platform/upstream/llvm.git
23 months ago[mlir][Transform] Fix ASAN heap-use-after-free
Nicolas Vasilache [Wed, 2 Nov 2022 16:39:16 +0000 (09:39 -0700)]
[mlir][Transform] Fix ASAN heap-use-after-free

23 months ago[mlir][scf] refactor scf structuralOpConversion to better support 1:N type conversion
Peiming Liu [Fri, 28 Oct 2022 23:39:47 +0000 (23:39 +0000)]
[mlir][scf] refactor scf structuralOpConversion to better support 1:N type conversion

This patch moves the 1:N type mapping into its own classes to allow better code reuse in D137100.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D137099

23 months ago[openmp][AMDGPU] - Correct getNumberOfBlocks calculation.
Ethan Stewart [Wed, 2 Nov 2022 16:37:42 +0000 (11:37 -0500)]
[openmp][AMDGPU] - Correct getNumberOfBlocks calculation.

This patch fixes the 6 amdgpu buildbot lit test failures
introduced by https://reviews.llvm.org/D135444.
      libomptarget :: amdgcn-amd-amdhsa :: mapping/reduction_implicit_map.cpp
      libomptarget :: amdgcn-amd-amdhsa :: offloading/cuda_no_devices.c
      libomptarget :: amdgcn-amd-amdhsa :: offloading/target-teams-atomic.c
      libomptarget :: amdgcn-amd-amdhsa-LTO :: mapping/reduction_implicit_map.cpp
      libomptarget :: amdgcn-amd-amdhsa-LTO :: offloading/cuda_no_devices.c
      libomptarget :: amdgcn-amd-amdhsa-LTO :: offloading/target-teams-atomic.c

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D137261

23 months ago[cmake][msvc] Enable standards-conforming preprocessor
Jan Svoboda [Wed, 2 Nov 2022 16:35:00 +0000 (09:35 -0700)]
[cmake][msvc] Enable standards-conforming preprocessor

Since we now only support Visual Studio 2019 16.7 and newer, we're able to use the /Zc:preprocessor flag that turns on the standards-conforming preprocessor. Among other things, it correctly expands __VA_ARGS__. This enables removal of some boilerplate in D135128.

Reviewed By: Bigcheese, thieta

Differential Revision: https://reviews.llvm.org/D135128

23 months ago[lldb] Update TestDump.test for gnuwin32's 'file' command output
Jonas Devlieghere [Wed, 2 Nov 2022 16:30:55 +0000 (09:30 -0700)]
[lldb] Update TestDump.test for gnuwin32's 'file' command output

Remove the colon from the CHECK line for the output of the file command.
On macOS and Linux, the file command uses a colon as the separator
between the path and the file type, but gnuwin32's file command uses a
semicolon.

23 months ago[mlir][Pass] Disallow mixing -pass-pipeline with other pass options
rkayaith [Tue, 11 Oct 2022 19:23:48 +0000 (15:23 -0400)]
[mlir][Pass] Disallow mixing -pass-pipeline with other pass options

Currently `-pass-pipeline` can be specified multiple times and mixed
with the individual `-pass-name` options. Removing this feature will
allow for including the pipeline anchor as part of the option
argument (see D134900).

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D135745

23 months ago[mlir] Don't mix -pass-pipeline with other pass options
rkayaith [Tue, 11 Oct 2022 21:29:39 +0000 (17:29 -0400)]
[mlir] Don't mix -pass-pipeline with other pass options

These are test updates required for D135745, which disallows mixing
`-pass-pipeline` and the individual `-pass-name` options.

Reviewed By: rriddle, mehdi_amini

Differential Revision: https://reviews.llvm.org/D135746

23 months ago[OpenMP][libomptarget] Fix AsyncInfoTy object in omp_target_memcpy
Kevin Sala Penadés [Wed, 2 Nov 2022 16:03:34 +0000 (12:03 -0400)]
[OpenMP][libomptarget] Fix AsyncInfoTy object in omp_target_memcpy

The AsyncInfoTy should be created in the same device as the async operation will be issued. In omp_target_memcpy, the AsyncInfoTy for the host to destination device transfer was created referring to the source device.

Reviewed By: tianshilei1992

Differential Revision: https://reviews.llvm.org/D137225

23 months ago[flang] Handle dispatch on nopass procedure from array element
Valentin Clement [Wed, 2 Nov 2022 15:52:05 +0000 (16:52 +0100)]
[flang] Handle dispatch on nopass procedure from array element

When a fir.dispatch is done with a NOPASS type-bound
procedure, the object in fir.dispatch is just used to gather the vtable.
Therefore, no transformation is done on it and the original entity
is used. The current code generation didn't expect the entity to be an array.
This patch update the code generation to be able to retrieve the vtable
accordingly.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D137255

2 years agollvm-reduce: Add explicit triple to test
Matt Arsenault [Wed, 2 Nov 2022 15:30:21 +0000 (08:30 -0700)]
llvm-reduce: Add explicit triple to test

Some host targets are managing to get through MIR parsing
without constructing their MachineFunctionInfo.

