Craig Topper [Wed, 26 Aug 2020 16:45:15 +0000 (09:45 -0700)]
[LegalizeTypes] Remove WidenVecRes_Shift and just use WidenVecRes_Binary
This function seems to allow for the shift amount to have a different type than the result, but I don't think we do that anywhere else for vector shifts. We also don't have any support for legalizing the shift amount alone if the result is legal and the shift amount type isn't. The code coverage report here shows this code as uncovered http://lab.llvm.org:8080/coverage/coverage-reports/coverage/Users/buildslave/jenkins/workspace/coverage/llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp.html
Differential Revision: https://reviews.llvm.org/D86475
Eduardo Caldas [Wed, 26 Aug 2020 08:51:21 +0000 (08:51 +0000)]
[SyntaxTree] Migrate `ParamatersAndQualifiers` to use the new List API
Fix: Add missing `List::getTerminationKind()`, `List::canBeEmpty()`,
`List::getDelimiterTokenKind()` for `CallArguments`.
Differential Revision: https://reviews.llvm.org/D86600
Kai Nacke [Wed, 20 May 2020 09:10:31 +0000 (11:10 +0200)]
[SystemZ/ZOS] Add header file to encapsulate use of <sysexits.h>
The non-standard header file `<sysexits.h>` provides some return values.
`EX_IOERR` is used to as a special value to signal a broken pipe to the clang driver.
On z/OS Unix System Services, this header file does not exists. This patch
- adds a check for `<sysexits.h>`, removing the dependency on `LLVM_ON_UNIX`
- adds a new header file `llvm/Support/ExitCodes`, which either includes
`<sysexits.h>` or defines `EX_IOERR`
- updates the users of `EX_IOERR` to include the new header file
Reviewed By: hubert.reinterpretcast
Differential Revision: https://reviews.llvm.org/D83472
Matt Morehouse [Wed, 26 Aug 2020 16:06:12 +0000 (09:06 -0700)]
[libFuzzer] Error and exit if user supplied fuzzer writeable directories don't exist
Currently, libFuzzer will exit with an error message if a non-existent
corpus directory is provided. However, if a user provides a non-existent
directory for the `artifact_prefix`, `exact_artifact_path`, or
`features_dir`, libFuzzer will continue execution but silently fail to
write artifacts/features.
To improve the user experience, this PR adds validation for the existence of
all user supplied directories before executing the main fuzzing loop. If they
don't exist, libFuzzer will exit with an error message.
Patch By: dgg5503
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D84808
Jeremy Morse [Wed, 26 Aug 2020 16:19:13 +0000 (17:19 +0100)]
Revert "[LiveDebugValues][NFC] Add instr-ref tests, adapt old tests"
Asan has picked up a memory leak, which will require some investigation:
http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fast/builds/45372/steps/check-llvm%20asan/logs/stdio
This reverts commit
60db26a66d5346b2feae279f6c928bde48078893.
Gi Vald [Tue, 25 Aug 2020 20:37:14 +0000 (22:37 +0200)]
Fix quiet mode in git-clang-format
Quiet mode is very useful for scripting, when only the diff format output is
required, or no output if not formatting is needed.
In case of no modified files, git-clang-format will output to screen even though
the quiet mode enabled.
This patch changes this behavior, so if quiet flag passes in - no output will be
available, even if no modified files exists.
Differential Revision: https://reviews.llvm.org/D85485
Owen Anderson [Wed, 26 Aug 2020 16:09:28 +0000 (16:09 +0000)]
Fix frame pointer layout on AArch64 Linux.
When floating point callee-saved registers were used, the frame pointer would
incorrectly point to the bottom of the CSR space (containing saved floating-point
registers), rather than to the frame record.
While all frame offsets were calculated consistently, resulting in working code,
this prevented stack walkers from being about to traverse the frame list.
