Unlike older ASICs GFX10+ have a lot of VGPRs. Therefore, it is possible
to achieve high occupancy even with all or almost all addressable VGPRs
used. Our scheduler was never tuned for this scenario. The VGPR Critical
Limit threshold always comes very high, even if maximum occupancy is
targeted. For example on gfx1100 it is set to 192 registers even with
the requested occupancy 16. As a result scheduler starts prioritizing
register pressure reduction very late and we easily end up spilling.

This patch makes VGPR critical limit similar to what we would have on
pre-gfx10 targets with much more limited VGPR budget while still trying
to maintain occupancy as it does now.

Pre-gfx10 ASICs shall not be affected as the limit shall be the same
as before, and on gfx10+ it shall only affect regions where we have
to spill.

Fixes: SWDEV-377300

Differential Revision: https://reviews.llvm.org/D141876

The Armv8.6-a and later architecture definitions included AES, SHA2,
SHA3 and SM4, but this did not have an effect when specifying
-march=armv8.6-a. The did not set preprocessor features
(https://godbolt.org/z/1YKad6M8e) or enable the relevant instructions
(like eor3 from sha3: https://godbolt.org/z/vY9v4MqvG). Similarly
architectures armv8 to armv8.5 defined +crypto, but this did not effect
the -march's, only the -mcpu with those architectures. I believe this
was working as intended.

After D141411 we now add the default features for architectures except
for +crypto, which has had the effect of enabling aes/sha2/sha3/sm4 when
-march=armv8.6-a is used. This patch removed those crypto features
again, going back to how things were before. It also removes the
AEK_CRYPTO feature from lower architecture levels, moving it to the cpus
that use it. This shouldn't make any changes, but a few extra tests have
been added for preprocessor features that have improved since llvm 15.

The -mcpu=ampere1 cpu is the only armv8.6+ cpu at present. For that, the
AES, SHA2 and SHA3 features have been re-added to the CPU definition to
keep it in-line with the gcc definition from
https://github.com/gcc-mirror/gcc/commit/db2f5d661239737157cf131de7d4df1c17d8d88d.

Differential Revision: https://reviews.llvm.org/D141606

Differential Revision: https://reviews.llvm.org/D142265

Differential Revision: https://reviews.llvm.org/D142039

This reverts commit 5474d7d9.
Wrong differential revision link was used.

When compiling clang/Lex/DirectoryLookup.h with option -Wbitfield-enum-conversion, we get the following warning:

DirectoryLookup.h:77:17: warning:
      bit-field 'DirCharacteristic' is not wide enough to store all enumerators of
      'CharacteristicKind' [-Wbitfield-enum-conversion]
      : u(Map), DirCharacteristic(DT), LookupType(LT_HeaderMap),

DirCharacteristic is a bitfield with 2 bits (4 values)
  /// DirCharacteristic - The type of directory this is: this is an instance of
  /// SrcMgr::CharacteristicKind.
  unsigned DirCharacteristic : 2;

Whereas SrcMgr::CharacterKind is an enum with 5 values:
enum CharacteristicKind {
  C_User,
  C_System,
  C_ExternCSystem,
  C_User_ModuleMap,
  C_System_ModuleMap
};

Solution is to increase DirCharacteristic bitfield from 2 to 3.
Patch by Dimitri van Heesch

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D142304

This reduces RISCV.td to mainly being a top level include file.

Reviewed By: asb, luismarques

Differential Revision: https://reviews.llvm.org/D142239

Differential Revision: https://reviews.llvm.org/D141606

This reverts commit 84c6129c.

Add the following intrinsic:
  SQCVT
  SQCVTU
  UQCVT

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

Reviewed By: kmclaughlin

Differential Revision: https://reviews.llvm.org/D142035

This patch moves a couple of helper functions from the global llvm::
namespace into the SCCPSolver class. This reduces the need for separate
SCCPSolver arguments and also limits the scope of those functions that
have quite generic names.

(The remaining isConstant and isOverdefined should ideally be removed)

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D142370

Add the following intrinsic:

  FCVT
  BFCVT
  FCVTZS
  FCVTZU
  SCVTF
  UCVTF

This patch also adds SelectCVTIntrinsic to handle the cases when the
intrinsic returns multiple (two or four) outputs

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

Reviewed By: kmclaughlin

Differential Revision: https://reviews.llvm.org/D142032

This option is useful for clang and clang-cl.

Differential Revision: https://reviews.llvm.org/D142367

Fixes #60184  https://github.com/llvm/llvm-project/issues/60184

Differential Revision: https://reviews.llvm.org/D142295

This patch reduces CMake configuration time drastically by removing a non-linear behavior.
Time to execute CMake configure step goes from 45s to 15s.

Differential Revision: https://reviews.llvm.org/D142374

Update the clang driver to include the following features as default for
the v8.9-A/v9.4-A architecture versions:

* FEAT_SPECRES2
* FEAT_CSSC
* FEAT_RASv2

Patch by Sam Elliott.

Reviewed By: lenary, tmatheson

Differential Revision: https://reviews.llvm.org/D141404

This introduces command line support (`+ite`) for the v9.4-A's
Instrumentation Extension (FEAT_ITE).

Patch by Son Tuan Vu.

