Commit Graph

414520 Commits

Author SHA1 Message Date
Roman Lebedev 3564551400
ProfileDataTests: fix BUILD_SHARED_LIBS build 2022-02-09 12:27:13 +03:00
Balazs Benics 5e0be95fce [analyzer][docs][NFC] Fix some broken links and some cosmetic changes
- We should report bugs to the GitHub Issues
- We should advocate using Discourse instead of the superseded cfe-dev
  mailing list.

There are a couple of other cosmetic changes such as preferring `using`
instead of `typedef` and quoting the `-analyze-function` parameter's
payload for escaping the braces.

Thanks @loic-joly-sonarsource for reporting this on Discord!

Reviewed By: NoQ, Szelethus

Differential Revision: https://reviews.llvm.org/D119245
2022-02-09 10:22:28 +01:00
Nikita Popov 36cae4299d Reapply [sanitizers] Avoid macro clash in SignalContext::WriteFlag (NFC)
D116208 may cause a macro clash on older versions of linux, where
fs.h defines a READ macro. This is resolved by switching to a more
typical casing style for non-macro symbols.

Reapplying with changes to the symbol names in various platform
specific code, which I missed previously.

Differential Revision: https://reviews.llvm.org/D118783
2022-02-09 10:22:05 +01:00
Nikolas Klauser a5b4479d3c [libc++] Add papers from February 2022 plenary meeting
Reviewed By: ldionne, Quuxplusone, #libc

Spies: arichardson, libcxx-commits

Differential Revision: https://reviews.llvm.org/D119189
2022-02-09 10:21:19 +01:00
Valentin Clement 903f6b7170
[flang] Add missing CFI case for REAL and COMPLEX
ISO_Fortran_binding.h was updated with missing entries for CFI
types for REAL and COMPLEX kinds 2,3,10,16. This patch updates TypeCode.h
to use these new types.

This patch is part of the upstreaming effort from fir-dev branch.

Reviewed By: jeanPerier

Differential Revision: https://reviews.llvm.org/D119283
2022-02-09 10:07:40 +01:00
Nikita Popov 34840c1a7d Revert "[sanitizers] Avoid macro clash in SignalContext::WriteFlag (NFC)"
This reverts commit fda29264f3.

This breaks the sanitizer build on windows, will reapply with
additional changes.
2022-02-09 10:07:23 +01:00
Qiu Chaofan c091792cd3 [NFC] Fix SSE3 intrinsics test for PowerPC
Previous test in ppc-pmmintrin.c did not check IR of intrinsic function
definition. Add them and simplify.

These tests shouldn't be auto-generated, because we don't want to check
wrapper functions.
2022-02-09 17:00:34 +08:00
Muhammad Omair Javaid cd817231ec [clang-cl] Bump default -fms-compatibility-version to 19.14
clang-cl MSVC required version is 19.20 now. Update the default
-fms-compatibility-version to 19.14.

Differential Revision: https://reviews.llvm.org/D114639
2022-02-09 13:54:25 +05:00
RVP 62e4a77746 [Support] Fix for two issues with clearing of the internal storage for cl::bits
This patch fixes two issues with clearing of the internal storage for cl::bits

1. The internal bits storage for cl::bits is uninitialized. This is a problem if a cl::bits option is not defined with static lifetime.
2. ResetAllOptionOccurrences does not reset cl::bits options.

The latter is also discussed in:

https://lists.llvm.org/pipermail/llvm-dev/2021-February/148299.html

Differential Revision: https://reviews.llvm.org/D119066
2022-02-09 09:46:46 +01:00
Nikita Popov fda29264f3 [sanitizers] Avoid macro clash in SignalContext::WriteFlag (NFC)
D116208 may cause a macro clash on older versions of linux, where
fs.h defines a READ macro. This is resolved by switching to a more
typical casing style for non-macro symbols.

Differential Revision: https://reviews.llvm.org/D118783
2022-02-09 09:43:28 +01:00
Nikita Popov ff0b391600 [NVPTX] Remove image/sampler special case in call lowering
I suspect that this is dead code. There is no test coverage for
this special case, and the struct type names this checks against
don't seem to match what OpenCL actually generates (which would be
%opencl.sampler_t rather than %struct._sampler_t for example).

Motivation for this change is that this code is incompatible with
opaque pointers -- simply deleting it is the simplest way of
making it compatible :)

Differential Revision: https://reviews.llvm.org/D119229
2022-02-09 09:40:27 +01:00
Andrzej Warzynski 69c3309d45 [flang][driver] Add support for `-emit-mlir`
This patch adds support for generating MLIR files in Flang's frontend
driver (i.e. `flang-new -fc1`). `-emit-fir` is added as an alias for
`-emit-mlir`. We may want to decide to split the two in the future.

