Kai Sasaki [Tue, 2 May 2023 02:47:26 +0000 (11:47 +0900)]
[mlir][tosa] Typo in transpose canonicalization message
NAKAMURA Takumi [Fri, 17 Mar 2023 20:04:22 +0000 (05:04 +0900)]
Introduce `llvm-min-tblgen` to build public header files
`llvm-min-tblgen` is capable of building `llvm/include/llvm`;
- `-gen-attrs`
- `-gen-directive-*`
- `-gen-intrinsics-*`
- `-gen-riscv-target-def`
`llvm-min-tblgen` is built and used only when `llvm-tblgen` is built in-tree.
This is not installed.
`llvm-tblgen` is built with complete set and may be installed.
`check-llvm` uses not `llvm-min-tblgen` but `llvm-tblgen`.
`LLVM_TABLEGEN_PROJECT` overrides the definition of `tablegen(project)`.
`LLVM_HEADERS` is used as the overridden prefix for LLVM header generators.
If `EXPORT` is not specified in `add_tablegen`, its tablegen is treated as internal.
Let `llvm-tblgen` depend on `intrinsics_gen`
Depends on D149072
Differential Revision: https://reviews.llvm.org/D146352
NAKAMURA Takumi [Thu, 27 Apr 2023 13:15:46 +0000 (22:15 +0900)]
[CMake] add_tablegen: Rename host tablegen to `${target}-host"
Since `llvm-min-tblgen` has been introduced, `llvm-min-tblgen-host`
would make more sense than `LLVM_HEADERS-tablegen-host`.
Depends on D146352
Differential Revision: https://reviews.llvm.org/D149343
Philip Reames [Tue, 2 May 2023 01:47:28 +0000 (18:47 -0700)]
[RISCV] Use vslide1down idiom for generic build_vector
We had previously been going through the stack.
A couple small notes:
We have the vslide1down idiom in a few other places. As a post patch, I plan to try to common the code a bit.
VF=2 case is still going through the splat + insert path. Picking the optimal sequence for this seems to be a bit fiddly (due to constant mat costs), so I restricted this to cases which would have previously hit the stack.
I'm only handling integer vectors for the moment. Mostly because I don't see the existing vfslide1down ISD nodes being in place. This will be an obvious followup.
One of the test diffs does expose a missing combine - a build_vector with a prefix coming from a vector extract sequence. The code after this is arguably worse (due to domain crossing vs stack store), but I think this is a narrow enough case to be non-blocking for now. Let me know if you disagree.
Differential Revision: https://reviews.llvm.org/D149263
Philip Reames [Tue, 2 May 2023 00:49:35 +0000 (17:49 -0700)]
[RISCV][InsertVSETVLI] Avoid VL toggles for extractelement patterns
Noticed this while looking at some SLP output. If we have an extractelement, we're probably using a slidedown into an destination with no contents. Given this, we can allow the slideup to use a larger VL and clobber tail elements of the destination vector. Doing this allows us to avoid vsetvli toggles in many fixed length vector examples.
Differential Revision: https://reviews.llvm.org/D148834
Ben Shi [Mon, 1 May 2023 08:33:40 +0000 (16:33 +0800)]
[ELF][test] Improve AVR tests
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D149318
KOMATA Manabu [Tue, 2 May 2023 01:34:09 +0000 (18:34 -0700)]
[clang][Driver] Fix crash with unsupported architectures in MinGW and CrossWindows
MinGW and CrossWindows crash if the target is an unsupported target architecture.
Changed it to emit an error message.
Fixes https://github.com/llvm/llvm-project/issues/59545
Reviewed By: MaskRay, mstorsjo
Differential Revision: https://reviews.llvm.org/D148944
Fangrui Song [Tue, 2 May 2023 01:22:49 +0000 (18:22 -0700)]
[Driver] Drop incorrect hint from err_target_unknown_triple
For an unknown target triple, the ", please use -triple or -arch" hint of the
diagnostic is almost always wrong.
For cc1, the error is due to a specified but unknown -triple.
For driver, -triple and -arch are not driver options.
Stefan Pintilie [Mon, 1 May 2023 17:11:47 +0000 (13:11 -0400)]
[PowerPC] Implement DFP add and sub instructions.
Add the following Decimal Floating Point (DFP) instructions for PowerPC.
dadd, daddq, dsub, dsubq
In order to add these instructions a new register class for a pair
of floating point registers is added.
This patch is only to allow the user to specify the instructions in
assembly. There is no scheduling or patterns for the instructions.
Reviewed By: amyk
Differential Revision: https://reviews.llvm.org/D148597
Nikolas Klauser [Fri, 28 Apr 2023 23:26:09 +0000 (16:26 -0700)]
[libc++][PSTL] Simplify pstl_config.h
This is done by removing configurations we don't support.
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D149499
LLVM GN Syncbot [Mon, 1 May 2023 23:27:58 +0000 (23:27 +0000)]
[gn build] Port
a50e54fbeb48
Artem Belevich [Thu, 27 Apr 2023 18:10:37 +0000 (11:10 -0700)]
[CUDA] Temporarily undefine __noinline__ when including bits/shared_ptr_base.h
This avoid CUDA compilation errors caused by CUDA headers defining __noinline__
and conflicting with __noinline__ use in libstdc++.
