BS, T, void_t<decltype(T::__bitset_enum_max), std::enable_if_t<std::is_same<BS, bs_t<T>>::value>>>
+{
+	using under = typename bs_base<T>::under;
+
+	static inline bool _op(bs_t<T>& left, T value)
+	{
+		return atomic_storage<under>::btc(reinterpret_cast<under&>(left), static_cast<uint>(static_cast<under>(value)));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+template<typename T>
+struct atomic_test_and_set<T, T, void_t<decltype(T::__bitset_set_type)>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline bool _op(T& left, T value)
+	{
+		return atomic_storage<under>::test_and_set(reinterpret_cast<under&>(left), static_cast<under>(value));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+template<typename T>
+struct atomic_test_and_reset<T, T, void_t<decltype(T::__bitset_set_type)>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline bool _op(T& left, T value)
+	{
+		return atomic_storage<under>::test_and_reset(reinterpret_cast<under&>(left), static_cast<under>(value));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+template<typename T>
+struct atomic_test_and_complement<T, T, void_t<decltype(T::__bitset_set_type)>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline bool _op(T& left, T value)
+	{
+		return atomic_storage<under>::test_and_complement(reinterpret_cast<under&>(left), static_cast<under>(value));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+// Binary '|' operator: bitwise OR
+template<typename T, typename = decltype(T::__bitwise_ops)>
+constexpr T operator |(T lhs, T rhs)
+{
+	return static_cast<T>(std::underlying_type_t<T>(lhs) | std::underlying_type_t<T>(rhs));
+}
+
+// Binary '&' operator: bitwise AND
+template<typename T, typename = decltype(T::__bitwise_ops)>
+constexpr T operator &(T lhs, T rhs)
+{
+	return static_cast<T>(std::underlying_type_t<T>(lhs) & std::underlying_type_t<T>(rhs));
+}
+
+// Binary '^' operator: bitwise XOR
+template<typename T, typename = decltype(T::__bitwise_ops)>
+constexpr T operator ^(T lhs, T rhs)
+{
+	return static_cast<T>(std::underlying_type_t<T>(lhs) ^ std::underlying_type_t<T>(rhs));
+}
+
+// Unary '~' operator: bitwise NEG
+template<typename T, typename = decltype(T::__bitwise_ops)>
+constexpr T operator ~(T value)
+{
+	return static_cast<T>(~std::underlying_type_t<T>(value));
+}
+
+// Bitwise OR assignment
+template<typename T, typename = decltype(T::__bitwise_ops)>
+inline T& operator |=(T& lhs, T rhs)
+{
+	reinterpret_cast<std::underlying_type_t<T>&>(lhs) |= std::underlying_type_t<T>(rhs);
+	return lhs;
+}
+
+// Bitwise AND assignment
+template<typename T, typename = decltype(T::__bitwise_ops)>
+inline T& operator &=(T& lhs, T rhs)
+{
+	reinterpret_cast<std::underlying_type_t<T>&>(lhs) &= std::underlying_type_t<T>(rhs);
+	return lhs;
+}
+
+// Bitwise XOR assignment
+template<typename T, typename = decltype(T::__bitwise_ops)>
+inline T& operator ^=(T& lhs, T rhs)
+{
+	reinterpret_cast<std::underlying_type_t<T>&>(lhs) ^= std::underlying_type_t<T>(rhs);
+	return lhs;
+}
+
+template<typename T, typename = decltype(T::__bitwise_ops)>
+constexpr bool test(T value)
+{
+	return std::underlying_type_t<T>(value) != 0;
+}
+
+template<typename T, typename = decltype(T::__bitwise_ops)>
+constexpr bool test(T lhs, T rhs)
+{
+	return (std::underlying_type_t<T>(lhs) & std::underlying_type_t<T>(rhs)) != 0;
+}
+
+template<typename T, typename = decltype(T::__bitwise_ops)>
+inline bool test_and_set(T& lhs, T rhs)
+{
+	return test_and_set(reinterpret_cast<std::underlying_type_t<T>&>(lhs), std::underlying_type_t<T>(rhs));
+}
+
+template<typename T, typename = decltype(T::__bitwise_ops)>
+inline bool test_and_reset(T& lhs, T rhs)
+{
+	return test_and_reset(reinterpret_cast<std::underlying_type_t<T>&>(lhs), std::underlying_type_t<T>(rhs));
+}
+
+template<typename T, typename = decltype(T::__bitwise_ops)>
+inline bool test_and_complement(T& lhs, T rhs)
+{
+	return test_and_complement(reinterpret_cast<std::underlying_type_t<T>&>(lhs), std::underlying_type_t<T>(rhs));
+}
+
+template<typename T>
+struct atomic_or<T, T, void_t<decltype(T::__bitwise_ops), std::enable_if_t<std::is_enum<T>::value>>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline T op1(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::fetch_or(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline T op2(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::or_fetch(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename T>
+struct atomic_and<T, T, void_t<decltype(T::__bitwise_ops), std::enable_if_t<std::is_enum<T>::value>>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline T op1(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::fetch_and(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline T op2(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::and_fetch(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename T>
+struct atomic_xor<T, T, void_t<decltype(T::__bitwise_ops), std::enable_if_t<std::is_enum<T>::value>>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline T op1(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::fetch_xor(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto fetch_op = &op1;
+
+	static inline T op2(T& left, T right)
+	{
+		return static_cast<T>(atomic_storage<under>::xor_fetch(reinterpret_cast<under&>(left), static_cast<under>(right)));
+	}
+
+	static constexpr auto op_fetch = &op2;
+	static constexpr auto atomic_op = &op2;
+};
+
+template<typename T>
+struct atomic_test_and_set<T, T, void_t<decltype(T::__bitwise_ops), std::enable_if_t<std::is_enum<T>::value>>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline bool _op(T& left, T value)
+	{
+		return atomic_storage<under>::test_and_set(reinterpret_cast<under&>(left), static_cast<under>(value));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+template<typename T>
+struct atomic_test_and_reset<T, T, void_t<decltype(T::__bitwise_ops), std::enable_if_t<std::is_enum<T>::value>>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline bool _op(T& left, T value)
+	{
+		return atomic_storage<under>::test_and_reset(reinterpret_cast<under&>(left), static_cast<under>(value));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};
+
+template<typename T>
+struct atomic_test_and_complement<T, T, void_t<decltype(T::__bitwise_ops), std::enable_if_t<std::is_enum<T>::value>>>
+{
+	using under = std::underlying_type_t<T>;
+
+	static inline bool _op(T& left, T value)
+	{
+		return atomic_storage<under>::test_and_complement(reinterpret_cast<under&>(left), static_cast<under>(value));
+	}
+
+	static constexpr auto fetch_op = &_op;
+	static constexpr auto op_fetch = &_op;
+	static constexpr auto atomic_op = &_op;
+};