Copying an MCInst isn't cheap (copies all operands) and the whole
instruction is only used for the Intel erratum mitigation, which is off
by default. In all other cases, the opcode alone suffices.

This slightly pessimizes code that uses moves to segment registers --
but that's uncommon and not performance-sensitive anyway.

As a related change, also call canPadInst() only when the result is
actually used, which is typically only the case in emitInstrEnd.

This gives a minor performance improvement.

This allows tested passes to depend on the AMX model in the function
info. Preparatory work for to adopt #94358 for other AMX passes.

Follow up of #94887.

Context:
https://github.com/llvm/llvm-project/pull/94887#pullrequestreview-2106213891

This was previously set to 4uops which was including the cost of extra register moves in the original test code.

This addresses a review comment for PR #94987 Because that PR is a big
automatic change, this change was moved in a separate one.

This is still not perfect, but an improvement in general.

Use DEFINE: %{res} = ... instead of $(cat ...). Rewrite one use of a
subshell to write to a temporary file instead.

Fixes #95036 #95033 #94933 #94930

This matches the other two functions AddOverflow and SubOverflow.

This was pointed out in PR #93940.

This patch adds APFloat type support for two FP6 data types,
E2M3 and E3M2. The definitions for the two formats are detailed
in section 5.3.2 of the OCP specification, which can be accessed here:

https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf



Signed-off-by: Durgadoss R <durgadossr@nvidia.com>

Without this gcc (9.3.0) warns with
 ../lib/Target/AArch64/GISel/AArch64CallLowering.cpp: In function 'unsigned int getCallOpcode(const llvm::MachineFunction&, bool, bool, std::optional<llvm::CallLowering::PtrAuthInfo>&, llvm::MachineRegisterInfo&)':
 ../lib/Target/AArch64/GISel/AArch64CallLowering.cpp:1025: warning: enumeral and non-enumeral type in conditional expression [-Wextra]
  1025 |       return IsIndirect ? getBLRCallOpcode(CallerF) : AArch64::BL;
       |

This commit adds a test for lea_rsp_pattern_p which was previously due
as FIXME.

This PR generates dwarf to extract the information about the arrays from
descriptor. The DWARF needs the offset of the fields like `lower_bound`
and `extent`. The getComponentOffset has been added to calculate
them which pushes the issue of host and target data size into
getDescFieldTypeModel.

As we use data layout now, some tests needed to be adjusted to have a
dummy data layout to avoid failure.

With this change in place, GDB is able show the assumed shape arrays
correctly.

  subroutine ff(n, m, arr)
    integer n, m
    integer :: arr(:, :)
    print *, arr
    do i = 1, n
      do j = 1, m
        arr(j, i) = (i * 5) + j + 10
      end do
    end do
    print *, arr
  end subroutine ff

Breakpoint 1, ff (n=4, m=3, arr=...) at test1.f90:13
13          print *, arr
(gdb) p arr
$1 = ((6, 7, 8, 9) (11, 12, 13, 14) (16, 17, 18, 19))
(gdb) ptype arr
type = integer (4,3)
(gdb) c
Continuing.
 6 7 8 9 11 12 13 14 16 17 18 19

The pass pipeline of some architecture splits register allocation phase
based on different register classes. As some analyses need to be
computed at the beginning of the register allocation and kept alive till
all values are assigned to some physical registers.

This poses challenge with objective of introducing StackSlotColoring
after partial virtual registers are assigned to physical registers, in
order to optimize stack slots usage.As this pass doesn't preserve few
analysis yet to be needed by the register allocation of the remaining
virtual registers, necessiating them to be kept preserved.

The patch includes a repro for a case where we were returning a null
`FieldDecl`
when calling `getReferencedDecls()` on the `InitListExpr` for a union.

Also, I noticed while working on this that `RecordInitListHelper` has a
bug
where it doesn't work correctly for empty unions. This patch also
includes a
repro and fix for this bug.

This avoids std::map, which is slow, and uses a StringMap. Section name,
group name, linked-to name and unique id are encoded into the key for
fast lookup.

This gives a measurable performance boost for applications that compile
many small object files (e.g., functions in JIT compilers).

---

Now also the second case works properly. That's what happens when you do
that last refactoring without re-running all tests... sorry.

This is one of the node kinds that should be considered an "original
initializer". The patch adds a test that was causing an assertion
failure in
`assert(Children.size() == 1)` without the fix.

