mlir-hlo

mlir-hlo copied to clipboard

use openxla mlir_hlo location

apivovarov

Hi, I'm getting started with this project and found that readme doesn't work well. 1st - building. It would help to add step of building llvm with `-DLLVM_ENABLE_PROJECTS="clang;lld"` after checking...

marcus-nilis

Is there a pass (sequence) that can lower the `mhlo.scatter` operation to standard MLIR dialects, such as `linalg` or `tensor`? The goal is to lower to the LLVM dialect and...

dime10

Failed to import repo as a third party dependency using bazel

2

Hi there, we're currently working on a graph-compiler based on [some front graph]-mhlo-linalg-[some backend]. When we tried to use mhlo as a third party dependency in another repo as following...

yyyokata

cudnn call support

1

Is there any plan to support cudnn/cublas call for convolution、dot computation?

feiyuvl

Missing mhlo canonicalization matching algebraic simplifier for empty reduce.

jpienaar

The project is tested continuously with CMake on Kokoro, but this is invisible outside of Google right now. It would be great if this CI was part of the publicly...

joker-eph

Failure to compile broadcast_in_dim

4

The following fails to compile ```mlir module @jit_encode_batch { func.func public @main(%arg0: tensor, %arg293: tensor

jpienaar

Using StableHLO for ONNX to HLO conversion

5

In light of the presentation on OpenXLA where they touted StableHLO as the preferred representation for exchanges between compilers/tools, would it make sense to migrate our support from ONNX to...

AlexandreEichenberger

Deduplicate reduction subcomputations when converting from MHLO to HLO

1

See https://github.com/google/jax/issues/7654 We should deduplicate reducers when converting from MHLO to HLO. e.g. compare: ``` In [1]: import jax In [2]: import jax.numpy as jnp In [3]: def f(x, y):...

hawkinsp