qcom-sv-binaries: Release initial sv binaries recipe - #2968
Conversation
Dmitry Baryshkov (lumag)
left a comment
Dmitry Baryshkov (lumag)
left a comment
There was a problem hiding this comment.
Please fix your editor configuration, commit message has incorrect text wrapping.
Also, please mention why SV API is better than OpenCV. Also, we already provide FastCV, why do we need another API?
8fd7e56 to
34a0568
Compare
Test Results 68 files - 51 319 suites - 396 9h 26m 48s ⏱️ - 6h 40m 24s For more details on these failures, see this check. Results for commit f34e690. ± Comparison against base commit 7659417. This pull request removes 24 tests.This pull request skips 1 test.♻️ This comment has been updated with latest results. |
Dmitry Baryshkov (lumag)
left a comment
Dmitry Baryshkov (lumag)
left a comment
There was a problem hiding this comment.
The SV API complements OpenCV and FastCV. OpenCV provides
general-purpose computer-vision algorithms.
What is the plan for the further integration? Will it be usable through the OpenCV? Where is the EVA kernel driver? Is it a part of qcom-next or will it come as a separate dlkm recipe?
| install -m 0644 ${S}/usr/lib/pkgconfig/qcom-sv-binaries.pc ${D}${libdir}/pkgconfig/ | ||
| install -m 0644 ${S}/usr/share/doc/${PN}/NOTICE ${D}${datadir}/doc/${PN} | ||
| install -m 0644 ${S}/usr/share/doc/${PN}/LICENSE.qcom-2.txt ${D}${datadir}/doc/${PN} | ||
| install -m 0644 ${S}/usr/include/sv/svBuffer.h ${D}${includedir}/sv/ |
There was a problem hiding this comment.
You can use wildcards to install those.
Add a recipe for the Snapdragon Vision API runtime libraries. The SV API provides a client interface to Qualcomm EVA hardware-accelerated vision features. It enables Camera, Video, and CV applications to configure EVA features, submit buffers, communicate with the EVA driver and firmware stack, and retrieve hardware-generated outputs and metadata. The SV API complements OpenCV and FastCV. OpenCV provides general-purpose computer-vision algorithms, while FastCV provides Qualcomm-optimized CV primitives. Neither provides the complete EVA-specific hardware, buffer, metadata, and firmware integration required by EVA workloads. Signed-off-by: Sandeep Singh <sasingh@qti.qualcomm.com>
34a0568 to
f34e690
Compare
For EVA/SV overlays enablement in meta-qcom , we need overlays exception jira to be approved , for which as per current guidance we are working on RFC preparation for driver architecture alignment discussions in upstream , HFI alignment discussions meant for upstream. The SV library may be used by OpenCV in the future; however, its current purpose is to provide communication with the EVA hardware driver. |
|
What would be the point of merging a half-baked solution? Yes, the kernel driver must come along with the binaries. Also, what is the proposed way to use those libraries? If SV isn't supported as an OpenCV backend, do we want to force customers to use SV directly? That doesn't sound like a viable solution. |
3 participants
Add a recipe for the Snapdragon Vision API runtime libraries.
The SV API provides a client interface to Qualcomm EVA
hardware-accelerated vision features. It enables Camera, Video,
and CV applications to configure EVA features, submit buffers,
communicate with the EVA driver and firmware stack, and retrieve
hardware-generated outputs and metadata.
The SV API complements OpenCV and FastCV. OpenCV provides
general-purpose computer-vision algorithms, while FastCV
provides Qualcomm-optimized CV primitives. Neither provides the
complete EVA-specific hardware, buffer, metadata, and firmware
integration required by EVA workloads.