River Riddle [Tue, 8 Mar 2022 08:03:27 +0000 (00:03 -0800)]
[mlir][NFC] Properly include PassDetail.h for FoldSubViewOpsPass
River Riddle [Tue, 8 Mar 2022 08:01:41 +0000 (00:01 -0800)]
[mlir][NFC] Remove dead forward declaration of FuncOp
Adrian Kuegel [Tue, 8 Mar 2022 07:55:55 +0000 (08:55 +0100)]
[mlir][Bazel] Adjust regarding StandardToLLVM -> FuncToLLVM rename.
Differential Revision: https://reviews.llvm.org/D121185
Petr Hosek [Tue, 8 Mar 2022 07:59:29 +0000 (23:59 -0800)]
Revert "[CMake][compiler-rt] Make CRT separately buildable"
This reverts commit
ed28461e2716fda004f4f55654ef33bc8c78f203 since
it broke the VE bot.
Arthur Eubanks [Mon, 7 Mar 2022 18:17:09 +0000 (10:17 -0800)]
[NewPM][Inliner] Make inlined calls to functions in same SCC as callee exponentially expensive
Introduce a new attribute "function-inline-cost-multiplier" which
multiplies the inline cost of a call site (or all calls to a callee) by
the multiplier.
When processing the list of calls created by inlining, check each call
to see if the new call's callee is in the same SCC as the original
callee. If so, set the "function-inline-cost-multiplier" attribute of
the new call site to double the original call site's attribute value.
This does not happen when the original call site is intra-SCC.
This is an alternative to D120584, which marks the call sites as
noinline.
Hopefully fixes PR45253.
Reviewed By: davidxl
Differential Revision: https://reviews.llvm.org/D121084
Lorenzo Chelini [Tue, 8 Mar 2022 07:30:20 +0000 (08:30 +0100)]
[MLIR][OpDSL] Silence warning (NFC)
Stanislav Mekhanoshin [Mon, 7 Mar 2022 23:53:16 +0000 (15:53 -0800)]
[AMDGPU] Set noclobber metadata on loads instead of cast to constant
A load via pointer cast to constant will return true from
pointsToConstantMemory which is not necessarily so.
Fixes: SWDEV-326463
Differential Revision: https://reviews.llvm.org/D121172
Petr Hosek [Mon, 28 Feb 2022 20:21:11 +0000 (12:21 -0800)]
[CMake][compiler-rt] Make CRT separately buildable
This is useful when building a complete toolchain to ensure that CRT
is built after builtins but before the rest of the compiler-rt.
Differential Revision: https://reviews.llvm.org/D120682
Zi Xuan Wu [Mon, 7 Mar 2022 10:28:13 +0000 (18:28 +0800)]
[CSKY] Enable TargetAsmStreamer/ELFStreamer and support asm directive of csky_attribute
Add support of parsing .csky_attribute directive and emit related target attributes in .csky.attribute section.
It does not emit attribute directive in assembly code, so only emit target attributes in ELF streamer.
In ELF streamer, it handles the header EFlag and the csky_attribute section which contains some attribute items.
The EFlag and attribute items are calculated from feature bits based on Subtarget.
Yuanfang Chen [Mon, 7 Mar 2022 20:42:09 +0000 (12:42 -0800)]
Reland "[clang][debug] port clang-cl /JMC flag to ELF"
This relands commit
731347431976509823e38329a96fcbc69fe98cd2.
It failed on Windows/Mac because `-fjmc` is only checked for ELF targets.
Check the flag unconditionally instead and issue a warning for non-ELF targets.
Valentin Clement [Tue, 8 Mar 2022 05:51:10 +0000 (06:51 +0100)]
[flang] Remove unused variable in Allocatable.cpp
Fix for buildbot failure: https://lab.llvm.org/buildbot/#/builders/180/builds/3532
Johannes Doerfert [Fri, 25 Feb 2022 00:50:37 +0000 (18:50 -0600)]
[OpenMP][FIX] Ensure flag to disable de-globalization works properly
If the user disables de-globalization we did not seed the AAHeapToShared
and AAHeapToStack but we still could end up with them through in-flight
lookups. With this patch we disable AAHeapToShared completely if the
user disabled de-globalization. Heap-2-stack is still run though.
Differential Revision: https://reviews.llvm.org/D121059
Johannes Doerfert [Fri, 18 Feb 2022 17:49:00 +0000 (11:49 -0600)]
[OpenMP][CUDA] Use one event pool per device
An event pool, similar to the stream pool, needs to be kept per device.
For one, events are associated with cuda contexts which means we cannot
destroy the former after the latter. Also, CUDA documentation states
streams and events need to be associated with the same context, which
we did not ensure at all.
Differential Revision: https://reviews.llvm.org/D120142
Johannes Doerfert [Thu, 17 Feb 2022 22:04:12 +0000 (16:04 -0600)]
[OpenMP] Allow to explicitly deinitialize device resources
There are two problems this patch tries to address:
1) We currently free resources in a random order wrt. plugin and
libomptarget destruction. This patch should ensure the CUDA plugin
is less fragile if something during the deinitialization goes wrong.
2) We need to support (hard) pause runtime calls eventually. This patch
allows us to free all associated resources, though we cannot
reinitialize the device yet.
