platform/upstream/llvm.git
12 months ago[ValueTracking] Correctly check addrspace of alloca
Nikita Popov [Fri, 23 Jun 2023 11:08:46 +0000 (13:08 +0200)]
[ValueTracking] Correctly check addrspace of alloca

The DataLayout alloca address space is the address space that should
be used when creating new allocas. However, not all allocas are
required to be in this address space. The isKnownNonZero() check
should work on the actual address space of the alloca, not the
default alloca address space.

12 months agoFix a failing assertion with implicit function definitions
Aaron Ballman [Fri, 23 Jun 2023 11:02:21 +0000 (07:02 -0400)]
Fix a failing assertion with implicit function definitions

When implicitly defining a function in C, we would try to find an
appropriate declaration context for the function to be declared within.
However, we did not account for GNU statement expressions, which
masquerade as a compound statement and can be used in other contexts
such as within structure member declarations.

Fixes https://github.com/llvm/llvm-project/issues/48579

12 months ago[Clang][Driver] Warn on invalid Arm or AArch64 baremetal target triple
Michael Platings [Wed, 21 Jun 2023 13:49:09 +0000 (14:49 +0100)]
[Clang][Driver] Warn on invalid Arm or AArch64 baremetal target triple

A common user mistake is specifying a target of aarch64-none-eabi or
arm-none-elf whereas the correct names are aarch64-none-elf &
arm-none-eabi. Currently if a target of aarch64-none-eabi is specified
then the Generic_ELF toolchain is used, unlike aarch64-none-elf which
will use the BareMetal toolchain. This is unlikely to be intended by the
user so issue a warning that the target is invalid.

The target parser is liberal in what input it accepts so invalid triples
may yield behaviour that's sufficiently close to what the user intended.
Therefore invalid triples were used in many tests. This change updates
those tests to use valid triples.
One test (gnu-mcount.c) relies on the Generic_ELF toolchain behaviour so
change it to explicitly specify aarch64-unknown-none-gnu as the target.

Reviewed By: peter.smith, DavidSpickett

Differential Revision: https://reviews.llvm.org/D153430

12 months ago[CostModel] Use operands argument in getInstructionCost in more places
Luke Lau [Thu, 22 Jun 2023 16:59:55 +0000 (16:59 +0000)]
[CostModel] Use operands argument in getInstructionCost in more places

The current instruction's pointer operand may be different from the one
specified in the Operands argument. We should use the pointer operand
from here instead in case the user has transformed it.

This manifested itself somewhere down the line in
https://reviews.llvm.org/D149889, but I haven't been able to create a
test case on its own yet unfortunately.

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D153574

12 months ago[SCEV] Use object size for allocas as well
Nikita Popov [Fri, 23 Jun 2023 10:28:59 +0000 (12:28 +0200)]
[SCEV] Use object size for allocas as well

The object size and alignment based restriction on the possible
allocation range also applies to allocas, not just globals, so
handle them as well.

We shouldn't really need any type restriction here at all, but
for now stay conservative.

12 months ago[SCEV] Store getValue() result in variable (NFC)
Nikita Popov [Fri, 23 Jun 2023 10:30:01 +0000 (12:30 +0200)]
[SCEV] Store getValue() result in variable (NFC)

12 months ago[AArch64] Add tests for double reducts of vector.reduce.fmaximum/fminimum. NFC
David Green [Fri, 23 Jun 2023 10:28:09 +0000 (11:28 +0100)]
[AArch64] Add tests for double reducts of vector.reduce.fmaximum/fminimum. NFC

Including some tests with mixed minnum/minimum reductions and removing the fast
from fmin/fmax reductions as those should not be needed.

12 months ago[symbolizer] Check existence of input file in GNU mode
Serge Pavlov [Fri, 23 Jun 2023 10:19:38 +0000 (17:19 +0700)]
[symbolizer] Check existence of input file in GNU mode

GNU addr2line exits immediately if it cannot open the file specified as
executable/relocatable. In contrast llvm-addr2line does not exit and, if
addresses are not specified in command line, waits for input on stdin. This
causes the test compiler-rt/test/asan/TestCases/Posix/asan-symbolize-bad-path.cc to block
forever on Gentoo (see https://reviews.llvm.org/rG27c4777f41d2ab204c1cf84ff1cccd5ba41354da#1190273).
To fix this issue the behavior llvm-addr2line now exits if
executable/relocatable file cannot be found.

It fixes https://github.com/llvm/llvm-project/issues/42099 (llvm-addr2line
does not exit when passed a non-existent file).

Differential Revision: https://reviews.llvm.org/D147652

12 months ago[ConstantRange] Calculate precise range for multiply by -1
Nikita Popov [Fri, 23 Jun 2023 10:15:02 +0000 (12:15 +0200)]
[ConstantRange] Calculate precise range for multiply by -1

These are pretty common in SCEV, so make sure we get a precise
result by mapping to the sub() operation.

12 months agoRevert "Revert "[CodeGen] Extend reduction support in ComplexDeinterleaving pass...
Igor Kirillov [Fri, 23 Jun 2023 09:55:34 +0000 (09:55 +0000)]
Revert "Revert "[CodeGen] Extend reduction support in ComplexDeinterleaving pass to support predication""

Adds the capability to recognize SelectInst that appear in the IR.
These instructions are generated during scalable vectorization for reduction
and when the code contains conditions inside the loop body or when
"-prefer-predicate-over-epilogue=predicate-dont-vectorize" is set.

Differential Revision: https://reviews.llvm.org/D152558

This reverts commit ab09654832dba5cef8baa6400fdfd3e4d1495624.

Reason: Reapplying after removing unnecessary default case in switch expression.

12 months ago[ARM] generate armv6m eXecute Only (XO) code
Ties Stuij [Fri, 23 Jun 2023 09:16:48 +0000 (10:16 +0100)]
[ARM] generate armv6m eXecute Only (XO) code

[ARM] generate armv6m eXecute Only (XO) code for immediates, globals

Previously eXecute Only (XO) support was implemented for targets that support
MOVW/MOVT (~armv7+). See: https://reviews.llvm.org/D27449

XO prevents the compiler from generating data accesses to code sections. This
patch implements XO codegen for armv6-M, which does not support MOVW/MOVT, and
must resort to the following general pattern to avoid loads:

    movs    r3, :upper8_15:foo
    lsls    r3, #8
    adds    r3, :upper0_7:foo
    lsls    r3, #8
    adds    r3, :lower8_15:foo
    lsls    r3, #8
    adds    r3, :lower0_7:foo
    ldr     r3, [r3]

This is equivalent to the code pattern generated by GCC.

