tvm_mlir_learn

tvm_mlir_learn

Learning notes and experiments for deep learning compilers.

This repository collects examples around TVM, MLIR, LLVM, TorchScript, Relay, code generation, scheduling, and compiler-guided kernel optimization. It is kept as a public learning archive for AI compiler systems.

Contents

• scheduler/: TVM scheduler examples and scheduling experiments.
• dataflow_controlflow/: small examples comparing data flow and control flow concepts.
• paper_reading/: notes for compiler and ML systems papers such as PET, Ansor, and MLIR-related work.
• relay/: Relay examples, custom pass experiments, and model deployment demos.
• codegen/: TVM code generation examples based on tensor expressions and Relay IR.
• torchscript/: TorchScript usage examples.
• optimize_gemm/: GEMM optimization experiments guided by compiler ideas.
• compile_tvm_in_docker.md: TVM build notes in Docker.

Related Repositories

• CUDA and GPU optimization: https://github.com/BBuf/how-to-optim-algorithm-in-cuda
• Deep learning framework notes: https://github.com/BBuf/how-to-learn-deep-learning-framework

Status

Legacy learning archive. I may still reference this repository, but new public-facing documentation will use English entry points.