You are viewing a plain text version of this content. The canonical link for it is here.
Posted to discuss-archive@tvm.apache.org by Sergio via TVM Discuss <no...@discuss.tvm.ai> on 2020/05/12 14:46:00 UTC

[TVM Discuss] [Questions] Question about touched memory as a feature for ML cost model


I have been studying the features extracted from the low-level AST in the paper:

"Learning to Optimize Tensor Programs"

https://arxiv.org/abs/1805.08166

I came across the example below, 

![image|469x162](upload://1Nwo3tRquN16jTKyfcot5zSU0s5.png) 

but I am having trouble understanding the numbers   | x | 8 |  8 | 64 |. Since I do not see "x" indexing "A". The same applies to | y | 64 | 64 | 64 |, since I do not see "y" indexing matrix "B"

I would really appreciate it if someone could explain this to me in more detail, and in general, the concept of touched memory





---
[Visit Topic](https://discuss.tvm.ai/t/question-about-touched-memory-as-a-feature-for-ml-cost-model/6655/1) 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/f66ca3f2ef8dce5da1970d61adaec127266151cd86d7ee07058eedd2b9dacae6).