The above relocations are new to LLVM and have been implemented in a parent
patch: https://reviews.llvm.org/D149443.

This patch limits itself to implementing codegen for this pattern and enabling
XO for armv6-M in the backend.

Separate patches will follow for:
- switch tables
- replacing specific loads from constant islands which are spread out over the
  ARM backend codebase. Amongst others: FastISel, call lowering, stack frames.

Reviewed By: john.brawn

Differential Revision: https://reviews.llvm.org/D152795

12 months ago[NFC] clang-format GlobalISelEmitter.cpp
pvanhout [Fri, 23 Jun 2023 09:42:51 +0000 (11:42 +0200)]
[NFC] clang-format GlobalISelEmitter.cpp

It was overdue for a clang-format run, and it avoids unrelated formatting changes sneaking into diffs.

12 months ago[NFC] Remove leftover `inline` on some RuleMatcher functions
pvanhout [Fri, 23 Jun 2023 09:39:56 +0000 (11:39 +0200)]
[NFC] Remove leftover `inline` on some RuleMatcher functions

Accidentally copy-pasted them into the .cpp while refactoring the file in D151432
Those functions are currently only used in the .cpp so it didn't cause an issue, but it causes an undefined reference if another file attempts to use them.

12 months ago[SVE ACLE] Remove DAG combines that are no longer relevant.
Jolanta Jensen [Mon, 12 Jun 2023 09:14:13 +0000 (09:14 +0000)]
[SVE ACLE] Remove DAG combines that are no longer relevant.

This patch removes DAG combines that are no longer relevant
because equivalent IR combines have been added.

Differential Revision: https://reviews.llvm.org/D153445

12 months ago[APFloat] Add APFloat semantic support for TF32
Jeremy Furtek [Fri, 23 Jun 2023 08:54:35 +0000 (10:54 +0200)]
[APFloat] Add APFloat semantic support for TF32

This diff adds APFloat support for a semantic that matches the TF32 data type
used by some accelerators (most notably GPUs from both NVIDIA and AMD).

For more information on the TF32 data type, see https://blogs.nvidia.com/blog/2020/05/14/tensorfloat-32-precision-format/.
Some intrinsics that support the TF32 data type were added in https://reviews.llvm.org/D122044.

For some discussion on supporting common semantics in `APFloat`, see similar
efforts for 8-bit formats at https://reviews.llvm.org/D146441, as well as
https://discourse.llvm.org/t/rfc-adding-the-amd-graphcore-maybe-others-float8-formats-to-apfloat/67969.

A subsequent diff will extend MLIR to use this data type. (Those changes are
not part of this diff to simplify the review process.)

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D151923

12 months ago[LegacyPM] Remove LowerMatrixIntrinsicsLegacyPass and LowerMatrixIntrinsicsMinimalLeg...
Kazu Hirata [Fri, 23 Jun 2023 08:32:38 +0000 (01:32 -0700)]
[LegacyPM] Remove LowerMatrixIntrinsicsLegacyPass and LowerMatrixIntrinsicsMinimalLegacyPass

Differential Revision: https://reviews.llvm.org/D153615

12 months ago[LegacyPM] Remove unused declaration initializeAnnotation2MetadataLegacyPass
Kazu Hirata [Fri, 23 Jun 2023 08:32:36 +0000 (01:32 -0700)]
[LegacyPM] Remove unused declaration initializeAnnotation2MetadataLegacyPass

The corresponding function definition was removed by:

  commit 773d663e4729f55d23cb04f78a9d003643f2cb37
  Author: Arthur Eubanks <aeubanks@google.com>
  Date:   Mon Feb 27 19:00:37 2023 -0800

12 months ago[SCCPSolver] Speed up SCCPSolver by avoiding repeated work list elements
Tamás Danyluk [Fri, 23 Jun 2023 08:14:25 +0000 (10:14 +0200)]
[SCCPSolver] Speed up SCCPSolver by avoiding repeated work list elements

If a value is already the last element of the worklist, then I think that we don't have to add it again, it is not needed to process it repeatedly.

For some long Triton-generated LLVM IR, this can cause a ~100x speedup.

Differential Revision: https://reviews.llvm.org/D153561

12 months ago[mlir][transform] don't wrap a warning into silenceable failure
Alex Zinenko [Thu, 22 Jun 2023 13:49:15 +0000 (13:49 +0000)]
[mlir][transform] don't wrap a warning into silenceable failure

Wrapping a warning into a silenceable failure will result in the warning
being interpreted as an error, which it is not.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D153546

12 months ago[mlir][transform] fix handle invalidation check for reentrant regions
Alex Zinenko [Thu, 22 Jun 2023 13:24:55 +0000 (13:24 +0000)]
[mlir][transform] fix handle invalidation check for reentrant regions

When exiting the scope of a region attached to a transform op, clean up
the handle invalidation checks assocaited with handles defined in this
region. Otherwise, these checks may trigger on the next entry to the
region while there is no incorrect usage.

Reviewed By: springerm

Differential Revision: https://reviews.llvm.org/D153545

12 months ago[clang][ASTImporter] Add import of CXXRewrittenBinaryOperator.
Balázs Kéri [Fri, 23 Jun 2023 07:31:44 +0000 (09:31 +0200)]
[clang][ASTImporter] Add import of CXXRewrittenBinaryOperator.

Fix for issue #62770.

Reviewed By: donat.nagy

Differential Revision: https://reviews.llvm.org/D153424

12 months ago[TargetLowering] Better code generation for ISD::SADDSAT/SSUBSAT when operand sign...
Dhruv Chawla [Thu, 22 Jun 2023 17:02:01 +0000 (22:32 +0530)]
[TargetLowering] Better code generation for ISD::SADDSAT/SSUBSAT when operand sign is known

When the sign of either of the operands is known, it is possible to
determine what the saturating value will be without having to compute it
using the sign bits.

