Sign Up
Log In
Log In
or
Sign Up
Places
All Projects
Status Monitor
Collapse sidebar
home:AndnoVember:test
OpenUSD
usd-tbb.patch
Overview
Repositories
Revisions
Requests
Users
Attributes
Meta
File usd-tbb.patch of Package OpenUSD
From 9443dc6121948c054ade923ab0ee4fddfa6214a1 Mon Sep 17 00:00:00 2001 From: Alex Fuller <boberfly@gmail.com> Date: Wed, 17 May 2023 16:27:47 +0200 Subject: [PATCH 01/14] oneTBB: tbb::atomic to std::atomic in pcp --- pxr/usd/pcp/mapExpression.cpp | 4 ++-- pxr/usd/pcp/mapExpression.h | 3 +-- pxr/usd/pcp/pch.h | 1 - 3 files changed, 3 insertions(+), 5 deletions(-) Index: OpenUSD-23.11/pxr/usd/pcp/mapExpression.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/usd/pcp/mapExpression.cpp +++ OpenUSD-23.11/pxr/usd/pcp/mapExpression.cpp @@ -238,7 +238,7 @@ PcpMapExpression::_Node::New( _Op op_, // Check for existing instance to re-use _NodeMap::accessor accessor; if (_nodeRegistry->map.insert(accessor, key) || - accessor->second->_refCount.fetch_and_increment() == 0) { + accessor->second->_refCount.fetch_add(1) == 0) { // Either there was no node in the table, or there was but it had // begun dying (another client dropped its refcount to 0). We have // to create a new node in the table. When the client that is @@ -388,7 +388,7 @@ intrusive_ptr_add_ref(PcpMapExpression:: void intrusive_ptr_release(PcpMapExpression::_Node* p) { - if (p->_refCount.fetch_and_decrement() == 1) + if (p->_refCount.fetch_sub(1) == 1) delete p; } Index: OpenUSD-23.11/pxr/usd/pcp/mapExpression.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/pcp/mapExpression.h +++ OpenUSD-23.11/pxr/usd/pcp/mapExpression.h @@ -30,7 +30,6 @@ #include <boost/intrusive_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/spin_mutex.h> #include <atomic> @@ -267,7 +266,7 @@ private: // data struct _NodeMap; static TfStaticData<_NodeMap> _nodeRegistry; - mutable tbb::atomic<int> _refCount; + mutable std::atomic<int> _refCount; mutable Value _cachedValue; mutable std::set<_Node*> _dependentExpressions; Value _valueForVariable; Index: OpenUSD-23.11/pxr/usd/pcp/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/pcp/pch.h +++ OpenUSD-23.11/pxr/usd/pcp/pch.h @@ -194,7 +194,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_hash_map.h> Index: OpenUSD-23.11/pxr/usd/usdGeom/bboxCache.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdGeom/bboxCache.cpp +++ OpenUSD-23.11/pxr/usd/usdGeom/bboxCache.cpp @@ -46,6 +46,7 @@ #include <tbb/enumerable_thread_specific.h> #include <algorithm> +#include <atomic> PXR_NAMESPACE_OPEN_SCOPE @@ -124,11 +125,24 @@ private: struct _PrototypeTask { - _PrototypeTask() : numDependencies(0) { } + _PrototypeTask() noexcept + : numDependencies(0) { } + + _PrototypeTask(const _PrototypeTask &other) noexcept + : dependentPrototypes(other.dependentPrototypes) + { + numDependencies.store(other.numDependencies.load()); + } + + _PrototypeTask(_PrototypeTask &&other) noexcept + : dependentPrototypes(std::move(other.dependentPrototypes)) + { + numDependencies.store(other.numDependencies.load()); + } // Number of dependencies -- prototype prims that must be resolved // before this prototype can be resolved. - tbb::atomic<size_t> numDependencies; + std::atomic<size_t> numDependencies; // List of prototype prims that depend on this prototype. std::vector<_PrimContext> dependentPrototypes; @@ -220,7 +234,7 @@ private: _PrototypeTask& dependentPrototypeData = prototypeTasks->find(dependentPrototype)->second; if (dependentPrototypeData.numDependencies - .fetch_and_decrement() == 1){ + .fetch_sub(1) == 1){ dispatcher->Run( &_PrototypeBBoxResolver::_ExecuteTaskForPrototype, this, dependentPrototype, prototypeTasks, xfCaches, @@ -1522,4 +1536,3 @@ UsdGeomBBoxCache::_PrimContext::ToString } PXR_NAMESPACE_CLOSE_SCOPE - Index: OpenUSD-23.11/pxr/usd/usdGeom/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdGeom/pch.h +++ OpenUSD-23.11/pxr/usd/usdGeom/pch.h @@ -181,7 +181,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usd/sdf/changeManager.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/usd/sdf/changeManager.cpp +++ OpenUSD-23.11/pxr/usd/sdf/changeManager.cpp @@ -34,7 +34,7 @@ #include "pxr/base/tf/instantiateSingleton.h" #include "pxr/base/tf/stackTrace.h" -#include <tbb/atomic.h> +#include <atomic> using std::string; using std::vector; @@ -150,9 +150,9 @@ Sdf_ChangeManager::_ProcessRemoveIfInert TF_VERIFY(data->outermostBlock); } -static tbb::atomic<size_t> & +static std::atomic<size_t> & _InitChangeSerialNumber() { - static tbb::atomic<size_t> value; + static std::atomic<size_t> value; value = 1; return value; } @@ -191,8 +191,8 @@ Sdf_ChangeManager::_SendNotices(_Data *d } // Obtain a serial number for this round of change processing. - static tbb::atomic<size_t> &changeSerialNumber = _InitChangeSerialNumber(); - size_t serialNumber = changeSerialNumber.fetch_and_increment(); + static std::atomic<size_t> &changeSerialNumber = _InitChangeSerialNumber(); + size_t serialNumber = changeSerialNumber.fetch_add(1); // Send global notice. SdfNotice::LayersDidChange(changes, serialNumber).Send(); Index: OpenUSD-23.11/pxr/usd/sdf/layer.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/usd/sdf/layer.cpp +++ OpenUSD-23.11/pxr/usd/sdf/layer.cpp @@ -214,7 +214,7 @@ SdfLayer::SdfLayer( _MarkCurrentStateAsClean(); } -SdfLayer::~SdfLayer() +SdfLayer::~SdfLayer() noexcept { TF_PY_ALLOW_THREADS_IN_SCOPE(); Index: OpenUSD-23.11/pxr/usd/sdf/layer.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/sdf/layer.h +++ OpenUSD-23.11/pxr/usd/sdf/layer.h @@ -98,7 +98,7 @@ class SdfLayer public: /// Destructor SDF_API - virtual ~SdfLayer(); + virtual ~SdfLayer() noexcept; // noexcept needed for std::atomic member /// Noncopyable SdfLayer(const SdfLayer&) = delete; Index: OpenUSD-23.11/pxr/usd/sdf/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/sdf/pch.h +++ OpenUSD-23.11/pxr/usd/sdf/pch.h @@ -225,7 +225,6 @@ #include <boost/variant.