Add tipps section.

docs/reference/env_variables.rst

@@ -49,7 +49,7 @@ details how to use the variables check the :doc:`Setting the number of CUs page
     * - | ``HSA_CU_MASK``
         | Sets the mask on a lower level of queue creation in the driver, 
         | this mask will also be set for queues being profiled.
-      -
+      - ``1:0-8``
 
     * - | ``ROC_GLOBAL_CU_MASK``
         | Sets the mask on queues created by the HIP or the OpenCL runtimes, 
@@ -60,6 +60,11 @@ details how to use the variables check the :doc:`Setting the number of CUs page
         | A list of device indices or UUIDs that will be exposed to applications.
       - ``0,GPU-DEADBEEFDEADBEEF``
 
+    * - | ``HIP_FORCE_QUEUE_PROFILING``
+        | Used to run the app as if it were run in rocprof. Forces command queue
+        | profiling on by default.
+      - ON, OFF
+
 Debug environment variables
 ===========================
 
@@ -152,3 +157,19 @@ next table.
         | Use remote memory for the global heap allocation
       - ``bool``
       - ``false``
+
+HIP environment variables tips
+==============================
+
+Please feel free to open PR and we will add new tips here.
+
+* The performance can be improved at Gromacs HIP backend, when the
+``ROC_ACTIVE_WAIT_TIMEOUT=0`` and ``ROC_USE_FGS_KERNARG=0`` environment
+  variables are set.
+* ``HSA_OVERRIDE_GFX_VERSION`` can be used for unsupported GPUs looks like
+  supported GPUS. `HSA_OVERRIDE_GFX_VERSION example. <https://github.com/ROCm/ROCm/issues/2216>`_
+* ``ROC_GLOBAL_CU_MASK`` and ``ROCR_VISIBLE_DEVICES`` can be used to see how a
+  kernel scale with differen CU numbers. If a kernel compute bounded, it scales
+  lineary.
+* ``AMD_LOG_LEVEL=3``, ``AMD_SERIALIZE_KERNEL=3`` and ``AMD_SERIALIZE_COPY=3``
+  can help to debug (without debugger) and see which kernel fails.