Stephen Tozer [Tue, 20 Dec 2022 10:01:56 +0000 (10:01 +0000)]
[DebugInfo] Fix: Variables that have no non-empty values being emitted when they have a DBG_VALUE_LIST
This patch fixes a simple bug where `DbgValueHistoryMap::hasNonEmptyLocation` was incorrectly handling DBG_VALUE_LIST instructions, treating empty values as non-empty, causing empty variables to be emitted into DWARF.
Reviewed By: Orlando
Differential Revision: https://reviews.llvm.org/D133925
LLVM GN Syncbot [Sun, 25 Dec 2022 21:05:12 +0000 (21:05 +0000)]
[gn build] Port
f0756e086010
Vitaly Buka [Sun, 25 Dec 2022 20:49:35 +0000 (12:49 -0800)]
Revert "[clang-format] Add an option to format integer literal separators"
Revert "[clang-format] Disable FixRanges in IntegerLiteralSeparatorTest"
Breaks buildbots, details in https://reviews.llvm.org/D140543
This reverts commit
879bd9146a2c9ea395abd7c1ebd0f76f414a4967.
This reverts commit
46c94e5067b5f396c24bb950505c79bc819bd4b8.
Craig Topper [Sun, 25 Dec 2022 20:40:53 +0000 (12:40 -0800)]
Revert "[RISCV] Enable the LocalStackSlotAllocation pass support."
This reverts commit
180397cdded67a8fdf56f92a0b70d32f0dac8af6.
This seems to cause llvm-testsuite failures.
Nikolas Klauser [Sun, 18 Dec 2022 23:02:44 +0000 (00:02 +0100)]
[libc++] Implement constexpr {isfinite, isinf, isnan, isnormal}
This starts implementing P0533
Reviewed By: Mordante, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D140277
Joseph Huber [Sun, 25 Dec 2022 15:47:04 +0000 (09:47 -0600)]
[OpenMP] Fix test on 32-bit platforms
Summary:
This test didn't specify the triple so it defaulted to the user's, if
this was 32-bit then it failed due to a diagnostic message.
Alexey Lapshin [Sun, 25 Dec 2022 14:25:18 +0000 (15:25 +0100)]
[NFC][ADT] Rename StringMapEntry *Create() into StringMapEntry *create.
Roman Lebedev [Sun, 25 Dec 2022 13:43:20 +0000 (16:43 +0300)]
[NFC][SupportTests] Adjust `UnsignedDivideUsingMagic()` for readability
Alexandros Lamprineas [Sun, 25 Dec 2022 08:05:21 +0000 (10:05 +0200)]
[IPSCCP] Enable specialization of functions.
This patch enables Function Specialization by default at all
optimization levels except Os, Oz.
Compilation Time Overhead:
--------------------------
Measured the Instruction Count increase (Geomean) for CTMark from
the llvm-testsuite as in https://llvm-compile-time-tracker.com.
* {-O3, Non-LTO}: +0.136% Instruction Count
* {-O3, LTO}: +0.346% Instruction Count
Performance Uplift:
-------------------
Measured +9.121% score increase for 505.mcf_r from SPEC Int 2017
(Tested on Neoverse N1 with -O3 + LTO)
Correctness Testing:
--------------------
* Passes bootstrap Clang with ASAN + LTO + FuncSpec aggressive options:
{ MaxClonesThreshold=10,
SmallFunctionThreshold=10,
AvgLoopIterationCount=30,
SpecializeOnAddresses=true,
EnableSpecializationForLiteralConstant=true,
FuncSpecializationMaxIters=10 }
* Builds Chromium and passes its unittests with the above options + ThinLTO.
For more info please refer to
https://discourse.llvm.org/t/rfc-should-we-enable-function-specialization/61518
Differential Revision: https://reviews.llvm.org/D140210
eopXD [Sat, 24 Dec 2022 13:12:35 +0000 (05:12 -0800)]
[Doc] Replace PYTHON_EXECUTABLE with Python3_EXECUTABLE
As topic, the variable to specify the python executable now should be this.
This is probably something that was left out in D78762.
