Autotvm related questions: two ops needs to be auto-tuned

Questions
#1

my desktop used intel Core™ i3-7100, the network has mobilenetv2 as backbone, and also some conv2d, when doing autotvm,I have two questions: 1. just set the target = “llvm”? not like the form of “llvm -mcpu=core-avx2”？
2. tasks = autotvm.task.extract_from_program(net, target=target,
params=params, ops=?)
ops should be set as ops=(relay.op.nn.conv2d,relay.op.nn.depthwise_conv2d_nchw)?
but I can only find the op of contrib_depthwise_conv2d_nchwc in the file of relay/op/nn/nn.py

#2
  1. If your CPU supports AVX-2, you should add the cpu flag.
  2. Depthwise conv2d should be covered by relay.op.nn.conv2d.
#3

I got it. OP2TOPI = {
tvm.relay.op.nn.conv2d: [topi.nn.conv2d, topi.nn.depthwise_conv2d_nchw, …] in the file of relay_integration.py, tvm.relay.op.nn.conv2d includes topi.nn.conv2d, topi.nn.depthwise_conv2d_nchw both. there is another question:
in the tune_kernels from tune_relay_x86.py, there are code lines as follows:
for i, tsk in enumerate(tasks):
prefix = "[Task %2d/%2d] " % (i+1, len(tasks))

converting conv2d tasks to conv2d_NCHWc tasks

op_name = tsk.workload[0]
if op_name == ‘conv2d’:
func_create = ‘topi_x86_conv2d_NCHWc’
elif op_name == ‘depthwise_conv2d_nchw’:
func_create = ‘topi_x86_depthwise_conv2d_NCHWc_from_nchw’
else:
raise ValueError(“Tuning {} is not supported on x86”.format(op_name))

else means till now, there are only two ops(conv2d, depthwise_conv2d_nchw) are supported auto-tunning for x86 CPU?

#4

is there a plan on support of Depthwise conv2d on intel graphics under the topi/python/topi/intel_graphics? thanks a lot!