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_hash_map.h> Index: OpenUSD-23.11/extras/usd/examples/usdObj/pch.h =================================================================== --- OpenUSD-23.11.orig/extras/usd/examples/usdObj/pch.h +++ OpenUSD-23.11/extras/usd/examples/usdObj/pch.h @@ -166,7 +166,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/extras/usd/examples/usdSchemaExamples/pch.h =================================================================== --- OpenUSD-23.11.orig/extras/usd/examples/usdSchemaExamples/pch.h +++ OpenUSD-23.11/extras/usd/examples/usdSchemaExamples/pch.h @@ -168,7 +168,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/base/plug/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/base/plug/pch.h +++ OpenUSD-23.11/pxr/base/plug/pch.h @@ -183,7 +183,6 @@ #include <boost/type_traits/remove_reference.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_vector.h> #include <tbb/enumerable_thread_specific.h> Index: OpenUSD-23.11/pxr/base/tf/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/base/tf/pch.h +++ OpenUSD-23.11/pxr/base/tf/pch.h @@ -242,7 +242,6 @@ #include <boost/variant.hpp> #include <boost/variant/get.hpp> #include <boost/variant/variant.hpp> -#include <tbb/atomic.h> #include <tbb/enumerable_thread_specific.h> #include <tbb/spin_mutex.h> #include <tbb/spin_rw_mutex.h> Index: OpenUSD-23.11/pxr/base/trace/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/base/trace/pch.h +++ OpenUSD-23.11/pxr/base/trace/pch.h @@ -178,7 +178,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_vector.h> Index: OpenUSD-23.11/pxr/base/vt/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/base/vt/pch.h +++ OpenUSD-23.11/pxr/base/vt/pch.h @@ -171,7 +171,6 @@ #include <boost/type_traits/is_same.hpp> #include <boost/type_traits/remove_reference.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_unordered_map.h> #include <tbb/enumerable_thread_specific.h> Index: OpenUSD-23.11/pxr/base/work/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/base/work/pch.h +++ OpenUSD-23.11/pxr/base/work/pch.h @@ -110,7 +110,6 @@ #include <boost/type_traits/is_enum.hpp> #include <boost/type_traits/is_same.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_vector.h> Index: OpenUSD-23.11/pxr/imaging/garch/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/garch/pch.h +++ OpenUSD-23.11/pxr/imaging/garch/pch.h @@ -145,7 +145,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/enumerable_thread_specific.h> #include <tbb/spin_rw_mutex.h> #ifdef PXR_PYTHON_SUPPORT_ENABLED Index: OpenUSD-23.11/pxr/imaging/glf/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/glf/pch.h +++ OpenUSD-23.11/pxr/imaging/glf/pch.h @@ -199,7 +199,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/imaging/hd/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hd/pch.h +++ OpenUSD-23.11/pxr/imaging/hd/pch.h @@ -152,7 +152,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility/enable_if.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/imaging/hdGp/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hdGp/pch.h +++ OpenUSD-23.11/pxr/imaging/hdGp/pch.h @@ -120,7 +120,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_map.h> Index: OpenUSD-23.11/pxr/imaging/hdMtlx/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hdMtlx/pch.h +++ OpenUSD-23.11/pxr/imaging/hdMtlx/pch.h @@ -139,7 +139,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_hash_map.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/imaging/hdSt/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hdSt/pch.h +++ OpenUSD-23.11/pxr/imaging/hdSt/pch.h @@ -170,7 +170,6 @@ #include <opensubdiv/osd/cpuVertexBuffer.h> #include <opensubdiv/osd/mesh.h> #include <opensubdiv/version.h> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/imaging/hdar/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hdar/pch.h +++ OpenUSD-23.11/pxr/imaging/hdar/pch.h @@ -131,7 +131,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/spin_mutex.h> Index: OpenUSD-23.11/pxr/imaging/hdsi/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hdsi/pch.h +++ OpenUSD-23.11/pxr/imaging/hdsi/pch.h @@ -119,7 +119,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/concurrent_queue.h> #ifdef PXR_PYTHON_SUPPORT_ENABLED #include "pxr/base/tf/pySafePython.h" Index: OpenUSD-23.11/pxr/imaging/hdx/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hdx/pch.h +++ OpenUSD-23.11/pxr/imaging/hdx/pch.h @@ -152,7 +152,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility/enable_if.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/imaging/hgiMetal/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hgiMetal/pch.h +++ OpenUSD-23.11/pxr/imaging/hgiMetal/pch.h @@ -141,7 +141,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility/enable_if.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/spin_mutex.h> Index: OpenUSD-23.11/pxr/imaging/plugin/hdEmbree/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/plugin/hdEmbree/pch.h +++ OpenUSD-23.11/pxr/imaging/plugin/hdEmbree/pch.h @@ -154,7 +154,6 @@ #include <embree3/rtcore.h> #include <embree3/rtcore_geometry.h> #include <embree3/rtcore_ray.h> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/imaging/plugin/hdStorm/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/plugin/hdStorm/pch.h +++ OpenUSD-23.11/pxr/imaging/plugin/hdStorm/pch.h @@ -141,7 +141,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility/enable_if.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/spin_mutex.h> Index: OpenUSD-23.11/pxr/imaging/plugin/hioOiio/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/plugin/hioOiio/pch.h +++ OpenUSD-23.11/pxr/imaging/plugin/hioOiio/pch.h @@ -199,7 +199,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usd/ar/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/ar/pch.h +++ OpenUSD-23.11/pxr/usd/ar/pch.h @@ -166,7 +166,6 @@ #include <boost/type_traits/is_same.hpp> #include <boost/type_traits/remove_reference.