Fixes at least SystemZ and SPARC (issue 58768)

2 years ago[libc] Add a doc about the libc fullbuild mode.
Siva Chandra Reddy [Mon, 31 Oct 2022 08:25:25 +0000 (01:25 -0700)]
[libc] Add a doc about the libc fullbuild mode.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D137061

2 years ago[libc] Add links to libc Discourse and to the libc Discord channel.
Siva Chandra Reddy [Wed, 2 Nov 2022 07:34:11 +0000 (00:34 -0700)]
[libc] Add links to libc Discourse and to the libc Discord channel.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D137229

2 years ago[libc] Add a contributing guide to the docs.
Siva Chandra Reddy [Fri, 28 Oct 2022 08:09:38 +0000 (01:09 -0700)]
[libc] Add a contributing guide to the docs.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D136961

2 years ago[libc] Add a porting guide to the docs.
Siva Chandra Reddy [Fri, 28 Oct 2022 07:12:08 +0000 (00:12 -0700)]
[libc] Add a porting guide to the docs.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D136960

2 years ago[RISCV] Fix cost of legal fixed length masked load and stores
Philip Reames [Wed, 2 Nov 2022 00:19:29 +0000 (17:19 -0700)]
[RISCV] Fix cost of legal fixed length masked load and stores

We can cost them the same way as a scalable masked load/store. By hitting the default path, we were costing them as if they were being scalarized. This is a significant over estimate.

Differential Revision: https://reviews.llvm.org/D137218

2 years agoFix Assignment Tracking docs error from 33c7ae55e729069be754f56c4d4606cdeddd377b
OCHyams [Wed, 2 Nov 2022 14:08:23 +0000 (14:08 +0000)]
Fix Assignment Tracking docs error from 33c7ae55e729069be754f56c4d4606cdeddd377b

link: https://lab.llvm.org/buildbot/#/builders/30/builds/27812
2 years ago[mlir] Use the same pipeline tuning options as clang for execution engine
Denys Shabalin [Wed, 2 Nov 2022 13:04:37 +0000 (14:04 +0100)]
[mlir] Use the same pipeline tuning options as clang for execution engine

This change make sure that ExecutionEngine's pass pipeline is identical to one
used by clang. Previously, SLPVectorization was not enabled which caused
differences in code...

...generation.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D137248

2 years ago[MIRVRegNamer] Avoid opcode hash collision
John Brawn [Fri, 28 Oct 2022 13:17:01 +0000 (14:17 +0100)]
[MIRVRegNamer] Avoid opcode hash collision

D121929 happens to cause CodeGen/MIR/AArch64/mirnamer.mir to fail due
to a hash collision caused by adding two extra opcodes. The collision
is only in the top 19 bits of the hashed opcode so fix this by just
using the whole hash (in fixed width hex for consistency) instead of
the top 5 decimal digits.

Differential Revision: https://reviews.llvm.org/D137155

2 years ago[MachineCSE] Allow PRE of instructions that read physical registers
John Brawn [Thu, 27 Oct 2022 13:14:57 +0000 (14:14 +0100)]
[MachineCSE] Allow PRE of instructions that read physical registers

Currently MachineCSE forbids PRE when the instruction reads a physical
register. Relax this so that it's allowed when the value being read is
the same as what would be read in the place the instruction would be
hoisted to.

This is being done in preparation for adding FPCR handling to the
AArch64 backend, in order to prevent it to from worsening the
generated code, but for targets that already have a similar register
it should improve things.

This patch affects code generation in several tests. The new code
looks better except for in Thumb2/LowOverheadLoops/memcall.ll where
we perform PRE but the LowOverheadLoops transformation then undoes
it. Also in AMDGPU/selectcc-opt.ll the CHECK makes things look worse,
but actually the function as a whole is better (as a MOV is PRE'd).

Differential Revision: https://reviews.llvm.org/D136675

2 years ago[CVP] Add test for icmp without constant operand (NFC)
Nikita Popov [Wed, 2 Nov 2022 13:47:57 +0000 (14:47 +0100)]
[CVP] Add test for icmp without constant operand (NFC)

We should still be able to fold comparisons based on ranged in
this case.

2 years ago[Assignment Tracking][1/*] Add initial docs for Assignment Tracking
OCHyams [Wed, 2 Nov 2022 13:42:52 +0000 (13:42 +0000)]
[Assignment Tracking][1/*] Add initial docs for Assignment Tracking

The Assignment Tracking debug-info feature is outlined in this RFC:

https://discourse.llvm.org/t/
  rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir

Add documentation outlining the intent and design.

2 years ago[InstCombine] fold mul with decremented "shl -1" factor (2nd try)
Sanjay Patel [Wed, 2 Nov 2022 13:15:11 +0000 (09:15 -0400)]
[InstCombine] fold mul with decremented "shl -1" factor (2nd try)

This is a corrected version of:
bc886e9b587b

I made a copy-paste error that created an "add" instead of the
intended "sub" on that attempt. The regression tests showed the
bug, but I overlooked that.

As I said in a comment on issue #58717, the bug reports resulting
from the botched patch confirm that the pattern does occur in
many real-world applications, so hopefully eliminating the multiply
results in better code.

I added one more regression test in this version of the patch,
and here's an Alive2 proof to show that exact example:
https://alive2.llvm.org/ce/z/dge7VC

Original commit message:

This is a sibling to:
6064e92b0a84
...but we canonicalize the shl+add to shl+xor,
so the pattern is different than I expected:
https://alive2.llvm.org/ce/z/8CX16e

I have not found any patterns that are safe
to propagate no-wrap, so that is not included
here.

Differential Revision: https://reviews.llvm.org/D137157

2 years ago[DebugInfo] Fix minor debug info bug in deleteDeadLoop
OCHyams [Thu, 20 Oct 2022 14:40:14 +0000 (15:40 +0100)]
[DebugInfo] Fix minor debug info bug in deleteDeadLoop

Using a DebugVariable as the set key rather than std::pair<DIVariable *,
DIExpression *> ensures we don't accidently confuse multiple instances of
inlined variables.