Reviewed By: JDevlieghere
Differential Revision: https://reviews.llvm.org/D140652
LLVM GN Syncbot [Sun, 25 Dec 2022 02:55:37 +0000 (02:55 +0000)]
[gn build] Port
46c94e5067b5
LLVM GN Syncbot [Sun, 25 Dec 2022 02:55:36 +0000 (02:55 +0000)]
[gn build] Port
066b492b747a
Owen Pan [Sun, 25 Dec 2022 02:47:23 +0000 (18:47 -0800)]
[clang-format] Disable FixRanges in IntegerLiteralSeparatorTest
The FixRanges unit test from
46c94e5067b5 breaks the build bots.
Disable it for now.
Kshitij Jain [Mon, 19 Dec 2022 04:07:42 +0000 (04:07 +0000)]
Adds support for GOT relocations to i386/ELF backend
This CR adds support for GOT relocations to the JITLink i386/ELF backend.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D140279
Kshitij Jain [Sun, 20 Nov 2022 04:09:29 +0000 (04:09 +0000)]
Updates and adds tests for i386/ELF JITLink backend
This CR modifies the existing 32 bit pcrel relocation test to
include the case when the relocation target might be present at
a smaller address than the address of the location that needs to be
patched.
Additionally, it adds a test for 16 bit absolute relocation.
Reviewed By: sunho
Differential Revision: https://reviews.llvm.org/D138372
Roman Lebedev [Sat, 24 Dec 2022 21:48:05 +0000 (00:48 +0300)]
[NFC] Add exhaustive test coverage for `{Un}signedDivisionByConstantInfo`
Use this wrapper if you want to try brute-forcing wider bit widths:
https://godbolt.org/z/3xGzTM881
I've brute-forced i16 for both signed and unsigned, and we're all good.
As mentioned in https://reviews.llvm.org/D140636
Owen Pan [Thu, 22 Dec 2022 09:21:17 +0000 (01:21 -0800)]
[clang-format] Add an option to format integer literal separators
Closes #58949.
Differential Revision: https://reviews.llvm.org/D140543
Florian Hahn [Sat, 24 Dec 2022 21:52:28 +0000 (21:52 +0000)]
[ConstraintElim] Add missing CHECK lines.
Florian Hahn [Sat, 24 Dec 2022 21:51:26 +0000 (21:51 +0000)]
[ConstraintElim] Add extra test with chained GEPs without inbounds.
Florian Hahn [Sat, 24 Dec 2022 21:46:46 +0000 (21:46 +0000)]
[ConstraintElim] Convert tests to use opaque pointers (NFC).
Florian Hahn [Sat, 24 Dec 2022 18:33:36 +0000 (18:33 +0000)]
[LV] Use SCEV to check if the trip count <= VF * UF.
Just comparing constant trip counts causes LV to miss cases where the
vector loop body only executes once.
The motivation for this is to remove the need for unrolling to remove
vector loop back-edges, if the body only executes once in more cases.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D133017
Roman Lebedev [Sat, 24 Dec 2022 16:18:50 +0000 (19:18 +0300)]
[NFC][Codegen][X86] Autogenerate check lines in shift-i256.ll
Roman Lebedev [Sat, 24 Dec 2022 16:12:27 +0000 (19:12 +0300)]
[NFC][Codegen][AVR] Make shift.ll autogenerate-able
Roman Lebedev [Sat, 24 Dec 2022 15:39:57 +0000 (18:39 +0300)]
[NFC][Codegen] Add tests with oversized shifts by non-byte-multiple
Roman Lebedev [Sat, 24 Dec 2022 15:39:48 +0000 (18:39 +0300)]
[NFC][Codegen] Rename tests for oversized shifts by byte multiple
Mark de Wever [Sun, 20 Mar 2022 12:40:02 +0000 (13:40 +0100)]
[libc++][chrono] Add calendar type formatters.
Some of the calendar types have landed before, this adds the missing
set. Note this does not complete the implementation of the chrono
formatters.
This removes the `chrono` header for some transitive include in C++17
mode. This is needed to avoid inclusion cycles.
Partially implements:
- P1361 Integration of chrono with text formatting
- P2372 Fixing locale handling in chrono formatters
Reviewed By: #libc, ldionne
Differential Revision: https://reviews.llvm.org/D137022
serge-sans-paille [Sat, 24 Dec 2022 11:53:50 +0000 (12:53 +0100)]
Fix lldb option handling since
e953ae5bbc313fd0cc980ce021d487e5b5199ea4 (part 2)
Option tables are no longer null-terminated.