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/concurrent_hash_map.h> #include <tbb/enumerable_thread_specific.h> #include <tbb/spin_rw_mutex.h> Index: OpenUSD-23.11/pxr/usd/ndr/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/ndr/pch.h +++ OpenUSD-23.11/pxr/usd/ndr/pch.h @@ -198,7 +198,6 @@ #include <boost/type_traits/remove_reference.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/enumerable_thread_specific.h> Index: OpenUSD-23.11/pxr/usd/plugin/usdAbc/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/plugin/usdAbc/pch.h +++ OpenUSD-23.11/pxr/usd/plugin/usdAbc/pch.h @@ -206,7 +206,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/plugin/usdDraco/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/plugin/usdDraco/pch.h +++ OpenUSD-23.11/pxr/usd/plugin/usdDraco/pch.h @@ -169,7 +169,6 @@ #include <draco/compression/encode.h> #include <draco/mesh/mesh.h> #include <draco/mesh/mesh_misc_functions.h> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usd/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usd/pch.h +++ OpenUSD-23.11/pxr/usd/usd/pch.h @@ -227,7 +227,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/utility/in_place_factory.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_hash_map.h> Index: OpenUSD-23.11/pxr/usd/usdHydra/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdHydra/pch.h +++ OpenUSD-23.11/pxr/usd/usdHydra/pch.h @@ -162,7 +162,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usdLux/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdLux/pch.h +++ OpenUSD-23.11/pxr/usd/usdLux/pch.h @@ -177,7 +177,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usdMedia/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdMedia/pch.h +++ OpenUSD-23.11/pxr/usd/usdMedia/pch.h @@ -170,7 +170,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usdMtlx/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdMtlx/pch.h +++ OpenUSD-23.11/pxr/usd/usdMtlx/pch.h @@ -193,7 +193,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_map.h> Index: OpenUSD-23.11/pxr/usd/usdPhysics/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdPhysics/pch.h +++ OpenUSD-23.11/pxr/usd/usdPhysics/pch.h @@ -181,7 +181,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usd/usdProc/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdProc/pch.h +++ OpenUSD-23.11/pxr/usd/usdProc/pch.h @@ -165,7 +165,6 @@ #include <boost/variant.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usdRender/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdRender/pch.h +++ OpenUSD-23.11/pxr/usd/usdRender/pch.h @@ -170,7 +170,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usdRi/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdRi/pch.h +++ OpenUSD-23.11/pxr/usd/usdRi/pch.h @@ -173,7 +173,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usdShade/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdShade/pch.h +++ OpenUSD-23.11/pxr/usd/usdShade/pch.h @@ -179,7 +179,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usd/usdSkel/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdSkel/pch.h +++ OpenUSD-23.11/pxr/usd/usdSkel/pch.h @@ -180,7 +180,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usd/usdUI/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdUI/pch.h +++ OpenUSD-23.11/pxr/usd/usdUI/pch.h @@ -168,7 +168,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usd/usdUtils/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdUtils/pch.h +++ OpenUSD-23.11/pxr/usd/usdUtils/pch.h @@ -215,7 +215,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usd/usdVol/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usdVol/pch.h +++ OpenUSD-23.11/pxr/usd/usdVol/pch.h @@ -170,7 +170,6 @@ #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/base/tf/testenv/error.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/base/tf/testenv/error.cpp +++ OpenUSD-23.11/pxr/base/tf/testenv/error.cpp @@ -29,7 +29,7 @@ #include "pxr/base/arch/functionLite.h" -#include <tbb/tbb_thread.h> +#include <thread> #define FILENAME "error.cpp" @@ -195,7 +195,7 @@ Test_TfErrorThreadTransport() printf("Creating TfErrorMark\n"); TfErrorMark m; printf("Launching thread\n"); - tbb::tbb_thread t([&transport]() { _ThreadTask(&transport); }); + std::thread t([&transport]() { _ThreadTask(&transport); }); TF_AXIOM(m.IsClean()); t.join(); printf("Thread completed, posting error.\n"); Index: OpenUSD-23.11/pxr/usd/usd/clipCache.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usd/clipCache.cpp +++ OpenUSD-23.11/pxr/usd/usd/clipCache.cpp @@ -218,10 +218,9 @@ Usd_ClipCache::PopulateClipsForPrim( const bool primHasClips = !allClips.empty(); if (primHasClips) { TRACE_SCOPE("Usd_ClipCache::PopulateClipsForPrim (primHasClips)"); - tbb::mutex::scoped_lock lock; - if (_concurrentPopulationContext) { - lock.acquire(_concurrentPopulationContext->_mutex); - } + std::unique_lock<std::mutex> lock = (_concurrentPopulationContext) ? + std::unique_lock<std::mutex>(_concurrentPopulationContext->_mutex) : + std::unique_lock<std::mutex>(); // Find nearest ancestor with clips specified. const std::vector<Usd_ClipSetRefPtr>* ancestralClips = nullptr; @@ -260,10 +259,10 @@ Usd_ClipCache::PopulateClipsForPrim( SdfLayerHandleSet Usd_ClipCache::GetUsedLayers() const { - tbb::mutex::scoped_lock lock; - if (_concurrentPopulationContext) { - lock.acquire(_concurrentPopulationContext->_mutex); - } + std::unique_lock<std::mutex> lock = (_concurrentPopulationContext) ? + std::unique_lock<std::mutex>(_concurrentPopulationContext->_mutex) : + std::unique_lock<std::mutex>(); + SdfLayerHandleSet layers; for (_ClipTable::iterator::value_type const &clipsListIter : _table){ for (Usd_ClipSetRefPtr const &clipSet : clipsListIter.second){ @@ -342,10 +341,9 @@ const std::vector<Usd_ClipSetRefPtr>& Usd_ClipCache::GetClipsForPrim(const SdfPath& path) const { TRACE_FUNCTION(); - tbb::mutex::scoped_lock lock; - if (_concurrentPopulationContext) { - lock.acquire(_concurrentPopulationContext->_mutex); - } + std::unique_lock<std::mutex> lock = (_concurrentPopulationContext) ? + std::unique_lock<std::mutex>(_concurrentPopulationContext->_mutex) : + std::unique_lock<std::mutex>(); return _GetClipsForPrim_NoLock(path); } Index: OpenUSD-23.11/pxr/usd/usd/clipCache.