Reviewed By: jryans

Differential Revision: https://reviews.llvm.org/D133303

2 years ago[BasicAA] Re-enable cs-cs-arm.ll test (PR58738)
Nikita Popov [Wed, 2 Nov 2022 13:21:38 +0000 (14:21 +0100)]
[BasicAA] Re-enable cs-cs-arm.ll test (PR58738)

Fixes https://github.com/llvm/llvm-project/issues/58738.

2 years ago[mlir][linalg] Fix crash in canonicalization pattern
Matthias Springer [Wed, 2 Nov 2022 11:36:54 +0000 (12:36 +0100)]
[mlir][linalg] Fix crash in canonicalization pattern

This crash was due to incorrect usage of `hasTensorSemantics`, which has changed recently with DestinationStyleOpInterface. An op has tensor semantics if all of its inits and inputs are tensors. Previously, only inits needed to be tensors.

Differential Revision: https://reviews.llvm.org/D137243

2 years ago[Test] Add test exposing crash in SimpleLoopUnswitch
Dmitry Makogon [Wed, 2 Nov 2022 12:55:56 +0000 (19:55 +0700)]
[Test] Add test exposing crash in SimpleLoopUnswitch

The test crashes with an assert in unswitchNontrivialInvariants.

2 years ago[InstCombine] use logical-and matcher to avoid crash
Sanjay Patel [Wed, 2 Nov 2022 12:21:42 +0000 (08:21 -0400)]
[InstCombine] use logical-and matcher to avoid crash

Follow-on to:
ec0b406e16c44f1554

This should prevent crashing for example like issue #58552
by not matching a select-of-vectors-with-scalar-condition.

The test that shows a regression seems unlikely to occur
in real code.

This also picks up an optimization in the case where a real
(bitwise) logic op is used. We could already convert some
similar select ops to real logic via impliesPoison(), so
we don't see more diffs on commuted tests. Using commutative
matchers (when safe) might also handle one of the TODO tests.

2 years ago[InstCombine] add tests for logical-and / logical-or folds; NFC
Sanjay Patel [Wed, 2 Nov 2022 12:14:44 +0000 (08:14 -0400)]
[InstCombine] add tests for logical-and / logical-or folds; NFC

Similar to 29661fe94bf12ced1 - there are matching
deficiencies and a potential crash lurking in
these patterns.

2 years ago[wasm] Always treat DWARF expression addresses as load addresses
Philip Pfaffe [Wed, 2 Nov 2022 10:51:18 +0000 (10:51 +0000)]
[wasm] Always treat DWARF expression addresses as load addresses

When resolving absolute addresses for DW_OP_addr or DW_OP_addrx, these are always load addresses rather than file addresses in wasm.

Reviewed By: DavidSpickett

Differential Revision: https://reviews.llvm.org/D135664

2 years agoReenable POSIX builtin library functions in gnu2x mode
Aaron Ballman [Wed, 2 Nov 2022 11:56:43 +0000 (07:56 -0400)]
Reenable POSIX builtin library functions in gnu2x mode

gnu17 and earlier modes automatically expose several POSIX C APIs, and
this was accidentally disabled for gnu2x in
7d644e1215b376ec5e915df9ea2eeb56e2d94626.

This restores the behavior for gnu2x mode (without changing the
behavior in C standards modes instead of GNU modes).

Fixes #56607

2 years ago[InstCombine] Remove redundant splats in InstCombineVectorOps
Matt Devereau [Wed, 12 Oct 2022 09:20:05 +0000 (09:20 +0000)]
[InstCombine] Remove redundant splats in InstCombineVectorOps

Splatting the first vector element of the result of a BinOp, where any of the
BinOp's operands are the result of a first vector element splat can be simplified to
splatting the first vector element of the result of the BinOp

Differential Revision: https://reviews.llvm.org/D135876

2 years ago[AMDGPU][NFC] Split MC tests into promoted from VOP1 to VOP3 and only VOP3
Mirko Brkusanin [Tue, 1 Nov 2022 15:32:55 +0000 (16:32 +0100)]
[AMDGPU][NFC] Split MC tests into promoted from VOP1 to VOP3 and only VOP3

Differential Revision: https://reviews.llvm.org/D136149

2 years ago[AMDGPU][NFC] Split MC tests into promoted from VOP2 to VOP3 and only VOP3
Mirko Brkusanin [Tue, 1 Nov 2022 15:22:45 +0000 (16:22 +0100)]
[AMDGPU][NFC] Split MC tests into promoted from VOP2 to VOP3 and only VOP3

Differential Revision: https://reviews.llvm.org/D136148

2 years ago[RuntimeDyld] Require x86_64-linux for IFUNC test
Moritz Sichert [Wed, 2 Nov 2022 11:08:52 +0000 (12:08 +0100)]
[RuntimeDyld] Require x86_64-linux for IFUNC test

This test was introduced by a667aa4de041816cb4865bce8f523228f2332ffa. It
assumes that it runs on x86_64 on linux, so require that for the test.

2 years ago[AMDGPU] Constify a couple of methods. NFC.
Jay Foad [Wed, 2 Nov 2022 11:03:59 +0000 (11:03 +0000)]
[AMDGPU] Constify a couple of methods. NFC.

2 years ago[LoopPredication][NFCI] Perform 'visited' check before pushing to worklist
Max Kazantsev [Wed, 2 Nov 2022 11:02:59 +0000 (18:02 +0700)]
[LoopPredication][NFCI] Perform 'visited' check before pushing to worklist

This prevents duplicates to be pushed into the stack and hypothetically
should reduce memory footprint on ugly cornercases with multiple repeating
duplicates in 'and' tree.

