SWDEV-270013 - Allocate kernel_arguments from kern_arg & finegrain pool instead of coarse grain.

Change-Id: Id4c6977934fdd6ef2311f6e75593801f1e51983c

rocclr/device/gpu/gpudevice.cpp

@@ -2144,7 +2144,7 @@ void Device::fillHwSampler(uint32_t state, void* hwState, uint32_t hwStateSize,
                      hwStateSize);
 }
 
-void* Device::hostAlloc(size_t size, size_t alignment, bool atomics) const {
+void* Device::hostAlloc(size_t size, size_t alignment, bool MemorySegment mem_seg) const {
   // for discrete gpu, we only reserve,no commit yet.
   return amd::Os::reserveMemory(NULL, size, alignment, amd::Os::MEM_PROT_NONE);
 }

rocclr/device/gpu/gpudevice.hpp

@@ -523,7 +523,7 @@ class Device : public NullDevice, public CALGSLDevice {
                      ) const;
 
   //! host memory alloc
-  virtual void* hostAlloc(size_t size, size_t alignment, bool atomics = false) const;
+  virtual void* hostAlloc(size_t size, size_t alignment, MemorySegment mem_seg = kNoAtomics) const;
 
   //! SVM allocation
   virtual void* svmAlloc(amd::Context& context, size_t size, size_t alignment,