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usd/clipCache.h +++ OpenUSD-23.11/pxr/usd/usd/clipCache.h @@ -30,7 +30,7 @@ #include "pxr/usd/usd/clipSet.h" #include "pxr/usd/sdf/pathTable.h" -#include <tbb/mutex.h> +#include <mutex> #include <vector> PXR_NAMESPACE_OPEN_SCOPE @@ -61,7 +61,7 @@ public: explicit ConcurrentPopulationContext(Usd_ClipCache &cache); ~ConcurrentPopulationContext(); Usd_ClipCache &_cache; - tbb::mutex _mutex; + std::mutex _mutex; }; /// Populate the cache with clips for \p prim. Returns true if clips Index: OpenUSD-23.11/pxr/usd/usd/crateFile.h =================================================================== --- OpenUSD-23.11.orig/pxr/usd/usd/crateFile.h +++ OpenUSD-23.11/pxr/usd/usd/crateFile.h @@ -349,12 +349,15 @@ private: bool operator!=(ZeroCopySource const &other) const { return !(*this == other); } - friend size_t tbb_hasher(ZeroCopySource const &z) { - return TfHash::Combine( - reinterpret_cast<uintptr_t>(z._addr), - z._numBytes - ); - } + + struct Hash { + inline size_t operator()(const ZeroCopySource& z) const { + return TfHash::Combine( + reinterpret_cast<uintptr_t>(z._addr), + z._numBytes + ); + } + }; // Return true if the refcount is nonzero. bool IsInUse() const { return _refCount; } @@ -422,7 +425,7 @@ private: ArchConstFileMapping _mapping; char const *_start; int64_t _length; - tbb::concurrent_unordered_set<ZeroCopySource> _outstandingRanges; + tbb::concurrent_unordered_set<ZeroCopySource, ZeroCopySource::Hash> _outstandingRanges; }; public: Index: OpenUSD-23.11/pxr/base/trace/concurrentList.h =================================================================== --- OpenUSD-23.11.orig/pxr/base/trace/concurrentList.h +++ OpenUSD-23.11/pxr/base/trace/concurrentList.h @@ -111,7 +111,7 @@ public: while (curNode) { Node* nodeToDelete = curNode; curNode = curNode->next; - _alloc.destroy(nodeToDelete); + nodeToDelete->~Node(); _alloc.deallocate(nodeToDelete, 1); } } @@ -130,7 +130,7 @@ public: /// the newly created item. iterator Insert() { Node* newNode = _alloc.allocate(1); - _alloc.construct(newNode); + new(newNode) Node(); // Add the node to the linked list in an atomic manner. do { Index: OpenUSD-23.11/pxr/usdImaging/plugin/usdShaders/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/plugin/usdShaders/pch.h +++ OpenUSD-23.11/pxr/usdImaging/plugin/usdShaders/pch.h @@ -160,7 +160,6 @@ #include <boost/unordered_map.hpp> #include <boost/utility.hpp> #include <boost/utility/enable_if.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usdImaging/usdAppUtils/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdAppUtils/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdAppUtils/pch.h @@ -173,7 +173,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_set.h> Index: OpenUSD-23.11/pxr/usdImaging/usdImaging/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdImaging/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdImaging/pch.h @@ -173,7 +173,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usdImaging/usdImaging/resolvedAttributeCache.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdImaging/resolvedAttributeCache.h +++ OpenUSD-23.11/pxr/usdImaging/usdImaging/resolvedAttributeCache.h @@ -283,22 +283,36 @@ private: // non-time varying data, entries may exist in the cache with invalid // values. The version is used to determine validity. struct _Entry { - _Entry() + _Entry() noexcept : value(Strategy::MakeDefault()) , version(_GetInitialEntryVersion()) { } _Entry(const query_type & query_, const value_type& value_, - unsigned version_) + unsigned version_) noexcept : query(query_) , value(value_) , version(version_) { } + _Entry(const _Entry &other) noexcept + : query(other.query) + , value(other.value) + { + version.store(other.version.load()); + } + + _Entry(_Entry &&other) noexcept + : query(std::move(other.query)) + , value(std::move(other.value)) + { + version.store(other.version.load()); + } + query_type query; value_type value; - tbb::atomic<unsigned> version; + std::atomic<unsigned> version; }; // Returns the version number for a valid cache entry @@ -338,7 +352,7 @@ private: // A serial number indicating the valid state of entries in the cache. When // an entry has an equal or greater value, the entry is valid. - tbb::atomic<unsigned> _cacheVersion; + std::atomic<unsigned> _cacheVersion; // Value overrides for a set of descendents. ValueOverridesMap _valueOverrides; @@ -357,7 +371,7 @@ UsdImaging_ResolvedAttributeCache<Strate // Note: _cacheVersion is not allowed to change during cache access. unsigned v = entry->version; if (v < _cacheVersion - && entry->version.compare_and_swap(_cacheVersion, v) == v) + && entry->version.compare_exchange_strong(v, _cacheVersion.load())) { entry->value = value; entry->version = _GetValidVersion(); @@ -377,7 +391,7 @@ typename UsdImaging_ResolvedAttributeCac UsdImaging_ResolvedAttributeCache<Strategy, ImplData>::_GetCacheEntryForPrim( const UsdPrim &prim) const { - typename _CacheMap::const_iterator it = _cache.find(prim); + typename _CacheMap::iterator it = _cache.find(prim); if (it != _cache.end()) { return &it->second; } Index: OpenUSD-23.11/pxr/usdImaging/usdImagingGL/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdImagingGL/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdImagingGL/pch.h @@ -186,7 +186,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_map.h> Index: OpenUSD-23.11/pxr/usdImaging/usdProcImaging/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdProcImaging/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdProcImaging/pch.h @@ -161,7 +161,6 @@ #include <boost/variant.