2 years ago[clangd] Index unscoped enums in class scope for code completion
Tom Praschan [Wed, 2 Nov 2022 11:50:50 +0000 (12:50 +0100)]
[clangd] Index unscoped enums in class scope for code completion

Fixes https://github.com/clangd/clangd/issues/1082

Differential Revision: https://reviews.llvm.org/D136925

2 years ago[mlir] Make `OperationFingerPrint` class public
Ivan Butygin [Mon, 31 Oct 2022 13:16:03 +0000 (14:16 +0100)]
[mlir] Make `OperationFingerPrint` class public

It can be useful to external users as well for detecting if there were any changes in IR between passes.

Differential Revision: https://reviews.llvm.org/D137078

2 years ago[MachineCombiner] Add `const` to `shouldReduceRegisterPressure` arguments. NFC
Anton Sidorenko [Tue, 1 Nov 2022 16:32:39 +0000 (19:32 +0300)]
[MachineCombiner] Add `const` to `shouldReduceRegisterPressure` arguments. NFC

Differential Revision: https://reviews.llvm.org/D137174

2 years ago[RuntimeDyld] Added support for relocation of indirect functions
Moritz Sichert [Wed, 20 Jan 2021 16:08:47 +0000 (17:08 +0100)]
[RuntimeDyld] Added support for relocation of indirect functions

In ELF, symbols of type STT_GNU_IFUNC need to be resolved by calling the
function at the symbol's address. This is implemented by adding special
stubs for all symbols of that type.

Differential Revision: https://reviews.llvm.org/D105465

2 years ago[ValueLattice] Use DL-aware folding in getCompare()
Nikita Popov [Wed, 2 Nov 2022 09:41:11 +0000 (10:41 +0100)]
[ValueLattice] Use DL-aware folding in getCompare()

Use DL-aware ConstantFoldCompareInstOperands() API instead of
ConstantExpr API. The practical effect of this is that SCCP can
now fold comparisons that require DL.

2 years ago[ValueLattice] Move getCompare() out of line (NFC)
Nikita Popov [Wed, 2 Nov 2022 09:33:44 +0000 (10:33 +0100)]
[ValueLattice] Move getCompare() out of line (NFC)

This is a fairly large method that is unlikely to benefit from
inlining.

2 years ago[SCCP] Add test for icmp that requires DL to fold (NFC)
Nikita Popov [Wed, 2 Nov 2022 09:30:04 +0000 (10:30 +0100)]
[SCCP] Add test for icmp that requires DL to fold (NFC)

2 years agoReapply [ValueLattice] Fix getCompare() for undef values
Nikita Popov [Tue, 1 Nov 2022 15:39:20 +0000 (16:39 +0100)]
Reapply [ValueLattice] Fix getCompare() for undef values

Relative to the previous attempt, this also updates the
ValueLattice unit tests.

-----

Resolve the TODO about incorrect getCompare() behavior. This can
be made more precise (e.g. by materializing the undef value and
performing constant folding on it), but for now just return an
unknown result to fix the correctness issue.

This should be NFC in terms of user-visible behavior, because the
only user of this method (SCCP) was already guarding against
UndefValue results.

2 years ago[AA] Remove some overloads (NFC)
Nikita Popov [Tue, 1 Nov 2022 15:23:26 +0000 (16:23 +0100)]
[AA] Remove some overloads (NFC)

Having all these instruction-specific overloads does not seem to
provide any compile-time benefit, so drop them in favor of the
generic methods accepting "const Instruction *". Only leave behind
the per-instruction AAQI overloads, which are part of the internal
implementation.

2 years ago[mlir][transform] Add PrintOp to transform dialect
Matthias Springer [Wed, 2 Nov 2022 09:09:46 +0000 (10:09 +0100)]
[mlir][transform] Add PrintOp to transform dialect

This op dumps the associated payload IR to stderr. It has proven useful for printf-style debugging.

Differential Revision: https://reviews.llvm.org/D137151

2 years ago[reland][libc] Switch to new implementation of mem* functions
Guillaume Chatelet [Wed, 2 Nov 2022 09:03:58 +0000 (09:03 +0000)]
[reland][libc] Switch to new implementation of mem* functions

The new framework makes it explicit which processor feature is being
used and allows for easier per platform customization:
 - ARM cpu now uses trivial implementations to reduce code size.
 - Memcmp, Bcmp and Memmove have been optimized for x86
 - Bcmp has been optimized for aarch64.

This is a reland of https://reviews.llvm.org/D135134 (b3f1d58, 028414881381)

Reviewed By: courbet

Differential Revision: https://reviews.llvm.org/D136595

2 years agoRevert "[FLANG] Fix MSVC + clang-cl build"
Muhammad Omair Javaid [Wed, 2 Nov 2022 09:06:43 +0000 (13:06 +0400)]
Revert "[FLANG] Fix MSVC + clang-cl build"

This reverts commit 870fbf8e300f3012470593ac087219a04d52857f.

2 years ago[ConstraintElimination] Do not crash on vector GEP in decomposeGEP
Bjorn Pettersson [Tue, 1 Nov 2022 18:55:23 +0000 (19:55 +0100)]
[ConstraintElimination] Do not crash on vector GEP in decomposeGEP

Commit 359bc5c541ae4b02 caused
 Assertion `isa<To>(Val) && "cast<Ty>() argument of incompatible type!"'
failures in decomposeGEP when the GEP pointer operand is a vector.

Fix is to use DataLayout::getIndexTypeSizeInBits when fetching the
index size, as it will use the scalar type in case of a ptr vector.

