Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix/master #612

Merged
merged 38 commits into from Aug 5, 2020
Merged

Fix/master #612

merged 38 commits into from Aug 5, 2020

Conversation

sklam
Copy link
Member

@sklam sklam commented Aug 5, 2020

Accidentally squash merged the last 4 PRs and lost their commit history. This PR undo the squash merges and re-merges the PRs.

esc and others added 30 commits June 16, 2020 09:50
The `shutil` import is no longer needed.
In Python we do prefer indentation to be done with 4 spaces.
LLVM 10 removes llvm::make_unique in favor of std::make_unique.
However, this requires C++14 and is therefore unsuitable for LLVM 9
that forces -std=c++11.  Update the code to use both conditionally.
This fixes all issues with LLVM 10.
This is needed as the llvmlite package we are building needs to
differentiate itself somehow from the 'other' dev builds.
Fixes numba#601: add the missing GPU hosts to the wheel build for llvmdev.
With LLVM 10 the following CMake line no longer appears to work.

```
llvm_map_components_to_libnames(llvm_libs all)
```

... and this commit is a suitable workaround.

Current working hypothesis is that this might have been caused by the
change in:

llvm/llvm-project@ab41180#diff-cebfde545aa260d20ca6b0cdc3da08daR270

However the docs at https://llvm.org/docs/CMake.html are not very clear
about this. And in addition the output of `llvm-config --components`
does list `all` as an option.

Many thanks to @angloyna for debugging this!!
The next llvmlite release will have a split LLVM dependency. The
`aarch64` target will use LLVM 9 and all other targets will use LLVM 10.
This reverts commit 3f66129.

Once numba/numba#6030 is merged, it will no
longer be necessary to disable the autoupgrade of atomic intrinsics for
NVPTX, because LLVM from llvmlite will not be used to optimize the IR
before sending it to NVVM.
The page at:

https://releases.llvm.org/10.0.1/docs/

Currently (Tue Aug  4 10:48:42 2020 GMT+2) returns a 404 - page not
found and this breaks our build as Sphinx can't setup the intersphinx
links.
As we are doing a split release for 0.34.0 and `aarch64` will need LLVM
9.0.* -- we allow to build with this version again.
Fix missing word and missing punctuation.

Co-authored-by: stuartarchibald <stuartarchibald@users.noreply.github.com>
Copy link
Member

@esc esc left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Errare humanum est. And kudos for the octomerge! :octocat:

@sklam sklam merged commit 468c16c into numba:master Aug 5, 2020
@sklam sklam deleted the fix/master branch August 5, 2020 20:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

4 participants