Break parts in custom entities and display

CMakeLists.txt

@@ -7,10 +7,11 @@ set(SIMD_LEVEL "avx2" CACHE STRING "SIMD processor acceleration (sse2, sse4.1, a
 option(USE_FMA "Use fma" 1)
 option(LOG_DEBUG "Show debug logs" 0)
 option(LOG_TRACE "Show trace logs" 0)
+option(NATIVE "Build with -march=native" 0)
 
 find_program(CCACHE_FOUND ccache)
 if(CCACHE_FOUND)
-	set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ccache)
+  set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ccache)
 endif(CCACHE_FOUND)
 
 add_subdirectory("deps/glfw")
@@ -29,14 +30,17 @@ add_compile_definitions(FIXED_WINDOW=${FIXED_WINDOW} LOG_DEBUG=${LOG_DEBUG} LOG_
 if(PROFILING)
 add_compile_definitions(TRACY_ENABLE=1)
 endif(PROFILING)
+if(NATIVE)
+  add_definitions(-march=native)
+endif()
 if(SIMD_LEVEL EQUAL "avx2")
-	add_compile_definitions(FN_COMPILE_AVX2=1)
+  add_compile_definitions(FN_COMPILE_AVX2=1)
 elseif(SIMD_LEVEL EQUAL "avx512f")
-	add_compile_definitions(FN_COMPILE_AVX512=1)
+  add_compile_definitions(FN_COMPILE_AVX512=1)
 endif()
 add_definitions(-m${SIMD_LEVEL})
 if(USE_FMA)
-	add_definitions(-mfma)
+  add_definitions(-mfma)
 endif(USE_FMA)
 
 file(GLOB_RECURSE CORE_SOURCES "src/core/*.cpp" "deps/tracy/TracyClient.cpp")
@@ -72,6 +76,13 @@ target_include_directories(univerxel-client PRIVATE ${CORE_HEADERS} ${CLIENT_HEA
 # Resource client files + default zstd.dict
 file(COPY resource/content DESTINATION ${CMAKE_BINARY_DIR})
 
+# Serialize entity model
+file(GLOB_RECURSE MODELS_SOURCES "src/client/render/api/Models.cpp")
+add_executable(generate_models EXCLUDE_FROM_ALL "src/tools/generate_models.cpp" ${MODELS_SOURCES})
+target_compile_features(generate_models PUBLIC cxx_std_17)
+target_link_libraries(generate_models glm::glm_static)
+target_include_directories(generate_models PRIVATE "deps/robin_hood" "deps/meshoptimizer")
+
 # Docs
 add_custom_target(docs
     COMMAND doxygen ${CMAKE_CURRENT_SOURCE_DIR}/Doxyfile

README.md

@@ -101,6 +101,7 @@ CMAKE_BUILD_TYPE | Level of optimization | `Release`
 PROFILING | Tracy profiling | `0`
 LOG_DEBUG | Debug logs | `0`
 LOG_TRACE | Trace logs | `0`
+NATIVE | Optimize for native CPU | `0`
 
 1. Build Make
 ```sh

TODO.md

@@ -34,7 +34,9 @@
   - [x] Transparency
   - [~] Entities
     - [ ] Collide
-    - [ ] Get models
+    - [x] Get models
+    - [ ] Save in world
+    - [ ] Rotate
   - [ ] Reduce compile unit count
   - [~] Review documentation
   - [ ] Clean kick
@@ -81,6 +83,7 @@
   - [x] QUIC protocol
   - [~] 8 for 1 contouring problem
   - [ ] Use in memory protocol (to replace server_handle)
+    - [ ] Remove SharedUniverse
   - [ ] Octree
     - [ ] Better Lod
   - [ ] VK
@@ -95,7 +98,8 @@
     - https://www.youtube.com/watch?v=iikdcAA7cww
     - Toon shading
     - Eye adaptation
-  - [ ] Correct sRGB conversions
+    - GLSL gamma correction
+      - https://github.com/skurmedel/shaders/blob/master/glsl/gamma_correct.glsl
   - [ ] Post processing
     - https://learnopengl.com/Advanced-OpenGL/Anti-Aliasing
     - Bloom

deps/robin_hood/robin_hood.h

@@ -6,7 +6,6 @@
 //                                      _/_____/
 //
 // Fast & memory efficient hashtable based on robin hood hashing for C++11/14/17/20
-// version 3.8.0
 // https://github.com/martinus/robin-hood-hashing
 //
 // Licensed under the MIT License <http://opensource.org/licenses/MIT>.
@@ -36,7 +35,7 @@
 
 // see https://semver.org/
 #define ROBIN_HOOD_VERSION_MAJOR 3 // for incompatible API changes
-#define ROBIN_HOOD_VERSION_MINOR 8 // for adding functionality in a backwards-compatible manner
+#define ROBIN_HOOD_VERSION_MINOR 9 // for adding functionality in a backwards-compatible manner
 #define ROBIN_HOOD_VERSION_PATCH 0 // for backwards-compatible bug fixes
 
 #include <algorithm>
@@ -132,46 +131,32 @@ static Counts& counts() {
 #endif
 
 // count leading/trailing bits
-#if ((defined __i386 || defined __x86_64__) && defined __BMI__) || defined _M_IX86 || defined _M_X64
+#if !defined(ROBIN_HOOD_DISABLE_INTRINSICS)
 #    ifdef _MSC_VER
+#        if ROBIN_HOOD(BITNESS) == 32
+#            define ROBIN_HOOD_PRIVATE_DEFINITION_BITSCANFORWARD() _BitScanForward
+#        else
+#            define ROBIN_HOOD_PRIVATE_DEFINITION_BITSCANFORWARD() _BitScanForward64
+#        endif
 #        include <intrin.h>
+#        pragma intrinsic(ROBIN_HOOD(BITSCANFORWARD))
+#        define ROBIN_HOOD_COUNT_TRAILING_ZEROES(x)                                       \
+            [](size_t mask) noexcept -> int {                                             \
+                unsigned long index;                                                      \
+                return ROBIN_HOOD(BITSCANFORWARD)(&index, mask) ? static_cast<int>(index) \
+                                                                : ROBIN_HOOD(BITNESS);    \
+            }(x)
 #    else
-#        include <x86intrin.h>
+#        if ROBIN_HOOD(BITNESS) == 32
+#            define ROBIN_HOOD_PRIVATE_DEFINITION_CTZ() __builtin_ctzl
+#            define ROBIN_HOOD_PRIVATE_DEFINITION_CLZ() __builtin_clzl
+#        else
+#            define ROBIN_HOOD_PRIVATE_DEFINITION_CTZ() __builtin_ctzll
+#            define ROBIN_HOOD_PRIVATE_DEFINITION_CLZ() __builtin_clzll
+#        endif
+#        define ROBIN_HOOD_COUNT_LEADING_ZEROES(x) ((x) ? ROBIN_HOOD(CLZ)(x) : ROBIN_HOOD(BITNESS))
+#        define ROBIN_HOOD_COUNT_TRAILING_ZEROES(x) ((x) ? ROBIN_HOOD(CTZ)(x) : ROBIN_HOOD(BITNESS))
 #    endif
-#    if ROBIN_HOOD(BITNESS) == 32
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_CTZ() _tzcnt_u32
-#    else
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_CTZ() _tzcnt_u64
-#    endif
-#    if defined __AVX2__ || defined __BMI__
-#        define ROBIN_HOOD_COUNT_TRAILING_ZEROES(x) ROBIN_HOOD(CTZ)(x)
-#    else
-#        define ROBIN_HOOD_COUNT_TRAILING_ZEROES(x) ROBIN_HOOD(CTZ)(x)
-#    endif
-#elif defined _MSC_VER
-#    if ROBIN_HOOD(BITNESS) == 32
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_BITSCANFORWARD() _BitScanForward
-#    else
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_BITSCANFORWARD() _BitScanForward64
-#    endif
-#    include <intrin.h>
-#    pragma intrinsic(ROBIN_HOOD(BITSCANFORWARD))
-#    define ROBIN_HOOD_COUNT_TRAILING_ZEROES(x)                                       \
-        [](size_t mask) noexcept -> int {                                             \
-            unsigned long index;                                                      \
-            return ROBIN_HOOD(BITSCANFORWARD)(&index, mask) ? static_cast<int>(index) \
-                                                            : ROBIN_HOOD(BITNESS);    \
-        }(x)
-#else
-#    if ROBIN_HOOD(BITNESS) == 32
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_CTZ() __builtin_ctzl
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_CLZ() __builtin_clzl
-#    else
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_CTZ() __builtin_ctzll
-#        define ROBIN_HOOD_PRIVATE_DEFINITION_CLZ() __builtin_clzll
-#    endif
-#    define ROBIN_HOOD_COUNT_LEADING_ZEROES(x) ((x) ? ROBIN_HOOD(CLZ)(x) : ROBIN_HOOD(BITNESS))
-#    define ROBIN_HOOD_COUNT_TRAILING_ZEROES(x) ((x) ? ROBIN_HOOD(CTZ)(x) : ROBIN_HOOD(BITNESS))
 #endif
 
 // fallthrough
@@ -195,6 +180,17 @@ static Counts& counts() {
 #    define ROBIN_HOOD_UNLIKELY(condition) __builtin_expect(condition, 0)
 #endif
 
+// detect if native wchar_t type is availiable in MSVC
+#ifdef _MSC_VER
+#    ifdef _NATIVE_WCHAR_T_DEFINED
+#        define ROBIN_HOOD_PRIVATE_DEFINITION_HAS_NATIVE_WCHART() 1
+#    else
+#        define ROBIN_HOOD_PRIVATE_DEFINITION_HAS_NATIVE_WCHART() 0
+#    endif
+#else
+#    define ROBIN_HOOD_PRIVATE_DEFINITION_HAS_NATIVE_WCHART() 1
+#endif
+
 // workaround missing "is_trivially_copyable" in g++ < 5.0
 // See https://stackoverflow.com/a/31798726/48181
 #if defined(__GNUC__) && __GNUC__ < 5
@@ -294,6 +290,13 @@ using index_sequence_for = make_index_sequence<sizeof...(T)>;
 
 namespace detail {
 
+// make sure we static_cast to the correct type for hash_int
+#if ROBIN_HOOD(BITNESS) == 64
+using SizeT = uint64_t;
+#else
+using SizeT = uint32_t;
+#endif
+
 template <typename T>
 T rotr(T x, unsigned k) {
     return (x >> k) | (x << (8U * sizeof(T) - k));
@@ -315,14 +318,14 @@ inline T reinterpret_cast_no_cast_align_warning(void const* ptr) noexcept {
 // make sure this is not inlined as it is slow and dramatically enlarges code, thus making other
 // inlinings more difficult. Throws are also generally the slow path.
 template <typename E, typename... Args>
-ROBIN_HOOD(NOINLINE)
+[[noreturn]] ROBIN_HOOD(NOINLINE)
 #if ROBIN_HOOD(HAS_EXCEPTIONS)
-void doThrow(Args&&... args) {
+    void doThrow(Args&&... args) {
     // NOLINTNEXTLINE(cppcoreguidelines-pro-bounds-array-to-pointer-decay)
     throw E(std::forward<Args>(args)...);
 }
 #else
-void doThrow(Args&&... ROBIN_HOOD_UNUSED(args) /*unused*/) {
+    void doThrow(Args&&... ROBIN_HOOD_UNUSED(args) /*unused*/) {
     abort();
 }
 #endif
@@ -388,7 +391,7 @@ public:
     void reset() noexcept {
         while (mListForFree) {
             T* tmp = *mListForFree;
-            free(mListForFree);
+            std::free(mListForFree);
             mListForFree = reinterpret_cast_no_cast_align_warning<T**>(tmp);
         }
         mHead = nullptr;
@@ -423,7 +426,7 @@ public:
         // calculate number of available elements in ptr
         if (numBytes < ALIGNMENT + ALIGNED_SIZE) {
             // not enough data for at least one element. Free and return.
-            free(ptr);
+            std::free(ptr);
         } else {
             add(ptr, numBytes);
         }
@@ -490,7 +493,7 @@ private:
         // alloc new memory: [prev |T, T, ... T]
         // std::cout << (sizeof(T*) + ALIGNED_SIZE * numElementsToAlloc) << " bytes" << std::endl;
         size_t const bytes = ALIGNMENT + ALIGNED_SIZE * numElementsToAlloc;
-        add(assertNotNull<std::bad_alloc>(malloc(bytes)), bytes);
+        add(assertNotNull<std::bad_alloc>(std::malloc(bytes)), bytes);
         return mHead;
     }
 
@@ -526,7 +529,7 @@ struct NodeAllocator<T, MinSize, MaxSize, true> {
 
     // we are not using the data, so just free it.
     void addOrFree(void* ptr, size_t ROBIN_HOOD_UNUSED(numBytes) /*unused*/) noexcept {
-        free(ptr);
+        std::free(ptr);
     }
 };
 
@@ -670,7 +673,7 @@ inline constexpr bool operator>=(pair<A, B> const& x, pair<A, B> const& y) {
     return !(x < y);
 }
 
-inline size_t hash_bytes(void const* ptr, size_t const len) noexcept {
+inline size_t hash_bytes(void const* ptr, size_t len) noexcept {
     static constexpr uint64_t m = UINT64_C(0xc6a4a7935bd1e995);
     static constexpr uint64_t seed = UINT64_C(0xe17a1465);
     static constexpr unsigned int r = 47;
@@ -730,8 +733,8 @@ inline size_t hash_int(uint64_t x) noexcept {
     //
     // Instead of shifts, we use rotations so we don't lose any bits.
     //
-    // Added a final multiplcation with a constant for more mixing. It is most important that the
+    // Added a final multiplcation with a constant for more mixing. It is most important that
-    // lower bits are well mixed.
+    // the lower bits are well mixed.
     auto h1 = x * UINT64_C(0xA24BAED4963EE407);
     auto h2 = detail::rotr(x, 32U) * UINT64_C(0x9FB21C651E98DF25);
     auto h = detail::rotr(h1 + h2, 32U);
@@ -739,14 +742,14 @@ inline size_t hash_int(uint64_t x) noexcept {
 }
 
 // A thin wrapper around std::hash, performing an additional simple mixing step of the result.
-template <typename T>
+template <typename T, typename Enable = void>
 struct hash : public std::hash<T> {
     size_t operator()(T const& obj) const
         noexcept(noexcept(std::declval<std::hash<T>>().operator()(std::declval<T const&>()))) {
         // call base hash
         auto result = std::hash<T>::operator()(obj);
         // return mixed of that, to be save against identity has
-        return hash_int(static_cast<uint64_t>(result));
+        return hash_int(static_cast<detail::SizeT>(result));
     }
 };
 
@@ -769,21 +772,29 @@ struct hash<std::basic_string_view<CharT>> {
 template <class T>
 struct hash<T*> {
     size_t operator()(T* ptr) const noexcept {
-        return hash_int(reinterpret_cast<size_t>(ptr));
+        return hash_int(reinterpret_cast<detail::SizeT>(ptr));
     }
 };
 
 template <class T>
 struct hash<std::unique_ptr<T>> {
     size_t operator()(std::unique_ptr<T> const& ptr) const noexcept {
-        return hash_int(reinterpret_cast<size_t>(ptr.get()));
+        return hash_int(reinterpret_cast<detail::SizeT>(ptr.get()));
     }
 };
 
 template <class T>
 struct hash<std::shared_ptr<T>> {
     size_t operator()(std::shared_ptr<T> const& ptr) const noexcept {
-        return hash_int(reinterpret_cast<size_t>(ptr.get()));
+        return hash_int(reinterpret_cast<detail::SizeT>(ptr.get()));
+    }
+};
+
+template <typename Enum>
+struct hash<Enum, typename std::enable_if<std::is_enum<Enum>::value>::type> {
+    size_t operator()(Enum e) const noexcept {
+        using Underlying = typename std::underlying_type<Enum>::type;
+        return hash<Underlying>{}(static_cast<Underlying>(e));
     }
 };
 
@@ -806,7 +817,9 @@ ROBIN_HOOD_HASH_INT(signed char);
 ROBIN_HOOD_HASH_INT(unsigned char);
 ROBIN_HOOD_HASH_INT(char16_t);
 ROBIN_HOOD_HASH_INT(char32_t);
+#if ROBIN_HOOD(HAS_NATIVE_WCHART)
 ROBIN_HOOD_HASH_INT(wchar_t);
+#endif
 ROBIN_HOOD_HASH_INT(short);
 ROBIN_HOOD_HASH_INT(unsigned short);
 ROBIN_HOOD_HASH_INT(int);
@@ -914,7 +927,8 @@ private:
     static constexpr size_t InitialNumElements = sizeof(uint64_t);
     static constexpr uint32_t InitialInfoNumBits = 5;
     static constexpr uint8_t InitialInfoInc = 1U << InitialInfoNumBits;
-    static constexpr uint8_t InitialInfoHashShift = sizeof(size_t) * 8 - InitialInfoNumBits;
+    static constexpr size_t InfoMask = InitialInfoInc - 1U;
+    static constexpr uint8_t InitialInfoHashShift = 0;
     using DataPool = detail::NodeAllocator<value_type, 4, 16384, IsFlat>;
 
     // type needs to be wider than uint8_t.
@@ -1247,7 +1261,7 @@ private:
         Iter operator++(int) noexcept {
             Iter tmp = *this;
             ++(*this);
-            return std::move(tmp);
+            return tmp;
         }
 
         reference operator*() const {
@@ -1278,13 +1292,29 @@ private:
                 mInfo += sizeof(size_t);
                 mKeyVals += sizeof(size_t);
             }
-#if ROBIN_HOOD(LITTLE_ENDIAN)
+#if defined(ROBIN_HOOD_DISABLE_INTRINSICS)
-            auto inc = ROBIN_HOOD_COUNT_TRAILING_ZEROES(n) / 8;
+            // we know for certain that within the next 8 bytes we'll find a non-zero one.
+            if (ROBIN_HOOD_UNLIKELY(0U == detail::unaligned_load<uint32_t>(mInfo))) {
+                mInfo += 4;
+                mKeyVals += 4;
+            }
+            if (ROBIN_HOOD_UNLIKELY(0U == detail::unaligned_load<uint16_t>(mInfo))) {
+                mInfo += 2;
+                mKeyVals += 2;
+            }
+            if (ROBIN_HOOD_UNLIKELY(0U == *mInfo)) {
+                mInfo += 1;
+                mKeyVals += 1;
+            }
 #else
+#    if ROBIN_HOOD(LITTLE_ENDIAN)
+            auto inc = ROBIN_HOOD_COUNT_TRAILING_ZEROES(n) / 8;
+#    else
             auto inc = ROBIN_HOOD_COUNT_LEADING_ZEROES(n) / 8;
-#endif
+#    endif
             mInfo += inc;
             mKeyVals += inc;
+#endif
         }
 
         friend class Table<IsFlat, MaxLoadFactor100, key_type, mapped_type, hasher, key_equal>;
@@ -1306,10 +1336,11 @@ private:
             typename std::conditional<std::is_same<::robin_hood::hash<key_type>, hasher>::value,
                                       ::robin_hood::detail::identity_hash<size_t>,
                                       ::robin_hood::hash<size_t>>::type;
-        *idx = Mix{}(WHash::operator()(key));
 
-        *info = mInfoInc + static_cast<InfoType>(*idx >> mInfoHashShift);
+        // the lower InitialInfoNumBits are reserved for info.
-        *idx &= mMask;
+        auto h = Mix{}(WHash::operator()(key));
+        *info = mInfoInc + static_cast<InfoType>((h & InfoMask) >> mInfoHashShift);
+        *idx = (h >> InitialInfoNumBits) & mMask;
     }
 
     // forwards the index by one, wrapping around at the end
@@ -1448,13 +1479,19 @@ public:
     using iterator = Iter<false>;
     using const_iterator = Iter<true>;
 
+    Table() noexcept(noexcept(Hash()) && noexcept(KeyEqual()))
+        : WHash()
+        , WKeyEqual() {
+        ROBIN_HOOD_TRACE(this)
+    }
+
     // Creates an empty hash map. Nothing is allocated yet, this happens at the first insert.
     // This tremendously speeds up ctor & dtor of a map that never receives an element. The
     // penalty is payed at the first insert, and not before. Lookup of this empty map works
     // because everybody points to DummyInfoByte::b. parameter bucket_count is dictated by the
     // standard, but we can ignore it.
     explicit Table(
-        size_t ROBIN_HOOD_UNUSED(bucket_count) /*unused*/ = 0, const Hash& h = Hash{},
+        size_t ROBIN_HOOD_UNUSED(bucket_count) /*unused*/, const Hash& h = Hash{},
         const KeyEqual& equal = KeyEqual{}) noexcept(noexcept(Hash(h)) && noexcept(KeyEqual(equal)))
         : WHash(h)
         , WKeyEqual(equal) {
@@ -1535,7 +1572,7 @@ public:
 
             auto const numElementsWithBuffer = calcNumElementsWithBuffer(o.mMask + 1);
             mKeyVals = static_cast<Node*>(detail::assertNotNull<std::bad_alloc>(
-                malloc(calcNumBytesTotal(numElementsWithBuffer))));
+                std::malloc(calcNumBytesTotal(numElementsWithBuffer))));
             // no need for calloc because clonData does memcpy
             mInfo = reinterpret_cast<uint8_t*>(mKeyVals + numElementsWithBuffer);
             mNumElements = o.mNumElements;
@@ -1583,12 +1620,12 @@ public:
             // no luck: we don't have the same array size allocated, so we need to realloc.
             if (0 != mMask) {
                 // only deallocate if we actually have data!
-                free(mKeyVals);
+                std::free(mKeyVals);
             }
 
             auto const numElementsWithBuffer = calcNumElementsWithBuffer(o.mMask + 1);
             mKeyVals = static_cast<Node*>(detail::assertNotNull<std::bad_alloc>(
-                malloc(calcNumBytesTotal(numElementsWithBuffer))));
+                std::malloc(calcNumBytesTotal(numElementsWithBuffer))));
 
             // no need for calloc here because cloneData performs a memcpy.
             mInfo = reinterpret_cast<uint8_t*>(mKeyVals + numElementsWithBuffer);
@@ -2058,8 +2095,13 @@ private:
                 }
             }
 
-            // don't destroy old data: put it into the pool instead
+            // this check is not necessary as it's guarded by the previous if, but it helps silence
-            DataPool::addOrFree(oldKeyVals, calcNumBytesTotal(oldMaxElementsWithBuffer));
+            // g++'s overeager "attempt to free a non-heap object 'map'
+            // [-Werror=free-nonheap-object]" warning.
+            if (oldKeyVals != reinterpret_cast_no_cast_align_warning<Node*>(&mMask)) {
+                // don't destroy old data: put it into the pool instead
+                DataPool::addOrFree(oldKeyVals, calcNumBytesTotal(oldMaxElementsWithBuffer));
+            }
         }
     }
 
@@ -2103,7 +2145,7 @@ private:
 
         // calloc also zeroes everything
         mKeyVals = reinterpret_cast<Node*>(detail::assertNotNull<std::bad_alloc>(
-            calloc(1, calcNumBytesTotal(numElementsWithBuffer))));
+            std::calloc(1, calcNumBytesTotal(numElementsWithBuffer))));
         mInfo = reinterpret_cast<uint8_t*>(mKeyVals + numElementsWithBuffer);
 
         // set sentinel
@@ -2290,7 +2332,7 @@ private:
         // reports a compile error: attempt to free a non-heap object ‘fm’
         // [-Werror=free-nonheap-object]
         if (mKeyVals != reinterpret_cast_no_cast_align_warning<Node*>(&mMask)) {
-            free(mKeyVals);
+            std::free(mKeyVals);
         }
     }
 

deps/tracy/AUTHORS

@@ -9,3 +9,6 @@ Dedmen Miller <dedmen@dedmen.de>                    (find zone bug fixes, improv
 Michał Cichoń <michcic@gmail.com>                   (OSX call stack decoding backport)
 Thales Sabino <thales@codeplay.com>                 (OpenCL support)
 Andrew Depke <andrewdepke@gmail.com>                (Direct3D 12 support)
+Simonas Kazlauskas <git@kazlauskas.me>              (OSX CI, external bindings)
+Jakub Žádník <kubouch@gmail.com>                    (csvexport utility)
+Andrey Voroshilov <andrew.voroshilov@gmail.com>     (multi-DLL fixes)

deps/tracy/Tracy.hpp

@@ -11,6 +11,9 @@
 #define ZoneNamedC(x,y,z)
 #define ZoneNamedNC(x,y,z,w)
 
+#define ZoneTransient(x,y)
+#define ZoneTransientN(x,y,z)
+
 #define ZoneScoped
 #define ZoneScopedN(x)
 #define ZoneScopedC(x)
@@ -50,12 +53,22 @@
 
 #define TracyAlloc(x,y)
 #define TracyFree(x)
+#define TracySecureAlloc(x,y)
+#define TracySecureFree(x)
+
+#define TracyAllocN(x,y,z)
+#define TracyFreeN(x,y)
+#define TracySecureAllocN(x,y,z)
+#define TracySecureFreeN(x,y)
 
 #define ZoneNamedS(x,y,z)
 #define ZoneNamedNS(x,y,z,w)
 #define ZoneNamedCS(x,y,z,w)
 #define ZoneNamedNCS(x,y,z,w,a)
 
+#define ZoneTransientS(x,y,z)
+#define ZoneTransientNS(x,y,z,w)
+
 #define ZoneScopedS(x)
 #define ZoneScopedNS(x,y)
 #define ZoneScopedCS(x,y)
@@ -63,6 +76,13 @@
 
 #define TracyAllocS(x,y,z)
 #define TracyFreeS(x,y)
+#define TracySecureAllocS(x,y,z)
+#define TracySecureFreeS(x,y)
+
+#define TracyAllocNS(x,y,z,w)
+#define TracyFreeNS(x,y,z)
+#define TracySecureAllocNS(x,y,z,w)
+#define TracySecureFreeNS(x,y,z)
 
 #define TracyMessageS(x,y,z)
 #define TracyMessageLS(x,y)
@@ -71,23 +91,32 @@
 
 #define TracyParameterRegister(x)
 #define TracyParameterSetup(x,y,z,w)
+#define TracyIsConnected false
 
 #else
 
+#include <string.h>
+
 #include "client/TracyLock.hpp"
 #include "client/TracyProfiler.hpp"
 #include "client/TracyScoped.hpp"
 
 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK
-#  define ZoneNamed( varname, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
+#  define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
-#  define ZoneNamedN( varname, name, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
+#  define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
-#  define ZoneNamedC( varname, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
+#  define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
-#  define ZoneNamedNC( varname, name, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
+#  define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), TRACY_CALLSTACK, active );
+
+#  define ZoneTransient( varname, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), nullptr, 0, TRACY_CALLSTACK, active );
+#  define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), TRACY_CALLSTACK, active );
 #else
-#  define ZoneNamed( varname, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
+#  define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
-#  define ZoneNamedN( varname, name, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
+#  define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
-#  define ZoneNamedC( varname, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
+#  define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
-#  define ZoneNamedNC( varname, name, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
+#  define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), active );
+
+#  define ZoneTransient( varname, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), nullptr, 0, active );
+#  define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), active );
 #endif
 
 #define ZoneScoped ZoneNamed( ___tracy_scoped_zone, true )
@@ -109,13 +138,13 @@
 
 #define FrameImage( image, width, height, offset, flip ) tracy::Profiler::SendFrameImage( image, width, height, offset, flip );
 
-#define TracyLockable( type, varname ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static const tracy::SourceLocationData srcloc { nullptr, #type " " #varname, __FILE__, __LINE__, 0 }; return &srcloc; }() };
+#define TracyLockable( type, varname ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, __FILE__, __LINE__, 0 }; return &srcloc; }() };
-#define TracyLockableN( type, varname, desc ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static const tracy::SourceLocationData srcloc { nullptr, desc, __FILE__, __LINE__, 0 }; return &srcloc; }() };
+#define TracyLockableN( type, varname, desc ) tracy::Lockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, __FILE__, __LINE__, 0 }; return &srcloc; }() };
-#define TracySharedLockable( type, varname ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static const tracy::SourceLocationData srcloc { nullptr, #type " " #varname, __FILE__, __LINE__, 0 }; return &srcloc; }() };
+#define TracySharedLockable( type, varname ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, __FILE__, __LINE__, 0 }; return &srcloc; }() };
-#define TracySharedLockableN( type, varname, desc ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static const tracy::SourceLocationData srcloc { nullptr, desc, __FILE__, __LINE__, 0 }; return &srcloc; }() };
+#define TracySharedLockableN( type, varname, desc ) tracy::SharedLockable<type> varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, __FILE__, __LINE__, 0 }; return &srcloc; }() };
 #define LockableBase( type ) tracy::Lockable<type>
 #define SharedLockableBase( type ) tracy::SharedLockable<type>
-#define LockMark( varname ) static const tracy::SourceLocationData __tracy_lock_location_##varname { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; varname.Mark( &__tracy_lock_location_##varname );
+#define LockMark( varname ) static constexpr tracy::SourceLocationData __tracy_lock_location_##varname { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; varname.Mark( &__tracy_lock_location_##varname );
 #define LockableName( varname, txt, size ) varname.CustomName( txt, size );
 
 #define TracyPlot( name, val ) tracy::Profiler::PlotData( name, val );
@@ -129,31 +158,55 @@
 #  define TracyMessageC( txt, size, color ) tracy::Profiler::MessageColor( txt, size, color, TRACY_CALLSTACK );
 #  define TracyMessageLC( txt, color ) tracy::Profiler::MessageColor( txt, color, TRACY_CALLSTACK );
 
-#  define TracyAlloc( ptr, size ) tracy::Profiler::MemAllocCallstack( ptr, size, TRACY_CALLSTACK );
+#  define TracyAlloc( ptr, size ) tracy::Profiler::MemAllocCallstack( ptr, size, TRACY_CALLSTACK, false );
-#  define TracyFree( ptr ) tracy::Profiler::MemFreeCallstack( ptr, TRACY_CALLSTACK );
+#  define TracyFree( ptr ) tracy::Profiler::MemFreeCallstack( ptr, TRACY_CALLSTACK, false );
+#  define TracySecureAlloc( ptr, size ) tracy::Profiler::MemAllocCallstack( ptr, size, TRACY_CALLSTACK, true );
+#  define TracySecureFree( ptr ) tracy::Profiler::MemFreeCallstack( ptr, TRACY_CALLSTACK, true );
+
+#  define TracyAllocN( ptr, size, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, TRACY_CALLSTACK, false, name );
+#  define TracyFreeN( ptr, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, TRACY_CALLSTACK, false, name );
+#  define TracySecureAllocN( ptr, size, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, TRACY_CALLSTACK, true, name );
+#  define TracySecureFreeN( ptr, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, TRACY_CALLSTACK, true, name );
 #else
 #  define TracyMessage( txt, size ) tracy::Profiler::Message( txt, size, 0 );
 #  define TracyMessageL( txt ) tracy::Profiler::Message( txt, 0 );
 #  define TracyMessageC( txt, size, color ) tracy::Profiler::MessageColor( txt, size, color, 0 );
 #  define TracyMessageLC( txt, color ) tracy::Profiler::MessageColor( txt, color, 0 );
 
-#  define TracyAlloc( ptr, size ) tracy::Profiler::MemAlloc( ptr, size );
+#  define TracyAlloc( ptr, size ) tracy::Profiler::MemAlloc( ptr, size, false );
-#  define TracyFree( ptr ) tracy::Profiler::MemFree( ptr );
+#  define TracyFree( ptr ) tracy::Profiler::MemFree( ptr, false );
+#  define TracySecureAlloc( ptr, size ) tracy::Profiler::MemAlloc( ptr, size, true );
+#  define TracySecureFree( ptr ) tracy::Profiler::MemFree( ptr, true );
+
+#  define TracyAllocN( ptr, size, name ) tracy::Profiler::MemAllocNamed( ptr, size, false, name );
+#  define TracyFreeN( ptr, name ) tracy::Profiler::MemFreeNamed( ptr, false, name );
+#  define TracySecureAllocN( ptr, size, name ) tracy::Profiler::MemAllocNamed( ptr, size, true, name );
+#  define TracySecureFreeN( ptr, name ) tracy::Profiler::MemFreeNamed( ptr, true, name );
 #endif
 
 #ifdef TRACY_HAS_CALLSTACK
-#  define ZoneNamedS( varname, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
+#  define ZoneNamedS( varname, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
-#  define ZoneNamedNS( varname, name, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
+#  define ZoneNamedNS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
-#  define ZoneNamedCS( varname, color, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
+#  define ZoneNamedCS( varname, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
-#  define ZoneNamedNCS( varname, name, color, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
+#  define ZoneNamedNCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
+
+#  define ZoneTransientS( varname, depth, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), nullptr, 0, depth, active );
+#  define ZoneTransientNS( varname, name, depth, active ) tracy::ScopedZone varname( __LINE__, __FILE__, strlen( __FILE__ ), __FUNCTION__, strlen( __FUNCTION__ ), name, strlen( name ), depth, active );
 
 #  define ZoneScopedS( depth ) ZoneNamedS( ___tracy_scoped_zone, depth, true )
 #  define ZoneScopedNS( name, depth ) ZoneNamedNS( ___tracy_scoped_zone, name, depth, true )
 #  define ZoneScopedCS( color, depth ) ZoneNamedCS( ___tracy_scoped_zone, color, depth, true )
 #  define ZoneScopedNCS( name, color, depth ) ZoneNamedNCS( ___tracy_scoped_zone, name, color, depth, true )
 
-#  define TracyAllocS( ptr, size, depth ) tracy::Profiler::MemAllocCallstack( ptr, size, depth );
+#  define TracyAllocS( ptr, size, depth ) tracy::Profiler::MemAllocCallstack( ptr, size, depth, false );
-#  define TracyFreeS( ptr, depth ) tracy::Profiler::MemFreeCallstack( ptr, depth );
+#  define TracyFreeS( ptr, depth ) tracy::Profiler::MemFreeCallstack( ptr, depth, false );
+#  define TracySecureAllocS( ptr, size, depth ) tracy::Profiler::MemAllocCallstack( ptr, size, depth, true );
+#  define TracySecureFreeS( ptr, depth ) tracy::Profiler::MemFreeCallstack( ptr, depth, true );
+
+#  define TracyAllocNS( ptr, size, depth, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, depth, false, name );
+#  define TracyFreeNS( ptr, depth, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, depth, false, name );
+#  define TracySecureAllocNS( ptr, size, depth, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, depth, true, name );
+#  define TracySecureFreeNS( ptr, depth, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, depth, true, name );
 
 #  define TracyMessageS( txt, size, depth ) tracy::Profiler::Message( txt, size, depth );
 #  define TracyMessageLS( txt, depth ) tracy::Profiler::Message( txt, depth );
@@ -165,6 +218,9 @@
 #  define ZoneNamedCS( varname, color, depth, active ) ZoneNamedC( varname, color, active )
 #  define ZoneNamedNCS( varname, name, color, depth, active ) ZoneNamedNC( varname, name, color, active )
 
+#  define ZoneTransientS( varname, depth, active ) ZoneTransient( varname, active )
+#  define ZoneTransientNS( varname, name, depth, active ) ZoneTransientN( varname, name, active )
+
 #  define ZoneScopedS( depth ) ZoneScoped
 #  define ZoneScopedNS( name, depth ) ZoneScopedN( name )
 #  define ZoneScopedCS( color, depth ) ZoneScopedC( color )
@@ -172,6 +228,13 @@
 
 #  define TracyAllocS( ptr, size, depth ) TracyAlloc( ptr, size )
 #  define TracyFreeS( ptr, depth ) TracyFree( ptr )
+#  define TracySecureAllocS( ptr, size, depth ) TracySecureAlloc( ptr, size )
+#  define TracySecureFreeS( ptr, depth ) TracySecureFree( ptr )
+
+#  define TracyAllocNS( ptr, size, depth, name ) TracyAlloc( ptr, size, name )
+#  define TracyFreeNS( ptr, depth, name ) TracyFree( ptr, name )
+#  define TracySecureAllocNS( ptr, size, depth, name ) TracySecureAlloc( ptr, size, name )
+#  define TracySecureFreeNS( ptr, depth, name ) TracySecureFree( ptr, name )
 
 #  define TracyMessageS( txt, size, depth ) TracyMessage( txt, size )
 #  define TracyMessageLS( txt, depth ) TracyMessageL( txt )
@@ -181,6 +244,7 @@
 
 #define TracyParameterRegister( cb ) tracy::Profiler::ParameterRegister( cb );
 #define TracyParameterSetup( idx, name, isBool, val ) tracy::Profiler::ParameterSetup( idx, name, isBool, val );
+#define TracyIsConnected tracy::GetProfiler().IsConnected()
 
 #endif
 

deps/tracy/TracyC.h

@@ -11,6 +11,11 @@
 extern "C" {
 #endif
 
+TRACY_API void ___tracy_set_thread_name( const char* name );
+
+#define TracyCSetThreadName( name ) ___tracy_set_thread_name( name );
+
+
 #ifndef TRACY_ENABLE
 
 typedef const void* TracyCZoneCtx;
@@ -26,6 +31,8 @@ typedef const void* TracyCZoneCtx;
 
 #define TracyCAlloc(x,y)
 #define TracyCFree(x)
+#define TracyCSecureAlloc(x,y)
+#define TracyCSecureFree(x)
 
 #define TracyCFrameMark
 #define TracyCFrameMarkNamed(x)
@@ -47,6 +54,8 @@ typedef const void* TracyCZoneCtx;
 
 #define TracyCAllocS(x,y,z)
 #define TracyCFreeS(x,y)
+#define TracyCSecureAllocS(x,y,z)
+#define TracyCSecureFreeS(x,y)
 
 #define TracyCMessageS(x,y,z)
 #define TracyCMessageLS(x,y)
@@ -81,8 +90,9 @@ struct ___tracy_c_zone_context
 // This struct, as visible to user, is immutable, so treat it as if const was declared here.
