532a436c6dc7d23156d351f128d68e0b970476f9
1 #########
2 Changelog
3 #########
5 1.3.0 [Processor Linux SDK 5.3]
6 ===============================
7 **Added**
9 #. Added DSP support for MNIST example.
11 **Changed**
13 #. PSDK 5.3 and TIDL-API 1.3 use a slightly modified TIDL network binary
14 format to support ONNX network import. TIDL-API 1.3, as well as
15 tidl-viewer utility, can read both network formats before and after the
16 change. Network binary in TIDL-API 1.2 and earlier is of size 483364
17 bytes, network binary in TIDL-API 1.3 is of size 484384 bytes.
19 #. Improved performance of concat layer on C66x DSP.
21 1.2.2 [Processor Linux SDK 5.2]
22 ===============================
23 **Added**
25 #. Updated API implementation to minimize TIDL API/OpenCL dispatch overhead using multiple execution contexts in the :term:`ExecutionObject`.
26 Refer to :ref:`mnist-example` example for details.
28 #. Execution Graph generation
30 Enable a two phase approach to generating execution graphs. Use the
31 following API function to enable timestamp generation:
33 .. code:: cpp
35 bool EnableTimeStamps(const std::string& file = "timestamp.log", size_t num_frames=32);
37 The generated log file can be viewed by using the execution_graph.py script. Refer to :ref:`execution-graph` for details.
39 #. 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``.
41 .. code:: bash
43 root@am57xx-evm:~# export PYTHONPATH=/home/root/tidl-api/tidl_api
44 root@am57xx-evm:~# python3
45 >>> import tidl
46 >>> help (tidl)
48 **Removed**
50 #. Configuration::enableInternalInput. Not used by the API.
52 #. Execution::GetExecutionObjects().
54 Use Execution::operator[] and Execution::GetNumExecutionObjects() instead.
55 See :ref:`examples` for usage.
57 #. The timing methods for host execution in EOPs and EOs:
59 * GetProcessTimeInMilliSeconds()
60 * GetHostProcessTimeInMilliSeconds()
62 These methods were replaced by a timestamp based approach because they were
63 no longer accurate with multiple ExecutionObject contexts and pipelining.
64 Refer to :ref:`execution-graph` for details.
66 1.1.0 [Processor Linux SDK 5.1]
67 ===============================
68 **Added**
70 #. :term:`ExecutionObjectPipeline` class to hide complexity of executing network across C66x/EVE
71 #. API methods for tracing outputs from intermediate network layers - see :ref:`network_layer_output`.
72 #. Support for updating layer group id assignment before execution - see :ref:`layer-group-override`.
73 #. Provide feedback to the user on parameter and network heap size requirements - see :ref:`sizing_device_heaps`.
76 1.0.0 [Processor Linux SDK 5.0]
77 ===============================
78 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.
80 **Supported AM57x Sitara Processors**
82 * `AM5749`_ (offload to EVEs and C66x DSPs)
83 * `AM571x`_ (offload to C66x DSPs)
84 * `AM5728`_ (offload to C66x DSPs)
85 * `AM5748`_ (offload to C66x DSPs)
87 **Supported Evaluation Modules (EVMs)**
89 * `AM574x IDK EVM`_
90 * `AM572x EVM`_
91 * `AM571x IDK EVM`_
94 .. _AM572x EVM: http://www.ti.com/tool/tmdsevm572x
95 .. _AM571x IDK EVM: http://www.ti.com/tool/tmdxidk5718
96 .. _AM574x IDK EVM: http://www.ti.com/tool/tmdsidk574
97 .. _AM571x: http://www.ti.com/processors/sitara/arm-cortex-a15/am57x/products.html#p2098=1%20C66x&p809=2;2
98 .. _AM5728: http://www.ti.com/product/AM5728
99 .. _AM5748: http://www.ti.com/product/am5748
100 .. _AM5749: http://www.ti.com/product/am5749
101 .. _AM574x: http://www.ti.com/processors/sitara/arm-cortex-a15/am57x/products.html#p2098=2%20C66x&p815=ECC