Reviewed By: lenary, tmatheson

Differential Revision: https://reviews.llvm.org/D141403

Reviewed By: peter.smith

Differential Revision: https://reviews.llvm.org/D142229

collapse/expand_shape are supposed to be expanded before we hit the
lowering code.
The expansion is done with the pass called expand-strided-metadata.

This patch is NFC in spirit but not in practice because
expand-strided-metadata won't try to accomodate for "invalid" strides
for dynamic sizes that are 1 at runtime.

The previous code was broken in that respect too, but differently: it
handled only the case of row-major layouts.
That whole part is being reworked separately.

Differential Revision: https://reviews.llvm.org/D136483

bazel: adapt for https://github.com/llvm/llvm-project/commit/a4699a43e42615281c96599d20977cabf10bfb9c

This patch contains several related changes:

1. We move to using TARGET_BUILTIN for the 128-bit system register
   builtins to give better error messages when d128 has not been
   enabled, or has been enabled in a per-function manner.

2. We now validate the inputs to the 128-bit system register builtins,
   like we validate the other system register builtins.

3. We update the list of named PSTATE accessors for MSR (immediate), and
   now correctly enforce the expected ranges of the immediates. There is
   a long comment about how we chose to do this to comply with the ACLE
   when most of the PSTATE accessors for MSR (immediate) have aliased
   system registers for MRS/MSR which expect different values. In short,
   the MSR (immediate) names are prioritised, rather than falling-back
   to the register form when the value is out of range.

Differential Revision: https://reviews.llvm.org/D140222

The verification of affine value classification for symbols was
expecting, incorrectly, that the dimension operand of `memref.dim` was
being produced by a constant-like operation. This is legacy of the
dimension being an attribute originally, and was never updated after it
was switched to be an operation. Treat such cases conservatively and
classify the value as non-symbol.

A more advanced version could attempt to check that the value would be a
valid symbol for all possible values the dimension attribute could take,
but this does not seem immediately useful.

Fixes #59993.

Reviewed By: bondhugula

Differential Revision: https://reviews.llvm.org/D142204

It should have an "Allocate" effect on entry block arguments of all
regions in addition to consuming the operand.

Also relax the assertion in transform-dialect-check-uses until we can
properly support region-based control flow.

Fixes #60075.

Reviewed By: springerm

Differential Revision: https://reviews.llvm.org/D142200

Fixes #60058  https://github.com/llvm/llvm-project/issues/60058
It hit assert when legalizePatternResult on success of ExpandIfCondition which did nothing just return success when if condition is constant.

Added RemoveConstantIfCondition to remove the if cond by getCanonicalizationPatterns.
Also remove the check for constant if cond in ExpandIfCondition and
change check ifCond to assert because only op with ifCond will need legalize in ConvertOpenACCToSCFPass

Differential Revision: https://reviews.llvm.org/D142286

This adds support for the v8.9-A/v9.4-A architectural extensions to be
used in .arch_extension assembly directives.

Patch by Sam Elliott.

Reviewed By: lenary, tmatheson

Differential Revision: https://reviews.llvm.org/D141402

This adds support for the missing `PIRE0_EL12` system register, part of
v8.9-A/v9.4-A's Permission Indirection Extension.

Patch by Son Tuan Vu.

Reviewed By: tmatheson

Differential Revision: https://reviews.llvm.org/D141400

Summary:
Fix a few minor warnings that show up in `libomptarget`.

Summary:
Recently AMD moved the "hsa.h" include to "hsa/hsa.h". This causes
several warning. This patch checks to see if we can include that one
instead. This should hopefully keep things backwards compatible while
silencing the warnings.

Summary:
These warnings are very loud considering they get repeated at least 30
times each build. This patch just silences them.

Make the first line a title and relative link
to the Markdown of the demo notebook.

Found by PVS-Studio - https://pvs-studio.com/en/blog/posts/cpp/1003/, N37.

The code you is using the bit mask NullabilityKindMask which is 0x3
(00000011 in binary) to clear the bits in the NullabilityPayload variable.
Since NullabilityPayload is a 64-bit variable and NullabilityKindMask is
only a 8-bit variable(0x3), it will only affect the last 8 bits of the
variable. The higher 56 bits will remain unchanged.

Differential Revision: https://reviews.llvm.org/D142334

The new methods return a range for easier iteration. Use them everywhere
instead of getImplicitUses, getNumImplicitUses, getImplicitDefs and
getNumImplicitDefs. A future patch will remove the old methods.

In some use cases the new methods are less efficient because they always
have to scan the whole uses/defs array to count its length, but that
will be fixed in a future patch by storing the number of implicit
uses/defs explicitly in MCInstrDesc. At that point there will be no need
to 0-terminate the arrays.

Differential Revision: https://reviews.llvm.org/D142215

When a `!fir.box<>` is passed as an actual argument to an optional
`!fir.class<>` dummy it needs a `fir.rebox` in order to propagate
the dynamic type information.
The `fir.rebox` needs to happen only on present argument.

Reviewed By: jeanPerier

Differential Revision: https://reviews.llvm.org/D142340

Also, move the XformToShuffleWithZero and combineCarryDiamond folds later after some of the more basic canonicalizations/combines (such as this) have had a chance to occur

Fixes the v8i1-masks.ll regression from D127115

Introduce signals to rank providers of a symbol.

Differential Revision: https://reviews.llvm.org/D139921