Fix T51568: CUDA error in viewport render after fix for for OpenCL

Seems re-loading module invalidates memory pointers by the looks of it,
which gives an error on the next kernel call.

Not sure how to move memory pointer from one CUDA module to another one,
so for now simply disabling kernel re-load for CUDA devices. Not ideal,
but better than failing render.

Feature-selective option for CUDA is not an official feature anyway.

1 files changed, 13 insertions, 0 deletions

diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index 26671bdde28..99537e9a983 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -225,6 +225,9 @@ public:
 		cuDevice = 0;
 		cuContext = 0;
 
+		cuModule = 0;
+		cuFilterModule = 0;
+
 		split_kernel = NULL;
 
 		need_bindless_mapping = false;
@@ -487,6 +490,16 @@ public:
 
 	bool load_kernels(const DeviceRequestedFeatures& requested_features)
 	{
+		/* TODO(sergey): Support kernels re-load for CUDA devices.
+		 *
+		 * Currently re-loading kernel will invalidate memory pointers,
+		 * causing problems in cuCtxSynchronize.
+		 */
+		if(cuFilterModule && cuModule) {
+			VLOG(1) << "Skipping kernel reload, not currently supported.";
+			return true;
+		}
+
 		/* check if cuda init succeeded */
 		if(cuContext == 0)
 			return false;