Differential Revision: https://reviews.llvm.org/D149364
Valentin Clement [Mon, 1 May 2023 23:23:46 +0000 (16:23 -0700)]
[flang][openacc] Remove unused variable
Fix buildbot failure after landing D149385
Jorge Gorbe Moya [Mon, 1 May 2023 22:57:53 +0000 (15:57 -0700)]
Re-land "[lldb] Make the libcxx unique_ptr prettyprinter support custom deleters."
This reverts commit
45351120105a7257ccb1e38ec1b1f8a452269da2.
Alex Langford [Fri, 28 Apr 2023 20:23:55 +0000 (13:23 -0700)]
[lldb] Change ObjectValueDictionary to use a StringMap
llvm has a structure for maps where the key's type is a string. Using
that also means that the keys for OptionValueDictionary don't stick
around forever in ConstString's StringPool (even after they are gone).
The only thing we lose here is ordering: iterating over the map where the keys
are ConstStrings guarantees that we iterate in alphabetical order.
StringMap makes no guarantees about the ordering when you iterate over
the entire map.
Differential Revision: https://reviews.llvm.org/D149482
Valentin Clement [Mon, 1 May 2023 23:15:34 +0000 (16:15 -0700)]
[flang][openacc] Add basic support for derived type component in data operand
Add lowering support for derived type component appearing
in a data clause to the newly added data operand operations.
Depends on D149297
Reviewed By: razvanlupusoru, jeanPerier
Differential Revision: https://reviews.llvm.org/D149385
Aart Bik [Mon, 1 May 2023 19:56:26 +0000 (12:56 -0700)]
[mlir][sparse] relax constraints on tensor.cast with pre-rewriting
Reviewed By: wrengr
Differential Revision: https://reviews.llvm.org/D149489
Shilei Tian [Mon, 1 May 2023 23:01:33 +0000 (19:01 -0400)]
[OpenMP] Make `libomptarget` link against `libomp`
In `libomptarget` we use a couple of functions from `libomp`, but we didn't link
`libomptarget` against `libomp`. That will not work on some platforms such
as macOS. A linker error will be encountered because those symbols are not resolved
at link time when building `libomptarget`. This patch simply makes `libomptarget`
link agains `libomp`, makes it a "user" of `libomp`. I think this will not break
the policies between `libomp` and `libomptarget`.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D149617
Mehdi Amini [Mon, 1 May 2023 22:55:58 +0000 (15:55 -0700)]
Revert "Introduce MLIR Op Properties"
This reverts commit
d572cd1b067f1177a981a4711bf2e501eaa8117b.
Some bots are broken and investigation is needed before relanding.
Shilei Tian [Mon, 1 May 2023 22:25:11 +0000 (18:25 -0400)]
[OpenMP] Handle function calls from `libomp` to `libomptarget` correctly
D132005 introduced function calls from `libomp` to `libomptarget` if offloading
is enabled. However, the external function declaration may not always work. For
example, it causes a link error on macOS. Currently it is guarded properly by
a macro, but in order to get OpenMP target offloading working on macOS, it has
to be handled correctly. This patch applies the same idea of how we support
target memory extension by using function pointer indirect call for that function.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D149557
Mehdi Amini [Sun, 26 Feb 2023 15:46:01 +0000 (10:46 -0500)]
Introduce MLIR Op Properties
This new features enabled to dedicate custom storage inline within operations.
This storage can be used as an alternative to attributes to store data that is
specific to an operation. Attribute can also be stored inside the properties
storage if desired, but any kind of data can be present as well. This offers
a way to store and mutate data without uniquing in the Context like Attribute.
See the OpPropertiesTest.cpp for an example where a struct with a
std::vector<> is attached to an operation and mutated in-place:
struct TestProperties {
int a = -1;
float b = -1.;
std::vector<int64_t> array = {-33};
};
More complex scheme (including reference-counting) are also possible.
The only constraint to enable storing a C++ object as "properties" on an
operation is to implement three functions:
- convert from the candidate object to an Attribute
- convert from the Attribute to the candidate object
- hash the object
Optional the parsing and printing can also be customized with 2 extra
functions.
A new options is introduced to ODS to allow dialects to specify:
let usePropertiesForAttributes = 1;
When set to true, the inherent attributes for all the ops in this dialect
will be using properties instead of being stored alongside discardable
attributes.
The TestDialect showcases this feature.
Another change is that we introduce new APIs on the Operation class
to access separately the inherent attributes from the discardable ones.
We envision deprecating and removing the `getAttr()`, `getAttrsDictionary()`,
and other similar method which don't make the distinction explicit, leading
to an entirely separate namespace for discardable attributes.
Differential Revision: https://reviews.llvm.org/D141742
Craig Topper [Mon, 1 May 2023 22:26:17 +0000 (15:26 -0700)]
[RISCV] Don't store CPUKind in CPUInfo. NFC
This field is never used today. If you have a pointer to the row
you can find the CPUKind by subtracting the pointer from the start
of the array.