This is a follow-up to https://reviews.llvm.org/D139881
serge-sans-paille [Sat, 24 Dec 2022 10:56:21 +0000 (11:56 +0100)]
Fix lldb option handling since
e953ae5bbc313fd0cc980ce021d487e5b5199ea4
Option tables are no longer null-terminated.
This is a follow-up to https://reviews.llvm.org/D139881
serge-sans-paille [Mon, 12 Dec 2022 16:02:15 +0000 (17:02 +0100)]
[clang] Use a StringRef instead of a raw char pointer to store builtin and call information
This avoids recomputing string length that is already known at compile
time.
It has a slight impact on preprocessing / compile time, see
https://llvm-compile-time-tracker.com/compare.php?from=
3f36d2d579d8b0e8824d9dd99bfa79f456858f88&to=
e49640c507ddc6615b5e503144301c8e41f8f434&stat=instructions:u
This is a recommit of
719d98dfa841c522d8d452f0685e503538415a53 that into
account a GGC issue (probably
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92181) when dealing with
intiailizer_list and constant expressions.
Workaround this by avoiding initializer list, at the expense of a
temporary plain old array.
Differential Revision: https://reviews.llvm.org/D139881
Vitaly Buka [Sat, 24 Dec 2022 05:29:03 +0000 (21:29 -0800)]
Revert "[DebugInfo] Variables with only empty values emitting when one is variadic"
Breaks HWASAN somehow.
Fails at
def915c39cc4e18b304c7a8c4761cc4531c3bc4b
https://lab.llvm.org/buildbot/#/builders/236/builds/1547
Pass at
def915c39cc4e18b304c7a8c4761cc4531c3bc4b^
https://lab.llvm.org/buildbot/#/builders/236/builds/1529
This reverts commit
def915c39cc4e18b304c7a8c4761cc4531c3bc4b.
Craig Topper [Sat, 24 Dec 2022 04:54:54 +0000 (20:54 -0800)]
[RISCV] Support the short-forward-branch predicated ops in RISCVSExtWRemoval.
Lang Hames [Sat, 24 Dec 2022 05:01:09 +0000 (21:01 -0800)]
[examples] Fix leaks in OrcV2 c-bindings examples.
rdar://
103599609
Phoebe Wang [Sat, 24 Dec 2022 03:40:59 +0000 (11:40 +0800)]
[X86][Reduce] Preserve fast math flags when change it. NFCI
@arsenm raised a good question that we should use a flag guard.
But I found it is not a problem as long as user uses intrinsics only: https://godbolt.org/z/WoYsqqjh3
Anyway, it is still nice to have.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D140467
Stella Stamenova [Sat, 24 Dec 2022 01:31:08 +0000 (17:31 -0800)]
Revert "Apply shortened printing/parsing form to linalg.reduce."
This reverts commit
281c2d49c929c2130e5f1f0e02d6e96dbf14494a.
This broke the windows mlir buildbot:
https://lab.llvm.org/buildbot/#/builders/13/builds/30167
Stella Stamenova [Sat, 24 Dec 2022 01:29:42 +0000 (17:29 -0800)]
Revert "[mlir][GPU] Add known_block_size and known_grid_size to gpu.func"
This reverts commit
85e38d7cd670371206f6067772dc822049d2cbd8.
This broke the windows mlir buildbot:
https://lab.llvm.org/buildbot/#/builders/13/builds/30180/steps/6/logs/stdio
Jez Ng [Sat, 24 Dec 2022 00:44:56 +0000 (19:44 -0500)]
[lld-macho] Standardize error messages
Errors / warnings that originate from a particular file should be of the
form `$file: $message`.
Reviewed By: #lld-macho, keith
Differential Revision: https://reviews.llvm.org/D140634
Roman Lebedev [Sat, 24 Dec 2022 00:32:37 +0000 (03:32 +0300)]
[NFC][Codegen] Add RISCV test coverage for D140638
Owen Anderson [Fri, 23 Dec 2022 04:24:49 +0000 (21:24 -0700)]
Resolve a long-standing FIXME in memcpyopt.
Inspecting the downstream use of the cpyAlign, it is clear that
`performCallSlotOptzn` is expecting it to represent the alignment
of the copy destination, not the minimum of the src and dest
alignments. This patch renames the parameter to make this more
obvious.