According to RVV spec:
> In general, the requirement is to support LMUL ≥ SEWMIN/ELEN,
> where SEWMIN is the narrowest supported SEW value and ELEN is
> the widest supported SEW value.
>
> For a given supported fractional LMUL setting, implementations
> must support SEW settings between SEWMIN and LMUL * ELEN, inclusive.

We print a warning if these requirements are not met.

Reviewers: kito-cheng, asb, frasercrmck, jrtc27, michaelmaitland, lukel97

Reviewed By: lukel97

Pull Request: https://github.com/llvm/llvm-project/pull/94313

This is a follow-up to #92506.

Use clang_target_link_libraries() instead of LINK_LIBS when linking
clang libraries. This ensures that in CLANG_LINK_CLANG_DYLIB mode we
link against libclang-cpp.so (instead of linking against both it and the
static libraries).

Most places were already doing this correctly, there were just a handful
of leftovers.

It seems that we have `B` extension again:
https://github.com/riscv/riscv-b

According to the spec, `B` extension represents the collection of
the `Zba`, `Zbb`, `Zbs` extensions.

Though it hasn't been ratified, I set its version to `1.0`.

Reverts llvm/llvm-project#86609

This change causes compile-time regressions for stage2 builds
(https://llvm-compile-time-tracker.com/compare.php?from=3254f31a66263ea9647c9547f1531c3123444fcd&to=c5978f1eb5eeca8610b9dfce1fcbf1f473911cd8&stat=instructions:u).
It also introduced unintended changes to `.text` which should be
addressed before relanding.

GCC usually doesn't warn about unrecognized -Wno-<foo> options, if no
diagnostics are printed. However if some diagnostics are printed, it
also mentions that there were unrecognized -Wno-<foo> options.

Before 4feae05c, we checked for whether
-Wnested-anon-types was supported, and added the -Wno-<foo> form if the
positive form of the option was supported.

As of GCC 14, -Wnested-anon-types isn't supported, thus limit the use of
the option to actual Clang (and still only while using the GCC
compatible driver).

This avoids unnecessary mentions about unrecognized -Wno-<foo> options
when building with GCC.

Add plumbing for the numerical sanitizer on Clang's side.

… instructions

Lower global references to ptrauth constants into `@AUTH` `MCExpr`'s.
The logic is common for MachO and ELF - test both.

---------

Co-authored-by: Ahmed Bougacha <ahmed@bougacha.org>

This is a follow up patch to #94652 and handles the lowering of the
reduce intrinsic with DIM argument and non scalar result.

Add comments and a test for delay-init libraries on macOS. I originally
added the support in 954d00e8 a month
ago, but without these additional clarifications.

rdar://126885033

They were assigned from calls to find_chunk_ptr_for_size which return
size_t now.

This addresses a clang-tidy suggestion.

This patch updates the lowering of `LaunchFuncOp` in GPU to LLVM to only
legalize the operation with the converted operands, effectively removing
the lowering used by the old serialization pipeline.
It also removes all remaining uses of the old gpu serialization
infrastructure in `gpu-to-llvm`.

See [Compilation overview | 'gpu' Dialect - MLIR
docs](https://mlir.llvm.org/docs/Dialects/GPU/#compilation-overview) for
additional information on the target attributes compilation pipeline
that replaced the old serialization pipeline.

This reverts commit c007883f.

Add sanitize_numerical_stability attribute.

Relanding this PR now that
https://github.com/llvm/llvm-project/pull/90503 has merged. with `FTAN`
landing in
[TargetLoweringBase.cpp:L1021](https://github.com/llvm/llvm-project/blob/main/llvm/lib/CodeGen/TargetLoweringBase.cpp#L1020C23-L1021C63


) There is now a llvm tan intrinsic 32\64\128 Expand case for all llvm
backends.

In LLVM, the `llvm.experimental.constrained.cos` and
`llvm.experimental.constrained.sin` intrinsics are used for performing
cosine and sine calculations with additional constraints on
floating-point operations. This behavior is expected for all
floating-point math intrinsics. This change adds these constraints for
the `tan` intrinsic.

-  `Builtins.td` - replace TanF128 with F16F128MathTemplate
- `CGBuiltin.cpp` - map existing tan builtins to `tan` and
`constrained_tan` intrinsic
-   `ConstrainedOps.def` map tan and constrained_tan  to an ISDOpcode.

resolves  #91421

---------

Co-authored-by: Farzon Lotfi <farzon@farzon.com>