Follow up patch will associate one event pool per device/context.
Differential Revision: https://reviews.llvm.org/D120089
Johannes Doerfert [Wed, 2 Mar 2022 19:34:24 +0000 (13:34 -0600)]
[OpenMP][NFCI] Use RAII lock guards in libomptarget where possible
Differential Revision: https://reviews.llvm.org/D121060
Uday Bondhugula [Mon, 7 Mar 2022 23:42:54 +0000 (05:12 +0530)]
NFC. Improve isInnermostAffineForOp - drop unnecessary check
Rewrite isInnermostAffineForOp utility to make it more direct/efficient.
Drop unnecessary check. NFC.
Differential Revision: https://reviews.llvm.org/D121170
Shao-Ce SUN [Tue, 8 Mar 2022 04:43:33 +0000 (12:43 +0800)]
[RISCV] Share PatFprFpr classes for F, D, and Zfh
Inspired by D115469
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D121066
Lang Hames [Tue, 8 Mar 2022 04:37:33 +0000 (20:37 -0800)]
[ORC][JITLink] Fix MachO absolute symbol handling, add test case.
This patch removes the unintended resolution of locally scoped absolute symbols
(which was causing unexpected definition errors).
It stops using the JITSymbolFlags::Absolute flag (it isn't set or used elsewhere,
and causes mismatch-flags asserts), and adds JITSymbolFlags::Exported to default
scoped absolute symbols.
Finally, we now set the scope of absolute symbols correctly in
MachOLinkGraphBuilder.
Christudasan Devadasan [Wed, 2 Mar 2022 06:46:29 +0000 (12:16 +0530)]
AMDGPU: Skip folding REG_SEQUENCE if found unknown regclasses for its users
Use TII::getRegClass to return a valid regclass or a nullptr
if the RC is unknown for a given OpIdx. This fixes a potential
crash occurred while getting the RC from a variadic instruction.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D120813
Phoebe Wang [Tue, 8 Mar 2022 03:18:41 +0000 (11:18 +0800)]
[X86] Add helper enum for ternary intrinsics
Reviewed By: RKSimon, LuoYuanke
Differential Revision: https://reviews.llvm.org/D120307
Philip Reames [Tue, 8 Mar 2022 03:11:32 +0000 (19:11 -0800)]
[SLP] Extract a helper for buildvector [nfc]
Yitzhak Mandelbaum [Mon, 7 Mar 2022 21:16:28 +0000 (21:16 +0000)]
[clang][dataflow] Fix nullptr dereferencing error.
When pre-initializing fields in the environment, the code assumed that all
fields of a struct would be initialized. However, given limits on value
construction, that assumption is incorrect. This patch changes the code to drop
that assumption and thereby avoid dereferencing a nullptr.
Differential Revision: https://reviews.llvm.org/D121158
jacquesguan [Wed, 23 Feb 2022 07:30:10 +0000 (15:30 +0800)]
[RISCV] Add patterns for vector widening floating-point reduction instructions.
Add patterns for vector widening floating-point reduction instructions.
Differential Revision: https://reviews.llvm.org/D120390
Philip Reames [Tue, 8 Mar 2022 02:52:51 +0000 (18:52 -0800)]
[SLP] Fix spelling in a lambda name [NFC]
Amir Ayupov [Tue, 8 Mar 2022 02:41:12 +0000 (18:41 -0800)]
[BOLT][CMAKE] Remove CMake 3.13.4 incompatible parameter
Remove `TYPE BIN` parameter that is introduced in CMake 3.14 and revert back to
the equivalent compatible form `DESTINATION ${CMAKE_INSTALL_BINDIR}`.
Addresses https://github.com/llvm/llvm-project/issues/54099
Reviewed By: rafauler
Differential Revision: https://reviews.llvm.org/D121012
Jim Ingham [Tue, 8 Mar 2022 00:38:52 +0000 (16:38 -0800)]
Change "target.save-jit-objects" to "target.save-jit-objects-dir".
The old command wrote to CWD, which doesn't always work, and if it
didn't, there was no workaround (and it crashed on failure). This
patch changed the setting to provide a directory to save the objects
to.
Differential Revision: https://reviews.llvm.org/D121036
Kai Luo [Tue, 8 Mar 2022 01:23:23 +0000 (01:23 +0000)]
[libcxx][AIX] Remove `LIBCXX-AIX-FIXME` in isalwayslockfree check after
86478c7ad8a7
The workaround has been removed via https://reviews.llvm.org/D119931, checks passed on AIX-32.
Reviewed By: ldionne, daltenty, jsji, #libc
Differential Revision: https://reviews.llvm.org/D121081
Peter Klausler [Fri, 25 Feb 2022 21:54:44 +0000 (13:54 -0800)]
[flang] Fix module file missing USE for shadowed derived type
When a module uses a derived type that is shadowed by a generic
interface, the module file was missing a USE statement for the
name. Detect and handle this situation.
Differential Revision: https://reviews.llvm.org/D121160
Peixin-Qiao [Tue, 8 Mar 2022 01:04:31 +0000 (09:04 +0800)]
[clang] Fix OpenMP critical hint parameter check
The paramemter of hint clause in OpenMP critical hint should be
non-negative. The omp_lock_hint_none is 0 in omp.h.
