Adding optional target_storage to hal.tensor.export. This allows frontends to control the storage for exported tensors by tying arguments to results. Today it's required that these do not alias but we could do fun things in the future with providing slabs to place multiple results into. The native bindings have been updated to support iree.abi.output arg attrs indicating the result an argument provides storage for. Other bindings may decide this on their own however they want. There's still inefficiencies here (a copy is almost always performed) but this allows us to avoid the allocation and eliding the copy is something that we eventually need to do anyway. Output buffers can now be passed to tools as inputs with leading `&` (indicating storage reference), ex: `--function_input=&4x8xf32`.
IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler and runtime that lowers Machine Learning (ML) models to a unified IR that scales up to meet the needs of the datacenter and down to satisfy the constraints and special considerations of mobile and edge deployments.
See our website for project details, user guides, and instructions on building from source.
IREE is still in its early phase. We have settled down on the overarching infrastructure and are actively improving various software components as well as project logistics. It is still quite far from ready for everyday use and is made available without any support at the moment. With that said, we welcome any kind of feedback on any communication channels!
See our website for more information.
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.