Craig Topper [Mon, 1 May 2023 22:26:09 +0000 (15:26 -0700)]
[RISCV] Remove INVALID from the list of CPUs in RISCVTargetParser. NFC
This value is never used outside and is only used as a sentinel
internally which we can solve with other means.
Craig Topper [Mon, 1 May 2023 22:26:03 +0000 (15:26 -0700)]
[RISCV] Remove or simplify some StringSwitches in RISCVTargetParser.cpp. NFC
We can iterate over the RISCVCPUInfo table instead of using a
separate StringSwitch.
Noah Goldstein [Mon, 1 May 2023 20:37:23 +0000 (15:37 -0500)]
Recommit "[ValueTracking] Use knownbits interface for determining if `div`/`rem` are safe to speculate" (2nd Try)
Add `poison` checks that where missing.
Reviewed By: nikic
Noah Goldstein [Mon, 1 May 2023 20:37:11 +0000 (15:37 -0500)]
[InstCombine] Fix bug in `FoldOpIntoSelect` where we would incorrectly fold `undef` as constant
D146349 Introduced the ability to use the information from the
`select` condition to deduce constants as we folded a binop into
select. I.e if the `select` cond was `icmp eq %A, 10`, then in the
true-arm of `select`, we would be able to replace usage of `A` with
`10`.
This is broken for vectors that contain `undef` elements. I.e with
`icmp eq %A, <10, undef>`, subsituting `<10, undef>` for `A` can
result in creating a more undefined result than we otherwise would
have.
We fix the issue with simply checking if the candidate constant for
substituting may contain `undef` elements and don't do it in that
case.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D149592
Nikolas Klauser [Fri, 28 Apr 2023 18:02:46 +0000 (11:02 -0700)]
[libc++][PSTL] Fix clang-tidy
Reviewed By: ldionne, #libc
Spies: libcxx-commits, miyuki, carlosgalvezp
Differential Revision: https://reviews.llvm.org/D149500
Shilei Tian [Mon, 1 May 2023 22:22:23 +0000 (18:22 -0400)]
Revert "[OpenMP] Handle function calls from `libomp` to `libomptarget` correctly"
This reverts commit
479e335fc37c06767654141358ea076ac066de11.
The assertion at `kmp_tasking.cpp(29)` is triggered.
Shilei Tian [Mon, 1 May 2023 22:16:32 +0000 (18:16 -0400)]
[OpenMP] Handle function calls from `libomp` to `libomptarget` correctly
D132005 introduced function calls from `libomp` to `libomptarget` if offloading
is enabled. However, the external function declaration may not always work. For
example, it causes a link error on macOS. Currently it is guarded properly by
a macro, but in order to get OpenMP target offloading working on macOS, it has
to be handled correctly. This patch applies the same idea of how we support
target memory extension by using function pointer indirect call for that function.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D149557
LLVM GN Syncbot [Mon, 1 May 2023 21:30:38 +0000 (21:30 +0000)]
[gn build] Port
ade9c3bdca2c
Louis Dionne [Mon, 1 May 2023 21:25:14 +0000 (17:25 -0400)]
[libc++][NFC] Fix typos in documentation
Jim Ingham [Mon, 24 Apr 2023 23:08:27 +0000 (16:08 -0700)]
Improve the help output for `type XXX delete` where XXX is summary, format,
etc...
Differential Revision: https://reviews.llvm.org/D148282
Corentin Jabot [Mon, 23 Jan 2023 21:05:48 +0000 (22:05 +0100)]
[Clang] Fix a crash when recursively callig a default member initializer.
This fixes a regression introduced by
ca61961380, that would lead
to a segfault due to stack exhaustion when recursively calling
a default member initializer.
Fixes #60082
I'm not able to get clang to emit a stack exhaustion warning,
which it seems like it should be able to.
Reviewed By: shafik
Differential Revision: https://reviews.llvm.org/D142401
Jorge Gorbe Moya [Mon, 1 May 2023 21:13:28 +0000 (14:13 -0700)]
Revert "[lldb] Make the libcxx unique_ptr prettyprinter support custom deleters."
This reverts commit
d366da97bd24ddfb91c9f260fa0aaf105d947652.
Nikolas Klauser [Mon, 1 May 2023 21:11:07 +0000 (14:11 -0700)]
[libc++][NFC] Fix synopsis in a few tests
Jonas Devlieghere [Mon, 1 May 2023 21:08:43 +0000 (14:08 -0700)]
[lldb] Add debugger.external-editor setting
Add a new setting (debugger.external-editor) to specify an external
editor. The setting takes precedence over the existing
LLDB_EXTERNAL_EDITOR environment variable.
Differential revision: https://reviews.llvm.org/D149565
Hanhan Wang [Fri, 28 Apr 2023 18:16:26 +0000 (11:16 -0700)]
[mlir][linalg] Relax masked_vectorize(pad) to allow zero low Values.
It only accepted zeros attributes before the revision. It now acccepts
constant-like zero values as well.