Reviewed By: Alexey Bataev
Differential Revision: https://reviews.llvm.org/D121101
Bixia Zheng [Mon, 7 Mar 2022 03:34:40 +0000 (19:34 -0800)]
[mlir][sparse][taco] Split the evaluate method into compile and compute.
This is to align with the PyTACO API better.
Modify an existing unit test to test the new routines.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D121083
Jessica Paquette [Mon, 7 Mar 2022 23:33:12 +0000 (15:33 -0800)]
Add JSON output option to llvm-remark-size-diff
This adds JSON output to llvm-remark-size-diff.
The goal here is to make it easy for external tools to consume output from
llvm-remark-size-diff. These tools could be used for automated size analysis.
(E.g. in CI).
To specify JSON output, use `--report_style=json`. JSON output can be
pretty-printed via `--pretty`.
With automation in mind, the schema looks like this:
```
"Files": {
"A": <filename_a>
"B": <filename_b>
},
"InBoth": [
{
"FunctionName": <function name>,
"InstCount": [
<count_in_a>,
<count_in_b>
],
"StackSize": [
<count_in_a>,
<count_in_b>
]
},
...
]
"OnlyInA": [
{
"FunctionName": <function name>,
"InstCount": [
<count_in_a>,
0
],
"StackSize": [
<count_in_a>,
0
]
},
...
]
"OnlyInB": [
{
"FunctionName": <function name>,
"InstCount": [
0,
<count_in_b>
],
"StackSize": [
0,
<count_in_b>
]
},
...
]
```
A few notes:
- Filenames are included, because tools may want to combine many outputs
together in some way (a big JSON file, a big CSV, or something.)
- Counts are represented as [a, b] so that a diff can be calculated via b - a.
The original counts may be useful for size analysis (e.g. was this function
extremely large before?) and so both are preserved.
- `OnlyInA` and `OnlyInB` have a 0 for one of the counts always. This is to
make it easier for tools to share code between `OnlyInA`, `OnlyInB`, and
`InBoth`.
Differential Revision: https://reviews.llvm.org/D121173
Sam Clegg [Mon, 7 Mar 2022 20:52:23 +0000 (12:52 -0800)]
[WebAssembly] Add new target feature in support of 'extended-const' proposal
We don't yet do anything when this feature is enabled, this change
just lays the ground work by accepting that there is such a feature.
See https://github.com/WebAssembly/extended-const
Differential Revision: https://reviews.llvm.org/D121151
Peter Klausler [Mon, 28 Feb 2022 18:24:58 +0000 (10:24 -0800)]
[flang] Make uninitialized allocatable components explicitly NULL() in structure constructors
When a structure constructor does not initialize an allocatable component,
ensure that the typed expression representation contains an explicit
NULL() for the component. Expression semantics already copies default
initialized expressions for nonallocatable components into structure
constructors. This change is expected to simplify lowering.
Differential Revision: https://reviews.llvm.org/D121162
Arthur Eubanks [Tue, 8 Mar 2022 00:24:36 +0000 (16:24 -0800)]
[NFC] Cleanup StandardInstrumentations
Louis Dionne [Mon, 7 Mar 2022 22:34:17 +0000 (17:34 -0500)]
Revert "[libc++] Remove extension to support allocator<const T>"
This reverts commit
bed3240bf7d196a17cc31f1c5b59b4721017e638.
I will need to add more tests for std::shared_ptr<T const> before
re-landing this.
Jacob Lambert [Fri, 7 Jan 2022 21:29:15 +0000 (13:29 -0800)]
[AMDGPU] Add gfx10 assembler directive to specify shared VGPR count
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D105507
Jez Ng [Mon, 7 Mar 2022 21:34:02 +0000 (16:34 -0500)]
Re-land [lld-macho][nfc] Don't use `stubsHelperIndex` in ICF hash
Previous attempt was commit
112135e77444e8c8106efa77416af09f4b9a5012 and
reverted in
d86d431814c836ec5579fc0b981e86c7fb81f2f2.
Peter Klausler [Wed, 23 Feb 2022 22:16:10 +0000 (14:16 -0800)]
[flang] Runtime validation of SPREAD(DIM=dim) argument
Crash when DIM= is not a valid dimension in the result.
Differential Revision: https://reviews.llvm.org/D121145
Groverkss [Mon, 7 Mar 2022 21:31:24 +0000 (03:01 +0530)]
[MLIR][Presburger][NFC] Cleanup PresburgerSet
This patch cleans up the interface to PresburgerSet. At a high level it does
the following changes:
- Move member functions around to have constructors at top and print/dump
at end.
- Move a private function to be a static function instead.
- Change member functions of type "getAllIntegerPolyhedron" to "getAllPolys"
instead.
- Improve documentation for PresburgerSet.
Reviewed By: arjunp
Differential Revision: https://reviews.llvm.org/D121027
Peter Klausler [Wed, 23 Feb 2022 19:45:21 +0000 (11:45 -0800)]
[flang] Use faster path for default formatted character input
Rather than reading default character variables in formatted
input one byte at a time via NextInField(), skip and read
them via blocks of available buffer data. This eliminates
a bottleneck that affected reads of large character values.
