Test TVM-based shift/add operators on computed shapes that appear in PVT

Please follow TVM documentation to compile. "set(USE_CUDA ON)" is necessary

Since we need to auto-tuning module, Python dependencies below is needed.

pip install tornado psutil 'xgboost>=1.1.0' cloudpickle

Please move to the corresponding folder first.

cd ./OPs_Speedups/Shift

1. build op and tune:

python Ansor_tune.py

2. build op and test latency

python latency_test.py

Please move to the corresponding folder first.

cd ./OPs_Speedups/Add

1. build op and tune:

python Ansor_tune.py

2. build op and test latency

python latency_test.py