Differential Revision: https://reviews.llvm.org/D153575

12 months ago[AArch64] Pre-commit test for D153575
Dhruv Chawla [Fri, 23 Jun 2023 07:11:08 +0000 (12:41 +0530)]
[AArch64] Pre-commit test for D153575

12 months ago[LegacyPM] Remove LoopAccessLegacyAnalysis
Kazu Hirata [Fri, 23 Jun 2023 07:36:39 +0000 (00:36 -0700)]
[LegacyPM] Remove LoopAccessLegacyAnalysis

Differential Revision: https://reviews.llvm.org/D153610

12 months agoFix bytecode reader/writer on big-endian platforms
Ulrich Weigand [Fri, 23 Jun 2023 07:17:53 +0000 (09:17 +0200)]
Fix bytecode reader/writer on big-endian platforms

This makes the bytecode reader/writer work on big-endian platforms.
The only problem was related to encoding of multi-byte integers,
where both reader and writer code make implicit assumptions about
endianness of the host platform.

This fixes the current test failures on s390x, and in addition allows
to remove the UNSUPPORTED markers from all other bytecode-related
test cases - they now also all pass on s390x.

Also adding a GFAIL_SKIP to the MultiModuleWithResource unit test,
as this still fails due to an unrelated endian bug regarding
decoding of external resources.

Differential Revision: https://reviews.llvm.org/D153567

Reviewed By: mehdi_amini, jpienaar, rriddle

12 months ago[clangd] Store offsets in MacroOccurrence
Haojian Wu [Fri, 23 Jun 2023 07:08:35 +0000 (09:08 +0200)]
[clangd] Store offsets in MacroOccurrence

Remove the existing `Rng` field.

From the review comment: https://reviews.llvm.org/D147034

Reviewed By: kadircet

Differential Revision: https://reviews.llvm.org/D153259

12 months ago[flang][hlfir] Simplify hlfir::convertToValue
Jean Perier [Fri, 23 Jun 2023 07:04:14 +0000 (09:04 +0200)]
[flang][hlfir] Simplify hlfir::convertToValue

Use hlfir::loadTrivialScalars to dereference pointer, allocatables, and
load numerical and logical scalars.

This has a small fallout on tests:

- load is done on the HLFIR entity (#0 of hlfir.declare) and not the FIR one (#1). This makes no difference at the FIR level (#1 and #0 only differs to account for assumed and explicit shape lower bounds).

- loadTrivialScalars get rids of allocatable fir.box for monomoprhic scalars
  (it is not needed). This exposed a bug in lowering of MERGE with
  a polymorphic and a monomorphic argument: when the monomorphic is not
  a fir.box, the polymorphic fir.class should not be reboxed but its
  address should be read.

Reviewed By: tblah

Differential Revision: https://reviews.llvm.org/D153252

12 months ago[clang][dataflow] Dump useful debugging information when we crash.
Martin Braenne [Thu, 22 Jun 2023 14:02:54 +0000 (14:02 +0000)]
[clang][dataflow] Dump useful debugging information when we crash.

- The AST of the function we're currently analyzing
- The CFG
- The CFG element we're currently processing

Reviewed By: ymandel

Differential Revision: https://reviews.llvm.org/D153549

12 months ago[mlir] Remove unused forward declaration QuantizedIntegerType
Kazu Hirata [Fri, 23 Jun 2023 06:45:53 +0000 (23:45 -0700)]
[mlir] Remove unused forward declaration QuantizedIntegerType

The declaration was added without a corresponding class definition by:

  commit 13bb8f491a1cb429226768cfd4ca6bcea3b938dd
  Author: Stella Laurenzo <laurenzo@google.com>
  Date:   Wed Apr 3 11:16:32 2019 -0700

12 months ago[Transforms] Remove unused forward declaration PredicateScalarEvolution
Kazu Hirata [Fri, 23 Jun 2023 06:45:52 +0000 (23:45 -0700)]
[Transforms] Remove unused forward declaration PredicateScalarEvolution

The declaration was added without a corresponding class definition by:

  commit a84064bcda1a737658d33e96ca58516d01af70a6
  Author: Florian Hahn <flo@fhahn.com>
  Date:   Wed Dec 21 22:02:31 2022 +0000

It is most likely a misspelling of PredicatedScalarEvolution.

12 months ago[llvm] Use llvm::is_contained (NFC)
Kazu Hirata [Fri, 23 Jun 2023 06:45:50 +0000 (23:45 -0700)]
[llvm] Use llvm::is_contained (NFC)

12 months ago[mlir][docgen] Change nested check
Jacques Pienaar [Fri, 23 Jun 2023 05:05:31 +0000 (22:05 -0700)]
[mlir][docgen] Change nested check

Allows for single op nested regions.

12 months ago[CodeGen] Remove unused declaration createVLIWSched
Kazu Hirata [Fri, 23 Jun 2023 04:21:29 +0000 (21:21 -0700)]
[CodeGen] Remove unused declaration createVLIWSched

The declaration was added without a corresponding function by:

  commit cc3bb85580189d4a004cfd9bd2d6286cd1c1169f
  Author: James Nagurne <j-nagurne@ti.com>
  Date:   Fri Oct 22 17:08:16 2021 -0500

12 months ago[libc] Factor specifics of packet type out of process
Jon Chesterfield [Fri, 23 Jun 2023 02:45:22 +0000 (03:45 +0100)]
[libc] Factor specifics of packet type out of process

NFC. Simplifies process slightly, gives more options for testing it.

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D153604

12 months ago[libc] Simplify access permissions, change to composition over inheritance
Jon Chesterfield [Fri, 23 Jun 2023 01:54:21 +0000 (02:54 +0100)]
[libc] Simplify access permissions, change to composition over inheritance

Private member variable minimises scope of access to Process

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D153603

12 months ago[compiler-rt][CMake] Multiarch build of XRay libraries
Oleksii Lozovskyi [Fri, 23 Jun 2023 01:41:07 +0000 (18:41 -0700)]
[compiler-rt][CMake] Multiarch build of XRay libraries

Instead of dumping all sources into RTXray object library with a weird
special case for x86, handle multiarch builds better. Build a separate
object library for each arch with its arch-specific sources, then link
in all those libraries.

This fixes the build on platforms that produce fat binaries, such as new
macOS which expects both x86_64 and aarch64 objects in the same library
since Apple Silicon is a thing.

This only enables building XRay support for Apple Silicon. It does not
actually work yet on macOS, neither on Intel nor on Apple Silicon CPUs.
Thus the tests are still disabled.

Reviewed By: MaskRay, phosek

Differential Revision: https://reviews.llvm.org/D153221

12 months ago[libc] Instantiate and sanity check rpc class
Jon Chesterfield [Fri, 23 Jun 2023 01:11:18 +0000 (02:11 +0100)]
[libc] Instantiate and sanity check rpc class

CMake plumbing cargo culted from other tests.
Minor changes to Process to allow statically allocating a buffer.

