Printing a float from MLIR

Beginners
1

I have a simple matrix multiplication written by hand in MLIR in the memref and arith dialects:

module {
  func.func @matmul(%M: index, %N: index, %K: index, %A: memref<?x?xf32>, %B: memref<?x?xf32>, %C: memref<?x?xf32>) {
    %c0 = arith.constant 0 : index
    
    // Load A[0, 0]
    %a = memref.load %A[%c0, %c0] : memref<?x?xf32>
    
    // Load B[0, 0]
    %b = memref.load %B[%c0, %c0] : memref<?x?xf32>
    
    // Multiply A[0, 0] and B[0, 0]
    %mul = arith.mulf %a, %b : f32
    
    // Store the result in C[0, 0]
    memref.store %mul, %C[%c0, %c0] : memref<?x?xf32>
    
    return
  }

  func.func @main() {
    %c2 = arith.constant 2 : index
    
    // Allocate memory for 2x2 matrices A, B, and C
    %A = memref.alloc(%c2, %c2) : memref<?x?xf32>
    %B = memref.alloc(%c2, %c2) : memref<?x?xf32>
    %C = memref.alloc(%c2, %c2) : memref<?x?xf32>

    // Initialize matrices A and B
    %f1 = arith.constant 1.0 : f32
    %f2 = arith.constant 2.0 : f32
    %f3 = arith.constant 3.0 : f32
    %f4 = arith.constant 4.0 : f32

    %c0 = arith.constant 0 : index
    %c1 = arith.constant 1 : index

    // Initialize matrix A
    memref.store %f1, %A[%c0, %c0] : memref<?x?xf32>
    memref.store %f2, %A[%c0, %c1] : memref<?x?xf32>
    memref.store %f3, %A[%c1, %c0] : memref<?x?xf32>
    memref.store %f4, %A[%c1, %c1] : memref<?x?xf32>

    // Initialize matrix B
    memref.store %f1, %B[%c0, %c0] : memref<?x?xf32>
    memref.store %f2, %B[%c0, %c1] : memref<?x?xf32>
    memref.store %f3, %B[%c1, %c0] : memref<?x?xf32>
    memref.store %f4, %B[%c1, %c1] : memref<?x?xf32>

    // Call matmul function
    call @matmul(%c2, %c2, %c2, %A, %B, %C) : (index, index, index, memref<?x?xf32>, memref<?x?xf32>, memref<?x?xf32>) -> ()

    // Print C[0, 0] using printf
    %result = memref.load %C[%c0, %c0] : memref<?x?xf32>
    // TODO: print %result

    // Free allocated memory
    memref.dealloc %A : memref<?x?xf32>
    memref.dealloc %B : memref<?x?xf32>
    memref.dealloc %C : memref<?x?xf32>

    return
  }
}

How do I print %result?

I am executing this file by converting it to LLVM and then using Clang to execute the LLVM file.

I can see that there are multiple functions to help with printing here llvm-project/mlir/include/mlir/ExecutionEngine/CRunnerUtils.h at main · llvm/llvm-project · GitHub, but maybe they need the vector dialect, so I tried the following:

%vec = vector.broadcast %result : f32 to vector<1xf32>
vector.print %vec : vector<1xf32>

but I was not able to use --convert-vector-to-llvm to lower it to LLVM.

2

I was able to solve this by using the prinMemreftF32 function, similar to here - llvm-project/mlir/test/mlir-cpu-runner/unranked-memref.mlir at main · llvm/llvm-project · GitHub