(It also exposed a problem with sequential reads with RECL=
set on the OPEN statement, so that's fixed too.)
Differential Revision: https://reviews.llvm.org/D121144
Jonathan Peyton [Mon, 7 Mar 2022 20:54:07 +0000 (14:54 -0600)]
[OpenMP][libomp] Fix register constraint for tpause and umwait
Register constraint switched to "=q" which means very specifically (from
https://gcc.gnu.org/onlinedocs/gcc/Machine-Constraints.html#Machine-Constraints)
> Any register accessible as rl. In 32-bit mode, a, b, c, and d; in 64-bit
mode, any integer register.
Older gcc versions (8.x and below) were trying to use esi or edi for the
8 bit flag variable, but it wound up displaying this error in the end:
kmp_lock.cpp: In function ‘void __kmp_spin_backoff(kmp_backoff_t*)’:
kmp_lock.cpp:2684:1: error: unsupported size for integer register
Hence the correct restriction is "=q" instead of "=r".
Fixes: https://github.com/llvm/llvm-project/issues/53309
Differential Revision: https://reviews.llvm.org/D120519
Hanhan Wang [Mon, 7 Mar 2022 20:52:03 +0000 (12:52 -0800)]
[mlir][Vector] Add patterns to reorder elementwise ops and broadcast/transpose ops.
In quantized comutation, there are casting ops around computation ops.
Reorder the ops to make reduce-to-contract actually work.
Reviewed By: ThomasRaoux
Differential Revision: https://reviews.llvm.org/D120760
Valentin Clement [Mon, 7 Mar 2022 20:22:28 +0000 (21:22 +0100)]
[flang] Lower allocate and deallocate statements
This patch add the lowering for the allocate
and the deallocate statements.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D121146
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Arthur O'Dwyer [Mon, 7 Mar 2022 16:46:06 +0000 (11:46 -0500)]
[libc++] Make common_iterator's proxy types into aggregates.
Saves one move in each case, which is basically nothing perf-wise;
this is more about simplifying the code.
Differential Revision: https://reviews.llvm.org/D121130
Ahmed Bougacha [Mon, 7 Mar 2022 20:43:37 +0000 (12:43 -0800)]
[sancov] Don't instrument calls to bitcast funcs: they're not indirect.
Currently, when instrumenting indirect calls, this uses
CallBase::getCalledFunction to determine whether a given callsite is
eligible.
However, that returns null if:
this is an indirect function invocation or the function signature
does not match the call signature.
So, we end up instrumenting direct calls where the callee is a bitcast
ConstantExpr, even though we presumably don't need to.
Use isIndirectCall to ignore those funky direct calls.
Differential Revision: https://reviews.llvm.org/D119594
Yuanfang Chen [Mon, 7 Mar 2022 20:40:15 +0000 (12:40 -0800)]
Revert "[clang][debug] port clang-cl /JMC flag to ELF"
This reverts commit
731347431976509823e38329a96fcbc69fe98cd2.
Break bots:
http://45.33.8.238/win/54551/step_7.txt
http://45.33.8.238/macm1/29590/step_7.txt
Martin Storsjö [Mon, 7 Mar 2022 18:32:52 +0000 (13:32 -0500)]
[libcxxabi] Fix cmake order dependency wrt dllexporting
If LIBCXX_ENABLE_SHARED isn't explicitly set on the cmake command
line, isn't set in the cache, and the libcxxabi project is configured
before libcxx, then LIBCXX_ENABLE_SHARED isn't defined yet. Once
the libcxx cmake project has been parsed, LIBCXX_ENABLE_SHARED would
have been set to its default value of ON.
This makes sure that the symbols are properly dllexported in such
a configuration scenario.
Differential Revision: https://reviews.llvm.org/D120982
Louis Dionne [Fri, 4 Mar 2022 14:54:29 +0000 (09:54 -0500)]
[libc++] Remove extension to support allocator<const T>
This extension is a portability trap for users, since no other standard
library supports it. Furthermore, the Standard explicitly allows
implementations to reject std::allocator<cv T>, so allowing it is
really going against the current.
This was discovered in D120684: this extension required `const_cast`ing
in `__construct_range_forward`, a fishy bit of code that can be removed
if we don't support the extension anymore.
Differential Revision: https://reviews.llvm.org/D120996
Stanislav Mekhanoshin [Thu, 3 Mar 2022 21:14:42 +0000 (13:14 -0800)]
[AMDGPU] new gfx940 fp atomics
Differential Revision: https://reviews.llvm.org/D121028
Peter Klausler [Wed, 23 Feb 2022 00:59:48 +0000 (16:59 -0800)]
[flang] Make per-argument intrinsic error messages more localized
A recent patch made it possible to emit more localized error messages
pertaining to actual arguments in non-intrinsic procedure references.
Use these new powers for good and make intrinsic error messages more
precise, too.