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D153594

12 months ago[DAG] Peek through trunc when combining select into shifts.
Amaury Séchet [Sat, 30 Apr 2022 23:45:28 +0000 (23:45 +0000)]
[DAG] Peek through trunc when combining select into shifts.

This fixes a regression in D127115

Depends on D127115

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D151916

12 months ago[m68k] Fix incorrect handling of TLS when matching addressing mode.
Sheng [Fri, 23 Jun 2023 00:30:53 +0000 (08:30 +0800)]
[m68k] Fix incorrect handling of TLS when matching addressing mode.

`TargetGlobalTLSAddress` is not considered and handled correctly when matching addressing mode, which leads to an incorrect result of instruction selection.

fixes #63162.

Reviewed By: myhsu

Differential Revision: https://reviews.llvm.org/D153103

12 months ago[nfc][msan] Clang-format includes
Vitaly Buka [Fri, 23 Jun 2023 00:13:11 +0000 (17:13 -0700)]
[nfc][msan] Clang-format includes

12 months ago[msan] Release origin pages with shadow
Vitaly Buka [Fri, 23 Jun 2023 00:05:02 +0000 (17:05 -0700)]
[msan] Release origin pages with shadow

12 months ago[test][hwasan] Reformat comments
Vitaly Buka [Fri, 23 Jun 2023 00:00:55 +0000 (17:00 -0700)]
[test][hwasan] Reformat comments

12 months ago[test][hwasan] Describe why HWASAN does not work
Vitaly Buka [Thu, 22 Jun 2023 23:58:56 +0000 (16:58 -0700)]
[test][hwasan] Describe why HWASAN does not work

12 months ago[lldb] Adjust for changes in objc runtime
Alex Langford [Thu, 22 Jun 2023 23:04:06 +0000 (16:04 -0700)]
[lldb] Adjust for changes in objc runtime

The Objective-C runtime and the shared cache has changed slightly.
Given a class_ro_t, the baseMethods ivar is now a pointer union and may
either be a method_list_t pointer or a pointer to a relative list of
lists. The entries of this relative list of lists are indexes that refer
to a specific image in the shared cache in addition to a pointer offset
to find the accompanying method_list_t. We have to go over each of these
entries, parse it, and then if the relevant image is loaded in the
process, we add those methods to the relevant clang Decl.

In order to determine if an image is loaded, the Objective-C runtime
exposes a symbol that lets us determine if a particular image is loaded.
We maintain a data structure SharedCacheImageHeaders to keep track of
that information.

There is a known issue where if an image is loaded after we create a
Decl for a class, the Decl will not have the relevant methods from that
image (i.e. for Categories).

rdar://107957209

Differential Revision: https://reviews.llvm.org/D153597

12 months agoRevert "AMDGPU: Use generic helper for skipping over allocas"
Matt Arsenault [Thu, 22 Jun 2023 22:15:19 +0000 (18:15 -0400)]
Revert "AMDGPU: Use generic helper for skipping over allocas"

This reverts commit aa7e09ebd38c5f23f6d7d6d8394a2aea04715ba9.

12 months agoAMDGPU: Use generic helper for skipping over allocas
Matt Arsenault [Thu, 22 Jun 2023 21:16:43 +0000 (17:16 -0400)]
AMDGPU: Use generic helper for skipping over allocas

12 months ago[MC] Detect out of range jumps further than 2^32 bytes
Daniel Hoekwater [Thu, 22 Jun 2023 03:56:14 +0000 (03:56 +0000)]
[MC] Detect out of range jumps further than 2^32 bytes

On AArch64, object files may be greater than 2^32 bytes. If an
offset is greater than the max value of a 32-bit unsigned integer,
LLVM silently truncates the offset. Instead, make it return an
error.

Differential Revision: https://reviews.llvm.org/D153494

12 months ago[MC] Suppress -Wunused-but-set-variable in -DLLVM_ENABLE_ASSERTIONS=off builds after...
Fangrui Song [Thu, 22 Jun 2023 21:51:08 +0000 (14:51 -0700)]
[MC] Suppress -Wunused-but-set-variable in -DLLVM_ENABLE_ASSERTIONS=off builds after D153096

12 months ago[RISCV] Document overview of vector pseudos [nfc]
Philip Reames [Thu, 22 Jun 2023 21:47:03 +0000 (14:47 -0700)]
[RISCV] Document overview of vector pseudos [nfc]

I tried to give a rough overview of our current pseudo structure. I'm mostly focused on the policy handling bits - since that's what I'm in the process of changing - but touched on the other dimensions in the process of framing it.

Differential Revision: https://reviews.llvm.org/D152937

12 months ago[BOLT] Fixing relative ordering of cold sections under multi-way function splitting
Shatian Wang [Thu, 22 Jun 2023 21:26:21 +0000 (14:26 -0700)]
[BOLT] Fixing relative ordering of cold sections under multi-way function splitting

Order code sections with names in the form of ".text.cold.i" based on the value of i

[Context] SplitFunctions.cpp implements splitting strategies that can potentially split each function into maximum N>2 fragments.
When such N-way splitting happens, new code sections with names ".text.cold.1", ..., ".text.cold.i", ... "text.cold.N-2" will be created
A section with name ".text.cold.i" contains the the (i+2)th fragment of each function.
As an example, if each function is splitted into N=3 fragments: hot, warm, cold, then code sections will now include
- a section with name ".text" containing hot fragments
- a section with name ".text.cold" containing warm fragments
- a section with name ".text.cold.1" containing cold fragments

The order of these new sections in the output binary currently depends on the order in which they are encountered by the emitter.
For example, under N=3-way splitting, if the first function is 2-way splitted into hot and cold and the second function is 3-way splitted into hot, warm, and cold
then the cold fragment is encountered first, resulting in the final section to be in the following order
.text (hot), .text.cold.1 (cold), .text.cold (warm)

The above is suboptimal because the distance of jumps/calls between the hot and the warm sections will be much bigger than when ordering the sections as follows
.text (hot), .text.cold (warm), .text.cold.1 (cold)

This diff orders the sections with names in the form of ".text.cold" or ".text.cold.i" based on the value of i (assuming the i-value of ".text.cold" is 0).

