Age | Commit message (Collapse) | Author |
|
Fix importing indexed models, the inclusion of header files was not
working.
Change-Id: Idab7e002abc31f57acc4eda6a2f1cf2defdfd5c4
|
|
If an enqueued inference is cancelled it is simply removed from the
queue. Both a cancel inference response with status done and a
inference response with status cancelled are sent back.
Also override the new operators to call in the FreeRTOS allocator
instead of malloc/free.
Change-Id: I243e678aa6b996084c9b9be1d1b00ffcecc75bc9
|
|
Infransctrure work to being able to send back and forth requests and
responses for inference cancellations. The implementation is stubbed
to always fail in this commit.
Change-Id: Id7848930b81a3f38fe6c05323ccca2edd6b5b5bf
|
|
Remove output queue and associated blocking calls. Instead write
directly in the mailbox ring buffer.
If inference queue is full do not block, instead bail out immeditely
and send back a message with rejected status.
Change-Id: Id08a39792791fe383f8c01bf28b07a293a49e9b0
|
|
Change-Id: I2753434badec7c5af2c19a2b32e5e808131ba519
|
|
Change-Id: I1ae0b1fa6d3c559afd10772778c16002ae93cc63
|
|
Add message for fetching meta data about built in network models.
Change-Id: I757094c20848d4cb018db68b0455297bb03be463
|
|
Support building a network model into the firmware binary. The model
can be placed by a scatter file or linker script in for example a
low lateny high bandwidth memory like SRAM.
Change-Id: Ic742abed65e20f0da4ded7adefb039389b68b767
|
|
Update sample applications to use an external context for the PMU
configuration. The external context stored in the InferenceJob will
be set as TFLu external context and will be returned in the
ethosu_inference_begin() and ethosu_inference_end() callbacks.
Change-Id: Ief1f0943e322c2b50e8b964017af59161f67de6b
|
|
This is an example application for communication with Ethos-U kernel driver
Change-Id: I1d602682cbfdf26a2b082a56ee6c38e93a49ab8d
|