A follow-up to D136133. It was mentioned in #58382 that there is a need
to test for DRs that have not been officially resolved yet. This patch
aims to replace original "hackery" with proper handling for such cases.
Highlights:
- Availability can be suffixed (further) with "open", "drafting", or
"review", e.g. // dr2565: 16 open, // drXXXX: 16 c++17 drafting
- Checks are implemented to ensure that this suffix corresponds to
actual issue status
- Non-resolved DRs are counted (stdout of make_cxx_dr_status)
- No changes made to cxx_dr_status.html
- 'c++20' availability suffix added
- Remove 'concurrency' status since it's no longer
on the list of statuses in CWG Active Issues
Differential Revision: https://reviews.llvm.org/D138901
When we sort entries for attempting to reorder scalars, need to use
actual vectorization factor, not the number of scalars. Otherwise the
compiler crashes, if the scalars has to be reordered.
Differential Revision: https://reviews.llvm.org/D138819
Relax linalg elementwise fusion check to allow mixed consumers. Producer is still required to be fully tensor to avoid potential memref aliasing.
Differential Revision: https://reviews.llvm.org/D138759
A follow-up to D138852. Apparently cxx_dr_status.html was changed
manually there, since make_cxx_dr_status script doesn't generate the
same HTML after that patch landed.
Differential Revision: https://reviews.llvm.org/D138895
This extension only adds builtin functions and thus doesn't need to be
included as an extension. Instead of a pragma, the builtin functions
of the extension can be exposed through enabling preprocessor defines.
This change is focussed on simplifying `Support/Host.h` to only do
target detection. In this case, this function is close in usage to
existing functions in `Support/Threading.h`, so I moved it into there.
The function is also renamed to `llvm::get_physical_cores()` to match
the style of threading's functions.
The big change here is that now if you have threading disabled,
`llvm::get_physical_cores()` will return -1, as if it had not been able
to work out the right info. This is due to how Threading.cpp includes
OS-specific code/headers. This seems ok, as if threading is disabled,
LLVM should not need to know the number of physical cores.
Differential Revision: https://reviews.llvm.org/D137836
When generating __clang_call_terminate use SetLLVMFunctionAttributes
to set the default function attributes, like we do for all the other
functions generated by clang. This fixes a problem where target
features from the command line weren't being applied to this function.
Differential Revision: https://reviews.llvm.org/D138679
As noticed in GH58962, we should only diagnose illegal overloads of member functions
when the ref qualifiers don't match if the trailing constraints are the same.
The fix is to move the existing constraints check earlier in Sema::IsOverload.
Closes https://github.com/llvm/llvm-project/issues/58962
Reviewed By: erichkeane
Differential Revision: https://reviews.llvm.org/D138749
As discussed on Issue #59217, under certain circumstances the DAG can generate duplicate MUL and MUL_LOHI nodes, often during MULO legalization.
This patch attempts to replace MUL nodes with additional uses of the LO result from the MUL_LOHI node
Differential Revision: https://reviews.llvm.org/D138790
To generate code compatible to streaming mode:
- enable custome lowering for TruncStore to avoid crashing
during legalizing TruncStore for non Integer vector.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D138720
The libc++ data formatter for `std::shared_ptr` allows any namespace, but the test asserts that it must be the default `__1` namespace. Relax the regex to allow anything that looks like `__.*` (although we use `__[^:]*` so we don't match arbitrarily long text).
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D129898
Should cover most of the tests for GVN, GVNHoist, GVNSink, GlobalOpt,
GlobalSplit, InstCombine, Reassociate, SROA and TailCallElim that
had not been updated earlier.
The layout is essentially just reversed from the stable std::string layout.
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D138850
Change how the binding tables are retrived. Use the newly lowered
fir.dispatch_table operations instead of the fir.global type infos.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D138903
As strange as it seems to our files-are-strings view of the world, some editors
that treat files as arrays of lines can get confused about whether the last line
has a newline or not.
The consequences of failing to handle a bad incremental update are catastrophic.
If an update would be valid except for a missing newline at end of file, pretend
one exists.
This fixes problems still present in neovim where deleting all text often leads
to a desync shortly afterwards: https://github.com/neovim/neovim/issues/17085
Differential Revision: https://reviews.llvm.org/D135508
Some headers in LLDB work only when considered as textual inclusion, but not if one attempts to use them on their own or with a different context.
- python-typemaps.h: uses Python definitions without using "Python.h".
- RISCVCInstructions.h uses RISC-V register enums without including the enums header.
- RISCVInstructions.h includes EmulateInstructionRISCV.h, but is unnecessary since we forward-declare EmulateInstructionRISCV anyway. Including the header is problematic because EmulateInstructionRISCV.h uses DecodeResult which isn't defined until later in RISCVInstructions.h.
This makes LLDB build cleanly with the "parse_headers" feature [1]. I'm not sure what the analagous CMake option is.
