Expanded environment variable document

Author: oleksandr-pavlyk

docs/doc_sources/beginners_guides/environment_variables.rst

@@ -4,7 +4,12 @@ Environment variables
 =====================
 
 Behavior of :py:mod:`dpctl` is affected by :dpcpp_envar:`environment variables <>` that
-affect DPC++ compiler runtime. Particularly, the varible ``ONEAPI_DEVICE_SELECTOR`` can be
+affect DPC++ compiler runtime.
+
+Variable ``ONEAPI_DEVICE_SELECTOR``
+-----------------------------------
+
+The varible ``ONEAPI_DEVICE_SELECTOR`` can be
 used to limit the choice of devices available to :py:mod:`dpctl`.
 
 As such, the device returned by :py:func:`select_default_device`, as well the behavior
@@ -37,3 +42,21 @@ This script may be executed on a CPU, or GPU as follows:
     # execute on GPU device
     ONEAPI_DEVICE_SELECTOR=*:gpu python run.py
     #   Output: Approximate value of integral: 48329. running on Device(level_zero:gpu:0)
+
+
+Variable ``SYCL_CACHE_PERSISTENT``
+----------------------------------
+
+The binaries implementing :py:mod:`dpctl.tensor` created using DPC++ compiler contain sections
+with standardized intermediate forms (e.g. `SPIR-V <https://www.khronos.org/spir/>`_) that must be
+further built using SYCL device drivers for execution on the specific target hardware.
+This step is known as just-in-time compiling (JIT-ing).
+
+By default, the result of JIT-ing persists for the duration of SYCL application, i.e. for the
+duration of the Python session where :py:mod:`dpctl.tensor` is used. Setting environment variable
+``SYCL_CACHE_PERSISTENT`` to value of ``1`` instructs DPC++ runtime to save the result of JIT-ing to
+disk and reuse it in subsequent Python sessions (assuming the variable remains to be set when sessions
+are started).
+
+Setting of the environment variable ``SYCL_CACHE_PERSISTENT`` improves times of function invocations,
+but requires sufficient disk space.