For a function not in an IR COMDAT, currently we place it into a nodeduplicate IR
COMDAT so that its text section and its associated .pseudo_probe section will be
in the same section group, which can be retained or discarded by the linker as a
unit. However, the section group wastes space.

After D153189 uses SHF_LINK_ORDER to ensure a .pseudo_probe section will be
discarded when its associated text section is discarded, we can remove the
nodeduplicate IR change.

In the following example, the .pseudo_probe associated with .text.f is discarded as expected.
```
clang -c -ffunction-sections -fpseudo-probe-for-profiling -xc =(printf 'void _start(){} void f(){}') -o a.o
ld.lld --gc-sections --print-gc-sections a.o
```

Reviewed By: hoy

Differential Revision: https://reviews.llvm.org/D153191

OpenBSD commit for reference:
https://github.com/openbsd/src/commit/7b407c478fab53a6d9a091887c828c3f7b3f8b46

The test fails SCEV verification, which cause the expensive check bots
to fail. Always run verification and mark as XFAIL until fixed.

to simplify code near __tls_get_addr parsing.

Add support for the bare pointer calling convention in the gpu-to-llvm
pass. This wasn't being exposed and is needed when GPU-compiled MLIR is
to be called with this convention.

Reviewed By: krzysz00

Differential Revision: https://reviews.llvm.org/D152477

This reverts commit abfeda5a.
and fe19036e.

The added assertion triggers during clang bootstrap builds. Revert while
I investigate.

Differential Revision: https://reviews.llvm.org/D153203

Add test for LSR miscompile.

- This was to make implementing jthread easier and requested in https://reviews.llvm.org/D151559

Differential Revision: https://reviews.llvm.org/D151792

It was previously present in the inline code block and did not work as a
hyperlink.

Reviewed By: yassingh

Differential Revision: https://reviews.llvm.org/D153061

GenerateTruncates at the moment creates extends/truncates for post-inc
uses of normalized expressions. For example, if an add rec of the form
{1,+,-1} is used outside the loop, the normalized form will use {1,+,-1}
instead of {0,+,-1}. When naively sign-extending the normalized
expression, it will get extended incorrectly to {1,+,-1} for the wider
type, if the backedge-taken count of the loop is 1.

To address this, the patch updates GenerateTruncates to check if the
LSRUse contains any fixups with PostIncLoops. If that's the case, first
de-normalize the expression, then perform the extend/truncate, then
normalize again.

There may be other places where similar checks are needed and the helper
can be generalized for those cases. I'd not be surprised if other subtle
mis-compiles are caused by this.

Fixes #38847.
Fixes #58039.
Fixes #62852.

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D153004

* Add the SHF_LINK_ORDER flag so that the .pseudo_probe section is discarded when the associated text section is discarded.
* Add unique ID so that with `clang -ffunction-sections -fno-unique-section-names`, there is one separate .pseudo_probe for each text section (disambiguated by `.section ....,unique,id` in assembly)

The changes allow .pseudo_probe GC even if we don't place instrumented functions
in an IR comdat (see `getOrCreateFunctionComdat` in SampleProfileProbe.cpp).

Reviewed By: hoy

Differential Revision: https://reviews.llvm.org/D153189

Differential Revision: https://reviews.llvm.org/D153139

This commit breaks up CodeGen/TargetInfo.cpp into a set of *.cpp files,
one file per target. There are no functional changes, mostly just code moving.

Non-code-moving changes are:
* A virtual destructor has been added to DefaultABIInfo to pin the vtable to a cpp file.
* A few methods of ABIInfo and DefaultABIInfo were split into declaration + definition
  in order to reduce the number of transitive includes.
* Several functions that used to be static have been placed in clang::CodeGen
  namespace so that they can be accessed from other cpp files.

RFC: https://discourse.llvm.org/t/rfc-splitting-clangs-targetinfo-cpp/69883

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D148094

Wrap calls to XXXTargetCodeGenInfo constructors into factory functions.
This allows moving implementations of TargetCodeGenInfo to dedicated cpp
files without a change.

