Store unique callstack payloads.

server/TracyWorker.cpp

@@ -982,7 +982,7 @@ void Worker::DispatchProcess( const QueueItem& ev, char*& ptr )
             AddSourceLocationPayload( ev.stringTransfer.ptr, ptr, sz );
             break;
         case QueueType::CallstackPayload:
-            //AddCallstackPayload( ev.stringTransfer.ptr, ptr, sz );
+            AddCallstackPayload( ev.stringTransfer.ptr, ptr, sz );
             break;
         default:
             assert( false );
@@ -1265,6 +1265,36 @@ void Worker::AddCustomString( uint64_t ptr, char* str, size_t sz )
     m_pendingCustomStrings.emplace( ptr, StoreString( str, sz ) );
 }
 
+void Worker::AddCallstackPayload( uint64_t ptr, char* _data, size_t sz )
+{
+    assert( m_pendingCallstacks.find( ptr ) == m_pendingCallstacks.end() );
+
+    const auto memsize = sizeof( VarArray<uint64_t> ) + sz;
+    auto mem = (char*)m_slab.AllocRaw( memsize );
+
+    auto data = (uint64_t*)mem;
+    memcpy( data, _data, sz );
+
+    auto arr = (VarArray<uint64_t>*)( mem + sz );
+    new(arr) VarArray<uint64_t>( sz / sizeof( uint64_t ), data );
+
+    uint32_t idx;
+    auto it = m_data.callstackMap.find( arr );
+    if( it == m_data.callstackMap.end() )
+    {
+        idx = m_data.callstackPayload.size();
+        m_data.callstackMap.emplace( arr, idx );
+        m_data.callstackPayload.push_back( arr );
+    }
+    else
+    {
+        idx = it->second;
+        m_slab.Unalloc( memsize );
+    }
+
+    m_pendingCallstacks.emplace( ptr, idx );
+}
+
 void Worker::InsertPlot( PlotData* plot, int64_t time, double val )
 {
     if( plot->data.empty() )

server/TracyWorker.hpp

@@ -17,6 +17,7 @@
 #include "tracy_flat_hash_map.hpp"
 #include "TracyEvent.hpp"
 #include "TracySlab.hpp"
+#include "TracyVarArray.hpp"
 
 namespace tracy
 {
@@ -98,6 +99,9 @@ class Worker
         bool sourceLocationZonesReady;
 #endif
 
+        flat_hash_map<VarArray<uint64_t>*, uint32_t, VarArrayHasherPOT<uint64_t>, VarArrayComparator<uint64_t>> callstackMap;
+        Vector<VarArray<uint64_t>*> callstackPayload;
+
         std::map<uint32_t, LockMap> lockMap;
 
         flat_hash_map<uint64_t, uint16_t, nohash<uint64_t>> threadMap;
@@ -243,6 +247,8 @@ private:
     void AddThreadString( uint64_t id, char* str, size_t sz );
     void AddCustomString( uint64_t ptr, char* str, size_t sz );
 
+    void AddCallstackPayload( uint64_t ptr, char* data, size_t sz );
+
     void InsertPlot( PlotData* plot, int64_t time, double val );
     void HandlePlotName( uint64_t name, char* str, size_t sz );
 
@@ -288,6 +294,7 @@ private:
     flat_hash_map<uint16_t, GpuCtxData*, nohash<uint16_t>> m_gpuCtxMap;
     flat_hash_map<uint64_t, StringLocation, nohash<uint64_t>> m_pendingCustomStrings;
     flat_hash_map<uint64_t, PlotData*, nohash<uint64_t>> m_pendingPlots;
+    flat_hash_map<uint64_t, uint32_t> m_pendingCallstacks;
     flat_hash_map<uint64_t, PlotData*, nohash<uint64_t>> m_plotMap;
     flat_hash_map<const char*, PlotData*, charutil::HasherPOT, charutil::Comparator> m_plotRev;
     flat_hash_map<uint64_t, int32_t, nohash<uint64_t>> m_pendingSourceLocationPayload;