SWDEV-513197 - Improve launch perf for Device Heap kernels

- If any kernel uses device heap, the launch needs to be preceeded by an
  init kernel, Save on the extra barrier packet launch/flush between the
init heap kernel and user kernel

Change-Id: I8ebc6246188200e5f673dc464bc76a53bcb8b7c6


[ROCm/clr commit: ca530c660b]

projects/clr/rocclr/device/rocm/rocblit.cpp

@@ -2537,9 +2537,10 @@ bool KernelBlitManager::initHeap(device::Memory* heap_to_initialize, device::Mem
   setArgument(kernels_[blitType], 2, sizeof(uint), &heap_size);
   setArgument(kernels_[blitType], 3, sizeof(uint), &number_of_initial_blocks);
   address parameters = captureArguments(kernels_[blitType]);
-  result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters, nullptr);
+  result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters,
+                                      nullptr, 0, nullptr, nullptr, true);
   releaseArguments(parameters);
-  gpu().releaseGpuMemoryFence();
+  gpu().Barriers().WaitCurrent();
   return result;
 }
 

projects/clr/rocclr/device/rocm/rocvirtual.cpp

@@ -851,10 +851,9 @@ bool VirtualGPU::processMemObjects(const amd::Kernel& kernel, const_address para
         WriteAqlArgAt(const_cast<address>(params), mem, sizeof(void*), it->second);
       }
       ClPrint(amd::LOG_INFO, amd::LOG_KERN,
-        "Arg%d: %s %s = val:%s", i, desc.typeName_.c_str(), desc.name_.c_str(),
-        (desc.size_ == 4) ? std::to_string(*reinterpret_cast<const int*>(srcArgPtr)).c_str() :
-        (desc.size_ == 8) ? std::to_string(*reinterpret_cast<const long long*>(srcArgPtr)).c_str()
-        : "");
+        "Arg%d: %s %s = val:0x%lx", i, desc.typeName_.c_str(), desc.name_.c_str(),
+        (desc.size_ == 4) ? *reinterpret_cast<const int*>(srcArgPtr) :
+        (desc.size_ == 8) ? *reinterpret_cast<const long long*>(srcArgPtr) : 0LL);
     }
     else if (desc.type_ == T_SAMPLER) {
       uint32_t index = desc.info_.arrayIndex_;