I believe this change is NFC, because the downstream code has
alignment checks such that it all works out in the end. I have not
been able to construct a test case that actually triggers a change
in output.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D140603
Roman Lebedev [Fri, 23 Dec 2022 21:27:40 +0000 (00:27 +0300)]
[NFC][Codegen] Tests with wide scalar shifts, for new potential legalization strategy
Craig Topper [Fri, 23 Dec 2022 21:33:20 +0000 (13:33 -0800)]
[Support] Use APInt::udivrem in DivisionByConstantInfo. NFC
David Blaikie [Fri, 23 Dec 2022 20:07:02 +0000 (20:07 +0000)]
Add explicit template instantiation declarations for existing explicit definitions.
Explicit instantiations should be declared. Found with -Wundefined-func-template.
Reviewed By: dblaikie, rriddle
Differential Revision: https://reviews.llvm.org/D140594
Jez Ng [Fri, 23 Dec 2022 19:50:58 +0000 (14:50 -0500)]
[lld-macho] Only fold private-label aliases that do not have flags
This will enable us to re-land {D139069}.
The issue with the original diff was that we were folding all
private-label symbols. We were not merging the symbol flags during this
folding; instead we just made all references to the folded symbol point
to its aliasee. This caused some flags to be incorrectly discarded. This
surfaced as code that was incorrectly stripped due to LLD dropping the
`.no_dead_strip` flag.
This diff fixes things by only folding flag-less private-label aliases.
Most (maybe all) of the `ltmp<N>` symbols that are generated by the MC
aarch64 backend are flag-less, so this conservative folding behavior
does the job.
Reviewed By: #lld-macho, thakis
Differential Revision: https://reviews.llvm.org/D140606
Jez Ng [Wed, 21 Dec 2022 22:44:45 +0000 (17:44 -0500)]
[reland][lld-macho] Private label aliases to weak symbols should not retain section data
This reverts commit
a650f2ec7a37cf1f495108bbb313e948c232c29c.
The crashes it was causing will be fixed by the stacked diff {D140606}.
Craig Topper [Fri, 23 Dec 2022 18:25:19 +0000 (10:25 -0800)]
[APInt] Move a comment from urem to srem where it belongs.
This comment talks about sign of the dividend and the result. I
think it belongs to the srem function.
This is the commit that added it
https://github.com/llvm/llvm-project/commit/
709a820a5338ca5c4b949f2f0286f7da39f00661
and the code change with it was for srem.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D140611
Craig Topper [Fri, 23 Dec 2022 18:24:35 +0000 (10:24 -0800)]
[APFloat] Remove workaround for old clang.
The comment says this is for clang 3.3. Our build requirements
are clang 5.0 or newer so I think we can remove this.
Reviewed By: nikic, RKSimon
Differential Revision: https://reviews.llvm.org/D140613
Roman Lebedev [Fri, 23 Dec 2022 17:13:23 +0000 (20:13 +0300)]
[NFC][Codegen][X86] Add codegen test coverage for the variably-indexed load of alloca w/zero upper half
Roman Lebedev [Fri, 23 Dec 2022 17:09:00 +0000 (20:09 +0300)]
[NFC][SROA] Variably-indexed load: add test variation w/ upper half of alloca being zeros
This is the actual pattern i'm looking at.
Alexandros Lamprineas [Thu, 22 Dec 2022 17:44:06 +0000 (17:44 +0000)]
[IPSCCP] Create a Pass parameter to control specialization of functions.
Required for D140210 in order to disable FuncSpec at {Os, Oz}
optimization levels.
Differential Revision: https://reviews.llvm.org/D140564
Pengxuan Zheng [Fri, 16 Dec 2022 18:32:35 +0000 (10:32 -0800)]
[lld-macho][test][nfc] Update stabs.s to use GMT time zone instead of UTC
This is to work around a singularity container issue we ran into recently. The
container fails to honor the time zone setting (TZ=UTC) when executing the touch
command. Replacing UTC with GMT worked correctly in the container. This change
does not change (at least not intentionally) the function of the test in any way
since GMT and UTC should be equivalent AFAIK. Please refer to the discussions in
D139980 for more background information.
