I failed to compile simple add to GPU executable code because ‘gpu-to-cubin’ pipeline was not found.
the commit id is 00a12585933ef63ff1204bf5cd265f0071d04642
is there anyone who knows about this?
> mlir-opt add.mlir \
-gpu-kernel-outlining \
-pass-pipeline='gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin)' \
-gpu-to-llvm
-o add.opt.mlir
MLIR Textual PassPipeline Parser:1:48: error: 'gpu-to-cubin' does not refer to a registered pass or pass pipeline
gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin ^
add.mlir
module {
func.func @add(%a: memref<?xf32>, %b: memref<?xf32>, %c: memref<?xf32>) {
affine.for %arg3 = 0 to 256 {
%0 = affine.load %a[%arg3] : memref<?xf32>
%1 = affine.load %b[%arg3] : memref<?xf32>
%2 = arith.addf %0, %1 : f32
affine.store %2, %c[%arg3] : memref<?xf32>
}
return
}
}