ARM DevSummit 2022
Here are some topics interesting.
That was so sad that they have shifted this conference into fully virtual, but totally understood the situation.
Adventures in Debugging
- using same interfaces to print out the error messages may be the problem since the buffer might not be emptied out from driver.
- time code is the worst, make sure unit tests cover future dates.
- this is really deep debugging including driver and hardware interface, i really see that was huge pain...
Arm for anything from Hypervisors to HPC workloads
- Introduction for Ampere Cloud Native Processor lineup.
- Ampere seems to be better for memory caching, linearly scalability and power consumption.
- Cloud service providers support Ampere cloud infrastructure recently.
- This will reduce the cost for cloud user and service provider from conservation of CPU workload more efficiently.
Building an Automotive Embedded Linux Image for Edge and Cloud Using Arm-based Graviton Instances, Yocto Project, and SOAFEE
- Introduction about SOAFEE and Yocto project.
- Amazon EC2 G5g instances, most cost-effective instances in Amazon EC2 for native automotive software development.
- Using Amazon cloud with Graviton 2, we can use the same image or binary to Edge IoT devices such as autonomous driving application.
Creating ML applications for embedded devices on Arm Virtual Hardware
- sending the data to the cloud could be expensive, this must be considered as architecture driver.
- security reason to deal with end user data or preferences, besides latency and reliability.
-
What is Arm Virtual Hardware? software model of the complete ARM IP. should be exactly same with silicon. - https://www.arm.com/products/development-tools/simulation/virtual-hardware
Developing Safe Software for Autonomous Systems
- Autonomous driving system is software, connected to IoT solutions. and it will be more.
- ISO 26262 for ADAS systems with high level of complexity and high safety integrity level.
- Essential capabilities for software compliance challenges for automated cars.
- Traceability and Transparency
- Analysis and Change Management
- Cross Discipline Collaboration
- Rapid Iterative Development
- Predictable Tool Qualification and Certification
Integration of Arm Software Test Library with the Wind River safety-certifiable Hypervisor
- Wind River Helix Virtualization Platform, that provides robust separation and isolation, fault containment.
- heterogenous systems for heterogenous projects at the same time.
- partitioning of CPU time, memory and device access via hypervisor.
- static XML easy system configuration to describe.
- Arm Software Test libraries (STL) complement Arm’s functional safety technology which supports systematic capability for ISO26262
SOAFEE-Based Software Defined Vehicle Development for Autonomous Driving
- EWAOL is the SOAFEE reference implementation, Edge Workload Abstraction and Orchestration Layer.
- Which can be deployed on AWS Graviton in the cloud and AVA developer platform provided by ADLINK.
- Autoware Open AD Kit Architecture
SOAFEE and the Heterogeneous Road to Autonomy
- Software complexity will be increasing significantly as software defined.
- Software tied to
BSP, non-standard interfaces and so on, so that application cannot be portable. - Basically SOAFEE is
Cloud Native+ Edge Architecture such as realtime, safety and so on. - VirtIO for device abstraction? https://wiki.libvirt.org/page/Virtio