[1] I didn't find public documentation but @MaskRay wrote this up: https://maskray.me/blog/2022-09-25-layering-check-with-clang#parse_headers
Reviewed By: labath, MaskRay
Differential Revision: https://reviews.llvm.org/D138310
The revision makes additional integer debug info parameters such as
line or column numbers optional (e.g. the line and column numbers 0 are
elided when printing, which is in line with the LLVM implementation).
Depends on D138405
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D138437
The zero-as-null-pointer-constant check should not fire if it is inside
a defaulted function, e.g. defaulted spaceship operators.
Add C++20 tests with spaceship operators.
Fixes#50221
Differential Revision: https://reviews.llvm.org/D138727
This patch updates the load insert point of the merged load in AggressiveInstCombine().
This is done to handle the reported test breaks by handling Alias Analysis correctly.
Differential Revision: https://reviews.llvm.org/D137201
After accepted in Kona, update the code to accept static operator[] as well.
No big code changes: accept this operator as static in SemaDeclCXX, update AST call generation in SemaOverload and update feature macros + tests accordingly.
Reviewed By: cor3ntin, erichkeane, #clang-language-wg
Differential Revision: https://reviews.llvm.org/D138387
Currently, the import of LLVMIR fails if the program contains debug
intrinsics. The revision adds support to import debug intrinsics that
have no debug expression attached and drops all debug intrinsics with a
non-empty debug expression. It also moves the existing debug intrinsics
into the "intr" namespace by deriving from LLVM_IntrOp.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D138405
Previously, we just used the platform name. This worked mostly OK, but
it required adding special handling for any unusual (and potentially
downstream) platform plugins, as evidenced by the hardcoding of the
qemu-user platform.
The current implementation was added in
D121605/21c5bb0a636c23ec75b13681c0a6fdb03ecd9c0d, which this essentially
reverts and goes back to the previous method of retrieving the platform
name from the platform triple (the "OS" field).
The motivation for D121605 was the ability to retrieve the process
without constructing an SBDebugger object (which would be necessary in a
world where SBPlatforms are managed by SBDebuggers). However, this world
did not arrive (mainly due to other commitments on my part), and I now
think that if we do want to go in that direction, that we should just
create a dummy/empty SBDebugger object for holding the initial
SBPlatform.
One benefit of D121605 was the unification of getPlatform and
getHostPlatform code paths, and I preserve that benefit by unifying them
in the other direction -- using the host SBPlatform for getHostPlatform.
Differential Revision: https://reviews.llvm.org/D138430
Fixes#59132 by mitigating the behavior of not setting underflow flag on RISC-V platform.
Reviewed By: vzakhari
Differential Revision: https://reviews.llvm.org/D138503
In revision B.q and before of the Armv8-M architecture reference
manual, the vector/scalar forms of the `vmla` and `vmlas` instructions
came in signed and unsigned integer forms, such as `vmla.s8 q0,q1,r2`
or `vmlas.u32 q3,q4,r5`.
Revision B.r has changed this. There are no longer signed and unsigned
versions of these instructions, since they were functionally identical
anyway. Now there is just `vmla.i8` (or `i16` or `i32`, and similarly
for `vmlas`). Bit 28 of the instruction encoding, which was previously
0 for signed or 1 for unsigned, is now expected to be 0 always.
This change updates LLVM to the new version of the architecture. The
obsoleted encodings for unsigned integers are now decoding errors, and
only the still-valid encoding is ever emitted. This shouldn't break
any existing assembly code, because the old signed and unsigned
versions of the mnemonic are still accepted by the assembler (which is
standard practice anyway for all signedness-agnostic MVE integer
instructions).
Reviewed By: dmgreen, lenary
Differential Revision: https://reviews.llvm.org/D138827
This variable is derived from LLVM_DEFAULT_TARGET_TRIPLE by default,
but using a separate variable allows additional normalization to be
performed if needed.
Differential Revision: https://reviews.llvm.org/D137451
Enables transposed gpu.subgroup_mma_load_matrix and updates the lowerings in Vector to GPU and GPU to SPIRV. Needed to enable B transpose matmuls lowering to wmma ops.
Taken over from author: stanley-nod <stanley@nod-labs.com>
Reviewed By: ThomasRaoux, antiagainst
Differential Revision: https://reviews.llvm.org/D138770
Make promote single iteration optional on affine-loop-normalize:
introduce a command-line flag and an argument on the utility. Disable it
by default since such a promotion isn't normally expected with loop
normalization: it could drop certain structure or information on the
loops that a user wanted to preserve.
Reviewed By: dcaballe
Differential Revision: https://reviews.llvm.org/D137605
Rename DEBUG_TYPE for LoopUtils -> loop-utils to be consistent with the
names used for the other files. Fix clang-tidy warnings on argument name
mismatches.
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D138752
`ShuffleBlockStrategy` will shuffle the instructions in a basic block without breaking the dependency of instructions.
It is implemented as a topological sort, only we randomly select instructions with no dependency.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D138339
Vlad Serebrennikov