Reviewed By: dcaballe
Differential Revision: https://reviews.llvm.org/D149474
Nikolas Klauser [Fri, 28 Apr 2023 23:00:14 +0000 (16:00 -0700)]
[libc++][PSTL] Implement std::fill{,_n}
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D149540
Philip Reames [Mon, 1 May 2023 20:54:09 +0000 (13:54 -0700)]
[LV] Refresh some auto-gen tests to reduce diff [nfc]
Teresa Johnson [Mon, 1 May 2023 18:04:59 +0000 (11:04 -0700)]
[MemProf] Recognize hot/cold operator new as replaceable allocations
Follow up to LLVM-side change to allow conversion to hot/cold hinted
operator new, not yet standard but supported by open source tcmalloc.
The LLVM change in
a35206d78280e0ebcf48cd21bc5fff5c3b9c73fa added the
necessary support to recognize these as library functions, and we
subsequently found that a clang-side change is needed to give them
builtin/nobuiltin attributes as appropriate so they have consistent
removeability.
Based on discussion with Richard Smith, converted the parameter type to
a reserved identifier (
39f7b48671dae5fbe3afc49f33f50c2b6340bb89) and
added support in this patch to recognize it in clang.
Differential Revision: https://reviews.llvm.org/D149600
Jorge Gorbe Moya [Tue, 18 Apr 2023 20:19:45 +0000 (13:19 -0700)]
[lldb] Make the libcxx unique_ptr prettyprinter support custom deleters.
The unique_ptr prettyprinter calls `GetValueOfLibCXXCompressedPair`,
which looks for a `__value_` child. However, when the second value in
the compressed pair is not an empty class, there are two `__value_`
children because `__compressed_pair` derives twice from
`__compressed_pair_elem`, one for each member of the pair. And then the
lookup fails because it's ambiguous.
This patch makes the following changes:
- Rename `GetValueOfLibCXXCompressedPair` to
`GetFirstValueOfLibCXXCompressedPair`, and add a similar function to
get the second value. Put both functions in
Plugin/Language/CPlusPlus/LibCxx.cpp because it seems inappropriate to
have libcxx-specific helpers separate from all the libcxx-dependent
code.
- Read the second value of the `__ptr_` pair and display a "deleter"
child in the unique_ptr synthetic child provider, when available.
- Add a test case for the non-empty deleter case.
Differential Revision: https://reviews.llvm.org/D148662
Craig Topper [Mon, 1 May 2023 20:00:05 +0000 (13:00 -0700)]
[RISCV] Merge RISCV::parseCPUKind and RISCV::checkCPUKind.
Similar for RISCV::parseTuneCPU and RISCV::checkTuneCPUKind.
This makes the CPUKind enum no longer part of the API. It wasn't
providing much value. It was only used to pass between the two
functions.
By removing it, we can remove a dependency on a tablegen generated
file from the RISCVTargetParser.h file. Then we can remove a
dependency from several CMakeLists.txt.
Arthur Eubanks [Tue, 25 Apr 2023 20:16:01 +0000 (13:16 -0700)]
Reland [clang] Canonicalize system headers in dependency file when -canonical-prefixes
Clang was writing paths to the dependency file that don't exist when using a sysroot with symlinks, causing everything to get rebuilt every time. This is reproducible on Linux by creating a symlink to '/', using that as the sysroot, and trying to build something with ninja that includes the C++ stdlib (e.g. a typical build of LLVM).
This fixes https://github.com/ninja-build/ninja/issues/1330 and somewhat matches gcc.
gcc canonicalizes system headers in dependency files under a -f[no-]canonical-system-headers, but it makes more sense to look at -canonical-prefixes.
D37954 was a previous attempt at this.
Fixed use of %T instead of %t in test, causing bots to fail the test on the initial commit.
Reviewed By: hans
Differential Revision: https://reviews.llvm.org/D149187
Nikolas Klauser [Sat, 29 Apr 2023 00:19:41 +0000 (17:19 -0700)]
[libc++][PSTL] Remove _PSTL_HIDE_FROM_ABI_{PUSH,POP} macros
Reviewed By: ldionne, #libc
Spies: libcxx-commits, miyuki
Differential Revision: https://reviews.llvm.org/D149556
LLVM GN Syncbot [Mon, 1 May 2023 19:12:51 +0000 (19:12 +0000)]
[gn build] Port
13fe6733019a
Craig Topper [Mon, 1 May 2023 19:05:16 +0000 (12:05 -0700)]
[RISCV] Move NTLH hint emission into RISCVAsmPrinter.cpp.
Rather than having a separate pass to add the hint instructions,
emit them directly into the streamer during asm printing.
Reviewed By: BeMg, kito-cheng
Differential Revision: https://reviews.llvm.org/D149511
Peiming Liu [Mon, 1 May 2023 18:58:22 +0000 (18:58 +0000)]
[mlir][sparse] fix compiler warning in release build.
Differential Revision: https://reviews.llvm.org/D149607
LLVM GN Syncbot [Mon, 1 May 2023 18:48:32 +0000 (18:48 +0000)]
[gn build] Port
a595b931f1f9
Ian Anderson [Mon, 17 Apr 2023 16:38:38 +0000 (09:38 -0700)]
[libc++] cuchar redeclares ::mbstate_t when it's in its own clang module
When cuchar is compiled independently on platforms that don't have <uchar.h>, it doesn't get a declaration of mbstate_t, and so makes an empty declaration for ::mbstate_t. That conflicts with the declarations in __mbstate_t.h and cwchar since both of those headers do get mbstate_t before making ::mbstate_t.
