[Codegen] Forward temporary DPS output copies (#24727) Destination-passing-style (DPS) operations update explicit init operands. After tensor-to-memref bufferization preserves tensor value semantics, an update of a dispatch output may be represented by a private temporary: %temp = memref.alloc() memref.copy %source, %temp linalg.generic ... outs(%temp) memref.copy %temp, %target Forward the final target through this chain: memref.copy %source, %target linalg.generic ... outs(%target) This removes the private allocation and copy-out. The copy-in is redirected to %target rather than removed because the DPS operation may still read its init operand. Reject the rewrite when the copy source or another DPS operand may alias the final target, or when an intervening operation may access that target. These checks prevent moving target writes ahead of reads that observed its previous contents. MLIR alias analysis handles ordinary local values but does not distinguish all IREE storage roots. Refine it with dispatch ABI facts: different HAL interface binding ordinals and different memref global symbols are disjoint; subspans of the same binding may overlap because their byte offsets can be dynamic; and local allocations are disjoint from bindings and globals. Unknown roots remain conservative. Run the pass after bufferization has made the allocation and copies explicit and before Linalg/LinalgExt lowering removes the DPS init operands needed to retarget the operation. --------- Signed-off-by: Yuwei Sun <yuweis2@illinois.edu> Signed-off-by: Hao Ren <rhao8608@gmail.com> Co-authored-by: Yuwei Sun <yuweis2@illinois.edu>
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.
Releases notes are published on GitHub releases.
| Package | Release status |
|---|---|
| GitHub release (stable) | |
| GitHub release (nightly) | |
iree-base-compiler | |
iree-base-runtime |
For more details on the release process, see https://iree.dev/developers/general/release-management/.
| Operating system | Build status |
|---|---|
| Linux | |
| macOS |
For the full list of workflows see https://iree.dev/developers/general/github-actions/.
See our website for more information.
Community meeting recordings: IREE YouTube channel
| Date | Title | Recording | Slides |
|---|---|---|---|
| 2025-06-10 | Data-Tiling in IREE: Achieving High Performance Through Compiler Design (AsiaLLVM) | recording | slides |
| 2025-05-17 | Introduction to GPU architecture and IREE's GPU CodeGen Pipeline | recording | slides |
| 2025-02-12 | The Long Tail of AI: SPIR-V in IREE and MLIR (Vulkanised) | recording | slides |
| 2024-10-01 | Unveiling the Inner Workings of IREE: An MLIR-Based Compiler for Diverse Hardware | recording | |
| 2021-06-09 | IREE Runtime Design Tech Talk | recording | slides |
| 2020-08-20 | IREE CodeGen (MLIR Open Design Meeting) | recording | slides |
| 2020-03-18 | Interactive HAL IR Walkthrough | recording | |
| 2020-01-31 | End-to-end MLIR Workflow in IREE (MLIR Open Design Meeting) | recording | slides |
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.