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D152941

12 months ago[BOLT] Remove unnecessary diagnostics
Maksim Panchenko [Tue, 20 Jun 2023 22:27:31 +0000 (15:27 -0700)]
[BOLT] Remove unnecessary diagnostics

When optimizations passes do not change anything, skip their diagnostics
output. NFC otherwise.

Reviewed By: Amir

Differential Revision: https://reviews.llvm.org/D153386

12 months ago[lldb] Fix variable name mismatch between signature and docs (NFC)
Jonas Devlieghere [Thu, 22 Jun 2023 15:57:30 +0000 (08:57 -0700)]
[lldb] Fix variable name mismatch between signature and docs (NFC)

The variable is named `bundle_dir` but the documentation referenced
`directory` which generated a warning.

12 months ago[DFSAN] Add support for _tolower
Tomasz Kuchta [Wed, 21 Jun 2023 21:24:33 +0000 (21:24 +0000)]
[DFSAN] Add support for _tolower

I noticed that in some cases _tolower shows as uninstrumented - I've added it as "functional" in the done_abilist.txt file

Reviewed by: browneee

Differential Revision: https://reviews.llvm.org/D153410

12 months ago[MC,x86-32] Remove a gold<2.34 workaround
Fangrui Song [Thu, 22 Jun 2023 20:44:15 +0000 (13:44 -0700)]
[MC,x86-32] Remove a gold<2.34 workaround

This workaround appears to apply with gold<2.34 -O2/-O3 (linker -O2, not
compiler driver -O2). This used to be more visible as we used -Wl,-O3 in
CMake, but the option is generally not recommended and has been removed
by d63016a86548e8231002a760bbe9eb817cd1eb00 (Dec 2021).

This finishes a workaround removal work started by D64327 (2019).

Link: https://github.com/llvm/llvm-project/issues/45269
12 months ago[CLANG] Fix Static Code Analyzer Concerns with bad bit right shift operation in getNV...
Manna, Soumi [Thu, 22 Jun 2023 20:24:53 +0000 (13:24 -0700)]
[CLANG] Fix Static Code Analyzer Concerns with bad bit right shift operation in getNVPTXLaneID()

In getNVPTXLaneID(CodeGenFunction &), the value of LaneIDBits is 4294967295 since function call llvm::Log2_32(CGF->getTarget()->getGridValue().GV_Warp_Size) might return 4294967295.

  unsigned LaneIDBits =
       llvm::Log2_32(CGF.getTarget().getGridValue().GV_Warp_Size);
  unsigned LaneIDMask = ~0u >> (32u - LaneIDBits);

The shift amount (32U - LaneIDBits) might be 33, So it has undefined behavior for right shifting by more than 31 bits.

This patch adds an assert to guard the LaneIDBits overflow issue with LaneIDMask value.

Reviewed By: tahonermann

Differential Revision: https://reviews.llvm.org/D151606

12 months ago[flang] Avoid crash in statement function error case
Peter Klausler [Thu, 22 Jun 2023 16:25:03 +0000 (09:25 -0700)]
[flang] Avoid crash in statement function error case

The predicate IsPureProcedure() crashes with infinite
recursion when presented with mutually recursive statement
functions -- an error case that should be recoverable.
Fix by adding a visited set.

Fixes bug https://github.com/llvm/llvm-project/issues/63231

Differential Revision: https://reviews.llvm.org/D153569

12 months ago[libc] Can build amdgpu libc even if rocm is missing
Jon Chesterfield [Thu, 22 Jun 2023 20:18:43 +0000 (21:18 +0100)]
[libc] Can build amdgpu libc even if rocm is missing

Clang defaults to failing to build if it can't find rocm device libs

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D153581

12 months ago[flang][openacc] Add lowering support for multi-dimensional arrays reduction
Valentin Clement [Thu, 22 Jun 2023 20:10:50 +0000 (13:10 -0700)]
[flang][openacc] Add lowering support for multi-dimensional arrays reduction

Lower multi-dimensional arrays reduction for add and mul operator.

Depends on D153448

Reviewed By: razvanlupusoru

Differential Revision: https://reviews.llvm.org/D153455

12 months ago[flang][openacc] Add lowering support for 1d array reduction for add/mul operator
Valentin Clement [Thu, 22 Jun 2023 20:09:15 +0000 (13:09 -0700)]
[flang][openacc] Add lowering support for 1d array reduction for add/mul operator

Lower 1d array reduction for add and mul operator. Multi-dimensional arrays and
other operator will follow.

Reviewed By: razvanlupusoru

Differential Revision: https://reviews.llvm.org/D153448

12 months ago[CLANG] Fix potential null pointer dereference bugs
Manna, Soumi [Thu, 22 Jun 2023 19:52:25 +0000 (12:52 -0700)]
[CLANG] Fix potential null pointer dereference bugs

This patch uses castAs instead of getAs which will assert if the type doesn't match and adds nullptr check if needed.

Also this patch improves the codes and passes I.getData() instead of doing a lookup in dumpVarDefinitionName()
since we're iterating over the same map in LocalVariableMap::dumpContex().

Reviewed By: aaron.ballman, aaronpuchert

Differential Revision: https://reviews.llvm.org/D153033

12 months ago[NFC][asan] Add FIXME for a posible optimization
Vitaly Buka [Thu, 22 Jun 2023 18:17:12 +0000 (11:17 -0700)]
[NFC][asan] Add FIXME for a posible optimization

12 months ago[Clang] Fix Static Code Analysis Concerns with copy without assign
Manna, Soumi [Thu, 22 Jun 2023 19:30:01 +0000 (12:30 -0700)]
[Clang] Fix Static Code Analysis Concerns with copy without assign

This patch adds missing assignment operator to the class which has user-defined copy constructor.

Reviewed By: tahonermann, aaronpuchert

Differential Revision: https://reviews.llvm.org/D150931

12 months ago[asan] Don't double poison secondary allocations
Vitaly Buka [Thu, 22 Jun 2023 06:18:25 +0000 (23:18 -0700)]
[asan] Don't double poison secondary allocations

Sanitizers allocate shadow and memory as MAP_NORESERVE.

User memory can stay this way and do not increase RSS as long as we
don't store there.

The shadow unpoisoning also can avoid RSS increase for zeroed pages.
However as soon we poison the shadow, we need the page in RSS.

To avoid unnececary RSS increase we should not poison memory just before
unpoisoning them.

Depends on D153497.

