P4 to Git Change 1381244 by gandryey@gera-w8 on 2017/03/03 17:58:38

SWDEV-107546 - [ROCm CQE][OCL][LC/HSAIL][mGPU][G] WF conf test "Buffers" fails in mGPU configs
	- Add MGPU coherency layer support

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocblit.cpp#14 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocblit.hpp#5 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#42 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.hpp#17 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocmemory.cpp#12 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocmemory.hpp#6 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#32 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/memory.cpp#125 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/memory.hpp#99 edit

rocclr/runtime/device/rocm/rocdevice.cpp

@@ -1382,9 +1382,12 @@ Device::createMemory(amd::Memory &owner) const
         return NULL;
     }
 
+    // Transfer data only if OCL context has one device.
+    // Cache coherency layer will update data for multiple devices
     if (!memory->isHostMemDirectAccess() && owner.asImage() &&
-        owner.parent() == NULL &&
-        (owner.getMemFlags() & (CL_MEM_COPY_HOST_PTR | CL_MEM_USE_HOST_PTR))) {
+        (owner.parent() == nullptr) &&
+        (owner.getMemFlags() & CL_MEM_COPY_HOST_PTR) &&
+        (owner.getContext().devices().size() == 1)) {
         // To avoid recurssive call to Device::createMemory, we perform
         // data transfer to the view of the image.
         amd::Image* imageView = owner.asImage()->createView(
@@ -1417,15 +1420,18 @@ Device::createMemory(amd::Memory &owner) const
                                       amd::Coord3D(0, 0, 0), imageView->getRegion(),
                                       0,
                                       0, true);
-        // Release host memory for single device, since runtime copied data
-        if ((owner.getMemFlags() & CL_MEM_COPY_HOST_PTR) &&
-            (owner.getContext().devices().size() == 1)) {
-            owner.setHostMem(nullptr);
-        }
+
+        // Release host memory, since runtime copied data
+        owner.setHostMem(nullptr);
 
         imageView->release();
     }
 
+    // Prepin sysmem buffer for possible data synchronization between CPU and GPU
+    if (!memory->isHostMemDirectAccess() && (owner.getHostMem() != nullptr)) {
+        memory->pinSystemMemory(owner.getHostMem(), owner.getSize());
+    }
+
     if (!result) {
         delete memory;
         return NULL;