According to the IR verifier, "Declaration[s] may not be in a Comdat!"

This is a re-commit of 76b3f0b4 with
updates to the test:
 * Force emission of the extra-module, to trigger the bug after D138264,
   by providing a second symbol @g, and making the comdat nodeduplicate.
   (Technically only one is needed, but two should be safer.)
 * Name the comdat $f to avoid failure on Windows:
   LLVM ERROR: Associative COMDAT symbol 'c' does not exist.

Differential Revision: https://reviews.llvm.org/D142443

This patch fixes #60307 issue. The 8bb4451a introduces the possibility
to unite overlapped or adjacent address ranges to keep address ranges
in an unambiguous state. The AddressRangesMap is used to normalize
address ranges. The AddressRangesMap keeps address ranges and the value
of the relocated address. For intersected range, it creates a united
range that keeps the last inserted mapping value. The same for adjusted ranges.
While it is OK to use the last inserted mapping value for intersected ranges
(as there is no way how to resolve ambiguity) It is not OK to use the
last inserted value for adjacent address ranges. Currently, two following
address ranges are united into a single one:

{0,24,17e685c} {24,d8,55afe20} -> {0,d8,55afe20}

To avoid the problem, the AddressRangesMap should not unite adjacent address ranges
with different relocated addresses. Instead, it should leave adjacent address ranges
as separate ranges. So, the rang...

Originally the test has a bug: in the input YAML `.symtab.dyn` section has a type
repeated twice with different values (SHT_SYMTAB and SHT_NOBITS). YAML parser
took SHT_NOBITS as a type of the section, so the test wasn't checking the desired
case.

This patch changes the test to verify that a section with SHT_SYMTAB type and
SHF_ALLOC flag is not removed when --strip-all-gnu flag present.

Differential Revision: https://reviews.llvm.org/D143086

This test is equivalent to another one, where we can infer a
reasonable iteration count. The only difference is that umin and zext
here are swapped, and rewriter fails here.

Adds intrinsics for the following SME2 instructions:

* sdot (multi-indexed, 2 and 4 vectors, 32-bit and 64-bit ZA)
* udot (multi-indexed, 2 and 4 vectors, 32-bit and 64-bit ZA)
* usdot (multi-indexed, 2 and 4 vectors)
* sudot (multi-indexed, 2 and 4 vectors)
* fdot (multi-indexed, 2 and 4 vectors)
* bfdot (multi-indexed, 2 and 4 vectors)

NOTE: These intrinsics are still in development and are subject to future changes.

Differential Revision: https://reviews.llvm.org/D142533

Differential Revision: https://reviews.llvm.org/D142388

The custom DCE in cg-rewrite is meant to get rid of fir.shape, fir.shift,
fir.shape_shift and fir.slice ops as well as their unused operands
before codegen (that does not lower those abstract operation to LLVM).

However, it turned out to be flowed in case some fir.shape operands were
unused outside of fir.shape and appeared several times as operands:
they were erased at the first appearance, causing the further attemp
to erase it to segfault (since the op IR storage was deallocated).

Instead of trying to fixing the custom DCE code, use mlir::runRegionDCE.

Differential Revision: https://reviews.llvm.org/D143247

Similary to D140209, the scope might need to be retrieved
from the typeSymbol. The test code was crashing because the
scope passed to CollectBindings was initially null.

Reviewed By: jeanPerier

Differential Revision: https://reviews.llvm.org/D143188

(A & B) ^ (A | C) --> A ? ~B : C

https://alive2.llvm.org/ce/z/KCBfXr
https://alive2.llvm.org/ce/z/Pm-zJN
https://alive2.llvm.org/ce/z/VT8uC2

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D141660

failure (instead of bool flag)

Differential Revision: https://reviews.llvm.org/D143171

Failure on Windows:
LLVM ERROR: Associative COMDAT symbol 'c' does not exist.

This reverts commit 76b3f0b4 while
I investigate the problem and a solution that still triggers the
original problem.

Top-level ModuleOps cannot be transformed with the GreedyPatternRewriteDriver since D141945 because they do not have an enclosing region that could be used as a scope. Make the scope optional inside GreedyPatternRewriteDriver, so that top-level ops can be processed when they are on the initial list of ops.

Note: This does not allow users to bypass the scoping mechanism by setting `config.scope = nullptr`.

Fixes #60462.

Differential Revision: https://reviews.llvm.org/D143151

Don't include it among the mandatory options; the automatically built
tools via a nested cmake build work fine these days
(in particular, since 93010544 /
https://reviews.llvm.org/D126313).

Clarify the directory path-to-host-bin into something more verbose,
to avoid ambiguity with LLVM_HOST_TRIPLE.

