This is the minimal example to look into the way to implement the hello-world kind of program with MLIR. The basic code structure is borrowed from standalone and Toy language in LLVM project.

• LLVM
• MLIR
• CMake
• Ninja

We need to build our own MLIR in the local machine in advance. Please follow the build instruction for MLIR here.

Please make sure to build LLVM project first according to the instruction.

mkdir build && cd build
cmake -G Ninja .. -DLLVM_DIR=/path/to/llvm-project/build/lib/cmake/llvm \
  -DMLIR_DIR=/path/to/llvm-project/build/lib/cmake/mlir

cmake --build . --target hello-opt

To run the test, check-hello target will be usable.

To build the documentation from the TableGen description of the dialect operations, run

cmake --build . --target mlir-doc

hello-opt will lower the MLIR into the bytecode of LLVM.

# Lower MLIR to LLVM IR
$ ./build/bin/hello-opt ./test/Hello/print.mlir > /path/to/print.ll

# Execute the code with LLVM interpreter
$ lli /path/to/print.ll 

1.000000 2.000000 3.000000
4.000000 5.000000 6.000000

constant

Constant operation turns a literal into an SSA value. The data is attached to the operation as an attribute. For example:

  %0 = "hello.constant"()
  { value = dense<[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]> : tensor<2x3xf64> }
  : () -> tensor<2x3xf64>

Interfaces: NoSideEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

print operation

Syntax:

operation ::= `hello.print` $input attr-dict `:` type($input)

The "print" builtin operation prints a given input tensor, and produces no results.

Interfaces: NoSideEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}