A new parent class for code-gen frontend actions is introduced:
`CodeGenAction`. We will be using this class to encapsulate logic shared
between all code-generation actions, but not required otherwise. For
now, it will:
 * run prescanning, parsing and semantic checks,
 * lower the input to MLIR.
`EmitObjAction` is updated to inherit from this class. This means that
the behaviour of `flang-new -fc1 -emit-obj` is also updated (previously,
it would just exit immediately). This change required
`flang/test/Driver/syntax-only.f90` to be updated.

For `-emit-fir`, a specialisation of `CodeGenAction` is introduced:
`EmitMLIRAction`. The key logic for this class is implemented in
`EmitMLIRAction::ExecuteAction`.

Differential Revision: https://reviews.llvm.org/D118985
2022-02-09 08:35:48 +00:00
Nikita Popov 68c1eeb4ba [ArgPromotion] Make implementation offset based
This rewrites ArgPromotion to be based on offsets rather than GEP
structure. We inspect all loads at constant offsets and remember
which types are loaded at which offsets. Then we promote based on
those types.

This generalizes ArgPromotion to work with bitcasted loads, and
is compatible with opaque pointers.

This patch also fixes incorrect handling of alignment during
argument promotion. Previously, the implementation only checked
that the pointer is dereferenceable, but was happy to speculate
overaligned loads. (I would have fixed this separately in advance,
but I found this hard to do with the previous implementation
approach).

Differential Revision: https://reviews.llvm.org/D118685
2022-02-09 09:35:01 +01:00
Jean Perier 47995a0ec9 [flang] catch implicit interface incompatibility with global scope symbol
Previously, when calling a procedure implicitly for which a global scope
procedure symbol with the same name existed, semantics resolved the
procedure name in the call to the global symbol without checking that
the symbol interface was compatible with the implicit interface of the
call.
This could cause expression rewrite and lowering to later badly process
the implicit call assuming a different result type or an explicit
interface. This could lead to lowering crash in case the actual argument
were incompatible with the dummies from the explicit interface.

Emit errors in the following problematic cases:
- If the result type from the symbol did not match the one from the
  implicit interface.
- If the symbol requires an explicit interface.

This patch still allows calling an F77 like procedure with different
actual argument types than the one it was defined with because it is
correctly supported in lowering and is a feature in some program
(it is a pointer cast). The two cases that won't be accepted have
little chance to make much sense. Results returning ABIs may differ
depending on the return types, and function that requires explicit
interface usually requires descriptors or specific processing that
is incompatible with implicit interfaces.

Note that this patch is not making a deep analysis, and it will only
catch mistakes if a global symbol and an implicit interface are
involved. Cases where the user provided a conflicting explicit
interface would still require a pass after name resolution to study
conflicts more deeply. But these cases will not crash lowering or
trigger expression rewrite to do weird things.

Differential Revision: https://reviews.llvm.org/D119274
2022-02-09 09:30:32 +01:00
Rainer Orth b63ca0a068 [Sanitizers][test] XFAIL long double tests on Solaris/sparc
As reported in Issue #41838, `clang` doesn't correctly implement `long
double` on 32-bit Solaris/SPARC: the psABI requires this to be an 128-bit
type.  Four sanitizer tests currently `FAIL` for this reason.

While there is a WIP patch to fix `clang` (D89130
<https://reviews.llvm.org/D89130>), it isn't complete yet and I've hit so
many brick walls while trying to finish it that I'm unsure if I ever will.

This patch therefore `XFAIL`s those tests in the meantime.

Tested on `sparcv9-sun-solaris2.11`.

Differential Revision: https://reviews.llvm.org/D119016
2022-02-09 09:20:21 +01:00
Haojian Wu fe932a88e9 [pseudo] Add first and follow set computation in Grammar.
These will be used when building parsing table for LR parsers.

Separate from https://reviews.llvm.org/D118196.