Differential Revision: https://reviews.llvm.org/D121126
Sanjay Patel [Mon, 7 Mar 2022 19:42:39 +0000 (14:42 -0500)]
[x86] reduce code duplication for select of X86ISD::CMP; NFC
Sanjay Patel [Mon, 7 Mar 2022 18:39:09 +0000 (13:39 -0500)]
[x86] add tests for mask+cmp; NFC
Sanjay Patel [Mon, 7 Mar 2022 17:26:16 +0000 (12:26 -0500)]
[x86] remove unnecessary decorations in test file; NFC
Jonas Devlieghere [Mon, 7 Mar 2022 20:12:49 +0000 (12:12 -0800)]
[lldb] Format g_color_tokens as a table (NFC)
This makes editing the table easier and mistakes more obvious.
Jonas Devlieghere [Mon, 7 Mar 2022 20:09:22 +0000 (12:09 -0800)]
[lldb] Support "bright" ANSI colors
Support ANSI escape codes for bright colors variants. Most modern
terminals support them. LLDB is not using them in any of its defaults,
but they're useful for people who want to modify their preferred ANSI
prefix/suffix.
Differential revision: https://reviews.llvm.org/D121131
Stanislav Mekhanoshin [Fri, 4 Mar 2022 19:49:34 +0000 (11:49 -0800)]
[AMDGPU] Add v_mov_b64 gfx940 opcode
Differential Revision: https://reviews.llvm.org/D121023
Florian Hahn [Mon, 7 Mar 2022 19:57:26 +0000 (19:57 +0000)]
[IVDescriptors] Bail out instead of asserting that order is expected.
When dealing with multiple phis that depend on each other, the order
might have been changed and may not match the expectation. If that
happens, bail out, rather than asserting.
Fixes https://github.com/llvm/llvm-project/issues/54218
Fixes https://github.com/llvm/llvm-project/issues/54233
Fixes https://github.com/llvm/llvm-project/issues/54254
Krzysztof Parzyszek [Mon, 7 Mar 2022 19:27:41 +0000 (11:27 -0800)]
[Hexagon] Handle v2f16 in build_vector in isel
Stanislav Mekhanoshin [Wed, 2 Mar 2022 23:27:25 +0000 (15:27 -0800)]
[AMDGPU] gfx940 uses new names for coherency bits
Differential Revision: https://reviews.llvm.org/D120855
Greg Clayton [Mon, 7 Mar 2022 19:31:21 +0000 (11:31 -0800)]
Fix buildbots after https://reviews.llvm.org/D120755.
This improves this test a lot because before when using the "attachCommands" to run the following commands:
(lldb) target create -d /path/to/a.out
(lldb) process launch
This was racy as it wasn't stopping the program at the entry point, and the process might run to completion before we can even debug it. With the recent changes to the "attachCommands" we were waiting for the process to stop, but the process might be exited already, and that _should_ have caused the attach to fail since there was no process to attach to. By adding "--stop-at-entry" to the process launch, we ensure this should be less racy and give us a valid process to attach to.
Siva Chandra Reddy [Mon, 7 Mar 2022 09:00:14 +0000 (09:00 +0000)]
[libc] Align the stack pointer in the start function.
The loader TLS test for x86_64, which now passes, has been enabled.
A future change should enable the test for aarch64 as well.
Reviewed By: jeffbailey
Differential Revision: https://reviews.llvm.org/D121091
Peter Klausler [Fri, 18 Feb 2022 19:15:08 +0000 (11:15 -0800)]
[flang] Avoid crash case in provenance mapping
When a contiguous range of a cooked character stream is being
mapped to a range of source provenance, the code was assuming
that the "end()" position of the input range -- being the character
immediately after the range -- would also follow the range's
source provenance. This isn't always the case.
Modify the code to work with the true last character of the
input range (at end()-1) and to also cope with cases when that
last position truly maps to an earlier provenance, which can happen
when the prescanner has inserted a space into the cooked character
stream.
Differential Revision: https://reviews.llvm.org/D121124
Austin Kerbow [Mon, 7 Mar 2022 19:31:12 +0000 (11:31 -0800)]
[AMDGPU] Fix uninitialized value after
8d0c34fd4f
Stanislav Mekhanoshin [Tue, 1 Mar 2022 21:39:57 +0000 (13:39 -0800)]
[AMDGPU] gfx940: support V_FMAMK_F32 and V_FMAAK_F32
Differential Revision: https://reviews.llvm.org/D120769
Mitch Phillips [Thu, 3 Mar 2022 18:48:05 +0000 (10:48 -0800)]
[MTE] Add NT_ANDROID_TYPE_MEMTAG
This ELF note is aarch64 and Android-specific. It specifies to the
dynamic loader that specific work should be scheduled to enable MTE
protection of stack and heap regions.
Current synthesis of the ".note.android.memtag" ELF note is done in the
Android build system. We'd like to move that to the compiler, and this
is the first step.
Reviewed By: MaskRay, jhenderson
Differential Revision: https://reviews.llvm.org/D119381
Craig Topper [Mon, 7 Mar 2022 19:12:17 +0000 (11:12 -0800)]
[RISCV] Rename 'SplatOperand' to 'ScalarOperand'. NFC
vslide1up/down have this flag set, but the value isn't a splat.
Rename for clarity.
Reviewed By: khchen
Differential Revision: https://reviews.llvm.org/D121037
Craig Topper [Sat, 5 Mar 2022 02:53:54 +0000 (18:53 -0800)]
[LegalizeTypes][ARM][X86] Change ExpandIntRes_ABS to use sra+xor+sub.