Reviewed By: int3, #lld-macho
Differential Revision: https://reviews.llvm.org/D140233
Dani Ferreira Franco Moura [Fri, 23 Dec 2022 13:53:13 +0000 (13:53 +0000)]
[clang][nullability] Remove old overload for getNullability()
Reviewed By: gribozavr2
Differential Revision: https://reviews.llvm.org/D140626
Jez Ng [Fri, 23 Dec 2022 05:18:11 +0000 (00:18 -0500)]
[lld-macho][test] Hyphenate lit's check-prefixes
For consistency.
Mark de Wever [Thu, 22 Dec 2022 19:09:24 +0000 (20:09 +0100)]
[libc++] LWG3738 Validates a missing precondition.
No real changes were needed, but add an assert for the pre-condition.
This implements:
- 3738 Missing preconditions for take_view constructor
Reviewed By: #libc, philnik
Differential Revision: https://reviews.llvm.org/D140568
Mark de Wever [Thu, 22 Dec 2022 19:35:36 +0000 (20:35 +0100)]
[libc++] LWG3745 noexcept for atomic_wait.
The noexcept was already implemented, this only updates the synposis and
adds tests to validate that the functions are noexcept.
This implements:
- LWG3745 std::atomic_wait and its friends lack noexcept
Reviewed By: #libc, philnik
Differential Revision: https://reviews.llvm.org/D140575
Alexey Bataev [Fri, 23 Dec 2022 16:00:50 +0000 (08:00 -0800)]
[TTI][NFC]Remove trailing spaces, NFC.
Nick Kreeger [Fri, 23 Dec 2022 15:55:15 +0000 (09:55 -0600)]
[mlir] Enable types to us custom assembly formats involving optional attributes.
Author: Laszlo Kindrat <laszlokindrat@gmail.com>
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D140322
Shilei Tian [Fri, 23 Dec 2022 15:42:29 +0000 (10:42 -0500)]
[NFC][OpenMP] Fix compile warning caused by using `std::move` on a local object on a `return` statement
Nikita Popov [Fri, 23 Dec 2022 15:33:28 +0000 (16:33 +0100)]
[LoopUnroll] Convert some tests to opaque pointers (NFC)
Roman Lebedev [Fri, 23 Dec 2022 15:10:39 +0000 (18:10 +0300)]
[DAGCombiner] `visitFREEZE()`: fix cycle breaking
Depending on the particular DAG, we might either create a `freeze`,
or not. And only in the former case, the cycle would be formed.
It would be nicer to have `ReplaceAllUsesOfValueWithIf()`,
like we have in IR, but we don't have that.
Fixes https://github.com/llvm/llvm-project/issues/59677
Matt Arsenault [Fri, 23 Dec 2022 03:38:10 +0000 (22:38 -0500)]
ValueTracking: Teach canCreateUndefOrPoison about saturating intrinsics
Matt Arsenault [Fri, 23 Dec 2022 03:37:12 +0000 (22:37 -0500)]
InstCombine: Add baseline tests for saturating poison handling
Nikolas Klauser [Sat, 13 Aug 2022 20:33:12 +0000 (22:33 +0200)]
[libc++] Add custom clang-tidy checks
Reviewed By: #libc, ldionne
Spies: jwakely, beanz, smeenai, cfe-commits, tschuett, avogelsgesang, Mordante, sstefan1, libcxx-commits, ldionne, mgorny, arichardson, miyuki
Differential Revision: https://reviews.llvm.org/D131963
Mikhail Goncharov [Fri, 23 Dec 2022 14:32:59 +0000 (15:32 +0100)]
fix warn-xparser test
for https://reviews.llvm.org/D140224
Nico Weber [Fri, 23 Dec 2022 14:29:38 +0000 (09:29 -0500)]
ibricchi [Fri, 23 Dec 2022 14:12:36 +0000 (09:12 -0500)]
[InlineAdvisor] Restructure advisor plugin unittest cmake
Move the plugin used in the unittest to test Inline Advisor Plugins
into a separate folder to clean up the cmake file for the analysis
tests.
Differential Revision: https://reviews.llvm.org/D140559
Roman Lebedev [Fri, 23 Dec 2022 14:14:22 +0000 (17:14 +0300)]
[DAGCombiner] `visitFREEZE()`: fix handling of no maybe-poison ops
The original code was confusing. It was stripping poison-generating flags,
but the comments were saying that doing so was a TODO.
If the poison-generating flags are present, then even if all operands
are guaranteed not to be undef or poison, the whole operation may still
produce undef or poison. We can still deal with that case,
and we already do deal with it in fact, by also dropping those flags.
