[GENX] Update GENX branch to LLVM 5e5a22c
#12112
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
This showed up when simplifying some large testcase, where the cfi directives became out of sync with the proc's they enclose. Now restricted to platforms that support .subsections_via_symbols. This reverts commit 797b68c. Fixes: #72802 Differential revision: https://reviews.llvm.org/D153167 rdar://111459507
This is part-2 change to improve codegen for vec_fabs. In this patch, v16f16 and v132f16 fabs are improved. There will be at least two followups patches after this one. 1) fixing the ISEL crash when fabs.v32f16 uses custom lowering with AVX512 2) better expansion for v16f16, v32f16 types on AVX1 subtargets.
We can use short forward branch to conditionally negate if the value is negative.
Update tests in "vector-contract-matvec-transforms.mlir" so that they
are consistent with similar tests in:
* "vector-contract-to-outerproduct-transforms.mlir".
This is to enable further refactoring in a follow-up patch, namely to:
* remove duplication (this will be much easier once consistent naming
is used),
* extend tests in "vector-contract-matvec-transforms.mlir" with cases
for scalable vectors,
* merge "vector-contract-matvec-transforms.mlir" and
"vector-contract-to-outerproduct-transforms.mlir" (there's no need
for 2 different files testing identical transformations).
Overview of changes in this patch:
1. Simplify the test by removing MemRef wrappers - this test verifies
Vector -> Vector transformations and MemRefs are not needed.
2. Use (m, k) indices instead of (i, j).
3. Rename function names.
This is part of a larger effort to improve test coverage for scalable
vectors in the Vector dialect. Implements #72834.
`atomic` is required to be followed by a special `atomic clause`, so this patch manages the parsing of that. We are representing each of the variants of the atomic construct as separate kinds, because they have distinct rules/application/etc, and this should make it easier to check rules in the future.
…ntrinisc (#70362)" This reverts commit f79676a.
… definitions and declarations (#72452) https://reviews.llvm.org/D128482 regressed certain cases of VTT emission which are no longer hidden with -fvisibility=hidden. Fix this regression by marking both declarations and definitions. Fixes [clang codegen][regression] VTT definitions missing dso_local/hidden/etc markings #72451
The bots have been running smoothly for a while. Check out https://libcxx.efcs.ca/cistats.html for more info.
Jakub Jelínek reports: As mentioned in https://gcc.gnu.org/PR112563, the new DECLARE_WRAPPER macro added in 37445e9 and ammended in 85d3873 doesn't work on SPARC/Solaris with Solaris as. While clang and GNU as when used from GCC seems to be forgiving on most architectures and allow both %function and @function (with the latter not being allowed on ARM/AArch64 I believe because @ is assembler comment start there), Solaris as doesn't allow the %function form. Fix it by using %function only for ARM. Co-developed-by: Jakub Jelínek <jakub@redhat.com> Reported-by: Jakub Jelínek <jakub@redhat.com> Closes: llvm/llvm-project#72970
These assertions can only be triggered by bugs in the algorithm's implementation; all user inputs should be handled gracefully.
OutputUnformattedBlock and InputUnformattedBlock are not used.
…2576) Build clang with the host compiler and ccache enabled in order to speed up the phase 1 builds. This helps reduce the amount of time spent running on the non-free builders.
…le reader and writer (#73026) Test fixture `MaybeSparseInstrProfTest` parameterize InstrProfWriter by whether output is sparse or not. This test fixture has 20 test cases, and 6 of them doesn't use profile reader and writer. Undo the parameterization for these test cases will reduce redundant tests. This is one clean-up PR. (A few more clean-ups to come soon, but they are not inter-dependent)
This will help users analyze whether high register usage is coming from inability of scheduler to reduce RP, or from sacrificing good RP to improve ILP.
The NonNeg flag was being Anded with the Exact flag.
Baremetal targets tend to implement their own runtime support for sanitizers. Clang driver gatekeeping of allowed sanitizer types is counter productive. This change allows anything that does not crash and burn in compilation, and leaves any potential runtime issues for the user to figure out.
This skips the build of all the unittests and llvm/clang tools, reducing the number of ninja targets from 4,826 to 3,816 in phase 1 and phase 2.
