Thank you for your reply.

[quote="haichen, post:10, topic:6161"]
The strategy to select implementations for `conv2d` op is defined at [here 
](https://github.com/apache/incubator-tvm/blob/master/python/tvm/relay/op/strategy/cuda.py#L91-L198)
[/quote]
Which function in AutoTVM will use this strategy?

[quote="puddingfjz, post:9, topic:6161"]
Also, are there any examples or tutorials on how to use these templates?
[/quote]
I want to ask how to use one template which is implemented by TVM to autotune 
an operator, and someone already helped me on this, but I [still have a 
question](https://discuss.tvm.ai/t/how-to-use-autotvm-with-manually-created-topi-computations/4895/12?u=puddingfjz),
 and maybe you can help.

[quote="puddingfjz, post:9, topic:6161"]
are there templates with more kinds of knobs?
[/quote]
By reading the documents and some code, I find that there are four functions 
which can define the ```config_space```: ```define_split, define_reorder, 
define_annotate, define_knob```. I also find that ```define_split, 
define_knob``` are the most used. Is it because ```loop tiling``` and ```loop 
unrolling``` is the most important or comman techniques to automatically 
optimize an operator? 

Thank you in advance.





---
[Visit 
Topic](https://discuss.tvm.ai/t/topi-winograd-convolution-performance-is-too-slow/6161/11)
 to respond.

You are receiving this because you enabled mailing list mode.

To unsubscribe from these emails, [click 
here](https://discuss.tvm.ai/email/unsubscribe/6ba5dd9ab417e09fc7837f247c659c135b9a55101039f68abfa4632839fcd40c).

Reply via email to