| commit | 7a27de5bdbf6a064af5f84c4fd82b7ef9a9476f7 | [log] [tgz] |
|---|---|---|
| author | Ben Vanik <ben.vanik@gmail.com> | Fri Oct 07 14:23:59 2022 -0700 |
| committer | Ben Vanik <ben.vanik@gmail.com> | Wed Oct 12 08:31:04 2022 -0700 |
| tree | 3b747763010d17db15a3f8622ccbb89c59571a9c | |
| parent | 888843cb272b15a532219b811bd3937b8ccba2b9 [diff] |
Enabling CPU async execution by fixing task system cmd buffer retains. When running async the task system needs to ensure all command buffers remain live until the entire submission retires. This ensures all resources used by commands (buffers/events/etc) as well as inlined command buffer data (update_buffer/etc) is valid. This also fixes a race where a retire signal would wake waiters that may try to tear down the VM context and modules prior to the retire cleanup dropping its resource references. This would lead to vm.buffer deinitialize trying to drop the rodata with live references -> an abort() call. Now we always release the resources before signaling in both success and failure cases.
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.