adjust according to performance profiling

optimize async sound loads by starting load early
allow for 0 decals

some more details on the threading changes:

with a queued renderer, having the character rendering be threaded has a lot of contention with locks,
especially in the queue.

cl_threaded_bone_setup 1 seems to work

client leaf system was causing hitches when it was not threaded.

Author: mastercoms

engine/audio/private/snd_wave_data.cpp

@@ -2071,19 +2071,20 @@ bool CWaveDataStreamAsync::IsReadyToMix()
 {
 	if ( IsPC() )
 	{
-		// If not async loaded, start mixing right away
-		if ( !m_source.IsAsyncLoad() && !snd_async_fullyasync.GetBool() )
-		{
-			return true;
-		}
-
+		// Notify load
 		bool bCacheValid;
 		bool bLoaded = wavedatacache->IsDataLoadCompleted( m_hCache, &bCacheValid );
 		if ( !bCacheValid )
 		{
 			wavedatacache->RestartDataLoad( &m_hCache, GetFileName(), m_dataSize, m_dataStart );
 		}
-		return bLoaded;
+
+		// Start mixing right away unless not loaded
+		if (bLoaded || !m_source.IsAsyncLoad() && !snd_async_fullyasync.GetBool())
+		{
+			return true;
+		}
+		return false;
 	}
 
 	if ( IsX360() )
@@ -2334,11 +2335,10 @@ bool CWaveDataMemoryAsync::IsReadyToMix()
 	if ( !m_source.IsAsyncLoad() && !snd_async_fullyasync.GetBool() )
 	{
 		// Wait until we're pending at least
-		if ( m_source.GetCacheStatus() == CAudioSource::AUDIO_NOT_LOADED )
+		if ( m_source.GetCacheStatus() != CAudioSource::AUDIO_NOT_LOADED )
 		{
-			return false;
+			return true;
 		}
-		return true;
 	}
 
 	if ( m_source.IsCached() )

engine/r_decal.cpp

@@ -823,18 +823,31 @@ static decal_t *R_DecalAlloc( int flags )
 
 	if ( iSlot == -1 )
 	{
-		iSlot = R_FindDynamicDecalSlot( g_iLastReplacedDynamic+1 );
-		if ( iSlot == -1 )
+		// if the user has requested 0 decals, let them
+		if (dynamicDecalLimit > 0)
+		{
+			iSlot = R_FindDynamicDecalSlot(g_iLastReplacedDynamic + 1);
+		    if ( iSlot == -1 )
+		    {
+			    if ( !bWarningOnce )
+			    {
+				    // Can't find a free slot. Just kill the first one.
+				    DevWarning( 1, "Exceeded MAX_DECALS (%d).\n", g_nMaxDecals );
+				    bWarningOnce = true;
+			    }
+			    iSlot = 0;
+		    }
+		}
+		else if (bPermanent)
 		{
-			if ( !bWarningOnce )
-			{
-				// Can't find a free slot. Just kill the first one.
-				DevWarning( 1, "Exceeded MAX_DECALS (%d).\n", g_nMaxDecals );
-				bWarningOnce = true;
-			}
 			iSlot = 0;
 		}
 
+		if (iSlot == -1)
+		{
+			return NULL;
+		}
+
 		R_DecalUnlink( s_aDecalPool[iSlot], host_state.worldbrush );
 		g_iLastReplacedDynamic = iSlot;
 	}
@@ -1700,6 +1713,11 @@ static void R_DecalCreate( decalinfo_t* decalinfo, SurfaceHandle_t surfID, float
 	}
 
 	pdecal = R_DecalAlloc( decalinfo->m_Flags );
+
+	if (!pdecal)
+	{
+		return;
+	}
 
 	pdecal->flags = decalinfo->m_Flags;
 	pdecal->color = decalinfo->m_Color;

game/client/c_baseanimating.cpp

@@ -2728,7 +2728,7 @@ ConVar cl_warn_thread_contested_bone_setup("cl_warn_thread_contested_bone_setup"
 // Marked this developmentonly because it currently crashes, and users are enabling it and complaining because of
 // course.  Once this actually works it should just be FCVAR_INTERNAL_USE.
 // UNDONE(mastercoms)
-ConVar cl_threaded_bone_setup("cl_threaded_bone_setup", "0", FCVAR_INTERNAL_USE,
+ConVar cl_threaded_bone_setup("cl_threaded_bone_setup", "1", FCVAR_INTERNAL_USE,
                               "Enable parallel processing of C_BaseAnimating::SetupBones()" );
 
 //-----------------------------------------------------------------------------

game/client/clientleafsystem.cpp

@@ -32,7 +32,7 @@ class VMatrix;  // forward decl
 static ConVar cl_drawleaf("cl_drawleaf", "-1", FCVAR_CHEAT );
 static ConVar r_PortalTestEnts( "r_PortalTestEnts", "1", FCVAR_CHEAT, "Clip entities against portal frustums." );
 static ConVar r_portalsopenall( "r_portalsopenall", "0", FCVAR_CHEAT, "Open all portals" );
-static ConVar cl_threaded_client_leaf_system("cl_threaded_client_leaf_system", "0"  );
+static ConVar cl_threaded_client_leaf_system("cl_threaded_client_leaf_system", "1"  );
 
 
 DEFINE_FIXEDSIZE_ALLOCATOR( CClientRenderablesList, 1, CUtlMemoryPool::GROW_SLOW );

game/client/viewrender.cpp

@@ -116,8 +116,7 @@ ConVar r_drawviewmodel( "r_drawviewmodel","1", FCVAR_CHEAT );
 #endif
 static ConVar r_drawtranslucentrenderables( "r_drawtranslucentrenderables", "1", FCVAR_CHEAT );
 static ConVar r_drawopaquerenderables( "r_drawopaquerenderables", "1", FCVAR_CHEAT );
-static ConVar r_threaded_renderables( "r_threaded_renderables", "1" );
-static ConVar r_threaded_character_rendering("r_threaded_character_rendering", "1");
+static ConVar r_threaded_renderables( "r_threaded_renderables", "0" );
 
 // FIXME: This is not static because we needed to turn it off for TF2 playtests
 ConVar r_DrawDetailProps( "r_DrawDetailProps", "1", FCVAR_NONE, "0=Off, 1=Normal, 2=Wireframe" );
@@ -4159,14 +4158,7 @@ void CRendering3dView::DrawOpaqueRenderables( ERenderDepthMode DepthMode )
 	//
 	// Draw NPCs now
 	//
-	if (r_threaded_character_rendering.GetBool())
-	{
-		DrawOpaqueRenderables_Parallel(arrRenderEntsNpcsFirst.Base(), numNpcs, DepthMode, "Draw NPCs");
-	}
-	else
-	{
-		DrawOpaqueRenderables_Range(arrRenderEntsNpcsFirst.Base(), arrRenderEntsNpcsFirst.Base() + numNpcs, DepthMode);
-	}
+	DrawOpaqueRenderables_Range(arrRenderEntsNpcsFirst.Base(), arrRenderEntsNpcsFirst.Base() + numNpcs, DepthMode);
 
 	//
 	// Ropes and particles