Reviewed By: thurston

Differential Revision: https://reviews.llvm.org/D153500

12 months ago[MC] Fold A-B when A's fragment precedes B's fragment
Fangrui Song [Thu, 22 Jun 2023 19:24:19 +0000 (12:24 -0700)]
[MC] Fold A-B when A's fragment precedes B's fragment

When the MCAssembler is non-null and the MCAsmLayout is null, we can fold A-B
when

* A and B are in the same fragment, or
* A's fragment suceeds B's fragment, and they are not separated by non-data fragments (D69411)

This patch allows folding when A's fragment precedes B's fragment so
that `9997b - . == 0` below can be evaluated as true:

```
nop
.arch_extension sec
9997:nop
// old behavior: error: expected absolute expression
.if 9997b - . == 0
.endif
```

Add a case to llvm/test/MC/ARM/directive-if-subtraction.s.
Note: for MCAsmStreamer, we cannot evaluate `.if . - 9997b == 0` at parse
time due to MCAsmStreamer::getAssemblerPtr returning nullptr (D45164).

Some Darwin tests check that this folding does not work. Add `.p2align 2` to
block some label difference folding or adjust the tests.

Reviewed By: nickdesaulniers

Differential Revision: https://reviews.llvm.org/D153096

12 months ago[LV] Add test with reduction start values that are/may be poison/undef.
Florian Hahn [Thu, 22 Jun 2023 19:15:21 +0000 (20:15 +0100)]
[LV] Add test with reduction start values that are/may be poison/undef.

Test cases for #62565.

12 months ago[Clang] Disable `libc` headers for offloading languages
Joseph Huber [Thu, 22 Jun 2023 19:03:18 +0000 (14:03 -0500)]
[Clang] Disable `libc` headers for offloading languages

These headers are currently broken when included from the offloading
languages like OpenMP, OpenCL, CUDA, and HIP. Turn this logic off so we
can compile these languages when the GPU libc is installed. I am
currently trying to remedy this and have made an RFC for it in libc,
see https://discourse.llvm.org/t/rfc-implementing-gpu-headers-in-the-llvm-c-library/71523.

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D153578

12 months ago[MC][test] Clean up MC/ARM/directive-if-subtraction.s
Fangrui Song [Thu, 22 Jun 2023 19:11:52 +0000 (12:11 -0700)]
[MC][test] Clean up MC/ARM/directive-if-subtraction.s

12 months ago[CLANG] Fix uninitialized scalar field issues
Manna, Soumi [Thu, 22 Jun 2023 19:05:30 +0000 (12:05 -0700)]
[CLANG] Fix uninitialized scalar field issues

Reviewed By: erichkeane, steakhal, tahonermann, shafik

Differential Revision: https://reviews.llvm.org/D150744

12 months ago[dataflow] Avoid copying environment
Sam McCall [Thu, 22 Jun 2023 02:27:20 +0000 (04:27 +0200)]
[dataflow] Avoid copying environment

This appears to be just an accidental copy rather than move from a scratch
variable.
As well as doing redundant work, these copies introduce extra SAT variables
which make debugging harder (each Enviroment has a unique FC token).

Example flow condition before:
```
(B0:1 = V15)
(B1:1 = V8)
(B2:1 = V10)
(B3:1 = (V4 & (!V7 => V6)))
(V10 = (B3:1 & !V7))
(V12 = B1:1)
(V13 = B2:1)
(V15 = (V12 | V13))
(V3 = V2)
(V4 = V3)
(V8 = (B3:1 & !!V7))
B0:1
V2
```

after:
```
(B0:1 = (V9 | V10))
(B1:1 = (B3:1 & !!V6))
(B2:1 = (B3:1 & !V6))
(B3:1 = (V3 & (!V6 => V5)))
(V10 = B2:1)
(V3 = V2)
(V9 = B1:1)
B0:1
V2
```

(with labelling from D153488)

There are also some more copies that can be avoided here (when multiple blocks
without terminating statements are joined), but they're less trivial, so I'll
put those in another patch.

Differential Revision: https://reviews.llvm.org/D153491

12 months agoRevert "Try to implement lambdas with inalloca parameters by forwarding without use...
Amy Huang [Thu, 22 Jun 2023 18:36:30 +0000 (11:36 -0700)]
Revert "Try to implement lambdas with inalloca parameters by forwarding without use of inallocas."

Causes a clang crash (see crbug.com/1457256).

This reverts commit 015049338d7e8e0e81f2ad2f94e5a43e2e3f5220.

12 months ago[AMDGPU] Add _e64_dpp asm suffix to docs
Joe Nash [Thu, 22 Jun 2023 15:31:50 +0000 (11:31 -0400)]
[AMDGPU] Add _e64_dpp asm suffix to docs

The _e64_dpp suffix can be added to an instruction to force the
AsmParser to encode it as VOP3 with DPP if possible on GFX11+. This has
been the behavior since GFX11 was introduced; this patch only updates
the documentation.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D153564

12 months agoRevert "[CodeGen] Extend reduction support in ComplexDeinterleaving pass to support...
Vitaly Buka [Thu, 22 Jun 2023 18:24:47 +0000 (11:24 -0700)]
Revert "[CodeGen] Extend reduction support in ComplexDeinterleaving pass to support predication"

ComplexDeinterleavingPass.cpp:1849:3: error: default label in switch which covers all enumeration values

This reverts commit 116953b82130df1ebd817b3587b16154f659c013.

12 months ago[RISCV] Sort the extensions in SupportedExtensions and SupportedExperimentalExtensions.
Craig Topper [Thu, 22 Jun 2023 18:01:11 +0000 (11:01 -0700)]
[RISCV] Sort the extensions in SupportedExtensions and SupportedExperimentalExtensions.

As the extension list continues to grow it probably makes sense
to use a binary search rather than linear search. Sorting the strings
will make this possible.

This also avoids any question about where to add new strings in
the tables.

Reviewed By: asb

Differential Revision: https://reviews.llvm.org/D153170

12 months ago[asan] Optimize Quarantine of secondary allocations
Vitaly Buka [Thu, 22 Jun 2023 05:44:56 +0000 (22:44 -0700)]
[asan] Optimize Quarantine of secondary allocations

For the secondary allocation we don't need poison and fill memory if we
skip quarantine, and we don't need to poison after quarantine. In both
cases the secondary allocator will unmap memory and unpoison the shadow
from get_allocator().Deallocate().

Depends on D153496.