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_hash_map.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usdImaging/usdRiPxrImaging/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdRiPxrImaging/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdRiPxrImaging/pch.h @@ -169,7 +169,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_map.h> Index: OpenUSD-23.11/pxr/usdImaging/usdSkelImaging/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdSkelImaging/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdSkelImaging/pch.h @@ -169,7 +169,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/blocked_range.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/usdImaging/usdVolImaging/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdVolImaging/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdVolImaging/pch.h @@ -167,7 +167,6 @@ #include <boost/utility/enable_if.hpp> #include <boost/variant.hpp> #include <boost/weak_ptr.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_queue.h> #include <tbb/concurrent_unordered_map.h> Index: OpenUSD-23.11/pxr/usdImaging/usdviewq/pch.h =================================================================== --- OpenUSD-23.11.orig/pxr/usdImaging/usdviewq/pch.h +++ OpenUSD-23.11/pxr/usdImaging/usdviewq/pch.h @@ -164,7 +164,6 @@ #include <boost/variant.hpp> #include <boost/vmd/is_empty.hpp> #include <boost/vmd/is_tuple.hpp> -#include <tbb/atomic.h> #include <tbb/cache_aligned_allocator.h> #include <tbb/concurrent_hash_map.h> #include <tbb/concurrent_queue.h> Index: OpenUSD-23.11/pxr/imaging/hd/dependencyForwardingSceneIndex.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/imaging/hd/dependencyForwardingSceneIndex.cpp +++ OpenUSD-23.11/pxr/imaging/hd/dependencyForwardingSceneIndex.cpp @@ -211,7 +211,7 @@ HdDependencyForwardingSceneIndex::_PrimD void HdDependencyForwardingSceneIndex::_ClearDependencies(const SdfPath &primPath) { - _AffectedPrimToDependsOnPathsEntryMap::const_iterator it = + _AffectedPrimToDependsOnPathsEntryMap::iterator it = _affectedPrimToDependsOnPathsMap.find(primPath); if (it == _affectedPrimToDependsOnPathsMap.end()) { return; Index: OpenUSD-23.11/pxr/base/work/testenv/testWorkThreadLimits.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/base/work/testenv/testWorkThreadLimits.cpp +++ OpenUSD-23.11/pxr/base/work/testenv/testWorkThreadLimits.cpp @@ -38,6 +38,10 @@ #include <set> #include <thread> +#if TBB_INTERFACE_VERSION_MAJOR >= 12 +#include <tbb/global_control.h> +#endif + using namespace std::placeholders; PXR_NAMESPACE_USING_DIRECTIVE @@ -56,16 +60,41 @@ _CountThreads(size_t begin, size_t end) _uniqueThreads->insert(std::this_thread::get_id()); } +static unsigned +_GetConcurrencyLimit() +{ +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + // For oneTBB, get limit in an arena with max concurrency as + // WorkSetConcurrencyLimit by itself no longer increases the concurrency + // beyond the number of cores by itself. + unsigned limit; + tbb::task_arena arena(tbb::global_control::active_value(tbb::global_control::max_allowed_parallelism)); + arena.execute([&]() { + limit = WorkGetConcurrencyLimit(); + }); + return limit; +#else + return WorkGetConcurrencyLimit(); +#endif +} + static size_t _ExpectedLimit(const int envVal, const size_t n) { // If envVal is non-zero, it wins over n! // envVal may also be a negative number, which means all but that many // cores. - return envVal ? + const size_t val = envVal ? (envVal < 0 ? std::max<int>(1, envVal+WorkGetPhysicalConcurrencyLimit()) : envVal) : n; + +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + // oneTBB has an internal limit of 256 + 1 threads. + return std::min<size_t>(val, 257); +#else + return val; +#endif } static void @@ -101,41 +130,41 @@ _TestArguments(const int envVal) // Set to maximum concurrency, which should remain within envVal. const int numCores = WorkGetPhysicalConcurrencyLimit(); WorkSetConcurrencyLimitArgument(numCores); - TF_AXIOM(WorkGetConcurrencyLimit() == _ExpectedLimit(envVal, numCores)); + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, numCores)); // n = 0, means "no change" WorkSetConcurrencyLimitArgument(0); - TF_AXIOM(WorkGetConcurrencyLimit() == _ExpectedLimit(envVal, numCores)); + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, numCores)); // n = 1 means no threading WorkSetConcurrencyLimitArgument(1); - TF_AXIOM(WorkGetConcurrencyLimit() == _ExpectedLimit(envVal, 1)); + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 1)); // n = 3 means 3 WorkSetConcurrencyLimitArgument(3); - TF_AXIOM(WorkGetConcurrencyLimit() == _ExpectedLimit(envVal, 3)); + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 3)); // n = 1000 means 1000 WorkSetConcurrencyLimitArgument(1000); - TF_AXIOM(WorkGetConcurrencyLimit() == _ExpectedLimit(envVal, 1000)); + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 1000)); // n = -1 means numCores - 1, with a minimum of 1 WorkSetConcurrencyLimitArgument(-1); - TF_AXIOM(WorkGetConcurrencyLimit() == + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, std::max(1, numCores-1))); // n = -3 means numCores - 3, with a minimum of 1 WorkSetConcurrencyLimitArgument(-3); - TF_AXIOM(WorkGetConcurrencyLimit() == + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, std::max(1, numCores-3))); // n = -numCores means 1 (no threading) WorkSetConcurrencyLimitArgument(-numCores); - TF_AXIOM(WorkGetConcurrencyLimit() == _ExpectedLimit(envVal, 1)); + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 1)); // n = -numCores*10 means 1 (no threading) WorkSetConcurrencyLimitArgument(-numCores*10); - TF_AXIOM(WorkGetConcurrencyLimit() == _ExpectedLimit(envVal, 1)); + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 1)); } struct _RawTBBCounter @@ -218,35 +247,35 @@ main(int argc, char **argv) // Test with full concurrency. std::cout << "Testing full concurrency...