Previously we used sra+add+xor if ADDCARRY is supported. This changes
to sra+xor+sub is SUBCARRY is available.
This is consistent with the recent change to the default expansion
in LegalizeDAG.
Differential Revision: https://reviews.llvm.org/D121039
River Riddle [Sat, 5 Mar 2022 03:45:00 +0000 (19:45 -0800)]
[mlir][NFC] Remove FuncOp overload of NestedPattern::match
This method is redundant with the Operation* overload, and is an artifact of when
FuncOp wasn't an operation.
River Riddle [Sat, 5 Mar 2022 03:41:15 +0000 (19:41 -0800)]
[mlir][NFC] Drop a few dead forward declarations of FuncOp
River Riddle [Tue, 1 Mar 2022 22:53:41 +0000 (14:53 -0800)]
[mlir][NFC] Rename StandardToLLVM to FuncToLLVM
The current StandardToLLVM conversion patterns only really handle
the Func dialect. The pass itself adds patterns for Arithmetic/CFToLLVM, but
those should be/will be split out in a followup. This commit focuses solely
on being an NFC rename.
Aside from the directory change, the pattern and pass creation API have been renamed:
* populateStdToLLVMFuncOpConversionPattern -> populateFuncToLLVMFuncOpConversionPattern
* populateStdToLLVMConversionPatterns -> populateFuncToLLVMConversionPatterns
* createLowerToLLVMPass -> createConvertFuncToLLVMPass
Differential Revision: https://reviews.llvm.org/D120778
Haojian Wu [Mon, 7 Mar 2022 09:47:38 +0000 (10:47 +0100)]
[pseudo] Strip comments for TokenStream.
Add a utility function to strip comments from a "raw" tokenstream. The
derived stream will be fed to the GLR parser (for early testing).
Differential Revision: https://reviews.llvm.org/D121092
AndreyChurbanov [Mon, 7 Mar 2022 19:24:18 +0000 (22:24 +0300)]
[OpenMP] libomp: implemented task priorities.
Before this patch task priorities were ignored, that was a valid implementation
as the task priority is a hint according to OpenMP specification.
Implemented shared list of sorted (high -> low) task deques one per task
priority value. Tasks execution changed to first check if priority tasks ready
for execution exist, and these tasks executed before others;
otherwise usual tasks execution mechanics work.
Differential Revision: https://reviews.llvm.org/D119676
Jinsong Ji [Mon, 7 Mar 2022 19:08:57 +0000 (14:08 -0500)]
[libc++][AIX][test] Enable put_double/long_double locale tests
AIX print -0.0 , inf, nan differently, which are causing the test
failures. We are OK for most other tests.
This patch remove the tests related these limitations conditionally on AIX,
so that we can enable the other tests to avoid losing test coverage.
The general direction is:
```
if strings don't differ between environments, keep the string literal "INF" and the padding, instead of folding them into variables.
```
Reviewed By: ldionne, #libc, Mordante
Differential Revision: https://reviews.llvm.org/D120448
Haojian Wu [Mon, 7 Mar 2022 09:13:12 +0000 (10:13 +0100)]
[pseudo] empty parameter-declaration should be allowed in lambda declarator.
This was an oversight, as we did a avoild-nullable modication to
parameter-declaration-clause.
Differential Revision: https://reviews.llvm.org/D121089
Valentin Clement [Mon, 7 Mar 2022 18:55:48 +0000 (19:55 +0100)]
[flang] Add lowering for host association
This patches adds the code to handle host association for
inner subroutines and functions.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D121134
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Nico Weber [Mon, 7 Mar 2022 18:50:38 +0000 (13:50 -0500)]
Revert "[lld-macho][nfc] Don't use `stubsHelperIndex` in ICF hash"
This reverts commit
112135e77444e8c8106efa77416af09f4b9a5012.
Breaks lld/test/MachO/{icf.s,cfstring-dedup.s,invalid/cfstring.s}
Peter Klausler [Fri, 18 Feb 2022 22:08:57 +0000 (14:08 -0800)]
[flang] Accommodate arrays with a zero-extent dimension in location folding
The index incrementation code used for FINDLOC, MAXLOC, and MINLOC folding
would crash if the array had a zero extent on the dimension selected with
a DIM= argument since the subscript passed to IncrementSubscripts would
have a value less than the lower bound. Fix, and add tests.
Differential Revision: https://reviews.llvm.org/D121125
Nico Weber [Mon, 7 Mar 2022 18:43:29 +0000 (13:43 -0500)]
Revert "[gn build] (manually) port
0c2b43ab8cb1 (X86MCTests)"
This reverts commit
336e3677eac3f0ab7af212505723ef5041bacb6c.
0c2b43ab8cb1 was reverted in
cf9b3ef941d4b4.
Maksim Panchenko [Mon, 7 Mar 2022 18:40:48 +0000 (10:40 -0800)]
Revert "[X86] Fix MCSymbolizer interface for X86Disassembler"
This reverts commit
0c2b43ab8cb1067dd1c7899094b824890803a7d2.
