v1.38 Adding missing __threadfence_system() (#70)

CHANGELOG.md

@@ -3,6 +3,11 @@
 Documentation for TransferBench is available at
 [https://rocm.docs.amd.com/projects/TransferBench](https://rocm.docs.amd.com/projects/TransferBench).
 
+## v1.38
+
+### Fixes
+* Adding missing threadfence which could cause non-fine-grained Transfers to report higher speeds
+
 ## v1.37
 
 ### Changes

src/include/EnvVars.hpp

@@ -29,7 +29,7 @@ THE SOFTWARE.
 #include "Compatibility.hpp"
 #include "Kernels.hpp"
 
-#define TB_VERSION "1.37"
+#define TB_VERSION "1.38"
 
 extern char const MemTypeStr[];
 extern char const ExeTypeStr[];

src/include/Kernels.hpp

@@ -247,6 +247,7 @@ GpuReduceKernel(SubExecParam* params)
   __syncthreads();
   if (threadIdx.x == 0)
   {
+    __threadfence_system();
     p.stopCycle  = wall_clock64();
     p.startCycle = startCycle;
     p.xccId      = xccId;