 typedef /*const*/ struct ___tracy_c_zone_context TracyCZoneCtx;
 
-TRACY_API uint64_t ___tracy_alloc_srcloc( uint32_t line, const char* source, const char* function );
+TRACY_API void ___tracy_init_thread(void);
-TRACY_API uint64_t ___tracy_alloc_srcloc_name( uint32_t line, const char* source, const char* function, const char* name, size_t nameSz );
+TRACY_API uint64_t ___tracy_alloc_srcloc( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz );
+TRACY_API uint64_t ___tracy_alloc_srcloc_name( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz );
 
 TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin( const struct ___tracy_source_location_data* srcloc, int active );
 TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_callstack( const struct ___tracy_source_location_data* srcloc, int depth, int active );
@@ -112,10 +122,10 @@ TRACY_API void ___tracy_emit_zone_value( TracyCZoneCtx ctx, uint64_t value );
 #define TracyCZoneValue( ctx, value ) ___tracy_emit_zone_value( ctx, value );
 
 
-TRACY_API void ___tracy_emit_memory_alloc( const void* ptr, size_t size );
+TRACY_API void ___tracy_emit_memory_alloc( const void* ptr, size_t size, int secure );
-TRACY_API void ___tracy_emit_memory_alloc_callstack( const void* ptr, size_t size, int depth );
+TRACY_API void ___tracy_emit_memory_alloc_callstack( const void* ptr, size_t size, int depth, int secure );
-TRACY_API void ___tracy_emit_memory_free( const void* ptr );
+TRACY_API void ___tracy_emit_memory_free( const void* ptr, int secure );
-TRACY_API void ___tracy_emit_memory_free_callstack( const void* ptr, int depth );
+TRACY_API void ___tracy_emit_memory_free_callstack( const void* ptr, int depth, int secure );
 
 TRACY_API void ___tracy_emit_message( const char* txt, size_t size, int callstack );
 TRACY_API void ___tracy_emit_messageL( const char* txt, int callstack );
@@ -123,16 +133,20 @@ TRACY_API void ___tracy_emit_messageC( const char* txt, size_t size, uint32_t co
 TRACY_API void ___tracy_emit_messageLC( const char* txt, uint32_t color, int callstack );
 
 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK
-#  define TracyCAlloc( ptr, size ) ___tracy_emit_memory_alloc_callstack( ptr, size, TRACY_CALLSTACK )
+#  define TracyCAlloc( ptr, size ) ___tracy_emit_memory_alloc_callstack( ptr, size, TRACY_CALLSTACK, 0 )
-#  define TracyCFree( ptr ) ___tracy_emit_memory_free_callstack( ptr, TRACY_CALLSTACK )
+#  define TracyCFree( ptr ) ___tracy_emit_memory_free_callstack( ptr, TRACY_CALLSTACK, 0 )
+#  define TracyCSecureAlloc( ptr, size ) ___tracy_emit_memory_alloc_callstack( ptr, size, TRACY_CALLSTACK, 1 )
+#  define TracyCSecureFree( ptr ) ___tracy_emit_memory_free_callstack( ptr, TRACY_CALLSTACK, 1 )
 
 #  define TracyCMessage( txt, size ) ___tracy_emit_message( txt, size, TRACY_CALLSTACK );
 #  define TracyCMessageL( txt ) ___tracy_emit_messageL( txt, TRACY_CALLSTACK );
 #  define TracyCMessageC( txt, size, color ) ___tracy_emit_messageC( txt, size, color, TRACY_CALLSTACK );
 #  define TracyCMessageLC( txt, color ) ___tracy_emit_messageLC( txt, color, TRACY_CALLSTACK );
 #else
-#  define TracyCAlloc( ptr, size ) ___tracy_emit_memory_alloc( ptr, size );
+#  define TracyCAlloc( ptr, size ) ___tracy_emit_memory_alloc( ptr, size, 0 );
-#  define TracyCFree( ptr ) ___tracy_emit_memory_free( ptr );
+#  define TracyCFree( ptr ) ___tracy_emit_memory_free( ptr, 0 );
+#  define TracyCSecureAlloc( ptr, size ) ___tracy_emit_memory_alloc( ptr, size, 1 );
+#  define TracyCSecureFree( ptr ) ___tracy_emit_memory_free( ptr, 1 );
 
 #  define TracyCMessage( txt, size ) ___tracy_emit_message( txt, size, 0 );
 #  define TracyCMessageL( txt ) ___tracy_emit_messageL( txt, 0 );
@@ -166,8 +180,10 @@ TRACY_API void ___tracy_emit_message_appinfo( const char* txt, size_t size );
 #  define TracyCZoneCS( ctx, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { NULL, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
 #  define TracyCZoneNCS( ctx, name, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,__LINE__) = { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,__LINE__), depth, active );
 
-#  define TracyCAllocS( ptr, size, depth ) ___tracy_emit_memory_alloc_callstack( ptr, size, depth )
+#  define TracyCAllocS( ptr, size, depth ) ___tracy_emit_memory_alloc_callstack( ptr, size, depth, 0 )
-#  define TracyCFreeS( ptr, depth ) ___tracy_emit_memory_free_callstack( ptr, depth )
+#  define TracyCFreeS( ptr, depth ) ___tracy_emit_memory_free_callstack( ptr, depth, 0 )
+#  define TracyCSecureAllocS( ptr, size, depth ) ___tracy_emit_memory_alloc_callstack( ptr, size, depth, 1 )
+#  define TracyCSecureFreeS( ptr, depth ) ___tracy_emit_memory_free_callstack( ptr, depth, 1 )
 
 #  define TracyCMessageS( txt, size, depth ) ___tracy_emit_message( txt, size, depth );
 #  define TracyCMessageLS( txt, depth ) ___tracy_emit_messageL( txt, depth );
@@ -181,6 +197,8 @@ TRACY_API void ___tracy_emit_message_appinfo( const char* txt, size_t size );
 
 #  define TracyCAllocS( ptr, size, depth ) TracyCAlloc( ptr, size )
 #  define TracyCFreeS( ptr, depth ) TracyCFree( ptr )
+#  define TracyCSecureAllocS( ptr, size, depth ) TracyCSecureAlloc( ptr, size )
+#  define TracyCSecureFreeS( ptr, depth ) TracyCSecureFree( ptr )
 
 #  define TracyCMessageS( txt, size, depth ) TracyCMessage( txt, size )
 #  define TracyCMessageLS( txt, depth ) TracyCMessageL( txt )

deps/tracy/TracyClient.cpp

@@ -15,6 +15,10 @@
 
 #ifdef TRACY_ENABLE
 
+#ifdef _MSC_VER
+#  pragma warning(push, 0)
+#endif
+
 #include "common/tracy_lz4.cpp"
 #include "client/TracyProfiler.cpp"
 #include "client/TracyCallstack.cpp"
@@ -42,6 +46,7 @@
 #ifdef _MSC_VER
 #  pragma comment(lib, "ws2_32.lib")
 #  pragma comment(lib, "dbghelp.lib")
+#  pragma warning(pop)
 #endif
 
 #endif

deps/tracy/TracyOpenGL.hpp

@@ -52,21 +52,21 @@ public:
 
 #define TracyGpuContext tracy::InitRPMallocThread(); tracy::GetGpuCtx().ptr = (tracy::GpuCtx*)tracy::tracy_malloc( sizeof( tracy::GpuCtx ) ); new(tracy::GetGpuCtx().ptr) tracy::GpuCtx;
 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK
-#  define TracyGpuNamedZone( varname, name, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active );
+#  define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active );
-#  define TracyGpuNamedZoneC( varname, name, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active );
+#  define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), TRACY_CALLSTACK, active );
 #  define TracyGpuZone( name ) TracyGpuNamedZoneS( ___tracy_gpu_zone, name, TRACY_CALLSTACK, true )
 #  define TracyGpuZoneC( name, color ) TracyGpuNamedZoneCS( ___tracy_gpu_zone, name, color, TRACY_CALLSTACK, true )
 #else
-#  define TracyGpuNamedZone( varname, name, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), active );
+#  define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), active );
-#  define TracyGpuNamedZoneC( varname, name, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), active );
+#  define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), active );
 #  define TracyGpuZone( name ) TracyGpuNamedZone( ___tracy_gpu_zone, name, true )
 #  define TracyGpuZoneC( name, color ) TracyGpuNamedZoneC( ___tracy_gpu_zone, name, color, true )
 #endif
 #define TracyGpuCollect tracy::GetGpuCtx().ptr->Collect();
 
 #ifdef TRACY_HAS_CALLSTACK
-#  define TracyGpuNamedZoneS( varname, name, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active );
+#  define TracyGpuNamedZoneS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active );
-#  define TracyGpuNamedZoneCS( varname, name, color, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active );
+#  define TracyGpuNamedZoneCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,__LINE__), depth, active );
 #  define TracyGpuZoneS( name, depth ) TracyGpuNamedZoneS( ___tracy_gpu_zone, name, depth, true )
 #  define TracyGpuZoneCS( name, color, depth ) TracyGpuNamedZoneCS( ___tracy_gpu_zone, name, color, depth, true )
 #else
@@ -110,7 +110,7 @@ public:
         MemWrite( &item->gpuNewContext.thread, thread );
         MemWrite( &item->gpuNewContext.period, period );
         MemWrite( &item->gpuNewContext.context, m_context );
-        MemWrite( &item->gpuNewContext.accuracyBits, (uint8_t)bits );
+        MemWrite( &item->gpuNewContext.flags, uint8_t( 0 ) );
         MemWrite( &item->gpuNewContext.type, GpuContextType::OpenGl );
 
 #ifdef TRACY_ON_DEMAND
@@ -215,6 +215,8 @@ public:
         const auto queryId = GetGpuCtx().ptr->NextQueryId();
         glQueryCounter( GetGpuCtx().ptr->TranslateOpenGlQueryId( queryId ), GL_TIMESTAMP );
 
+        GetProfiler().SendCallstack( depth );
+
         const auto thread = GetThreadHandle();
         TracyLfqPrepare( QueueType::GpuZoneBeginCallstack );
         MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() );
@@ -223,8 +225,6 @@ public:
         MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) );
         MemWrite( &item->gpuZoneBegin.context, GetGpuCtx().ptr->GetId() );
         TracyLfqCommit;
-
-        GetProfiler().SendCallstack( depth );
     }
 
     tracy_force_inline ~GpuCtxScope()

deps/tracy/TracyVulkan.hpp

@@ -4,6 +4,7 @@
 #if !defined TRACY_ENABLE
 
 #define TracyVkContext(x,y,z,w) nullptr
+#define TracyVkContextCalibrated(x,y,z,w,a,b) nullptr
 #define TracyVkDestroy(x)
 #define TracyVkNamedZone(c,x,y,z,w)
 #define TracyVkNamedZoneC(c,x,y,z,w,a)
@@ -19,10 +20,9 @@
 namespace tracy
 {
 class VkCtxScope {};
-class VkCtx;
 }
 
-using TracyVkCtx = tracy::VkCtx*;
+using TracyVkCtx = void*;
 
 #else
 
@@ -43,16 +43,36 @@ class VkCtx
     enum { QueryCount = 64 * 1024 };
 
 public:
-    VkCtx( VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf )
+    VkCtx( VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf, PFN_vkGetPhysicalDeviceCalibrateableTimeDomainsEXT _vkGetPhysicalDeviceCalibrateableTimeDomainsEXT, PFN_vkGetCalibratedTimestampsEXT _vkGetCalibratedTimestampsEXT )
         : m_device( device )
+        , m_timeDomain( VK_TIME_DOMAIN_DEVICE_EXT )
         , m_context( GetGpuCtxCounter().fetch_add( 1, std::memory_order_relaxed ) )
         , m_head( 0 )
         , m_tail( 0 )
         , m_oldCnt( 0 )
         , m_queryCount( QueryCount )
+        , m_vkGetCalibratedTimestampsEXT( _vkGetCalibratedTimestampsEXT )
     {
         assert( m_context != 255 );
 
+        if( _vkGetPhysicalDeviceCalibrateableTimeDomainsEXT && _vkGetCalibratedTimestampsEXT )
+        {
+            uint32_t num;
+            _vkGetPhysicalDeviceCalibrateableTimeDomainsEXT( physdev, &num, nullptr );
+            if( num > 4 ) num = 4;
+            VkTimeDomainEXT data[4];
+            _vkGetPhysicalDeviceCalibrateableTimeDomainsEXT( physdev, &num, data );
+            for( uint32_t i=0; i<num; i++ )
+            {
+                // TODO VK_TIME_DOMAIN_CLOCK_MONOTONIC_RAW_EXT
+                if( data[i] == VK_TIME_DOMAIN_QUERY_PERFORMANCE_COUNTER_EXT )
+                {
+                    m_timeDomain = data[i];
+                    break;
+                }
+            }
+        }
+
         VkPhysicalDeviceProperties prop;
         vkGetPhysicalDeviceProperties( physdev, &prop );
         const float period = prop.limits.timestampPeriod;
@@ -82,21 +102,56 @@ public:
         vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE );
         vkQueueWaitIdle( queue );
 
-        vkBeginCommandBuffer( cmdbuf, &beginInfo );
+        int64_t tcpu, tgpu;
-        vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, m_query, 0 );
+        if( m_timeDomain == VK_TIME_DOMAIN_DEVICE_EXT )
-        vkEndCommandBuffer( cmdbuf );
+        {
-        vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE );
+            vkBeginCommandBuffer( cmdbuf, &beginInfo );
-        vkQueueWaitIdle( queue );
+            vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, m_query, 0 );
+            vkEndCommandBuffer( cmdbuf );
+            vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE );
+            vkQueueWaitIdle( queue );
 
-        int64_t tcpu = Profiler::GetTime();
+            tcpu = Profiler::GetTime();
-        int64_t tgpu;
+            vkGetQueryPoolResults( device, m_query, 0, 1, sizeof( tgpu ), &tgpu, sizeof( tgpu ), VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT );
-        vkGetQueryPoolResults( device, m_query, 0, 1, sizeof( tgpu ), &tgpu, sizeof( tgpu ), VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT );
 
-        vkBeginCommandBuffer( cmdbuf, &beginInfo );
+            vkBeginCommandBuffer( cmdbuf, &beginInfo );
-        vkCmdResetQueryPool( cmdbuf, m_query, 0, 1 );
+            vkCmdResetQueryPool( cmdbuf, m_query, 0, 1 );
-        vkEndCommandBuffer( cmdbuf );
+            vkEndCommandBuffer( cmdbuf );
-        vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE );
+            vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE );
-        vkQueueWaitIdle( queue );
+            vkQueueWaitIdle( queue );
+        }
+        else
+        {
+            enum { NumProbes = 32 };
+
+            VkCalibratedTimestampInfoEXT spec[2] = {
+                { VK_STRUCTURE_TYPE_CALIBRATED_TIMESTAMP_INFO_EXT, nullptr, VK_TIME_DOMAIN_DEVICE_EXT },
+                { VK_STRUCTURE_TYPE_CALIBRATED_TIMESTAMP_INFO_EXT, nullptr, m_timeDomain },
+            };
+            uint64_t ts[2];
+            uint64_t deviation[NumProbes];
+            for( int i=0; i<NumProbes; i++ )
+            {
+                _vkGetCalibratedTimestampsEXT( device, 2, spec, ts, deviation+i );
+            }
+            uint64_t minDeviation = deviation[0];
+            for( int i=1; i<NumProbes; i++ )
+            {
+                if( minDeviation > deviation[i] )
+                {
+                    minDeviation = deviation[i];
+                }
+            }
+            m_deviation = minDeviation * 3 / 2;
+
+            m_qpcToNs = int64_t( 1000000000. / GetFrequencyQpc() );
+
+            Calibrate( device, m_prevCalibration, tgpu );
+            tcpu = Profiler::GetTime();
+        }
+
+        uint8_t flags = 0;
+        if( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT ) flags |= GpuContextCalibration;
 
         auto item = Profiler::QueueSerial();
         MemWrite( &item->hdr.type, QueueType::GpuNewContext );
@@ -105,7 +160,7 @@ public:
         memset( &item->gpuNewContext.thread, 0, sizeof( item->gpuNewContext.thread ) );
         MemWrite( &item->gpuNewContext.period, period );
         MemWrite( &item->gpuNewContext.context, m_context );
-        MemWrite( &item->gpuNewContext.accuracyBits, uint8_t( 0 ) );
+        MemWrite( &item->gpuNewContext.flags, flags );
         MemWrite( &item->gpuNewContext.type, GpuContextType::Vulkan );
 
 #ifdef TRACY_ON_DEMAND
@@ -133,6 +188,8 @@ public:
         {
             vkCmdResetQueryPool( cmdbuf, m_query, 0, m_queryCount );
             m_head = m_tail = 0;
+            int64_t tgpu;
+            if( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT ) Calibrate( m_device, m_prevCalibration, tgpu );
             return;
         }
 #endif
@@ -164,6 +221,25 @@ public:
             Profiler::QueueSerialFinish();
         }
 
+        if( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT )
+        {
+            int64_t tgpu, tcpu;
+            Calibrate( m_device, tcpu, tgpu );
+            const auto refCpu = Profiler::GetTime();
+            const auto delta = tcpu - m_prevCalibration;
+            if( delta > 0 )
+            {
+                m_prevCalibration = tcpu;
+                auto item = Profiler::QueueSerial();
+                MemWrite( &item->hdr.type, QueueType::GpuCalibration );
+                MemWrite( &item->gpuCalibration.gpuTime, tgpu );
+                MemWrite( &item->gpuCalibration.cpuTime, refCpu );
+                MemWrite( &item->gpuCalibration.cpuDelta, delta );
+                MemWrite( &item->gpuCalibration.context, m_context );
+                Profiler::QueueSerialFinish();
+            }
+        }
+
         vkCmdResetQueryPool( cmdbuf, m_query, m_tail, cnt );
 
         m_tail += cnt;
@@ -184,8 +260,35 @@ private:
         return m_context;
     }
 
+    tracy_force_inline void Calibrate( VkDevice device, int64_t& tCpu, int64_t& tGpu )
+    {
+        assert( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT );
+        VkCalibratedTimestampInfoEXT spec[2] = {
+            { VK_STRUCTURE_TYPE_CALIBRATED_TIMESTAMP_INFO_EXT, nullptr, VK_TIME_DOMAIN_DEVICE_EXT },
+            { VK_STRUCTURE_TYPE_CALIBRATED_TIMESTAMP_INFO_EXT, nullptr, m_timeDomain },
+        };
+        uint64_t ts[2];
+        uint64_t deviation;
+        do
+        {
+            m_vkGetCalibratedTimestampsEXT( device, 2, spec, ts, &deviation );
+        }
+        while( deviation > m_deviation );
+
+#if defined _WIN32 || defined __CYGWIN__
+        tGpu = ts[0];
+        tCpu = ts[1] * m_qpcToNs;
+#else
+        assert( false );
+#endif
+    }
+
     VkDevice m_device;
     VkQueryPool m_query;
+    VkTimeDomainEXT m_timeDomain;
+    uint64_t m_deviation;
+    int64_t m_qpcToNs;
+    int64_t m_prevCalibration;
     uint8_t m_context;
 
     unsigned int m_head;
@@ -194,6 +297,8 @@ private:
     unsigned int m_queryCount;
 
     int64_t* m_res;
+
+    PFN_vkGetCalibratedTimestampsEXT m_vkGetCalibratedTimestampsEXT;
 };
 
 class VkCtxScope
@@ -237,6 +342,8 @@ public:
         const auto queryId = ctx->NextQueryId();
         vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, ctx->m_query, queryId );
 
+        GetProfiler().SendCallstack( depth );
+
         auto item = Profiler::QueueSerial();
         MemWrite( &item->hdr.type, QueueType::GpuZoneBeginCallstackSerial );
         MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() );
@@ -245,8 +352,6 @@ public:
         MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) );
         MemWrite( &item->gpuZoneBegin.context, ctx->GetId() );
         Profiler::QueueSerialFinish();
-
-        GetProfiler().SendCallstack( depth );
     }
 
     tracy_force_inline ~VkCtxScope()
@@ -272,11 +377,11 @@ private:
     VkCtx* m_ctx;
 };
 
-static inline VkCtx* CreateVkContext( VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf )
+static inline VkCtx* CreateVkContext( VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf, PFN_vkGetPhysicalDeviceCalibrateableTimeDomainsEXT gpdctd, PFN_vkGetCalibratedTimestampsEXT gct )
 {
     InitRPMallocThread();
     auto ctx = (VkCtx*)tracy_malloc( sizeof( VkCtx ) );
-    new(ctx) VkCtx( physdev, device, queue, cmdbuf );
+    new(ctx) VkCtx( physdev, device, queue, cmdbuf, gpdctd, gct );
     return ctx;
 }
 
@@ -290,24 +395,25 @@ static inline void DestroyVkContext( VkCtx* ctx )
 
 using TracyVkCtx = tracy::VkCtx*;
 
-#define TracyVkContext( physdev, device, queue, cmdbuf ) tracy::CreateVkContext( physdev, device, queue, cmdbuf );
+#define TracyVkContext( physdev, device, queue, cmdbuf ) tracy::CreateVkContext( physdev, device, queue, cmdbuf, nullptr, nullptr );
+#define TracyVkContextCalibrated( physdev, device, queue, cmdbuf, gpdctd, gct ) tracy::CreateVkContext( physdev, device, queue, cmdbuf, gpdctd, gct );
 #define TracyVkDestroy( ctx ) tracy::DestroyVkContext( ctx );
 #if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK
-#  define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, TRACY_CALLSTACK, active );
+#  define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, TRACY_CALLSTACK, active );
-#  define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, TRACY_CALLSTACK, active );
+#  define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, TRACY_CALLSTACK, active );
 #  define TracyVkZone( ctx, cmdbuf, name ) TracyVkNamedZoneS( ctx, ___tracy_gpu_zone, cmdbuf, name, TRACY_CALLSTACK, true )
 #  define TracyVkZoneC( ctx, cmdbuf, name, color ) TracyVkNamedZoneCS( ctx, ___tracy_gpu_zone, cmdbuf, name, color, TRACY_CALLSTACK, true )
 #else
-#  define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, active );
+#  define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, active );
-#  define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, active );
+#  define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, active );
 #  define TracyVkZone( ctx, cmdbuf, name ) TracyVkNamedZone( ctx, ___tracy_gpu_zone, cmdbuf, name, true )
 #  define TracyVkZoneC( ctx, cmdbuf, name, color ) TracyVkNamedZoneC( ctx, ___tracy_gpu_zone, cmdbuf, name, color, true )
 #endif
 #define TracyVkCollect( ctx, cmdbuf ) ctx->Collect( cmdbuf );
 
 #ifdef TRACY_HAS_CALLSTACK
-#  define TracyVkNamedZoneS( ctx, varname, cmdbuf, name, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, depth, active );
+#  define TracyVkNamedZoneS( ctx, varname, cmdbuf, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, depth, active );
-#  define TracyVkNamedZoneCS( ctx, varname, cmdbuf, name, color, depth, active ) static const tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, depth, active );
+#  define TracyVkNamedZoneCS( ctx, varname, cmdbuf, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,__LINE__) { name, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,__LINE__), cmdbuf, depth, active );
 #  define TracyVkZoneS( ctx, cmdbuf, name, depth ) TracyVkNamedZoneS( ctx, ___tracy_gpu_zone, cmdbuf, name, depth, true )
 #  define TracyVkZoneCS( ctx, cmdbuf, name, color, depth ) TracyVkNamedZoneCS( ctx, ___tracy_gpu_zone, cmdbuf, name, color, depth, true )
 #else

deps/tracy/client/TracyArmCpuTable.hpp

@@ -1,10 +1,8 @@
-#ifdef _MSC_VER
-#  pragma warning(disable:4996)
-#endif
-
 namespace tracy
 {
 
+#if defined __linux__ && defined __ARM_ARCH
+
 static const char* DecodeArmImplementer( uint32_t v )
 {
     static char buf[16];
@@ -16,6 +14,7 @@ static const char* DecodeArmImplementer( uint32_t v )
     case 0x44: return "DEC";
     case 0x46: return "Fujitsu";
     case 0x48: return "HiSilicon";
+    case 0x49: return "Infineon";
     case 0x4d: return "Motorola";
     case 0x4e: return "Nvidia";
     case 0x50: return "Applied Micro";
@@ -27,6 +26,7 @@ static const char* DecodeArmImplementer( uint32_t v )
     case 0x66: return "Faraday";
     case 0x68: return "HXT";
     case 0x69: return "Intel";
+    case 0xc0: return "Ampere Computing";
     default: break;
     }
     sprintf( buf, "0x%x", v );
@@ -75,6 +75,7 @@ static const char* DecodeArmPart( uint32_t impl, uint32_t part )
         case 0xc60: return " Cortex-M0+";
         case 0xd00: return " AArch64 simulator";
         case 0xd01: return " Cortex-A32";
+        case 0xd02: return " Cortex-A34";
         case 0xd03: return " Cortex-A53";
         case 0xd04: return " Cortex-A35";
         case 0xd05: return " Cortex-A55";
@@ -91,6 +92,10 @@ static const char* DecodeArmPart( uint32_t impl, uint32_t part )
         case 0xd13: return " Cortex-R52";
         case 0xd20: return " Cortex-M23";
         case 0xd21: return " Cortex-M33";
+        case 0xd40: return " Zeus";
+        case 0xd41: return " Cortex-A78";
+        case 0xd43: return " Cortex-A65AE";
+        case 0xd44: return " Cortex-X1";
         case 0xd4a: return " Neoverse E1";
         default: break;
         }
@@ -110,6 +115,13 @@ static const char* DecodeArmPart( uint32_t impl, uint32_t part )
         case 0xa2: return " ThunderX 81XX";
         case 0xa3: return " ThunderX 83XX";
         case 0xaf: return " ThunderX2 99xx";
+        case 0xb0: return " OcteonTX2";
+        case 0xb1: return " OcteonTX2 T98";
+        case 0xb2: return " OcteonTX2 T96";
+        case 0xb3: return " OcteonTX2 F95";
+        case 0xb4: return " OcteonTX2 F95N";
+        case 0xb5: return " OcteonTX2 F95MM";
+        case 0xb8: return " ThunderX3 T110";
         default: break;
         }
     case 0x44:
@@ -212,6 +224,8 @@ static const char* DecodeArmPart( uint32_t impl, uint32_t part )
     return buf;
 }
 
+#elif defined __APPLE__ && TARGET_OS_IPHONE == 1
+
 static const char* DecodeIosDevice( const char* id )
 {
     static const char* DeviceTable[] = {
@@ -252,6 +266,7 @@ static const char* DecodeIosDevice( const char* id )
         "iPhone12,1", "iPhone 11",
         "iPhone12,3", "iPhone 11 Pro",
         "iPhone12,5", "iPhone 11 Pro Max",
+        "iPhone12,8", "iPhone SE 2nd Gen",
         "iPad1,1", "iPad (A1219/A1337)",
         "iPad2,1", "iPad 2 (A1395)",
         "iPad2,2", "iPad 2 (A1396)",
@@ -302,6 +317,10 @@ static const char* DecodeIosDevice( const char* id )
         "iPad8,6", "iPad Pro 12.9\" 3rd gen (A1876)",
         "iPad8,7", "iPad Pro 12.9\" 3rd gen (A1895/A1983/A2014)",
         "iPad8,8", "iPad Pro 12.9\" 3rd gen (A1895/A1983/A2014)",
+        "iPad8,9", "iPad Pro 11\" 2nd gen (Wifi)",
+        "iPad8,10", "iPad Pro 11\" 2nd gen (Wifi+Cellular)",
+        "iPad8,11", "iPad Pro 12.9\" 4th gen (Wifi)",
+        "iPad8,12", "iPad Pro 12.9\" 4th gen (Wifi+Cellular)",
         "iPad11,1", "iPad Mini 5th gen (A2133)",
         "iPad11,2", "iPad Mini 5th gen (A2124/A2125/A2126)",
         "iPad11,3", "iPad Air 3rd gen (A2152)",
@@ -325,4 +344,6 @@ static const char* DecodeIosDevice( const char* id )
     return id;
 }
 
+#endif
+
 }

deps/tracy/client/TracyCallstack.cpp

@@ -222,9 +222,9 @@ static const char* GetModuleName( uint64_t addr )
     return "[unknown]";
 }
 
-SymbolData DecodeSymbolAddress( uint64_t ptr )
+CallstackSymbolData DecodeSymbolAddress( uint64_t ptr )
 {
-    SymbolData sym;
+    CallstackSymbolData sym;
     IMAGEHLP_LINE64 line;
     DWORD displacement = 0;
     line.SizeOfStruct = sizeof(IMAGEHLP_LINE64);
@@ -242,9 +242,9 @@ SymbolData DecodeSymbolAddress( uint64_t ptr )
     return sym;
 }
 
-SymbolData DecodeCodeAddress( uint64_t ptr )
+CallstackSymbolData DecodeCodeAddress( uint64_t ptr )
 {
-    SymbolData sym;
+    CallstackSymbolData sym;
     const auto proc = GetCurrentProcess();
     bool done = false;
 
@@ -442,14 +442,10 @@ const char* DecodeCallstackPtrFast( uint64_t ptr )
 
 static int SymbolAddressDataCb( void* data, uintptr_t pc, uintptr_t lowaddr, const char* fn, int lineno, const char* function )
 {
-    auto& sym = *(SymbolData*)data;
+    auto& sym = *(CallstackSymbolData*)data;
     if( !fn )
     {
-        const char* symloc = nullptr;
+        sym.file = "[unknown]";
-        Dl_info dlinfo;
-        if( dladdr( (void*)pc, &dlinfo ) ) symloc = dlinfo.dli_fname;
-        if( !symloc ) symloc = "[unknown]";
-        sym.file = symloc;
         sym.line = 0;
         sym.needFree = false;
     }
@@ -465,20 +461,20 @@ static int SymbolAddressDataCb( void* data, uintptr_t pc, uintptr_t lowaddr, con
 
 static void SymbolAddressErrorCb( void* data, const char* /*msg*/, int /*errnum*/ )
 {
-    auto& sym = *(SymbolData*)data;
+    auto& sym = *(CallstackSymbolData*)data;
     sym.file = "[unknown]";
     sym.line = 0;
     sym.needFree = false;
 }
 
-SymbolData DecodeSymbolAddress( uint64_t ptr )
+CallstackSymbolData DecodeSymbolAddress( uint64_t ptr )
 {
-    SymbolData sym;
+    CallstackSymbolData sym;
     backtrace_pcinfo( cb_bts, ptr, SymbolAddressDataCb, SymbolAddressErrorCb, &sym );
     return sym;
 }
 
-SymbolData DecodeCodeAddress( uint64_t ptr )
+CallstackSymbolData DecodeCodeAddress( uint64_t ptr )
 {
     return DecodeSymbolAddress( ptr );
 }
@@ -494,14 +490,12 @@ static int CallstackDataCb( void* /*data*/, uintptr_t pc, uintptr_t lowaddr, con
     if( !fn && !function )
     {
         const char* symname = nullptr;
-        const char* symloc = nullptr;
         auto vptr = (void*)pc;
         ptrdiff_t symoff = 0;
 
         Dl_info dlinfo;
         if( dladdr( vptr, &dlinfo ) )
         {
-            symloc = dlinfo.dli_fname;
             symname = dlinfo.dli_sname;
             symoff = (char*)pc - (char*)dlinfo.dli_saddr;
 
@@ -518,7 +512,6 @@ static int CallstackDataCb( void* /*data*/, uintptr_t pc, uintptr_t lowaddr, con
         }
 
         if( !symname ) symname = "[unknown]";
-        if( !symloc ) symloc = "[unknown]";
 
         if( symoff == 0 )
         {
@@ -536,15 +529,7 @@ static int CallstackDataCb( void* /*data*/, uintptr_t pc, uintptr_t lowaddr, con
             cb_data[cb_num].name = name;
         }
 
-        char buf[32];
+        cb_data[cb_num].file = CopyString( "[unknown]" );
-        const auto addrlen = sprintf( buf, " [%p]", (void*)pc );
-        const auto loclen = strlen( symloc );
-        auto loc = (char*)tracy_malloc( loclen + addrlen + 1 );
-        memcpy( loc, symloc, loclen );
-        memcpy( loc + loclen, buf, addrlen );
-        loc[loclen + addrlen] = '\0';
-        cb_data[cb_num].file = loc;
-
         cb_data[cb_num].line = 0;
     }
     else
@@ -652,16 +637,16 @@ const char* DecodeCallstackPtrFast( uint64_t ptr )
     return ret;
 }
 
-SymbolData DecodeSymbolAddress( uint64_t ptr )
+CallstackSymbolData DecodeSymbolAddress( uint64_t ptr )
 {
     const char* symloc = nullptr;
     Dl_info dlinfo;
     if( dladdr( (void*)ptr, &dlinfo ) ) symloc = dlinfo.dli_fname;
     if( !symloc ) symloc = "[unknown]";
-    return SymbolData { symloc, 0, false };
+    return CallstackSymbolData { symloc, 0, false };
 }
 
-SymbolData DecodeCodeAddress( uint64_t ptr )
+CallstackSymbolData DecodeCodeAddress( uint64_t ptr )
 {
     return DecodeSymbolAddress( ptr );
 }
@@ -717,15 +702,7 @@ CallstackEntryData DecodeCallstackPtr( uint64_t ptr )
         cb.name = name;
     }
 
-    char buf[32];
+    cb.file = CopyString( "[unknown]" );
-    const auto addrlen = sprintf( buf, " [%p]", (void*)ptr );
-    const auto loclen = strlen( symloc );
-    auto loc = (char*)tracy_malloc( loclen + addrlen + 1 );
-    memcpy( loc, symloc, loclen );
-    memcpy( loc + loclen, buf, addrlen );
-    loc[loclen + addrlen] = '\0';
-    cb.file = loc;
-
     cb.symLen = 0;
     cb.symAddr = (uint64_t)symaddr;
 

deps/tracy/client/TracyCallstack.hpp

@@ -22,7 +22,7 @@
 namespace tracy
 {
 
-struct SymbolData
+struct CallstackSymbolData
 {
     const char* file;
     uint32_t line;
@@ -45,8 +45,8 @@ struct CallstackEntryData
     const char* imageName;
 };
 
-SymbolData DecodeSymbolAddress( uint64_t ptr );
+CallstackSymbolData DecodeSymbolAddress( uint64_t ptr );
-SymbolData DecodeCodeAddress( uint64_t ptr );
+CallstackSymbolData DecodeCodeAddress( uint64_t ptr );
 const char* DecodeCallstackPtrFast( uint64_t ptr );
 CallstackEntryData DecodeCallstackPtr( uint64_t ptr );
 void InitCallstack();

deps/tracy/client/TracyDxt1.cpp

@@ -172,6 +172,12 @@ static tracy_force_inline uint64_t ProcessRGB( const uint8_t* src )
         return uint64_t( to565( src[0], src[1], src[2] ) ) << 16;
     }
 
+    __m128i amask = _mm_set1_epi32( 0xFFFFFF );
+    px0 = _mm_and_si128( px0, amask );
+    px1 = _mm_and_si128( px1, amask );
+    px2 = _mm_and_si128( px2, amask );
+    px3 = _mm_and_si128( px3, amask );
+
     __m128i min0 = _mm_min_epu8( px0, px1 );
     __m128i min1 = _mm_min_epu8( px2, px3 );
     __m128i min2 = _mm_min_epu8( min0, min1 );
@@ -492,6 +498,12 @@ static tracy_force_inline void ProcessRGB_AVX( const uint8_t* src, char*& dst )
         return;
     }
 
+    __m256i amask = _mm256_set1_epi32( 0xFFFFFF );
+    px0 = _mm256_and_si256( px0, amask );
+    px1 = _mm256_and_si256( px1, amask );
+    px2 = _mm256_and_si256( px2, amask );
+    px3 = _mm256_and_si256( px3, amask );
+
     __m256i min0 = _mm256_min_epu8( px0, px1 );
     __m256i min1 = _mm256_min_epu8( px2, px3 );
     __m256i min2 = _mm256_min_epu8( min0, min1 );

deps/tracy/client/TracyLock.hpp

@@ -23,7 +23,8 @@ public:
     {
         assert( m_id != std::numeric_limits<uint32_t>::max() );
 
-        TracyLfqPrepare( QueueType::LockAnnounce );
+        auto item = Profiler::QueueSerial();
+        MemWrite( &item->hdr.type, QueueType::LockAnnounce );
         MemWrite( &item->lockAnnounce.id, m_id );
         MemWrite( &item->lockAnnounce.time, Profiler::GetTime() );
         MemWrite( &item->lockAnnounce.lckloc, (uint64_t)srcloc );
@@ -31,7 +32,7 @@ public:
 #ifdef TRACY_ON_DEMAND
         GetProfiler().DeferItem( *item );
 #endif
-        TracyLfqCommit;
+        Profiler::QueueSerialFinish();
     }
 
     LockableCtx( const LockableCtx& ) = delete;
@@ -39,14 +40,14 @@ public:
 
     tracy_force_inline ~LockableCtx()
     {
-        TracyLfqPrepare( QueueType::LockTerminate );
+        auto item = Profiler::QueueSerial();
+        MemWrite( &item->hdr.type, QueueType::LockTerminate );
         MemWrite( &item->lockTerminate.id, m_id );
         MemWrite( &item->lockTerminate.time, Profiler::GetTime() );
-        MemWrite( &item->lockTerminate.type, LockType::Lockable );
 #ifdef TRACY_ON_DEMAND
         GetProfiler().DeferItem( *item );
 #endif
-        TracyLfqCommit;
+        Profiler::QueueSerialFinish();
     }
 
     tracy_force_inline bool BeforeLock()
@@ -69,7 +70,6 @@ public:
         MemWrite( &item->lockWait.thread, GetThreadHandle() );
         MemWrite( &item->lockWait.id, m_id );
         MemWrite( &item->lockWait.time, Profiler::GetTime() );
-        MemWrite( &item->lockWait.type, LockType::Lockable );
         Profiler::QueueSerialFinish();
         return true;
     }
@@ -155,16 +155,18 @@ public:
 
     tracy_force_inline void CustomName( const char* name, size_t size )
     {
-        auto ptr = (char*)tracy_malloc( size+1 );
+        assert( size < std::numeric_limits<uint16_t>::max() );
+        auto ptr = (char*)tracy_malloc( size );
         memcpy( ptr, name, size );
-        ptr[size] = '\0';
+        auto item = Profiler::QueueSerial();
-        TracyLfqPrepare( QueueType::LockName );
+        MemWrite( &item->hdr.type, QueueType::LockName );
-        MemWrite( &item->lockName.id, m_id );
+        MemWrite( &item->lockNameFat.id, m_id );
-        MemWrite( &item->lockName.name, (uint64_t)ptr );
+        MemWrite( &item->lockNameFat.name, (uint64_t)ptr );
+        MemWrite( &item->lockNameFat.size, (uint16_t)size );
 #ifdef TRACY_ON_DEMAND
         GetProfiler().DeferItem( *item );
 #endif
-        TracyLfqCommit;
+        Profiler::QueueSerialFinish();
     }
 
 private:
@@ -236,17 +238,16 @@ public:
     {
         assert( m_id != std::numeric_limits<uint32_t>::max() );
 
-        TracyLfqPrepare( QueueType::LockAnnounce );
+        auto item = Profiler::QueueSerial();
+        MemWrite( &item->hdr.type, QueueType::LockAnnounce );
         MemWrite( &item->lockAnnounce.id, m_id );
         MemWrite( &item->lockAnnounce.time, Profiler::GetTime() );
         MemWrite( &item->lockAnnounce.lckloc, (uint64_t)srcloc );
         MemWrite( &item->lockAnnounce.type, LockType::SharedLockable );
-
 #ifdef TRACY_ON_DEMAND
         GetProfiler().DeferItem( *item );
 #endif
-
+        Profiler::QueueSerialFinish();
-        TracyLfqCommit;
     }
 
     SharedLockableCtx( const SharedLockableCtx& ) = delete;
@@ -254,16 +255,14 @@ public:
 
     tracy_force_inline ~SharedLockableCtx()
     {
-        TracyLfqPrepare( QueueType::LockTerminate );
+        auto item = Profiler::QueueSerial();
+        MemWrite( &item->hdr.type, QueueType::LockTerminate );
         MemWrite( &item->lockTerminate.id, m_id );
         MemWrite( &item->lockTerminate.time, Profiler::GetTime() );
-        MemWrite( &item->lockTerminate.type, LockType::SharedLockable );
-
 #ifdef TRACY_ON_DEMAND
         GetProfiler().DeferItem( *item );
 #endif
-
+        Profiler::QueueSerialFinish();
-        TracyLfqCommit;
     }
 
     tracy_force_inline bool BeforeLock()
@@ -286,7 +285,6 @@ public:
         MemWrite( &item->lockWait.thread, GetThreadHandle() );
         MemWrite( &item->lockWait.id, m_id );
         MemWrite( &item->lockWait.time, Profiler::GetTime() );
-        MemWrite( &item->lockWait.type, LockType::SharedLockable );
         Profiler::QueueSerialFinish();
         return true;
     }
@@ -369,7 +367,6 @@ public:
         MemWrite( &item->lockWait.thread, GetThreadHandle() );
         MemWrite( &item->lockWait.id, m_id );
         MemWrite( &item->lockWait.time, Profiler::GetTime() );
-        MemWrite( &item->lockWait.type, LockType::SharedLockable );
         Profiler::QueueSerialFinish();
         return true;
     }
@@ -455,16 +452,18 @@ public:
 
     tracy_force_inline void CustomName( const char* name, size_t size )
     {
-        auto ptr = (char*)tracy_malloc( size+1 );
+        assert( size < std::numeric_limits<uint16_t>::max() );
+        auto ptr = (char*)tracy_malloc( size );
         memcpy( ptr, name, size );
-        ptr[size] = '\0';
+        auto item = Profiler::QueueSerial();
-        TracyLfqPrepare( QueueType::LockName );
+        MemWrite( &item->hdr.type, QueueType::LockName );
-        MemWrite( &item->lockName.id, m_id );
+        MemWrite( &item->lockNameFat.id, m_id );
-        MemWrite( &item->lockName.name, (uint64_t)ptr );
+        MemWrite( &item->lockNameFat.name, (uint64_t)ptr );
+        MemWrite( &item->lockNameFat.size, (uint16_t)size );
 #ifdef TRACY_ON_DEMAND
         GetProfiler().DeferItem( *item );
 #endif
-        TracyLfqCommit;
+        Profiler::QueueSerialFinish();
     }
 
 private:

deps/tracy/client/TracyProfiler.cpp

@@ -37,6 +37,10 @@
 #  include <sys/sysctl.h>
 #endif
 
+#if defined __APPLE__
+#  include "TargetConditionals.h"
+#endif
+
 #include <algorithm>
 #include <assert.h>
 #include <atomic>
@@ -80,11 +84,6 @@
 #  endif
 #endif
 
-#if defined TRACY_HW_TIMER && __ARM_ARCH >= 6 && !defined TARGET_OS_IOS
-#  include <signal.h>
-#  include <setjmp.h>
-#endif
-
 #if defined _WIN32 || defined __CYGWIN__
 #  include <lmcons.h>
 extern "C" typedef LONG (WINAPI *t_RtlGetVersion)( PRTL_OSVERSIONINFOW );
@@ -93,9 +92,6 @@ extern "C" typedef BOOL (WINAPI *t_GetLogicalProcessorInformationEx)( LOGICAL_PR
 #  include <unistd.h>
 #  include <limits.h>
 #endif
-#if defined __APPLE__
-#  include "TargetConditionals.h"
-#endif