Nico Weber [Mon, 7 Mar 2022 18:36:44 +0000 (13:36 -0500)]
[gn build] (manually) port
0c2b43ab8cb1 (X86MCTests)
Arthur O'Dwyer [Wed, 2 Mar 2022 22:22:55 +0000 (17:22 -0500)]
[libc++] [ranges] Fix `decltype(auto) ranges::iter_move`.
See
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92894#c3
https://reviews.llvm.org/D119589#inline-1151299
Differential Revision: https://reviews.llvm.org/D120417
Nico Weber [Mon, 7 Mar 2022 14:42:35 +0000 (09:42 -0500)]
[clang-format] Fix namespace end comments in ObjC++ files too
See also
d96ae867351ec.
Differential Revision: https://reviews.llvm.org/D121112
Arthur O'Dwyer [Fri, 4 Mar 2022 23:24:18 +0000 (18:24 -0500)]
[libc++] Overhaul std::quoted; fix its relationship to character traits.
Move `__quoted_output_proxy` into the one file that uses it.
A `const char*` has no associated traits class, so `std::quoted("literal")`
should be printable into any basic_ostream regardless of traits.
Use hidden-friend `operator<<` and `operator>>`, since we're permitted to.
(The exact signature is unspecified because the class itself is unspecified.)
We shouldn't support `std::quoted("literal")` in C++03 or C++11 mode.
(We do need `std::__quoted(s)` and `std::__quoted(cs)` in C++11 mode,
because they're used by `std::__fs::filesystem::path`.)
Differential Revision: https://reviews.llvm.org/D120135
Maksim Panchenko [Mon, 7 Mar 2022 18:27:28 +0000 (10:27 -0800)]
[X86] Fix MCSymbolizer interface for X86Disassembler
Fix a number of issues with MCSymbolizer::tryAddingSymbolicOperand()
in X86Disassembler:
* Pass instruction size instead of immediate size.
* Correctly adjust the value of PC-relative operands.
* Set operand offset to zero when the operand is specified
implicitly.
Reviewed By: Amir, skan
Differential Revision: https://reviews.llvm.org/D121065
David Blaikie [Mon, 7 Mar 2022 17:56:51 +0000 (17:56 +0000)]
Simplify OpenMP Lambda use
* Use default ref capture for non-escaping lambdas (this makes
maintenance easier by allowing new uses, removing uses, having
conditional uses (such as in assertions) not require updates to an
explicit capture list)
* Simplify addPrivate API not to take a lambda, since it calls it
unconditionally/immediately anyway - most callers are simply passing
in a named value or short expression anyway and the lambda syntax just
adds noise/overhead
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D121077
Yuanfang Chen [Mon, 7 Mar 2022 18:15:37 +0000 (10:15 -0800)]
[clang][debug] port clang-cl /JMC flag to ELF
The motivation is to enable the MSVC-style JMC instrumentation usable by a ELF-based
debugger. Since there is no prior experience implementing JMC feature for ELF-based
debugger, it might be better to just reuse existing MSVC-style JMC instrumentation.
For debuggers that support both ELF&COFF (like lldb), the JMC implementation might
be shared between ELF&COFF. If this is found to inadequate, it is pretty low-cost
switching to alternatives.
Implementation:
- The '-fjmc' is already a driver and cc1 flag. Wire it up for ELF in the driver.
- Refactor the JMC instrumentation pass a little bit.
- The ELF handling is different from MSVC in two places:
* the flag section name is ".just.my.code" instead of ".msvcjmc"
* the way default function is provided: MSVC uses /alternatename; ELF uses weak function.
Based on D118428.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D119910
Bixia Zheng [Fri, 4 Mar 2022 21:25:10 +0000 (13:25 -0800)]
[mlir][sparse][taco] Add more unit tests.
These unit tests resides in an internal repository. Porting the tests to the
public repository.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D121021
Yi Kong [Mon, 7 Mar 2022 18:01:15 +0000 (02:01 +0800)]
Revert "Add CMake option not to build BOLT tests"
This reverts commit
d8f4d54664171bac46e6c8446df04850b40cc4f7.
Merged by accident.
Yi Kong [Mon, 7 Mar 2022 17:59:06 +0000 (01:59 +0800)]
Fix the comment for EnableNoundefAttrs, NFC
Yi Kong [Wed, 2 Mar 2022 16:43:24 +0000 (00:43 +0800)]
Add CMake option not to build BOLT tests
Diego Caballero [Mon, 7 Mar 2022 17:40:21 +0000 (17:40 +0000)]
[mlir][Vector] Improve default lowering of vector transpose operations
The default lowering of vector transpose operations generates a large sequence of
scalar extract/insert operations, one pair for each scalar element in the input tensor.
In other words, the vector transpose is scalarized. However, there are transpose
patterns where one or more adjacent high-order dimensions are not transposed (for
example, in the transpose pattern [1, 0, 2, 3], dimensions 2 and 3 are not transposed).
This patch improves the lowering of those cases by not scalarizing them and extracting/
inserting a full n-D vector, where 'n' is the number of adjacent high-order dimensions
not being transposed. By doing so, we prevent the scalarization of the code and generate a
more performant vector version.