Change `__mbstate_t.h` to just get the underlying declaration for mbstate_t and not make a declaration for ::mbstate_t. Include __mbstate_t.h in uchar.h and wchar.h when their next headers aren't available so that at least mbstate_t gets defined.
Add __std_mbstate_t.h to declare ::mbstate_t for headers that need that when cuchar or cwchar aren't available (because either _LIBCPP_HAS_NO_WIDE_CHARACTERS or _LIBCPP_CXX03_LANG).
Reviewed By: ldionne, Mordante, #libc, philnik
Differential Revision: https://reviews.llvm.org/D148542
Jakub Kuderski [Mon, 1 May 2023 18:31:20 +0000 (14:31 -0400)]
[mlir][arith] Add narrowing patterns to commute more vector ops
This commutes the extension (`arith.extsi`, `arith.extui`) over the
following vector ops: `vector.broadcast`, `vector.shape_cast`,
`vector.transpose`, `vector.flat_transpose`.
I focused on these as I saw them getting created by vector unroll
patterns. Maybe except `vector.flat_transpose`.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D149534
Jakub Kuderski [Mon, 1 May 2023 18:29:01 +0000 (14:29 -0400)]
[mlir][arith] Add narrowing patterns for other insertion ops
Allow to commute extension ops over `vector.insertelement` and
`vector.insert_strided_slice`.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D149509
Valentin Clement [Mon, 1 May 2023 18:28:20 +0000 (11:28 -0700)]
[flang][openacc][NFC] Fix typos segement -> segment
Noah Goldstein [Mon, 1 May 2023 18:20:09 +0000 (13:20 -0500)]
Revert "[ValueTracking] Use knownbits interface for determining if `div`/`rem` are safe to speculate"
Appears to be causing out-of-tree test failures. Reverting while the
issue is investigated.
This reverts commit
fbc7fcf5ae26fad7db3e7c861fc6447e02b39cf0.
Peiming Liu [Fri, 21 Apr 2023 20:06:36 +0000 (20:06 +0000)]
[mlir][sparse] extend unpack operation to support unpacking a batched COO type
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D149103
Arthur Eubanks [Mon, 1 May 2023 17:55:23 +0000 (10:55 -0700)]
Revert "[clang] Canonicalize system headers in dependency file when -canonical-prefixes"
This reverts commit
8f0dd4ef3ed2137d1e2554204127434fc46ee190.
Causes bot failures: https://lab.llvm.org/buildbot/#builders/139/builds/39964
Philip Reames [Mon, 1 May 2023 17:38:44 +0000 (10:38 -0700)]
[LAA] Add command line flag to disable unit stride speculation
This is purely so that we can expose and work through downstream codegen issues. My intention is to see if we can get this disabled by default, but that requires fixing a bunch of downstream issues first.
Arthur Eubanks [Tue, 25 Apr 2023 20:16:01 +0000 (13:16 -0700)]
[clang] Canonicalize system headers in dependency file when -canonical-prefixes
Clang was writing paths to the dependency file that don't exist when using a sysroot with symlinks, causing everything to get rebuilt every time. This is reproducible on Linux by creating a symlink to '/', using that as the sysroot, and trying to build something with ninja that includes the C++ stdlib (e.g. a typical build of LLVM).
This fixes https://github.com/ninja-build/ninja/issues/1330 and somewhat matches gcc.
gcc canonicalizes system headers in dependency files under a -f[no-]canonical-system-headers, but it makes more sense to look at -canonical-prefixes.
D37954 was a previous attempt at this.
Reviewed By: hans
Differential Revision: https://reviews.llvm.org/D149187
Jakub Kuderski [Mon, 1 May 2023 17:31:30 +0000 (13:31 -0400)]
[mlir][arith][NFC] Simplify narrowing patterns with a wrapper type
Add a new wraper type that represents either of `ExtSIOp` or `ExtUIOp`.
This is to simplify the code by using a single type, so that we do not
have to use templates or branching to handle both extension kinds.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D149485
Alan Zhao [Mon, 1 May 2023 17:26:05 +0000 (10:26 -0700)]
Fix ReleaseNotes.rst unresolved merge conflict
Douglas Yung [Mon, 1 May 2023 17:22:53 +0000 (10:22 -0700)]
Revert "Host: generalise `GetXcodeSDKPath`"
This reverts commit
ade3c6a6a88ed3a9b06c076406f196da9d3cc1b9.
This breaks the build with GCC and affects at least 2 build bots:
https://lab.llvm.org/buildbot/#/builders/217/builds/20568
https://lab.llvm.org/buildbot/#/builders/243/builds/5576
Philip Reames [Mon, 1 May 2023 17:16:32 +0000 (10:16 -0700)]
[LAA] Rework overflow checking in getPtrStride [nfc]
The previous code structure and comments were exceedingly confusing. I have multiple times looked at this code and suspected a bug. This time, I decided to take the time to reflow the code and comment out why it is correct.
