Read HCC_OPT_FLUSH and optimize dispatch accordingly.

If HCC is in this mode, we can use less aggressive flushes in some
cases.

hipamd/src/hip_hcc.cpp

@@ -97,6 +97,8 @@ int HIP_COHERENT_HOST_ALLOC = 0;
 // USE_ HIP_SYNC_HOST_ALLOC
 int HIP_SYNC_HOST_ALLOC = 1;
 
+int HCC_OPT_FLUSH = 0;
+
 
 
 
@@ -1204,6 +1206,9 @@ void HipReadEnv()
 
     READ_ENV_I(release, HIP_COHERENT_HOST_ALLOC, 0, "If set, all host memory will be allocated as fine-grained system memory.  This allows threadfence_system to work but prevents host memory from being cached on GPU which may have performance impact.");
 
+
+    READ_ENV_I(release, HCC_OPT_FLUSH, 0, "Note this flag also impact HCC.  When set, use agent-scope flush rather than system-scope flush when possible.");
+
     // Some flags have both compile-time and runtime flags - generate a warning if user enables the runtime flag but the compile-time flag is disabled.
     if (HIP_DB && !COMPILE_HIP_DB) {
         fprintf (stderr, "warning: env var HIP_DB=0x%x but COMPILE_HIP_DB=0.  (perhaps enable COMPILE_HIP_DB in src code before compiling?)\n", HIP_DB);