Paradoxically, this patch shouldn't improve the performance of transpose operations if
we are using LLVM. The LLVM pipeline is able to optimize away some of the extract/insert
operations and the SLP vectorizer is converting the scalar operations back to its vector
form. However, scalarizing a vector version of the code in MLIR and relying on the SLP
vectorizer to reconstruct the vector code again is highly undesirable for several reasons.
Reviewed By: nicolasvasilache, ThomasRaoux
Differential Revision: https://reviews.llvm.org/D120601
Stanislav Gatev [Fri, 4 Mar 2022 11:03:29 +0000 (11:03 +0000)]
[clang][dataflow] Extend flow conditions from block terminators
This is part of the implementation of the dataflow analysis framework.
See "[RFC] A dataflow analysis framework for Clang AST" on cfe-dev.
Reviewed-by: ymandel, xazax.hun
Differential Revision: https://reviews.llvm.org/D120984
Jez Ng [Mon, 7 Mar 2022 17:35:49 +0000 (12:35 -0500)]
[lld-macho][nfc] Reduce size of icfEqClass hash
... from a `uint64_t` to a `uint32_t`. (LLD-ELF uses a `uint32_t` too.)
About a 1.7% reduction in peak RSS when linking chromium_framework on my
3.2 GHz 16-Core Intel Xeon W Mac Pro, and no stat sig change in wall
time.
</Users/jezng/test2.sh ["before"]> </Users/jezng/test2.sh ["after"]> difference (95% CI)
RSS
1003036672.000 ± 9891065.259
985539505.231 ±
10272748.749 [ -2.3% .. -1.2%]
samples 27 26
base diff difference (95% CI)
sys_time 1.277 ± 0.023 1.277 ± 0.024 [ -0.9% .. +0.9%]
user_time 6.682 ± 0.046 6.598 ± 0.043 [ -1.6% .. -0.9%]
wall_time 5.904 ± 0.062 5.895 ± 0.063 [ -0.7% .. +0.4%]
samples 46 28
No appreciable change (~0.01%) in number of `equals` comparisons either:
Before:
ld64.lld: ICF needed 8 iterations
ld64.lld: equalsConstant() called 701643 times
ld64.lld: equalsVariable() called 3438526 times
After:
ld64.lld: ICF needed 8 iterations
ld64.lld: equalsConstant() called 701729 times
ld64.lld: equalsVariable() called 3438526 times
Reviewed By: #lld-macho, MaskRay, thakis
Differential Revision: https://reviews.llvm.org/D121052
Jez Ng [Mon, 7 Mar 2022 08:02:40 +0000 (03:02 -0500)]
[lld-macho][nfc] Don't use `stubsHelperIndex` in ICF hash
The existing hashing of stubsHelperIndex has mostly been a no-op* for
some time now (ever since we made ICF run before dylib symbols get their
stubs indices assigned). I guess we could consider hashing the name +
filename of the DylibSymbol instead, but I'm not sure the overhead's
worth it... moreover, LLD/ELF only hashes their Defined symbols as well.
*: Technically it does change the hash value since stubsHelperIndex is
initialized to `UINT32_MAX` by default. But since all stubsHelperIndex
values are the same at when ICF runs, they don't add any useful
information to the hash.
Jez Ng [Mon, 7 Mar 2022 04:27:35 +0000 (23:27 -0500)]
[lld-macho][nfc] Rename isec -> referentIsec to avoid shadowing
I found the shadowing a bit confusing
Jez Ng [Mon, 7 Mar 2022 17:35:15 +0000 (12:35 -0500)]
[lld-macho][nfc] Track # of ICF calls to `equals*` methods
This is debug code that is disabled by default. It'll provide a easy way
to figure out the impact (if any) of tweaking ICF's hashing algorithm
(since a poor quality hash will result in many more `equals*` calls).
Reviewed By: #lld-macho, oontvoo
Differential Revision: https://reviews.llvm.org/D121051
Jez Ng [Sat, 5 Mar 2022 01:22:40 +0000 (20:22 -0500)]
[lld-macho][nfc] Use llvm::function_ref instead of std::function
Jez Ng [Mon, 7 Mar 2022 17:35:12 +0000 (12:35 -0500)]
[lld-macho][nfc] Remove file statics from ICF.cpp
This gets us closer to the [LLD-as-a-library goal][1].
[1]: https://lists.llvm.org/pipermail/llvm-dev/2021-June/151184.html
Reviewed By: #lld-macho, thakis
Differential Revision: https://reviews.llvm.org/D121050
Roman Lebedev [Mon, 7 Mar 2022 17:25:09 +0000 (20:25 +0300)]
UpdateTestChecks: fix handling of UTC with spaces
We can't just split by space, that's not going to give us the same
argv we'd have gotten from the shell, it could be in a string,
we must actually parse that as argv.
zhongyunde [Mon, 7 Mar 2022 16:31:07 +0000 (00:31 +0800)]
[AArch64] Perform first active true vector combine
Materialize : i1 = extract_vector_elt t37, Constant:i64<0>
... into: "ptrue p, all" + PTEST
Test bit of lane 0 can use P register directly, and the instruction “pture all”
is loop invariant, which will beneficial to SVE after hoisting out the loop.
Reviewed By: david-arm, paulwalker-arm
Differential Revision: https://reviews.llvm.org/D120891