1 #########
2 Changelog
3 #########
5 1.2.1 [Processor Linux SDK 5.2]
6 ===============================
7 **Added**
9 #. Updated API implementation to minimize TIDL API/OpenCL dispatch overhead using multiple execution contexts in the :term:`ExecutionObject`.
10 Refer to :ref:`mnist-example` example for details.
12 #. Execution Graph generation
14 Enable a two phase approach to generating execution graphs. Use the
15 following API function to enable timestamp generation:
17 .. code::
19 bool EnableTimeStamps(const std::string& file = "timestamp.log", size_t num_frames=32);
21 The generated log file can be viewed by using the execution_graph.py script. Refer to :ref:`execution-graph` for details.
23 #. Added Python 3 bindings for TIDL API. See the ``examples/pybind`` directory for examples of using the Python bindings. Set PYTHONPATH to the location of ``tidl.so``.
25 .. code::
27 root@am57xx-evm:~# export PYTHONPATH=/home/root/tidl-api/tidl_api
28 root@am57xx-evm:~# python3
29 >>> import tidl
30 >>> help (tidl)
32 **Removed**
34 #. Configuration::enableInternalInput. Not used by the API.
36 #. Execution::GetExecutionObjects().
38 Use Execution::operator[] and Execution::GetNumExecutionObjects() instead.
39 See :ref:`examples` for usage.
41 #. The timing methods for host execution in EOPs and EOs:
43 * GetProcessTimeInMilliSeconds()
44 * GetHostProcessTimeInMilliSeconds()
46 These methods were replaced by a timestamp based approach because they were
47 no longer accurate with multiple ExecutionObject contexts and pipelining.
48 Refer to :ref:`execution-graph` for details.
50 1.1.0 [Processor Linux SDK 5.1]
51 ===============================
52 **Added**
54 #. :term:`ExecutionObjectPipeline` class to hide complexity of executing network across C66x/EVE
55 #. API methods for tracing outputs from intermediate network layers - see :ref:`network_layer_output`.
56 #. Support for updating layer group id assignment before execution - see :ref:`layer-group-override`.
57 #. Provide feedback to the user on parameter and network heap size requirements - see :ref:`sizing_device_heaps`.
60 1.0.0 [Processor Linux SDK 5.0]
61 ===============================
62 First release of the TI Deep Learning API. TIDL API brings deep learning to the edge by enabling applications to leverage TI's proprietary, highly optimized CNN/DNN implementation on the EVE and C66x DSP compute engines. TIDL will initially target Vision/2D use cases.
64 **Supported AM57x Sitara Processors**
66 * `AM5749`_ (offload to EVEs and C66x DSPs)
67 * `AM571x`_ (offload to C66x DSPs)
68 * `AM5728`_ (offload to C66x DSPs)
69 * `AM5748`_ (offload to C66x DSPs)
71 **Supported Evaluation Modules (EVMs)**
73 * `AM574x IDK EVM`_
74 * `AM572x EVM`_
75 * `AM571x IDK EVM`_
78 .. _AM572x EVM: http://www.ti.com/tool/tmdsevm572x
79 .. _AM571x IDK EVM: http://www.ti.com/tool/tmdxidk5718
80 .. _AM574x IDK EVM: http://www.ti.com/tool/tmdsidk574
81 .. _AM571x: http://www.ti.com/processors/sitara/arm-cortex-a15/am57x/products.html#p2098=1%20C66x&p809=2;2
82 .. _AM5728: http://www.ti.com/product/AM5728
83 .. _AM5748: http://www.ti.com/product/am5748
84 .. _AM5749: http://www.ti.com/product/am5749
85 .. _AM574x: http://www.ti.com/processors/sitara/arm-cortex-a15/am57x/products.html#p2098=2%20C66x&p815=ECC