Differential Revision: https://reviews.llvm.org/D137185

2 years ago[AArch64][SVE2] Add the SVE2.1 ld1q gather & st1q scatter instructions
David Sherwood [Tue, 1 Nov 2022 15:09:26 +0000 (15:09 +0000)]
[AArch64][SVE2] Add the SVE2.1 ld1q gather & st1q scatter instructions

This patch adds the assembly/disassembly for the following instructions:

ld1q : Gather load quadwords
st1q : Scatter store quadwords

The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2022-09

Differential Revision: https://reviews.llvm.org/D137167

2 years ago[libc] Improve testing of mem functions
Guillaume Chatelet [Thu, 27 Oct 2022 17:01:14 +0000 (17:01 +0000)]
[libc] Improve testing of mem functions

This patch extracts the testing logic from `op_tests.cpp` into
`memory_check_utils.h` so we can reuse it for mem* function integration
tests.

This makes testing consistent and thorough.
For instance this catches a bug that got unnoticed during submission of
D136595 and D135134. Integration test for memcmp was only testing a
single size.

This also leverages ASAN to make sure that data is not read / written
outside permitted boundaries

Differential Revision: https://reviews.llvm.org/D136865

2 years ago[AArch64][SVE2] Add the SVE2.1 while & pext predicate pair instructions
David Sherwood [Wed, 26 Oct 2022 10:40:02 +0000 (10:40 +0000)]
[AArch64][SVE2] Add the SVE2.1 while & pext predicate pair instructions

This patch adds the assembly/disassembly for the following
predicate pair instructions:

pext:    Set pair of predicates from predicate-as-counter
whilelt: While incrementing signed scalar less than scalar
whilele: While incrementing signed scalar less than or equal to scalar
whilegt: While incrementing signed scalar greater than scalar
whilege: While incrementing signed scalar greater than or equal to scalar
whilelo: While incrementing unsigned scalar lower than scalar
whilels: While incrementing unsigned scalar lower or same as scalar
whilehs: While decrementing unsigned scalar higher or same as scalar
whilehi: While decrementing unsigned scalar higher than scalar

The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2022-09

Differential Revision: https://reviews.llvm.org/D136759

2 years ago[FLANG] Fix MSVC + clang-cl build
Muhammad Omair Javaid [Mon, 31 Oct 2022 20:53:16 +0000 (00:53 +0400)]
[FLANG] Fix MSVC + clang-cl build

Flang build on windows with MSVC environment and clang-cl compiler
requires clang_rt.builtin.${target} library. This patch allows us to
locate and include this link library. This is mostly needed for flang
runtime and associated unittests. Clang implicitly uses this library
while calling lld-link via clang driver however a standalone call to
lld-link doesnt link against clang_rt library on its own.

Differential Revision: https://reviews.llvm.org/D137112

2 years ago[CLANG] XFAIL c-strings.c & volatile-1.c AArch64/Windows
Muhammad Omair Javaid [Wed, 2 Nov 2022 08:29:21 +0000 (12:29 +0400)]
[CLANG] XFAIL c-strings.c & volatile-1.c AArch64/Windows

c-strings.c and volatile-1.c failing due to alignment issue on WoA. I am
going to mark them as XFAIL for now.

2 years ago[asan][test] Convert some tests to use opaque pointers
Fangrui Song [Wed, 2 Nov 2022 08:26:48 +0000 (01:26 -0700)]
[asan][test] Convert some tests to use opaque pointers

2 years ago[X86] Use default attributes for more intrinsics
Nikita Popov [Tue, 1 Nov 2022 13:24:41 +0000 (14:24 +0100)]
[X86] Use default attributes for more intrinsics

This is a continuation of D136939, adding the default attributes
(nosync, nofree, nocallback and willreturn) to more readnone
X86 intrinsics.

This still doesn't cover all of them...

Differential Revision: https://reviews.llvm.org/D137163

2 years ago[LoongArch] Inline one-time-used variable and format codes. NFC.
gonglingqin [Tue, 1 Nov 2022 11:31:30 +0000 (19:31 +0800)]
[LoongArch] Inline one-time-used variable and format codes. NFC.

2 years ago[flang] lower intrinsic constants to HLFIR
Jean Perier [Wed, 2 Nov 2022 07:36:45 +0000 (08:36 +0100)]
[flang] lower intrinsic constants to HLFIR

Use the utility to lower Constant<T> that was split from current lowering
in https://reviews.llvm.org/D136955.

The difference in HLFIR is the addition of a fir.declare on constant
outlined in memory so that all the information about them is available.

Lowering to HLFIR is enabled in Brideg::genExprValue to allow testing
of scalar constant lowering.

Differential Revision: https://reviews.llvm.org/D137084

2 years ago[libc] Add definitions of a few missing macros and types.
Siva Chandra Reddy [Tue, 1 Nov 2022 22:27:21 +0000 (22:27 +0000)]
[libc] Add definitions of a few missing macros and types.

2 years ago[hwasan] Remove no-op setDSOLocal. NFC
Fangrui Song [Wed, 2 Nov 2022 07:17:16 +0000 (00:17 -0700)]
[hwasan] Remove no-op setDSOLocal. NFC

PrivateLinkage and HiddenVisibility are implicitly dso_local.

2 years ago[CSKY] Fix the adjustFixupValue of fixup_csky_pcrel_uimm7_scale4
Zi Xuan Wu (Zeson) [Mon, 31 Oct 2022 10:07:37 +0000 (18:07 +0800)]
[CSKY] Fix the adjustFixupValue of fixup_csky_pcrel_uimm7_scale4

The logic to calculate the offset of lrw16 is complex and the Value
before adjustFixupValue should in range of 0 <= (Value >> 2) <= 0xfe.

It also influences the relax condition.

2 years ago[CSKY] Fix .bss directive compling error created by --save-temps
Zi Xuan Wu (Zeson) [Fri, 28 Oct 2022 02:26:32 +0000 (10:26 +0800)]
[CSKY] Fix .bss directive compling error created by --save-temps

.bss without symbol and num following can't be compiled in CSKY target, which is consistent to GCC behavior.
So when --save-temps generates .bss, we should UsesELFSectionDirectiveForBSS to avoid such error.