The only suspect (to me) case left is that an underaligned access with a unit stride (in terms of the access type) might miss the undefined null pointer when wrapping. This is unlikely to be an issue for C/C++ code with real page sizes, so I'm not bothering to fully convince myself whether that case is correct or not.
William Huang [Sat, 29 Apr 2023 00:30:25 +0000 (00:30 +0000)]
[llvm-profdata] Deprecate Compact Binary Sample Profile Format
Remove support for compact binary sample profile format
Reviewed By: davidxl, wenlei
Differential Revision: https://reviews.llvm.org/D149400
Alan Zhao [Thu, 13 Apr 2023 18:04:59 +0000 (11:04 -0700)]
[clang] Fix overly aggressive lifetime checks for parenthesized aggregate initialization
Before this patch, initialized class members would have the LifetimeKind
LK_MemInitializer, which does not allow for binding a temporary to a
reference. Binding to a temporary however is allowed in parenthesized
aggregate initialization, even if it leads to a dangling reference. To
fix this, we create a new EntityKind, EK_ParenAggInitMember, which has
LifetimeKind LK_FullExpression.
This patch does *not* attempt to diagnose dangling references as a
result of using this feature.
This patch also refactors TryOrBuildParenListInitialization(...) to
accomodate creating different InitializedEntity objects.
Fixes #61567
[0]: https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p0960r3.html
Reviewed By: shafik
Differential Revision: https://reviews.llvm.org/D148274
Slava Zakharin [Mon, 1 May 2023 16:53:45 +0000 (09:53 -0700)]
[flang][hlfir] Fixed boxchar variables emboxing.
We have to unbox the boxchar variables in order to get to
the raw address that can be used for emboxing.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D149473
Slava Zakharin [Mon, 1 May 2023 16:06:02 +0000 (09:06 -0700)]
[flang][hlfir] Fixed lowering for optional dummy.
We have to keep it as a box, since taking box_addr of the optional
box may be invalid.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D149505
Yingwei Zheng [Mon, 1 May 2023 16:06:26 +0000 (00:06 +0800)]
[InstCombine] Combine const GEP chains
This patch reverts rGae739aefd7473517d3f08b5c8d08a66c7f469198 to address performance regressions reported by our [CI](https://github.com/dtcxzyw/llvm-ci/issues/137) after rG2ec1d0f427c7822540352c0c14d057e7bfe4f77b.
For example:
```
define ptr @const_gep_chain(ptr %p, i64 %a) {
%p1 = getelementptr inbounds i8, ptr %p, i64 %a
%p2 = getelementptr inbounds i8, ptr %p1, i64 1
%p3 = getelementptr inbounds i8, ptr %p2, i64 2
%p4 = getelementptr inbounds i8, ptr %p3, i64 3
ret ptr %p4
}
```
The last three GEPs will not be folded since rG2ec1d0f427c7822540352c0c14d057e7bfe4f77b.
I think it is appropriate to remove this code because there is no compile-time regression reported in our benchmarks.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D149240
Alan Zhao [Thu, 27 Apr 2023 22:31:51 +0000 (15:31 -0700)]
[clang] Fix default initializers being ignored when initializing templated aggregate types
Previously, when checking whether an in-class initializer exists when
performing parenthesized aggregate initialization, Clang checks that the
output of FieldDecl::getInClassInitializer() is non-null. This is
incorrect; if the field is part of a templated type, then
getInClassInitializer() will return nullptr if we haven't called
Sem::BuildCXXDefaultInitExpr(...) before, even if
FieldDecl::hasInClassInitializer() returns true. The end result is that
Clang incorrectly ignores the in class initializer and
value-initializes the field. The fix therefore is to instead call
FieldDecl::hasInClassInitializer(), which is what we do for braced init
lists [0].
Before this patch, Clang does correctly recognize the in-class field
initializer in certain cases. This is Sema::BuildCXXDefaultInitExpr(...)
populates the in class initializer of the corresponding FieldDecl
object. Therefore, if that method was previously called with the same
FieldDecl object, as can happen with a decltype(...) or a braced list
initialization, FieldDecl::getInClassInitializer() will return a
non-null expression, and the field becomes properly initialized.
Fixes 62266
[0]: https://github.com/llvm/llvm-project/blob/
be5f35e24f4c15caf3c4aeccddc54c52560c28a0/clang/lib/Sema/SemaInit.cpp#L685
Reviewed By: shafik
Differential Revision: https://reviews.llvm.org/D149389
LLVM GN Syncbot [Mon, 1 May 2023 16:04:47 +0000 (16:04 +0000)]
[gn build] Port
824391693cbb
Nikolas Klauser [Thu, 27 Apr 2023 23:28:22 +0000 (16:28 -0700)]
[libc++][PSTL] Implement std::find{,_if,_if_not}
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D149539
Philip Reames [Mon, 1 May 2023 15:32:28 +0000 (08:32 -0700)]
[LAA] Use early return [nfc]
Muiez Ahmed [Mon, 1 May 2023 15:25:14 +0000 (11:25 -0400)]
[SystemZ][z/OS] Mark ecsymbols test as unsupported
The llvm-lib tool is not yet supported on z/OS.