Differential Revision: https://reviews.llvm.org/D118990
2022-02-09 09:16:27 +01:00
Rainer Orth 0e4b214b8c [sanitizer_common] Don't try to unmap unaligned memory
Enabling `sanitizer_common` tests on Solaris (D91606
<https://reviews.llvm.org/D91606>) and SPARC (D91608
<https://reviews.llvm.org/D91608>) uncovered a sparcv9 failure

  SanitizerCommon-Unit :: ./Sanitizer-sparcv9-Test/CompactRingBuffer.int64

like this:

  [ RUN      ] CompactRingBuffer.int64
  ==24576==ERROR: SanitizerTool failed to deallocate 0x2000 (8192) bytes at address 0xffffffff7f59b000
  ==24576==Sanitizer CHECK failed: /vol/llvm/src/llvm-project/local/compiler-rt/lib/sanitizer_common/sanitizer_posix.cpp:61 (("unable to unmap" && 0)) != (0) (0, 0)

The problem is that the original allocation via
`MmapAlignedOrDieOnFatalError` is for 4 kB, but the Solaris/sparcv9
pagesize is 8 kB.  So the initial allocation is for 12 kB, rounded to a
multiple of the pagesize.  Afterwards, the unneeded rest is unmapped again,
but this fails since the address is not pagesize-aligned.

This patch avoids this by aligning the end of the mapping to the pagesize.

With D91827 <https://reviews.llvm.org/D91827> added, the test `PASS`es on
`sparcv9-sun-solaris2.11`.

Differential Revision: https://reviews.llvm.org/D91615
2022-02-09 09:15:41 +01:00
Rainer Orth 0d4b6f1f4b [sanitizer_common] Fix DenseMapCustomTest.DefaultMinReservedSizeTest on SPARC
As described in Issue #53523, the
`DenseMapCustomTest.DefaultMinReservedSizeTest` test FAILs on Solaris/SPARC
(both 32 and 64-bit):

  /vol/llvm/src/llvm-project/local/compiler-rt/lib/sanitizer_common/tests/sanitizer_dense_map_test.cpp:399:
Failure
  Expected: (MemorySize) != (Map.getMemorySize()), actual: 8192 vs 8192

This happens because SPARC, unlike many other CPUs, uses an 8 kB pagesize.

Fixed by incorporating the pagesize into the calculations of
`ExpectedInitialBucketCount` and derived values.

Tested on `sparcv9-sun-solaris2.11`, `amd64-pc-solaris2.11`, and
`x86_64-pc-linux-gnu`.

Differential Revision: https://reviews.llvm.org/D118771
2022-02-09 09:12:02 +01:00
Tobias Hieta 7cca34acc4 [llvm-rc]: Find <target>-clang over just clang
This patch makes llvm-rc/windres prefer <target>-clang over
clang when doing it's preprocessing. This is so that we can
have a .cfg file for <target> and configure sysroot and other
important flags.

Config files not picked up with clang --target=<target>
automatically.

We only look for <target>-clang in the same dir as llvm-windres
and not for all PATHs to minimize the change.

Reviewed By: mstorsjo

Differential Revision: https://reviews.llvm.org/D119219
2022-02-09 09:04:53 +01:00
Fraser Cormack 6449bea508 [RISCV] Select unmasked RVV pseudos in a DAG post-process
This patch drops TableGen patterns matching all-ones masked RVV pseudos
in the case where there are fallback patterns matching the generic
masked forms to "_MASK" pseudos. This optimization is now performed with
a SelectionDAG post-processing step which peephole-optimizes these same
pseudos with all-ones masks and swaps them out to their unmasked
pseudos.

This cuts our generated ISel table down by around ~5% (~110kB) in lieu
of a far smaller auto-generated table to help with the peephole.

This only targets our custom RISCVISD::*_VL binary operator nodes, which
use the one form for both masked and unmasked variants. A similar
approach could be used for our intrinsics but we'd need to do some work,
e.g., to represent unmasked intrinsics as true-masked intrinsics at the
IR or ISel level. At a rough estimate, this could save us a further 9%
on the size of our ISel table for the binary intrinsic patterns alone.

There is no observable impact on our tests.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D118810
2022-02-09 07:50:15 +00:00
Saiyedul Islam 4db88a54b6
[OpenMP][Clang] Move partial support of reverse offload to a future version
OpenMP Spec 5.2 requires unimplemented requires clauses to produce
compile time error termination. Moving current partial support of
reverse_offload to a distant future version 9.9 so that existing
code can be tested and maintained until a complete implementation
is available.

Reviewed By: ABataev

Differential Revision: https://reviews.llvm.org/D119256
2022-02-09 07:11:41 +00:00
Fangrui Song 99580e29d8 [ELF] --warn-backrefs: suppress warnings for backward references within the archive 2022-02-08 21:45:55 -08:00
Jonas Devlieghere 43d3d886a0 [llvm] Recognize arm64 as target-aarch64 in lit
Recognize arm64 triples as AArch64 so we can XFAIL/skip tests using
target-aarch64.

