rocr/driver: add scratch memory allocation into driver interface

Add AllocateScratchMemory interface to Driver base class and implement it
in both KFD and XDNA drivers. This change encapsulates the low-level
scratch memory allocation details within driver implementations, making
the code more maintainable and the interface cleaner.

The main changes include:
- Add AllocateScratchMemory virtual method to Driver interface
- Implement the interface in KfdDriver with existing allocation logic
- Add stub implementation in XdnaDriver
- Update GpuAgent to use the new interface instead of direct KMT calls

Signed-off-by: Honglei Huang <Honglei1.Huang@amd.com>


[ROCm/ROCR-Runtime commit: da8dd9e1e3]

projects/rocr-runtime/runtime/hsa-runtime/core/driver/kfd/amd_kfd_driver.cpp

@@ -574,6 +574,21 @@ hsa_status_t KfdDriver::SetTrapHandler(uint32_t node_id, const void* base, uint6
   return HSA_STATUS_SUCCESS;
 }
 
+hsa_status_t KfdDriver::AllocateScratchMemory(uint32_t node_id, uint64_t size, void** mem) const {
+  assert(mem);
+  assert(size > 0);
+
+  HsaMemFlags flags = {};
+  flags.ui32.Scratch = 1;
+  flags.ui32.HostAccess = 1;
+
+  void* ptr = AllocateKfdMemory(flags, node_id, size);
+  if (ptr == nullptr) return HSA_STATUS_ERROR_OUT_OF_RESOURCES;
+
+  *mem = ptr;
+  return HSA_STATUS_SUCCESS;
+}
+
 hsa_status_t KfdDriver::GetDeviceHandle(uint32_t node_id, void** device_handle) const {
   assert(device_handle);
 

projects/rocr-runtime/runtime/hsa-runtime/core/driver/xdna/amd_xdna_driver.cpp

@@ -877,6 +877,10 @@ hsa_status_t XdnaDriver::SetTrapHandler(uint32_t node_id, const void* base, uint
   return HSA_STATUS_ERROR;
 }
 
+hsa_status_t XdnaDriver::AllocateScratchMemory(uint32_t node_id, uint64_t size, void** mem) const {
+  return HSA_STATUS_ERROR;
+}
+
 hsa_status_t XdnaDriver::GetDeviceHandle(uint32_t node_id, void** device_handle) const {
   return HSA_STATUS_ERROR;
 }