Hi Everyone, my goal is to use the execution engine to JIT MLIR kernels generated from a DSL from C++. To do so, I plan to have a wrapper that takes in the raw pointers and convert them to tensors and finally call a kernel. For instance func @wrap_8(%a: f32, %x: memref<8xf32>, %y: memref<8xf32>, %…

Memref.tensor_store vs bufferization.to_memref + memref.copy

monellz April 20, 2022, 5:16pm 2

I manually write a pass to remove it…
I also wonder about a better solution

Topic		Replies	Views
[MLIR] lower bufferization.to_memref to LLVM MLIR llvm , mlir	1	367	May 26, 2024
Lowering ::mlir::memref::TensorStoreOp to LLVM IR MLIR llvm	4	634	June 8, 2023
How to bufferize `memref.tensor_store` op? MLIR	0	246	February 7, 2023
Tensor to memref conversion (a.k.a. bufferize) question MLIR	17	3207	November 10, 2020
[mlir][One-Shot Bufferizer] Customizable tensor <-> memref conversions MLIR	4	511	February 13, 2024