SWDEV-246687 - Do not use std::vector reference as class member cuMask_

The current implementation creates default reference in the stack and assigns it to class member cuMasks_, so whenever the content of the stack changes, cuMask_ would change.

Change-Id: Iefab63c335d504b83c4ae90bd34ae76c6afb8f3c

rocclr/device/rocm/rocvirtual.cpp

@@ -492,7 +492,7 @@ bool VirtualGPU::dispatchGenericAqlPacket(
   // NOTE: need multiple packets to dispatch the performance counter
   //       packet blob of the legacy devices (gfx8)
   for (uint i = 0; i < size; i++, index++, packet++) {
-    AqlPacket* aql_loc = &((AqlPacket*)(gpu_queue_->base_address))[index & queueMask]; 
+    AqlPacket* aql_loc = &((AqlPacket*)(gpu_queue_->base_address))[index & queueMask];
     *aql_loc = *packet;
     if (header != 0) {
       packet_store_release(reinterpret_cast<uint32_t*>(aql_loc), header, rest);
@@ -503,7 +503,7 @@ bool VirtualGPU::dispatchGenericAqlPacket(
             "group_seg_size=%zu, kernel_obj=0x%zx, kernarg_address=0x%zx, completion_signal=0x%zx",
             std::this_thread::get_id(), gpu_queue_,
             header, extractAqlBits(header, HSA_PACKET_HEADER_TYPE, HSA_PACKET_HEADER_WIDTH_TYPE),
-            extractAqlBits(header, HSA_PACKET_HEADER_BARRIER, 
+            extractAqlBits(header, HSA_PACKET_HEADER_BARRIER,
                            HSA_PACKET_HEADER_WIDTH_BARRIER),
             extractAqlBits(header, HSA_PACKET_HEADER_SCACQUIRE_FENCE_SCOPE,
                            HSA_PACKET_HEADER_WIDTH_SCACQUIRE_FENCE_SCOPE),

rocclr/device/rocm/rocvirtual.hpp

@@ -368,7 +368,7 @@ class VirtualGPU : public device::VirtualDevice {
   uint16_t dispatchPacketHeader_;
 
   //!< bit-vector representing the CU mask. Each active bit represents using one CU
-  const std::vector<uint32_t>& cuMask_;
+  const std::vector<uint32_t> cuMask_;
   amd::CommandQueue::Priority priority_; //!< The priority for the hsa queue
 };
 

rocclr/platform/commandqueue.hpp

@@ -131,14 +131,14 @@ class CommandQueue : public RuntimeObject {
         context_(context),
         cuMask_(cuMask){}
 
-  Properties properties_;             //!< Queue properties
-  uint rtCUs_;                        //!< The number of used RT compute units
-  Priority priority_;                 //!< Queue priority
-  Monitor queueLock_;                 //!< Lock protecting the queue
-  Monitor lastCmdLock_;               //!< Lock protecting the last queued command
-  Device& device_;                    //!< The device
-  SharedReference<Context> context_;  //!< The context of this command queue
-  const std::vector<uint32_t>& cuMask_;  //!< The CU mask
+  Properties properties_;               //!< Queue properties
+  uint rtCUs_;                          //!< The number of used RT compute units
+  Priority priority_;                   //!< Queue priority
+  Monitor queueLock_;                   //!< Lock protecting the queue
+  Monitor lastCmdLock_;                 //!< Lock protecting the last queued command
+  Device& device_;                      //!< The device
+  SharedReference<Context> context_;    //!< The context of this command queue
+  const std::vector<uint32_t> cuMask_;  //!< The CU mask
 
  private:
   //! Disable copy constructor