# Node ID f14a633b3ac81f14c82ebedc4af60d9360a8069c
# Parent  01a0e9fbaad4274aa1d11b7bf1919a52816f38dd
diff --git a/src/arch/hsail/insts/decl.hh b/src/arch/hsail/insts/decl.hh
--- a/src/arch/hsail/insts/decl.hh
+++ b/src/arch/hsail/insts/decl.hh
@@ -941,7 +941,11 @@
         initiateAcc(GPUDynInstPtr gpuDynInst)
         {
             Wavefront *wave = gpuDynInst->wavefront();
-            wave->computeUnit->injectGlobalMemFence(gpuDynInst);
+            Request* req = new Request(0, 0, 0, 0,
+                                       wave->computeUnit->masterId(), 0,
+                                       gpuDynInst->wfDynId, -1);
+            gpuDynInst->setRequestFlags(req, true);
+            wave->computeUnit->injectGlobalMemFence(gpuDynInst, false, req);
         }
 
         void
@@ -963,6 +967,7 @@
                 gpuDynInst->kern_id = w->kern_id;
                 gpuDynInst->cu_id = w->computeUnit->cu_id;
 
+                gpuDynInst->s_type = SEG_GLOBAL;
                 gpuDynInst->memoryOrder =
                     getGenericMemoryOrder(memFenceMemOrder);
                 gpuDynInst->scope =
diff --git a/src/gpu-compute/compute_unit.hh b/src/gpu-compute/compute_unit.hh
--- a/src/gpu-compute/compute_unit.hh
+++ b/src/gpu-compute/compute_unit.hh
@@ -300,8 +300,7 @@
     virtual void init();
     void sendRequest(GPUDynInstPtr gpuDynInst, int index, PacketPtr pkt);
     void sendSyncRequest(GPUDynInstPtr gpuDynInst, int index, PacketPtr pkt);
-    void injectGlobalMemFence(GPUDynInstPtr gpuDynInst,
-                              bool kernelLaunch=true,
+    void injectGlobalMemFence(GPUDynInstPtr gpuDynInst, bool kernelLaunch,
                               RequestPtr req=nullptr);
     void handleMemPacket(PacketPtr pkt, int memport_index);
     bool processTimingPacket(PacketPtr pkt);