diff --git a/include/hcc_detail/hip_hcc.h b/include/hcc_detail/hip_hcc.h
index 3ad1efddeb..8b9f1db97b 100644
--- a/include/hcc_detail/hip_hcc.h
+++ b/include/hcc_detail/hip_hcc.h
@@ -32,7 +32,7 @@ THE SOFTWARE.
 // #define USE_MEMCPYTOSYMBOL
 //
 //Use the new HCC accelerator_view::copy instead of am_copy
-#define USE_AV_COPY (__hcc_workweek__ >= 16351) 
+#define USE_AV_COPY (__hcc_workweek__ >= 16351)
 
 // Compile peer-to-peer support.
 // >= 2 : use HCC hc:accelerator::get_is_peer
@@ -457,7 +457,10 @@ typedef uint64_t SeqNum_t ;
     // Use this if we already have the stream critical data mutex:
     void                 wait(LockedAccessor_StreamCrit_t &crit, bool assertQueueEmpty=false);
 
-    void launchModuleKernel(hsa_signal_t signal, uint32_t blockDimX, uint32_t blockDimY, uint32_t blockDimZ, uint32_t gridDimX, uint32_t gridDimY, uint32_t gridDimZ, uint32_t sharedMemBytes, void *kernarg, size_t kernSize, uint64_t kernel);
+    void launchModuleKernel(hc::accelerator_view av, hsa_signal_t signal,
+                            uint32_t blockDimX, uint32_t blockDimY, uint32_t blockDimZ,
+                            uint32_t gridDimX, uint32_t gridDimY, uint32_t gridDimZ,
+                            uint32_t sharedMemBytes, void *kernarg, size_t kernSize, uint64_t kernel);
 
     // Non-threadsafe accessors - must be protected by high-level stream lock with accessor passed to function.
     SIGSEQNUM            lastCopySeqId (LockedAccessor_StreamCrit_t &crit) const { return crit->_last_copy_signal ? crit->_last_copy_signal->_sigId : 0; };
diff --git a/samples/0_Intro/module_api/Makefile b/samples/0_Intro/module_api/Makefile
index db270beaa0..8981938e91 100644
--- a/samples/0_Intro/module_api/Makefile
+++ b/samples/0_Intro/module_api/Makefile
@@ -3,11 +3,14 @@ ifeq (,$(HIP_PATH))
 	HIP_PATH=../../..
 endif
 HIPCC=$(HIP_PATH)/bin/hipcc
+OPT=
 HIP_PLATFORM=$(shell $(HIP_PATH)/bin/hipconfig --compiler)
 
 ifeq (${HIP_PLATFORM}, hcc)
 	GENCODEOBJECT_FLAGS=--target-isa=fiji
 
+all: runKernel.hip.out
+
 vcpy_isa.compile: vcpy_isa.cpp
 	$(HIPCC) --genco $(GENCODEOBJECT_FLAGS) vcpy_isa.cpp -o vcpy_isa.co
 
@@ -29,5 +32,5 @@ endif
 all: vcpy_isa.compile runKernel.hip.out
 
 runKernel.hip.out: runKernel.cpp
-	$(HIPCC) runKernel.cpp -o runKernel.hip.out
+	$(HIPCC) $(OPT) runKernel.cpp -o runKernel.hip.out
 
diff --git a/src/hip_hcc.cpp b/src/hip_hcc.cpp
index cb47dafc7d..1954b31c70 100644
--- a/src/hip_hcc.cpp
+++ b/src/hip_hcc.cpp
@@ -137,7 +137,7 @@ ihipCtx_t * ihipGetPrimaryCtx(unsigned deviceIndex)
 };
 
 
-static thread_local ihipCtx_t *tls_defaultCtx = nullptr;  
+static thread_local ihipCtx_t *tls_defaultCtx = nullptr;
 void ihipSetTlsDefaultCtx(ihipCtx_t *ctx)
 {
     tls_defaultCtx = ctx;
@@ -512,7 +512,10 @@ int ihipStream_t::preCopyCommand(LockedAccessor_StreamCrit_t &crit, ihipSignal_t
 }
 
 
-void ihipStream_t::launchModuleKernel(hsa_signal_t signal,
+// Precursor: the stream is already locked,specifically so this routine can enqueue work into the specified av.
+void ihipStream_t::launchModuleKernel(
+                        hc::accelerator_view av,
+                        hsa_signal_t signal,
                         uint32_t blockDimX,
                         uint32_t blockDimY,
                         uint32_t blockDimZ,
@@ -526,11 +529,6 @@ void ihipStream_t::launchModuleKernel(hsa_signal_t signal,
     hsa_status_t status;
     void *kern;
 
-    // Lock stream to prevent other threads from enqueueing kernels at same time.
-    LockedAccessor_StreamCrit_t crit (_criticalData);
-
-    hc::accelerator_view av = crit->_av;
-
     hsa_amd_memory_pool_t *pool = reinterpret_cast<hsa_amd_memory_pool_t*>(av.get_hsa_kernarg_region());
     status = hsa_amd_memory_pool_allocate(*pool, kernSize, 0, &kern);
     status = hsa_amd_agents_allow_access(1, (hsa_agent_t*)av.get_hsa_agent(), 0, kern);
@@ -1344,11 +1342,11 @@ hipStream_t ihipSyncAndResolveStream(hipStream_t stream)
     }
 }
 
-void ihipPrintKernelLaunch(const char *kernelName, const grid_launch_parm *lp, const hipStream_t stream) 
+void ihipPrintKernelLaunch(const char *kernelName, const grid_launch_parm *lp, const hipStream_t stream)
 {
     std::string streamString = ToString(stream);
     fprintf(stderr, KGRN "<<hip-api: hipLaunchKernel '%s' gridDim:(%d,%d,%d) groupDim:(%d,%d,%d) groupMem:+%d %s\n" KNRM, \
-            kernelName, lp->grid_dim.x, lp->grid_dim.y, lp->grid_dim.z, lp->group_dim.x, lp->group_dim.y, lp->group_dim.z, 
+            kernelName, lp->grid_dim.x, lp->grid_dim.y, lp->grid_dim.z, lp->group_dim.x, lp->group_dim.y, lp->group_dim.z,
             lp->dynamic_group_mem_bytes, streamString.c_str());\
 }
 
diff --git a/src/hip_module.cpp b/src/hip_module.cpp
index 456d3725b8..28c65b6669 100644
--- a/src/hip_module.cpp
+++ b/src/hip_module.cpp
@@ -221,6 +221,7 @@ hipError_t hipModuleGetFunction(hipFunction_t *hfunc, hipModule_t hmod,
     return ihipModuleGetFunction(hfunc, hmod, name);
 }
 
+
 hipError_t hipModuleLaunchKernel(hipFunction_t f,
             uint32_t gridDimX, uint32_t gridDimY, uint32_t gridDimZ,
             uint32_t blockDimX, uint32_t blockDimY, uint32_t blockDimZ,
@@ -268,7 +269,7 @@ Kernel argument preparation.
 /*
   Launch AQL packet
 */
-        hStream->launchModuleKernel(signal, blockDimX, blockDimY, blockDimZ,
+        hStream->launchModuleKernel(*lp.av, signal, blockDimX, blockDimY, blockDimZ,
                   gridDimX, gridDimY, gridDimZ, sharedMemBytes, config[1], kernSize, f->kernel);
 
 /*
@@ -279,7 +280,7 @@ Kernel argument preparation.
 
 
         ihipPostLaunchKernel(hStream, lp);
-                
+
     }
 
     return ihipLogStatus(ret);