Differential Revision: https://reviews.llvm.org/D149591
Zarko Todorovski [Mon, 1 May 2023 14:30:29 +0000 (10:30 -0400)]
Set EnableAIXExtendedAltivecABI in BackendUtils from LangOpts
Fix a bug where after
github.com/llvm/llvm-project/commit/
68dd51421f16f1e17cd453cb1730fcca99a6cfb7
refactor where we are not passing -mabi=vec-extabi to th backend.
Philip Reames [Mon, 1 May 2023 14:32:55 +0000 (07:32 -0700)]
[RISCV] Rewrite isLegalElementTypeForRVV in terms of ValueTypes [nfc]
This was briefly mentioned as a possibility in review discussion on D149369.
One slightly surprising bit to call out - these interfaces can get invoked with non-vector typed arguments. LoopVectorizer likes to call the costing interfaces with scalar types when unrolling, but not vectorizing. I found that surprising, not sure if others do.
Differential Revision: https://reviews.llvm.org/D149462
Florian Hahn [Mon, 1 May 2023 14:03:15 +0000 (15:03 +0100)]
[VPlan] Remove DeadInsts arg from VPInstructionsToVPRecipes (NFC)
The argument isn't used. VPlan-based dead recipe removal can be used
instead.
Louis Dionne [Mon, 1 May 2023 13:48:13 +0000 (09:48 -0400)]
[clang] Temporarily enable debug output when uploading artifacts during CI jobs
This is an attempt to troubleshoot our excessive usage of network in
Clang jobs on Buildkite.
Doru Bercea [Wed, 26 Apr 2023 14:41:10 +0000 (10:41 -0400)]
Emit info message when use_device_address variable does not have a device counterpart.
Aaron Ballman [Mon, 1 May 2023 11:31:03 +0000 (07:31 -0400)]
Fix the clang sphinx bot
Addresses the issue found in:
https://lab.llvm.org/buildbot/#/builders/92/builds/43478
Christian Sigg [Mon, 1 May 2023 10:04:32 +0000 (12:04 +0200)]
[bazel] Fix layering check after 0610e2f
`BytecodeWriter.h` is now included in `IR.h`, and the corresponding dependency should be moved to `header_deps`. This change fixes Google's internal build, which seems to have a stricter layering check than bazel's, but it's still the right thing to do.
Differential Revision: https://reviews.llvm.org/D149576
Kadir Cetinkaya [Mon, 1 May 2023 10:04:24 +0000 (12:04 +0200)]
[clangd][IncludeCleaner] Use a proper comparator for deduplicating findings
Previous one didn't have strict weak ordering guarantees.
Advenam Tacet [Mon, 1 May 2023 05:13:15 +0000 (22:13 -0700)]
[compiler-rt][ASan] Fix description of __sanitizer_annotate_contiguous_container
This revision updates the description of
`__sanitizer_annotate_contiguous_container` in includes. Possibilites of
the function were changed in D132522 and it supports:
- unaligned beginning,
- shared first/last granule with other objects.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D149341
Jacques Pienaar [Mon, 1 May 2023 05:11:02 +0000 (22:11 -0700)]
[mlir][bytecode] Return error instead of min version
Can't return a well-formed IR output while enabling version to be bumped
up during emission. Previously it would return min version but
potentially invalid IR which was confusing, instead make it return
error and abort immediately instead.
Differential Revision: https://reviews.llvm.org/D149569
Vitaly Buka [Mon, 1 May 2023 04:59:52 +0000 (21:59 -0700)]
[test][HWASAN] Auto-generate a few tests
Vitaly Buka [Mon, 1 May 2023 03:59:05 +0000 (20:59 -0700)]
Revert "[NFC][HWASAN] Handle tags as Int8"
More tests need updates.
This reverts commit
e876ba5db98830db427395ed9b3718d20bf519fb.
Shilei Tian [Mon, 1 May 2023 03:34:31 +0000 (23:34 -0400)]
[OpenMP] Only enable version script if supported
The linker flag `--version-script` may not be supported by all linkers, such as
macOS's linker. `libomp` is already capable of detecting whether the linker supports
it and append the linker flag accordingly. Since currently we assume `libomptarget`
only works on Linux, we don't do the check accordingly. This patch simply adds
the check before adding it to linker flag. This will be the first patch to make
OpenMP target offloading work on macOS. Note that CMake files in `plugins` are
not touched before they are going to be removed pretty soon anyway.
Reviewed By: jhuber6
Differential Revision: https://reviews.llvm.org/D149555
Vitaly Buka [Sun, 30 Apr 2023 18:53:51 +0000 (11:53 -0700)]
[NFC][HWASAN] Handle tags as Int8
Vitaly Buka [Mon, 1 May 2023 02:48:32 +0000 (19:48 -0700)]
[NFC][HWASAN] Rename local variable
Tim Schmielau [Mon, 1 May 2023 01:54:20 +0000 (07:24 +0530)]
Basic documentation of -mrecip=... option
The documentation is still rather terse because it needs to fit
onto a single line of clang --help output.
