Allow concurrent copies in blit kernel path.

Also removed an unnecessary cache flush in dependency barrier packet.

Change-Id: I573df3bdf0a10df0bcd78025672c44038f8091ff


[ROCm/ROCR-Runtime commit: 4647a5454d]

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_blit_kernel.cpp

@@ -625,9 +625,8 @@ hsa_status_t BlitKernel::SubmitLinearCopyCommand(
   // Insert barrier packets to handle dependent signals.
   // Barrier bit keeps signal checking traffic from competing with a copy.
   const uint16_t kBarrierPacketHeader = (HSA_PACKET_TYPE_BARRIER_AND << HSA_PACKET_HEADER_TYPE) |
-      (1 << HSA_PACKET_HEADER_BARRIER) |
       (HSA_FENCE_SCOPE_NONE << HSA_PACKET_HEADER_SCACQUIRE_FENCE_SCOPE) |
-      (HSA_FENCE_SCOPE_AGENT << HSA_PACKET_HEADER_SCRELEASE_FENCE_SCOPE);
+      (HSA_FENCE_SCOPE_NONE << HSA_PACKET_HEADER_SCRELEASE_FENCE_SCOPE);
 
   hsa_barrier_and_packet_t barrier_packet = {0};
   barrier_packet.header = HSA_PACKET_TYPE_INVALID;
@@ -807,7 +806,6 @@ void BlitKernel::PopulateQueue(uint64_t index, uint64_t code_handle, void* args,
 
   static const uint16_t kDispatchPacketHeader =
       (HSA_PACKET_TYPE_KERNEL_DISPATCH << HSA_PACKET_HEADER_TYPE) |
-      (((completion_signal.handle != 0) ? 1 : 0) << HSA_PACKET_HEADER_BARRIER) |
       (HSA_FENCE_SCOPE_SYSTEM << HSA_PACKET_HEADER_SCACQUIRE_FENCE_SCOPE) |
       (HSA_FENCE_SCOPE_SYSTEM << HSA_PACKET_HEADER_SCRELEASE_FENCE_SCOPE);