Sjoerd Meijer [Wed, 26 Aug 2020 15:55:25 +0000 (16:55 +0100)]
[LV] Fallback strategies if tail-folding fails
This implements 2 different vectorisation fallback strategies if tail-folding
fails: 1) don't vectorise at all, or 2) vectorise using a scalar epilogue. This
can be controlled with option -prefer-predicate-over-epilogue, that has been
changed to take a numeric value corresponding to the tail-folding preference
and preferred fallback.
Patch by: Pierre van Houtryve, Sjoerd Meijer.
Differential Revision: https://reviews.llvm.org/D79783
Jeremy Morse [Tue, 25 Aug 2020 15:43:47 +0000 (16:43 +0100)]
[LiveDebugValues][NFC] Add instr-ref tests, adapt old tests
This patch adds a few tests in DebugInfo/MIR/InstrRef/ of interesting
behaviour that the instruction referencing implementation of
LiveDebugValues has. Mostly, these tests exist to ensure that if you
give the "-experimental-debug-variable-locations" command line switch,
the right implementation runs; and to ensure it behaves the same way as
the VarLoc LiveDebugValues implementation.
I've also touched roughly 30 other tests, purely to make the tests less
rigid about what output to accept. DBG_VALUE instructions are usually
printed with a trailing !debug-location indicating its scope:
!debug-location !1234
However InstrRefBasedLDV produces new DebugLoc instances on the fly,
meaning there sometimes isn't a numbered node when they're printed,
making the output:
!debug-location !DILocation(line: 0, blah blah)
Which causes a ton of these tests to fail. This patch removes checks for
that final part of each DBG_VALUE instruction. None of them appear to
be actually checking the scope is correct, just that it's present, so
I don't believe there's any loss in coverage here.
Differential Revision: https://reviews.llvm.org/D83054
Louis Dionne [Wed, 26 Aug 2020 15:32:40 +0000 (11:32 -0400)]
[libc++] Use xcrun to find Ninja in the macOS backdeployment CI too
Alexandre Ganea [Wed, 26 Aug 2020 15:32:17 +0000 (11:32 -0400)]
[Clang] Fix tests following rG087047144210
Raphael Isemann [Wed, 26 Aug 2020 14:37:35 +0000 (16:37 +0200)]
[clang][NFC] Properly fix a GCC warning in ASTImporterTest.cpp
Follow up to
c9b45ce1fd97531c228e092bedee719b971f82a3 which just defined
the function instead of just 'using' the function from the base class (thanks
David).
Kirill Bobyrev [Wed, 26 Aug 2020 15:08:00 +0000 (17:08 +0200)]
[clangd] Use string[] for allCommitCharacters
As per LSP specification, allCommitCharacters should be string[] instead of
string:
https://microsoft.github.io/language-server-protocol/specification#textDocument_completion
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D86604
Louis Dionne [Wed, 26 Aug 2020 14:44:38 +0000 (10:44 -0400)]
[libc++] Always run Ninja through xcrun in the macOS CI scripts
Ninja isn't installed by default on OSX, so run it through xcrun to find
the one in the developer tools if needed.
Haojian Wu [Wed, 26 Aug 2020 14:46:24 +0000 (16:46 +0200)]
[clangd] Enable recovery-ast-type by default.
Differential Revision: https://reviews.llvm.org/D86602
Hans Wennborg [Wed, 26 Aug 2020 14:44:55 +0000 (16:44 +0200)]
Bump -len_control value in fuzzer-custommutator.test (PR47286)
to make the test more stable, as suggested by mmoroz.
Zachary Henkel [Wed, 26 Aug 2020 14:30:33 +0000 (16:30 +0200)]
Fix failing tests after VCTOOLSDIR change
Switch from hardcoded x64 arch to a regex in the target triple
Differential revision: https://reviews.llvm.org/D86622
Raphael Isemann [Wed, 26 Aug 2020 09:49:28 +0000 (11:49 +0200)]
[lldb][NFC] Simplify string literal in GDBRemoteCommunicationClient
Jay Foad [Wed, 26 Aug 2020 13:53:41 +0000 (14:53 +0100)]
[AMDGPU] Make more use of Subtarget reference in SIInstrInfo
AndreyChurbanov [Wed, 26 Aug 2020 13:52:46 +0000 (16:52 +0300)]
[OpenMP] Fix build on macOS sdk 10.12 and newer
Patch by nihui (Ni Hui)
Differential Revision: https://reviews.llvm.org/D76755
Tue Ly [Tue, 28 Jul 2020 05:41:36 +0000 (01:41 -0400)]
[libc] Add implementations for sqrt, sqrtf, and sqrtl.
