[Bf-blender-cvs] [6b1a4fc] master: Cycle CUDA: revert the f1aeb2ccf4 and 84f958754 busywait fixes for now.

[Brecht Van Lommel]

Commit: 6b1a4fc66ef4e3197601318ce4c36db2c8359b98
Author: Brecht Van Lommel
Date:   Wed Feb 19 15:59:15 2014 +0100
https://developer.blender.org/rB6b1a4fc66ef4e3197601318ce4c36db2c8359b98

Cycle CUDA: revert the f1aeb2ccf4 and 84f958754 busywait fixes for now.

It's unclear what kind of impact they have on performance at the moment, so I
rather play it safe and postpone this for 2.71.

Ref T38679, Ref T38712

===================================================================

M	intern/cycles/device/device_cuda.cpp

===================================================================

diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index fa65aef..0fbb48c 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -41,8 +41,6 @@ public:
 	CUdevice cuDevice;
 	CUcontext cuContext;
 	CUmodule cuModule;
-	CUstream cuStream;
-	CUevent tileDone;
 	map<device_ptr, bool> tex_interp_map;
 	int cuDevId;
 	int cuDevArchitecture;
@@ -209,9 +207,6 @@ public:
 		if(cuda_error_(result, "cuCtxCreate"))
 			return;
 
-		cuda_assert(cuStreamCreate(&cuStream, 0))
-		cuda_assert(cuEventCreate(&tileDone, 0x1))
-
 		int major, minor;
 		cuDeviceComputeCapability(&major, &minor, cuDevId);
 		cuDevArchitecture = major*100 + minor*10;
@@ -228,8 +223,6 @@ public:
 	{
 		task_pool.stop();
 
-		cuda_assert(cuEventDestroy(tileDone))
-		cuda_assert(cuStreamDestroy(cuStream))
 		cuda_assert(cuCtxDestroy(cuContext))
 	}
 
@@ -652,7 +645,9 @@ public:
 
 		cuda_assert(cuFuncSetCacheConfig(cuPathTrace, CU_FUNC_CACHE_PREFER_L1))
 		cuda_assert(cuFuncSetBlockShape(cuPathTrace, xthreads, ythreads, 1))
-		cuda_assert(cuLaunchGridAsync(cuPathTrace, xblocks, yblocks, cuStream))
+		cuda_assert(cuLaunchGrid(cuPathTrace, xblocks, yblocks))
+
+		cuda_assert(cuCtxSynchronize())
 
 		cuda_pop_context();
 	}
@@ -986,8 +981,6 @@ public:
 
 					task->update_progress(tile);
 				}
-				cuda_assert(cuEventRecord(tileDone, cuStream ))
-				cuda_assert(cuEventSynchronize(tileDone))
 
 				task->release_tile(tile);
 			}