But at least it lists what the user can specify and documents the
non-obvious syntax.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D149456
Vitaly Buka [Mon, 1 May 2023 02:07:57 +0000 (19:07 -0700)]
[NFC][HWASAN] Set constant type from another operand
Vitaly Buka [Sun, 30 Apr 2023 20:58:25 +0000 (13:58 -0700)]
[test][HWASAN] Remove update_test_checks note
This test is not autogenerated.
Vitaly Buka [Mon, 1 May 2023 01:10:40 +0000 (18:10 -0700)]
[HWASAN] Untag before tagging alloca pointers
This is folloup to
b5595836, which missed the
Replacemen variable.
Before
b5595836 the code assumed that alloca
ptrs are not tagged so tagging is implemented
as simple OR.
So this patch completes support of tagged SP
by passing untagged alloca pointers into
tagPointer.
Joshua Cao [Mon, 1 May 2023 01:17:48 +0000 (18:17 -0700)]
[LangRef] Fix broken tail call optimization link
max [Sun, 30 Apr 2023 22:46:29 +0000 (17:46 -0500)]
[MLIR][tensor] generate default builder for FromElementsOp
Removed builder is the same as default builder, with the added benefit that python bindings will be generated for the default builder.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D149508
Valentin Churavy [Sun, 30 Apr 2023 23:19:17 +0000 (19:19 -0400)]
Don't loop unswitch vector selects
Otherwise we could produce `br <2x i1>` which are of course not legal.
```
Branch condition is not 'i1' type!
br <2 x i1> %cond.fr1, label %entry.split.us, label %entry.split
%cond.fr1 = freeze <2 x i1> %cond
LLVM ERROR: Broken module found, compilation aborted!
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
Stack dump:
0. Program arguments: /home/vchuravy/builds/llvm/bin/opt -passes=simple-loop-unswitch<nontrivial> -S
```
Fixes change introduced by https://reviews.llvm.org/D138526
Reviewed By: caojoshua
Differential Revision: https://reviews.llvm.org/D149560
sstwcw [Sun, 30 Apr 2023 22:26:41 +0000 (22:26 +0000)]
[clang-format] Recognize Verilog type dimension in module header
We had the function `verilogGroupDecl` for that. However, the type
name would be incorrectly annotated in `isStartOfName` when it was not
a C++ keyword and followed another identifier.
Reviewed By: HazardyKnusperkeks, owenpan, MyDeveloperDay
Differential Revision: https://reviews.llvm.org/D149352
sstwcw [Sun, 30 Apr 2023 22:22:31 +0000 (22:22 +0000)]
[clang-format] Correctly format goto labels followed by blocks
There doesn't seem to be an issue on GitHub. But previously, a space
would be inserted before the goto colon in the code below.
switch (x) {
case 0:
goto_0: {
action();
break;
}
}
Previously, the colon following a goto label would be annotated as
`TT_InheritanceColon`. A goto label followed by an opening brace
wasn't recognized. It is easy to add another line to have
`spaceRequiredBefore` function recognize the case, but I believed it
is more proper to avoid doing the same thing in `UnwrappedLineParser`
and `TokenAnnotator`. So now the label colons would be labeled in
`UnwrappedLineParser`, and `spaceRequiredBefore` would rely on that.
Previously we had the type `TT_GotoLabelColon` intended for both goto
labels and case labels. But since handling of goto labels and case
labels differ somewhat, I split it into separate types for goto and
case labels.
This patch doesn't change the behavior for case labels. I added the
lines annotating case labels because they would previously be
mistakenly annotated as `TT_InheritanceColon` just like goto labels.
And since I added the annotations, the checks for the `case` and
`default` keywords in `spaceRequiredBefore` are not necessary anymore.
Reviewed By: MyDeveloperDay
Differential Revision: https://reviews.llvm.org/D148484
Richard Smith [Sun, 30 Apr 2023 22:13:22 +0000 (15:13 -0700)]
Remove bug fix release note for a bug that was never released.
In its place, add a description of a bug that the change introducing
the bug in question inadvertently fixed.
See BZ 62361 for more details.
LLVM GN Syncbot [Sun, 30 Apr 2023 20:27:26 +0000 (20:27 +0000)]
[gn build] Port
68ee014b15d3
Nikolas Klauser [Thu, 19 Jan 2023 19:10:11 +0000 (20:10 +0100)]
[libc++][PSTL] Implement std::for_each{, _n}
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D149381
Florian Hahn [Sun, 30 Apr 2023 20:12:15 +0000 (21:12 +0100)]
[LV] Clarify comment for selectVectorizationFactor (NFC).
The comment is stale, as UserVF is handled before selectVectorizationFactor
is called. Clarify the comment by remove the mention of UserVF.
Suggested as independent improvement in D143938.
Florian Hahn [Sun, 30 Apr 2023 20:10:55 +0000 (21:10 +0100)]
[LV] Remove loop arg from CM::isCandidateForEpilogueVectorization (NFC)
LVP operates on the loop it stores in TheLoop. Use it instead of the
argument, to be in line with other member functions.
Suggested as independent improvement in D143938.