Refs. https://github.com/llvm/llvm-project/issues/59676
Roman Lebedev [Fri, 23 Dec 2022 14:02:14 +0000 (17:02 +0300)]
[DAGCombiner] `visitFREEZE()`: restore previous behaviour on no maybe-poison operands
Lack of such operands implies that the op might be poison-producing due to
it's flags. We seem to drop them already, but the comments are confusing.
Fixes https://github.com/llvm/llvm-project/issues/59676
Ilya Andreev [Tue, 13 Sep 2022 13:01:56 +0000 (09:01 -0400)]
[RISCV] Combine comparison and logic ops
Two comparison operations and a logical operation are combined into selection using MIN or MAX and comparison operation.
For optimization to be applied conditions have to be satisfied:
1. In comparison operations has to be the one common operand.
2. Supports only signed and unsigned integers.
3. Comparison has to be the same with respect to common operand.
4. There are no more users of comparison except logic operation.
5. Every combination of comparison and AND, OR are supported.
It will convert
%l0 = %a < %c
%l1 = %b < %c
%res = %l0 or %l1
into
%sel = min(%a, %b)
%res = %sel < %c
It supports several comparison operations (<, <=, >, >=), signed, unsigned values and different order of operands if they do not violate conditions.
Differential Revision: https://reviews.llvm.org/D134277
Ilya Andreev [Tue, 13 Sep 2022 13:01:56 +0000 (09:01 -0400)]
[RISCV][test] Combine comparison and logic ops
Two comparison operations and a logical operation are combined into selection using MIN or MAX and comparison operation.
For optimization to be applied conditions have to be satisfied:
1. In comparison operations has to be the one common operand.
2. Supports only signed or unsigned integers.
3. Comparison has to be the same with respect to common operand.
4. There are no more users of comparison except logic operation.
5. Every combination of comparison and AND, OR are supported.
It will convert
%l0 = %a < %c
%l1 = %b < %c
%res = %l0 or %l1
into
%sel = min(%a, %b)
%res = %sel < %c
It supports several comparison operations (<, <=, >, >=), signed, unsigned values and different order of operands if they do not violate conditions.
Nico Weber [Fri, 23 Dec 2022 14:08:56 +0000 (09:08 -0500)]
Dmitri Gribenko [Fri, 23 Dec 2022 14:03:48 +0000 (15:03 +0100)]
[clang] Remove deprecated ControlFlowContext::build()
Reviewed By: merrymeerkat
Differential Revision: https://reviews.llvm.org/D140625
Luke Drummond [Tue, 20 Dec 2022 00:05:46 +0000 (00:05 +0000)]
[NFC][NVPTX] Remove dead override
After
68f2218e1e, NVPTXTargetObjectFile::Initialize is an empty wrapper
of the parent method. Get rid of it.
Differential Revision: https://reviews.llvm.org/D140397
Aliia Khasanova [Fri, 23 Dec 2022 13:38:29 +0000 (14:38 +0100)]
Apply shortened printing/parsing form to linalg.reduce.
Differential Revision: https://reviews.llvm.org/D140622
Dmitri Gribenko [Fri, 23 Dec 2022 13:26:34 +0000 (14:26 +0100)]
[clang] Migrate away from a deprecated Clang CFG factory function
Reviewed By: merrymeerkat
Differential Revision: https://reviews.llvm.org/D140620
Florian Hahn [Fri, 23 Dec 2022 12:51:20 +0000 (12:51 +0000)]
[LV] Move exit cond simplification to separate transform.
This sets the stage for D133017 by moving out the code that performs
VPlan based simplifications to a separate transform that takes the
chosen VF & UF as arguments.
The main advantage is that this transform runs before any changes to
the CFG are being made. This allows using SCEV without worrying about
making queries while the IR is in an incomplete state.
Note that this patch switches the reasoning to use SCEV, but still only
simplifies loops with constant trip counts. Using SCEV here is needed to
access the backedge taken count, because the trip count IR value has not
been created yet.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D135017
serge-sans-paille [Fri, 23 Dec 2022 12:25:58 +0000 (13:25 +0100)]
Revert "[clang] Use a StringRef instead of a raw char pointer to store builtin and call information"
There are still remaining issues with GCC 12, see for instance
https://lab.llvm.org/buildbot/#/builders/93/builds/12669
This reverts commit
5ce4e92264102de21760c94db9166afe8f71fcf6.