2 years ago[compiler-rt] Switch from llvm-config to find_package(LLVM)
Michał Górny [Sat, 29 Oct 2022 17:03:37 +0000 (19:03 +0200)]
[compiler-rt] Switch from llvm-config to find_package(LLVM)

Replace the use of the deprecated `llvm-config` tool with LLVM's CMake
files for detecting LLVM in standalone builds.

Differential Revision: https://reviews.llvm.org/D137024

2 years ago[llvm-debuginfo-analyzer] Fix memory leak reported by sanitizers.
Carlos Alberto Enciso [Tue, 1 Nov 2022 12:41:04 +0000 (12:41 +0000)]
[llvm-debuginfo-analyzer] Fix memory leak reported by sanitizers.

The sanitizer reported memory leak issues; the command line
used in the test case is:

llvm-debuginfo-analyzer --attribute=level
                        --print=instructions
                        pr-incorrect-instructions-dwarf-clang.o

When dealing with logical instruction lines associated with
an artificial logical scope, skip the process of finding
their enclosing scope. Just add them to the scope.

Create logical debug lines only if the command line specifies:
  --print=lines or --print=elements or --print=all

Reviewed By: jryans, vitalybuka

Differential Revision: https://reviews.llvm.org/D137156

2 years ago[clang][modules][deps] System module maps might not be affecting
Jan Svoboda [Wed, 2 Nov 2022 04:57:42 +0000 (21:57 -0700)]
[clang][modules][deps] System module maps might not be affecting

The dependency scanner relies on the module map filtering logic in `ASTWriter`. The algorithm currently considers all system module maps affecting, which is not only sub-optimal, but can also cause failures when building a module explicitly (see attached test case).

This patch applies the same filtering logic to system module maps.

Reviewed By: Bigcheese

Differential Revision: https://reviews.llvm.org/D136007

2 years ago[AArch64] Improve codegen for shifted mask op
chenglin.bi [Wed, 2 Nov 2022 05:11:35 +0000 (13:11 +0800)]
[AArch64] Improve codegen for shifted mask op

The special case for bit extraction pattern is  `((x >> C) & mask) << C`.
It can be combined to `x & (mask << C)` by return true in isDesirableToCommuteWithShift.

Fix: #56427

Reviewed By: dmgreen

Differential Revision: https://reviews.llvm.org/D136014

2 years ago[mlir][vector] Fix distribution of scf.for with value coming from above
Thomas Raoux [Tue, 1 Nov 2022 06:25:47 +0000 (06:25 +0000)]
[mlir][vector] Fix distribution of scf.for with value coming from above

When a value used in the forOp is defined outside the region but within
the parent warpOp we need to return and distribute the value to pass it
to new operations created within the loop.
Also simplify the lambda interface.

Differential Revision: https://reviews.llvm.org/D137146

2 years ago[Attributor][NFCI] Move MemIntrinsic handling into the initializer
Johannes Doerfert [Fri, 7 Oct 2022 18:43:23 +0000 (11:43 -0700)]
[Attributor][NFCI] Move MemIntrinsic handling into the initializer

The handling for MemIntrinsic is not dependent on optimistic
information, no need to put it in update for now. Added a TODO though.

2 years ago[Attributor][NFC] Hide verbose output behind `attributor-verbose`
Johannes Doerfert [Fri, 7 Oct 2022 18:37:37 +0000 (11:37 -0700)]
[Attributor][NFC] Hide verbose output behind `attributor-verbose`

2 years ago[OpenMP] Utilize the "non-uniform-workgroup" to simplify DeviceRTL
Johannes Doerfert [Fri, 7 Oct 2022 13:34:29 +0000 (06:34 -0700)]
[OpenMP] Utilize the "non-uniform-workgroup" to simplify DeviceRTL

OpenMP offloading always uses uniform workgroups, see
https://reviews.llvm.org/D135374. The runtime doesn't need to handle
non-uniform workgroups at all either.

Differential Revision: https://reviews.llvm.org/D135444

2 years agoRevert "[RISCV] Enable the LocalStackSlotAllocation pass support"
Craig Topper [Wed, 2 Nov 2022 03:03:41 +0000 (20:03 -0700)]
Revert "[RISCV] Enable the LocalStackSlotAllocation pass support"

This reverts commit 82c820b95cf7ec284baf182cf838ca9e26758098.

This failed llvm-testsuite in our downstream and a similar issue
was reported by @rogfer01.

2 years ago[clang][modules] Account for non-affecting inputs in `ASTWriter`
Jan Svoboda [Wed, 2 Nov 2022 02:23:51 +0000 (19:23 -0700)]
[clang][modules] Account for non-affecting inputs in `ASTWriter`

In D106876, we stopped serializing module map files that didn't affect compilation of the current module.

However, since each `SourceLocation` is simply an offset into `SourceManager`'s global buffer of concatenated input files in, these need to be adjusted during serialization. Otherwise, they can incorrectly point after the buffer or into subsequent input file.

This patch starts adjusting `SourceLocation`s, `FileID`s and other `SourceManager` offsets in `ASTWriter`.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D136624

2 years ago[clang][modules] NFCI: Avoid unnecessary serialization logic for non-affecting files
Jan Svoboda [Wed, 2 Nov 2022 02:19:06 +0000 (19:19 -0700)]
[clang][modules] NFCI: Avoid unnecessary serialization logic for non-affecting files

This patch delays some `ASTWriter` logic until after we've checked whether the source location entry we're serializing as an affecting file or not.

