Merge/hpce/develop #166
Draft
Merge/hpce/develop #166
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
…RWTH-Projects/clang into aurora_offloading_prototype
Conflicts: tools/sotoc/test/arrays/global_array_static.c tools/sotoc/test/arrays/static_variable_length_array_2d.c tools/sotoc/test/arrays/variable_length_array_2d.c
Fix generated target region parameters for multidimensional VLAs Closes #28 See merge request NEC-RWTH-Projects/clang!21
Previously we collected all variables declared and used in for loops and made them into target region parameters. Now we check wether these variables where declared outside the target region first.
Filter out variables which should not be params Closes #33 See merge request NEC-RWTH-Projects/clang!22
…inary) are the file permissions. This patch fixes this problem.
Previously we just assume we have 1 but until veosinfo works we just assume that we have 8 (because thats what we have here).
….de/NEC-RWTH-Projects/clang into aurora_offloading_prototype
CMake caches proved unrealiable. Running cmake separately on the runtimes seems to give a more controlled build that is less affacted by changes to the LLVM runtime/project build system. Technically: Revert "Revert "Revert "[clang][ve] expect libc++ headers installed at top-level of the Clang prefix (this happens also to be the default install target)""" This reverts commit 683f410.
VE doesn't support madvise. VE doesn't support paging, therefore VE's madvise syscall returns success always. VEOS doesn't define MADV_DONTNEED, so add definitions to InstrProfilingUtil.c.
Changes to make compiler-rt tests to pass on the VE target. These are: * No consistent denormal support on VE. We skip denormal fp inputs in builtin tests. * `madvise` unsupported on VE. * Instruction alignment requirements. Differential Revision: https://reviews.llvm.org/D113093
Before, SDNodes were tagged as pack-legalized by tracking them in a std of SDNodes. This is fragile, as SDNodes objects may be reused even in the same SelectionDAG. This new implementation wraps the VL in a 'LEGALAVL' SDNode to indicate that it refers to packs of 64bit elements. This is more robust and will be the basis for upstreaming.
Reported by @saudet on github: #24 (comment)
This reverts commit 84956a0.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This merges
hpce/developintodevelopseeking to close the separatehpce/developline.This includes sotoc, the RWTH prototype for VH->VE offloading with a source-to-source transformation on the clang AST and compilation of the generated source code with ncc.
The RV integration is factored out into the
feature/rv_integrationbranch and not part of this merge.Currently, there are issues with the OpenMP clang tests (that report mismatches in the clang version used to compile the pre-compiled test headers (PCH).