Add ___tracy_init_thread function

This function needs to be called for each non-main thread before calling
the `___tracy_alloc_` functions.

Alternative way to achieve this could be initializing the allocator
transparently in the `___tracy_alloc_*` calls.

TracyC.h

@@ -81,6 +81,7 @@ 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 void ___tracy_init_thread(void);
 TRACY_API uint64_t ___tracy_alloc_srcloc( uint32_t line, const char* source, const char* function );
 TRACY_API uint64_t ___tracy_alloc_srcloc_name( uint32_t line, const char* source, const char* function, const char* name, size_t nameSz );
 

client/TracyProfiler.cpp

@@ -3068,6 +3068,13 @@ TRACY_API void ___tracy_emit_message_appinfo( const char* txt, size_t size ) { t
 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 ); }
 
+// 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) { (void)tracy::s_rpmalloc_thread_init; }
+
 #ifdef __cplusplus
 }
 #endif

manual/tracy.tex

@@ -1477,6 +1477,9 @@ couleur=black!5,
 logo=\bcbombe
 ]{Important}
 Since you are directly calling the profiler functions here, you will need to take care of manually disabling the code, if the \texttt{TRACY\_ENABLE} macro is not defined.
+
+Additionally, make sure you have called \texttt{\_\_\_tracy\_init\_thread} before calling the
+\textttt{alloc} functions for the first time on a new non-main thread.
 \end{bclogo}
 
 \subsubsection{Memory profiling}