 #if defined __linux__
 #  include <sys/sysinfo.h>
 #  include <sys/utsname.h>
@@ -231,7 +227,11 @@ static int64_t SetupHwTimer()
         const char* noCheck = getenv( "TRACY_NO_INVARIANT_CHECK" );
         if( !noCheck || noCheck[0] != '1' )
         {
+#if defined _WIN32 || defined __CYGWIN__
             InitFailure( "CPU doesn't support invariant TSC.\nDefine TRACY_NO_INVARIANT_CHECK=1 to ignore this error, *if you know what you are doing*.\nAlternatively you may rebuild the application with the TRACY_TIMER_QPC define to use lower resolution timer." );
+#else
+            InitFailure( "CPU doesn't support invariant TSC.\nDefine TRACY_NO_INVARIANT_CHECK=1 to ignore this error, *if you know what you are doing*." );
+#endif
         }
     }
 #endif
@@ -270,6 +270,7 @@ static const char* GetProcessName()
     return processName;
 }
 
+#if defined __linux__ && defined __ARM_ARCH
 static uint32_t GetHex( char*& ptr, int skip )
 {
     uint32_t ret;
@@ -287,6 +288,7 @@ static uint32_t GetHex( char*& ptr, int skip )
     ptr = end;
     return ret;
 }
+#endif
 
 static const char* GetHostInfo()
 {
@@ -540,6 +542,8 @@ static char s_crashText[1024];
 
 LONG WINAPI CrashFilter( PEXCEPTION_POINTERS pExp )
 {
+    if( !GetProfiler().IsConnected() ) return EXCEPTION_CONTINUE_SEARCH;
+
     const unsigned ec = pExp->ExceptionRecord->ExceptionCode;
     auto msgPtr = s_crashText;
     switch( ec )
@@ -590,12 +594,12 @@ LONG WINAPI CrashFilter( PEXCEPTION_POINTERS pExp )
     }
 
     {
+        GetProfiler().SendCallstack( 60, "KiUserExceptionDispatcher" );
+
         TracyLfqPrepare( QueueType::CrashReport );
         item->crashReport.time = Profiler::GetTime();
         item->crashReport.text = (uint64_t)s_crashText;
         TracyLfqCommit;
-
-        GetProfiler().SendCallstack( 60, "KiUserExceptionDispatcher" );
     }
 
     HANDLE h = CreateToolhelp32Snapshot( TH32CS_SNAPTHREAD, 0 );
@@ -822,12 +826,12 @@ static void CrashHandler( int signal, siginfo_t* info, void* /*ucontext*/ )
     }
 
     {
+        GetProfiler().SendCallstack( 60, "__kernel_rt_sigreturn" );
+
         TracyLfqPrepare( QueueType::CrashReport );
         item->crashReport.time = Profiler::GetTime();
         item->crashReport.text = (uint64_t)s_crashText;
         TracyLfqCommit;
-
-        GetProfiler().SendCallstack( 60, "__kernel_rt_sigreturn" );
     }
 
     DIR* dp = opendir( "/proc/self/task" );
@@ -863,7 +867,7 @@ static void CrashHandler( int signal, siginfo_t* info, void* /*ucontext*/ )
 
 enum { QueuePrealloc = 256 * 1024 };
 
-static Profiler* s_instance;
+static Profiler* s_instance = nullptr;
 static Thread* s_thread;
 static Thread* s_compressThread;
 
@@ -871,6 +875,19 @@ static Thread* s_compressThread;
 static Thread* s_sysTraceThread = nullptr;
 #endif
 
+TRACY_API bool ProfilerAvailable() { return s_instance != nullptr; }
+
+TRACY_API int64_t GetFrequencyQpc()
+{
+#if defined _WIN32 || defined __CYGWIN__
+    LARGE_INTEGER t;
+    QueryPerformanceFrequency( &t );
+    return t.QuadPart;
+#else
+    return 0;
+#endif
+}
+
 #ifdef TRACY_DELAYED_INIT
 struct ThreadNameData;
 TRACY_API moodycamel::ConcurrentQueue<QueueItem>& GetQueue();
@@ -911,6 +928,25 @@ struct ProfilerThreadData
 #  endif
 };
 
+#  ifdef TRACY_MANUAL_LIFETIME
+ProfilerData* s_profilerData = nullptr;
+TRACY_API void StartupProfiler()
+{
+    s_profilerData = new ProfilerData;
+    s_profilerData->profiler.SpawnWorkerThreads();
+}
+static ProfilerData& GetProfilerData()
+{
+    assert(s_profilerData);
+    return *s_profilerData;
+}
+TRACY_API void ShutdownProfiler()
+{
+    delete s_profilerData;
+    s_profilerData = nullptr;
+    rpmalloc_finalize();
+}
+#  else
 static std::atomic<int> profilerDataLock { 0 };
 static std::atomic<ProfilerData*> profilerData { nullptr };
 
@@ -932,6 +968,7 @@ static ProfilerData& GetProfilerData()
     }
     return *ptr;
 }
+#  endif
 
 static ProfilerThreadData& GetProfilerThreadData()
 {
@@ -953,10 +990,12 @@ std::atomic<ThreadNameData*>& GetThreadNameData() { return GetProfilerData().thr
 TRACY_API LuaZoneState& GetLuaZoneState() { return GetProfilerThreadData().luaZoneState; }
 #  endif
 
+#  ifndef TRACY_MANUAL_LIFETIME
 namespace
 {
     const auto& __profiler_init = GetProfiler();
 }
+#  endif
 
 #else
 TRACY_API void InitRPMallocThread()
@@ -1044,8 +1083,8 @@ Profiler::Profiler()
     , m_fiQueue( 16 )
     , m_fiDequeue( 16 )
     , m_frameCount( 0 )
-#ifdef TRACY_ON_DEMAND
     , m_isConnected( false )
+#ifdef TRACY_ON_DEMAND
     , m_connectionId( 0 )
     , m_deferredQueue( 64*1024 )
 #endif
@@ -1081,6 +1120,13 @@ Profiler::Profiler()
         m_userPort = atoi( userPort );
     }
 
+#if !defined(TRACY_DELAYED_INIT) || !defined(TRACY_MANUAL_LIFETIME)
+    SpawnWorkerThreads();
+#endif
+}
+
+void Profiler::SpawnWorkerThreads()
+{
     s_thread = (Thread*)tracy_malloc( sizeof( Thread ) );
     new(s_thread) Thread( LaunchWorker, this );
 
@@ -1172,6 +1218,8 @@ void Profiler::Worker()
     s_profilerTid = syscall( SYS_gettid );
 #endif
 
+    ThreadExitHandler threadExitHandler;
+
     SetThreadName( "Tracy Profiler" );
 
 #ifdef TRACY_DATA_PORT
@@ -1223,6 +1271,12 @@ void Profiler::Worker()
     uint8_t cpuArch = CpuArchUnknown;
 #endif
 
+#ifdef TRACY_NO_CODE_TRANSFER
+    uint8_t codeTransfer = 0;
+#else
+    uint8_t codeTransfer = 1;
+#endif
+
 #if defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64
     uint32_t regs[4];
     char manufacturer[12];
@@ -1250,6 +1304,7 @@ void Profiler::Worker()
     MemWrite( &welcome.onDemand, onDemand );
     MemWrite( &welcome.isApple, isApple );
     MemWrite( &welcome.cpuArch, cpuArch );
+    MemWrite( &welcome.codeTransfer, codeTransfer );
     memcpy( welcome.cpuManufacturer, manufacturer, 12 );
     MemWrite( &welcome.cpuId, cpuId );
     memcpy( welcome.programName, procname, pnsz );
@@ -1318,6 +1373,11 @@ void Profiler::Worker()
 #ifndef TRACY_NO_EXIT
             if( !m_noExit && ShouldExit() )
             {
+                if( m_broadcast )
+                {
+                    broadcastMsg.activeTime = -1;
+                    m_broadcast->Send( broadcastPort, &broadcastMsg, broadcastLen );
+                }
                 m_shutdownFinished.store( true, std::memory_order_relaxed );
                 return;
             }
@@ -1335,12 +1395,20 @@ void Profiler::Worker()
                 {
                     lastBroadcast = t;
                     const auto ts = std::chrono::duration_cast<std::chrono::seconds>( std::chrono::system_clock::now().time_since_epoch() ).count();
-                    broadcastMsg.activeTime = uint32_t( ts - m_epoch );
+                    broadcastMsg.activeTime = int32_t( ts - m_epoch );
+                    assert( broadcastMsg.activeTime >= 0 );
                     m_broadcast->Send( broadcastPort, &broadcastMsg, broadcastLen );
                 }
             }
         }
 
+        if( m_broadcast )
+        {
+            lastBroadcast = 0;
+            broadcastMsg.activeTime = -1;
+            m_broadcast->Send( broadcastPort, &broadcastMsg, broadcastLen );
+        }
+
         // Handshake
         {
             char shibboleth[HandshakeShibbolethSize];
@@ -1378,8 +1446,8 @@ void Profiler::Worker()
         const auto currentTime = GetTime();
         ClearQueues( token );
         m_connectionId.fetch_add( 1, std::memory_order_release );
-        m_isConnected.store( true, std::memory_order_release );
 #endif
+        m_isConnected.store( true, std::memory_order_release );
 
         HandshakeStatus handshake = HandshakeWelcome;
         m_sock->Send( &handshake, sizeof( handshake ) );
@@ -1403,16 +1471,19 @@ void Profiler::Worker()
         for( auto& item : m_deferredQueue )
         {
             uint64_t ptr;
+            uint16_t size;
             const auto idx = MemRead<uint8_t>( &item.hdr.idx );
             switch( (QueueType)idx )
             {
             case QueueType::MessageAppInfo:
-                ptr = MemRead<uint64_t>( &item.message.text );
+                ptr = MemRead<uint64_t>( &item.messageFat.text );
-                SendString( ptr, (const char*)ptr, QueueType::CustomStringData );
+                size = MemRead<uint16_t>( &item.messageFat.size );
+                SendSingleString( (const char*)ptr, size );
                 break;
             case QueueType::LockName:
-                ptr = MemRead<uint64_t>( &item.lockName.name );
+                ptr = MemRead<uint64_t>( &item.lockNameFat.name );
-                SendString( ptr, (const char*)ptr, QueueType::CustomStringData );
+                size = MemRead<uint16_t>( &item.lockNameFat.size );
+                SendSingleString( (const char*)ptr, size );
                 break;
             default:
                 break;
@@ -1469,8 +1540,8 @@ void Profiler::Worker()
         }
         if( ShouldExit() ) break;
 
-#ifdef TRACY_ON_DEMAND
         m_isConnected.store( false, std::memory_order_release );
+#ifdef TRACY_ON_DEMAND
         m_bufferOffset = 0;
         m_bufferStart = 0;
 #endif
@@ -1592,6 +1663,8 @@ void Profiler::Worker()
 
 void Profiler::CompressWorker()
 {
+    ThreadExitHandler threadExitHandler;
+
     SetThreadName( "Tracy DXT1" );
     while( m_timeBegin.load( std::memory_order_relaxed ) == 0 ) std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) );
     rpmalloc_thread_initialize();
@@ -1631,12 +1704,12 @@ void Profiler::CompressWorker()
                 tracy_free( fi->image );
 
                 TracyLfqPrepare( QueueType::FrameImage );
-                MemWrite( &item->frameImage.image, (uint64_t)etc1buf );
+                MemWrite( &item->frameImageFat.image, (uint64_t)etc1buf );
-                MemWrite( &item->frameImage.frame, fi->frame );
+                MemWrite( &item->frameImageFat.frame, fi->frame );
-                MemWrite( &item->frameImage.w, w );
+                MemWrite( &item->frameImageFat.w, w );
-                MemWrite( &item->frameImage.h, h );
+                MemWrite( &item->frameImageFat.h, h );
                 uint8_t flip = fi->flip;
-                MemWrite( &item->frameImage.flip, flip );
+                MemWrite( &item->frameImageFat.flip, flip );
                 TracyLfqCommit;
 
                 fi++;
@@ -1664,7 +1737,7 @@ static void FreeAssociatedMemory( const QueueItem& item )
     {
     case QueueType::ZoneText:
     case QueueType::ZoneName:
-        ptr = MemRead<uint64_t>( &item.zoneText.text );
+        ptr = MemRead<uint64_t>( &item.zoneTextFat.text );
         tracy_free( (void*)ptr );
         break;
     case QueueType::Message:
@@ -1674,7 +1747,7 @@ static void FreeAssociatedMemory( const QueueItem& item )
 #ifndef TRACY_ON_DEMAND
     case QueueType::MessageAppInfo:
 #endif
-        ptr = MemRead<uint64_t>( &item.message.text );
+        ptr = MemRead<uint64_t>( &item.messageFat.text );
         tracy_free( (void*)ptr );
         break;
     case QueueType::ZoneBeginAllocSrcLoc:
@@ -1683,30 +1756,27 @@ static void FreeAssociatedMemory( const QueueItem& item )
         tracy_free( (void*)ptr );
         break;
     case QueueType::CallstackMemory:
-        ptr = MemRead<uint64_t>( &item.callstackMemory.ptr );
-        tracy_free( (void*)ptr );
-        break;
     case QueueType::Callstack:
-        ptr = MemRead<uint64_t>( &item.callstack.ptr );
+        ptr = MemRead<uint64_t>( &item.callstackFat.ptr );
         tracy_free( (void*)ptr );
         break;
     case QueueType::CallstackAlloc:
-        ptr = MemRead<uint64_t>( &item.callstackAlloc.nativePtr );
+        ptr = MemRead<uint64_t>( &item.callstackAllocFat.nativePtr );
         tracy_free( (void*)ptr );
-        ptr = MemRead<uint64_t>( &item.callstackAlloc.ptr );
+        ptr = MemRead<uint64_t>( &item.callstackAllocFat.ptr );
         tracy_free( (void*)ptr );
         break;
     case QueueType::CallstackSample:
-        ptr = MemRead<uint64_t>( &item.callstackSample.ptr );
+        ptr = MemRead<uint64_t>( &item.callstackSampleFat.ptr );
         tracy_free( (void*)ptr );
         break;
     case QueueType::FrameImage:
-        ptr = MemRead<uint64_t>( &item.frameImage.image );
+        ptr = MemRead<uint64_t>( &item.frameImageFat.image );
         tracy_free( (void*)ptr );
         break;
 #ifndef TRACY_ON_DEMAND
     case QueueType::LockName:
-        ptr = MemRead<uint64_t>( &item.lockName.name );
+        ptr = MemRead<uint64_t>( &item.lockNameFat.name );
         tracy_free( (void*)ptr );
         break;
 #endif
@@ -1779,6 +1849,7 @@ Profiler::DequeueStatus Profiler::Dequeue( moodycamel::ConsumerToken& token )
             while( sz-- > 0 )
             {
                 uint64_t ptr;
+                uint16_t size;
                 auto idx = MemRead<uint8_t>( &item->hdr.idx );
                 if( idx < (int)QueueType::Terminate )
                 {
@@ -1786,21 +1857,29 @@ Profiler::DequeueStatus Profiler::Dequeue( moodycamel::ConsumerToken& token )
                     {
                     case QueueType::ZoneText:
                     case QueueType::ZoneName:
-                        ptr = MemRead<uint64_t>( &item->zoneText.text );
+                        ptr = MemRead<uint64_t>( &item->zoneTextFat.text );
-                        SendString( ptr, (const char*)ptr, QueueType::CustomStringData );
+                        size = MemRead<uint16_t>( &item->zoneTextFat.size );
+                        SendSingleString( (const char*)ptr, size );
                         tracy_free( (void*)ptr );
                         break;
                     case QueueType::Message:
-                    case QueueType::MessageColor:
                     case QueueType::MessageCallstack:
+                        ptr = MemRead<uint64_t>( &item->messageFat.text );
+                        size = MemRead<uint16_t>( &item->messageFat.size );
+                        SendSingleString( (const char*)ptr, size );
+                        tracy_free( (void*)ptr );
+                        break;
+                    case QueueType::MessageColor:
                     case QueueType::MessageColorCallstack:
-                        ptr = MemRead<uint64_t>( &item->message.text );
+                        ptr = MemRead<uint64_t>( &item->messageColorFat.text );
-                        SendString( ptr, (const char*)ptr, QueueType::CustomStringData );
+                        size = MemRead<uint16_t>( &item->messageColorFat.size );
+                        SendSingleString( (const char*)ptr, size );
                         tracy_free( (void*)ptr );
                         break;
                     case QueueType::MessageAppInfo:
-                        ptr = MemRead<uint64_t>( &item->message.text );
+                        ptr = MemRead<uint64_t>( &item->messageFat.text );
-                        SendString( ptr, (const char*)ptr, QueueType::CustomStringData );
+                        size = MemRead<uint16_t>( &item->messageFat.size );
+                        SendSingleString( (const char*)ptr, size );
 #ifndef TRACY_ON_DEMAND
                         tracy_free( (void*)ptr );
 #endif
@@ -1815,54 +1894,44 @@ Profiler::DequeueStatus Profiler::Dequeue( moodycamel::ConsumerToken& token )
                         ptr = MemRead<uint64_t>( &item->zoneBegin.srcloc );
                         SendSourceLocationPayload( ptr );
                         tracy_free( (void*)ptr );
-                        idx++;
-                        MemWrite( &item->hdr.idx, idx );
                         break;
                     }
                     case QueueType::Callstack:
-                        ptr = MemRead<uint64_t>( &item->callstack.ptr );
+                        ptr = MemRead<uint64_t>( &item->callstackFat.ptr );
                         SendCallstackPayload( ptr );
                         tracy_free( (void*)ptr );
-                        idx++;
-                        MemWrite( &item->hdr.idx, idx );
                         break;
                     case QueueType::CallstackAlloc:
-                        ptr = MemRead<uint64_t>( &item->callstackAlloc.nativePtr );
+                        ptr = MemRead<uint64_t>( &item->callstackAllocFat.nativePtr );
                         if( ptr != 0 )
                         {
                             CutCallstack( (void*)ptr, "lua_pcall" );
                             SendCallstackPayload( ptr );
                             tracy_free( (void*)ptr );
                         }
-                        ptr = MemRead<uint64_t>( &item->callstackAlloc.ptr );
+                        ptr = MemRead<uint64_t>( &item->callstackAllocFat.ptr );
                         SendCallstackAlloc( ptr );
                         tracy_free( (void*)ptr );
-                        idx++;
-                        MemWrite( &item->hdr.idx, idx );
                         break;
                     case QueueType::CallstackSample:
                     {
-                        ptr = MemRead<uint64_t>( &item->callstackSample.ptr );
+                        ptr = MemRead<uint64_t>( &item->callstackSampleFat.ptr );
                         SendCallstackPayload64( ptr );
                         tracy_free( (void*)ptr );
-                        int64_t t = MemRead<int64_t>( &item->callstackSample.time );
+                        int64_t t = MemRead<int64_t>( &item->callstackSampleFat.time );
                         int64_t dt = t - refCtx;
                         refCtx = t;
-                        MemWrite( &item->callstackSample.time, dt );
+                        MemWrite( &item->callstackSampleFat.time, dt );
-                        idx++;
-                        MemWrite( &item->hdr.idx, idx );
                         break;
                     }
                     case QueueType::FrameImage:
                     {
-                        ptr = MemRead<uint64_t>( &item->frameImage.image );
+                        ptr = MemRead<uint64_t>( &item->frameImageFat.image );
-                        const auto w = MemRead<uint16_t>( &item->frameImage.w );
+                        const auto w = MemRead<uint16_t>( &item->frameImageFat.w );
-                        const auto h = MemRead<uint16_t>( &item->frameImage.h );
+                        const auto h = MemRead<uint16_t>( &item->frameImageFat.h );
                         const auto csz = size_t( w * h / 2 );
                         SendLongString( ptr, (const char*)ptr, csz, QueueType::FrameImageData );
                         tracy_free( (void*)ptr );
-                        idx++;
-                        MemWrite( &item->hdr.idx, idx );
                         break;
                     }
                     case QueueType::ZoneBegin:
@@ -1882,13 +1951,6 @@ Profiler::DequeueStatus Profiler::Dequeue( moodycamel::ConsumerToken& token )
                         MemWrite( &item->zoneEnd.time, dt );
                         break;
                     }
-                    case QueueType::LockName:
-                        ptr = MemRead<uint64_t>( &item->lockName.name );
-                        SendString( ptr, (const char*)ptr, QueueType::CustomStringData );
-#ifndef TRACY_ON_DEMAND
-                        tracy_free( (void*)ptr );
-#endif
-                        break;
                     case QueueType::GpuZoneBegin:
                     case QueueType::GpuZoneBeginCallstack:
                     {
@@ -2056,11 +2118,9 @@ Profiler::DequeueStatus Profiler::DequeueSerial()
                 switch( (QueueType)idx )
                 {
                 case QueueType::CallstackMemory:
-                    ptr = MemRead<uint64_t>( &item->callstackMemory.ptr );
+                    ptr = MemRead<uint64_t>( &item->callstackFat.ptr );
                     SendCallstackPayload( ptr );
                     tracy_free( (void*)ptr );
-                    idx++;
-                    MemWrite( &item->hdr.idx, idx );
                     break;
                 case QueueType::LockWait:
                 case QueueType::LockSharedWait:
@@ -2089,8 +2149,20 @@ Profiler::DequeueStatus Profiler::DequeueSerial()
                     MemWrite( &item->lockRelease.time, dt );
                     break;
                 }
+                case QueueType::LockName:
+                {
+                    ptr = MemRead<uint64_t>( &item->lockNameFat.name );
+                    uint16_t size = MemRead<uint16_t>( &item->lockNameFat.size );
+                    SendSingleString( (const char*)ptr, size );
+#ifndef TRACY_ON_DEMAND
+                    tracy_free( (void*)ptr );
+#endif
+                    break;
+                }
                 case QueueType::MemAlloc:
+                case QueueType::MemAllocNamed:
                 case QueueType::MemAllocCallstack:
+                case QueueType::MemAllocCallstackNamed:
                 {
                     int64_t t = MemRead<int64_t>( &item->memAlloc.time );
                     int64_t dt = t - refSerial;
@@ -2099,7 +2171,9 @@ Profiler::DequeueStatus Profiler::DequeueSerial()
                     break;
                 }
                 case QueueType::MemFree:
+                case QueueType::MemFreeNamed:
                 case QueueType::MemFreeCallstack:
+                case QueueType::MemFreeCallstackNamed:
                 {
                     int64_t t = MemRead<int64_t>( &item->memFree.time );
                     int64_t dt = t - refSerial;
@@ -2166,11 +2240,10 @@ bool Profiler::SendData( const char* data, size_t len )
     return m_sock->Send( m_lz4Buf, lz4sz + sizeof( lz4sz_t ) ) != -1;
 }
 
-void Profiler::SendString( uint64_t str, const char* ptr, QueueType type )
+void Profiler::SendString( uint64_t str, const char* ptr, size_t len, QueueType type )
 {
     assert( type == QueueType::StringData ||
             type == QueueType::ThreadName ||
-            type == QueueType::CustomStringData ||
             type == QueueType::PlotName ||
             type == QueueType::FrameName ||
             type == QueueType::ExternalName ||
@@ -2180,7 +2253,6 @@ void Profiler::SendString( uint64_t str, const char* ptr, QueueType type )
     MemWrite( &item.hdr.type, type );
     MemWrite( &item.stringTransfer.ptr, str );
 
-    auto len = strlen( ptr );
     assert( len <= std::numeric_limits<uint16_t>::max() );
     auto l16 = uint16_t( len );
 
@@ -2191,6 +2263,36 @@ void Profiler::SendString( uint64_t str, const char* ptr, QueueType type )
     AppendDataUnsafe( ptr, l16 );
 }
 
+void Profiler::SendSingleString( const char* ptr, size_t len )
+{
+    QueueItem item;
+    MemWrite( &item.hdr.type, QueueType::SingleStringData );
+
+    assert( len <= std::numeric_limits<uint16_t>::max() );
+    auto l16 = uint16_t( len );
+
+    NeedDataSize( QueueDataSize[(int)QueueType::SingleStringData] + sizeof( l16 ) + l16 );
+
+    AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::SingleStringData] );
+    AppendDataUnsafe( &l16, sizeof( l16 ) );
+    AppendDataUnsafe( ptr, l16 );
+}
+
+void Profiler::SendSecondString( const char* ptr, size_t len )
+{
+    QueueItem item;
+    MemWrite( &item.hdr.type, QueueType::SecondStringData );
+
+    assert( len <= std::numeric_limits<uint16_t>::max() );
+    auto l16 = uint16_t( len );
+
+    NeedDataSize( QueueDataSize[(int)QueueType::SecondStringData] + sizeof( l16 ) + l16 );
+
+    AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::SecondStringData] );
+    AppendDataUnsafe( &l16, sizeof( l16 ) );
+    AppendDataUnsafe( ptr, l16 );
+}
+
 void Profiler::SendLongString( uint64_t str, const char* ptr, size_t len, QueueType type )
 {
     assert( type == QueueType::FrameImageData ||
@@ -2234,16 +2336,17 @@ void Profiler::SendSourceLocationPayload( uint64_t _ptr )
     MemWrite( &item.hdr.type, QueueType::SourceLocationPayload );
     MemWrite( &item.stringTransfer.ptr, _ptr );
 
-    const auto len = *((uint32_t*)ptr);
+    uint16_t len;
-    assert( len <= std::numeric_limits<uint16_t>::max() );
+    memcpy( &len, ptr, sizeof( len ) );
-    assert( len > 4 );
+    assert( len > 2 );
-    const auto l16 = uint16_t( len - 4 );
+    len -= 2;
+    ptr += 2;
 
-    NeedDataSize( QueueDataSize[(int)QueueType::SourceLocationPayload] + sizeof( l16 ) + l16 );
+    NeedDataSize( QueueDataSize[(int)QueueType::SourceLocationPayload] + sizeof( len ) + len );
 
     AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::SourceLocationPayload] );
-    AppendDataUnsafe( &l16, sizeof( l16 ) );
+    AppendDataUnsafe( &len, sizeof( len ) );
-    AppendDataUnsafe( ptr + 4, l16 );
+    AppendDataUnsafe( ptr, len );
 }
 
 void Profiler::SendCallstackPayload( uint64_t _ptr )
@@ -2304,15 +2407,15 @@ void Profiler::SendCallstackAlloc( uint64_t _ptr )
     MemWrite( &item.hdr.type, QueueType::CallstackAllocPayload );
     MemWrite( &item.stringTransfer.ptr, _ptr );
 
-    const auto len = *((uint32_t*)ptr);
+    uint16_t len;
-    assert( len <= std::numeric_limits<uint16_t>::max() );
+    memcpy( &len, ptr, 2 );
-    const auto l16 = uint16_t( len );
+    ptr += 2;
 
-    NeedDataSize( QueueDataSize[(int)QueueType::CallstackAllocPayload] + sizeof( l16 ) + l16 );
+    NeedDataSize( QueueDataSize[(int)QueueType::CallstackAllocPayload] + sizeof( len ) + len );
 
     AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::CallstackAllocPayload] );
-    AppendDataUnsafe( &l16, sizeof( l16 ) );
+    AppendDataUnsafe( &len, sizeof( len ) );
-    AppendDataUnsafe( ptr + 4, l16 );
+    AppendDataUnsafe( ptr, len );
 }
 
 void Profiler::SendCallstackFrame( uint64_t ptr )
@@ -2321,13 +2424,12 @@ void Profiler::SendCallstackFrame( uint64_t ptr )
     const auto frameData = DecodeCallstackPtr( ptr );
 
     {
-        SendString( uint64_t( frameData.imageName ), frameData.imageName, QueueType::CustomStringData );
+        SendSingleString( frameData.imageName );
 
         QueueItem item;
         MemWrite( &item.hdr.type, QueueType::CallstackFrameSize );
         MemWrite( &item.callstackFrameSize.ptr, ptr );
         MemWrite( &item.callstackFrameSize.size, frameData.size );
-        MemWrite( &item.callstackFrameSize.imageName, (uint64_t)frameData.imageName );
 
         AppendData( &item, QueueDataSize[(int)QueueType::CallstackFrameSize] );
     }
@@ -2336,23 +2438,14 @@ void Profiler::SendCallstackFrame( uint64_t ptr )
     {
         const auto& frame = frameData.data[i];
 
-        SendString( uint64_t( frame.name ), frame.name, QueueType::CustomStringData );
+        SendSingleString( frame.name );
-        SendString( uint64_t( frame.file ), frame.file, QueueType::CustomStringData );
+        SendSecondString( frame.file );
 
         QueueItem item;
         MemWrite( &item.hdr.type, QueueType::CallstackFrame );
-        MemWrite( &item.callstackFrame.name, (uint64_t)frame.name );
-        MemWrite( &item.callstackFrame.file, (uint64_t)frame.file );
         MemWrite( &item.callstackFrame.line, frame.line );
         MemWrite( &item.callstackFrame.symAddr, frame.symAddr );
-        if( frame.symLen > ( 1 << 24 ) )
+        MemWrite( &item.callstackFrame.symLen, frame.symLen );
-        {
-            memset( item.callstackFrame.symLen, 0, 3 );
-        }
-        else
-        {
-            memcpy( item.callstackFrame.symLen, &frame.symLen, 3 );
-        }
 
         AppendData( &item, QueueDataSize[(int)QueueType::CallstackFrame] );
 
@@ -2383,7 +2476,7 @@ bool Profiler::HandleServerQuery()
     case ServerQueryThreadString:
         if( ptr == m_mainThread )
         {
-            SendString( ptr, "Main thread", QueueType::ThreadName );
+            SendString( ptr, "Main thread", 11, QueueType::ThreadName );
         }
         else
         {
@@ -2418,9 +2511,11 @@ bool Profiler::HandleServerQuery()
     case ServerQuerySymbol:
         HandleSymbolQuery( ptr );
         break;
+#ifndef TRACY_NO_CODE_TRANSFER
     case ServerQuerySymbolCode:
         HandleSymbolCodeQuery( ptr, extra );
         break;
+#endif
     case ServerQueryCodeLocation:
         SendCodeLocation( ptr );
         break;
@@ -2517,9 +2612,6 @@ void Profiler::HandleDisconnect()
 void Profiler::CalibrateTimer()
 {
 #ifdef TRACY_HW_TIMER
-#  if !defined TARGET_OS_IOS && __ARM_ARCH >= 6
-    m_timerMul = 1.;
-#  else
     std::atomic_signal_fence( std::memory_order_acq_rel );
     const auto t0 = std::chrono::high_resolution_clock::now();
     const auto r0 = GetTime();
@@ -2534,7 +2626,6 @@ void Profiler::CalibrateTimer()
     const auto dr = r1 - r0;
 
     m_timerMul = double( dt ) / double( dr );
-#  endif
 #else
     m_timerMul = 1.;
 #endif
@@ -2542,7 +2633,7 @@ void Profiler::CalibrateTimer()
 
 void Profiler::CalibrateDelay()
 {
-    enum { Iterations = 50000 };
+    constexpr int Iterations = 50000;
 
     auto mindiff = std::numeric_limits<int64_t>::max();
     for( int i=0; i<Iterations * 10; i++ )
@@ -2557,7 +2648,7 @@ void Profiler::CalibrateDelay()
 #ifdef TRACY_DELAYED_INIT
     m_delay = m_resolution;
 #else
-    enum { Events = Iterations * 2 };   // start + end
+    constexpr int Events = Iterations * 2;   // start + end
     static_assert( Events < QueuePrealloc, "Delay calibration loop will allocate memory in queue" );
 
     static const tracy::SourceLocationData __tracy_source_location { nullptr, __FUNCTION__,  __FILE__, (uint32_t)__LINE__, 0 };
@@ -2692,12 +2783,16 @@ void Profiler::ReportTopology()
         sprintf( path, "%s%i/topology/physical_package_id", basePath, i );
         char buf[1024];
         FILE* f = fopen( path, "rb" );
+        if( !f )
+        {
+            tracy_free( cpuData );
+            return;
+        }
         auto read = fread( buf, 1, 1024, f );
         buf[read] = '\0';
         fclose( f );
         cpuData[i].package = uint32_t( atoi( buf ) );
         cpuData[i].thread = i;
-
         sprintf( path, "%s%i/topology/core_id", basePath, i );
         f = fopen( path, "rb" );
         read = fread( buf, 1, 1024, f );
@@ -2730,11 +2825,10 @@ void Profiler::ReportTopology()
 void Profiler::SendCallstack( int depth, const char* skipBefore )
 {
 #ifdef TRACY_HAS_CALLSTACK
+    TracyLfqPrepare( QueueType::Callstack );
     auto ptr = Callstack( depth );
     CutCallstack( ptr, skipBefore );
-
+    MemWrite( &item->callstackFat.ptr, (uint64_t)ptr );
-    TracyLfqPrepare( QueueType::Callstack );
-    MemWrite( &item->callstack.ptr, (uint64_t)ptr );
     TracyLfqCommit;
 #endif
 }
@@ -2800,11 +2894,10 @@ void Profiler::HandleSymbolQuery( uint64_t symbol )
 #ifdef TRACY_HAS_CALLSTACK
     const auto sym = DecodeSymbolAddress( symbol );
 
-    SendString( uint64_t( sym.file ), sym.file, QueueType::CustomStringData );
+    SendSingleString( sym.file );
 
     QueueItem item;
     MemWrite( &item.hdr.type, QueueType::SymbolInformation );
-    MemWrite( &item.symbolInformation.file, uint64_t( sym.file ) );
     MemWrite( &item.symbolInformation.line, sym.line );
     MemWrite( &item.symbolInformation.symAddr, symbol );
 
@@ -2824,12 +2917,11 @@ void Profiler::SendCodeLocation( uint64_t ptr )
 #ifdef TRACY_HAS_CALLSTACK
     const auto sym = DecodeCodeAddress( ptr );
 
-    SendString( uint64_t( sym.file ), sym.file, QueueType::CustomStringData );
+    SendSingleString( sym.file );
 
     QueueItem item;
     MemWrite( &item.hdr.type, QueueType::CodeInformation );
     MemWrite( &item.codeInformation.ptr, ptr );
-    MemWrite( &item.codeInformation.file, uint64_t( sym.file ) );
     MemWrite( &item.codeInformation.line, sym.line );
 
     AppendData( &item, QueueDataSize[(int)QueueType::CodeInformation] );
@@ -2900,14 +2992,13 @@ TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_callstack( const struct ___trac
         TracyLfqCommitC;
     }
 #endif
+    tracy::GetProfiler().SendCallstack( depth );
     {
         TracyLfqPrepareC( tracy::QueueType::ZoneBeginCallstack );
         tracy::MemWrite( &item->zoneBegin.time, tracy::Profiler::GetTime() );
         tracy::MemWrite( &item->zoneBegin.srcloc, (uint64_t)srcloc );
         TracyLfqCommitC;
     }
-
-    tracy::GetProfiler().SendCallstack( depth );
     return ctx;
 }
 
@@ -2966,14 +3057,13 @@ TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_alloc_callstack( uint64_t srclo
         TracyLfqCommitC;
     }
 #endif
+    tracy::GetProfiler().SendCallstack( depth );
     {
         TracyLfqPrepareC( tracy::QueueType::ZoneBeginAllocSrcLocCallstack );
         tracy::MemWrite( &item->zoneBegin.time, tracy::Profiler::GetTime() );
         tracy::MemWrite( &item->zoneBegin.srcloc, srcloc );
         TracyLfqCommitC;
     }
-
-    tracy::GetProfiler().SendCallstack( depth );
     return ctx;
 }
 
@@ -2996,10 +3086,10 @@ TRACY_API void ___tracy_emit_zone_end( TracyCZoneCtx ctx )
 
 TRACY_API void ___tracy_emit_zone_text( TracyCZoneCtx ctx, const char* txt, size_t size )
 {
+    assert( size < std::numeric_limits<uint16_t>::max() );
     if( !ctx.active ) return;
-    auto ptr = (char*)tracy::tracy_malloc( size+1 );
+    auto ptr = (char*)tracy::tracy_malloc( size );
     memcpy( ptr, txt, size );
-    ptr[size] = '\0';
 #ifndef TRACY_NO_VERIFY
     {
         TracyLfqPrepareC( tracy::QueueType::ZoneValidation );
@@ -3009,17 +3099,18 @@ TRACY_API void ___tracy_emit_zone_text( TracyCZoneCtx ctx, const char* txt, size
 #endif
     {
         TracyLfqPrepareC( tracy::QueueType::ZoneText );
-        tracy::MemWrite( &item->zoneText.text, (uint64_t)ptr );
+        tracy::MemWrite( &item->zoneTextFat.text, (uint64_t)ptr );
+        tracy::MemWrite( &item->zoneTextFat.size, (uint16_t)size );
         TracyLfqCommitC;
     }
 }
 
 TRACY_API void ___tracy_emit_zone_name( TracyCZoneCtx ctx, const char* txt, size_t size )
 {
+    assert( size < std::numeric_limits<uint16_t>::max() );
     if( !ctx.active ) return;
-    auto ptr = (char*)tracy::tracy_malloc( size+1 );
+    auto ptr = (char*)tracy::tracy_malloc( size );
     memcpy( ptr, txt, size );
-    ptr[size] = '\0';
 #ifndef TRACY_NO_VERIFY
     {
         TracyLfqPrepareC( tracy::QueueType::ZoneValidation );
@@ -3029,7 +3120,8 @@ TRACY_API void ___tracy_emit_zone_name( TracyCZoneCtx ctx, const char* txt, size
 #endif
     {
         TracyLfqPrepareC( tracy::QueueType::ZoneName );
-        tracy::MemWrite( &item->zoneText.text, (uint64_t)ptr );
+        tracy::MemWrite( &item->zoneTextFat.text, (uint64_t)ptr );
+        tracy::MemWrite( &item->zoneTextFat.size, (uint16_t)size );
         TracyLfqCommitC;
     }
 }
@@ -3051,10 +3143,10 @@ TRACY_API void ___tracy_emit_zone_value( TracyCZoneCtx ctx, uint64_t value )
     }
 }
 
-TRACY_API void ___tracy_emit_memory_alloc( const void* ptr, size_t size ) { tracy::Profiler::MemAlloc( ptr, size ); }
+TRACY_API void ___tracy_emit_memory_alloc( const void* ptr, size_t size, int secure ) { tracy::Profiler::MemAlloc( ptr, size, secure != 0 ); }
-TRACY_API void ___tracy_emit_memory_alloc_callstack( const void* ptr, size_t size, int depth ) { tracy::Profiler::MemAllocCallstack( ptr, size, depth ); }
+TRACY_API void ___tracy_emit_memory_alloc_callstack( const void* ptr, size_t size, int depth, int secure ) { tracy::Profiler::MemAllocCallstack( ptr, size, depth, secure != 0 ); }
-TRACY_API void ___tracy_emit_memory_free( const void* ptr ) { tracy::Profiler::MemFree( ptr ); }
+TRACY_API void ___tracy_emit_memory_free( const void* ptr, int secure ) { tracy::Profiler::MemFree( ptr, secure != 0 ); }
-TRACY_API void ___tracy_emit_memory_free_callstack( const void* ptr, int depth ) { tracy::Profiler::MemFreeCallstack( ptr, depth ); }
+TRACY_API void ___tracy_emit_memory_free_callstack( const void* ptr, int depth, int secure ) { tracy::Profiler::MemFreeCallstack( ptr, depth, secure != 0 ); }
 TRACY_API void ___tracy_emit_frame_mark( const char* name ) { tracy::Profiler::SendFrameMark( name ); }
 TRACY_API void ___tracy_emit_frame_mark_start( const char* name ) { tracy::Profiler::SendFrameMark( name, tracy::QueueType::FrameMarkMsgStart ); }
 TRACY_API void ___tracy_emit_frame_mark_end( const char* name ) { tracy::Profiler::SendFrameMark( name, tracy::QueueType::FrameMarkMsgEnd ); }
@@ -3065,8 +3157,27 @@ TRACY_API void ___tracy_emit_messageL( const char* txt, int callstack ) { tracy:
 TRACY_API void ___tracy_emit_messageC( const char* txt, size_t size, uint32_t color, int callstack ) { tracy::Profiler::MessageColor( txt, size, color, callstack ); }
 TRACY_API void ___tracy_emit_messageLC( const char* txt, uint32_t color, int callstack ) { tracy::Profiler::MessageColor( txt, color, callstack ); }
 TRACY_API void ___tracy_emit_message_appinfo( const char* txt, size_t size ) { tracy::Profiler::MessageAppInfo( txt, size ); }
-TRACY_API uint64_t ___tracy_alloc_srcloc( uint32_t line, const char* source, const char* function ) { return tracy::Profiler::AllocSourceLocation( line, source, function ); }
+
-TRACY_API uint64_t ___tracy_alloc_srcloc_name( uint32_t line, const char* source, const char* function, const char* name, size_t nameSz ) { return tracy::Profiler::AllocSourceLocation( line, source, function, name, nameSz ); }
+TRACY_API uint64_t ___tracy_alloc_srcloc( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz ) {
+    return tracy::Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz );
+}
+
+TRACY_API uint64_t ___tracy_alloc_srcloc_name( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz ) {
+    return tracy::Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz );
+}
+
+// thread_locals are not initialized on thread creation. At least on GNU/Linux. Instead they are
+// initialized on their first ODR-use. This means that the allocator is not automagically
+// initialized every time a thread is created. As thus, expose to the C API users a simple API to
+// call every time they create a thread. Here we can then put all sorts of per-thread
+// initialization.
