From a632d9e2a3f3af6ddecc92d491a713eb5e42dcd8 Mon Sep 17 00:00:00 2001
From: Bartosz Taudul <wolf.pld@gmail.com>
Date: Tue, 26 Mar 2019 22:06:00 +0100
Subject: [PATCH] Add zone vector cache.

Zone children will be now collected in staging vectors. When the zone is
ended (and no children can be added anymore to it), a size-fitted vector
is allocated using slab allocation. The over-allocated vector is then
put into cache for use in future zones.

This is only active for vectors <= 8192 elements, or 64 KB (chosen
arbitrarily), to reduce time spent on copying memory.

Overall, this change should have the following effects:
- System memory allocation pressure reduction, due to re-usage of
  vectors, which eliminates the need for constant growth.
- Reduction of memory usage, because children vectors are now fitted to
  required size.
- Slight increase of zone processing time, due to memory copying?
---
 server/TracyWorker.cpp | 27 ++++++++++++++++++++++++++-
 server/TracyWorker.hpp |  2 ++
 2 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/server/TracyWorker.cpp b/server/TracyWorker.cpp
index c6f603ae..eb1c4f8f 100644
--- a/server/TracyWorker.cpp
+++ b/server/TracyWorker.cpp
@@ -1978,7 +1978,18 @@ void Worker::NewZone( ZoneEvent* zone, uint64_t thread )
         if( back->child < 0 )
         {
             back->child = int32_t( m_data.zoneChildren.size() );
-            m_data.zoneChildren.push_back( Vector<ZoneEvent*>( zone ) );
+            if( m_data.zoneVectorCache.empty() )
+            {
+                m_data.zoneChildren.push_back( Vector<ZoneEvent*>( zone ) );
+            }
+            else
+            {
+                Vector<ZoneEvent*> vze = std::move( m_data.zoneVectorCache.back_and_pop() );
+                assert( !vze.empty() );
+                vze.clear();
+                vze.push_back_non_empty( zone );
+                m_data.zoneChildren.push_back( std::move( vze ) );
+            }
         }
         else
         {
@@ -2613,6 +2624,20 @@ void Worker::ProcessZoneEnd( const QueueZoneEnd& ev )
 
     m_data.lastTime = std::max( m_data.lastTime, zone->end );
 
+    if( zone->child >= 0 )
+    {
+        auto& childVec = m_data.zoneChildren[zone->child];
+        const auto sz = childVec.size();
+        if( sz <= 8 * 1024 )
+        {
+            Vector<ZoneEvent*> fitVec;
+            fitVec.reserve_exact( sz, m_slab );
+            memcpy( fitVec.data(), childVec.data(), sz * sizeof( ZoneEvent* ) );
+            std::swap( fitVec, childVec );
+            m_data.zoneVectorCache.push_back( std::move( fitVec ) );
+        }
+    }
+
 #ifndef TRACY_NO_STATISTICS
     auto timeSpan = zone->end - zone->start;
     if( timeSpan > 0 )
diff --git a/server/TracyWorker.hpp b/server/TracyWorker.hpp
index 5168cf66..748a3da5 100644
--- a/server/TracyWorker.hpp
+++ b/server/TracyWorker.hpp
@@ -178,6 +178,8 @@ private:
         Vector<Vector<ZoneEvent*>> zoneChildren;
         Vector<Vector<GpuEvent*>> gpuChildren;
 
+        Vector<Vector<ZoneEvent*>> zoneVectorCache;
+
         CrashEvent crashEvent;
     };