Open main menu

DAVE Developer's Wiki β

Changes

no edit summary
==Running the application==
Verificare se quanto sotto è applicabile
In order to have reproducible and reliable results, some measures were taken:
* All the files required to run the test—the executable, the image files, etc.—are stored on a tmpfs RAM disk in order to make file system/storage medium overhead neglectable.
Before illustrating the results of the inference by running the two C++ applications, it is useful to show some information about the DPU. This can be done by using the '''dexplorerDExplorer''' status tool, which provides the DNNDK version and the DPUrunning mode configuration:  ===DExplorer===
<pre>
Channel Augmentation : Enabled
Average Pool : Enabled
</pre>
</pre>===DDump===
Furthermore, it is possible to dump some information encapsulated inside the DPU ELF file, such as the DPU Kernel name and general info and the DPU architecture info which are useful for analysis and debugging purposes. This can be done by using the '''ddumpDDump'''tool as illustrated as follows:
<pre>
</pre>
'''DNNDK coarse ===Coarse grained profiling'''using DNNDK low level API===
<pre>
6) green_apple : 3.44132e-14
_______________________________________________________________
 
</pre>
 '''DNNDK fine ===Fine grained profiling''' using DPU profile modeDNNDK low level API===
<pre>
6) green_apple : 3.44132e-14
_______________________________________________________________
 
</pre>
'''===Profiling analysis with DSight===DSight is the DNNDK performance profiling tool which is used as a visual performance analysis tool for neural network models. By running the DNNDK application with profile as the DPU running mode configuration, a .prof log file is produced. This file can be parsed and processed with DSight, obtaining an HTML web page, providing a visual format chart showing DPU cores'''utilization and scheduling efficiency.
[[File:Xilinx DSight.png|thumb|center|1000px|DESCRIPTION]]
'''vaitrace''' + ''' vaiProfiler''' per ===Profiling VARThigh-level APIs with Vitis AI Profiler===
<pre>
 
------------------------------------------------
image name : red_apple_1.jpg
tot correct : 1
tot wrong : 0
 
</pre>
dave_user
207
edits