+TRACY_API void ___tracy_init_thread(void) {
+#ifdef TRACY_DELAYED_INIT
+    (void)tracy::GetProfilerThreadData();
+#else
+    (void)tracy::s_rpmalloc_thread_init;
+#endif
+}
 
 #ifdef __cplusplus
 }

deps/tracy/client/TracyProfiler.hpp

@@ -5,6 +5,7 @@
 #include <atomic>
 #include <stdint.h>
 #include <string.h>
+#include <time.h>
 
 #include "tracy_concurrentqueue.h"
 #include "TracyCallstack.hpp"
@@ -24,11 +25,11 @@
 #  include <mach/mach_time.h>
 #endif
 
-#if defined _WIN32 || defined __CYGWIN__ || ( ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) && !defined __ANDROID__ ) || __ARM_ARCH >= 6
+#if defined _WIN32 || defined __CYGWIN__ || ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) || ( defined TARGET_OS_IOS && TARGET_OS_IOS == 1 )
 #  define TRACY_HW_TIMER
 #endif
 
-#if !defined TRACY_HW_TIMER || ( defined __ARM_ARCH && __ARM_ARCH >= 6 && !defined CLOCK_MONOTONIC_RAW )
+#if !defined TRACY_HW_TIMER
   #include <chrono>
 #endif
 
@@ -41,6 +42,10 @@
 
 namespace tracy
 {
+#if defined(TRACY_DELAYED_INIT) && defined(TRACY_MANUAL_LIFETIME)
+void StartupProfiler();
+void ShutdownProfiler();
+#endif
 
 class GpuCtx;
 class Profiler;
@@ -59,6 +64,8 @@ TRACY_API std::atomic<uint8_t>& GetGpuCtxCounter();
 TRACY_API GpuCtxWrapper& GetGpuCtx();
 TRACY_API uint64_t GetThreadHandle();
 TRACY_API void InitRPMallocThread();
+TRACY_API bool ProfilerAvailable();
+TRACY_API int64_t GetFrequencyQpc();
 
 struct SourceLocationData
 {
@@ -106,7 +113,7 @@ class Profiler
     struct FrameImageQueueItem
     {
         void* image;
-        uint64_t frame;
+        uint32_t frame;
         uint16_t w;
         uint16_t h;
         uint8_t offset;
@@ -117,19 +124,13 @@ public:
     Profiler();
     ~Profiler();
 
+    void SpawnWorkerThreads();
+
     static tracy_force_inline int64_t GetTime()
     {
 #ifdef TRACY_HW_TIMER
 #  if defined TARGET_OS_IOS && TARGET_OS_IOS == 1
         return mach_absolute_time();
-#  elif defined __ARM_ARCH && __ARM_ARCH >= 6
-#    ifdef CLOCK_MONOTONIC_RAW
-        struct timespec ts;
-        clock_gettime( CLOCK_MONOTONIC_RAW, &ts );
-        return int64_t( ts.tv_sec ) * 1000000000ll + int64_t( ts.tv_nsec );
-#    else
-        return std::chrono::duration_cast<std::chrono::nanoseconds>( std::chrono::high_resolution_clock::now().time_since_epoch() ).count();
-#    endif
 #  elif defined _WIN32 || defined __CYGWIN__
 #    ifdef TRACY_TIMER_QPC
         return GetTimeQpc();
@@ -144,9 +145,17 @@ public:
         uint64_t rax, rdx;
         asm volatile ( "rdtsc" : "=a" (rax), "=d" (rdx) );
         return ( rdx << 32 ) + rax;
+#  else
+#    error "TRACY_HW_TIMER detection logic needs fixing"
 #  endif
 #else
+#  if defined __linux__ && defined CLOCK_MONOTONIC_RAW
+        struct timespec ts;
+        clock_gettime( CLOCK_MONOTONIC_RAW, &ts );
+        return int64_t( ts.tv_sec ) * 1000000000ll + int64_t( ts.tv_nsec );
+#  else
         return std::chrono::duration_cast<std::chrono::nanoseconds>( std::chrono::high_resolution_clock::now().time_since_epoch() ).count();
+#  endif
 #endif
     }
 
@@ -197,6 +206,7 @@ public:
     static tracy_force_inline void SendFrameImage( const void* image, uint16_t w, uint16_t h, uint8_t offset, bool flip )
     {
         auto& profiler = GetProfiler();
+        assert( profiler.m_frameCount.load( std::memory_order_relaxed ) < std::numeric_limits<uint32_t>::max() );
 #ifdef TRACY_ON_DEMAND
         if( !profiler.IsConnected() ) return;
 #endif
@@ -207,7 +217,7 @@ public:
         profiler.m_fiLock.lock();
         auto fi = profiler.m_fiQueue.prepare_next();
         fi->image = ptr;
-        fi->frame = profiler.m_frameCount.load( std::memory_order_relaxed ) - offset;
+        fi->frame = uint32_t( profiler.m_frameCount.load( std::memory_order_relaxed ) - offset );
         fi->w = w;
         fi->h = h;
         fi->flip = flip;
@@ -269,18 +279,19 @@ public:
 
     static tracy_force_inline void Message( const char* txt, size_t size, int callstack )
     {
+        assert( size < std::numeric_limits<uint16_t>::max() );
 #ifdef TRACY_ON_DEMAND
         if( !GetProfiler().IsConnected() ) return;
 #endif
-        TracyLfqPrepare( callstack == 0 ? QueueType::Message : QueueType::MessageCallstack );
-        auto ptr = (char*)tracy_malloc( size+1 );
-        memcpy( ptr, txt, size );
-        ptr[size] = '\0';
-        MemWrite( &item->message.time, GetTime() );
-        MemWrite( &item->message.text, (uint64_t)ptr );
-        TracyLfqCommit;
-
         if( callstack != 0 ) tracy::GetProfiler().SendCallstack( callstack );
+
+        TracyLfqPrepare( callstack == 0 ? QueueType::Message : QueueType::MessageCallstack );
+        auto ptr = (char*)tracy_malloc( size );
+        memcpy( ptr, txt, size );
+        MemWrite( &item->messageFat.time, GetTime() );
+        MemWrite( &item->messageFat.text, (uint64_t)ptr );
+        MemWrite( &item->messageFat.size, (uint16_t)size );
+        TracyLfqCommit;
     }
 
     static tracy_force_inline void Message( const char* txt, int callstack )
@@ -288,31 +299,32 @@ public:
 #ifdef TRACY_ON_DEMAND
         if( !GetProfiler().IsConnected() ) return;
 #endif
-        TracyLfqPrepare( callstack == 0 ? QueueType::MessageLiteral : QueueType::MessageLiteralCallstack );
-        MemWrite( &item->message.time, GetTime() );
-        MemWrite( &item->message.text, (uint64_t)txt );
-        TracyLfqCommit;
-
         if( callstack != 0 ) tracy::GetProfiler().SendCallstack( callstack );
+
+        TracyLfqPrepare( callstack == 0 ? QueueType::MessageLiteral : QueueType::MessageLiteralCallstack );
+        MemWrite( &item->messageLiteral.time, GetTime() );
+        MemWrite( &item->messageLiteral.text, (uint64_t)txt );
+        TracyLfqCommit;
     }
 
     static tracy_force_inline void MessageColor( const char* txt, size_t size, uint32_t color, int callstack )
     {
+        assert( size < std::numeric_limits<uint16_t>::max() );
 #ifdef TRACY_ON_DEMAND
         if( !GetProfiler().IsConnected() ) return;
 #endif
-        TracyLfqPrepare( callstack == 0 ? QueueType::MessageColor : QueueType::MessageColorCallstack );
-        auto ptr = (char*)tracy_malloc( size+1 );
-        memcpy( ptr, txt, size );
-        ptr[size] = '\0';
-        MemWrite( &item->messageColor.time, GetTime() );
-        MemWrite( &item->messageColor.text, (uint64_t)ptr );
-        MemWrite( &item->messageColor.r, uint8_t( ( color       ) & 0xFF ) );
-        MemWrite( &item->messageColor.g, uint8_t( ( color >> 8  ) & 0xFF ) );
-        MemWrite( &item->messageColor.b, uint8_t( ( color >> 16 ) & 0xFF ) );
-        TracyLfqCommit;
-
         if( callstack != 0 ) tracy::GetProfiler().SendCallstack( callstack );
+
+        TracyLfqPrepare( callstack == 0 ? QueueType::MessageColor : QueueType::MessageColorCallstack );
+        auto ptr = (char*)tracy_malloc( size );
+        memcpy( ptr, txt, size );
+        MemWrite( &item->messageColorFat.time, GetTime() );
+        MemWrite( &item->messageColorFat.text, (uint64_t)ptr );
+        MemWrite( &item->messageColorFat.r, uint8_t( ( color       ) & 0xFF ) );
+        MemWrite( &item->messageColorFat.g, uint8_t( ( color >> 8  ) & 0xFF ) );
+        MemWrite( &item->messageColorFat.b, uint8_t( ( color >> 16 ) & 0xFF ) );
+        MemWrite( &item->messageColorFat.size, (uint16_t)size );
+        TracyLfqCommit;
     }
 
     static tracy_force_inline void MessageColor( const char* txt, uint32_t color, int callstack )
@@ -320,26 +332,27 @@ public:
 #ifdef TRACY_ON_DEMAND
         if( !GetProfiler().IsConnected() ) return;
 #endif
-        TracyLfqPrepare( callstack == 0 ? QueueType::MessageLiteralColor : QueueType::MessageLiteralColorCallstack );
-        MemWrite( &item->messageColor.time, GetTime() );
-        MemWrite( &item->messageColor.text, (uint64_t)txt );
-        MemWrite( &item->messageColor.r, uint8_t( ( color       ) & 0xFF ) );
-        MemWrite( &item->messageColor.g, uint8_t( ( color >> 8  ) & 0xFF ) );
-        MemWrite( &item->messageColor.b, uint8_t( ( color >> 16 ) & 0xFF ) );
-        TracyLfqCommit;
-
         if( callstack != 0 ) tracy::GetProfiler().SendCallstack( callstack );
+
+        TracyLfqPrepare( callstack == 0 ? QueueType::MessageLiteralColor : QueueType::MessageLiteralColorCallstack );
+        MemWrite( &item->messageColorLiteral.time, GetTime() );
+        MemWrite( &item->messageColorLiteral.text, (uint64_t)txt );
+        MemWrite( &item->messageColorLiteral.r, uint8_t( ( color       ) & 0xFF ) );
+        MemWrite( &item->messageColorLiteral.g, uint8_t( ( color >> 8  ) & 0xFF ) );
+        MemWrite( &item->messageColorLiteral.b, uint8_t( ( color >> 16 ) & 0xFF ) );
+        TracyLfqCommit;
     }
 
     static tracy_force_inline void MessageAppInfo( const char* txt, size_t size )
     {
+        assert( size < std::numeric_limits<uint16_t>::max() );
         InitRPMallocThread();
-        auto ptr = (char*)tracy_malloc( size+1 );
+        auto ptr = (char*)tracy_malloc( size );
         memcpy( ptr, txt, size );
-        ptr[size] = '\0';
         TracyLfqPrepare( QueueType::MessageAppInfo );
-        MemWrite( &item->message.time, GetTime() );
+        MemWrite( &item->messageFat.time, GetTime() );
-        MemWrite( &item->message.text, (uint64_t)ptr );
+        MemWrite( &item->messageFat.text, (uint64_t)ptr );
+        MemWrite( &item->messageFat.size, (uint16_t)size );
 
 #ifdef TRACY_ON_DEMAND
         GetProfiler().DeferItem( *item );
@@ -348,8 +361,9 @@ public:
         TracyLfqCommit;
     }
 
-    static tracy_force_inline void MemAlloc( const void* ptr, size_t size )
+    static tracy_force_inline void MemAlloc( const void* ptr, size_t size, bool secure )
     {
+        if( secure && !ProfilerAvailable() ) return;
 #ifdef TRACY_ON_DEMAND
         if( !GetProfiler().IsConnected() ) return;
 #endif
@@ -360,8 +374,9 @@ public:
         GetProfiler().m_serialLock.unlock();
     }
 
-    static tracy_force_inline void MemFree( const void* ptr )
+    static tracy_force_inline void MemFree( const void* ptr, bool secure )
     {
+        if( secure && !ProfilerAvailable() ) return;
 #ifdef TRACY_ON_DEMAND
         if( !GetProfiler().IsConnected() ) return;
 #endif
@@ -372,8 +387,9 @@ public:
         GetProfiler().m_serialLock.unlock();
     }
 
-    static tracy_force_inline void MemAllocCallstack( const void* ptr, size_t size, int depth )
+    static tracy_force_inline void MemAllocCallstack( const void* ptr, size_t size, int depth, bool secure )
     {
+        if( secure && !ProfilerAvailable() ) return;
 #ifdef TRACY_HAS_CALLSTACK
         auto& profiler = GetProfiler();
 #  ifdef TRACY_ON_DEMAND
@@ -385,16 +401,17 @@ public:
         auto callstack = Callstack( depth );
 
         profiler.m_serialLock.lock();
-        SendMemAlloc( QueueType::MemAllocCallstack, thread, ptr, size );
         SendCallstackMemory( callstack );
+        SendMemAlloc( QueueType::MemAllocCallstack, thread, ptr, size );
         profiler.m_serialLock.unlock();
 #else
-        MemAlloc( ptr, size );
+        MemAlloc( ptr, size, secure );
 #endif
     }
 
-    static tracy_force_inline void MemFreeCallstack( const void* ptr, int depth )
+    static tracy_force_inline void MemFreeCallstack( const void* ptr, int depth, bool secure )
     {
+        if( secure && !ProfilerAvailable() ) return;
 #ifdef TRACY_HAS_CALLSTACK
         auto& profiler = GetProfiler();
 #  ifdef TRACY_ON_DEMAND
@@ -406,11 +423,85 @@ public:
         auto callstack = Callstack( depth );
 
         profiler.m_serialLock.lock();
-        SendMemFree( QueueType::MemFreeCallstack, thread, ptr );
         SendCallstackMemory( callstack );
+        SendMemFree( QueueType::MemFreeCallstack, thread, ptr );
         profiler.m_serialLock.unlock();
 #else
-        MemFree( ptr );
+        MemFree( ptr, secure );
+#endif
+    }
+
+    static tracy_force_inline void MemAllocNamed( const void* ptr, size_t size, bool secure, const char* name )
+    {
+        if( secure && !ProfilerAvailable() ) return;
+#ifdef TRACY_ON_DEMAND
+        if( !GetProfiler().IsConnected() ) return;
+#endif
+        const auto thread = GetThreadHandle();
+
+        GetProfiler().m_serialLock.lock();
+        SendMemName( name );
+        SendMemAlloc( QueueType::MemAllocNamed, thread, ptr, size );
+        GetProfiler().m_serialLock.unlock();
+    }
+
+    static tracy_force_inline void MemFreeNamed( const void* ptr, bool secure, const char* name )
+    {
+        if( secure && !ProfilerAvailable() ) return;
+#ifdef TRACY_ON_DEMAND
+        if( !GetProfiler().IsConnected() ) return;
+#endif
+        const auto thread = GetThreadHandle();
+
+        GetProfiler().m_serialLock.lock();
+        SendMemName( name );
+        SendMemFree( QueueType::MemFreeNamed, thread, ptr );
+        GetProfiler().m_serialLock.unlock();
+    }
+
+    static tracy_force_inline void MemAllocCallstackNamed( const void* ptr, size_t size, int depth, bool secure, const char* name )
+    {
+        if( secure && !ProfilerAvailable() ) return;
+#ifdef TRACY_HAS_CALLSTACK
+        auto& profiler = GetProfiler();
+#  ifdef TRACY_ON_DEMAND
+        if( !profiler.IsConnected() ) return;
+#  endif
+        const auto thread = GetThreadHandle();
+
+        InitRPMallocThread();
+        auto callstack = Callstack( depth );
+
+        profiler.m_serialLock.lock();
+        SendCallstackMemory( callstack );
+        SendMemName( name );
+        SendMemAlloc( QueueType::MemAllocCallstackNamed, thread, ptr, size );
+        profiler.m_serialLock.unlock();
+#else
+        MemAlloc( ptr, size, secure );
+#endif
+    }
+
+    static tracy_force_inline void MemFreeCallstackNamed( const void* ptr, int depth, bool secure, const char* name )
+    {
+        if( secure && !ProfilerAvailable() ) return;
+#ifdef TRACY_HAS_CALLSTACK
+        auto& profiler = GetProfiler();
+#  ifdef TRACY_ON_DEMAND
+        if( !profiler.IsConnected() ) return;
+#  endif
+        const auto thread = GetThreadHandle();
+
+        InitRPMallocThread();
+        auto callstack = Callstack( depth );
+
+        profiler.m_serialLock.lock();
+        SendCallstackMemory( callstack );
+        SendMemName( name );
+        SendMemFree( QueueType::MemFreeCallstackNamed, thread, ptr );
+        profiler.m_serialLock.unlock();
+#else
+        MemFree( ptr, secure );
 #endif
     }
 
@@ -419,7 +510,7 @@ public:
 #ifdef TRACY_HAS_CALLSTACK
         auto ptr = Callstack( depth );
         TracyLfqPrepare( QueueType::Callstack );
-        MemWrite( &item->callstack.ptr, (uint64_t)ptr );
+        MemWrite( &item->callstackFat.ptr, (uint64_t)ptr );
         TracyLfqCommit;
 #endif
     }
@@ -445,12 +536,12 @@ public:
 
     static bool ShouldExit();
 
-#ifdef TRACY_ON_DEMAND
     tracy_force_inline bool IsConnected() const
     {
         return m_isConnected.load( std::memory_order_acquire );
     }
 
+#ifdef TRACY_ON_DEMAND
     tracy_force_inline uint64_t ConnectionId() const
     {
         return m_connectionId.load( std::memory_order_acquire );
@@ -468,11 +559,16 @@ public:
     void RequestShutdown() { m_shutdown.store( true, std::memory_order_relaxed ); m_shutdownManual.store( true, std::memory_order_relaxed ); }
     bool HasShutdownFinished() const { return m_shutdownFinished.load( std::memory_order_relaxed ); }
 
-    void SendString( uint64_t ptr, const char* str, QueueType type );
+    void SendString( uint64_t str, const char* ptr, QueueType type ) { SendString( str, ptr, strlen( ptr ), type ); }
+    void SendString( uint64_t str, const char* ptr, size_t len, QueueType type );
+    void SendSingleString( const char* ptr ) { SendSingleString( ptr, strlen( ptr ) ); }
+    void SendSingleString( const char* ptr, size_t len );
+    void SendSecondString( const char* ptr ) { SendSecondString( ptr, strlen( ptr ) ); }
+    void SendSecondString( const char* ptr, size_t len );
 
 
     // Allocated source location data layout:
-    //  4b  payload size
+    //  2b  payload size
     //  4b  color
     //  4b  source line
     //  fsz function name
@@ -483,30 +579,36 @@ public:
 
     static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, const char* function )
     {
-        const auto fsz = strlen( function );
+        return AllocSourceLocation( line, source, function, nullptr, 0 );
-        const auto ssz = strlen( source );
-        const uint32_t sz = uint32_t( 4 + 4 + 4 + fsz + 1 + ssz + 1 );
-        auto ptr = (char*)tracy_malloc( sz );
-        memcpy( ptr, &sz, 4 );
-        memset( ptr + 4, 0, 4 );
-        memcpy( ptr + 8, &line, 4 );
-        memcpy( ptr + 12, function, fsz+1 );
-        memcpy( ptr + 12 + fsz + 1, source, ssz + 1 );
-        return uint64_t( ptr );
     }
 
     static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, const char* function, const char* name, size_t nameSz )
     {
-        const auto fsz = strlen( function );
+        return AllocSourceLocation( line, source, strlen(source), function, strlen(function), name, nameSz );
-        const auto ssz = strlen( source );
+    }
-        const uint32_t sz = uint32_t( 4 + 4 + 4 + fsz + 1 + ssz + 1 + nameSz );
+
+    static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz )
+    {
+        return AllocSourceLocation( line, source, sourceSz, function, functionSz, nullptr, 0 );
+    }
+
+    static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz )
+    {
+        const auto sz32 = uint32_t( 2 + 4 + 4 + functionSz + 1 + sourceSz + 1 + nameSz );
+        assert( sz32 <= std::numeric_limits<uint16_t>::max() );
+        const auto sz = uint16_t( sz32 );
         auto ptr = (char*)tracy_malloc( sz );
-        memcpy( ptr, &sz, 4 );
+        memcpy( ptr, &sz, 2 );
-        memset( ptr + 4, 0, 4 );
+        memset( ptr + 2, 0, 4 );
-        memcpy( ptr + 8, &line, 4 );
+        memcpy( ptr + 6, &line, 4 );
-        memcpy( ptr + 12, function, fsz+1 );
+        memcpy( ptr + 10, function, functionSz );
-        memcpy( ptr + 12 + fsz + 1, source, ssz + 1 );
+        ptr[10 + functionSz] = '\0';
-        memcpy( ptr + 12 + fsz + 1 + ssz + 1, name, nameSz );
+        memcpy( ptr + 10 + functionSz + 1, source, sourceSz );
+        ptr[10 + functionSz + 1 + sourceSz] = '\0';
+        if( nameSz != 0 )
+        {
+            memcpy( ptr + 10 + functionSz + 1 + sourceSz + 1, name, nameSz );
+        }
         return uint64_t( ptr );
     }
 
@@ -575,14 +677,14 @@ private:
 #ifdef TRACY_HAS_CALLSTACK
         auto item = GetProfiler().m_serialQueue.prepare_next();
         MemWrite( &item->hdr.type, QueueType::CallstackMemory );
-        MemWrite( &item->callstackMemory.ptr, (uint64_t)ptr );
+        MemWrite( &item->callstackFat.ptr, (uint64_t)ptr );
         GetProfiler().m_serialQueue.commit_next();
 #endif
     }
 
     static tracy_force_inline void SendMemAlloc( QueueType type, const uint64_t thread, const void* ptr, size_t size )
     {
-        assert( type == QueueType::MemAlloc || type == QueueType::MemAllocCallstack );
+        assert( type == QueueType::MemAlloc || type == QueueType::MemAllocCallstack || type == QueueType::MemAllocNamed || type == QueueType::MemAllocCallstackNamed );
 
         auto item = GetProfiler().m_serialQueue.prepare_next();
         MemWrite( &item->hdr.type, type );
@@ -605,7 +707,7 @@ private:
 
     static tracy_force_inline void SendMemFree( QueueType type, const uint64_t thread, const void* ptr )
     {
-        assert( type == QueueType::MemFree || type == QueueType::MemFreeCallstack );
+        assert( type == QueueType::MemFree || type == QueueType::MemFreeCallstack || type == QueueType::MemFreeNamed || type == QueueType::MemFreeCallstackNamed );
 
         auto item = GetProfiler().m_serialQueue.prepare_next();
         MemWrite( &item->hdr.type, type );
@@ -615,6 +717,15 @@ private:
         GetProfiler().m_serialQueue.commit_next();
     }
 
+    static tracy_force_inline void SendMemName( const char* name )
+    {
+        assert( name );
+        auto item = GetProfiler().m_serialQueue.prepare_next();
+        MemWrite( &item->hdr.type, QueueType::MemNamePayload );
+        MemWrite( &item->memName.name, (uint64_t)name );
+        GetProfiler().m_serialQueue.commit_next();
+    }
+
 #if ( defined _WIN32 || defined __CYGWIN__ ) && defined TRACY_TIMER_QPC
     static int64_t GetTimeQpc();
 #endif
@@ -655,8 +766,8 @@ private:
     TracyMutex m_fiLock;
 
     std::atomic<uint64_t> m_frameCount;
-#ifdef TRACY_ON_DEMAND
     std::atomic<bool> m_isConnected;
+#ifdef TRACY_ON_DEMAND
     std::atomic<uint64_t> m_connectionId;
 
     TracyMutex m_deferredLock;

deps/tracy/client/TracyRingBuffer.hpp

@@ -0,0 +1,116 @@
+namespace tracy
+{
+
+template<size_t Size>
+class RingBuffer
+{
+public:
+    RingBuffer( int fd )
+        : m_fd( fd )
+    {
+        const auto pageSize = uint32_t( getpagesize() );
+        assert( Size >= pageSize );
+        assert( __builtin_popcount( Size ) == 1 );
+        m_mapSize = Size + pageSize;
+        auto mapAddr = mmap( nullptr, m_mapSize, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0 );
+        if( !mapAddr )
+        {
+            m_fd = 0;
+            close( fd );
+            return;
+        }
+        m_metadata = (perf_event_mmap_page*)mapAddr;
+        assert( m_metadata->data_offset == pageSize );
+        m_buffer = ((char*)mapAddr) + pageSize;
+    }
+
+    ~RingBuffer()
+    {
+        if( m_metadata ) munmap( m_metadata, m_mapSize );
+        if( m_fd ) close( m_fd );
+    }
+
+    RingBuffer( const RingBuffer& ) = delete;
+    RingBuffer& operator=( const RingBuffer& ) = delete;
+
+    RingBuffer( RingBuffer&& other )
+    {
+        memcpy( (char*)&other, (char*)this, sizeof( RingBuffer ) );
+        m_metadata = nullptr;
+        m_fd = 0;
+    }
+
+    RingBuffer& operator=( RingBuffer&& other )
+    {
+        memcpy( (char*)&other, (char*)this, sizeof( RingBuffer ) );
+        m_metadata = nullptr;
+        m_fd = 0;
+        return *this;
+    }
+
+    bool IsValid() const { return m_metadata != nullptr; }
+
+    void Enable()
+    {
+        ioctl( m_fd, PERF_EVENT_IOC_ENABLE, 0 );
+    }
+
+    bool HasData() const
+    {
+        const auto head = LoadHead();
+        return head > m_metadata->data_tail;
+    }
+
+    void Read( void* dst, uint64_t offset, uint64_t cnt )
+    {
+        auto src = ( m_metadata->data_tail + offset ) % Size;
+        if( src + cnt <= Size )
+        {
+            memcpy( dst, m_buffer + src, cnt );
+        }
+        else
+        {
+            const auto s0 = Size - src;
+            memcpy( dst, m_buffer + src, s0 );
+            memcpy( (char*)dst + s0, m_buffer, cnt - s0 );
+        }
+    }
+
+    void Advance( uint64_t cnt )
+    {
+        StoreTail( m_metadata->data_tail + cnt );
+    }
+
+    bool CheckTscCaps() const
+    {
+        return m_metadata->cap_user_time_zero;
+    }
+
+    int64_t ConvertTimeToTsc( int64_t timestamp ) const
+    {
+        assert( m_metadata->cap_user_time_zero );
+        const auto time = timestamp - m_metadata->time_zero;
+        const auto quot = time / m_metadata->time_mult;
+        const auto rem = time % m_metadata->time_mult;
+        return ( quot << m_metadata->time_shift ) + ( rem << m_metadata->time_shift ) / m_metadata->time_mult;
+    }
+
+private:
+    uint64_t LoadHead() const
+    {
+        return std::atomic_load_explicit( (const volatile std::atomic<uint64_t>*)&m_metadata->data_head, std::memory_order_acquire );
+    }
+
+    void StoreTail( uint64_t tail )
+    {
+        std::atomic_store_explicit( (volatile std::atomic<uint64_t>*)&m_metadata->data_tail, tail, std::memory_order_release );
+    }
+
+    perf_event_mmap_page* m_metadata;
+    char* m_buffer;
+
+    size_t m_mapSize;
+    int m_fd;
+};
+
+}

deps/tracy/client/TracyScoped.hpp

@@ -1,6 +1,7 @@
 #ifndef __TRACYSCOPED_HPP__
 #define __TRACYSCOPED_HPP__
 
+#include <limits>
 #include <stdint.h>
 #include <string.h>
 
@@ -15,6 +16,11 @@ namespace tracy
 class ScopedZone
 {
 public:
+    ScopedZone( const ScopedZone& ) = delete;
+    ScopedZone( ScopedZone&& ) = delete;
+    ScopedZone& operator=( const ScopedZone& ) = delete;
+    ScopedZone& operator=( ScopedZone&& ) = delete;
+
     tracy_force_inline ScopedZone( const SourceLocationData* srcloc, bool is_active = true )
 #ifdef TRACY_ON_DEMAND
         : m_active( is_active && GetProfiler().IsConnected() )
@@ -43,12 +49,50 @@ public:
 #ifdef TRACY_ON_DEMAND
         m_connectionId = GetProfiler().ConnectionId();
 #endif
+        GetProfiler().SendCallstack( depth );
+
         TracyLfqPrepare( QueueType::ZoneBeginCallstack );
         MemWrite( &item->zoneBegin.time, Profiler::GetTime() );
         MemWrite( &item->zoneBegin.srcloc, (uint64_t)srcloc );
         TracyLfqCommit;
+    }
 
+    tracy_force_inline ScopedZone( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, bool is_active = true )
+#ifdef TRACY_ON_DEMAND
+        : m_active( is_active && GetProfiler().IsConnected() )
+#else
+        : m_active( is_active )
+#endif
+    {
+        if( !m_active ) return;
+#ifdef TRACY_ON_DEMAND
+        m_connectionId = GetProfiler().ConnectionId();
+#endif
+        TracyLfqPrepare( QueueType::ZoneBeginAllocSrcLoc );
+        const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz );
+        MemWrite( &item->zoneBegin.time, Profiler::GetTime() );
+        MemWrite( &item->zoneBegin.srcloc, srcloc );
+        TracyLfqCommit;
+    }
+
+    tracy_force_inline ScopedZone( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, int depth, bool is_active = true )
+#ifdef TRACY_ON_DEMAND
+        : m_active( is_active && GetProfiler().IsConnected() )
+#else
+        : m_active( is_active )
+#endif
+    {
+        if( !m_active ) return;
+#ifdef TRACY_ON_DEMAND
+        m_connectionId = GetProfiler().ConnectionId();
+#endif
         GetProfiler().SendCallstack( depth );
+
+        TracyLfqPrepare( QueueType::ZoneBeginAllocSrcLocCallstack );
+        const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz );
+        MemWrite( &item->zoneBegin.time, Profiler::GetTime() );
+        MemWrite( &item->zoneBegin.srcloc, srcloc );
+        TracyLfqCommit;
     }
 
     tracy_force_inline ~ScopedZone()
@@ -64,29 +108,31 @@ public:
 
     tracy_force_inline void Text( const char* txt, size_t size )
     {
+        assert( size < std::numeric_limits<uint16_t>::max() );
         if( !m_active ) return;
 #ifdef TRACY_ON_DEMAND
         if( GetProfiler().ConnectionId() != m_connectionId ) return;
 #endif
-        auto ptr = (char*)tracy_malloc( size+1 );
+        auto ptr = (char*)tracy_malloc( size );
         memcpy( ptr, txt, size );
-        ptr[size] = '\0';
         TracyLfqPrepare( QueueType::ZoneText );
-        MemWrite( &item->zoneText.text, (uint64_t)ptr );
+        MemWrite( &item->zoneTextFat.text, (uint64_t)ptr );
+        MemWrite( &item->zoneTextFat.size, (uint16_t)size );
         TracyLfqCommit;
     }
 
     tracy_force_inline void Name( const char* txt, size_t size )
     {
+        assert( size < std::numeric_limits<uint16_t>::max() );
         if( !m_active ) return;
 #ifdef TRACY_ON_DEMAND
         if( GetProfiler().ConnectionId() != m_connectionId ) return;
 #endif
-        auto ptr = (char*)tracy_malloc( size+1 );
+        auto ptr = (char*)tracy_malloc( size );
         memcpy( ptr, txt, size );
-        ptr[size] = '\0';
         TracyLfqPrepare( QueueType::ZoneName );
-        MemWrite( &item->zoneText.text, (uint64_t)ptr );
+        MemWrite( &item->zoneTextFat.text, (uint64_t)ptr );
+        MemWrite( &item->zoneTextFat.size, (uint16_t)size );
         TracyLfqCommit;
     }
 

deps/tracy/client/TracySysTrace.cpp

@@ -21,23 +21,28 @@
 #    include "../common/TracyAlloc.hpp"
 #    include "../common/TracySystem.hpp"
 #    include "TracyProfiler.hpp"
+#    include "TracyThread.hpp"
 
 namespace tracy
 {
 
-DEFINE_GUID ( /* ce1dbfb4-137e-4da6-87b0-3f59aa102cbc */
+struct __declspec(uuid("{ce1dbfb4-137e-4da6-87b0-3f59aa102cbc}")) PERFINFOGUID;
-    PerfInfoGuid,
+static const auto PerfInfoGuid = __uuidof(PERFINFOGUID);
-    0xce1dbfb4,
+
-    0x137e,
+struct __declspec(uuid("{802EC45A-1E99-4B83-9920-87C98277BA9D}")) DXGKRNLGUID;
-    0x4da6,
+static const auto DxgKrnlGuid = __uuidof(DXGKRNLGUID);
-    0x87, 0xb0, 0x3f, 0x59, 0xaa, 0x10, 0x2c, 0xbc
+
-);
 
 static TRACEHANDLE s_traceHandle;
 static TRACEHANDLE s_traceHandle2;
 static EVENT_TRACE_PROPERTIES* s_prop;
 static DWORD s_pid;
 
+static EVENT_TRACE_PROPERTIES* s_propVsync;
+static TRACEHANDLE s_traceHandleVsync;
+static TRACEHANDLE s_traceHandleVsync2;
+Thread* s_threadVsync = nullptr;
+
 struct CSwitch
 {
     uint32_t    newThreadId;
@@ -85,6 +90,19 @@ struct StackWalkEvent
     uint64_t stack[192];
 };
 
+struct VSyncInfo
+{
+    void*       dxgAdapter;
+    uint32_t    vidPnTargetId;
+    uint64_t    scannedPhysicalAddress;
+    uint32_t    vidPnSourceId;
+    uint32_t    frameNumber;
+    int64_t     frameQpcTime;
+    void*       hFlipDevice;
+    uint32_t    flipType;
+    uint64_t    flipFenceId;
+};
+
 #ifdef __CYGWIN__
 extern "C" typedef DWORD (WINAPI *t_GetProcessIdOfThread)( HANDLE );
 extern "C" typedef DWORD (WINAPI *t_GetProcessImageFileNameA)( HANDLE, LPSTR, DWORD );
