From 3da84d1579722e77aacedf95276f78874e14d41f Mon Sep 17 00:00:00 2001
From: Bartosz Taudul <wolf@nereid.pl>
Date: Mon, 31 May 2021 22:38:22 +0200
Subject: [PATCH] Hide rpmalloc init behind thread local boolean.

---
 TracyClient.cpp          |  1 +
 client/TracyAlloc.cpp    | 35 +++++++++++++++++++++++++++++++++++
 client/TracyProfiler.cpp |  2 ++
 common/TracyAlloc.hpp    | 27 ++++-----------------------
 4 files changed, 42 insertions(+), 23 deletions(-)
 create mode 100644 client/TracyAlloc.cpp

diff --git a/TracyClient.cpp b/TracyClient.cpp
index 838c3bd8..67f99cd1 100644
--- a/TracyClient.cpp
+++ b/TracyClient.cpp
@@ -27,6 +27,7 @@
 #include "common/TracySocket.cpp"
 #include "client/tracy_rpmalloc.cpp"
 #include "client/TracyDxt1.cpp"
+#include "client/TracyAlloc.cpp"
 
 #if TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 3 || TRACY_HAS_CALLSTACK == 4 || TRACY_HAS_CALLSTACK == 6
 #  include "libbacktrace/alloc.cpp"
diff --git a/client/TracyAlloc.cpp b/client/TracyAlloc.cpp
new file mode 100644
index 00000000..eb94dd8d
--- /dev/null
+++ b/client/TracyAlloc.cpp
@@ -0,0 +1,35 @@
+#ifdef TRACY_ENABLE
+
+#include <atomic>
+
+#include "../common/TracyAlloc.hpp"
+#include "../common/TracyYield.hpp"
+
+namespace tracy
+{
+
+extern std::atomic<int> RpInitDone;
+extern std::atomic<int> RpInitLock;
+
+TRACY_API void InitRpmallocPlumbing()
+{
+    const auto done = RpInitDone.load( std::memory_order_acquire );
+    if( !done )
+    {
+        int expected = 0;
+        while( !RpInitLock.compare_exchange_weak( expected, 1, std::memory_order_release, std::memory_order_relaxed ) ) { expected = 0; YieldThread(); }
+        const auto done = RpInitDone.load( std::memory_order_acquire );
+        if( !done )
+        {
+            rpmalloc_initialize();
+            RpInitDone.store( 1, std::memory_order_release );
+        }
+        RpInitLock.store( 0, std::memory_order_release );
+    }
+    rpmalloc_thread_initialize();
+    RpThreadInitDone = true;
+}
+
+}
+
+#endif
diff --git a/client/TracyProfiler.cpp b/client/TracyProfiler.cpp
index 128d42d6..6a722cd1 100644
--- a/client/TracyProfiler.cpp
+++ b/client/TracyProfiler.cpp
@@ -977,6 +977,7 @@ TRACY_API void ShutdownProfiler()
 #  else
 std::atomic<int> RpInitDone { 0 };
 std::atomic<int> RpInitLock { 0 };
+thread_local bool RpThreadInitDone = false;
 static std::atomic<int> profilerDataLock { 0 };
 static std::atomic<ProfilerData*> profilerData { nullptr };
 
@@ -1097,6 +1098,7 @@ thread_local ThreadHandleWrapper init_order(104) s_threadHandle { detail::GetThr
 static InitTimeWrapper init_order(101) s_initTime { SetupHwTimer() };
 std::atomic<int> init_order(102) RpInitDone( 0 );
 std::atomic<int> init_order(102) RpInitLock( 0 );
+thread_local bool RpThreadInitDone = false;
 moodycamel::ConcurrentQueue<QueueItem> init_order(103) s_queue( QueuePrealloc );
 std::atomic<uint32_t> init_order(104) s_lockCounter( 0 );
 std::atomic<uint8_t> init_order(104) s_gpuCtxCounter( 0 );
diff --git a/common/TracyAlloc.hpp b/common/TracyAlloc.hpp
index 1981c09a..27861c97 100644
--- a/common/TracyAlloc.hpp
+++ b/common/TracyAlloc.hpp
@@ -4,9 +4,8 @@
 #include <stdlib.h>
 
 #ifdef TRACY_ENABLE
-#  include <atomic>
+#  include "TracyApi.h"
 #  include "TracyForceInline.hpp"
-#  include "TracyYield.hpp"
 #  include "../client/tracy_rpmalloc.hpp"
 #endif
 
@@ -14,30 +13,12 @@ namespace tracy
 {
 
 #ifdef TRACY_ENABLE
-extern std::atomic<int> RpInitDone;
-extern std::atomic<int> RpInitLock;
-
-namespace
-{
-static inline void InitRpmallocPlumbing()
-{
-    int expected = 0;
-    while( !RpInitLock.compare_exchange_weak( expected, 1, std::memory_order_release, std::memory_order_relaxed ) ) { expected = 0; YieldThread(); }
-    const auto done = RpInitDone.load( std::memory_order_acquire );
-    if( !done )
-    {
-        rpmalloc_initialize();
-        RpInitDone.store( 1, std::memory_order_release );
-    }
-    RpInitLock.store( 0, std::memory_order_release );
-}
+extern thread_local bool RpThreadInitDone;
+TRACY_API void InitRpmallocPlumbing();
 
 static tracy_force_inline void InitRpmalloc()
 {
-    const auto done = RpInitDone.load( std::memory_order_acquire );
-    if( !done ) InitRpmallocPlumbing();
-    rpmalloc_thread_initialize();
-}
+    if( !RpThreadInitDone ) InitRpmallocPlumbing();
 }
 #endif