Differential revision: https://reviews.llvm.org/D119169
2022-02-08 21:29:21 -08:00
Fangrui Song a163cdf6b5 [docs] Re-generate ClangCommandLineReference.rst 2022-02-08 20:54:52 -08:00
Sameer Sahasrabuddhe c6a6b57902 [AMDGPU] [NFC] Fix incorrect use of bitwise operator.
Differential Revision: https://reviews.llvm.org/D119308
2022-02-08 22:12:54 -05:00
Yaxun (Sam) Liu 1d97cb1f6e [HIP] Emit amdgpu_code_object_version module flag
code object version determines ABI, therefore should not be mixed.

This patch emits amdgpu_code_object_version module flag in LLVM IR
based on code object version (default 4).

The amdgpu_code_object_version value is code object version times 100.

LLVM IR with different amdgpu_code_object_version module flag cannot
be linked.

The -cc1 option -mcode-object-version=none is for ROCm device library use
only, which supports multiple ABI.

Reviewed by: Artem Belevich

Differential Revision: https://reviews.llvm.org/D119026
2022-02-08 21:58:40 -05:00
Zakk Chen cfe7f69036 [RISCV][NFC] Refactor RISCVISAInfo.
1. Remove computeDefaultABIFromArch and add computeDefaultABI in
RISCVISAInfo.
2. Add parseFeatureBits which may used in D118333.

Differential Revision: https://reviews.llvm.org/D119250
2022-02-08 18:37:43 -08:00
jacquesguan 5e71bbfb6c [RISCV] Add patterns for vector widening floating-point fused multiply-add instructions
Add patterns for vector widening floating-point fused multiply-add instructions.

Differential Revision: https://reviews.llvm.org/D117546
2022-02-09 10:34:39 +08:00
Jon Roelofs 8440ec66f4 [docs] Document -ffinite-math-only, -fhonor-{infinities,infinites,nans} 2022-02-08 18:10:17 -08:00
Bill Wendling deaf22bc0e [X86] Implement -fzero-call-used-regs option
The "-fzero-call-used-regs" option tells the compiler to zero out
certain registers before the function returns. It's also available as a
function attribute: zero_call_used_regs.

The two upper categories are:

  - "used": Zero out used registers.
  - "all": Zero out all registers, whether used or not.

The individual options are:

  - "skip": Don't zero out any registers. This is the default.
  - "used": Zero out all used registers.
  - "used-arg": Zero out used registers that are used for arguments.
  - "used-gpr": Zero out used registers that are GPRs.
  - "used-gpr-arg": Zero out used GPRs that are used as arguments.
  - "all": Zero out all registers.
  - "all-arg": Zero out all registers used for arguments.
  - "all-gpr": Zero out all GPRs.
  - "all-gpr-arg": Zero out all GPRs used for arguments.

This is used to help mitigate Return-Oriented Programming exploits.

Reviewed By: nickdesaulniers

Differential Revision: https://reviews.llvm.org/D110869
2022-02-08 17:42:54 -08:00
Richard Smith dc8f4e118d [cxx_status] Add a couple of Feb 2022 library papers that need compiler support. 2022-02-08 17:25:52 -08:00
LLVM GN Syncbot f556743434 [gn build] Port 216575e581 2022-02-09 01:18:18 +00:00
Nico Weber 5167f3a73a [gn build] (manually) port 216575e581 2022-02-08 20:17:40 -05:00
Nico Weber 94064aaeb1 [gn build] (manually) port 4a6553f4c2 some more 2022-02-08 20:17:40 -05:00
Mircea Trofin 2868c57caf [nfc][mlgo][regalloc] Add the url to a reference pre-trained model 2022-02-08 16:57:24 -08:00
Nico Weber ae92365397 [gn build] (manually) port 817d897b57 (LIBCXX_ABI_UNSTABLE) 2022-02-08 19:55:18 -05:00
Alexander Yermolovich 0d9921daad [BOLT][DWARF] Remove caching of ranges/abbrevs
Removing caching of ranges/abbrevs to simplify the code.
Before we were doing it to get around a gdb limitation.
FBD34015613

Reviewed By: Amir, maksfb

Differential Revision: https://reviews.llvm.org/D119276
2022-02-08 16:37:40 -08:00
Matt Arsenault 5af0f097ba GlobalISel: Constant fold G_PTR_ADD
Some globals lower to literal addresses on AMDGPU.