@@ -168,9 +186,9 @@ void WINAPI EventRecordCallback( PEVENT_RECORD record )
                     memcpy( trace, &sz, sizeof( uint64_t ) );
                     memcpy( trace+1, sw->stack, sizeof( uint64_t ) * sz );
                     TracyLfqPrepare( QueueType::CallstackSample );
-                    MemWrite( &item->callstackSample.time, sw->eventTimeStamp );
+                    MemWrite( &item->callstackSampleFat.time, sw->eventTimeStamp );
-                    MemWrite( &item->callstackSample.thread, (uint64_t)sw->stackThread );
+                    MemWrite( &item->callstackSampleFat.thread, (uint64_t)sw->stackThread );
-                    MemWrite( &item->callstackSample.ptr, (uint64_t)trace );
+                    MemWrite( &item->callstackSampleFat.ptr, (uint64_t)trace );
                     TracyLfqCommit;
                 }
             }
@@ -181,6 +199,136 @@ void WINAPI EventRecordCallback( PEVENT_RECORD record )
     }
 }
 
+static constexpr const char* VsyncName[] = {
+    "[0] Vsync",
+    "[1] Vsync",
+    "[2] Vsync",
+    "[3] Vsync",
+    "[4] Vsync",
+    "[5] Vsync",
+    "[6] Vsync",
+    "[7] Vsync",
+    "Vsync"
+};
+
+static uint32_t VsyncTarget[8] = {};
+
+void WINAPI EventRecordCallbackVsync( PEVENT_RECORD record )
+{
+#ifdef TRACY_ON_DEMAND
+    if( !GetProfiler().IsConnected() ) return;
+#endif
+
+    const auto& hdr = record->EventHeader;
+    assert( hdr.ProviderId.Data1 == 0x802EC45A );
+    assert( hdr.EventDescriptor.Id == 0x0011 );
+
+    const auto vs = (const VSyncInfo*)record->UserData;
+
+    int idx = 0;
+    do
+    {
+        if( VsyncTarget[idx] == 0 )
+        {
+            VsyncTarget[idx] = vs->vidPnTargetId;
+            break;
+        }
+        else if( VsyncTarget[idx] == vs->vidPnTargetId )
+        {
+            break;
+        }
+    }
+    while( ++idx < 8 );
+
+    TracyLfqPrepare( QueueType::FrameMarkMsg );
+    MemWrite( &item->frameMark.time, hdr.TimeStamp.QuadPart );
+    MemWrite( &item->frameMark.name, uint64_t( VsyncName[idx] ) );
+    TracyLfqCommit;
+}
+
+static void SetupVsync()
+{
+#if _WIN32_WINNT >= _WIN32_WINNT_WINBLUE
+    const auto psz = sizeof( EVENT_TRACE_PROPERTIES ) + MAX_PATH;
+    s_propVsync = (EVENT_TRACE_PROPERTIES*)tracy_malloc( psz );
+    memset( s_propVsync, 0, sizeof( EVENT_TRACE_PROPERTIES ) );
+    s_propVsync->LogFileMode = EVENT_TRACE_REAL_TIME_MODE;
+    s_propVsync->Wnode.BufferSize = psz;
+#ifdef TRACY_TIMER_QPC
+    s_propVsync->Wnode.ClientContext = 1;
+#else
+    s_propVsync->Wnode.ClientContext = 3;
+#endif
+    s_propVsync->LoggerNameOffset = sizeof( EVENT_TRACE_PROPERTIES );
+    strcpy( ((char*)s_propVsync) + sizeof( EVENT_TRACE_PROPERTIES ), "TracyVsync" );
+
+    auto backup = tracy_malloc( psz );
+    memcpy( backup, s_propVsync, psz );
+
+    const auto controlStatus = ControlTraceA( 0, "TracyVsync", s_propVsync, EVENT_TRACE_CONTROL_STOP );
+    if( controlStatus != ERROR_SUCCESS && controlStatus != ERROR_WMI_INSTANCE_NOT_FOUND )
+    {
+        tracy_free( backup );
+        tracy_free( s_propVsync );
+        return;
+    }
+
+    memcpy( s_propVsync, backup, psz );
+    tracy_free( backup );
+
+    const auto startStatus = StartTraceA( &s_traceHandleVsync, "TracyVsync", s_propVsync );
+    if( startStatus != ERROR_SUCCESS )
+    {
+        tracy_free( s_propVsync );
+        return;
+    }
+
+    EVENT_FILTER_EVENT_ID fe = {};
+    fe.FilterIn = TRUE;
+    fe.Count = 1;
+    fe.Events[0] = 0x0011;  // VSyncDPC_Info
+
+    EVENT_FILTER_DESCRIPTOR desc = {};
+    desc.Ptr = (ULONGLONG)&fe;
+    desc.Size = sizeof( fe );
+    desc.Type = EVENT_FILTER_TYPE_EVENT_ID;
+
+    ENABLE_TRACE_PARAMETERS params = {};
+    params.Version = ENABLE_TRACE_PARAMETERS_VERSION_2;
+    params.EnableProperty = EVENT_ENABLE_PROPERTY_IGNORE_KEYWORD_0;
+    params.SourceId = s_propVsync->Wnode.Guid;
+    params.EnableFilterDesc = &desc;
+    params.FilterDescCount = 1;
+
+    uint64_t mask = 0x4000000000000001;   // Microsoft_Windows_DxgKrnl_Performance | Base
+    EnableTraceEx2( s_traceHandleVsync, &DxgKrnlGuid, EVENT_CONTROL_CODE_ENABLE_PROVIDER, TRACE_LEVEL_INFORMATION, mask, mask, 0, &params );
+
+    char loggerName[MAX_PATH];
+    strcpy( loggerName, "TracyVsync" );
+
+    EVENT_TRACE_LOGFILEA log = {};
+    log.LoggerName = loggerName;
+    log.ProcessTraceMode = PROCESS_TRACE_MODE_REAL_TIME | PROCESS_TRACE_MODE_EVENT_RECORD | PROCESS_TRACE_MODE_RAW_TIMESTAMP;
+    log.EventRecordCallback = EventRecordCallbackVsync;
+
+    s_traceHandleVsync2 = OpenTraceA( &log );
+    if( s_traceHandleVsync2 == (TRACEHANDLE)INVALID_HANDLE_VALUE )
+    {
+        CloseTrace( s_traceHandleVsync );
+        tracy_free( s_propVsync );
+        return;
+    }
+
+    s_threadVsync = (Thread*)tracy_malloc( sizeof( Thread ) );
+    new(s_threadVsync) Thread( [] (void*) {
+        ThreadExitHandler threadExitHandler;
+        SetThreadPriority( GetCurrentThread(), THREAD_PRIORITY_TIME_CRITICAL );
+        SetThreadName( "Tracy Vsync" );
+        ProcessTrace( &s_traceHandleVsync2, 1, nullptr, nullptr );
+    }, nullptr );
+#endif
+}
+
 bool SysTraceStart( int64_t& samplingPeriod )
 {
     if( !_GetThreadDescription ) _GetThreadDescription = (t_GetThreadDescription)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "GetThreadDescription" );
@@ -233,6 +381,8 @@ bool SysTraceStart( int64_t& samplingPeriod )
 #endif
     s_prop->Wnode.Guid = SystemTraceControlGuid;
     s_prop->BufferSize = 1024;
+    s_prop->MinimumBuffers = std::thread::hardware_concurrency() * 4;
+    s_prop->MaximumBuffers = std::thread::hardware_concurrency() * 6;
     s_prop->LoggerNameOffset = sizeof( EVENT_TRACE_PROPERTIES );
     memcpy( ((char*)s_prop) + sizeof( EVENT_TRACE_PROPERTIES ), KERNEL_LOGGER_NAME, sizeof( KERNEL_LOGGER_NAME ) );
 
@@ -242,6 +392,7 @@ bool SysTraceStart( int64_t& samplingPeriod )
     const auto controlStatus = ControlTrace( 0, KERNEL_LOGGER_NAME, s_prop, EVENT_TRACE_CONTROL_STOP );
     if( controlStatus != ERROR_SUCCESS && controlStatus != ERROR_WMI_INSTANCE_NOT_FOUND )
     {
+        tracy_free( backup );
         tracy_free( s_prop );
         return false;
     }
@@ -288,17 +439,29 @@ bool SysTraceStart( int64_t& samplingPeriod )
         return false;
     }
 
+    SetupVsync();
+
     return true;
 }
 
 void SysTraceStop()
 {
+    if( s_threadVsync )
+    {
+        CloseTrace( s_traceHandleVsync2 );
+        CloseTrace( s_traceHandleVsync );
+        s_threadVsync->~Thread();
+        tracy_free( s_threadVsync );
+    }
+
     CloseTrace( s_traceHandle2 );
     CloseTrace( s_traceHandle );
 }
 
 void SysTraceWorker( void* ptr )
 {
+    ThreadExitHandler threadExitHandler;
+    SetThreadPriority( GetCurrentThread(), THREAD_PRIORITY_TIME_CRITICAL );
     SetThreadName( "Tracy SysTrace" );
     ProcessTrace( &s_traceHandle2, 1, 0, 0 );
     ControlTrace( 0, KERNEL_LOGGER_NAME, s_prop, EVENT_TRACE_CONTROL_STOP );
@@ -323,7 +486,7 @@ void SysTraceSendExternalName( uint64_t thread )
             auto ret = wcstombs( buf, tmp, 256 );
             if( ret != 0 )
             {
-                GetProfiler().SendString( thread, buf, QueueType::ExternalThreadName );
+                GetProfiler().SendString( thread, buf, ret, QueueType::ExternalThreadName );
                 threadSent = true;
             }
         }
@@ -351,9 +514,10 @@ void SysTraceSendExternalName( uint64_t thread )
                                 if( (uint64_t)ptr >= (uint64_t)info.lpBaseOfDll && (uint64_t)ptr <= (uint64_t)info.lpBaseOfDll + (uint64_t)info.SizeOfImage )
                                 {
                                     char buf2[1024];
-                                    if( _GetModuleBaseNameA( phnd, modules[i], buf2, 1024 ) != 0 )
+                                    const auto modlen = _GetModuleBaseNameA( phnd, modules[i], buf2, 1024 );
+                                    if( modlen != 0 )
                                     {
-                                        GetProfiler().SendString( thread, buf2, QueueType::ExternalThreadName );
+                                        GetProfiler().SendString( thread, buf2, modlen, QueueType::ExternalThreadName );
                                         threadSent = true;
                                     }
                                 }
@@ -367,7 +531,7 @@ void SysTraceSendExternalName( uint64_t thread )
         CloseHandle( hnd );
         if( !threadSent )
         {
-            GetProfiler().SendString( thread, "???", QueueType::ExternalThreadName );
+            GetProfiler().SendString( thread, "???", 3, QueueType::ExternalThreadName );
             threadSent = true;
         }
         if( pid != 0 )
@@ -381,7 +545,7 @@ void SysTraceSendExternalName( uint64_t thread )
             }
             if( pid == 4 )
             {
-                GetProfiler().SendString( thread, "System", QueueType::ExternalName );
+                GetProfiler().SendString( thread, "System", 6, QueueType::ExternalName );
                 return;
             }
             else
@@ -407,9 +571,9 @@ void SysTraceSendExternalName( uint64_t thread )
 
     if( !threadSent )
     {
-        GetProfiler().SendString( thread, "???", QueueType::ExternalThreadName );
+        GetProfiler().SendString( thread, "???", 3, QueueType::ExternalThreadName );
     }
-    GetProfiler().SendString( thread, "???", QueueType::ExternalName );
+    GetProfiler().SendString( thread, "???", 3, QueueType::ExternalName );
 }
 
 }
@@ -428,8 +592,15 @@ void SysTraceSendExternalName( uint64_t thread )
 #    include <string.h>
 #    include <unistd.h>
 #    include <atomic>
+#    include <thread>
+#    include <linux/perf_event.h>
+#    include <linux/version.h>
+#    include <sys/mman.h>
+#    include <sys/ioctl.h>
 
 #    include "TracyProfiler.hpp"
+#    include "TracyRingBuffer.hpp"
+#    include "TracyThread.hpp"
 
 #    ifdef __ANDROID__
 #      include "TracySysTracePayload.hpp"
@@ -449,6 +620,173 @@ static const char BufferSizeKb[] = "buffer_size_kb";
 static const char TracePipe[] = "trace_pipe";
 
 static std::atomic<bool> traceActive { false };
+static Thread* s_threadSampling = nullptr;
+static int s_numCpus = 0;
+
+static constexpr size_t RingBufSize = 64*1024;
+static RingBuffer<RingBufSize>* s_ring = nullptr;
+
+static int perf_event_open( struct perf_event_attr* hw_event, pid_t pid, int cpu, int group_fd, unsigned long flags )
+{
+    return syscall( __NR_perf_event_open, hw_event, pid, cpu, group_fd, flags );
+}
+
+static void SetupSampling( int64_t& samplingPeriod )
+{
+#ifndef CLOCK_MONOTONIC_RAW
+    return;
+#endif
+
+    samplingPeriod = 100*1000;
+
+    s_numCpus = (int)std::thread::hardware_concurrency();
+    s_ring = (RingBuffer<RingBufSize>*)tracy_malloc( sizeof( RingBuffer<RingBufSize> ) * s_numCpus );
+
+    perf_event_attr pe = {};
+
+    pe.type = PERF_TYPE_SOFTWARE;
+    pe.size = sizeof( perf_event_attr );
+    pe.config = PERF_COUNT_SW_CPU_CLOCK;
+
+    pe.sample_freq = 10000;
+    pe.sample_type = PERF_SAMPLE_TID | PERF_SAMPLE_TIME | PERF_SAMPLE_CALLCHAIN;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION( 4, 8, 0 )
+    pe.sample_max_stack = 127;
+#endif
+    pe.exclude_callchain_kernel = 1;
+
+    pe.disabled = 1;
+    pe.freq = 1;
+#if !defined TRACY_HW_TIMER || !( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 )
+    pe.use_clockid = 1;
+    pe.clockid = CLOCK_MONOTONIC_RAW;
+#endif
+
+    for( int i=0; i<s_numCpus; i++ )
+    {
+        const int fd = perf_event_open( &pe, -1, i, -1, 0 );
+        if( fd == -1 )
+        {
+            for( int j=0; j<i; j++ ) s_ring[j].~RingBuffer<RingBufSize>();
+            tracy_free( s_ring );
+            return;
+        }
+        new( s_ring+i ) RingBuffer<RingBufSize>( fd );
+    }
+
+    s_threadSampling = (Thread*)tracy_malloc( sizeof( Thread ) );
+    new(s_threadSampling) Thread( [] (void*) {
+        ThreadExitHandler threadExitHandler;
+        SetThreadName( "Tracy Sampling" );
+        sched_param sp = { 5 };
+        pthread_setschedparam( pthread_self(), SCHED_FIFO, &sp );
+        uint32_t currentPid = (uint32_t)getpid();
+#if defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 )
+        for( int i=0; i<s_numCpus; i++ )
+        {
+            if( !s_ring[i].CheckTscCaps() )
+            {
+                for( int j=0; j<s_numCpus; j++ ) s_ring[j].~RingBuffer<RingBufSize>();
+                tracy_free( s_ring );
+                const char* err = "Tracy Profiler: sampling is disabled due to non-native scheduler clock. Are you running under a VM?";
+                Profiler::MessageAppInfo( err, strlen( err ) );
+                return;
+            }
+        }
+#endif
+        for( int i=0; i<s_numCpus; i++ ) s_ring[i].Enable();
+        for(;;)
+        {
+            bool hadData = false;
+            for( int i=0; i<s_numCpus; i++ )
+            {
+                if( !traceActive.load( std::memory_order_relaxed ) ) break;
+                if( !s_ring[i].HasData() ) continue;
+                hadData = true;
+
+                perf_event_header hdr;
+                s_ring[i].Read( &hdr, 0, sizeof( perf_event_header ) );
+                if( hdr.type == PERF_RECORD_SAMPLE )
+                {
+                    uint32_t pid, tid;
+                    uint64_t t0;
+                    uint64_t cnt;
+
+                    auto offset = sizeof( perf_event_header );
+                    s_ring[i].Read( &pid, offset, sizeof( uint32_t ) );
+                    if( pid == currentPid )
+                    {
+                        offset += sizeof( uint32_t );
+                        s_ring[i].Read( &tid, offset, sizeof( uint32_t ) );
+                        offset += sizeof( uint32_t );
+                        s_ring[i].Read( &t0, offset, sizeof( uint64_t ) );
+                        offset += sizeof( uint64_t );
+                        s_ring[i].Read( &cnt, offset, sizeof( uint64_t ) );
+                        offset += sizeof( uint64_t );
+
+                        auto trace = (uint64_t*)tracy_malloc( ( 1 + cnt ) * sizeof( uint64_t ) );
+                        s_ring[i].Read( trace+1, offset, sizeof( uint64_t ) * cnt );
+
+                        // remove non-canonical pointers
+                        do
+                        {
+                            const auto test = (int64_t)trace[cnt];
+                            const auto m1 = test >> 63;
+                            const auto m2 = test >> 47;
+                            if( m1 == m2 ) break;
+                        }
+                        while( --cnt > 0 );
+                        for( uint64_t j=1; j<cnt; j++ )
+                        {
+                            const auto test = (int64_t)trace[j];
+                            const auto m1 = test >> 63;
+                            const auto m2 = test >> 47;
+                            if( m1 != m2 ) trace[j] = 0;
+                        }
+
+                        // skip kernel frames
+                        uint64_t j;
+                        for( j=0; j<cnt; j++ )
+                        {
+                            if( (int64_t)trace[j+1] >= 0 ) break;
+                        }
+                        if( j == cnt )
+                        {
+                            tracy_free( trace );
+                        }
+                        else
+                        {
+                            if( j > 0 )
+                            {
+                                cnt -= j;
+                                memmove( trace+1, trace+1+j, sizeof( uint64_t ) * cnt );
+                            }
+                            memcpy( trace, &cnt, sizeof( uint64_t ) );
+
+#if defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 )
+                            t0 = s_ring[i].ConvertTimeToTsc( t0 );
+#endif
+
+                            TracyLfqPrepare( QueueType::CallstackSample );
+                            MemWrite( &item->callstackSampleFat.time, t0 );
+                            MemWrite( &item->callstackSampleFat.thread, (uint64_t)tid );
+                            MemWrite( &item->callstackSampleFat.ptr, (uint64_t)trace );
+                            TracyLfqCommit;
+                        }
+                    }
+                }
+                s_ring[i].Advance( hdr.size );
+            }
+            if( !hadData )
+            {
+                std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) );
+            }
+        }
+
+        for( int i=0; i<s_numCpus; i++ ) s_ring[i].~RingBuffer<RingBufSize>();
+        tracy_free( s_ring );
+    }, nullptr );
+}
 
 #ifdef __ANDROID__
 static bool TraceWrite( const char* path, size_t psz, const char* val, size_t vsz )
@@ -525,6 +863,10 @@ void SysTraceInjectPayload()
 
 bool SysTraceStart( int64_t& samplingPeriod )
 {
+#ifndef CLOCK_MONOTONIC_RAW
+    return false;
+#endif
+
     if( !TraceWrite( TracingOn, sizeof( TracingOn ), "0", 2 ) ) return false;
     if( !TraceWrite( CurrentTracer, sizeof( CurrentTracer ), "nop", 4 ) ) return false;
     TraceWrite( TraceOptions, sizeof( TraceOptions ), "norecord-cmd", 13 );
@@ -533,12 +875,12 @@ bool SysTraceStart( int64_t& samplingPeriod )
     TraceWrite( TraceOptions, sizeof( TraceOptions ), "noannotate", 11 );
 #if defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 )
     if( !TraceWrite( TraceClock, sizeof( TraceClock ), "x86-tsc", 8 ) ) return false;
-#elif __ARM_ARCH >= 6
+#else
     if( !TraceWrite( TraceClock, sizeof( TraceClock ), "mono_raw", 9 ) ) return false;
 #endif
     if( !TraceWrite( SchedSwitch, sizeof( SchedSwitch ), "1", 2 ) ) return false;
     if( !TraceWrite( SchedWakeup, sizeof( SchedWakeup ), "1", 2 ) ) return false;
-    if( !TraceWrite( BufferSizeKb, sizeof( BufferSizeKb ), "512", 4 ) ) return false;
+    if( !TraceWrite( BufferSizeKb, sizeof( BufferSizeKb ), "4096", 5 ) ) return false;
 
 #if defined __ANDROID__ && ( defined __aarch64__ || defined __ARM_ARCH )
     SysTraceInjectPayload();
@@ -547,6 +889,8 @@ bool SysTraceStart( int64_t& samplingPeriod )
     if( !TraceWrite( TracingOn, sizeof( TracingOn ), "1", 2 ) ) return false;
     traceActive.store( true, std::memory_order_relaxed );
 
+    SetupSampling( samplingPeriod );
+
     return true;
 }
 
@@ -554,23 +898,27 @@ void SysTraceStop()
 {
     TraceWrite( TracingOn, sizeof( TracingOn ), "0", 2 );
     traceActive.store( false, std::memory_order_relaxed );
+    if( s_threadSampling )
+    {
+        s_threadSampling->~Thread();
+        tracy_free( s_threadSampling );
+    }
 }
 
-static uint64_t ReadNumber( const char*& ptr )
+static uint64_t ReadNumber( const char*& data )
 {
-    uint64_t val = 0;
+    auto ptr = data;
+    assert( *ptr >= '0' && *ptr <= '9' );
+    uint64_t val = *ptr++ - '0';
     for(;;)
     {
-        if( *ptr >= '0' && *ptr <= '9' )
+        const uint8_t v = uint8_t( *ptr - '0' );
-        {
+        if( v > 9 ) break;
-            val = val * 10 + ( *ptr - '0' );
+        val = val * 10 + v;
-            ptr++;
+        ptr++;
-        }
-        else
-        {
-            return val;
-        }
     }
+    data = ptr;
+    return val;
 }
 
 static uint8_t ReadState( char state )
@@ -674,7 +1022,7 @@ static void HandleTraceLine( const char* line )
 
 #if defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 )
     const auto time = ReadNumber( line );
-#elif __ARM_ARCH >= 6
+#else
     const auto ts = ReadNumber( line );
     line++;      // '.'
     const auto tus = ReadNumber( line );
@@ -774,19 +1122,16 @@ static void ProcessTraceLines( int fd )
         line = buf;
         for(;;)
         {
-            auto next = line;
+            auto next = (char*)memchr( line, '\n', end - line );
-            while( next < end && *next != '\n' ) next++;
+            if( !next )
-            next++;
-            if( next >= end )
             {
                 const auto lsz = end - line;
                 memmove( buf, line, lsz );
                 line = buf + lsz;
                 break;
             }
-
             HandleTraceLine( line );
-            line = next;
+            line = ++next;
         }
         if( rd < 64*1024 )
         {
@@ -799,6 +1144,7 @@ static void ProcessTraceLines( int fd )
 
 void SysTraceWorker( void* ptr )
 {
+    ThreadExitHandler threadExitHandler;
     SetThreadName( "Tracy SysTrace" );
     int pipefd[2];
     if( pipe( pipefd ) == 0 )
@@ -812,6 +1158,8 @@ void SysTraceWorker( void* ptr )
             if( dup2( pipefd[1], STDOUT_FILENO ) >= 0 )
             {
                 close( pipefd[1] );
+                sched_param sp = { 4 };
+                pthread_setschedparam( pthread_self(), SCHED_FIFO, &sp );
 #if defined __ANDROID__ && ( defined __aarch64__ || defined __ARM_ARCH )
                 execlp( "su", "su", "-c", "/data/tracy_systrace", (char*)nullptr );
 #endif
@@ -823,6 +1171,8 @@ void SysTraceWorker( void* ptr )
         {
             // parent
             close( pipefd[1] );
+            sched_param sp = { 5 };
+            pthread_setschedparam( pthread_self(), SCHED_FIFO, &sp );
             ProcessTraceLines( pipefd[0] );
             close( pipefd[0] );
         }
@@ -856,14 +1206,10 @@ static void ProcessTraceLines( int fd )
         const auto end = buf + rd;
         for(;;)
         {
-            auto next = line;
+            auto next = (char*)memchr( line, '\n', end - line );
-            while( next < end && *next != '\n' ) next++;
+            if( !next ) break;
-            if( next == end ) break;
-            assert( *next == '\n' );
-            next++;
-
             HandleTraceLine( line );
-            line = next;
+            line = ++next;
         }
     }
 
@@ -872,6 +1218,7 @@ static void ProcessTraceLines( int fd )
 
 void SysTraceWorker( void* ptr )
 {
+    ThreadExitHandler threadExitHandler;
     SetThreadName( "Tracy SysTrace" );
     char tmp[256];
     memcpy( tmp, BasePath, sizeof( BasePath ) - 1 );
@@ -879,6 +1226,8 @@ void SysTraceWorker( void* ptr )
 
     int fd = open( tmp, O_RDONLY );
     if( fd < 0 ) return;
+    sched_param sp = { 5 };
+    pthread_setschedparam( pthread_self(), SCHED_FIFO, &sp );
     ProcessTraceLines( fd );
     close( fd );
 }
@@ -900,7 +1249,7 @@ void SysTraceSendExternalName( uint64_t thread )
     }
     else
     {
-        GetProfiler().SendString( thread, "???", QueueType::ExternalThreadName );
+        GetProfiler().SendString( thread, "???", 3, QueueType::ExternalThreadName );
     }
 
     sprintf( fn, "/proc/%" PRIu64 "/status", thread );
@@ -909,7 +1258,7 @@ void SysTraceSendExternalName( uint64_t thread )
     {
         int pid = -1;
         size_t lsz = 1024;
-        auto line = (char*)malloc( lsz );
+        auto line = (char*)tracy_malloc( lsz );
         for(;;)
         {
             auto rd = getline( &line, &lsz, f );
@@ -920,7 +1269,7 @@ void SysTraceSendExternalName( uint64_t thread )
                 break;
             }
         }
-        free( line );
+        tracy_free( line );
         fclose( f );
         if( pid >= 0 )
         {
@@ -944,7 +1293,7 @@ void SysTraceSendExternalName( uint64_t thread )
             }
         }
     }
-    GetProfiler().SendString( thread, "???", QueueType::ExternalName );
+    GetProfiler().SendString( thread, "???", 3, QueueType::ExternalName );
 }
 
 }

deps/tracy/client/TracyThread.hpp

@@ -7,9 +7,24 @@
 #  include <pthread.h>
 #endif
 
+#ifdef TRACY_MANUAL_LIFETIME
+#  include "tracy_rpmalloc.hpp"
+#endif
+
 namespace tracy
 {
 
+class ThreadExitHandler
+{
+public:
+    ~ThreadExitHandler()
+    {
+#ifdef TRACY_MANUAL_LIFETIME
+        rpmalloc_thread_finalize();
+#endif
+    }
+};
+
 #if defined _WIN32 || defined __CYGWIN__
 
 class Thread

deps/tracy/client/tracy_concurrentqueue.h

@@ -62,24 +62,6 @@
 namespace tracy
 {
 
-// Exceptions
-#ifndef MOODYCAMEL_EXCEPTIONS_ENABLED
-#if (defined(_MSC_VER) && defined(_CPPUNWIND)) || (defined(__GNUC__) && defined(__EXCEPTIONS)) || (!defined(_MSC_VER) && !defined(__GNUC__))
-#define MOODYCAMEL_EXCEPTIONS_ENABLED
-#endif
-#endif
-#ifdef MOODYCAMEL_EXCEPTIONS_ENABLED
-#define MOODYCAMEL_TRY try
-#define MOODYCAMEL_CATCH(...) catch(__VA_ARGS__)
-#define MOODYCAMEL_RETHROW throw
-#define MOODYCAMEL_THROW(expr) throw (expr)
-#else
-#define MOODYCAMEL_TRY if (true)
-#define MOODYCAMEL_CATCH(...) else if (false)
-#define MOODYCAMEL_RETHROW
-#define MOODYCAMEL_THROW(expr)
-#endif
-
 #ifndef MOODYCAMEL_NOEXCEPT
 #if !defined(MOODYCAMEL_EXCEPTIONS_ENABLED)
 #define MOODYCAMEL_NOEXCEPT

deps/tracy/client/tracy_rpmalloc.cpp

@@ -1372,7 +1372,7 @@ _memory_allocate_heap(void) {
 		heap = (heap_t*)_memory_map((1 + (sizeof(heap_t) >> _memory_page_size_shift)) * _memory_page_size, &align_offset);
 		if (!heap)
 			return heap;
-		memset(heap, 0, sizeof(heap_t));
+		memset((char*)heap, 0, sizeof(heap_t));
 		heap->align_offset = align_offset;
 
 		//Get a new heap ID

deps/tracy/common/TracyProtocol.hpp

@@ -9,8 +9,8 @@ namespace tracy
 
 constexpr unsigned Lz4CompressBound( unsigned isize ) { return isize + ( isize / 255 ) + 16; }
 
-enum : uint32_t { ProtocolVersion = 35 };
+enum : uint32_t { ProtocolVersion = 42 };
-enum : uint32_t { BroadcastVersion = 1 };
+enum : uint16_t { BroadcastVersion = 2 };
 
 using lz4sz_t = uint32_t;
 
@@ -87,6 +87,7 @@ struct WelcomeMessage
     uint8_t onDemand;
     uint8_t isApple;
     uint8_t cpuArch;
+    uint8_t codeTransfer;
     char cpuManufacturer[12];
     uint32_t cpuId;
     char programName[WelcomeMessageProgramNameSize];
@@ -107,10 +108,10 @@ enum { OnDemandPayloadMessageSize = sizeof( OnDemandPayloadMessage ) };
 
 struct BroadcastMessage
 {
-    uint32_t broadcastVersion;
+    uint16_t broadcastVersion;
+    uint16_t listenPort;
     uint32_t protocolVersion;
-    uint32_t listenPort;
+    int32_t activeTime;        // in seconds
-    uint32_t activeTime;        // in seconds
     char programName[WelcomeMessageProgramNameSize];
 };
 

deps/tracy/common/TracyQueue.hpp

@@ -16,19 +16,12 @@ enum class QueueType : uint8_t
     MessageColorCallstack,
     MessageAppInfo,
     ZoneBeginAllocSrcLoc,
-    ZoneBeginAllocSrcLocLean,
     ZoneBeginAllocSrcLocCallstack,
-    ZoneBeginAllocSrcLocCallstackLean,
     CallstackMemory,
-    CallstackMemoryLean,
     Callstack,
-    CallstackLean,
     CallstackAlloc,
-    CallstackAllocLean,
     CallstackSample,
-    CallstackSampleLean,
     FrameImage,
-    FrameImageLean,
     ZoneBegin,
     ZoneBeginCallstack,
     ZoneEnd,
@@ -40,9 +33,13 @@ enum class QueueType : uint8_t
     LockSharedRelease,
     LockName,
     MemAlloc,
+    MemAllocNamed,
     MemFree,
+    MemFreeNamed,
     MemAllocCallstack,
+    MemAllocCallstackNamed,
     MemFreeCallstack,
+    MemFreeCallstackNamed,
     GpuZoneBegin,
     GpuZoneBeginCallstack,
     GpuZoneEnd,
@@ -56,6 +53,7 @@ enum class QueueType : uint8_t
     Terminate,
     KeepAlive,
     ThreadContext,
+    GpuCalibration,
     Crash,
     CrashReport,
     ZoneValidation,
@@ -82,9 +80,11 @@ enum class QueueType : uint8_t
     ParamSetup,
     ParamPingback,
     CpuTopology,
+    SingleStringData,
+    SecondStringData,
+    MemNamePayload,
     StringData,
     ThreadName,
-    CustomStringData,
     PlotName,
     SourceLocationPayload,
     CallstackPayload,
@@ -140,15 +140,15 @@ struct QueueFrameMark
     uint64_t name;      // ptr
 };
 
-struct QueueFrameImageLean
+struct QueueFrameImage
 {
-    uint64_t frame;
+    uint32_t frame;
     uint16_t w;
     uint16_t h;
     uint8_t flip;
 };
 
-struct QueueFrameImage : public QueueFrameImageLean
+struct QueueFrameImageFat : public QueueFrameImage
 {
     uint64_t image;     // ptr
 };
@@ -164,9 +164,10 @@ struct QueueSourceLocation
     uint8_t b;
 };
 
-struct QueueZoneText
+struct QueueZoneTextFat
 {
     uint64_t text;      // ptr
+    uint16_t size;
 };
 
 enum class LockType : uint8_t
@@ -187,7 +188,6 @@ struct QueueLockTerminate
 {
     uint32_t id;
     int64_t time;
-    LockType type;
 };
 
 struct QueueLockWait
@@ -195,7 +195,6 @@ struct QueueLockWait
     uint64_t thread;
     uint32_t id;
     int64_t time;
-    LockType type;
 };
 
 struct QueueLockObtain
@@ -222,7 +221,12 @@ struct QueueLockMark
 struct QueueLockName
 {
     uint32_t id;
+};
+
+struct QueueLockNameFat : public QueueLockName
+{
     uint64_t name;      // ptr
+    uint16_t size;
 };
 
 enum class PlotDataType : uint8_t
@@ -248,7 +252,6 @@ struct QueuePlotData
 struct QueueMessage
 {
     int64_t time;
-    uint64_t text;      // ptr
 };
 
 struct QueueMessageColor : public QueueMessage
@@ -258,6 +261,28 @@ struct QueueMessageColor : public QueueMessage
     uint8_t b;
 };
 
+struct QueueMessageLiteral : public QueueMessage
+{
+    uint64_t text;      // ptr
+};
+
+struct QueueMessageColorLiteral : public QueueMessageColor
+{
+    uint64_t text;      // ptr
+};
+
+struct QueueMessageFat : public QueueMessage
+{
+    uint64_t text;      // ptr
+    uint16_t size;
+};
+
+struct QueueMessageColorFat : public QueueMessageColor
+{
+    uint64_t text;      // ptr
+    uint16_t size;
+};
+
 // Don't change order, only add new entries at the end, this is also used on trace dumps!