\n"; WorkSetMaximumConcurrencyLimit(); - TF_AXIOM(WorkGetConcurrencyLimit() == + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, WorkGetPhysicalConcurrencyLimit())); _TestThreadLimit(envVal, WorkGetPhysicalConcurrencyLimit()); // Test with no concurrency. std::cout << "Testing turning off concurrency...\n"; WorkSetConcurrencyLimit(1); - TF_AXIOM(WorkGetConcurrencyLimit() == + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 1)); _TestThreadLimit(envVal, 1); // Test with 2 threads. std::cout << "Testing with 2 threads...\n"; WorkSetConcurrencyLimit(2); - TF_AXIOM(WorkGetConcurrencyLimit() == + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 2)); _TestThreadLimit(envVal, 2); // Test with 4 threads. std::cout << "Testing with 4 threads...\n"; WorkSetConcurrencyLimit(4); - TF_AXIOM(WorkGetConcurrencyLimit() == + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 4)); _TestThreadLimit(envVal, 4); // Test with 1000 threads. std::cout << "Testing with 1000 threads...\n"; WorkSetConcurrencyLimit(1000); - TF_AXIOM(WorkGetConcurrencyLimit() == + TF_AXIOM(_GetConcurrencyLimit() == _ExpectedLimit(envVal, 1000)); _TestThreadLimit(envVal, 1000); Index: OpenUSD-23.11/pxr/base/work/threadLimits.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/base/work/threadLimits.cpp +++ OpenUSD-23.11/pxr/base/work/threadLimits.cpp @@ -29,9 +29,18 @@ #include "pxr/base/tf/envSetting.h" -#include <tbb/task_scheduler_init.h> +// Blocked range is not used in this file, but this header happens to pull in +// the TBB version header in a way that works in all TBB versions. +#include <tbb/blocked_range.h> #include <tbb/task_arena.h> +#if TBB_INTERFACE_VERSION_MAJOR >= 12 +#include <tbb/global_control.h> +#include <tbb/info.h> +#else +#include <tbb/task_scheduler_init.h> +#endif + #include <algorithm> #include <atomic> @@ -58,16 +67,25 @@ TF_DEFINE_ENV_SETTING( PXR_NAMESPACE_OPEN_SCOPE -// We create a task_scheduler_init instance at static initialization time if -// PXR_WORK_THREAD_LIMIT is set to a nonzero value. Otherwise this stays NULL. -static tbb::task_scheduler_init *_tbbTaskSchedInit; +// We create a global_control or task_scheduler_init instance at static +// initialization time if PXR_WORK_THREAD_LIMIT is set to a nonzero value. +// Otherwise this stays NULL. +#if TBB_INTERFACE_VERSION_MAJOR >= 12 +static tbb::global_control *_tbbGlobalControl = nullptr; +#else +static tbb::task_scheduler_init *_tbbTaskSchedInit = nullptr; +#endif unsigned WorkGetPhysicalConcurrencyLimit() { // Use TBB here, since it pays attention to the affinity mask on Linux and // Windows. +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + return tbb::info::default_concurrency(); +#else return tbb::task_scheduler_init::default_num_threads(); +#endif } // This function always returns an actual thread count >= 1. @@ -123,7 +141,11 @@ Work_InitializeThreading() // previously initialized by the hosting environment (e.g. if we are running // as a plugin to another application.) if (settingVal) { +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + _tbbGlobalControl = new tbb::global_control(tbb::global_control::max_allowed_parallelism, threadLimit); +#else _tbbTaskSchedInit = new tbb::task_scheduler_init(threadLimit); +#endif } } static int _forceInitialization = (Work_InitializeThreading(), 0); @@ -153,6 +175,11 @@ WorkSetConcurrencyLimit(unsigned n) threadLimit = WorkGetConcurrencyLimit(); } + +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + delete _tbbGlobalControl; + _tbbGlobalControl = new tbb::global_control(tbb::global_control::max_allowed_parallelism, threadLimit); +#else // Note that we need to do some performance testing and decide if it's // better here to simply delete the task_scheduler_init object instead // of re-initializing it. If we decide that it's better to re-initialize @@ -168,6 +195,7 @@ WorkSetConcurrencyLimit(unsigned n) } else { _tbbTaskSchedInit = new tbb::task_scheduler_init(threadLimit); } +#endif } void @@ -185,7 +213,14 @@ WorkSetConcurrencyLimitArgument(int n) unsigned WorkGetConcurrencyLimit() { +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + // The effective concurrency requires taking into account both the + // task_arena and internal thread pool size set by global_control. + // https://github.com/oneapi-src/oneTBB/issues/405 + return std::min<unsigned>(tbb::global_control::active_value(tbb::global_control::max_allowed_parallelism), tbb::this_task_arena::max_concurrency()); +#else return tbb::this_task_arena::max_concurrency(); +#endif } bool Index: OpenUSD-23.11/pxr/base/work/dispatcher.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/base/work/dispatcher.cpp +++ OpenUSD-23.11/pxr/base/work/dispatcher.cpp @@ -32,27 +32,42 @@ WorkDispatcher::WorkDispatcher() tbb::task_group_context::isolated, tbb::task_group_context::concurrent_wait | tbb::task_group_context::default_traits) +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + , _taskGroup(_context) +#endif { _waitCleanupFlag.clear(); - + +#if TBB_INTERFACE_VERSION_MAJOR < 12 // The concurrent_wait flag used with the task_group_context ensures // the ref count will remain at 1 after all predecessor tasks are // completed, so we don't need to keep resetting it in Wait(). _rootTask = new(tbb::task::allocate_root(_context)) tbb::empty_task; _rootTask->set_ref_count(1); +#endif } -WorkDispatcher::~WorkDispatcher() +WorkDispatcher::~WorkDispatcher() noexcept { Wait(); + +#if TBB_INTERFACE_VERSION_MAJOR < 12 tbb::task::destroy(*_rootTask); +#endif } void WorkDispatcher::Wait() { // Wait for tasks to complete. +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + // The native task_group::wait() has a comment saying its call to the + // context reset method is not thread safe. So we bypass that implementation + // and do our own synchronization to ensure it is called once. + tbb::detail::d1::wait(_taskGroup.get_internal_wait_context(), _context); +#else _rootTask->wait_for_all(); +#endif // If we take the flag from false -> true, we do the cleanup. if (_waitCleanupFlag.test_and_set() == false) { @@ -73,7 +88,11 @@ WorkDispatcher::Wait() void WorkDispatcher::Cancel() { +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + _taskGroup.cancel(); +#else _context.cancel_group_execution(); +#endif } /* static */ Index: OpenUSD-23.11/pxr/base/work/dispatcher.h =================================================================== --- OpenUSD-23.11.orig/pxr/base/work/dispatcher.h +++ OpenUSD-23.11/pxr/base/work/dispatcher.h @@ -33,8 +33,15 @@ #include "pxr/base/tf/errorMark.h" #include "pxr/base/tf/errorTransport.h" +// Blocked range is not used in this file, but this header happens to pull in +// the TBB version header in a way that works in all TBB versions. +#include <tbb/blocked_range.h> #include <tbb/concurrent_vector.h> +#if TBB_INTERFACE_VERSION_MAJOR >= 12 +#include <tbb/task_group.h> +#else #include <tbb/task.h> +#endif #include <functional> #include <type_traits> @@ -79,7 +86,7 @@ public: WORK_API WorkDispatcher(); /// Wait() for any pending tasks to complete, then destroy the dispatcher. - WORK_API ~WorkDispatcher(); + WORK_API ~WorkDispatcher() noexcept; // noexcept needed for tbb::task_group WorkDispatcher(WorkDispatcher const &) = delete; WorkDispatcher &operator=(WorkDispatcher const &) = delete; @@ -103,7 +110,11 @@ public: template <class Callable> inline void Run(Callable &&c) { +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + _taskGroup.run(std::move(_InvokerTask<typename std::remove_reference<Callable>::type>(std::move(c), &_errors))); +#else _rootTask->spawn(_MakeInvokerTask(std::forward<Callable>(c))); +#endif } template <class Callable, class A0, class ... Args> @@ -136,12 +147,38 @@ private: // Function invoker helper that wraps the invocation with an ErrorMark so we // can transmit errors that occur back to the thread that Wait() s for tasks // to complete. +#if TBB_INTERFACE_VERSION_MAJOR >= 12 template <class Fn> - struct _InvokerTask : public tbb::task { + struct _InvokerTask { explicit _InvokerTask(Fn &&fn, _ErrorTransports *err) - : _fn(std::move(fn)), _errors(err) {} + : _fn(std::make_unique<Fn>(std::move(fn))), _errors(err) {} explicit _InvokerTask(Fn const &fn, _ErrorTransports *err) + : _fn(std::make_unique<Fn>(std::move(fn))), _errors(err) {} + + // Ensure only moves happen, no copies or assignments. + _InvokerTask(_InvokerTask &&other) = default; + _InvokerTask(const _InvokerTask &other) = delete; + _InvokerTask &operator=(const _InvokerTask &other) = delete; + _InvokerTask &operator=(_InvokerTask &&other) = delete; + + void operator()() const { + TfErrorMark m; + (*_fn)(); + if (!m.IsClean()) + WorkDispatcher::_TransportErrors(m, _errors); + } + private: + std::unique_ptr<Fn> _fn; + _ErrorTransports *_errors; + }; +#else + template <class Fn> + struct _InvokerTask : public tbb::task { + explicit _InvokerTask(Fn &&fn, _ErrorTransports *err) + : _fn(std::move(fn)), _errors(err) {} + + explicit _InvokerTask(Fn const &fn, _ErrorTransports *err) : _fn(fn), _errors(err) {} virtual tbb::task* execute() { @@ -164,16 +201,30 @@ private: _InvokerTask<typename std::remove_reference<Fn>::type>( std::forward<Fn>(fn), &_errors); } +#endif // Helper function that removes errors from \p m and stores them in a new // entry in \p errors. WORK_API static void _TransportErrors(const TfErrorMark &m, _ErrorTransports *errors); - // Task group context and associated root task that allows us to cancel - // tasks invoked directly by this dispatcher. + // Task group context to run tasks in. tbb::task_group_context _context; +#if TBB_INTERFACE_VERSION_MAJOR >= 12 + // Custom task group that lets us implement thread safe concurrent wait. + class _TaskGroup : public tbb::task_group { + public: + _TaskGroup(tbb::task_group_context& ctx) : tbb::task_group(ctx) {} + tbb::detail::d1::wait_context& get_internal_wait_context() { + return m_wait_ctx; + } + }; + + _TaskGroup _taskGroup; +#else + // Root task that allows us to cancel tasks invoked directly by this dispatcher. tbb::empty_task* _rootTask; +#endif // The error transports we use to transmit errors in other threads back to // this thread. Index: OpenUSD-23.11/pxr/base/work/testenv/testWorkLoops.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/base/work/testenv/testWorkLoops.cpp +++ OpenUSD-23.11/pxr/base/work/testenv/testWorkLoops.cpp @@ -35,6 +35,7 @@ #include <functional> #include <cstdio> +#include <cstring> #include <numeric> #include <iostream> #include <vector> Index: OpenUSD-23.11/pxr/base/work/testenv/testWorkReduce.cpp =================================================================== --- OpenUSD-23.11.orig/pxr/base/work/testenv/testWorkReduce.cpp +++ OpenUSD-23.11/pxr/base/work/testenv/testWorkReduce.cpp @@ -35,6 +35,7 @@ #include <functional> #include <cstdio> +#include <cstring> #include <iostream> #include <vector> Index: OpenUSD-23.11/build_scripts/build_usd.py =================================================================== --- OpenUSD-23.11.orig/build_scripts/build_usd.py +++ OpenUSD-23.11/build_scripts/build_usd.py @@ -939,13 +939,19 @@ elif MacOS(): else: TBB_URL = "https://github.com/oneapi-src/oneTBB/archive/refs/tags/v2020.3.zip" +TBB_2021_URL = "https://github.com/oneapi-src/oneTBB/archive/refs/tags/v2021.9.0.zip" + def InstallTBB(context, force, buildArgs): - if Windows(): - InstallTBB_Windows(context, force, buildArgs) - elif MacOS(): - InstallTBB_MacOS(context, force, buildArgs) + if context.tbbVersion == "2021": + with