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D150215

Replace getAs with castAs and add assert if needed.
Differential Revision: https://reviews.llvm.org/D152977

`LoongArchISD::CSRWR` has two results. The first is the result of
`loongarch.csrwr.[wd]` intrinsic and the second is the chain. But
currently the chain is not processed correctly when creating this
node, resulting in the `csrwr` instruction being optimized out when
the result is not used by anyone [1]. `LoongArchISD::CSRXCHG` has
the same issue.

This patch addresses this issue.

[1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/loongarch/include/asm/loongarch.h?h=v6.4-rc6#n219

Reviewed By: hev

Differential Revision: https://reviews.llvm.org/D153120

Reviewed By: xry111

Differential Revision: https://reviews.llvm.org/D153119

The patch for promised interfaces (a5ef51d7) doesn't register all
extensions in the CAPI's `mlirRegisterAllDialects()` function.  This is
used by the MLIR Python bindings, causing downstream users of the Python
bindings to terminate abruptly.  This patch adds the call to register
all extensions.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D153174

Improve D151719.
(xor (setcc constant, y, setlt), 1) -> (setcc y, constant + 1, setlt)
https://alive2.llvm.org/ce/z/BZNEia

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D152128

This makes the scope and ordering arguments actually do something.
Also add some new OpenCL tests since the existing HIP tests didn't
cover address spaces.

[MC] Restore a special case to support limited A-B folding when A/B are in the same fragment being laided out

Add subsection-if.s to test what we can fold (in the same fragment) and what we cannot.

Fix https://github.com/ClangBuiltLinux/linux/issues/1876
Fixes: 4bdc7f7a

If the first token of an annotated line already has a computed Newlines,
reuse it to avoid potential overlapping whitespace replacements before
preprocessor branching directives.

Fixes #62892.

Differential Revision: https://reviews.llvm.org/D151954

The code was using the tail policy being "agnostic" to select a instruction whose semantics were "undefined". This was almost always fine (as the pass through operand was usually implicit_def), but could in theory lead to a miscompile. I don't actually have a test case as it requires a later transform to exploit the wrong tail policy state, and I couldn't easily figure out to get vsetvli insertion to miscompile given the wrong state. This was spotted by inspection, and it may be a miscompile in theory only at the moment.

Note that this may cause regressions if there are instructions for which we either don't have a _TU pseudo form, or the _TU pseudo form is missing a policy operand. When I was first looking at this, I saw exactly that, and D153067 exists to add the missing policy operand I noticed.

As a later follow up, I want to always force the use of _TU, but it seemed good to fix the bug, then driven the _TU transition in a separate patch.

Differential Revision: https://reviews.llvm.org/D153070

This change adds a policy operand to the helper class which is used for binary ops like vadd, but also, possibly surprisingly, some of the vslide variants. This allows us to represent the tail agnostic state with this pseudo family - previously, we could only represent tail undefined and tail undisturbed. (Since these don't have a mask, they're always mask undefined.)

This is NFC because no current producer uses the tail agnostic state. This will change in an upcoming change to doPeepholeMaskedRVV.

Differential Revision: https://reviews.llvm.org/D153067

It looks like scalable vector support broke vectorization for 0-D
tensors and we didn't have any test coverting that case. This patch
provides a fix and a test.

Differential Revision: https://reviews.llvm.org/D153181

Replaces verifyFormat(..., getGoogleStyle()) with
verifyGoogleFormat(...) in FormatTest.cpp.

Don't use an alignment larger than the vector size.

We can just explicitly check if the new unmasked pseudo takes a policy
op, rather than implicitly relying on I->UnmaskedTUPseudo ==
I->UnmaskedPseudo. Split out from another patch to make the diff more
readable.

Differential Revision: https://reviews.llvm.org/D152961

Differential Revision: https://reviews.llvm.org/D152960

This continues towards the goal spelled out in https://discourse.llvm.org/t/riscv-transition-in-vector-pseudo-structure-policy-variants/71295. This patch switches all the binary operations (no widen, no narrow, but both int and FP) to use the _TU + implicit_def passthrough form.  Change is mechanical.

This only changes the unmasked variants. Masked variants will still go through doPeepholeMaskedRVV and end up in the unsuffixed/TA form. Fixing that will be a separate change.

Differential Revision: https://reviews.llvm.org/D152940