 enum class GpuContextType : uint8_t
 {
@@ -268,6 +293,11 @@ enum class GpuContextType : uint8_t
     Direct3D12
 };
 
+enum GpuContextFlags : uint8_t
+{
+    GpuContextCalibration   = 1 << 0
+};
+
 struct QueueGpuNewContext
 {
     int64_t cpuTime;
@@ -275,7 +305,7 @@ struct QueueGpuNewContext
     uint64_t thread;
     float period;
     uint8_t context;
-    uint8_t accuracyBits;
+    GpuContextFlags flags;
     GpuContextType type;
 };
 
@@ -303,6 +333,19 @@ struct QueueGpuTime
     uint8_t context;
 };
 
+struct QueueGpuCalibration
+{
+    int64_t gpuTime;
+    int64_t cpuTime;
+    int64_t cpuDelta;
+    uint8_t context;
+};
+
+struct QueueMemNamePayload
+{
+    uint64_t name;
+};
+
 struct QueueMemAlloc
 {
     int64_t time;
@@ -318,29 +361,24 @@ struct QueueMemFree
     uint64_t ptr;
 };
 
-struct QueueCallstackMemory
+struct QueueCallstackFat
 {
     uint64_t ptr;
 };
 
-struct QueueCallstack
+struct QueueCallstackAllocFat
-{
-    uint64_t ptr;
-};
-
-struct QueueCallstackAlloc
 {
     uint64_t ptr;
     uint64_t nativePtr;
 };
 
-struct QueueCallstackSampleLean
+struct QueueCallstackSample
 {
     int64_t time;
     uint64_t thread;
 };
 
-struct QueueCallstackSample : public QueueCallstackSampleLean
+struct QueueCallstackSampleFat : public QueueCallstackSample
 {
     uint64_t ptr;
 };
@@ -349,21 +387,17 @@ struct QueueCallstackFrameSize
 {
     uint64_t ptr;
     uint8_t size;
-    uint64_t imageName;
 };
 
 struct QueueCallstackFrame
 {
-    uint64_t name;
-    uint64_t file;
     uint32_t line;
     uint64_t symAddr;
-    char symLen[3];
+    uint32_t symLen;
 };
 
 struct QueueSymbolInformation
 {
-    uint64_t file;
     uint32_t line;
     uint64_t symAddr;
 };
@@ -371,7 +405,6 @@ struct QueueSymbolInformation
 struct QueueCodeInformation
 {
     uint64_t ptr;
-    uint64_t file;
     uint32_t line;
 };
 
@@ -460,9 +493,9 @@ struct QueueItem
         QueueStringTransfer stringTransfer;
         QueueFrameMark frameMark;
         QueueFrameImage frameImage;
-        QueueFrameImage frameImageLean;
+        QueueFrameImageFat frameImageFat;
         QueueSourceLocation srcloc;
-        QueueZoneText zoneText;
+        QueueZoneTextFat zoneTextFat;
         QueueLockAnnounce lockAnnounce;
         QueueLockTerminate lockTerminate;
         QueueLockWait lockWait;
@@ -470,20 +503,26 @@ struct QueueItem
         QueueLockRelease lockRelease;
         QueueLockMark lockMark;
         QueueLockName lockName;
+        QueueLockNameFat lockNameFat;
         QueuePlotData plotData;
         QueueMessage message;
         QueueMessageColor messageColor;
+        QueueMessageLiteral messageLiteral;
+        QueueMessageColorLiteral messageColorLiteral;
+        QueueMessageFat messageFat;
+        QueueMessageColorFat messageColorFat;
         QueueGpuNewContext gpuNewContext;
         QueueGpuZoneBegin gpuZoneBegin;
         QueueGpuZoneEnd gpuZoneEnd;
         QueueGpuTime gpuTime;
+        QueueGpuCalibration gpuCalibration;
         QueueMemAlloc memAlloc;
         QueueMemFree memFree;
-        QueueCallstackMemory callstackMemory;
+        QueueMemNamePayload memName;
-        QueueCallstack callstack;
+        QueueCallstackFat callstackFat;
-        QueueCallstackAlloc callstackAlloc;
+        QueueCallstackAllocFat callstackAllocFat;
         QueueCallstackSample callstackSample;
-        QueueCallstackSampleLean callstackSampleLean;
+        QueueCallstackSampleFat callstackSampleFat;
         QueueCallstackFrameSize callstackFrameSize;
         QueueCallstackFrame callstackFrame;
         QueueSymbolInformation symbolInformation;
@@ -504,27 +543,20 @@ struct QueueItem
 enum { QueueItemSize = sizeof( QueueItem ) };
 
 static constexpr size_t QueueDataSize[] = {
-    sizeof( QueueHeader ) + sizeof( QueueZoneText ),
+    sizeof( QueueHeader ),                                  // zone text
-    sizeof( QueueHeader ) + sizeof( QueueZoneText ),        // zone name
+    sizeof( QueueHeader ),                                  // zone name
     sizeof( QueueHeader ) + sizeof( QueueMessage ),
     sizeof( QueueHeader ) + sizeof( QueueMessageColor ),
     sizeof( QueueHeader ) + sizeof( QueueMessage ),         // callstack
     sizeof( QueueHeader ) + sizeof( QueueMessageColor ),    // callstack
     sizeof( QueueHeader ) + sizeof( QueueMessage ),         // app info
-    sizeof( QueueHeader ) + sizeof( QueueZoneBegin ),       // allocated source location, not for network transfer
+    sizeof( QueueHeader ) + sizeof( QueueZoneBeginLean ),   // allocated source location
-    sizeof( QueueHeader ) + sizeof( QueueZoneBeginLean ),   // lean allocated source location
+    sizeof( QueueHeader ) + sizeof( QueueZoneBeginLean ),   // allocated source location, callstack
-    sizeof( QueueHeader ) + sizeof( QueueZoneBegin ),       // allocated source location, callstack, not for network transfer
+    sizeof( QueueHeader ),                                  // callstack memory
-    sizeof( QueueHeader ) + sizeof( QueueZoneBeginLean ),   // lean allocated source location, callstack
+    sizeof( QueueHeader ),                                  // callstack
-    sizeof( QueueHeader ) + sizeof( QueueCallstackMemory ), // not for network transfer
+    sizeof( QueueHeader ),                                  // callstack alloc
-    sizeof( QueueHeader ),                                  // lean callstack memory
+    sizeof( QueueHeader ) + sizeof( QueueCallstackSample ),
-    sizeof( QueueHeader ) + sizeof( QueueCallstack ),       // not for network transfer
+    sizeof( QueueHeader ) + sizeof( QueueFrameImage ),
-    sizeof( QueueHeader ),                                  // lean callstack
-    sizeof( QueueHeader ) + sizeof( QueueCallstackAlloc ),  // not for network transfer
-    sizeof( QueueHeader ),                                  // lean callstack alloc
-    sizeof( QueueHeader ) + sizeof( QueueCallstackSample ), // not for network transfer
-    sizeof( QueueHeader ) + sizeof( QueueCallstackSampleLean ),
-    sizeof( QueueHeader ) + sizeof( QueueFrameImage ),      // not for network transfer
-    sizeof( QueueHeader ) + sizeof( QueueFrameImageLean ),
     sizeof( QueueHeader ) + sizeof( QueueZoneBegin ),
     sizeof( QueueHeader ) + sizeof( QueueZoneBegin ),       // callstack
     sizeof( QueueHeader ) + sizeof( QueueZoneEnd ),
@@ -536,9 +568,13 @@ static constexpr size_t QueueDataSize[] = {
     sizeof( QueueHeader ) + sizeof( QueueLockRelease ),     // shared
     sizeof( QueueHeader ) + sizeof( QueueLockName ),
     sizeof( QueueHeader ) + sizeof( QueueMemAlloc ),
+    sizeof( QueueHeader ) + sizeof( QueueMemAlloc ),        // named
     sizeof( QueueHeader ) + sizeof( QueueMemFree ),
+    sizeof( QueueHeader ) + sizeof( QueueMemFree ),         // named
     sizeof( QueueHeader ) + sizeof( QueueMemAlloc ),        // callstack
+    sizeof( QueueHeader ) + sizeof( QueueMemAlloc ),        // callstack, named
     sizeof( QueueHeader ) + sizeof( QueueMemFree ),         // callstack
+    sizeof( QueueHeader ) + sizeof( QueueMemFree ),         // callstack, named
     sizeof( QueueHeader ) + sizeof( QueueGpuZoneBegin ),
     sizeof( QueueHeader ) + sizeof( QueueGpuZoneBegin ),    // callstack
     sizeof( QueueHeader ) + sizeof( QueueGpuZoneEnd ),
@@ -553,6 +589,7 @@ static constexpr size_t QueueDataSize[] = {
     sizeof( QueueHeader ),                                  // terminate
     sizeof( QueueHeader ),                                  // keep alive
     sizeof( QueueHeader ) + sizeof( QueueThreadContext ),
+    sizeof( QueueHeader ) + sizeof( QueueGpuCalibration ),
     sizeof( QueueHeader ),                                  // crash
     sizeof( QueueHeader ) + sizeof( QueueCrashReport ),
     sizeof( QueueHeader ) + sizeof( QueueZoneValidation ),
@@ -564,10 +601,10 @@ static constexpr size_t QueueDataSize[] = {
     sizeof( QueueHeader ) + sizeof( QueueLockAnnounce ),
     sizeof( QueueHeader ) + sizeof( QueueLockTerminate ),
     sizeof( QueueHeader ) + sizeof( QueueLockMark ),
-    sizeof( QueueHeader ) + sizeof( QueueMessage ),         // literal
+    sizeof( QueueHeader ) + sizeof( QueueMessageLiteral ),
-    sizeof( QueueHeader ) + sizeof( QueueMessageColor ),    // literal
+    sizeof( QueueHeader ) + sizeof( QueueMessageColorLiteral ),
-    sizeof( QueueHeader ) + sizeof( QueueMessage ),         // literal, callstack
+    sizeof( QueueHeader ) + sizeof( QueueMessageLiteral ),  // callstack
-    sizeof( QueueHeader ) + sizeof( QueueMessageColor ),    // literal, callstack
+    sizeof( QueueHeader ) + sizeof( QueueMessageColorLiteral ), // callstack
     sizeof( QueueHeader ) + sizeof( QueueGpuNewContext ),
     sizeof( QueueHeader ) + sizeof( QueueCallstackFrameSize ),
     sizeof( QueueHeader ) + sizeof( QueueCallstackFrame ),
@@ -579,10 +616,12 @@ static constexpr size_t QueueDataSize[] = {
     sizeof( QueueHeader ) + sizeof( QueueParamSetup ),
     sizeof( QueueHeader ),                                  // param pingback
     sizeof( QueueHeader ) + sizeof( QueueCpuTopology ),
+    sizeof( QueueHeader ),                                  // single string data
+    sizeof( QueueHeader ),                                  // second string data
+    sizeof( QueueHeader ) + sizeof( QueueMemNamePayload ),
     // keep all QueueStringTransfer below
     sizeof( QueueHeader ) + sizeof( QueueStringTransfer ),  // string data
     sizeof( QueueHeader ) + sizeof( QueueStringTransfer ),  // thread name
-    sizeof( QueueHeader ) + sizeof( QueueStringTransfer ),  // custom string data
     sizeof( QueueHeader ) + sizeof( QueueStringTransfer ),  // plot name
     sizeof( QueueHeader ) + sizeof( QueueStringTransfer ),  // allocated source location payload
     sizeof( QueueHeader ) + sizeof( QueueStringTransfer ),  // callstack payload

deps/tracy/common/TracySocket.cpp

@@ -1,4 +1,5 @@
 #include <assert.h>
+#include <inttypes.h>
 #include <new>
 #include <stdio.h>
 #include <stdlib.h>
@@ -106,36 +107,39 @@ Socket::~Socket()
     }
 }
 
-bool Socket::Connect( const char* addr, int port )
+bool Socket::Connect( const char* addr, uint16_t port )
 {
     assert( !IsValid() );
 
     if( m_ptr )
     {
         const auto c = connect( m_connSock, m_ptr->ai_addr, m_ptr->ai_addrlen );
-        assert( c == -1 );
+        if( c == -1 )
-#if defined _WIN32 || defined __CYGWIN__
-        const auto err = WSAGetLastError();
-        if( err == WSAEALREADY || err == WSAEINPROGRESS ) return false;
-        if( err != WSAEISCONN )
         {
-            freeaddrinfo( m_res );
+#if defined _WIN32
-            closesocket( m_connSock );
+            const auto err = WSAGetLastError();
-            m_ptr = nullptr;
+            if( err == WSAEALREADY || err == WSAEINPROGRESS ) return false;
-            return false;
+            if( err != WSAEISCONN )
-        }
+            {
+                freeaddrinfo( m_res );
+                closesocket( m_connSock );
+                m_ptr = nullptr;
+                return false;
+            }
 #else
-        if( errno == EALREADY || errno == EINPROGRESS ) return false;
+            const auto err = errno;
-        if( errno != EISCONN )
+            if( err == EALREADY || err == EINPROGRESS ) return false;
-        {
+            if( err != EISCONN )
-            freeaddrinfo( m_res );
+            {
-            close( m_connSock );
+                freeaddrinfo( m_res );
-            m_ptr = nullptr;
+                close( m_connSock );
-            return false;
+                m_ptr = nullptr;
-        }
+                return false;
+            }
 #endif
+        }
 
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
         u_long nonblocking = 0;
         ioctlsocket( m_connSock, FIONBIO, &nonblocking );
 #else
@@ -156,7 +160,7 @@ bool Socket::Connect( const char* addr, int port )
     hints.ai_socktype = SOCK_STREAM;
 
     char portbuf[32];
-    sprintf( portbuf, "%i", port );
+    sprintf( portbuf, "%" PRIu16, port );
 
     if( getaddrinfo( addr, portbuf, &hints, &res ) != 0 ) return false;
     int sock = 0;
@@ -167,7 +171,7 @@ bool Socket::Connect( const char* addr, int port )
         int val = 1;
         setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) );
 #endif
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
         u_long nonblocking = 1;
         ioctlsocket( sock, FIONBIO, &nonblocking );
 #else
@@ -180,7 +184,7 @@ bool Socket::Connect( const char* addr, int port )
         }
         else
         {
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
             const auto err = WSAGetLastError();
             if( err != WSAEWOULDBLOCK )
             {
@@ -203,7 +207,7 @@ bool Socket::Connect( const char* addr, int port )
     freeaddrinfo( res );
     if( !ptr ) return false;
 
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
     u_long nonblocking = 0;
     ioctlsocket( sock, FIONBIO, &nonblocking );
 #else
@@ -215,6 +219,48 @@ bool Socket::Connect( const char* addr, int port )
     return true;
 }
 
+bool Socket::ConnectBlocking( const char* addr, uint16_t port )
+{
+    assert( !IsValid() );
+    assert( !m_ptr );
+
+    struct addrinfo hints;
+    struct addrinfo *res, *ptr;
+
+    memset( &hints, 0, sizeof( hints ) );
+    hints.ai_family = AF_UNSPEC;
+    hints.ai_socktype = SOCK_STREAM;
+
+    char portbuf[32];
+    sprintf( portbuf, "%" PRIu16, port );
+
+    if( getaddrinfo( addr, portbuf, &hints, &res ) != 0 ) return false;
+    int sock = 0;
+    for( ptr = res; ptr; ptr = ptr->ai_next )
+    {
+        if( ( sock = socket( ptr->ai_family, ptr->ai_socktype, ptr->ai_protocol ) ) == -1 ) continue;
+#if defined __APPLE__
+        int val = 1;
+        setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) );
+#endif
+        if( connect( sock, ptr->ai_addr, ptr->ai_addrlen ) == -1 )
+        {
+#ifdef _WIN32
+            closesocket( sock );
+#else
+            close( sock );
+#endif
+            continue;
+        }
+        break;
+    }
+    freeaddrinfo( res );
+    if( !ptr ) return false;
+
+    m_sock.store( sock, std::memory_order_relaxed );
+    return true;
+}
+
 void Socket::Close()
 {
     const auto sock = m_sock.load( std::memory_order_relaxed );
@@ -247,7 +293,7 @@ int Socket::GetSendBufSize()
 {
     const auto sock = m_sock.load( std::memory_order_relaxed );
     int bufSize;
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
     int sz = sizeof( bufSize );
     getsockopt( sock, SOL_SOCKET, SO_SNDBUF, (char*)&bufSize, &sz );
 #else
@@ -306,6 +352,24 @@ int Socket::Recv( void* _buf, int len, int timeout )
     }
 }
 
+int Socket::ReadUpTo( void* _buf, int len, int timeout )
+{
+    const auto sock = m_sock.load( std::memory_order_relaxed );
+    auto buf = (char*)_buf;
+
+    int rd = 0;
+    while( len > 0 )
+    {
+        const auto res = recv( sock, buf, len, 0 );
+        if( res == 0 ) break;
+        if( res == -1 ) return -1;
+        len -= res;
+        rd += res;
+        buf += res;
+    }
+    return rd;
+}
+
 bool Socket::Read( void* buf, int len, int timeout )
 {
     auto cbuf = (char*)buf;
@@ -383,33 +447,45 @@ ListenSocket::~ListenSocket()
     if( m_sock != -1 ) Close();
 }
 
-bool ListenSocket::Listen( int port, int backlog )
+static int addrinfo_and_socket_for_family( uint16_t port, int ai_family, struct addrinfo** res )
+{
+    struct addrinfo hints;
+    memset( &hints, 0, sizeof( hints ) );
+    hints.ai_family = ai_family;
+    hints.ai_socktype = SOCK_STREAM;
+#ifndef TRACY_ONLY_LOCALHOST
+    const char* onlyLocalhost = getenv( "TRACY_ONLY_LOCALHOST" );
+    if( !onlyLocalhost || onlyLocalhost[0] != '1' )
+    {
+        hints.ai_flags = AI_PASSIVE;
+    }
+#endif
+    char portbuf[32];
+    sprintf( portbuf, "%" PRIu16, port );
+    if( getaddrinfo( nullptr, portbuf, &hints, res ) != 0 ) return -1;
+    int sock = socket( (*res)->ai_family, (*res)->ai_socktype, (*res)->ai_protocol );
+    if (sock == -1) freeaddrinfo( *res );
+    return sock;
+}
+
+bool ListenSocket::Listen( uint16_t port, int backlog )
 {
     assert( m_sock == -1 );
 
-    struct addrinfo* res;
+    struct addrinfo* res = nullptr;
-    struct addrinfo hints;
 
-    memset( &hints, 0, sizeof( hints ) );
+#ifndef TRACY_ONLY_IPV4
-    hints.ai_family = AF_INET6;
+    const char* onlyIPv4 = getenv( "TRACY_ONLY_IPV4" );
-    hints.ai_socktype = SOCK_STREAM;
+    if( !onlyIPv4 || onlyIPv4[0] != '1' )
-#ifndef TRACY_ONLY_LOCALHOST
+    {
-    hints.ai_flags = AI_PASSIVE;
+        m_sock = addrinfo_and_socket_for_family( port, AF_INET6, &res );
+    }
 #endif
-
-    char portbuf[32];
-    sprintf( portbuf, "%i", port );
-
-    if( getaddrinfo( nullptr, portbuf, &hints, &res ) != 0 ) return false;
-
-    m_sock = socket( res->ai_family, res->ai_socktype, res->ai_protocol );
     if (m_sock == -1)
     {
         // IPV6 protocol may not be available/is disabled. Try to create a socket
         // with the IPV4 protocol
-        hints.ai_family = AF_INET;
+        m_sock = addrinfo_and_socket_for_family( port, AF_INET, &res );
-        if( getaddrinfo( nullptr, portbuf, &hints, &res ) != 0 ) return false;
-        m_sock = socket( res->ai_family, res->ai_socktype, res->ai_protocol );
         if( m_sock == -1 ) return false;
     }
 #if defined _WIN32 || defined __CYGWIN__
@@ -483,7 +559,7 @@ UdpBroadcast::~UdpBroadcast()
     if( m_sock != -1 ) Close();
 }
 
-bool UdpBroadcast::Open( const char* addr, int port )
+bool UdpBroadcast::Open( const char* addr, uint16_t port )
 {
     assert( m_sock == -1 );
 
@@ -495,7 +571,7 @@ bool UdpBroadcast::Open( const char* addr, int port )
     hints.ai_socktype = SOCK_DGRAM;
 
     char portbuf[32];
-    sprintf( portbuf, "%i", port );
+    sprintf( portbuf, "%" PRIu16, port );
 
     if( getaddrinfo( addr, portbuf, &hints, &res ) != 0 ) return false;
     int sock = 0;
@@ -506,7 +582,7 @@ bool UdpBroadcast::Open( const char* addr, int port )
         int val = 1;
         setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) );
 #endif
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
         unsigned long broadcast = 1;
         if( setsockopt( sock, SOL_SOCKET, SO_BROADCAST, (const char*)&broadcast, sizeof( broadcast ) ) == -1 )
 #else
@@ -541,7 +617,7 @@ void UdpBroadcast::Close()
     m_sock = -1;
 }
 
-int UdpBroadcast::Send( int port, const void* data, int len )
+int UdpBroadcast::Send( uint16_t port, const void* data, int len )
 {
     assert( m_sock != -1 );
     struct sockaddr_in addr;
@@ -563,8 +639,10 @@ IpAddress::~IpAddress()
 
 void IpAddress::Set( const struct sockaddr& addr )
 {
-#if __MINGW32__
+#if defined _WIN32 && ( !defined NTDDI_WIN10 || NTDDI_VERSION < NTDDI_WIN10 )
-    auto ai = (struct sockaddr_in*)&addr;
+    struct sockaddr_in tmp;
+    memcpy( &tmp, &addr, sizeof( tmp ) );
+    auto ai = &tmp;
 #else
     auto ai = (const struct sockaddr_in*)&addr;
 #endif
@@ -585,7 +663,7 @@ UdpListen::~UdpListen()
     if( m_sock != -1 ) Close();
 }
 
-bool UdpListen::Listen( int port )
+bool UdpListen::Listen( uint16_t port )
 {
     assert( m_sock == -1 );
 
@@ -596,14 +674,14 @@ bool UdpListen::Listen( int port )
     int val = 1;
     setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) );
 #endif
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
     unsigned long reuse = 1;
     setsockopt( m_sock, SOL_SOCKET, SO_REUSEADDR, (const char*)&reuse, sizeof( reuse ) );
 #else
     int reuse = 1;
     setsockopt( m_sock, SOL_SOCKET, SO_REUSEADDR, &reuse, sizeof( reuse ) );
 #endif
-#if defined _WIN32 || defined __CYGWIN__
+#if defined _WIN32
     unsigned long broadcast = 1;
     if( setsockopt( sock, SOL_SOCKET, SO_BROADCAST, (const char*)&broadcast, sizeof( broadcast ) ) == -1 )
 #else
@@ -649,14 +727,14 @@ void UdpListen::Close()
     m_sock = -1;
 }
 
-const char* UdpListen::Read( size_t& len, IpAddress& addr )
+const char* UdpListen::Read( size_t& len, IpAddress& addr, int timeout )
 {
     static char buf[2048];
 
     struct pollfd fd;
     fd.fd = (socket_t)m_sock;
     fd.events = POLLIN;
-    if( poll( &fd, 1, 10 ) <= 0 ) return nullptr;
+    if( poll( &fd, 1, timeout ) <= 0 ) return nullptr;
 
     sockaddr sa;
     socklen_t salen = sizeof( struct sockaddr );

deps/tracy/common/TracySocket.hpp

@@ -23,12 +23,14 @@ public:
     Socket( int sock );
     ~Socket();
 
-    bool Connect( const char* addr, int port );
+    bool Connect( const char* addr, uint16_t port );
+    bool ConnectBlocking( const char* addr, uint16_t port );
     void Close();
 
     int Send( const void* buf, int len );
     int GetSendBufSize();
 
+    int ReadUpTo( void* buf, int len, int timeout );
     bool Read( void* buf, int len, int timeout );
 
     template<typename ShouldExit>
@@ -74,7 +76,7 @@ public:
     ListenSocket();
     ~ListenSocket();
 
-    bool Listen( int port, int backlog );
+    bool Listen( uint16_t port, int backlog );
     Socket* Accept();
     void Close();
 
@@ -93,10 +95,10 @@ public:
     UdpBroadcast();
     ~UdpBroadcast();
 
-    bool Open( const char* addr, int port );
+    bool Open( const char* addr, uint16_t port );
     void Close();
 
-    int Send( int port, const void* data, int len );
+    int Send( uint16_t port, const void* data, int len );
 
     UdpBroadcast( const UdpBroadcast& ) = delete;
     UdpBroadcast( UdpBroadcast&& ) = delete;
@@ -134,10 +136,10 @@ public:
     UdpListen();
     ~UdpListen();
 
-    bool Listen( int port );
+    bool Listen( uint16_t port );
     void Close();
 
-    const char* Read( size_t& len, IpAddress& addr );
+    const char* Read( size_t& len, IpAddress& addr, int timeout );
 
     UdpListen( const UdpListen& ) = delete;
     UdpListen( UdpListen&& ) = delete;

deps/tracy/common/TracySystem.cpp

@@ -237,3 +237,13 @@ TRACY_API const char* GetThreadName( uint64_t id )
 }
 
 }
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+TRACY_API void ___tracy_set_thread_name( const char* name ) { tracy::SetThreadName( name ); }
+
+#ifdef __cplusplus
+}
+#endif

resource/content/shaders/Voxel.fs

@@ -218,5 +218,4 @@ void main() {
     float ratio = exp(vs.Depth * 0.69)-1;
     color = mix(color, vec4(pow(FogColor, vec3(2.2)), 1), clamp(ratio, 0, 1));
 #endif
-    color = pow(color, vec4(vec3(1.0 / 2.2), 1));
 }

src/client/Client.cpp

@@ -144,44 +144,30 @@ void Client::run(server_handle* const localHandle) {
                 frustum = {camera.getFrustum()};
             }
             const auto offset = state.position.raw_as_long();
-            { // Chunks
+            std::vector<glm::vec3> occlusion;
-                std::vector<glm::vec3> occlusion;
+            if (options.culling > 0) {
+                const auto ratio = options.culling * 2;
+                occlusion.reserve(glm::pow2(ratio * 2 - 1));
+                const auto [ch, cv] = player.getAngles();
+                const auto max_v = tan(options.camera.fov / 2.), max_h = Window::RATIO * max_v;
+                for(int iv = -ratio + 1; iv < ratio; iv++) {
+                    const auto v = cv + max_v * iv / ratio;
+                    for(int ih = -ratio + 1; ih < ratio; ih++) {
+                        const auto h = ch + max_h * ih / ratio;
+                        occlusion.emplace_back(cos(v) * sin(h), sin(v), cos(v) * cos(h));
+                    }
+                }
+            }
+            { // Solid
+                const auto pass = pipeline->beginWorldPass(true);
+                const auto draw = [&](glm::mat4 model, render::LodModel *const buffer, const contouring::Abstract::area_info &area, const voxel_pos &pos) {
+                    reports.models_count++;
+                    reports.tris_count += pass(buffer, model, glm::vec4(pos, std::get<1>(area)), std::get<2>(area));
+                };
                 if (options.culling > 0) {
-                    const auto ratio = options.culling * 2;
+                    state.contouring->getModels(draw, player.position, options.camera.far, occlusion, offset, options.voxel_density, true);
-                    occlusion.reserve(glm::pow2(ratio * 2 - 1));
+                } else {
-                    const auto [ch, cv] = player.getAngles();
+                    state.contouring->getModels(draw, frustum, offset, options.voxel_density, true);
-                    const auto max_v = tan(options.camera.fov / 2.), max_h = Window::RATIO * max_v;
-                    for(int iv = -ratio + 1; iv < ratio; iv++) {
-                        const auto v = cv + max_v * iv / ratio;
-                        for(int ih = -ratio + 1; ih < ratio; ih++) {
-                            const auto h = ch + max_h * ih / ratio;
-                            occlusion.emplace_back(cos(v) * sin(h), sin(v), cos(v) * cos(h));
-                        }
-                    }
-                }
-                { // Solid
-                    const auto pass = pipeline->beginWorldPass(true);
-                    const auto draw = [&](glm::mat4 model, render::LodModel *const buffer, const contouring::Abstract::area_info &area, const voxel_pos &pos) {
-                        reports.models_count++;
-                        reports.tris_count += pass(buffer, model, glm::vec4(pos, std::get<1>(area)), std::get<2>(area));
-                    };
-                    if (options.culling > 0) {
-                        state.contouring->getModels(draw, player.position, options.camera.far, occlusion, offset, options.voxel_density, true);
-                    } else {
-                        state.contouring->getModels(draw, frustum, offset, options.voxel_density, true);
-                    }
-                }
-                if (options.renderer.voxel.transparency) {
-                    const auto pass = pipeline->beginWorldPass(false);
-                    const auto draw = [&](glm::mat4 model, render::LodModel *const buffer, const contouring::Abstract::area_info &area, const voxel_pos &pos) {
-                        reports.models_count++;
-                        reports.tris_count += pass(buffer, model, glm::vec4(pos, std::get<1>(area)), std::get<2>(area));
-                    };
-                    if (options.culling > 0) {
-                        state.contouring->getModels(draw, player.position, options.camera.far, occlusion, offset, options.voxel_density, false);
-                    } else {
-                        state.contouring->getModels(draw, frustum, offset, options.voxel_density, false);
-                    }
                 }
             }
             { // Entities
@@ -203,6 +189,18 @@ void Client::run(server_handle* const localHandle) {
                     reports.tris_count += pass(model, options.editor.tool.shape, color);
                 }
             }
+            if (options.renderer.voxel.transparency) {
+                const auto pass = pipeline->beginWorldPass(false);
+                const auto draw = [&](glm::mat4 model, render::LodModel *const buffer, const contouring::Abstract::area_info &area, const voxel_pos &pos) {
+                    reports.models_count++;
+                    reports.tris_count += pass(buffer, model, glm::vec4(pos, std::get<1>(area)), std::get<2>(area));
+                };
+                if (options.culling > 0) {
+                    state.contouring->getModels(draw, player.position, options.camera.far, occlusion, offset, options.voxel_density, false);
+                } else {
+                    state.contouring->getModels(draw, frustum, offset, options.voxel_density, false);
+                }
+            }
             pipeline->postProcess();
             render::UI::Get()->render();
             pipeline->endFrame();

src/client/contouring/Abstract.hpp

@@ -31,6 +31,12 @@ namespace contouring {
         virtual void onNotify(const area_<chunk_pos> &pos, const chunk_pos &offset, const world::ChunkContainer &data) = 0;
         /// Display ImGui config
         virtual void onGui() = 0;
+        /// Register entity from model
+        virtual void onEntityLoad(size_t id, std::istream&) = 0;
+        /// Register entity from area
+        virtual void onEntityLoad(size_t id, const glm::ucvec3& size, const std::vector<std::shared_ptr<world::Chunk>>&) = 0;
+        /// Free entity model
+        virtual void onEntityUnload(size_t id) = 0;
         /// Get options
         virtual std::string getOptions() const = 0;
         /// Get camera recommended far range

src/client/contouring/FlatDualMC.cpp

@@ -11,6 +11,12 @@
 #include "optimizer.hpp"
 
 namespace contouring {
+    constexpr auto PACK = [](const render::VertexData &v) {
+        return render::PackedVertexData(meshopt_quantizeHalf(v.Position.x), meshopt_quantizeHalf(v.Position.y),
+            meshopt_quantizeHalf(v.Position.z), v.Material,
+            meshopt_quantizeHalf(v.Normal.x), meshopt_quantizeHalf(v.Normal.y),
+            meshopt_quantizeHalf(v.Normal.z));
+    };
     const std::vector<std::pair<float, float>> LEVELS = {{.001f, 9e-1f}, {.01f, 5e-1f}, {.1f, 1e-1f}, {.2, 1e-2f}, {.5, 5e-3f}};
 
     FlatDualMC::FlatDualMC(const std::string &str) : Abstract() {
@@ -43,10 +49,31 @@ namespace contouring {
 #endif
                 std::vector<render::VertexData> tmp;
                 while (running) {
-                    std::pair<area_<chunk_pos>, surrounding::corners> ctx;
+                    if (entity_area_job_t job; entityLoadQueue.pop(job)) {
-                    loadQueue.wait();
+                        ZoneScopedN("Entity");
-                    if (loadQueue.pop(ctx)) {
+                        render::Model::Data data;
-                        ZoneScopedN("ProcessContouring");
+                        typeof(render::Model::Data::indices) idx;
+                        for (uint8_t x = 0; x < job.size.x; x++) {
+                        for (uint8_t y = 0; y < job.size.y; y++) {
+                        for (uint8_t z = 0; z < job.size.z; z++) {
+                            surrounding::corners sur;
+                            surrounding::load(sur, job.size, glm::ucvec3(x, y, z), job.area);
+                            idx.clear();
+                            render(sur, idx, tmp, Layer::Both); //MAYBE: write inplace with sub-vector
+                            simplify(idx, tmp);
+
+                            data.indices.reserve(data.indices.size() + idx.size());
+                            const auto idx_start = data.vertices.size();
+                            std::transform(idx.begin(), idx.end(), std::back_inserter(data.indices), [&](glm::u16 i) { return i + idx_start; });
+                            data.vertices.reserve(data.vertices.size() + tmp.size());
+                            const auto vert_offset = glm::vec3(x * CHUNK_LENGTH, y * CHUNK_LENGTH, z * CHUNK_LENGTH);
+                            std::transform(tmp.begin(), tmp.end(), std::back_inserter(data.vertices), [&](render::VertexData v) {
+                                v.Position += vert_offset; return PACK(v); });
+                        }}}
+                        optimize_fetch(data);
+                        entityLoadedQueue.emplace(job.id, data);
+                    } else if (std::pair<area_<chunk_pos>, surrounding::corners> ctx; loadQueue.pop(ctx)) {
+                        ZoneScopedN("Chunk");
                         std::pair<render::LodModel::LodData, render::LodModel::LodData> data;
                         if (transparency) {
                             render(ctx.second, data.first, tmp, Layer::Solid);
@@ -56,26 +83,12 @@ namespace contouring {
                         }
                         //TODO: direct upload with vulkan
                         loadedQueue.emplace(ctx.first, data);
+                    } else {
+                        loadQueue.wait();
                     }
                 }
             });
         }
-
-        { //TODO: render for chunk(s)
-            auto pack = [](const glm::vec3 &pos) {
-                return render::PackedVertexData(meshopt_quantizeHalf(pos.x), meshopt_quantizeHalf(pos.y),
-                    meshopt_quantizeHalf(pos.z), world::materials::textures_map[7],
-                    meshopt_quantizeHalf(pos.x), meshopt_quantizeHalf(pos.y),
-                    meshopt_quantizeHalf(pos.z));
-            };
-            playerModel = render::Model::Create(render::Model::Data({
-                pack({1, 0, 0}), pack({0, 1, 0}), pack({0, 0, 1}),
-                pack({-1, 0, 0}), pack({0, -1, 0}), pack({0, 0, -1})
-            }, {
-                0, 1, 2, 0, 5, 1, 0, 2, 4, 0, 4, 5,
-                1, 3, 2, 5, 3, 1, 2, 3, 4, 4, 3, 5
-            }));
-        }
     }
     FlatDualMC::~FlatDualMC() {
         running = false;
@@ -85,7 +98,7 @@ namespace contouring {
             if (worker.joinable())
                 worker.join();
         }
-        
+
         //TODO: prefer unique_ptr
         for(auto& buffer: buffers) {
             for(auto& val: buffer.second.second) {
@@ -149,12 +162,36 @@ namespace contouring {
         }
     }
 
+    void FlatDualMC::onEntityLoad(size_t id, std::istream &in) {
+        if(!entities.get(id))
+            entities.set_emplace(id, nullptr);
+        entityLoadedQueue.emplace(id, render::Model::Data(in));
+    }
+    void FlatDualMC::onEntityLoad(size_t id, const glm::ucvec3& size, const std::vector<std::shared_ptr<world::Chunk>>& area) {
+        if(!entities.get(id))
+            entities.set_emplace(id, nullptr);
+        entityLoadQueue.emplace(entity_area_job_t{id, size, area});
+        loadQueue.notify_one();
+    }
+    void FlatDualMC::onEntityUnload(size_t id) {
+        entities.erase(id);
+    }
+
     void FlatDualMC::update(const voxel_pos& pos, const world::client::area_map& areas) {
         ZoneScopedN("Ct");
-        std::pair<area_<chunk_pos>, std::pair<render::LodModel::LodData, render::LodModel::LodData>> out;
+        TracyPlot("CtLoad", static_cast<int64_t>(loadQueue.size() + entityLoadQueue.size()));
-        TracyPlot("CtLoad", static_cast<int64_t>(loadQueue.size()));
         //MAYBE: clear out of range loadQueue.trim(keepDistance * keepDistance)
-        TracyPlot("CtLoaded", static_cast<int64_t>(loadedQueue.size()));
+        TracyPlot("CtLoaded", static_cast<int64_t>(loadedQueue.size() + entityLoadedQueue.size()));
+
+        {
+            std::pair<size_t, render::Model::Data> out;
+            for(auto handle = entityLoadedQueue.extractor(); handle.first(out);) {
+                if (auto entity = entities.get(out.first))
+                    *entity = render::Model::Create(out.second);
+            }
+        }
+
+        std::pair<area_<chunk_pos>, std::pair<render::LodModel::LodData, render::LodModel::LodData>> out;
         for(auto handle = loadedQueue.extractor(); handle.first(out);) {
             const auto bufferSolid = out.second.first.first.empty() ? NULL : render::LodModel::Create(out.second.first).release();
             const auto bufferTrans = out.second.second.first.empty() ? NULL : render::LodModel::Create(out.second.second).release();
@@ -265,7 +302,7 @@ namespace contouring {
         }
     }
 
-    void FlatDualMC::render(const surrounding::corners &surrounding, render::LodModel::LodData &out, std::vector<render::VertexData> &tmp, Layer layer) const {
+    void FlatDualMC::render(const surrounding::corners &surrounding, typeof(render::Model::Data::indices) &out, std::vector<render::VertexData> &tmp, Layer layer) const {
         const int SIZE = CHUNK_LENGTH + 3;
         std::array<dualmc::DualMC<float>::Point, SIZE * SIZE * SIZE> grid;
         {
@@ -288,11 +325,10 @@ namespace contouring {
             for (size_t i = 0; i < surrounding.size(); i++) {
                 auto &edits = surrounding[i]->getEdits();
                 auto offset = glm::ivec3(surrounding::g_corner_offsets[i]) * CHUNK_LENGTH;
-                for (auto it = edits.end(); it != edits.begin();) {
+                for (auto it = edits.begin(); it != edits.end(); ++it) {
-                    it--;
+                    auto p = offset + glm::ivec3(glm::fromIdx(it->first));
-                    auto p = offset + glm::ivec3(glm::fromIdx(it->idx));
                     if(p.x < SIZE && p.y < SIZE && p.z < SIZE) {
-                        setCell(p.x, p.y, p.z, it->value);
+                        setCell(p.x, p.y, p.z, it->second.value);
                     }
                 }
             }
@@ -305,7 +341,6 @@ namespace contouring {
             builder.buildTris(&grid.front(), SIZE, SIZE, SIZE, iso, world::materials::textures_map.cbegin(), world::materials::roughness.cbegin(),
                 manifold, dmc_vertices, dmc_tris);
 
-            auto &data = out.first;
             tmp.clear();
             tmp.reserve(dmc_vertices.size());
             constexpr auto HALF_MANTISSA = 10;
@@ -314,20 +349,20 @@ namespace contouring {
                     meshopt_quantizeFloat(v.z, HALF_MANTISSA)), v.w, glm::vec3(0));
             });
 
-            data.indices.reserve(dmc_tris.size() * 3);
+            out.reserve(dmc_tris.size() * 3);
             for (const auto& t: dmc_tris) {
                 glm::vec3 edge1 = tmp[t.i1].Position - tmp[t.i0].Position;
                 glm::vec3 edge2 = tmp[t.i2].Position - tmp[t.i0].Position;
                 glm::vec3 normal = glm::normalize(glm::cross(edge1, edge2));
 
                 if(!reordering || glm::length2(edge1) > glm::length2(edge2)) {
-                    data.indices.push_back(t.i0);
+                    out.push_back(t.i0);
-                    data.indices.push_back(t.i1);
+                    out.push_back(t.i1);
-                    data.indices.push_back(t.i2);
+                    out.push_back(t.i2);
                 } else {
-                    data.indices.push_back(t.i2);
+                    out.push_back(t.i2);
-                    data.indices.push_back(t.i0);
+                    out.push_back(t.i0);
-                    data.indices.push_back(t.i1);
+                    out.push_back(t.i1);
                 }
 
                 tmp[t.i0].Normal += normal;
@@ -338,17 +373,15 @@ namespace contouring {
             for(auto& v: tmp) {
                 v.Normal = glm::normalize(v.Normal);
             }
-
