This repository was archived by the owner on Jan 25, 2023. It is now read-only.
Update master with upstream #1
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…nto fix-for-3789
With this change, `-opt=0` is now correctly passed to NVVM when `opt=0` is in the `options` dict passed to `CompilationUnit.compile()`.
This commit moves the creation of the `CompileUnit` for `NvvmDIBuilder` to the moment before it is added to the module. At this point, the list of subprograms is populated, so the compile unit correctly refers to all the subprograms, instead of referring to empty metadata, which happened when the compile unit was created prior to any of the subprograms. The behaviour differs between the `DIBuilder` used for the CPU target, so a new method, `initialize()`, is added to the `AbstractDIBuilder` to facilitate this change without breaking the `DIBuilder`.
This enables / disables NVVM optimizations on the per-kernel basis.
… binop to lhs of the binop.
…eductions in prange are not interpreted as inplace binops. Unify the calltypes of the binop creates in the inplace_binop parfor body.
- remove unnecessary unpickling of OmittedArg.
# Conflicts: # numba/core/serialize.py
Co-authored-by: Siu Kwan Lam <[email protected]>
Co-authored-by: stuartarchibald <[email protected]>
Co-authored-by: Siu Kwan Lam <[email protected]>
Reduce objmode and unpickling overhead
Add prefer_literal option to overload API
…_containers Fix overspecialized containers
Remove f-strings in setup.py
Fixes invalid C prototype in helper function.
As title, workaround for NumPy ABI compatibility issue. Closes #6041
LLVM 9 onwards auto-upgrade the atomic intrinsics into an atomicrmw form unsupported by NVVM. This commit prevents the auto-upgrade by changing the "nvvm" component of the intrinsic to "numba_nvvm" until after it has been parsed and dumped by LLVM. With thanks to Stuart Archibald for the suggestion and initial patch, and Val Haenel for catching the issue in the buildfarm prior to release.
As title. On windows relpath does not work across drives as they technically do not share the same root. A utility function is added to provide a "safe" relpath like function for use in future. Previous fix was #5319.
As title.
As title.
As title, the ctor initial value was already an initial value, it didn't need the getattr on it. Fixes #6082
As title.
As title.
…r-svml Fix SVML tests with LLVM 10 and AVX512
Duplicate NumPy's PyArray_DescrCheck macro
CUDA: Prevent auto-upgrade of atomic intrinsics
Fix issue with cross drive use and relpath.
Fix bug in initial value unify.
Update CHANGE_LOG for 0.51.0
reazulhoque
pushed a commit
that referenced
this pull request
Mar 30, 2021
More constant inference for 6822.
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.