Make texture node threaded

Quite trivial idea -- just pass tread ID to the texture sampling function.

Implemented as a TLS to avoid passing huge amount of extra contexts around.
Should be working on all platforms, but compilation test is required.

Reviewers: juicyfruit, campbellbarton

Reviewed By: campbellbarton

Differential Revision: https://developer.blender.org/D1831

4 files changed, 30 insertions, 7 deletions

diff --git a/source/blender/compositor/intern/COM_CPUDevice.cpp b/source/blender/compositor/intern/COM_CPUDevice.cpp
index c7c3f7769fe..a5824ec5248 100644
--- a/source/blender/compositor/intern/COM_CPUDevice.cpp
+++ b/source/blender/compositor/intern/COM_CPUDevice.cpp
@@ -22,6 +22,12 @@
 
 #include "COM_CPUDevice.h"
 
+CPUDevice::CPUDevice(int thread_id)
+  : Device(),
+    m_thread_id(thread_id)
+{
+}
+
 void CPUDevice::execute(WorkPackage *work)
 {
 	const unsigned int chunkNumber = work->getChunkNumber();
diff --git a/source/blender/compositor/intern/COM_CPUDevice.h b/source/blender/compositor/intern/COM_CPUDevice.h
index 3dc8fff66a3..d12666593d4 100644
--- a/source/blender/compositor/intern/COM_CPUDevice.h
+++ b/source/blender/compositor/intern/COM_CPUDevice.h
@@ -31,11 +31,18 @@
  */
 class CPUDevice : public Device {
 public:
+	CPUDevice(int thread_id);
+
 	/**
 	 * @brief execute a WorkPackage
 	 * @param work the WorkPackage to execute
 	 */
 	void execute(WorkPackage *work);
+
+	int thread_id() { return m_thread_id; }
+
+protected:
+	int m_thread_id;
 };
 
 #endif
diff --git a/source/blender/compositor/intern/COM_WorkScheduler.cpp b/source/blender/compositor/intern/COM_WorkScheduler.cpp
index fc6ea1299cf..4c85f11f655 100644
--- a/source/blender/compositor/intern/COM_WorkScheduler.cpp
+++ b/source/blender/compositor/intern/COM_WorkScheduler.cpp
@@ -50,7 +50,8 @@
 
 
 /// @brief list of all CPUDevices. for every hardware thread an instance of CPUDevice is created
-static vector<CPUDevice *> g_cpudevices;
+static vector<CPUDevice*> g_cpudevices;
+static ThreadLocal(CPUDevice*) g_thread_device;
 
 #if COM_CURRENT_THREADING_MODEL == COM_TM_QUEUE
 /// @brief list of all thread for every CPUDevice in cpudevices a thread exists
@@ -153,9 +154,9 @@ int COM_isHighlightedbNode(bNode *bnode)
 #if COM_CURRENT_THREADING_MODEL == COM_TM_QUEUE
 void *WorkScheduler::thread_execute_cpu(void *data)
 {
-	Device *device = (Device *)data;
+	CPUDevice *device = (CPUDevice *)data;
 	WorkPackage *work;
-	
+	BLI_thread_local_set(g_thread_device, device);
 	while ((work = (WorkPackage *)BLI_thread_queue_pop(g_cpuqueue))) {
 		HIGHLIGHT(work);
 		device->execute(work);
@@ -310,18 +311,20 @@ void WorkScheduler::initialize(bool use_opencl, int num_cpu_threads)
 			device->deinitialize();
 			delete device;
 		}
-
+		if (g_cpuInitialized) {
+			BLI_thread_local_delete(g_thread_device);
+		}
 		g_cpuInitialized = false;
 	}
 
 	/* initialize CPU threads */
 	if (!g_cpuInitialized) {
 		for (int index = 0; index < num_cpu_threads; index++) {
-			CPUDevice *device = new CPUDevice();
+			CPUDevice *device = new CPUDevice(index);
 			device->initialize();
 			g_cpudevices.push_back(device);
 		}
-
+		BLI_thread_local_create(g_thread_device);
 		g_cpuInitialized = true;
 	}
 
@@ -407,7 +410,7 @@ void WorkScheduler::deinitialize()
 			device->deinitialize();
 			delete device;
 		}
-
+		BLI_thread_local_delete(g_thread_device);
 		g_cpuInitialized = false;
 	}
 
@@ -450,3 +453,8 @@ void WorkScheduler::deinitialize()
 	}
 }
 
+int WorkScheduler::current_thread_id()
+{
+	CPUDevice *device = (CPUDevice *)BLI_thread_local_get(g_thread_device);
+	return device->thread_id();
+}
diff --git a/source/blender/compositor/intern/COM_WorkScheduler.h b/source/blender/compositor/intern/COM_WorkScheduler.h
index 27afdf6efd0..67d3fc87ce1 100644
--- a/source/blender/compositor/intern/COM_WorkScheduler.h
+++ b/source/blender/compositor/intern/COM_WorkScheduler.h
@@ -113,6 +113,8 @@ public:
 	 */
 	static bool hasGPUDevices();
 
+	static int current_thread_id();
+
 #ifdef WITH_CXX_GUARDEDALLOC
 	MEM_CXX_CLASS_ALLOC_FUNCS("COM:WorkScheduler")
 #endif