commit | 993201d3d33e5bf170084f0fc8a8ceeab507e6be | [log] [tgz] |
---|---|---|
author | Scott Todd <scotttodd@google.com> | Tue Nov 09 12:51:51 2021 -0800 |
committer | GitHub <noreply@github.com> | Tue Nov 09 12:51:51 2021 -0800 |
tree | 3469c5616d41fb570e0a204e965c75de5c313cf9 | |
parent | 36c24c9e109c50f4e94146502c094b804a0ad2be [diff] |
Expand embedded linker discovery and improve error messages. (#7575) This path didn't _quite_ work out of the box for development builds. This expands the search space to make automatic discovery more likely. The LLVM backend's embedded linker requires `lld` to be discoverable and we make an attempt to build it from source: https://github.com/google/iree/blob/222828129144c8bfb603b4647aa969afb64a5a45/iree/tools/CMakeLists.txt#L395-L401 However, the `lld` binary ends up in `iree-build\third_party\llvm-project\llvm\bin` for builds (`install` does put `lld` next to the other IREE tools). This now searches that path in the build tree and includes more information in error messages: ``` Searching environment for one of these tools: [iree-lld.exe, lld.exe, ld.lld.exe, lld-link.exe] Searching for tool at path 'D:\dev\projects\iree\..\iree-build\iree\tools' Searching for tool at path 'D:\dev\projects\iree\..\iree-build\iree\tools/../../third_party/llvm-project/llvm/bin/' (success) Found tool in build tree at path "D:\dev\projects\iree-build\third_party\llvm-project\llvm\bin\lld.exe" (failure) error: embedded linker tool (typically `lld`) not found after searching: * -iree-llvm-embedded-linker-path= flag * IREE_LLVMAOT_EMBEDDED_LINKER_PATH environment variable * common locations at relative file paths * system PATH Run with -debug-only=llvmaot-linker for search details ``` We do also have `iree-lld` now, but that lives in [llvm-external-projects/iree-compiler-api/python/](https://github.com/google/iree/blob/main/llvm-external-projects/iree-compiler-api/python/), not [iree/tools/](https://github.com/google/iree/tree/main/iree/tools/). A workaround we _don't_ want to encourage is to install `lld` directly at the system level. Using the built-from-source binary has less risk of compatibility issues.
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.