Differential Revision: https://reviews.llvm.org/D84726
Jay Foad [Wed, 26 Aug 2020 08:41:23 +0000 (09:41 +0100)]
[LegalizeTypes] Add ROTL/ROTR to ScalarizeVectorResult.
We can scalarize these just like any other binary operation.
Fixes https://bugs.llvm.org/show_bug.cgi?id=47303 caused by D77152.
Differential Revision: https://reviews.llvm.org/D86601
Dibya Ranjan Mishra [Wed, 26 Aug 2020 13:21:34 +0000 (09:21 -0400)]
[Support] Allow printing the stack trace only for a given depth
Differential Revision: https://reviews.llvm.org/D85458
Matt Arsenault [Wed, 26 Aug 2020 13:15:25 +0000 (09:15 -0400)]
AMDGPU: Use Subtarget reference in SIInstrInfo
Zachary Henkel [Wed, 26 Aug 2020 12:58:39 +0000 (14:58 +0200)]
Add clang-cl "vctoolsdir" option to specify the location of the msvc toolchain
Add an option to directly specify where the msvc toolchain lives for
clang-cl and avoid unwanted file and registry probes.
Differential revision: https://reviews.llvm.org/D85998
Matt Arsenault [Tue, 25 Aug 2020 23:53:32 +0000 (19:53 -0400)]
AMDGPU/GlobalISel: Tolerate negated control flow intrinsic outputs
If the condition output is negated, swap the branch targets. This is
similar to what SelectionDAG does for when SelectionDAGBuilder
decides to invert the condition and swap the branches.
This is leaving behind a dead constant def for some reason.
Matt Arsenault [Sun, 16 Aug 2020 15:49:41 +0000 (11:49 -0400)]
GlobalISel: Combine G_ADD of G_PTRTOINT to G_PTR_ADD
This produces less work for addressing mode matching. I think this is
safe since I don't think machine IR is supposed to give the same
aliasing properties as getelementptr in the IR.
Jay Foad [Wed, 26 Aug 2020 10:41:39 +0000 (11:41 +0100)]
[AMDGPU][GlobalISel] Eliminate barrier if workgroup size is not greater than wavefront size
If a workgroup size is known to be not greater than wavefront size
the s_barrier instruction is not needed since all threads are guaranteed
to come to the same point at the same time.
This is the same optimization that was implemented for SelectionDAG in
D31731.
Differential Revision: https://reviews.llvm.org/D86609
Xing GUO [Wed, 26 Aug 2020 12:34:36 +0000 (20:34 +0800)]
[DWARFYAML] Make the unit_length and header_length fields optional.
This patch makes the unit_length and header_length fields of line tables
optional. yaml2obj is able to infer them for us.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D86590
QingShan Zhang [Wed, 26 Aug 2020 12:26:21 +0000 (12:26 +0000)]
[Scheduling] Implement a new way to cluster loads/stores
Before calling target hook to determine if two loads/stores are clusterable,
we put them into different groups to avoid fake cluster due to dependency.
For now, we are putting the loads/stores into the same group if they have
the same predecessor. We assume that, if two loads/stores have the same
predecessor, it is likely that, they didn't have dependency for each other.
However, one SUnit might have several predecessors and for now, we just
pick up the first predecessor that has non-data/non-artificial dependency,
which is too arbitrary. And we are struggling to fix it.
So, I am proposing some better implementation.