Depends on D137214.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D137216

2 years ago[libc++] Use stack buffers for uninitialized storage in tests.
Konstantin Varlamov [Wed, 2 Nov 2022 02:20:08 +0000 (19:20 -0700)]
[libc++] Use stack buffers for uninitialized storage in tests.

This makes the tests more minimal, and in particular it avoids relying on a complete `<cstdlib>`, which may not be available on all platforms.

Differential Revision: https://reviews.llvm.org/D137188

2 years ago[clang] NFC: Extract lower-level SourceManager functions
Jan Svoboda [Wed, 2 Nov 2022 02:11:24 +0000 (19:11 -0700)]
[clang] NFC: Extract lower-level SourceManager functions

This is a prep-patch for D136624 which allows querying `SourceManager` with raw offsets instead of `SourceLocation`s.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D137215

2 years ago[clang][modules] NFCI: Scaffolding for serialization of adjusted SourceManager offsets
Jan Svoboda [Wed, 2 Nov 2022 02:06:54 +0000 (19:06 -0700)]
[clang][modules] NFCI: Scaffolding for serialization of adjusted SourceManager offsets

This patch is a NFC prep for D136624, where we start adjusting offsets into `SourceManager`.

Depends on D137213.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D137214

2 years ago[clang][modules] NFCI: Pragma diagnostic mappings: write/read FileID instead of Sourc...
Jan Svoboda [Wed, 2 Nov 2022 01:57:41 +0000 (18:57 -0700)]
[clang][modules] NFCI: Pragma diagnostic mappings: write/read FileID instead of SourceLocation

For pragma diagnostic mappings, we always write/read `SourceLocation` with offset 0. This is equivalent to just writing a `FileID`, which is exactly what this patch starts doing.

Depends on D137211.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D137213

2 years ago[clang][modules] NFCI: Unify FileID writing/reading
Jan Svoboda [Wed, 2 Nov 2022 01:45:55 +0000 (18:45 -0700)]
[clang][modules] NFCI: Unify FileID writing/reading

This patch adds new functions for writing/reading `FileID`s and uses them to replace some ad-hoc code.

Reviewed By: dexonsmith

Differential Revision: https://reviews.llvm.org/D137211

2 years agoLangRef: Attempt to formulate some rules for addrspacecast
Matt Arsenault [Tue, 18 Jun 2019 23:03:09 +0000 (19:03 -0400)]
LangRef: Attempt to formulate some rules for addrspacecast

Some work will be needed to comply with these rules for non-integral
pointers.

2 years agoRevert "[Sanitizers] Modified __aarch64__ to use the 64 bit version of the allocator."
Vitaly Buka [Wed, 2 Nov 2022 00:04:10 +0000 (17:04 -0700)]
Revert "[Sanitizers] Modified __aarch64__ to use the 64 bit version of the allocator."

We need to land asan_allocator.h patches first.

This reverts commit 75a8cdbc0659dc20746f31721cc48ef00c49f746.

2 years ago[mlir][sparse] Fix windows build.
bixia1 [Tue, 1 Nov 2022 23:55:24 +0000 (16:55 -0700)]
[mlir][sparse] Fix windows build.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D137210

2 years ago[clang][modules] NFC: Remove unused lambda argument
Jan Svoboda [Tue, 1 Nov 2022 23:53:46 +0000 (16:53 -0700)]
[clang][modules] NFC: Remove unused lambda argument

2 years ago[llvm-profgen] Fix a typo in collectProfiledFunctions
Hongtao Yu [Tue, 1 Nov 2022 22:17:57 +0000 (15:17 -0700)]
[llvm-profgen] Fix a typo in collectProfiledFunctions

As titled. The change should have minimal impact since the targets of branch samples are mostly covered by range samples.

Reviewed By: wenlei, wlei

Differential Revision: https://reviews.llvm.org/D137203

2 years agoFix DenseMap with APInt keys
Weverything [Tue, 4 Oct 2022 21:34:10 +0000 (14:34 -0700)]
Fix DenseMap with APInt keys

The empty key value for APInt was colliding with a valid zero-width
APInt.  Change the internal value of empty key and tombstone values
for APInt to avoid this collision.

Fixes: https://github.com/llvm/llvm-project/issues/58013

Differential Revision: https://reviews.llvm.org/D135741

2 years ago[libc][Obvious] Fix a typo in a CMake file.
Siva Chandra Reddy [Tue, 1 Nov 2022 23:18:34 +0000 (23:18 +0000)]
[libc][Obvious] Fix a typo in a CMake file.

2 years ago[libc] Skip setjmp tests under sanitizers.
Siva Chandra Reddy [Tue, 1 Nov 2022 23:13:57 +0000 (23:13 +0000)]
[libc] Skip setjmp tests under sanitizers.

2 years ago[AArch64][GlobalISel] Add a simple cross-regclass copy optimization post-selection.
Amara Emerson [Tue, 25 Oct 2022 03:51:24 +0000 (20:51 -0700)]
[AArch64][GlobalISel] Add a simple cross-regclass copy optimization post-selection.

This does some trivial cross-regclass folding, where we can either do some extra
constraining to eliminate the copy or modify uses to use a smaller regclass.

There are minor code size improvements on average.

