fix the device performance timing checks for clReleaseContext

The device performance timing checks were only supposed to happen
when the context reference count went to zero, but because the
reference count was only queried when call logging was enabled, it
was happening much more often.  Fix by querying (and updating) the
context reference count unconditionally.

Author: bashbaug

intercept/src/dispatch.cpp

@@ -625,9 +625,9 @@ CL_API_ENTRY cl_int CL_API_CALL CLIRN(clReleaseContext)(
     {
         GET_ENQUEUE_COUNTER();
 
-        cl_uint ref_count =
-            pIntercept->config().CallLogging ?
-            pIntercept->getRefCount( context ) : 0;
+        // Note: we use the ref count to for device performance timing checks,
+        // so get it unconditionally.
+        cl_uint ref_count = pIntercept->getRefCount( context );
         CALL_LOGGING_ENTER( "[ ref count = %d ] context = %p",
             ref_count,
             context );
@@ -639,7 +639,8 @@ CL_API_ENTRY cl_int CL_API_CALL CLIRN(clReleaseContext)(
         HOST_PERFORMANCE_TIMING_END();
         CHECK_ERROR( retVal );
         ADD_OBJECT_RELEASE( context );
-        CALL_LOGGING_EXIT( retVal, "[ ref count = %d ]", --ref_count );
+        --ref_count;
+        CALL_LOGGING_EXIT( retVal, "[ ref count = %d ]", ref_count );
         DEVICE_PERFORMANCE_TIMING_CHECK_CONDITIONAL( ref_count == 0 );
         FLUSH_CHROME_TRACE_BUFFERING_CONDITIONAL( ref_count == 0 );