Differential Revision: https://reviews.llvm.org/D142960

Setting LLVM_HOST_TRIPLE propagates the information to a few more
places than if only setting LLVM_TARGET_ARCH and
LLVM_DEFAULT_TARGET_TRIPLE, while both of those settings get their
defaults implied from LLVM_HOST_TRIPLE if they're not overridden.

Differential Revision: https://reviews.llvm.org/D142404

Differential Revision: https://reviews.llvm.org/D143244

Fix the insert point when expanding affine apply and handle cases with
symbols. Also add missing precondition to dynamic shape vectorization.

Differential Revision: https://reviews.llvm.org/D143243

This commit introduces a structured representation of loop metadata to
the LLVM dialect. This attribute explicitly models all known `!llvm.loop`
metadata fields and groups them by introducing nested attributes for each
namespace.

The new attribute replaces the LoopOptionAttr that could only model a
limited subset of loop metadata.

Reviewed By: gysit

Differential Revision: https://reviews.llvm.org/D143064

According to the IR verifier, "Declaration[s] may not be in a Comdat!"

Differential Revision: https://reviews.llvm.org/D142443

Reviewed By: RKSimon

Differential Revision: https://reviews.llvm.org/D143169

Clang implicitly adds the `[[gnu::const]]` attribute to many math
functions from the C standard library functions. Since this behavior is
Clang-specific, make sure the test only runs on Clang.

Differential Revision: https://reviews.llvm.org/D143033

This re-commits part of c79635cc. It is reverted since it contains
platform-inconsistent constant. Now the patch only contains constant
defined in DeclBase.h so it should be platform-independent. And this
should be still helpful.

Reviewed By: erichkeane

Differential Revision: https://reviews.llvm.org/D141992

Override hasAndNotCompare() to use more `ANDN` instead of using `AND`
and `NOT`.
This patch enables the following transforms:
(X & Y) == Y ---> (~X & Y) == 0
(X & Y) != Y ---> (~X & Y) != 0.

Differential Revision: https://reviews.llvm.org/D143037

`check_loc` is not used if ITT is disabled or debug is off, causing a
compiler warning.

Reviewed By: jlpeyton

Differential Revision: https://reviews.llvm.org/D143004

Summary:
  This patch introduces a mechanism to check the code object version from the module flag, This avoids checking from command line.
In case the module flag is missing, we use the current default code object version supported in the compiler.

For tools whose inputs are not IR, we may need other approach (directive, for example) to check the code
object version, That will be in a separate patch later.

For LIT tests update, we directly add module flag if there is only a single code object version associated with all checks in one file.
In cause of multiple code object version in one file, we use the "sed" method to "clone" the checks to achieve the goal.

Reviewer: arsenm

Differential Revision:
  https://reviews.llvm.org/D14313

If DAZ is enabled, a test against fcZero is not equivalent to a compare
with 0.

Use right prefix order to get merging.

Also drop -verify-machineinstrs and add -amdgpu-enable-delay-alu=0

This is NFC as it stands, since other combines will effectively
prevent this from being reachable. This will avoid regressions in a
future change which tries to make better use of select source
modifiers.

Didn't bother with the GlobalISel part for now, since the baseline
combine doesn't seem to work on the existing test.

Close https://github.com/llvm/llvm-project/issues/60275

The root cause of issue 60275 is the imbalance of
PushExpressionEvaluationContext() and PopExpressionEvaluationContext().

See
https://github.com/llvm/llvm-project/blob/f1c4f927f7c15b5efdc3589c050fd0513bf6b303/clang/lib/Parse/Parser.cpp#L1396-L1437

We will PushExpressionEvaluationContext() in ActOnStartOfFunctionDef()
in line 1396 and we should pop it in ActOnFinishFunctionBody later.
However if we skip the function body in line 1402, the expression
evaluation context will not be popped. Then here is the issue report. I
fix the issue by inserting codes to pop the expression evaluation
context explicitly if the function body is skipped. Maybe this looks
like an ad-hoc fix. But if we want to fix this in a pretty way, we
should refactor the current framework for pushing and popping expression
evaluation contexts. Currently there are 23
PushExpressionEvaluationContext() callsities and 21
PopExpressionEvaluationContext() callsites in the code. And it seems not
easy to balance them well and fast. So I suggest to land this fix first.
At least it can prevent the crash.

Reviewed By: cor3ntin

Differential Revision: https://reviews.llvm.org/D143053

Vector store on P8 little endian will have swap instruction added before
the store in PPCISelLowring. If the vector is generated by splat, the
swap instruction can be eliminated.

Reviewed By: shchenz

Differential Revision: https://reviews.llvm.org/D139691

As described in D45395 `This has been modeled after ARM's tail call opt.`
ARM's abi seems to limit weak symbol.

I did not find the limitation for RISCV. (Please correct me if I am wrong)

gcc seems to use the tail-call opt: https://godbolt.org/z/bjWE68n5o

Reviewed By: MaskRay

Differential Revision: https://reviews.llvm.org/D143137