P4 to Git Change 1104627 by gandryey@gera-dev-w7 on 2014/12/11 12:01:30

EPR #411058 - [CQE OCL][Lnx][QR][CZ]MultiDevice_Context fails in 2.0 conformance wimpyfull due to CL# 1101352
	- The detection of different map types is overcomplicated with possibility of multiple maps and multithreading environment. Thus keep USWC indirect map optimization based on the allocation flags.

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpumemory.cpp#114 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpumemory.hpp#46 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#342 edit


[ROCm/clr commit: 593d1e3b8d]

projects/clr/rocclr/runtime/device/gpu/gpuvirtual.cpp

@@ -1090,8 +1090,7 @@ VirtualGPU::submitMapMemory(amd::MapMemoryCommand& vcmd)
         vcmd.mapFlags(), vcmd.isEntireMemory());
 
     // If we have host memory, use it
-    if ((memory->owner()->getHostMem() != NULL) &&
-        memory->isDirectMap(vcmd.mapFlags())) {
+    if ((memory->owner()->getHostMem() != NULL) && memory->isDirectMap()) {
         if (!memory->isHostMemDirectAccess()) {
             // Make sure GPU finished operation before
             // synchronization with the backing store
@@ -1176,8 +1175,7 @@ VirtualGPU::submitUnmapMemory(amd::UnmapMemoryCommand& vcmd)
     amd::Memory* owner = memory->owner();
 
     // We used host memory
-    if ((owner->getHostMem() != NULL) &&
-        memory->isDirectMap(memory->isUnmapRead() ? CL_MAP_READ : 0)) {
+    if ((owner->getHostMem() != NULL) && memory->isDirectMap()) {
         if (memory->isUnmapWrite()) {
             // Target is the backing store, so sync
             owner->signalWrite(NULL);