rocr: add specific flag for blit kernel object

so that aql-to-pm4 conversion could verify the validity of the kernel
object.

Signed-off-by: Flora Cui <flora.cui@amd.com>

Author: fcui-amd

libhsakmt/include/hsakmt/hsakmttypes.h

@@ -586,7 +586,8 @@ typedef struct _HsaMemFlags
             unsigned int GTTAccess:     1;  // default = 0; If 1: The caller indicates this memory will be mapped to GART for MES
 					    // KFD will allocate GTT memory with the Preferred_node set as gpu_id for GART mapping
             unsigned int Contiguous:	1; // Allocate contiguous VRAM
-            unsigned int Reserved:      9;
+            unsigned int ExecuteBlit:	1; // default = 0; If 1: The caller indicates that the memory is for blit kernel object.
+            unsigned int Reserved:      8;
 
         } ui32;
         HSAuint32 Value;

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

@@ -241,6 +241,9 @@ KfdDriver::AllocateMemory(const core::MemoryRegion &mem_region,
             ? 1
             : kmt_alloc_flags.ui32.Uncached);
 
+  kmt_alloc_flags.ui32.ExecuteBlit =
+    !!(alloc_flags & core::MemoryRegion::AllocateExecutableBlitKernelObject);
+
   if (m_region.IsLocalMemory()) {
     // Allocate physically contiguous memory. AllocateKfdMemory function call
     // will fail if this flag is not supported in KFD.

runtime/hsa-runtime/core/inc/memory_region.h

@@ -106,6 +106,9 @@ class MemoryRegion : public Checked<0x9C961F19EE175BB3> {
     AllocateGTTAccess = (1 << 9),
     AllocateContiguous = (1 << 10), // Physically contiguous memory
     AllocateUncached = (1 << 11),   // Uncached memory
+    // this flag is ignored by Thunk and only used for emulator/dxg to track code-object
+    // allocations in AQL to PM4 conversion.
+    AllocateExecutableBlitKernelObject = (1 << 12),
   };
 
   typedef uint32_t AllocateFlags;

runtime/hsa-runtime/core/runtime/amd_gpu_agent.cpp

@@ -378,7 +378,8 @@ void GpuAgent::AssembleShader(const char* func_name, AssembleTarget assemble_tar
       (assemble_target == AssembleTarget::AQL ? sizeof(amd_kernel_code_t) : 0);
   code_buf_size = AlignUp(header_size + asic_shader->size, 0x1000);
 
-  code_buf = system_allocator()(code_buf_size, 0x1000, core::MemoryRegion::AllocateExecutable);
+  code_buf = system_allocator()(code_buf_size, 0x1000,
+    core::MemoryRegion::AllocateExecutable | core::MemoryRegion::AllocateExecutableBlitKernelObject);
   assert(code_buf != NULL && "Code buffer allocation failed");
 
   memset(code_buf, 0, code_buf_size);