Reviewed By: thurston

Differential Revision: https://reviews.llvm.org/D153497

12 months agoInstCombine: Fold select of ldexp to ldexp of select
Matt Arsenault [Wed, 17 May 2023 09:18:51 +0000 (10:18 +0100)]
InstCombine: Fold select of ldexp to ldexp of select

The select-of-different-exp pattern appears in the device
libraries. I haven't seen the select-of-values case.

12 months agoInstCombine: Add some baseline tests for ldexp combines
Matt Arsenault [Wed, 17 May 2023 08:10:39 +0000 (09:10 +0100)]
InstCombine: Add some baseline tests for ldexp combines

12 months ago[Headers] Fix up some conditionals
Paul Robinson [Thu, 22 Jun 2023 17:15:16 +0000 (10:15 -0700)]
[Headers] Fix up some conditionals

12 months ago[LSR] Return nullptr from getExpr if the result isn't invertible.
Florian Hahn [Thu, 22 Jun 2023 18:10:48 +0000 (19:10 +0100)]
[LSR] Return nullptr from getExpr if the result isn't invertible.

getExpr is missing a check to make sure the result is invertible.
This can lead to incorrect results, so return nullptr in those cases
like in other places in IVUsers.

Fixes #62660.

Reviewed By: qcolombet

Differential Revision: https://reviews.llvm.org/D153202

12 months ago[opt] Exposing the parameters of LoopRotate to the -passes interface
Yann Girsberger [Thu, 22 Jun 2023 17:57:46 +0000 (10:57 -0700)]
[opt] Exposing the parameters of LoopRotate to the -passes interface

There is a gap between running opt -Oz and running opt -passes="OZ_PASSES" where OZ_PASSES is taken from running opt -Oz -print-pipeline-passes.

One of the reasons causing this is that -Oz uses non-default setting for LoopRotate but LoopRotate does not expose its settings when printing the pipeline.

This commit fixes this by exposing LoopRotates parameters.

Reviewed By: aeubanks

Differential Revision: https://reviews.llvm.org/D153437

12 months agoRevert "[llvm-exegesis] Add ability to assign perf counters to specific PID"
Aiden Grossman [Thu, 22 Jun 2023 18:01:57 +0000 (18:01 +0000)]
Revert "[llvm-exegesis] Add ability to assign perf counters to specific PID"

Revert "[llvm-exegesis] Introduce Subprocess Executor Mode"

This reverts commit 5e9173c43a9b97c8614e36d6f754317f731e71e9.
This reverts commit 4d618b52f6e05e41d35f56653cb36bf7d4dc794e.

Reverting the PID commit as it is currently breaking MinGW builds and
the way I'm checking for the presence of pid_t needs to be fixed and I
need to do some testing. The subprocess executor mode patch is a
dependent patch so also needs to be reverted and also needs some work as
it is currently failing tests where libpfm is installed and the kernel
version is less than 5.6.

12 months ago[llvm] Refactor BalancedPartitioning for fixing build failure with MSVC
Kamlesh Kumar [Thu, 22 Jun 2023 17:41:09 +0000 (23:11 +0530)]
[llvm] Refactor BalancedPartitioning for fixing build failure with MSVC

Fix build failure on windows system with msvc toolchain

Reviewed By: ellis

Differential Revision: https://reviews.llvm.org/D153318

12 months ago[lldb][Windows] Fix ZipFileResolver tests
Kazuki Sakamoto [Wed, 21 Jun 2023 00:30:29 +0000 (17:30 -0700)]
[lldb][Windows] Fix ZipFileResolver tests

D152759 introduced the Android .zip so file support, but it only considered
POSIX path. The code also runs on Windows, so the path could be Windows path.
Support both patterns on Windows.

Differential Revision: https://reviews.llvm.org/D153390

12 months ago[NFC][asan] Add const to QuarantineCallback methods
Vitaly Buka [Thu, 22 Jun 2023 17:34:19 +0000 (10:34 -0700)]
[NFC][asan] Add const to QuarantineCallback methods

12 months ago[NFC][asan] Extract FillChunk
Vitaly Buka [Thu, 22 Jun 2023 17:26:21 +0000 (10:26 -0700)]
[NFC][asan] Extract FillChunk

12 months ago[NFC][asan] Add QuarantineCallback::{PreQuarantine,RecyclePassThrough}
Vitaly Buka [Thu, 22 Jun 2023 05:26:01 +0000 (22:26 -0700)]
[NFC][asan] Add QuarantineCallback::{PreQuarantine,RecyclePassThrough}

Reviewed By: thurston

Differential Revision: https://reviews.llvm.org/D153496

12 months ago[RISCV] Improve SiFive7 for reductions and ordered reductions
Michael Maitland [Wed, 21 Jun 2023 21:49:05 +0000 (14:49 -0700)]
[RISCV] Improve SiFive7 for reductions and ordered reductions

Since the scheduling resources for reductions and ordered reductions now
account for LMUL and SEW, we can modify the Latency and ResourceCycles
for these resoruces.

* Most reductions take a total of approx `vl*SEW/DLEN + 5*(4 + log2(DLEN/SEW))`
  cycles.
* Ordered floating-point reductions take a total of approx `5*vl` cycles.

This commit re-commits 208fc34c65d648e869d7d3ba0dfcbca90942cda0. It was
failing because it used wrong version of SchedSEWSet.

Differential Revision: https://reviews.llvm.org/D153474

12 months ago[clang] Add a namespace for interesting identifiers.
Zahira Ammarguellat [Thu, 15 Jun 2023 19:44:25 +0000 (15:44 -0400)]
[clang] Add a namespace for interesting identifiers.

Differential Revision: https://reviews.llvm.org/D146148

12 months agoRevert "[RISCV] Improve SiFive7 for reductions and ordered reductions"
Michael Maitland [Thu, 22 Jun 2023 17:19:04 +0000 (10:19 -0700)]
Revert "[RISCV] Improve SiFive7 for reductions and ordered reductions"

This reverts commit 208fc34c65d648e869d7d3ba0dfcbca90942cda0.

Reverting because build failure.

12 months ago[RISCV] Improve SiFive7 for reductions and ordered reductions
Michael Maitland [Wed, 21 Jun 2023 21:49:05 +0000 (14:49 -0700)]
[RISCV] Improve SiFive7 for reductions and ordered reductions

Since the scheduling resources for reductions and ordered reductions now
account for LMUL and SEW, we can modify the Latency and ResourceCycles
for these resoruces.

* Most reductions take a total of approx `vl*SEW/DLEN + 5*(4 + log2(DLEN/SEW))`
  cycles.
* Ordered floating-point reductions take a total of approx `5*vl` cycles.

