| commit | 14927d15c19710bcdd4d630e62b21428424d6ef6 | [log] [tgz] |
|---|---|---|
| author | Ben Vanik <ben.vanik@gmail.com> | Mon Feb 12 13:14:58 2024 -0800 |
| committer | GitHub <noreply@github.com> | Mon Feb 12 13:14:58 2024 -0800 |
| tree | 83626330dc41b2f953ce60fea54d5ba3627c300d | |
| parent | 97224402dae8d80a903813dcbd135cc1ade1d965 [diff] | |
| parent | 30901f584c9e1cd03329f6265627e4610c98fcb2 [diff] |
Replacing the ancient vm_util with function_io/function_util. (#16351) The new code is written using iree_io_stream_t instead of directly using files so that we have a consistent I/O interface. This also will allow us to use the flag handling code on platforms without stdio or from tools where we may want to avoid hitting the filesystem. The old `iree/base/internal/file_io.h` utilities aren't dead yet as they're currently the only place we support memory mapping files. Future changes will add a platform-specific file-backed stream type that allows for mapping so that we can change/enhance all APIs dealing with files to support mapping and bare metal execution without the need for the giant IREE_FILE_IO_ENABLE flag. The existing flag parsing is mostly compatible modulo changing numpy multi-array expansion to the explicit form of `--input=*foo.npy` to indicate that all remaining arrays in a file should be inserted. The parsing now supports consuming multiple inputs from a single binary file or singular arrays from a numpy file by using the same syntax as `--output=` has supported with `@` indicating start of file and `+` indicating continuation. To support all of this two new stream types were added: `iree_io_stdio_stream_t` using stdio to read/write files and `iree_io_vec_stream_t` using a growable vector of blocks (ala iovec) that allocates new blocks of memory as the stream is written.
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.