Program                                       size.__text
                                              before         after           diff
tramp3d-v4/tramp3d-v4                         366000.00      366012.00       0.0%
mafft/pairlocalalign                          248196.00      248188.00      -0.0%
7zip/7zip-benchmark                           568612.00      568592.00      -0.0%
kimwitu++/kc                                  434704.00      434676.00      -0.0%
Bullet/bullet                                 456128.00      456096.00      -0.0%
sqlite3/sqlite3                               284136.00      284100.00      -0.0%
ClamAV/clamscan                               381492.00      381396.00      -0.0%
SPASS/SPASS                                   412052.00      411944.00      -0.0%
lencod/lencod                                 428060.00      427912.00      -0.0%
consumer-typeset/consumer-typeset             413148.00      411116.00      -0.5%
                           Geomean difference                               -0.1%

Differential Revision: https://reviews.llvm.org/D136793

2 years ago[libc] Add x86_64 implementation of setjmp and longjmp.
Siva Chandra Reddy [Tue, 1 Nov 2022 06:18:12 +0000 (06:18 +0000)]
[libc] Add x86_64 implementation of setjmp and longjmp.

Reviewed By: michaelrj

Differential Revision: https://reviews.llvm.org/D137147

2 years ago[mlir][sparse] Add rewriting rule for the convert operator.
bixia1 [Mon, 31 Oct 2022 04:55:25 +0000 (21:55 -0700)]
[mlir][sparse] Add rewriting rule for the convert operator.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D136301

2 years agoHonor LLVM_LIBDIR_SUFFIX
serge-sans-paille [Sun, 30 Oct 2022 17:25:04 +0000 (18:25 +0100)]
Honor LLVM_LIBDIR_SUFFIX

Some distribution install libraries under lib64. LLVM supports this
through LLVM_LIBDIR_SUFFIX, have bolt do the same.

Differential Revision: https://reviews.llvm.org/D137039

2 years ago[mlir][TOSA]Add optional attributes to TOSA custom op
Eric Kunze [Tue, 1 Nov 2022 22:09:56 +0000 (15:09 -0700)]
[mlir][TOSA]Add optional attributes to TOSA custom op

The implementation_attrs attr allows passing of backend specific
attributes to TOSA custom ops.

Also adds a config option to avoid namespace collisions on the
identifier.

Reviewed By: rsuderman

Differential Revision: https://reviews.llvm.org/D137133

2 years ago[llvm-c] Harmonize usage of unwrap (NFC)
Aaron Puchert [Tue, 1 Nov 2022 21:26:40 +0000 (22:26 +0100)]
[llvm-c] Harmonize usage of unwrap (NFC)

Some places were using unwrap<T>(x) = cast<T>(unwrap(x)), even though
the unwrapped value already had type T. Removing the template argument
makes it clear that no cast is intended.

Other places were using cast<T>(unwrap(x)), we replace that with the
shorthand unwrap<T>(x) for consistency.

2 years agoAdd tests which need right Insert Point for merged load
bipmis [Tue, 1 Nov 2022 21:49:44 +0000 (21:49 +0000)]
Add tests which need right Insert Point for merged load

2 years ago[mlir][Linalg] Drop usage of tileWithLinalgTilingOptions in the structured.tile transform
Nicolas Vasilache [Thu, 14 Jul 2022 15:55:03 +0000 (08:55 -0700)]
[mlir][Linalg] Drop usage of tileWithLinalgTilingOptions in the structured.tile transform

This is on a path to deprecation.
Context: https://discourse.llvm.org/t/psa-retire-tileandfuselinalgops-method/63850

As the interface-based transformation is more generic, some additional folding of AffineMin/MaxOp and some extra canonicalizations are needed.
This can be further evolved.

Differential Revision: https://reviews.llvm.org/D137195

2 years agoPatch up attributes on a newly enabled test
Paul Robinson [Tue, 1 Nov 2022 21:14:13 +0000 (14:14 -0700)]
Patch up attributes on a newly enabled test

2 years ago[lldb] Document QSetDetachOnError packet
Felipe de Azevedo Piovezan [Fri, 28 Oct 2022 15:39:17 +0000 (11:39 -0400)]
[lldb] Document QSetDetachOnError packet

The packet was introduced in 106d02866d4d54b09c08e6a12915dba58e709294.

Differential Revision: https://reviews.llvm.org/D136958

2 years agoFix release note indentation; NFC
Aaron Ballman [Tue, 1 Nov 2022 20:54:08 +0000 (16:54 -0400)]
Fix release note indentation; NFC

2 years ago[flang] Add atomic_or to list of intrinsics
Katherine Rasmussen [Mon, 31 Oct 2022 17:53:04 +0000 (10:53 -0700)]
[flang] Add atomic_or to list of intrinsics

Add the atomic subroutine, atomic_or, to the list of
intrinsic subroutines, add its last dummy argument to a check
for a coindexed-object, and update test.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D137097

2 years ago[lit][REQUIRES] Fix some tests with incorrect REQUIRES clauses
Paul Robinson [Mon, 24 Oct 2022 15:34:30 +0000 (08:34 -0700)]
[lit][REQUIRES] Fix some tests with incorrect REQUIRES clauses

These weren't running anywhere because of bad specifications.
One test has bit-rotted and had to be XFAILed, the rest are okay.

Differential Revision: https://reviews.llvm.org/D136612

2 years ago[Sanitizers] Modified __aarch64__ to use the 64 bit version of the allocator.
Kirill Stoimenov [Mon, 31 Oct 2022 23:29:39 +0000 (23:29 +0000)]
[Sanitizers] Modified __aarch64__ to use the 64 bit version of the allocator.

This change will switch SizeClassAllocator32 to SizeClassAllocator64 on ARM. This might potentially affect ARM platforms with 39-bit address space. This addresses [[ https://github.com/google/sanitizers/issues/703  | issues/703  ]], but unlike [[ https://reviews.llvm.org/D60243 | D60243 ]] it defaults to 64 bit allocator.

Reviewed By: vitalybuka, MaskRay

Differential Revision: https://reviews.llvm.org/D137136