Remove tophub from being the default configuration when compiling

[tkonolige]

Prompted by https://github.com/apache/tvm/pull/10474, this RFC proposes removing tophub from the default compilation workflow. If you rely on tophub being the implicit default, please comment on this RFC.

@masahi @comaniac

[tkonolige]

I ran with and without tophub on a selection of models in tvm.relay.testing:

resnet
With Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.5029       0.1609       1.1974       0.1431       0.4449

Without Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.0885       0.0842       0.1020       0.0823       0.0072

densenet
With Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   1.7610       1.9728       1.9733       0.9131       0.4239

Without Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   4.3721       5.5795       5.5805       0.5085       1.9678

inception_v3
With Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   4.5442       6.2537       6.2550       0.3917       2.3228

Without Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
  43.6982      64.7263      64.9195       0.3586      26.9297

mlp
With Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.0184       0.0185       0.0197       0.0172       0.0010

Without Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.0154       0.0152       0.0167       0.0148       0.0007

mobilenet
With Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.0872       0.0832       0.1042       0.0816       0.0086

Without Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.0844       0.0840       0.0896       0.0816       0.0028

squeezenet
With Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.1058       0.1051       0.1103       0.1033       0.0024

Without Tophub
Execution time summary:
 mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
   0.1056       0.1051       0.1084       0.1040       0.0016

[comaniac]

I ran with and without tophub on a selection of models in tvm.relay.testing:

Thanks for the experiments. What's your target device?

[tkonolige]

@comaniac rtx-3070