This may be wrong for non-integral address spaces. I'm wondering if we
should just allow regular G_ADD to use pointer types, and reserve
G_PTR_ADD for non-integral address spaces.
2022-02-08 19:21:06 -05:00
Alexandre Ganea bb8be26a7e [LLD] Fix issue in HIP due to unspecified order of evaluation of the function object
This fixes the issue raised in https://reviews.llvm.org/D108850#3303452

Before C++17, the function object is evaluated in a unspecified order. In the following example: https://godbolt.org/z/8ao4vdsr7 the function object is either evaluated before or after the arguments, depending on the compiler. With MSVC and /std:c++14 the function object is evaluated after the arguments; with clang and gcc, it is evaluated before. With C++17, the function object is guaranteed to be evaluated before the arguments, see: https://riptutorial.com/cplusplus/example/19369/evaluation-order-of-function-arguments

In our case, the issue was that the `args` conversion to `ArrayRef` was evaluated before the lambda call `link`, which internally was calling `parseFlavor()`, which in turned modified `args`. We ended with an `ArrayRef` argument that reflected the previous contents of `args`.

Add coverage for `-flavor` which we didn't have before.

Differential Revision: https://reviews.llvm.org/D119278
2022-02-08 19:12:15 -05:00
Alexandre Ganea 1e661e583d [MLIR] Temporary workaround for calling the LLD ELF driver as-a-lib
This fixes the situation described in https://github.com/llvm/llvm-project/issues/53475 with a repro exposed by https://github.com/ROCmSoftwarePlatform/D108850-lld-bug-reproduction

This is purposely just a workaround to unblock users. This could be transplanted to the release/14.x branch if need be. A proper fix will later be provided in https://reviews.llvm.org/D119049.

Differential Revision: https://reviews.llvm.org/D119277
2022-02-08 19:12:15 -05:00
tyb0807 9b81d2fae8 Revert "[AArch64] ACLE feature macro for Armv8.8-A MOPS"
This reverts commit d379ec9908.
2022-02-09 00:10:09 +00:00
Joe Loser d7095e8035
[libc++] Remove outdated synopsis parts in experimental/functional
Remove mentions of `experimental::function`, its operators, etc. They are no
longer in `experimental/functional`.

Differential Revision: https://reviews.llvm.org/D119204
2022-02-08 19:06:35 -05:00
Snehasish Kumar cb81545e7d [memprof] Add LLVM_DEBUG for unused var in RawMemProfReader.cpp.
The ContainingSegment variable is only used to check whether we found
the address at this time. When building without Asserts this emits a
warning. So for now wrap this code in LLVM_DEBUG to avoid the warning.
2022-02-08 16:02:24 -08:00
Matt Arsenault 2af4a554fe GlobalISel: Constant fold FP bin ops in MIRBuilder
Might as well handle these if we're going to handle the integer ops
here.
2022-02-08 18:51:10 -05:00
Matt Arsenault 930f2498d4 GlobalISel: Constant fold integer min/max opcodes 2022-02-08 18:50:35 -05:00
Krzysztof Parzyszek 0792161c00 [Hexagon] Fix operation actions for v128f16
There were more cases of operations that should have been "Custom" for
v128f16, but ended up "Legal" (e.g. load and store).
2022-02-08 15:28:37 -08:00
Mogball 740e832644 [mlir][ods] Attribute and type formats: support whitespaces
Supports whitespace elements: ` ` and `\\n` as well as the "empty" whitespace `` that removes an otherwise printed space.

Depends on D118208

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D118210
2022-02-08 23:27:36 +00:00
Matt Arsenault 0877fbcc16 GlobalISel: Add FoldBinOpIntoSelect combine
This will do the combine in cases that should fold, but don't
now. e.g. we're relying on the CSEMIRBuilder's incomplete constant
folding. For instance it doesn't handle FP operations or vectors (and
we don't have separate constant folding combines either to catch
them).
2022-02-08 18:17:21 -05:00
Matt Arsenault 5847d5fb24 AMDGPU/GlobalISel: Add baseline test for binop fold into select combine 2022-02-08 18:17:21 -05:00
Daniel Thornburgh dcd4950d42 [Symbolizer] Add Build ID flag to llvm-symbolizer.
This adds a --build-id=<hex build ID> flag to llvm-symbolizer. If --obj
is unspecified, this will attempt to look up the provided build ID using
whatever mechanisms are available to the Symbolizer (typically,
debuginfod). The semantics are then as if the found binary were given
using the --obj flag.

Reviewed By: jhenderson, phosek

Differential Revision: https://reviews.llvm.org/D118633
2022-02-08 23:08:18 +00:00