diff options
author | Ajay Jayaraj | 2018-08-30 14:25:11 -0500 |
---|---|---|
committer | Ajay Jayaraj | 2018-08-30 16:28:59 -0500 |
commit | 6b03a3b668a0d0c299b886ba3802b850bc674ebf (patch) | |
tree | 820431c0a222a453e8e16880d614de1a92c8b3f1 /docs | |
parent | 9463c750203e2015b89d53328e4e0ed832399b39 (diff) | |
download | tidl-api-6b03a3b668a0d0c299b886ba3802b850bc674ebf.tar.gz tidl-api-6b03a3b668a0d0c299b886ba3802b850bc674ebf.tar.xz tidl-api-6b03a3b668a0d0c299b886ba3802b850bc674ebf.zip |
Added example to illustrate pipelining across EOs
two_eo_per_frame is a simple example to illustrate using
ExecutionObjectPipeline to split processing a single frame across EVE
and DSP.
(MCT-1048)
Diffstat (limited to 'docs')
-rw-r--r-- | docs/source/example.rst | 50 |
1 files changed, 29 insertions, 21 deletions
diff --git a/docs/source/example.rst b/docs/source/example.rst index 1b16d01..b72d6a0 100644 --- a/docs/source/example.rst +++ b/docs/source/example.rst | |||
@@ -2,29 +2,37 @@ | |||
2 | Examples | 2 | Examples |
3 | ******** | 3 | ******** |
4 | 4 | ||
5 | +---------------------+-----------------------------------------------------+ | 5 | +---------------------+----------------------------------------------------------------+ |
6 | | Example | Description | | 6 | | Example | Description | |
7 | +---------------------+-----------------------------------------------------+ | 7 | +---------------------+----------------------------------------------------------------+ |
8 | | one_eo_per_frame | Simple example to illustrate processing a single | | 8 | | one_eo_per_frame | Simple example to illustrate processing a single | |
9 | | | frame with one :term:`EO` using the j11_v2 network. | | 9 | | | frame with one :term:`EO` using the j11_v2 network. | |
10 | | | The per-frame processing time for this network is | | 10 | | | Per-frame processing time for this network is farily similar | |
11 | | | fairly similar across EVE and C66x DSP. The example | | 11 | | | across EVE and C66x DSP. The enables frame processing to be | |
12 | | | parallelizes frame processing across all available | | 12 | | | parallelized by distributing frames across all available EVE | |
13 | | | EVE and C66x cores. | | 13 | | | and C66x cores. | |
14 | +---------------------+-----------------------------------------------------+ | 14 | +---------------------+----------------------------------------------------------------+ |
15 | | imagenet | Classification | | 15 | | two_eo_per_frame | Simple example to illustrate processing a single | |
16 | +---------------------+-----------------------------------------------------+ | 16 | | | frame with two :term:`EOs<EO>` using the j11_v2 network. | |
17 | | segmentation | Pixel level segmentation | | 17 | +---------------------+----------------------------------------------------------------+ |
18 | +---------------------+-----------------------------------------------------+ | 18 | | imagenet | Classification | |
19 | | ssd_multibox | Object detection | | 19 | +---------------------+----------------------------------------------------------------+ |
20 | +---------------------+-----------------------------------------------------+ | 20 | | segmentation | Pixel level segmentation | |
21 | | tidl_classification | Classification | | 21 | +---------------------+----------------------------------------------------------------+ |
22 | +---------------------+-----------------------------------------------------+ | 22 | | ssd_multibox | Object detection | |
23 | | test | Unit test. Tests supported networks on C66x and EVE | | 23 | +---------------------+----------------------------------------------------------------+ |
24 | +---------------------+-----------------------------------------------------+ | 24 | | tidl_classification | Classification | |
25 | +---------------------+----------------------------------------------------------------+ | ||
26 | | layer_output | Illustrates using TIDL APIs to access output buffers | | ||
27 | | | of intermediate :term:`Layer`s in the network. | | ||
28 | +---------------------+----------------------------------------------------------------+ | ||
29 | | test | Unit test. Tests supported networks on C66x and EVE | | ||
30 | +---------------------+----------------------------------------------------------------+ | ||
25 | 31 | ||
26 | The examples included in the tidl-api package demonstrate three categories of | 32 | The examples included in the tidl-api package demonstrate three categories of |
27 | deep learning networks: classification, segmentation and object detection. ``imagenet`` and ``segmentation`` can run on AM57x processors with either EVE or C66x cores. ``ssd_multibox`` requires AM57x processors with both EVE and C66x. The performance | 33 | deep learning networks: classification, segmentation and object detection. |
34 | ``imagenet`` and ``segmentation`` can run on AM57x processors with either EVE or C66x cores. | ||
35 | ``ssd_multibox`` requires AM57x processors with both EVE and C66x. The performance | ||
28 | numbers that we present here were obtained on an AM5729 EVM, which | 36 | numbers that we present here were obtained on an AM5729 EVM, which |
29 | includes 2 Arm Cortex-A15 cores running at 1.5GHz, 4 EVE cores at 535MHz, and | 37 | includes 2 Arm Cortex-A15 cores running at 1.5GHz, 4 EVE cores at 535MHz, and |
30 | 2 DSP cores at 750MHz. | 38 | 2 DSP cores at 750MHz. |