Luke Drummond [Mon, 19 Dec 2022 14:59:12 +0000 (14:59 +0000)]
[NFC][NVPTX] Remove dead comment and commented code
A confusing comment after the last return statement in
`NVPTXAsmPrinter::doFinalization` referred to a preprocessor macro
(NVISA) that has never existed since the NVPTX backend has been a part
of upstream llvm - as far as the pickaxe will tell me anyway. Thus I've
removed it.
Differential Revision: https://reviews.llvm.org/D140399
Matthias Springer [Fri, 23 Dec 2022 12:01:00 +0000 (13:01 +0100)]
[mlir] Add option to limit number of pattern rewrites in CanonicalizerPass
The greedy pattern rewriter consists of two nested loops. `config.maxIterations` (which configurable on the CanonicalizerPass) controls the maximum number of iterations of the outer loop.
```
/// This specifies the maximum number of times the rewriter will iterate
/// between applying patterns and simplifying regions. Use `kNoLimit` to
/// disable this iteration limit.
int64_t maxIterations = 10;
```
This change adds `config.maxNumRewrites` which controls the maximum number of pattern rewrites within an iteration. (It effectively control the maximum number of iterations of the inner loop.)
This flag is meant for debugging and useful in cases where one or multiple faulty patterns can be applied indefinitely, resulting in an infinite loop.
Differential Revision: https://reviews.llvm.org/D140525
Nikita Popov [Fri, 23 Dec 2022 11:46:41 +0000 (12:46 +0100)]
[VE] Convert test to opaque pointers (NFC)
There is a minor codegen regression here (an extra and instruction).
The reason is that CGP only eliminates fallthrough branches if it
has made some other kind of change, and with opaque pointers that
other change does not occur.
Ideally, we should probably always try to eliminate fallthroughs,
but this runs into the problem that performing a dummy fallthrough
is a common pattern in tests for forcing SDAG to select them
separately, so it's not quite that simple.
serge-sans-paille [Mon, 12 Dec 2022 16:02:15 +0000 (17:02 +0100)]
[clang] Use a StringRef instead of a raw char pointer to store builtin and call information
This avoids recomputing string length that is already known at compile
time.
It has a slight impact on preprocessing / compile time, see
https://llvm-compile-time-tracker.com/compare.php?from=
3f36d2d579d8b0e8824d9dd99bfa79f456858f88&to=
e49640c507ddc6615b5e503144301c8e41f8f434&stat=instructions:u
This is a recommit of
719d98dfa841c522d8d452f0685e503538415a53 with a
change to llvm/utils/TableGen/OptParserEmitter.cpp to cope with GCC bug
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108158
Differential Revision: https://reviews.llvm.org/D139881
Florian Hahn [Fri, 23 Dec 2022 11:44:42 +0000 (11:44 +0000)]
[LV] Assert that the executed plan contains selected VF & UF (NFC).
Add assertion to ensure the executed plan is valid for the selected VF
and UF.
Haojian Wu [Thu, 22 Dec 2022 22:23:54 +0000 (23:23 +0100)]
[clang] Fix a clang crash on invalid code in C++20 mode.
This crash is a combination of recovery-expr + new SemaInit.cpp code
introduced by by https://reviews.llvm.org/D129531.
Differential Revision: https://reviews.llvm.org/D140587
Dani Ferreira Franco Moura [Fri, 23 Dec 2022 10:04:56 +0000 (10:04 +0000)]
Remove empty header file.
Reviewed By: gribozavr2, merrymeerkat
Differential Revision: https://reviews.llvm.org/D140483
Nikita Popov [Fri, 23 Dec 2022 10:41:57 +0000 (11:41 +0100)]
[VE] Name instructions in test (NFC)
serge-sans-paille [Fri, 23 Dec 2022 10:36:56 +0000 (11:36 +0100)]
Revert "[clang] Use a StringRef instead of a raw char pointer to store builtin and call information"
Failing builds: https://lab.llvm.org/buildbot#builders/9/builds/19030
This is GCC specific and has been reported upstream: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108158
This reverts commit
719d98dfa841c522d8d452f0685e503538415a53.
serge-sans-paille [Mon, 12 Dec 2022 16:02:15 +0000 (17:02 +0100)]
[clang] Use a StringRef instead of a raw char pointer to store builtin and call information
This avoids recomputing string length that is already known at compile
time.
