Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.

libcpu extension not found

pkhan10
New Contributor I
1,333 Views

hello ,
as per new release the support of libcpu extension is deprecated
but as you can see the samples are still dependent on cpu_extensionScreenshot from 2020-05-09 17-45-09.png

even these are the only shared_objects i can find in my inference engine folder, and none of them is for CPU

Screenshot from 2020-05-09 17-48-59.pngis there a way using which I can still build the samples or demoes with libcpu_extension.so can you please share whole cmake command to do that.

0 Kudos
3 Replies
Munesh_Intel
Moderator
1,333 Views

Hi Prateek,

CPU extensions library was removed, and extensions were moved into the plugin since Intel® Distribution of OpenVINO™ Toolkit 2020.1.

Information about CPU Kernels Extensibility is available at ‘How to Implement Custom CPU Layers’ page.

Link: https://docs.openvinotoolkit.org/2020.2/_docs_IE_DG_Extensibility_DG_CPU_Kernel.html

 

CMake scripts to build an extension library is available at 'Build Extension Library Using CMake*' page.

Link: https://docs.openvinotoolkit.org/2020.2/extension_build.html

 

Regards,

Munesh

0 Kudos
pkhan10
New Contributor I
1,333 Views

hello munesh..
i used same script run got a .so file in build folder

when i try to run object_detection_demo_ssd_async
i get following error... 

 

RuntimeError: dlSym cannot locate method 'CreateExtension': /media/prateek/shared_space/Notebooks/abg/openvino/build_demos/cpu_extension/build/libcpu_extension.so: undefined symbol: CreateExtension
 

0 Kudos
Munesh_Intel
Moderator
1,333 Views

Hi Prateek,

Samples are not dependent on CPU extensions, as they have already been moved into CPU plugin. The CPU extensions are optional and are required for CPU custom layers only.

If you are using the object_detection_demo_ssd_async with pre-trained models from Open Model Zoo repository (as well as in the open_model_zoo folder in OpenVINO™ Toolkit), there is no need to use ‘AddExtension’, since the extensions are loaded automatically while loading the CPU plugin. This also applies to custom pre-trained models where all the primitives are supported.

For custom pre-trained models that have primitives that are not supported by CPU plugin, you can build an extension library and use the ‘AddExtension’ method of the general plugin interface to load your primitives.

 

Regards,

Munesh

0 Kudos
Reply