Juneyoung Lee

Well, I couldn't solve the problem :/ 2019년 1월 18일 (금) 오후 4:06에 Tony Tong 님이 작성: > how did you solve it , dude, i come to the same...

A related issue: from1DIdx and to1DIdx are using urem/udiv, but this seems fishy

Related: LLVM's pointer can have a negative offset from the address of an allocation But it is not clear whether this should be supported in MemRef.

``` LinAlg is probably the right place to begin with. The very first step could be formalizing LingAlg dialect esp. structured ones linalg-fusion-for-tensor-ops could be an initial example that we...

@makesource will look into either linalg.subview+load or tensor_load + extract to load. This will require encoding memref type, which is a memory buffer that is either read-only or not.

We can support a few bufferize transformations as well. They are not changing function signatures. - `-linalg-bufferize ` (https://github.com/llvm/llvm-project/blob/bc8acf2ce8ad6e8c9b1d97b2e02d3f4ad26e1d9d/mlir/test/Dialect/Linalg/bufferize.mlir#L1) - `-tensor-bufferize` (https://github.com/llvm/llvm-project/blob/bc8acf2ce8ad6e8c9b1d97b2e02d3f4ad26e1d9d/mlir/test/Dialect/Tensor/bufferize.mlir#L1) - `-tensor-constant-bufferize` (https://github.com/llvm/llvm-project/blob/bc8acf2ce8ad6e8c9b1d97b2e02d3f4ad26e1d9d/mlir/test/Dialect/Standard/tensor-constant-bufferize.mlir#L1)

To match loops doing reduction, we can use `matchReduction` (https://reviews.llvm.org/D110303)

Tiling: `-linalg-tile`

TOSA to LinAlg: `-tosa-to-linalg` (updated)

Buffer opts: - buffer-deallocation - buffer-hoisting - buffer-loop-hoisting - promote-buffers-to-stack - buffer-results-to-out-params