CurrentWorkingDirectory(DownloadURL(TBB_2021_URL, context, force)): + RunCMake(context, force, buildArgs) else: - InstallTBB_Linux(context, force, buildArgs) + if Windows(): + InstallTBB_Windows(context, force, buildArgs) + elif MacOS(): + InstallTBB_MacOS(context, force, buildArgs) + else: + InstallTBB_Linux(context, force, buildArgs) def InstallTBB_Windows(context, force, buildArgs): with CurrentWorkingDirectory(DownloadURL(TBB_URL, context, force, @@ -1202,17 +1208,20 @@ BLOSC = Dependency("Blosc", InstallBLOSC # OpenVDB OPENVDB_URL = "https://github.com/AcademySoftwareFoundation/openvdb/archive/refs/tags/v9.1.0.zip" - -# OpenVDB v9.1.0 requires TBB 2019.0 or above, but this script installs -# TBB 2018 on macOS Intel systems for reasons documented above. So we -# keep OpenVDB at the version specified for the VFX Reference Platform -# CY2021, which is the last version that supported 2018. -OPENVDB_INTEL_URL = "https://github.com/AcademySoftwareFoundation/openvdb/archive/refs/tags/v8.2.0.zip" +OPENVDB_8_URL = "https://github.com/AcademySoftwareFoundation/openvdb/archive/refs/tags/v8.2.0.zip" +OPENVDB_10_URL = "https://github.com/AcademySoftwareFoundation/openvdb/archive/refs/tags/v10.0.1.zip" def InstallOpenVDB(context, force, buildArgs): openvdb_url = OPENVDB_URL - if MacOS() and not apple_utils.IsTargetArm(context): - openvdb_url = OPENVDB_INTEL_URL + # OpenVDB 10 is the first release to support oneTBB. + if context.tbbVersion == "2021": + openvdb_url = OPENVDB_10_URL + # OpenVDB v9.1.0 requires TBB 2019.0 or above, but this script installs + # TBB 2018 on macOS Intel systems for reasons documented above. So we + # keep OpenVDB at the version specified for the VFX Reference Platform + # CY2021, which is the last version that supported 2018. + elif MacOS() and not apple_utils.IsTargetArm(context): + openvdb_url = OPENVDB_8_URL with CurrentWorkingDirectory(DownloadURL(openvdb_url, context, force)): extraArgs = [ @@ -1942,6 +1951,13 @@ subgroup.add_argument("--no-openvdb", de action="store_false", help="Disable OpenVDB support in imaging (default)") subgroup = group.add_mutually_exclusive_group() +subgroup.add_argument("--onetbb", dest="enable_onetbb", action="store_true", + default=False, + help="Use new oneAPI TBB version") +subgroup.add_argument("--no-onetbb", dest="enable_onetbb", + action="store_false", + help="Use old TBB version (default)") +subgroup = group.add_mutually_exclusive_group() subgroup.add_argument("--usdview", dest="build_usdview", action="store_true", default=True, help="Build usdview (default)") @@ -2114,6 +2130,9 @@ class InstallContext: self.buildTutorials = args.build_tutorials self.buildTools = args.build_tools + # - TBB + self.tbbVersion = "2021" if args.enable_onetbb else "2019" + # - Imaging self.buildImaging = (args.build_imaging == IMAGING or args.build_imaging == USD_IMAGING) @@ -2378,6 +2397,7 @@ summaryMsg += """\ Python support {buildPython} Python Debug: {debugPython} Python docs: {buildPythonDocs} + TBB version: {tbbVersion} Documentation {buildDocs} Tests {buildTests} Examples {buildExamples} @@ -2439,6 +2459,7 @@ summaryMsg = summaryMsg.format( buildPython=("On" if context.buildPython else "Off"), debugPython=("On" if context.debugPython else "Off"), buildPythonDocs=("On" if context.buildPythonDocs else "Off"), + tbbVersion=context.tbbVersion, buildDocs=("On" if context.buildDocs else "Off"), buildTests=("On" if context.buildTests else "Off"), buildExamples=("On" if context.buildExamples else "Off"), Index: OpenUSD-23.11/cmake/modules/FindTBB.cmake =================================================================== --- OpenUSD-23.11.orig/cmake/modules/FindTBB.cmake +++ OpenUSD-23.11/cmake/modules/FindTBB.cmake @@ -197,7 +197,12 @@ if(NOT TBB_FOUND) ################################## if(TBB_INCLUDE_DIRS) - file(READ "${TBB_INCLUDE_DIRS}/tbb/tbb_stddef.h" _tbb_version_file) + # Use new oneTBB version header if it exists. + if(EXISTS "${TBB_INCLUDE_DIRS}/oneapi/tbb/version.h") + file(READ "${TBB_INCLUDE_DIRS}/oneapi/tbb/version.h" _tbb_version_file) + else() + file(READ "${TBB_INCLUDE_DIRS}/tbb/tbb_stddef.h" _tbb_version_file) + endif() string(REGEX REPLACE ".*#define TBB_VERSION_MAJOR ([0-9]+).*" "\\1" TBB_VERSION_MAJOR "${_tbb_version_file}") string(REGEX REPLACE ".*#define TBB_VERSION_MINOR ([0-9]+).*" "\\1" @@ -211,15 +216,22 @@ if(NOT TBB_FOUND) # Find TBB components ################################## + # oneTBB on Windows has interface version in the name. + if(WIN32 AND TBB_INTERFACE_VERSION GREATER_EQUAL 12000) + set(_tbb_library_name tbb12) + else() + set(_tbb_library_name tbb) + endif() + if(TBB_VERSION VERSION_LESS 4.3) - set(TBB_SEARCH_COMPOMPONENTS tbb_preview tbbmalloc tbb) + set(TBB_SEARCH_COMPOMPONENTS tbb_preview tbbmalloc ${_tbb_library_name}) else() - set(TBB_SEARCH_COMPOMPONENTS tbb_preview tbbmalloc_proxy tbbmalloc tbb) + set(TBB_SEARCH_COMPOMPONENTS tbb_preview tbbmalloc_proxy tbbmalloc ${_tbb_library_name}) endif() # Find each component foreach(_comp ${TBB_SEARCH_COMPOMPONENTS}) - if(";${TBB_FIND_COMPONENTS};tbb;" MATCHES ";${_comp};") + if(";${TBB_FIND_COMPONENTS};" MATCHES ";${_comp};") # Search for the libraries find_library(TBB_${_comp}_LIBRARY_RELEASE ${_comp} @@ -227,7 +239,7 @@ if(NOT TBB_FOUND) PATHS ${TBB_DEFAULT_SEARCH_DIR} ENV LIBRARY_PATH PATH_SUFFIXES ${TBB_LIB_PATH_SUFFIX}) - find_library(TBB_${_comp}_LIBRARY_DEBUG ${_comp}_debug + find_library(TBB_${_comp}_LIBRARY_DEBUG ${_comp} HINTS ${TBB_LIBRARY} ${TBB_SEARCH_DIR} PATHS ${TBB_DEFAULT_SEARCH_DIR} ENV LIBRARY_PATH PATH_SUFFIXES ${TBB_LIB_PATH_SUFFIX})
Locations
Projects
Search
Status Monitor
Help
OpenBuildService.org
Documentation
API Documentation
Code of Conduct
Contact
Support
@OBShq
Terms
openSUSE Build Service is sponsored by
The Open Build Service is an
openSUSE project
.
Sign Up
Log In
Places
Places
All Projects
Status Monitor