Differential Revision: https://reviews.llvm.org/D153474

12 months ago[RISCV] Improve SiFive7 for loads and stores
Michael Maitland [Wed, 21 Jun 2023 21:40:41 +0000 (14:40 -0700)]
[RISCV] Improve SiFive7 for loads and stores

* Unit-stride loads and stores can operate at the full bandwidth of the
memory pipe. The memory pipe is DLEN bits wide.

* Strided loads and stores operate at one element per cycle and should
be scheduled accordingly.

* Indexed loads and stores operate at one element per cycle, and they
stall the machine until all addresses have been generated, so they
cannot be scheduled.

* Unit stride seg2 load is number of DLEN parts

* seg3-8 are one segment per cycle, unless the segment is larger
than DLEN in which each segment takes multiple cycles.

Differential Revision: https://reviews.llvm.org/D153475

12 months ago[libc] Move fences into outbox/wait-for-ownership test
Jon Chesterfield [Thu, 22 Jun 2023 17:14:40 +0000 (18:14 +0100)]
[libc] Move fences into outbox/wait-for-ownership test

Also moves the wait-until-inbox-changes test into a shared method.

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D153573

12 months ago[asan] Don't quarantine large blocks
Vitaly Buka [Thu, 22 Jun 2023 05:02:43 +0000 (22:02 -0700)]
[asan] Don't quarantine large blocks

Almost NFC, as blocks over max quarantine size will trigger immediate
drain anyway. In followup patches we can optimize passthrough case.

Reviewed By: thurston

Differential Revision: https://reviews.llvm.org/D153495

12 months ago[XRay] Make xray_instr_map compatible with Mach-O
Fangrui Song [Thu, 22 Jun 2023 17:03:17 +0000 (10:03 -0700)]
[XRay] Make xray_instr_map compatible with Mach-O

The `__DATA,xray_instr_map` section has label differences like
`.quad Lxray_sled_0-Ltmp0` that is represented as a pair of UNSIGNED and SUBTRACTOR relocations.

LLVM integrated assembler attempts to rewrite A-B into A-B'+offset where B' can
be included in the symbol table. B' is called an atom and should be a
non-temporary symbol in the same section. However, since `xray_instr_map` does
not define a non-temporary symbol, the SUBTRACTOR relocation will have no
associated symbol, and its `r_extern` value will be 0. Therefore, we will see
linker errors like:

    error: SUBTRACTOR relocation must be extern at offset 0 of __DATA,xray_instr_map in a.o

To fix this issue, we need to define a non-temporary symbol in the section. We
can accomplish this by renaming `Lxray_sleds_start0` to `lxray_sleds_start0`
("L" to "l").

`lxray_sleds_start0` serves as the atom for this dead-strippable subsection.
With the `S_ATTR_LIVE_SUPPORT` attribute, `ld -dead_strip` will retain
subsections that reference live functions.

Special thanks to Oleksii Lozovskyi for reporting the issue and providing
initial analysis.

Differential Revision: https://reviews.llvm.org/D153239

12 months ago[NFC] Fix potential dereferencing of nullptr.
Sindhu Chittireddy [Mon, 19 Jun 2023 02:31:53 +0000 (19:31 -0700)]
[NFC] Fix potential dereferencing of nullptr.

Replace getAs with castAs and add assert if needed.
Differential revision: https://reviews.llvm.org/D153236

12 months ago[CodeGen] Extend reduction support in ComplexDeinterleaving pass to support predication
Igor Kirillov [Fri, 9 Jun 2023 16:03:22 +0000 (16:03 +0000)]
[CodeGen] Extend reduction support in ComplexDeinterleaving pass to support predication

Adds the capability to recognize SelectInst that appear in the IR.
These instructions are generated during scalable vectorization for reduction
and when the code contains conditions inside the loop body or when
"-prefer-predicate-over-epilogue=predicate-dont-vectorize" is set.

Differential Revision: https://reviews.llvm.org/D152558

12 months ago[libc][NFC] Simplify return value logic in set_thread_ptr()
Jun Zhang [Thu, 22 Jun 2023 16:38:41 +0000 (00:38 +0800)]
[libc][NFC] Simplify return value logic in set_thread_ptr()

Signed-off-by: Jun Zhang <jun@junz.org>
Differential Revision: https://reviews.llvm.org/D153572

12 months ago[libc] Add memory fences to device-local locking calls
Jon Chesterfield [Thu, 22 Jun 2023 16:46:08 +0000 (17:46 +0100)]
[libc] Add memory fences to device-local locking calls

This makes the interface less error prone. The acquire was previously
forgotten. Release is currently missing if recv() is the last operation made
before close.

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D153571

12 months ago[RISCV] Use GPR register class for RV64 ZDInx. Remove GPRF64 register class.
Craig Topper [Thu, 22 Jun 2023 16:38:46 +0000 (09:38 -0700)]
[RISCV] Use GPR register class for RV64 ZDInx. Remove GPRF64 register class.

The GPRF64 has the same spill size as GPR and is only used for RV64.
There's no real reason to have it as a separate class other than
for type inference for isel patterns in tablegen.

This patch adds f64 to the GPR register class when XLen=64. I use
f32 when XLen=32 even though we don't make use of it just to avoid
the oddity.

isel patterns have been updated to fix the lack of type infererence.

I might do similar for GPRF16 and GPRF32 or I might change them to
use an optimized spill size instead of always using XLen.

Reviewed By: asb

Differential Revision: https://reviews.llvm.org/D153110

12 months ago[vscode-mlir] bump vsce version to 2.19.0
Yuanfang Chen [Wed, 21 Jun 2023 21:45:35 +0000 (14:45 -0700)]
[vscode-mlir] bump vsce version to 2.19.0

For https://bugs.chromium.org/p/llvm/issues/detail?id=46

Differential Revision: https://reviews.llvm.org/D153473

12 months ago[RISCV] Move Zca/Zcb/Zcd/Zcf/Zcmp/Zcmt out of experimental status.
Craig Topper [Thu, 22 Jun 2023 16:22:58 +0000 (09:22 -0700)]
[RISCV] Move Zca/Zcb/Zcd/Zcf/Zcmp/Zcmt out of experimental status.

According to https://wiki.riscv.org/display/HOME/Recently+Ratified+Extensions
these were ratified in April 2023.

Reviewed By: VincentWu

Differential Revision: https://reviews.llvm.org/D153161