Target parameter for i.MX8M

Hi, currently I am doing the tutorial https://docs.tvm.ai/tutorials/frontend/from_darknet.html on the i.MX8M and I am not sure how to choose the right parameter for target and target_host in tvm.relay.build. Currently I have

target = tvm.target.create('llvm -device=arm_cpu -target=aarch64-linux-gnu')

but I get still following messages:

Cannot find config for target=llvm -device=arm_cpu -target=aarch64-linux-gnu, workload=('conv2d', (1, 1024, 13, 13, 'float32'), (512, 1024, 1, 1, 'float32'), (1, 1), (0, 0, 0, 0), (1, 1), 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.
Cannot find config for target=llvm -device=arm_cpu -target=aarch64-linux-gnu, workload=('conv2d', (1, 512, 13, 13, 'float32'), (1024, 512, 3, 3, 'float32'), (1, 1), (1, 1, 1, 1), (1, 1), 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.

Is there any specific parameter for i.MX8M? Thanks in advance

You should tune it (https://docs.tvm.ai/tutorials/autotvm/tune_relay_arm.html) . Besides you should add -mcpu=cortex-a53 -mattr=+neon to target. I have experience of iMX8M that if we don’t add -mcpu, we couldn’t get the best performance.

Thanks for the answer. In my case tuning seems to be a bit overhead because I only need it for a small demo but would like to know more about the right parameter for i.MX8M. Are there no pre defined configs for i.MX8M? I have now:

Cannot find config for target=llvm -device=arm_cpu -target=aarch64-linux-gnu -mcpu=cortex-a53 -mattr=+neon, workload=('conv2d', (1, 384, 52, 52, 'float32'), (128, 384, 1, 1, 'float32'), (1, 1), (0, 0, 0, 0), (1, 1), 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.

Which parameter do you have normally for target?