-            out.second = simplify_lod(data.indices, tmp, loadedLevels);
-            std::transform(tmp.begin(), tmp.end(), std::back_inserter(data.vertices), [](const render::VertexData &v) {
-                return render::PackedVertexData(meshopt_quantizeHalf(v.Position.x), meshopt_quantizeHalf(v.Position.y),
-                    meshopt_quantizeHalf(v.Position.z), v.Material,
-                    meshopt_quantizeHalf(v.Normal.x), meshopt_quantizeHalf(v.Normal.y),
-                    meshopt_quantizeHalf(v.Normal.z));
-            });
-            optimize_fetch(data);
         }
     }
+    void FlatDualMC::render(const surrounding::corners &surrounding, render::LodModel::LodData &out, std::vector<render::VertexData> &tmp, Layer layer) const {
+        render(surrounding, out.first.indices, tmp, layer);
+
+        out.second = simplify_lod(out.first.indices, tmp, loadedLevels);
+        std::transform(tmp.begin(), tmp.end(), std::back_inserter(out.first.vertices), PACK);
+        optimize_fetch(out.first);
+    }
 
     void FlatDualMC::getModels(draw_call out, const std::optional<geometry::Frustum> &frustum, const glm::llvec3& offset, int density, bool solid) {
         const auto scaling = glm::scale(glm::mat4(1), glm::vec3(1.f / density));
@@ -386,8 +419,9 @@ namespace contouring {
         }
     }
 
-    render::Model* FlatDualMC::getEntityModel(size_t) {
+    render::Model* FlatDualMC::getEntityModel(size_t id) {
-        return playerModel.get();
+        if (auto ptr = entities.get(id))
+            return ptr->get();
+        return nullptr;
     }
-
 }

src/client/contouring/FlatDualMC.hpp

@@ -31,6 +31,13 @@ namespace contouring {
         /// @note notify for chunks entering view while moving
         void onNotify(const area_<chunk_pos> &, const chunk_pos &, const world::ChunkContainer &) override;
 
+        /// Register entity from model
+        void onEntityLoad(size_t id, std::istream&) override;
+        /// Register entity from area
+        void onEntityLoad(size_t id, const glm::ucvec3& size, const std::vector<std::shared_ptr<world::Chunk>>&) override;
+        /// Free entity model
+        void onEntityUnload(size_t id) override;
+
         /// Get buffers in frustum with model matrices
         /// @note buffers invalidated after update
         void getModels(draw_call draw, const std::optional<geometry::Frustum> &frustum, const glm::llvec3 &offset, int density, bool solid) override;
@@ -44,10 +51,13 @@ namespace contouring {
         //FIXME: use unique_ptr
         robin_hood::unordered_map<area_id, robin_hood::pair<area_info, robin_hood::unordered_map<chunk_pos, std::pair<render::LodModel *, render::LodModel *>>>> buffers;
 
-        std::unique_ptr<render::Model> playerModel;
+        data::generational::view_vector<std::unique_ptr<render::Model>> entities;
 
         safe_priority_queue_map<area_<chunk_pos>, surrounding::corners, int, area_hash> loadQueue;
         safe_queue<std::pair<area_<chunk_pos>, std::pair<render::LodModel::LodData, render::LodModel::LodData>>> loadedQueue;
+        struct entity_area_job_t { size_t id; glm::ucvec3 size; std::vector<std::shared_ptr<world::Chunk>> area; };
+        safe_queue<entity_area_job_t> entityLoadQueue;
+        safe_queue<std::pair<size_t, render::Model::Data>> entityLoadedQueue;
 
         bool running = true;
         std::vector<std::thread> workers;
@@ -75,6 +85,7 @@ namespace contouring {
             return static_cast<int>(a) & static_cast<int>(b);
         }
 
+        void render(const surrounding::corners &surrounding, typeof(render::Model::Data::indices)& idx, std::vector<render::VertexData>& ver, Layer layer) const;
         void render(const surrounding::corners &surrounding, render::LodModel::LodData& out, std::vector<render::VertexData>& tmp, Layer layer) const;
     };
 }

src/client/contouring/optimizer.hpp

@@ -35,4 +35,10 @@ inline std::vector<size_t> simplify_lod(std::vector<I> &indices, const std::vect
     indices.reserve(indices.size() + full.size());
     indices.insert(indices.end(), full.begin(), full.end());
     return offsets;
+}
+template <typename I>
+inline void simplify(std::vector<I> &indices, const std::vector<render::VertexData> &vertices) {
+    ZoneScopedN("LOD");
+    simplify_buffer(indices, indices, vertices);
+    meshopt_optimizeVertexCache(indices.data(), indices.data(), indices.size(), vertices.size());
 }

src/client/contouring/surrounding.cpp

@@ -27,4 +27,14 @@ namespace contouring::surrounding {
         }
         return true;
     }
+    void load(corners &out, const glm::ucvec3 &areaSize, const glm::ucvec3 &chunkPos, const std::vector<std::shared_ptr<world::Chunk>> &area) {
+        for (size_t i = 0; i < 8; i++) {
+            const auto pos = chunkPos + glm::ucvec3(g_corner_offsets[i]);
+            if (pos.x < areaSize.x && pos.y < areaSize.y && pos.z < areaSize.z) {
+                out[i] = std::dynamic_pointer_cast<world::client::EdittableChunk>(area.at(glm::toIdx(pos, areaSize)));
+            } else {
+                out[i] = world::client::EMPTY_CHUNK;
+            }
+        }
+    }
 }

src/client/contouring/surrounding.hpp

@@ -16,4 +16,5 @@ namespace contouring::surrounding {
     };
 
     bool load(corners &out, const chunk_pos &chunkPos, const world::ChunkContainer &chunks);
+    void load(corners &out, const glm::ucvec3 &areaSize, const glm::ucvec3 &chunkPos, const std::vector<std::shared_ptr<world::Chunk>> &area);
 }

src/client/render/UI.cpp

@@ -226,7 +226,7 @@ UI::Actions UI::draw(config::client::options &options, state::state &state, cons
             }
             ImGui::EndCombo();
         }
-        ImGui::SliderInt("Radius", &options.editor.tool.radius, 0, 10);
+        ImGui::SliderInt("Radius", &options.editor.tool.radius, 1, 15);
         const auto shapeId = (size_t)options.editor.tool.shape;
         if (ImGui::BeginCombo("Shape", world::action::SHAPES[shapeId].c_str())) {
             for (size_t i = 0; i < world::action::SHAPES.size(); i++) {

src/client/render/api/Models.cpp

@@ -43,7 +43,7 @@ const std::pair<std::vector<glm::vec3>, std::vector<glm::vec4>> ColoredShape::LI
 void indexVBO(const std::vector<PackedVertexData> &in_vertices, std::vector<glm::u16> &out_indices, std::vector<PackedVertexData> &out_vertices);
 
 Model::Data::Data(const std::vector<PackedVertexData> &vs, const std::vector<glm::u16> &indices): indices(indices), vertices(vs) { }
-Model::Data::Data(std::ifstream& in) {
+Model::Data::Data(std::istream& in) {
     {
         size_t i_size;
         in.read(reinterpret_cast<char *>(&i_size), sizeof(i_size));
@@ -55,7 +55,7 @@ Model::Data::Data(std::ifstream& in) {
     vertices.resize(v_size, PackedVertexData(0, 0, 0, 0, 0, 0, 0, 0));
     in.read(reinterpret_cast<char *>(vertices.data()), sizeof(PackedVertexData) * v_size);
 }
-void Model::Data::serialize(std::ofstream& out) {
+void Model::Data::serialize(std::ostream& out) {
     {
         size_t i_size = indices.size();
         out.write(reinterpret_cast<char *>(&i_size), sizeof(i_size));

src/client/render/api/Models.hpp

@@ -70,7 +70,7 @@ public:
         Data() { }
         Data(const std::vector<PackedVertexData> &vertices, const std::vector<glm::u16> &indices);
         Data(const std::vector<PackedVertexData> &vertices) { index(vertices); }
-        Data(std::ifstream &in);
+        Data(std::istream &in);
 
         void index(const std::vector<PackedVertexData> &vertices);
         bool empty() const {
@@ -81,7 +81,7 @@ public:
             vertices.clear();
         }
 
-        void serialize(std::ofstream &out);
+        void serialize(std::ostream &out);
     };
 
     static _FORCE_INLINE_ std::unique_ptr<Model> Create(const Data& data) {

src/client/render/gl/Renderer.cpp

@@ -101,13 +101,12 @@ void Renderer::loadUI(Window& w) {
 void Renderer::beginFrame() {
     TracyGpuZone("Render");
     glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+    glEnable(GL_FRAMEBUFFER_SRGB);
 }
 
 std::function<size_t(render::LodModel *const, glm::mat4, glm::vec4, float)> Renderer::beginWorldPass(bool solid) {
-    if (solid) {
+    WorldPass->useIt();
-        WorldPass->useIt();
+    WorldPass->start(this);
-        WorldPass->start(this);
-    }
     return [&](render::LodModel *const buf, glm::mat4 model, glm::vec4 sph, float curv) {
         WorldPass->setup(model, sph, curv);
         return dynamic_cast<LodModel *const>(buf)->draw();
@@ -133,6 +132,7 @@ std::function<size_t(glm::mat4, world::action::Shape, glm::vec4)> Renderer::begi
 }
 
 void Renderer::postProcess() {
+    glDisable(GL_FRAMEBUFFER_SRGB);
     if(SkyEnable) {
         SkyPass->draw(this);
     }

src/client/render/vk/CommandCenter.hpp

@@ -11,7 +11,7 @@
 namespace tracy {
     class VkCtx;
 }
-typedef tracy::VkCtx* TracyVkCtx;
+typedef tracy::VkCtx* TracyVkCtxPtr;
 
 namespace render::vk {
 class SwapChain;
@@ -54,7 +54,7 @@ private:
     VkQueue graphicsQueue;
     VkCommandPool graphicsPool;
     std::vector<VkCommandBuffer> graphicsBuffers;
-    TracyVkCtx tracyCtx;
+    TracyVkCtxPtr tracyCtx;
 
     BufferGroup uniformBuffers;
 

src/client/render/vk/Pipeline.cpp

@@ -65,9 +65,9 @@ Pipeline::Pipeline(VkDevice device, const PhysicalDeviceInfo &info, const render
         if (hasSamples) {
             colorDepthSubpass.pResolveAttachments = &colorAttachmentResolveRef;
         }
-        std::array<VkSubpassDescription, 3> subpasses = {colorDepthSubpass, colorDepthSubpass, colorDepthSubpass};
+        std::array<VkSubpassDescription, 4> subpasses = {colorDepthSubpass, colorDepthSubpass, colorDepthSubpass, colorDepthSubpass};
 
-        std::array<VkSubpassDependency, 5> dependencies{};
+        std::array<VkSubpassDependency, 7> dependencies{};
         dependencies[0].srcSubpass = VK_SUBPASS_EXTERNAL;
         dependencies[0].dstSubpass = 0;
         dependencies[0].srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
@@ -108,6 +108,22 @@ Pipeline::Pipeline(VkDevice device, const PhysicalDeviceInfo &info, const render
         dependencies[4].dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
         dependencies[4].dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;
 
+        dependencies[5].srcSubpass = 2;
+        dependencies[5].dstSubpass = 3;
+        dependencies[5].srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
+        dependencies[5].srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
+        dependencies[5].dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
+        dependencies[5].dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
+        dependencies[5].dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;
+
+        dependencies[6].srcSubpass = 2;
+        dependencies[6].dstSubpass = 3;
+        dependencies[6].srcStageMask = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
+        dependencies[6].srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+        dependencies[6].dstStageMask = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
+        dependencies[6].dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+        dependencies[6].dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;
+
         VkRenderPassCreateInfo renderPassInfo{};
         renderPassInfo.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO;
         std::vector<VkAttachmentDescription> attachments = {colorAttachment, depthAttachment};
@@ -507,6 +523,99 @@ Pipeline::Pipeline(VkDevice device, const PhysicalDeviceInfo &info, const render
             FATAL("Failed to create graphics pipeline!");
         }
     }
+    { // Transparent World pipeline
+        VkPushConstantRange pushRange{};
+        pushRange.offset = 0;
+        pushRange.size = sizeof(UniqueCurvaturePush);
+        pushRange.stageFlags = VK_SHADER_STAGE_VERTEX_BIT;
+        setLayout(transparentWorldPass, {voxelDescriptorSet}, {pushRange});
+
+        struct SpeData {
+            bool fog;
+            bool pbr;
+            bool triplanar;
+            bool stochastic;
+            bool blend;
+            bool curvature;
+            bool curv_depth;
+            int32_t unitSize;
+        } speData;
+        std::array<VkSpecializationMapEntry, 8> speIndex;
+        speData.fog = options.voxel.fog;
+        speIndex[0].constantID = 0;
+        speIndex[0].offset = offsetof(SpeData, fog);
+        speIndex[0].size = sizeof(SpeData::fog);
+        speData.pbr = options.voxel.pbr;
+        speIndex[1].constantID = 1;
+        speIndex[1].offset = offsetof(SpeData, pbr);
+        speIndex[1].size = sizeof(SpeData::pbr);
+        speData.triplanar = options.voxel.triplanar;
+        speIndex[2].constantID = 2;
+        speIndex[2].offset = offsetof(SpeData, triplanar);
+        speIndex[2].size = sizeof(SpeData::triplanar);
+        speData.stochastic = options.voxel.stochastic;
+        speIndex[3].constantID = 3;
+        speIndex[3].offset = offsetof(SpeData, stochastic);
+        speIndex[3].size = sizeof(SpeData::stochastic);
+        speData.blend = options.voxel.blend;
+        speIndex[4].constantID = 4;
+        speIndex[4].offset = offsetof(SpeData, blend);
+        speIndex[4].size = sizeof(SpeData::blend);
+        speData.curvature = options.voxel.curvature;
+        speIndex[5].constantID = 5;
+        speIndex[5].offset = offsetof(SpeData, curvature);
+        speIndex[5].size = sizeof(SpeData::curvature);
+        speData.curv_depth = options.voxel.curv_depth;
+        speIndex[6].constantID = 6;
+        speIndex[6].offset = offsetof(SpeData, curv_depth);
+        speIndex[6].size = sizeof(SpeData::curv_depth);
+        speData.unitSize = 8; //TODO: load from world.voxel_density
+        speIndex[7].constantID = 16;
+        speIndex[7].offset = offsetof(SpeData, unitSize);
+        speIndex[7].size = sizeof(SpeData::unitSize);
+        VkSpecializationInfo specialization{};
+        specialization.dataSize = sizeof(SpeData);
+        specialization.pData = &speData;
+        specialization.mapEntryCount = speIndex.size();
+        specialization.pMapEntries = speIndex.data();
+
+        auto withGeometry = options.voxel.geometry && info.features.geometryShader;
+        auto shaderStages = setShaders(transparentWorldPass, withGeometry ? "Voxel.geo" : "Voxel", withGeometry, &specialization);
+
+        VkPipelineVertexInputStateCreateInfo vertexInputInfo{};
+        vertexInputInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;
+
+        auto bindingDescription = Model::getBindingDescription();
+        vertexInputInfo.vertexBindingDescriptionCount = 1;
+        vertexInputInfo.pVertexBindingDescriptions = &bindingDescription;
+        auto attributeDescriptions = Model::getAttributeDescriptions();
+        vertexInputInfo.vertexAttributeDescriptionCount = attributeDescriptions.size();
+        vertexInputInfo.pVertexAttributeDescriptions = attributeDescriptions.data();
+
+        VkGraphicsPipelineCreateInfo pipelineInfo{};
+        pipelineInfo.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO;
+        pipelineInfo.stageCount = shaderStages.size();
+        pipelineInfo.pStages = shaderStages.data();
+        pipelineInfo.pVertexInputState = &vertexInputInfo;
+        pipelineInfo.pInputAssemblyState = &trisInputAssembly;
+        pipelineInfo.pViewportState = &viewportState;
+        pipelineInfo.pRasterizationState = &rasterizer;
+        pipelineInfo.pMultisampleState = &multisampling;
+        pipelineInfo.pDepthStencilState = &depthStencil;
+        pipelineInfo.pColorBlendState = &colorBlending;
+        pipelineInfo.pDynamicState = nullptr;
+
+        pipelineInfo.layout = transparentWorldPass.layout;
+        pipelineInfo.renderPass = renderPass;
+        pipelineInfo.subpass = 2;
+
+        pipelineInfo.basePipelineHandle = VK_NULL_HANDLE;
+        pipelineInfo.basePipelineIndex = -1;
+
+        if (vkCreateGraphicsPipelines(device, VK_NULL_HANDLE, 1, &pipelineInfo, ALLOC, &transparentWorldPass.pipeline) != VK_SUCCESS) {
+            FATAL("Failed to create graphics pipeline!");
+        }
+    }
     { // Sky pipeline
         setLayout(skyPass, {skyDescriptorSet}, {});
         auto shaderStages = setShaders(skyPass, "Sky");
@@ -537,7 +646,7 @@ Pipeline::Pipeline(VkDevice device, const PhysicalDeviceInfo &info, const render
 
         pipelineInfo.layout = skyPass.layout;
         pipelineInfo.renderPass = renderPass;
-        pipelineInfo.subpass = 2;
+        pipelineInfo.subpass = 3;
 
         pipelineInfo.basePipelineHandle = VK_NULL_HANDLE;
         pipelineInfo.basePipelineIndex = -1;

src/client/render/vk/Pipeline.hpp

@@ -26,6 +26,7 @@ public:
     constexpr const Subpass &getEntitiesPass() const { return worldPass; }
     constexpr VkDescriptorSetLayout getIndicDescriptorSet() const { return indicDescriptorSet; }
     constexpr const Subpass &getIndicPass() const { return indicPass; }
+    constexpr const Subpass& getTransparentWorldPass() const { return transparentWorldPass; }
     constexpr VkDescriptorSetLayout getSkyDescriptorSet() const { return skyDescriptorSet; }
     constexpr const Subpass& getSkyPass() const { return skyPass; }
 
@@ -42,6 +43,7 @@ private:
 
     Subpass worldPass;
     Subpass indicPass;
+    Subpass transparentWorldPass;
     Subpass skyPass;
 
     VkRenderPass uiRenderPass;

src/client/render/vk/Renderer.cpp

@@ -454,10 +454,8 @@ void Renderer::beginFrame() {
 std::function<size_t(render::LodModel *const, glm::mat4, glm::vec4, float)> Renderer::beginWorldPass(bool solid) {
     assert(currentImage < swapChain->getImageViews().size());
 
-    auto& pass = pipeline->getWorldPass();
+    auto& pass = solid ? pipeline->getWorldPass() : pipeline->getTransparentWorldPass();
-    if (solid) {
+    commandCenter->startWorldPass(currentImage, pass);
-        commandCenter->startWorldPass(currentImage, pass);
-    }
     return [&](render::LodModel *const rBuffer, glm::mat4 model, glm::vec4 sphere, float curv) {
         auto buffer = dynamic_cast<render::vk::LodModel *const>(rBuffer);
         buffer->setLastUse(currentImage);

src/client/world/Chunk.hpp

@@ -13,13 +13,13 @@ public:
 
     std::optional<Faces> update(float deltaTime, bool animate) override {
         for (auto it = futureEdits.begin(); it != futureEdits.end();) {
-            it->second -= deltaTime;
+            it->second.second -= deltaTime;
-            if (it->second <= 0 && animate) {
+            if (it->second.second <= 0 && animate) {
-                invalidate(it->first.idx);
+                invalidate(it->first);
-                edits.emplace_back(it->first);
+                edits.emplace(it->first, it->second.first);
                 it = futureEdits.erase(it);
             } else {
-                it++;
+                ++it;
             }
         }
         return EdittableChunk::update(deltaTime, animate);
@@ -32,7 +32,7 @@ public:
     }
 
     /// Add future edit
-    void addFutureEdit(Chunk::Edit edit, float delay) { futureEdits.emplace_back(edit, delay); }
+    void addFutureEdit(Chunk::Edit edit, float delay) { futureEdits.insert_or_assign(edit.idx, std::make_pair(EditBody(edit), delay)); }
 
 protected:
     /// Is temporary average
@@ -41,8 +41,7 @@ protected:
     bool isMajorant = false;
 
     /// Temporary animated changes
-    /// MAYBE: sort by delay
+    robin_hood::unordered_map<chunk_voxel_idx, std::pair<Chunk::EditBody, float>> futureEdits;
-    std::vector<std::pair<Chunk::Edit, float>> futureEdits;
 };
 
 /// Chunk full of air

src/client/world/DistantUniverse.cpp

@@ -9,6 +9,7 @@
 #include "../../core/net/io.hpp"
 #include "../../core/utils/logger.hpp"
 #include "Chunk.hpp"
+#include <random>
 
 using namespace world::client;
 
@@ -29,6 +30,8 @@ void DistantUniverse::update(voxel_pos pos, float deltaTime) {
 
     { // Update alive areas
         ZoneScopedN("World");
+        auto rng = std::mt19937(std::rand());
+        const auto contouringThreshold = rng.max() / (1 + contouring->getQueueSize());
         for (auto& area: areas) {
             ZoneScopedN("Area");
             const bool chunkChangeArea = (false && area.second->move(glm::vec3(deltaTime))) || chunkChange; // TODO: area.velocity
@@ -41,7 +44,7 @@ void DistantUniverse::update(voxel_pos pos, float deltaTime) {
                     if (glm::length2(diff - it_c->first) > glm::pow2(options.keepDistance)) {
                         it_c = chunks.erase(it_c);
                     } else {
-                        if(const auto neighbors = std::dynamic_pointer_cast<Chunk>(it_c->second)->update(deltaTime, true /*MAYBE: random update*/)) {
+                        if(const auto neighbors = std::dynamic_pointer_cast<Chunk>(it_c->second)->update(deltaTime, rng() < contouringThreshold)) {
                             contouring->onUpdate(std::make_pair(area.first, it_c->first), diff, chunks, neighbors.value());
                         }
                         ++it_c;
@@ -155,7 +158,9 @@ bool DistantUniverse::onPacket(const data::out_view& buf, net::PacketFlags) {
             packet.read(serverDistance);
             auto predictable = false;
             packet.read(predictable);
-            if (!predictable) {
+            if (predictable) {
+                packet.read(floodfillLimit);
+            } else {
                 options.editHandling = false;
                 options.editPrediction = false;
             }
@@ -290,24 +295,8 @@ bool DistantUniverse::onPacket(const data::out_view& buf, net::PacketFlags) {
             if (!fill)
                 break;
 
-            if(const auto it = areas.find(fill->pos.first); it != areas.end()) {
+            world::iterator::ApplySplit<Chunk>(areas, *fill, floodfillLimit, [](std::shared_ptr<Chunk> &ck, chunk_pos, chunk_voxel_idx idx,
-                auto &chunks = it->second->setChunks();
+                Voxel, Voxel next, float delay) { ck->apply(Chunk::Edit{next, delay, idx}); });
-                auto iterator = world::iterator::Get(fill->shape, fill->radius);
-                world::iterator::pair point;
-                while (iterator->next(point)) {
-                    const voxel_pos offset = point.first;
-                    const auto split = glm::splitIdx(fill->pos.second + offset);
-                    if(chunks.inRange(split.first)) {
-                        if(const auto chunk = it->second->setChunks().findInRange(split.first)) {
-                            auto ck = std::dynamic_pointer_cast<Chunk>(chunk.value());
-                            auto prev = ck->get(split.second);
-                            const auto next = prev.filled(fill->val, point.second);
-                            const auto delay = glm::length2(offset) / fill->radius * .05f;
-                            ck->apply(Chunk::Edit{split.second, next, delay});
-                        }
-                    }
-                }
-            }
 
             break;
         }
@@ -346,16 +335,21 @@ bool DistantUniverse::onPacket(const data::out_view& buf, net::PacketFlags) {
             entities.clear();
             while(!packet.isFull()) {
                 size_t index;
-                glm::vec3 size;
+                glm::usvec3 size;
                 glm::vec3 scale;
-                if(packet.read(index) && packet.read(size) && packet.read(scale)) {
+                uint8_t flags;
-                    entities.set_emplace(index, size, scale);
+                if(packet.read(index) && packet.read(size) && packet.read(scale) && packet.read(flags)) {
+                    const bool permanant = (flags & (1 << 0)) != 0;
+                    const bool area = (flags & (1 << 1)) != 0;
+                    entities.set(index, Entity(size, scale, permanant, area));
+                    //MAYBE: just update
                 }
             }
             break;
         }
 
         case server_packet_type::ENTITIES: {
+            entities.remove([](size_t, Entity &entity) { entity.instances.clear(); return false; });
             while(!packet.isFull()) {
                 size_t entity_idx;
                 size_t count;
@@ -363,7 +357,8 @@ bool DistantUniverse::onPacket(const data::out_view& buf, net::PacketFlags) {
                     break;
 
                 if (auto entity = entities.get(entity_idx)) {
-                    entity->instances.clear();
+                    if (count == 0)
+                        continue;
                     for (size_t i = 0; i < count && !packet.isFull(); i++) {
                         size_t idx;
                         glm::ifvec3 pos;
@@ -371,11 +366,85 @@ bool DistantUniverse::onPacket(const data::out_view& buf, net::PacketFlags) {
                         if (packet.read(idx) && packet.read(pos) && packet.read(vel))
                             entity->instances.set_emplace(idx, Universe::Entity::Instance{pos, vel});
                     }
+                    if (dict.has_value()) {
+                        if (auto area = std::get_if<Universe::Entity::area_t>(&entity->shape)) {
+                            if (area->capacity())
+                                continue;
+
+                            const auto scale = glm::lvec3(1) + glm::divide(entity->size);
+                            area->reserve(scale.x * scale.y * scale.z);
+                        } else if (auto model = std::get_if<Universe::Entity::model_t>(&entity->shape)) {
+                            if (model->capacity())
+                                continue;
+
+                            model->reserve(8);
+                        }
+                        peer.send(PacketWriter::Of(client_packet_type::MISSING_ENTITY, entity_idx));
+                    }
                 } else {
                     LOG_W("Unknown entity type " << entity_idx);
                     packet.skip(count * (sizeof(size_t) + sizeof(glm::ifvec3) + sizeof(glm::vec3)));
                 }
             }
+            entities.remove([&](size_t id, const Entity &entity) {
+                if (!entity.permanant && entity.instances.empty()) {
+                    contouring->onEntityUnload(id);
+                    return true;
+                }
+                return false;
+            });
+            break;
+        }
+
+        case server_packet_type::ENTITY_SHAPE: {
+            assert(dict.has_value());
+
+            size_t id;
+            bool is_area;
+            if (!packet.read(id) || !packet.read(is_area))
+                break;
+
+            if (auto it = entities.get(id)) {
+                if (is_area) {
+                    auto area = std::get_if<Universe::Entity::area_t>(&it->shape);
+                    assert(area);
+                    area->clear();
+                    glm::i64 i = 0;
+                    while (!packet.isFull()) {
+                        i++;
+                        size_t size = 0;
+                        if (!packet.read(size))
+                            break;
+
+                        std::vector<char> buffer;
+                        if(auto err = dict.value().decompress(packet.readPart(size), buffer)) {
+                            LOG_E("Corrupted entity chunk packet " << err.value());
+                            break;
+                        }
+                        data::vec_istream idata(buffer);
+                        std::istream iss(&idata);
+                        area->push_back(std::make_shared<Chunk>(iss));
+                    }
+                    auto scale = glm::lvec3(1) + glm::divide(it->size);
+                    if (i != scale.x * scale.y * scale.z) {
+                        LOG_E("Corrupted entity area");
+                        break;
+                    }
+                    contouring->onEntityLoad(id, scale, *area);
+                } else {
+                    auto model = std::get_if<Universe::Entity::model_t>(&it->shape);
+                    assert(model);
+                    auto remain = packet.readAll();
+                    model->resize(remain.size());
+                    //MAYBE: avoid storing model
+                    memcpy(model->data(), remain.data(), remain.size());
+                    data::vec_istream idata(*model);
+                    std::istream iss(&idata);
+                    contouring->onEntityLoad(id, iss);
+                }
+            } else {
+                LOG_W("Shape for unknown entity type " << id);
+            }
             break;
         }
 
@@ -395,27 +464,9 @@ void DistantUniverse::emit(const action::packet &action) {
         peer.send(net::PacketWriter::Of(net::client_packet_type::FILL_SHAPE, *fill));
         if (options.editPrediction) {
             ZoneScopedN("Fill");
-            const auto keepDelay = 10 + (peer.getRTT() / 20000.f); // 10s + 50RTT
+            const auto keepDelay = 5 + (peer.getRTT() / 20000.f); // 5s + 50RTT
-            if(const auto it = areas.find(fill->pos.first); it != areas.end()) {
+            world::iterator::ApplySplit<Chunk>(areas, *fill, floodfillLimit, [&](std::shared_ptr<Chunk> &ck, chunk_pos, chunk_voxel_idx idx,
-                auto &chunks = it->second->setChunks();
+                Voxel, Voxel next, float delay) { ck->addFutureEdit(Chunk::Edit{next, keepDelay - delay * 2, idx}, delay); });
-                auto iterator = world::iterator::Get(fill->shape, fill->radius);
-                world::iterator::pair point;
-                while (iterator->next(point)) {
-                    const voxel_pos offset = point.first;
-                    const auto split = glm::splitIdx(fill->pos.second + offset);
-                    if(chunks.inRange(split.first)) {
-                        if(const auto chunk = it->second->setChunks().findInRange(split.first)) {
-                            auto ck = std::dynamic_pointer_cast<Chunk>(chunk.value());
-                            auto prev = ck->get(split.second);
-                            const auto next = prev.filled(fill->val, point.second);
-                            if(prev.value != next.value) {
-                                const auto delay = glm::length2(offset) / fill->radius * .05f;
-                                ck->addFutureEdit(Chunk::Edit{split.second, next, keepDelay - delay * 2}, delay);
-                            }
-                        }
-                    }
-                }
-            }
         }
     } else {
         LOG_W("Bad action " << action.index());

src/client/world/DistantUniverse.hpp

@@ -35,9 +35,19 @@ namespace world::client {
 
         /// Entities without generation
         struct Entity {
-            Entity(glm::vec3 size, glm::vec3 scale): size(size), scale(scale) { }
+            Entity(glm::usvec3 size, glm::vec3 scale, bool permanant, bool area): size(size), scale(scale), permanant(permanant) {
-            glm::vec3 size;
+                if (area) {
+                    shape = Universe::Entity::area_t();
+                } else {
+                    shape = Universe::Entity::model_t();
+                }
+            }
+
+            std::variant<Universe::Entity::area_t, Universe::Entity::model_t> shape;
+            glm::usvec3 size;
             glm::vec3 scale;
+            bool permanant;
+
             data::generational::view_vector<Universe::Entity::Instance> instances;
         };
         data::generational::view_vector<Entity> entities;
@@ -51,6 +61,7 @@ namespace world::client {
 
         options options;
         ushort serverDistance;
+        size_t floodfillLimit = CHUNK_SIZE;
 
         net::client::Client peer;
     };

src/client/world/LocalUniverse.cpp

@@ -1,6 +1,7 @@
 #include "LocalUniverse.hpp"
 
 #include "../../core/data/math.hpp"
+#include "../../core/data/mem.hpp"
 #include "../contouring/Abstract.hpp"
 #include "../Window.hpp"
 #include "../../core/utils/logger.hpp"
@@ -32,6 +33,23 @@ void LocalUniverse::update(voxel_pos pos, float) {
         onMessage(handle->message.value());
         handle->message = std::nullopt;
     }
+    if (handle->entityChange) {
+        handle->entities->iter([&](entity_id eId, const Entity &entity) {
+            if (entity.instances.size() == 0)
+                return;
+
+            if (auto area = std::get_if<Entity::area_t>(&entity.shape)) {
+                contouring->onEntityLoad(eId.index, glm::lvec3(1) + glm::divide(entity.size), *area);
+            } else {
+                auto model = std::get_if<Entity::model_t>(&entity.shape);
+                assert(model);
+                data::vec_istream idata(*model);
+                std::istream iss(&idata);
+                contouring->onEntityLoad(eId.index, iss);
+            }
+        });
+        handle->entityChange = false;
+    }
 
     const auto cur_chunk = glm::divide(pos);
     const auto chunkChange = cur_chunk != last_chunk;

src/core/data/generational.hpp

@@ -133,6 +133,12 @@ namespace data::generational {
             assert(contains(idx));
             return entries[idx.index].value.value();
         }
+        T* directly_at(size_t index) {
+            if (index >= entries.size() || !entries.at(index).value.has_value())
+                return nullptr;
+
+            return &entries[index].value.value();
+        }
 
         bool contains(id idx) const {
             return idx.index < entries.size() &&
@@ -279,6 +285,19 @@ namespace data::generational {
             return false;
         }
 
+        void erase(size_t i) {
+            if(i < this->size())
+                this->at(i) = std::nullopt;
+        }
+        template<typename extractor>
+        void remove(extractor fn) {
+            for (size_t i = 0; i < this->size(); i++) {
+                if(this->at(i).has_value() && fn(i, this->at(i).value())) {
+                    erase(i);
+                }
+            }
+        }
+
         size_t count() const {
             return std::count_if(this->begin(), this->end(),
                 [](const std::optional<T> &e) { return e.has_value(); });

src/core/data/math.hpp

@@ -60,9 +60,15 @@ namespace glm {
     constexpr idx inline toIdx(glm::u8 x, glm::u8 y, glm::u8 z) {
         return (x * IDX_LENGTH + y) * IDX_LENGTH + z;
     }
+    constexpr idx inline toIdx(glm::u8 x, glm::u8 y, glm::u8 z, glm::u8 sy, glm::u8 sz) {
+        return x * sy * sz + y * sz + z;
+    }
     constexpr idx inline toIdx(ucvec3 pos) {
         return toIdx(pos.x, pos.y, pos.z);
     }
+    constexpr idx inline toIdx(ucvec3 pos, ucvec3 size) {
+        return toIdx(pos.x, pos.y, pos.z, size.y, size.z);
+    }
     constexpr std::pair<lvec3, idx> inline splitIdx(const llvec3 &value, const ucvec3 &m = ucvec3(IDX_LENGTH)) {
         return {divide(value, m), toIdx(rem(value.x, m.x), rem(value.y, m.y), rem(value.z, m.z))};
     }

src/core/data/mem.hpp

@@ -9,8 +9,8 @@ namespace data {
 
 /// Map vector to istream
 struct vec_istream: std::streambuf {
-    vec_istream(std::vector<char> &vec) {
+    vec_istream(const std::vector<char> &vec) {
-        this->setg(&vec[0], &vec[0], &vec[0] + vec.size());
+        this->setg((char*)&vec[0], (char*)&vec[0], (char*)&vec[0] + vec.size());
     }
 };
 

src/core/net/data.hpp

@@ -53,19 +53,22 @@ enum class server_packet_type: uint8_t {
     EDITS = 20,
 
     /// Declare entities types
-    /// {size_t(index), vec3(size), vec3(scale)}
+    /// {size_t(index), usvec3(size), vec3(scale), flags(000000, area, permanant)}[]
-    /// TODO: zstd<chunk rle>
     ENTITY_TYPES = 32,
 
     /// Update entities instances position and velocity
     /// {size_t(entity), size_t(count), {size_t(index), ifvec3(pos), vec3(velocity)}[]}[]
     ENTITIES = 33,
 
+    /// Entity type model data
+    /// {size_t(entity), bool(area), render::Data::Model || {size_t(size), zstd<chunk rle>}[]}
+    ENTITY_SHAPE = 34,
+
     /// World compression dictionary
     /// zstd dict
     COMPRESSION = 64,
     /// Server capabilities
-    /// ushort(loadDistance), bool(predictable)
+    /// ushort(loadDistance), bool(predictable), (if predictable) size_t(floodfillLimit)
     //MAYBE: use uint8_t flags
     CAPABILITIES = 65,
 
@@ -79,6 +82,10 @@ enum class client_packet_type: uint8_t {
     /// actions::FillShape
     FILL_SHAPE = 0,
 
+    /// Request entity chunks
+    /// size_t(id)
+    MISSING_ENTITY = 6,
+
     /// Request missing regions
     /// area_id, region_pos[]
     MISSING_REGIONS = 7,

src/core/net/io.hpp

@@ -10,7 +10,7 @@ namespace net {
 /// Helper to write allocated packets
 class PacketWriter final {
 public:
-    PacketWriter(size_t size): buffer((uint8_t*)malloc(size), size) { }
+    PacketWriter(size_t size): buffer((uint8_t*)malloc(size), size), visible_size(size) { }
     PacketWriter(uint8_t type, size_t data_size): PacketWriter(sizeof(type) + data_size) {
         write(type);
     }
@@ -29,21 +29,56 @@ public:
         write(&d, sizeof(d));
     }
 
+    constexpr size_t getCursor() const { return buffer.cur; }
+    void writeAt(size_t pos, const void* data, size_t len) {
+        assert(pos + len <= buffer.siz);
+        memcpy(buffer.ptr + pos, data, len);
+    }
+    template<typename D>
+    void writeAt(size_t pos, const D& d) {
+        writeAt(pos, &d, sizeof(d));
+    }
+
+    void reserve(size_t target) {
+        if (target >= buffer.siz - buffer.cur) {
+            const auto size = target + buffer.cur;
+            buffer.ptr = (uint8_t *)realloc(buffer.ptr, size);
+            buffer.siz = size;
+        }
+    }
+    void resize(size_t target) {
+        reserve(target);
+        visible_size = target;
+    }
+
+    void writePush(const void* data, size_t len) {
+        if (buffer.cur + len > buffer.siz)
+            resize(buffer.cur + len);
+        write(data, len);
+    }
+    template<typename D>
+    void writePush(const D& d) {
+        writePush(&d, sizeof(d));
+    }
+
     struct varying_part {
-        varying_part(data::in_view &buffer): buffer(buffer), visible_size(0) { }
+        varying_part(data::in_view &buffer, size_t& p_size): buffer(buffer), parent_size(p_size), visible_size(0) { }
         ~varying_part() {
-            buffer.siz = buffer.cur + visible_size;
+            buffer.cur += visible_size;
-            buffer.cur = buffer.siz;
+            parent_size += visible_size;
         }
 
