MLIR

[ytgui]

https://github.com/tensorflow/mlir

https://github.com/bosskwei/temp/issues/29 https://github.com/bosskwei/temp/issues/31 https://github.com/bosskwei/temp/issues/35

2020.01 now mlir become part of llvm, reopen this issue

[ytgui]

Overview

• "Multi-Level IR" or "Multi-dimensional Loop IR" or "Machine Learning IR"
• similarities to traditional three-address SSA representations (like LLVM IR or SIL)
• but polyhedral loop optimization as first-class concepts
• generate target-specific code for high performance data parallel systems

  
  // function define
  func @mul(%A: tensor<100x?xf32>, %B: tensor<?x50xf32>) -> (tensor<100x50xf32>) {
  ...
  }

// Compute the inner dimension of %A using the dim operation. %n = dim %A, 1 : tensor<100x?xf32>

// Allocate addressable "buffers" and copy tensors %A and %B into them. %A_m = alloc(%n) : memref<100x?xf32> tensor_store %A to %A_m : memref<100x?xf32>

[ytgui]

No time and interest on this project, so close it.

[ytgui]

mlir.llvm.org

• MLIR project is a novel approach to building reusable and extensible compiler infrastructure.
• MLIR aims to address software fragmentation, improve compilation for heterogeneous hardware, significantly reduce the cost of building domain specific compilers, and aid in connecting existing compilers together.

arxiv.org/abs/2002.11054