The local runtime experiment is essential to understand the cost of mixre runtime.
[x] implement the serialized version of the example workloads with/without acceleration; it is to show the need of the edge-support with HW accelerator for XR workloads.
[x] implement the workloads with mixre runtime as a pipeline: the trade-off between the cost/benefit of parallelism in the local
[x] measure the latencies/throughput of the serialized version without GPU acceleration on the client
[x] measure the latencies/throughput of the mixre version without GPU acceleration on the client
[x] Compared to the serialized version, mixre on client does not show the benefit of parallelism. I need to check the CPU usage.
[x] measure the latencies/throughput of the serialized version with GPU acceleration on the server
[x] measure the latencies/throughput of the mixre version with GPU acceleration on the server
The local runtime experiment is essential to understand the cost of mixre runtime.