         data::in_view &buffer;
+        size_t &parent_size;
         size_t visible_size;
 
         constexpr size_t size() const { return visible_size; }
         void* data() { return buffer.writeTo(0); }
         void reserve(size_t target) {
             if (target >= buffer.siz - buffer.cur) {
-                buffer.ptr = (uint8_t*)realloc(buffer.ptr, target + buffer.cur);
+                const auto size = target + buffer.cur;
+                buffer.ptr = (uint8_t *)realloc(buffer.ptr, size);
+                buffer.siz = size;
             }
         }
         void resize(size_t target) {
@@ -53,17 +88,17 @@ public:
     };
     /// Only from resize, write, resize down
     varying_part varying() {
-        return varying_part(buffer);
+        return varying_part(buffer, visible_size);
     }
 
     bool isFull() const {
-        return buffer.isDone();
+        return buffer.cur >= visible_size;
     }
 
     [[nodiscard]]
     data::out_buffer finish() {
         assert(isFull());
-        data::out_buffer cpy(buffer.ptr, buffer.siz);
+        data::out_buffer cpy(buffer.ptr, visible_size);
         buffer.ptr = nullptr;
         buffer.siz = 0;
         return cpy;
@@ -94,6 +129,7 @@ public:
 
 private:
     data::in_view buffer;
+    size_t visible_size;
 };
 
 /// Helper to read out_view
@@ -121,9 +157,11 @@ public:
         return !isFull();
     }
 
+    data::out_view readPart(size_t size) {
+        return data::out_view(buffer.readFrom(size), size);
+    }
     data::out_view readAll() {
-        const size_t remain = buffer.size() - buffer.cur;
+        return readPart(buffer.size() - buffer.cur);
-        return data::out_view(buffer.readFrom(remain), remain);
     }
 
     bool isFull() const {

src/core/server_handle.hpp

@@ -14,4 +14,5 @@ struct server_handle {
     std::function<world::Universe::ray_result(const geometry::Ray &ray)> raycast;
     std::optional<voxel_pos> teleport;
     std::optional<std::string> message;
+    bool entityChange = false;
 };

src/core/world/Chunk.hpp

@@ -13,11 +13,13 @@ namespace world {
         Chunk(std::istream& str, bool rle = RLE);
         virtual ~Chunk();
 
-        struct Edit {
+        struct EditBody {
-            chunk_voxel_idx idx;
             Voxel value;
             float delay;
         };
+        struct Edit: EditBody {
+            chunk_voxel_idx idx;
+        };
 
         /// Get voxel from index
         inline const Voxel& get(chunk_voxel_idx idx) const {

src/core/world/EdittableChunk.cpp

@@ -12,12 +12,12 @@ EdittableChunk::~EdittableChunk() { }
 std::optional<Faces> EdittableChunk::update(float deltaTime, bool animate) {
     ZoneScopedN("Chunk");
     for(auto it = edits.begin(); it != edits.end();) {
-        it->delay -= deltaTime;
+        it->second.delay -= deltaTime;
-        if(it->delay <= 0 && animate) {
+        if(it->second.delay <= 0 && animate) {
-            invalidate(it->idx);
+            invalidate(it->first);
             it = edits.erase(it);
         } else {
-            it++;
+            ++it;
         }
     }
 
@@ -42,8 +42,9 @@ void EdittableChunk::apply(const Edit& edit) {
     const auto prev = voxels[edit.idx];
     if(prev.value != edit.value.value) {
         voxels[edit.idx] = edit.value;
+        edits.erase(edit.idx);
         if(edit.delay > 0) {
-            edits.emplace_back<Edit>({edit.idx, prev, edit.delay});
+            edits.emplace(edit.idx, EditBody{prev, edit.delay});
         } else {
             invalidate(edit.idx);
         }

src/core/world/EdittableChunk.hpp

@@ -22,8 +22,9 @@ namespace world::client {
 
         void apply(const Chunk::Edit &edit);
 
+        using edits_t = robin_hood::unordered_map<chunk_voxel_idx, EditBody>;
         /// Get pending changes
-        const std::vector<Chunk::Edit> &getEdits() const { return edits; }
+        const edits_t &getEdits() const { return edits; }
 
         static std::optional<chunk_voxel_idx> getNeighborIdx(chunk_voxel_idx idx, Face dir);
 
@@ -31,8 +32,7 @@ namespace world::client {
         EdittableChunk();
 
         /// Animated changes
-        /// MAYBE: sort by delay
+        edits_t edits;
-        std::vector<Chunk::Edit> edits;
         /// Require update
         bool upToDate = true;
         /// Neighbors to update

src/core/world/Universe.hpp

@@ -56,9 +56,21 @@ namespace world {
 
         /// Entities commun properties
         struct Entity {
-            Entity(const glm::vec3& size = glm::vec3(1), const glm::vec3& scale = glm::vec3(1)): size(size), scale(scale) { };
+            /// Linear cube of Chunk
-            glm::vec3 size;
+            /// Size: 1+divide(size)
+            using area_t = std::vector<std::shared_ptr<Chunk>>;
+            /// render::Model::Data serialized
+            using model_t = std::vector<char>;
+
+            Entity(const glm::usvec3& size = glm::usvec3(1), const glm::vec3& scale = glm::vec3(1), bool permanant = true):
+                size(size), scale(scale), permanant(permanant) { }
+            Entity(const model_t& model, const glm::usvec3& size = glm::usvec3(1), const glm::vec3& scale = glm::vec3(1), bool permanant = true):
+                size(size), scale(scale), permanant(permanant) { shape = model; }
+
+            std::variant<area_t, model_t> shape;
+            glm::usvec3 size;
             glm::vec3 scale;
+            bool permanant;
             struct Instance {
                 glm::ifvec3 pos;
                 glm::vec3 velocity;

src/core/world/Voxel.hpp

@@ -12,12 +12,12 @@ namespace world {
         uint16_t value;
         using material_t = uint_fast16_t;
         using density_t = uint_fast8_t;
-        constexpr static const density_t DENSITY_MAX = (1 << 3)-1;
+        static constexpr density_t DENSITY_MAX = (1 << 3)-1;
-        constexpr static const material_t MATERIAL_MAX = (1 << 12)-1;
+        static constexpr material_t MATERIAL_MAX = (1 << 12)-1;
 
-        constexpr static const uint16_t DENSITY_MASK = 0b0111;
+        static constexpr uint16_t DENSITY_MASK = 0b0111;
-        constexpr static const uint16_t MATERIAL_MASK = 0b0111'1111'1111'1000;
+        static constexpr uint16_t MATERIAL_MASK = 0b0111'1111'1111'1000;
-        constexpr static const uint16_t SWAP_MASK = 0b1000'0000'0000'0000;
+        static constexpr uint16_t SWAP_MASK = 0b1000'0000'0000'0000;
 
         Voxel(uint16_t value = 0): value(value) { }
         Voxel(material_t material, density_t density, bool swap = false) {
@@ -43,7 +43,7 @@ namespace world {
         }
         /// Quantity of element on [0, 1]
         constexpr inline float density_ratio() const {
-            return density() * 1.f / world::Voxel::DENSITY_MAX;
+            return density() * (1.f / world::Voxel::DENSITY_MAX);
         }
 
         /// Swap value

src/core/world/iterators.cpp

@@ -2,7 +2,7 @@
 
 using namespace world::iterator;
 
-std::unique_ptr<Abstract> world::iterator::Get(world::action::Shape shape, uint16_t radius) {
+std::unique_ptr<AbstractFill> world::iterator::Get(world::action::Shape shape, uint16_t radius) {
     switch (shape) {
     case world::action::Shape::Cube:
         return std::make_unique<Cube>(radius);
@@ -11,7 +11,18 @@ std::unique_ptr<Abstract> world::iterator::Get(world::action::Shape shape, uint1
     case world::action::Shape::SmoothSphere:
         return std::make_unique<SmoothSphere>(radius);
     default:
-        return std::unique_ptr<Abstract>(nullptr);
+        return std::unique_ptr<AbstractFill>(nullptr);
+    }
+}
+std::unique_ptr<AbstractBorder> world::iterator::GetBorder(world::action::Shape shape, uint16_t radius) {
+    switch (shape) {
+    case world::action::Shape::Cube:
+        return std::make_unique<CubeBorder>(radius);
+    case world::action::Shape::RawSphere:
+    case world::action::Shape::SmoothSphere:
+        return std::make_unique<SphereBorder>(radius);
+    default:
+        return std::unique_ptr<AbstractBorder>(nullptr);
     }
 }
 
@@ -36,6 +47,33 @@ bool Cube::next(pair& out) {
     }
     return true;
 }
+bool CubeBorder::next(glm::ivec3& out) {
+    if (face > 5)
+        return false;
+
+    out = pos;
+
+    const auto getAxis = [&](glm::ivec3 &in, int offset) -> int& {
+        const auto v = (face / 2 + offset) % 3;
+        return v == 0 ? in.x : (v == 1 ? in.y : in.z);
+    };
+    const auto r0_max = radius - (face > 1);
+    if (auto& r0 = getAxis(pos, 0); r0 < r0_max) {
+        r0++;
+    } else {
+        r0 = -r0_max;
+        const auto r1_max = radius - (face > 3);
+        if (auto &r1 = getAxis(pos, 1); r1 < r1_max) {
+            r1++;
+        } else {
+            r1 = -r1_max;
+            getAxis(pos, 2) = -radius;
+            face++;
+            getAxis(pos, 2) = (face % 2 == 0) ? -radius : radius;
+        }
+    }
+    return true;
+}
 
 bool RawSphere::next(pair& out) {
     const auto delta = [](int radius, int pos) -> int {
@@ -94,4 +132,24 @@ bool SmoothSphere::next(pair& out) {
 
     out.first = pos;
     return true;
+}
+
+constexpr float PI = 3.14159;
+bool SphereBorder::next(glm::ivec3& out) {
+    if (i >= max_i)
+        return false;
+
+    const float si = sin(PI * i / (float)max_i);
+    const float sj = sin(2 * PI * j / (float)max_j);
+    const float ci = cos(PI * i / (float)max_i);
+    const float cj = cos(2 * PI * j / (float)max_j);
+    out = glm::vec3(si * cj, si * sj, ci) * (radius + .5f);
+
+    if (j < max_j - 2) {
+        j++;
+    } else {
+        j = 0;
+        i++;
+    }
+    return true;
 }

src/core/world/iterators.hpp

@@ -2,23 +2,141 @@
 
 #include "position.h"
 #include "actions.hpp"
+#include <queue>
 
 namespace world::iterator {
 
 using pair = std::pair<glm::ivec3, float>;
 
-class Abstract {
+class AbstractFill {
 public:
     virtual bool next(pair&) = 0;
 
 protected:
     static constexpr uint32_t Diam(uint16_t radius) { return radius * 2 + 1; }
 };
+class AbstractBorder {
+public:
+    virtual bool next(glm::ivec3&) = 0;
+};
 
 /// From -radius to radius
-std::unique_ptr<Abstract> Get(action::Shape, uint16_t radius);
+std::unique_ptr<AbstractFill> Get(action::Shape, uint16_t radius);
+/// Border sampling from -radius to radius
+std::unique_ptr<AbstractBorder> GetBorder(action::Shape, uint16_t radius);
 
-class Cube final: public Abstract {
+template<typename Chunk, typename Area>
+inline bool GetChunk(Area& chunks, const std::pair<chunk_pos, chunk_voxel_idx>& split, Voxel& out,
+    typename std::shared_ptr<Chunk>& ck = nullptr, chunk_pos& ck_pos = chunk_pos(INT32_MAX)
+) {
+    if (split.first != ck_pos && chunks.inRange(split.first)) {
+        if(auto it = chunks.find(split.first); it != chunks.end()) {
+            ck = std::dynamic_pointer_cast<Chunk>(it->second);
+            ck_pos = split.first;
+        }
+    }
+    if (split.first == ck_pos) {
+        out = ck->get(split.second);
+        return true;
+    }
+    return false;
+}
+
+/// Apply shape on areas
+template<typename Chunk, typename area_map, typename CB>
+inline void Apply(area_map &areas, action::FillShape fill, const CB& callback) {
+    const auto it = areas.find(fill.pos.first);
+    if (it == areas.end())
+        return;
+
+    auto &chunks = it->second->setChunks();
+    auto iterator = Get(fill.shape, fill.radius);
+    pair point;
+    typename std::shared_ptr<Chunk> ck = nullptr;
+    chunk_pos ck_pos = chunk_pos(INT32_MAX);
+    while (iterator->next(point)) {
+        const voxel_pos offset = point.first;
+        const auto split = glm::splitIdx(fill.pos.second + offset);
+        if (Voxel prev; GetChunk<Chunk>(chunks, split, prev, ck, ck_pos)) {
+            const auto next = prev.filled(fill.val, point.second);
+            if (prev.value != next.value) {
+                callback(ck, ck_pos, split.second, prev, next, glm::length2(offset) / fill.radius * .05f);
+            }
+        }
+    }
+}
+/// Find parts using flood fill
+template<typename Chunk, typename area_map, typename CB>
+inline bool Split(area_map &areas, action::FillShape fill, size_t floodFillLimit, const CB& callback) {
+    if (floodFillLimit == 0 || fill.val.is_solid())
+        return false;
+
+    const auto it = areas.find(fill.pos.first);
+    if (it == areas.end())
+        return false;
+
+    // Check for subpart break
+    robin_hood::unordered_set<voxel_pos> joints;
+    auto &chunks = it->second->setChunks();
+    auto iterator = GetBorder(fill.shape, fill.radius+1);
+    glm::ivec3 point;
+    typename std::shared_ptr<Chunk> ck = nullptr;
+    chunk_pos ck_pos = chunk_pos(INT32_MAX);
+    const auto cleanVoxel = Voxel(world::materials::AIR, fill.val.density());
+    while (iterator->next(point)) {
+        const auto full = fill.pos.second + voxel_pos(point);
+        if (Voxel v; GetChunk<Chunk>(chunks, glm::splitIdx(full), v, ck, ck_pos) && v.is_solid())
+            joints.insert(full);
+    }
+    bool added = false;
+    while (!joints.empty()) {
+        std::queue<voxel_pos> todo;
+        todo.push(*joints.begin());
+        robin_hood::unordered_set<voxel_pos> part;
+        part.insert(todo.front());
+        size_t i = 0;
+        for (; !todo.empty() && i < floodFillLimit; i++) {
+            const auto full = todo.front();
+            todo.pop();
+            joints.erase(full);
+            if (Voxel v; GetChunk<Chunk>(chunks, glm::splitIdx(full), v, ck, ck_pos) && v.is_solid()) {
+                constexpr std::array<voxel_pos, 6> DIRS = {
+                    voxel_pos(1, 0, 0), voxel_pos(-1, 0, 0), voxel_pos(0, 1, 0),
+                    voxel_pos(0, -1, 0), voxel_pos(0, 0, 1), voxel_pos(0, 0, -1)};
+                    //MAYBE: diag
+                for (auto dir : DIRS) {
+                    const auto off = full + dir;
+                    if (part.insert(off).second)
+                        todo.push(off);
+                }
+            }
+        }
+        if (todo.empty()) {
+            added = true;
+            callback(part, cleanVoxel, chunks, ck, ck_pos, it->second);
+        }
+    }
+    return added;
+}
+
+/// Call Apply then Split
+template<typename Chunk, typename area_map, typename CB>
+inline void ApplySplit(area_map &areas, action::FillShape fill, size_t floodFillLimit, const CB& callback) {
+    Apply<Chunk>(areas, fill, callback);
+    Split<Chunk>(areas, fill, floodFillLimit, [&](const robin_hood::unordered_set<voxel_pos>& part, Voxel next,
+        world::ChunkContainer& chunks, std::shared_ptr<Chunk>& ck, chunk_pos& ck_pos, std::shared_ptr<Area>&) {
+        for(auto full: part) {
+            const auto split = glm::splitIdx(full);
+            if (Voxel prev; world::iterator::GetChunk<Chunk>(chunks, split, prev, ck, ck_pos) && prev.is_solid()) {
+                if (prev.value != next.value) {
+                    callback(ck, ck_pos, split.second, prev, next, fill.radius * .05f);
+                }
+            }
+        }
+    });
+}
+
+class Cube final: public AbstractFill {
 public:
     bool next(pair&) override;
 
@@ -28,9 +146,20 @@ private:
     const uint16_t radius;
     glm::ivec3 pos;
 };
+class CubeBorder final: public AbstractBorder {
+public:
+    bool next(glm::ivec3&) override;
+
+    CubeBorder(uint16_t radius): radius(radius), pos(-radius, -radius, -radius), face(0) { }
+
+private:
+    const uint16_t radius;
+    glm::ivec3 pos;
+    int face;
+};
 
 /// Interger sphere
-class RawSphere final: public Abstract {
+class RawSphere final: public AbstractFill {
 public:
     bool next(pair&) override;
 
@@ -42,9 +171,20 @@ private:
     int dy;
     int dz;
 };
+class SphereBorder final: public AbstractBorder {
+public:
+    bool next(glm::ivec3&) override;
+
+    SphereBorder(uint16_t radius): radius(radius), max_i(radius*3), max_j(radius*4), i(0), j(0) { }
+
+private:
+    const uint16_t radius;
+    const uint max_i, max_j;
+    uint i, j;
+};
 
 /// Anti-aliased sphere
-class SmoothSphere final: public Abstract {
+class SmoothSphere final: public AbstractFill {
 public:
     bool next(pair&) override;
 

src/core/world/models.hpp

@@ -0,0 +1,8 @@
+#pragma once
+
+#include <vector>
+#include <stdint.h>
+
+namespace world::models {
+    static const std::vector<char> player = {24,0,0,0,0,0,0,0,0,0,1,0,2,0,0,0,5,0,1,0,0,0,2,0,4,0,0,0,4,0,5,0,1,0,3,0,2,0,5,0,3,0,1,0,2,0,3,0,4,0,4,0,3,0,5,0,6,0,0,0,0,0,0,0,0,60,0,0,0,0,8,0,0,60,0,0,0,0,0,0,0,0,0,60,0,0,8,0,0,0,0,60,0,0,0,0,0,0,0,0,0,60,8,0,0,0,0,0,0,60,0,0,0,-68,0,0,0,0,8,0,0,-68,0,0,0,0,0,0,0,0,0,-68,0,0,8,0,0,0,0,-68,0,0,0,0,0,0,0,0,0,-68,8,0,0,0,0,0,0,-68,0,0};
+}

src/server/config.hpp

@@ -25,6 +25,7 @@ public:
         world.loadDistance = config["world"]["load_distance"].value_or(world.loadDistance);
         world.keepDistance = config["world"]["keep_distance"].value_or(world.keepDistance);
         world.folderPath = config["world"]["path"].value_or(world.folderPath);
+        world.floodFillLimit = config["world"]["max_part_size"].value_or<int64_t>(world.floodFillLimit);
     }
 
     toml::table save() {
@@ -43,7 +44,8 @@ public:
             {"world", toml::table({
                 {"load_distance", world.loadDistance},
                 {"keep_distance", world.keepDistance},
-                {"path", world.folderPath}
+                {"path", world.folderPath},
+                {"max_part_size", (int64_t)world.floodFillLimit}
             })}
         });
     }

src/server/net/Server.hpp

@@ -58,6 +58,14 @@ public:
             call(&peer);
         }
     }
+    template<typename P>
+    bool anyPeer(P predicate) {
+        for(auto& peer: peers) {
+            if(predicate(&peer))
+                return true;
+        }
+        return false;
+    }
 
 private:
     std::forward_list<Peer> peers;

src/server/world/Chunk.hpp

@@ -9,6 +9,7 @@ namespace world::server {
     /// Server size chunk
     class Chunk: public virtual world::Chunk {
     public:
+        Chunk(): world::Chunk() { }
         Chunk(const chunk_pos &pos, const std::unique_ptr<generator::Abstract> &rnd);
         Chunk(std::istream& str, bool rle = RLE);
         virtual ~Chunk();
@@ -27,9 +28,6 @@ namespace world::server {
         /// Voxel swap bit indicate perfect majority
         Voxel write(std::ostream& str, bool rle = RLE) const;
 
-    protected:
-        Chunk(): world::Chunk() { }
-
     private:
         /// Modified by player
         bool modified = false;

src/server/world/SharedParts.hpp

@@ -9,6 +9,7 @@ namespace world::server {
 // Server and Client merged chunk
 class SharedChunk final: public Chunk, public world::client::EdittableChunk {
 public:
+    SharedChunk(): world::Chunk(), Chunk(), world::client::EdittableChunk() { }
     SharedChunk(const chunk_pos &pos, const std::unique_ptr<generator::Abstract> &rnd): world::Chunk(), Chunk(pos, rnd), world::client::EdittableChunk() { }
     SharedChunk(std::istream &str, bool rle = RLE): world::Chunk(str, rle), Chunk(), world::client::EdittableChunk() { }
 
@@ -16,8 +17,9 @@ public:
     std::optional<Item> replace(chunk_voxel_idx idx, const Voxel &val, float delay = 0) override {
         const auto res = voxels[idx];
         set(idx, val);
+        edits.erase(idx);
         if(delay > 0) {
-            edits.emplace_back<Edit>({idx, res, delay});
+            edits.emplace(idx, EditBody{res, delay});
         } else {
             invalidate(idx);
         }

src/server/world/SharedUniverse.cpp

@@ -30,6 +30,7 @@ SharedUniverse::SharedUniverse(const options &o, server_handle *const localHandl
     localHandle->raycast = std::function([&](const geometry::Ray& ray){
         return this->raycast(ray); });
     localHandle->running = true;
+    localHandle->entityChange = true;
 }
 SharedUniverse::~SharedUniverse() {
     LOG_D("Breaking shared universe");
@@ -50,6 +51,10 @@ void SharedUniverse::broadcastMessage(const std::string& text) {
     localHandle->message = text;
     Universe::broadcastMessage(text);
 }
+void SharedUniverse::broadcastEntities() {
+    localHandle->entityChange = true;
+    Universe::broadcastEntities();
+}
 
 std::shared_ptr<Chunk> SharedUniverse::createChunk(const chunk_pos &pos, const std::unique_ptr<generator::Abstract> &rnd) const {
     return std::make_shared<SharedChunk>(pos, rnd);
@@ -57,3 +62,6 @@ std::shared_ptr<Chunk> SharedUniverse::createChunk(const chunk_pos &pos, const s
 std::shared_ptr<Chunk> SharedUniverse::createChunk(std::istream &str) const {
     return std::make_shared<SharedChunk>(str);
 }
+std::shared_ptr<Chunk> SharedUniverse::createChunk() const {
+    return std::make_shared<SharedChunk>();
+}

src/server/world/SharedUniverse.hpp

@@ -15,11 +15,13 @@ namespace world::server {
     protected:
         std::shared_ptr<Chunk> createChunk(const chunk_pos &pos, const std::unique_ptr<generator::Abstract> &rnd) const override;
         std::shared_ptr<Chunk> createChunk(std::istream &str) const override;
+        std::shared_ptr<Chunk> createChunk() const override;
 
         void loadChunk(area_<chunk_pos>, chunk_pos, const world::ChunkContainer &) override;
         void updateChunk(area_map::iterator &, world::ChunkContainer::iterator &, chunk_pos, float deltaTime) override;
 
         void broadcastMessage(const std::string&) override;
+        void broadcastEntities() override;
 
     private:
         server_handle *const localHandle;

src/server/world/Universe.cpp

@@ -9,6 +9,7 @@
 #include "../../core/world/raycast.hpp"
 #include "../../core/world/iterators.hpp"
 #include "../../core/world/actions.hpp"
+#include "../../core/world/models.hpp"
 #include "../../core/net/io.hpp"
 
 using namespace world::server;
@@ -66,7 +67,7 @@ Universe::Universe(const Universe::options &options): host(options.connection,
 
     {
         [[maybe_unused]]
-        const auto type_id = entities.emplace(glm::vec3(1), glm::vec3(2));
+        const auto type_id = entities.emplace(world::models::player, glm::usvec3(1), glm::vec3(2));
         assert(type_id == PLAYER_ENTITY_ID);
     }
 
@@ -231,7 +232,7 @@ void Universe::pull() {
         if (data == nullptr)
             return;
 
-        if (data->pendingEdits.empty() && peer->queueSize(net::server::queue::EDIT) == 0) {
+        if (!data->pendingEdits.empty() && peer->queueSize(net::server::queue::EDIT) == 0) {
             peer->send(net::PacketWriter::Of(net::server_packet_type::EDITS, data->pendingEdits.front()));
             data->pendingEdits.pop();
         }
@@ -411,7 +412,7 @@ void Universe::update(float deltaTime) {
     { // Update entities
         ZoneScopedN("Entities");
         size_t item_count = 0;
-        entities.for_each([&](entity_id type, Entity &val) {
+        entities.remove([&](entity_id type, Entity &val) {
             val.instances.remove([&](entity_id, Entity::Instance &inst) {
                 if (type == PLAYER_ENTITY_ID) {
                     //MAYBE: update players ?
@@ -428,6 +429,7 @@ void Universe::update(float deltaTime) {
                 //MAYBE: Store in region ?
                 //MAYBE: Save to files
             });
+            return !val.permanant && val.instances.empty();
         });
         TracyPlot("EntityCount", static_cast<int64_t>(item_count));
 
@@ -472,6 +474,9 @@ std::optional<uint16_t> Universe::onConnect(net::server::Peer* peer) {
         auto packet = PacketWriter(server_packet_type::CAPABILITIES, sizeof(loadDistance) + sizeof(bool));
         packet.write(loadDistance);
         packet.write(PREDICTABLE);
+        if constexpr (PREDICTABLE) {
+            packet.write(floodFillLimit);
+        }
         peer->send(packet.finish());
     }
 
@@ -485,17 +490,7 @@ std::optional<uint16_t> Universe::onConnect(net::server::Peer* peer) {
         peer->send(packet.finish());
         movedPlayers.insert(client->instanceId);
     }
-    {
+    broadcastEntities();
-        constexpr auto ITEM_SIZE = sizeof(entity_id::index) + sizeof(glm::vec3) * 2;
-        auto packet = PacketWriter(server_packet_type::ENTITY_TYPES, ITEM_SIZE * entities.size());
-        entities.iter([&](entity_id id, const Entity &entity) {
-            packet.write(id.index);
-            packet.write(entity.size);
-            packet.write(entity.scale);
-        });
-        //MAYBE: sendBroadcast
-        peer->send(packet.finish(), net::server::queue::ENTITY);
-    }
     broadcastMessage("> Player" + std::to_string(client->instanceId.index) + " has joined us");
     broadcastAreas();
     return std::nullopt;
@@ -620,8 +615,9 @@ bool Universe::onPacket(net::server::Peer *peer, const data::out_view &buf, net:
                         if (!packet.read(cpos))
                             break;
                         const auto dist = glm::length2(areaOffset - cpos);
-                        if (dist <= glm::pow2(loadDistance) && chunks.inRange(cpos) && chunks.findInRange(cpos).has_value()) {
+                        if (dist <= glm::pow2(loadDistance) && chunks.inRange(cpos)) {
-                            data->pushChunk(std::make_pair(id, cpos), dist);
+                            if (chunks.findInRange(cpos).has_value())
+                                data->pushChunk(std::make_pair(id, cpos), dist);
                         } else {
                             LOG_T("Request out of range chunk");
                         }
@@ -632,6 +628,45 @@ bool Universe::onPacket(net::server::Peer *peer, const data::out_view &buf, net:
             }
             break;
         }
+        case client_packet_type::MISSING_ENTITY: {
+            size_t id;
+            if (!packet.read(id))
+                break;
+
+            if (auto entity = entities.directly_at(id)) {
+                if (const auto area = std::get_if<Entity::area_t>(&entity->shape)) {
+                    auto packet = PacketWriter(server_packet_type::ENTITY_SHAPE, sizeof(id) + sizeof(bool) + sizeof(size_t));
+                    packet.write(id);
+                    packet.write(true);
+                    for (auto it = area->begin(); it != area->end(); ++it) {
+                        std::ostringstream out;
+                        std::dynamic_pointer_cast<Chunk>(*it)->write(out);
+                        auto size_pos = packet.getCursor();
+                        size_t size = 0;
+                        packet.writePush(size);
+                        {
+                            auto vec = packet.varying();
+                            dict_write_ctx.compress(out.str(), vec);
+                            size = vec.size();
+                        }
+                        packet.writeAt(size_pos, size);
+                    }
+                    peer->send(packet.finish(), net::server::CHUNK);
+                } else {
+                    const auto model = std::get_if<Entity::model_t>(&entity->shape);
+                    assert(model);
+                    auto packet = PacketWriter(server_packet_type::ENTITY_SHAPE, sizeof(id) + sizeof(bool) + model->size());
+                    //MAYBE: prefix model with id+false
+                    packet.write(id);
+                    packet.write(false);
+                    packet.write(model->data(), model->size());
+                    peer->send(packet.finish(), net::server::CHUNK);
+                }
+            } else {
+                LOG_T("Bad entity request " << id);
+            }
+            break;
+        }
         default:
             LOG_T("Bad packet from " << peer->getAddress());
             break;
@@ -664,6 +699,22 @@ data::out_buffer Universe::serializeChunk(area_<chunk_pos> id, const std::shared
 void Universe::broadcastMessage(const std::string& text) {
     host.sendBroadcast(net::PacketWriter::Of(net::server_packet_type::MESSAGE, text.data(), text.size()));
 }
+void Universe::broadcastEntities() {
+    constexpr auto ITEM_SIZE = sizeof(entity_id::index) + sizeof(glm::usvec3) + sizeof(glm::vec3) + sizeof(uint8_t);
+    auto packet = net::PacketWriter(net::server_packet_type::ENTITY_TYPES, ITEM_SIZE * entities.size());
+    entities.iter([&](entity_id id, const Entity &entity) {
+        packet.write(id.index);
+        packet.write(entity.size);
+        packet.write(entity.scale);
+        uint8_t flags = 0;
+        if (entity.permanant)
+            flags |= 1;
+        if (std::holds_alternative<Entity::area_t>(entity.shape))
+            flags |= 2;
+        packet.write(flags);
+    });
+    host.sendBroadcast(packet.finish(), net::server::queue::ENTITY);
+}
 
 void Universe::updateChunk(area_map::iterator &, world::ChunkContainer::iterator &, chunk_pos, float /*deltaTime*/) {}
 void Universe::loadChunk(area_<chunk_pos>, chunk_pos, const world::ChunkContainer &) {}
@@ -671,6 +722,7 @@ void Universe::loadChunk(area_<chunk_pos>, chunk_pos, const world::ChunkContaine
 void Universe::setOptions(const Universe::options& options) {
     loadDistance = options.loadDistance;
     keepDistance = options.keepDistance;
+    floodFillLimit = options.floodFillLimit;
 }
 
 Universe::ray_result Universe::raycast(const geometry::Ray &ray) const {
@@ -692,59 +744,83 @@ bool Universe::isAreaFree(const area_<voxel_pos> &pos, const geometry::Shape sha
 world::ItemList Universe::set(const area_<voxel_pos>& pos, int radius, action::Shape shape, const Voxel& val) {
     ZoneScopedN("Fill");
     ItemList list;
-    if(const auto it = areas.find(pos.first); it != areas.end()) {
+    const bool stupidClient = host.anyPeer([&](net::server::Peer *peer) {
-        robin_hood::unordered_map<chunk_pos, std::vector<Chunk::Edit>> edits;
+        auto data = peer->getCtx<net_client>();
-        auto &chunks = it->second->setChunks();
+        return data && !data->handleEdits;
-        auto iterator = world::iterator::Get(shape, radius);
+    });
-        world::iterator::pair point;
+    robin_hood::unordered_map<chunk_pos, std::vector<Chunk::Edit>> edits;
-        while (iterator->next(point)) {
+    const auto fill = world::action::FillShape(pos, val, shape, radius);
-            const voxel_pos offset = point.first;
+    world::iterator::Apply<Chunk>(areas, fill,
-            const auto split = glm::splitIdx(pos.second + offset);
+        [&](std::shared_ptr<Chunk>& ck, chunk_pos ck_pos, chunk_voxel_idx idx, Voxel /*prev*/, Voxel next, float delay) {
-            if(chunks.inRange(split.first)) {
+            if (stupidClient)
-                if(const auto chunk = it->second->setChunks().findInRange(split.first)) {
+                edits[ck_pos].push_back(Chunk::Edit{next, delay, idx});
-                    auto ck = std::dynamic_pointer_cast<Chunk>(chunk.value());
+            //TODO: apply break table
-                    auto prev = ck->get(split.second);
+            //TODO: inventory
-                    const auto next = prev.filled(val, point.second);
+            ck->replace(idx, next, delay);
-                    if(prev.value != next.value) {
+    });
-                        //TODO: apply break table
+    if (world::iterator::Split<Chunk>(areas, fill, floodFillLimit, [&](const robin_hood::unordered_set<voxel_pos>& part, Voxel cleanVoxel,
-                        //TODO: inventory
+        world::ChunkContainer& chunks, std::shared_ptr<Chunk>& ck, chunk_pos& ck_pos, std::shared_ptr<Area>& it
-                        const auto delay = glm::length2(offset) / radius * .05f;
+    ){
-                        edits[split.first].push_back(Chunk::Edit{split.second, next, delay});
+        voxel_pos min = voxel_pos(INT64_MAX);
-                        ck->replace(split.second, next, delay);
+        voxel_pos max = voxel_pos(INT64_MIN);
-                    }
+        for(auto full: part) {
+            min = glm::min(min, full);
+            max = glm::max(max, full);
+        }
+        const auto size = max - min;
+        const chunk_pos scale = chunk_pos(1) + glm::divide(size);
+        const auto type = entities.emplace(size, glm::vec3(1), false);
+        auto &entity = entities.at(type);
+        auto &area = std::get<Entity::area_t>(entity.shape);
+        {
+            const auto chunk_count = scale.x * scale.y * scale.z;
+            area.reserve(chunk_count);
+            for (long i = 0; i < chunk_count; i++)
+                area.push_back(createChunk());
+        }
+        std::shared_ptr<Chunk> ck_dest = nullptr;
+        chunk_pos ck_pos_dest = chunk_pos(INT32_MAX);
+        for(auto full: part) {
+            const auto split = glm::splitIdx(full);
+            if (Voxel v; world::iterator::GetChunk<Chunk>(chunks, split, v, ck, ck_pos) && v.is_solid()) {
+                if (stupidClient)
+                    edits[ck_pos].push_back(Chunk::Edit{cleanVoxel, fill.radius * .05f, split.second});
+
+                ck->replace(split.second, cleanVoxel);
+                const auto spl = glm::splitIdx(full - min);
+                if (spl.first != ck_pos_dest) {
+                    ck_dest = std::dynamic_pointer_cast<Chunk>(area.at(glm::toIdx(spl.first, scale)));
+                    assert(ck_dest);
+                    ck_pos_dest = spl.first;
                 }
+                ck_dest->set(spl.second, v);
             }
         }
+        entity.instances.emplace(Entity::Instance{it->getOffset() + min, glm::vec3(0)});
+    })) { broadcastEntities(); }
 
-        bool stupidClient = false;
+    if (stupidClient && !edits.empty()) {
+        ZoneScopedN("Packet");
+        size_t size = sizeof(area_id);
+        for(const auto& part: edits) {
+            size += sizeof(chunk_pos);
+            size += sizeof(chunk_voxel_idx);
+            size += sizeof(Chunk::Edit) * part.second.size();
+        }
+        auto packet = net::PacketWriter(net::server_packet_type::EDITS, size);
+        packet.write(pos.first);
+        for(const auto& part: edits) {
+            packet.write(part.first);
+            packet.write<chunk_voxel_idx>(part.second.size());
+            packet.write(part.second.data(), part.second.size() * sizeof(Chunk::Edit));
+        }
+        auto buffer = packet.finish();
         host.iterPeers([&](net::server::Peer *peer) {
+            //MAYBE: only in range
             auto data = peer->getCtx<net_client>();
             if (data && !data->handleEdits)
-                stupidClient = true;
+                peer->send(buffer, net::server::queue::CHUNK);
         });
-        if (stupidClient) {
-            ZoneScopedN("Packet");
-            size_t size = sizeof(area_id);
-            for(const auto& part: edits) {
-                size += sizeof(chunk_pos);
-                size += sizeof(chunk_voxel_idx);
-                size += sizeof(Chunk::Edit) * part.second.size();
-            }
-            auto packet = net::PacketWriter(net::server_packet_type::EDITS, size);
-            packet.write(pos.first);
-            for(const auto& part: edits) {
-                packet.write(part.first);
-                packet.write<chunk_voxel_idx>(part.second.size());
-                packet.write(part.second.data(), part.second.size() * sizeof(Chunk::Edit));
-            }
-            auto buffer = packet.finish();
-            host.iterPeers([&](net::server::Peer *peer) {
-                //MAYBE: only in range
-                auto data = peer->getCtx<net_client>();
-                if (data && !data->handleEdits)
-                    peer->send(buffer, net::server::queue::CHUNK);
-            });
-        }
     }
     return list;
 }
@@ -806,4 +882,7 @@ std::shared_ptr<Chunk> Universe::createChunk(const chunk_pos &pos, const std::un
 }
 std::shared_ptr<Chunk> Universe::createChunk(std::istream &str) const {
     return std::make_shared<Chunk>(str);
+}
+std::shared_ptr<Chunk> Universe::createChunk() const {
+    return std::make_shared<Chunk>();
 }

src/server/world/Universe.hpp

@@ -30,6 +30,9 @@ namespace world::server {
             std::string folderPath = "world";
 
             net::exposure connection = net::exposure{"", 4242, 1, "content/key.pem", "content/cert.pem"};
+
+            /// Iteration count in part break search
+            size_t floodFillLimit = CHUNK_SIZE;
         };
 
         Universe(const options &);
@@ -76,11 +79,13 @@ namespace world::server {
         void broadcastAreas();
         data::out_buffer serializeChunk(area_<chunk_pos>, const std::shared_ptr<Chunk>&);
         virtual void broadcastMessage(const std::string &);
+        virtual void broadcastEntities();
 
         using area_map = robin_hood::unordered_map<area_id, std::shared_ptr<Area>>;
 
         virtual std::shared_ptr<Chunk> createChunk(const chunk_pos &pos, const std::unique_ptr<generator::Abstract> &rnd) const;
         virtual std::shared_ptr<Chunk> createChunk(std::istream &str) const;
+        virtual std::shared_ptr<Chunk> createChunk() const;
 
         virtual void updateChunk(area_map::iterator&, world::ChunkContainer::iterator&, chunk_pos, float deltaTime);
         virtual void loadChunk(area_<chunk_pos>, chunk_pos, const world::ChunkContainer &);
@@ -109,6 +114,7 @@ namespace world::server {
         ushort loadDistance;
         ushort keepDistance;
         std::string folderPath;
+        size_t floodFillLimit;
 
         net::server::Server host;
 

src/tools/generate_models.cpp

@@ -0,0 +1,36 @@
+/**
+ * \file generate_models.cpp
+ * \brief Serialize model for core/world/models.hpp
+ */
+
+#define STANDALONE 1
+#include "../client/render/api/Models.hpp"
+#include "../core/world/materials.hpp"
+#include <meshoptimizer.h>
+#include <sstream>
+#include <iostream>
+
+/// Entry point
+int main(int, char *[]){
+    const auto pack = [](const glm::vec3 &pos) {
+    return render::PackedVertexData(meshopt_quantizeHalf(pos.x), meshopt_quantizeHalf(pos.y),
+        meshopt_quantizeHalf(pos.z), world::materials::textures_map[7],
+        meshopt_quantizeHalf(pos.x), meshopt_quantizeHalf(pos.y),
+        meshopt_quantizeHalf(pos.z));
+    };
+    auto model = render::Model::Data({
+        pack({1, 0, 0}), pack({0, 1, 0}), pack({0, 0, 1}),
+        pack({-1, 0, 0}), pack({0, -1, 0}), pack({0, 0, -1})
+    }, {
+        0, 1, 2, 0, 5, 1, 0, 2, 4, 0, 4, 5,
+        1, 3, 2, 5, 3, 1, 2, 3, 4, 4, 3, 5
+    });
+
+    std::ostringstream sstr;
+    model.serialize(sstr);
+    for (auto& el: sstr.str()) {
+        std::cout << (int)el << ',';
+    }
+    std::cout << std::endl;
+    return 0;
+}