It has a slight impact on preprocessing / compile time, see
https://llvm-compile-time-tracker.com/compare.php?from=
3f36d2d579d8b0e8824d9dd99bfa79f456858f88&to=
e49640c507ddc6615b5e503144301c8e41f8f434&stat=instructions:u
Differential Revision: https://reviews.llvm.org/D139881
Lang Hames [Fri, 23 Dec 2022 09:08:50 +0000 (01:08 -0800)]
[examples] Direct HowToUseJIT readers to HowToUseLLJIT instead.
HowToUseJIT describes the older APIs. We want to discourage their use in new
projects.
Nikita Popov [Fri, 23 Dec 2022 09:07:59 +0000 (10:07 +0100)]
[Docs] Clarify typed pointers support timeline
As there have been a couple of questions about this recently, this
gives a hard timeline on typed pointers support.
Given that we are about a month away from LLVM 16 branching, I think
we should retain best-effort typed pointer support in LLVM 16 even
if we get all tests migrated before that point.
Conversely, regardless of what the actual test migration state will
be at that point, I believe we should un-support typed pointers as
a matter of policy immediately after branching. Once release/16.x
has been branched, typed pointers on main will no longer be
supported (and can be actively broken). We only need to keep
not-yet-migrated tests working, if there are any left at that point.
Differential Revision: https://reviews.llvm.org/D140487
Nikita Popov [Fri, 23 Dec 2022 09:06:50 +0000 (10:06 +0100)]
[LoopDeletion] Convert tests to opaque pointers (NFC)
Nikita Popov [Fri, 23 Dec 2022 09:03:38 +0000 (10:03 +0100)]
[VectorCombine] Convert tests to opaque pointers (NFC)
Nikita Popov [Fri, 23 Dec 2022 09:01:56 +0000 (10:01 +0100)]
[SLP] Convert some tests to opaque pointers (NFC)
Nikita Popov [Fri, 23 Dec 2022 08:58:33 +0000 (09:58 +0100)]
[GVN] Convert some tests to opaque pointers (NFC)
Nikita Popov [Fri, 23 Dec 2022 08:57:22 +0000 (09:57 +0100)]
[BDCE] Convert tests to opaque pointers (NFC)
Nikita Popov [Fri, 23 Dec 2022 08:55:29 +0000 (09:55 +0100)]
[Attributor] Convert some tests to opaque pointers (NFC)
These were converted without adjustments.
Carlos Galvez [Fri, 23 Dec 2022 08:47:28 +0000 (08:47 +0000)]
[clang-tidy][NFC] Remove custom isInAnonymousNamespace matchers
Since now the same matcher exists in ASTMatchers.
Nikita Popov [Fri, 23 Dec 2022 08:48:36 +0000 (09:48 +0100)]
[ArgPromotion] Convert tests to opaque pointers (NFC)
update_test_checks was rerun for some of those, because we use
a different GEP representation with opaque pointers.
Nikita Popov [Fri, 23 Dec 2022 08:47:48 +0000 (09:47 +0100)]
[AggressiveInstCombine] Convert tests to opaque pointers (NFC)
liqinweng [Fri, 23 Dec 2022 08:01:29 +0000 (16:01 +0800)]
[MLIR][Arith] Remove unused assertions
We shouldn't be checking things that are guaranteed by the op's verifier.
Reviewed By: benshi001
Differential Revision: https://reviews.llvm.org/D140610
Craig Topper [Fri, 23 Dec 2022 06:48:12 +0000 (22:48 -0800)]
[Support] Use inplace APInt operators in DivisionByConstantInfo. NFC
Reduces the number of temporary APInts that get created and
copy/moved from.
Carlos Galvez [Mon, 19 Dec 2022 18:34:35 +0000 (18:34 +0000)]
[ASTMatchers] Add isInAnonymousNamespace narrowing matcher
Used in a couple clang-tidy checks so it could be extracted
out as its own matcher.
Differential Revision: https://reviews.llvm.org/D140328
Freddy Ye [Fri, 23 Dec 2022 06:53:33 +0000 (14:53 +0800)]
[X86] Add reduce_*_ep[i|u]8/16 series intrinsics.
Reviewed By: pengfei, skan
Differential Revision: https://reviews.llvm.org/D140531