…acePrinter. (#73029) Make some methods of StackTracPrinter that will have a common implementation, non virtual.
Adds a new Implementation of StackTracePrinter that only emits symbolizer markup. Currently this change only affects Fuchsia OS. Should be NFC.
… test of value profiles (#73038) This patch factor out the common code among three similar test cases. The input data and test logic are pretty similar. Parameterize the differences (prof-weight and endianness) as advised in llvm/llvm-project#72611. - Remove duplicated tests
removed two unused methods, removed obsoleted FIXME
Remove the run lines that check for the FIR lowering. HLFIR lowering produce the same check lines.
…ebugger (#71564) Use "__this" in DataMemberRecord, make vs debugger can be parsed normally Fixes #71562
…ests (#73035) HLFIR lowering as been set by default now and FIR lowering support will be removed in the near future. This patch removes the specific FIR check lines on enter/exit data tests.
Update regex to _explicitly_ show which exp versions are added. The previous regex used `exp[^e]` to avoid matching calls like: `@llvm.experimental.stepvector`. Note: ArmPL Mappings for scalable types are not yet utilized (eg, `llvm.exp10.nxv2f64`, `llvm.exp10.nxv4f32`), as `replace-with-veclib` pass needs improvements.
…(#72526) The code in the CloneInstructionsIntoPredec... function modified by this patch has a long history that dates back to 2011, see d715ec8. There, when folding branches, all dbg.value intrinsics seen when folding would be saved and then re-inserted at the end of whatever was folded. Over the last 12 years this behaviour has been preserved. However, IMO it's bad behaviour. If we have: inst1 dbg.value1 inst2 dbg.value2 And we fold that sequence into a different block, then we would want the instructions and variable assignments to appear in the same order. However because of this old behaviour, the dbg.values are sunk, and we get: inst1 inst2 dbg.value1 dbg.value2 This clustering of dbg.values can make assignments to the same variable invisible, as well as reducing the coverage of other assignments. This patch relaxes the CloneInstructions... function and allows it to clone and update dbg.values in-place, causing them to appear in the original order in the destination block. I've added some extra dbg.values to the updated test: without the changes to the pass, the dbg.values sink into a blob ahead of the select. The RemoveDIs code can't cope with this right now so I've removed the "--try..." flag, restored in a commit to land in a couple of hours. (Metadata changes to make the LLVM-IR parser not drop the debug-info for it being out of date. The RemoveDIs related RUN line has been removed because it was spuriously passing due to the debug-info being dropped).
…n type of function. (#69724) Import of a function with `auto` return type that is expanded to a `SubstTemplateTypeParmType` could fail if the function itself is the template specialization where the parameter was replaced.
Explicitly include some headers or forward-declare types, in preparation for removing an include that pulls in many transitive headers.
If GOTSym is not defined, we cannot call `GOTSym.getBlock()`. It failed with: ``` Assertion failed: (Base->isDefined() && "Not a defined symbol"), function getBlock, file JITLink.h, line 554. ```
Manually add clobbers for various register combinations to tests. This highlights incorrectly performing shrink-wrapping, with StoreSwiftAsyncContext expansion clobbering a live register.
The new libcxx workflow are run and failing in non llvm repo. Skip it similar to other workflow.
There have been some minor but pervasive changes to the generated CHECK lines, so regenerate all of them, to minimize future diffs.
Avoid some spurious changes in a future patch.
…ction (#73311) Floating point properties are a combination of target OS, target architecture and compiler support. - Adding target OS detection, - Moving floating point type detection to its own file. This is in preparation of adding support for `_Float16` which requires testing compiler **version** and target architecture.
whitneywhtsang
force-pushed
the
merge
branch
2 times, most recently
from
12c04ad to
aa06c2e
Compare
pengtu
approved these changes
Dec 8, 2023
Signed-off-by: Tsang, Whitney <whitney.tsang@intel.com>
Signed-off-by: Tsang, Whitney <whitney.tsang@intel.com>
|
Merge when triton-dse repo is ready to move to 34f165d6633713a2c9b667927c3baaf018157ef1. |
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.
No description provided.