Whats the lowered loop program in TVM paper

Hi, I recently read TVM: An Automated End-to-End Optimizing Compiler for Deep Learning

However, I’m not clear about lowered loop program. Could anyone give an explanation?

For each schedule configuration, we use an ML model that takes the lowered loop program as input and predicts its running time on a given hardware back-end.

From section 5.2 ML-Based Cost Model

Thanks in advance. :star_struck:

According to the source code of xgboost cost model in Autotvm, the input are certain features extracted from lowered statements of a certain schedule. Here lower is a process of tvm which transforms a schedule to IR, maybe lowered loop program refers to the IR.

1 Like

Greatly thanks a lot.