haowhsu-quic

Make flatcc cross-compile deterministic

11

As mentioned in https://github.com/pytorch/executorch/pull/4297, the original flow makes host / cross build happen concurrently. This change moves host build process into cmake configuring stage and refine related dependencies. Test Plan:...

Retire deep copy of edge_program in to_backend method

3

Qualcomm lowering flow is no longer requiring deep copy if https://github.com/pytorch/executorch/pull/4297 merged. Could consider removing this.

encode delegate binaries for perf enhancement

1

This draft has some overlapped parts with https://github.com/pytorch/executorch/pull/4163 in order to demo use case. For key changes, please check files under `exir`, `runtime/executor`, `schema`.

Qualcomm AI Engine Direct - fix conv2d to meet QNN constraint

5

Summary: - fix breakage of conv2d without bias in newer QNN SDK (>=2.24) - add test cases for conv2d - refactor some keywords