1. Collect all the loads/stores that has memory info first to reduce the complexity.
2. Sort these loads/stores so that we can stop the seeking as early as possible.
3. For each load/store, seeking for the first non-dependency instruction with the
sorted order, and check if they can cluster or not.
Reviewed By: Jay Foad
Differential Revision: https://reviews.llvm.org/D85517
River Riddle [Wed, 26 Aug 2020 12:12:07 +0000 (05:12 -0700)]
[mlir][PDL] Add a PDL Interpreter Dialect
The PDL Interpreter dialect provides a lower level abstraction compared to the PDL dialect, and is targeted towards low level optimization and interpreter code generation. The dialect operations encapsulates low-level pattern match and rewrite "primitives", such as navigating the IR (Operation::getOperand), creating new operations (OpBuilder::create), etc. Many of the operations within this dialect also fuse branching control flow with some form of a predicate comparison operation. This type of fusion reduces the amount of work that an interpreter must do when executing.
An example of this representation is shown below:
```mlir
// The following high level PDL pattern:
pdl.pattern : benefit(1) {
%resultType = pdl.type
%inputOperand = pdl.input
%root, %results = pdl.operation "foo.op"(%inputOperand) -> %resultType
pdl.rewrite %root {
pdl.replace %root with (%inputOperand)
}
}
// May be represented in the interpreter dialect as follows:
module {
func @matcher(%arg0: !pdl.operation) {
pdl_interp.check_operation_name of %arg0 is "foo.op" -> ^bb2, ^bb1
^bb1:
pdl_interp.return
^bb2:
pdl_interp.check_operand_count of %arg0 is 1 -> ^bb3, ^bb1
^bb3:
pdl_interp.check_result_count of %arg0 is 1 -> ^bb4, ^bb1
^bb4:
%0 = pdl_interp.get_operand 0 of %arg0
pdl_interp.is_not_null %0 : !pdl.value -> ^bb5, ^bb1
^bb5:
%1 = pdl_interp.get_result 0 of %arg0
pdl_interp.is_not_null %1 : !pdl.value -> ^bb6, ^bb1
^bb6:
pdl_interp.record_match @rewriters::@rewriter(%0, %arg0 : !pdl.value, !pdl.operation) : benefit(1), loc([%arg0]), root("foo.op") -> ^bb1
}
module @rewriters {
func @rewriter(%arg0: !pdl.value, %arg1: !pdl.operation) {
pdl_interp.replace %arg1 with(%arg0)
pdl_interp.return
}
}
}
```
Differential Revision: https://reviews.llvm.org/D84579
Georgii Rymar [Tue, 25 Aug 2020 14:44:39 +0000 (17:44 +0300)]
[llvm/Object] - Make dyn_cast<XCOFFObjectFile> work as it should.
Currently, `dyn_cast<XCOFFObjectFile>` always does cast and returns a pointer,
even when we pass `ELF`/`Wasm`/`Mach-O` or `COFF` instead of `XCOFF`.
It happens because `XCOFFObjectFile` class does not implement `classof`.
I've fixed it and added a unit test.
Differential revision: https://reviews.llvm.org/D86542
David Green [Wed, 26 Aug 2020 12:03:46 +0000 (13:03 +0100)]
[ARM] Increase MVE gather/scatter cost by MVECostFactor.
MVE Gather scatter codegeneration is looking a lot better than it used
to, but still has some issues. The instructions we currently model as 1
cycle per element, which is a bit low for some cases. Increasing the
cost by the MVECostFactor brings them in-line with our other instruction
costs. This will have the effect of only generating then when the extra
benefit is more likely to overcome some of the issues. Notably in
running out of registers and vectorizing loops that could otherwise be
SLP vectorized.
In the short-term whilst we look at other ways of dealing with those
more directly, we can increase the costs of gathers to make them more
likely to be beneficial when created.
Differential Revision: https://reviews.llvm.org/D86444
Georgii Rymar [Wed, 26 Aug 2020 11:56:40 +0000 (14:56 +0300)]
[llvm-readobj][test] - Commit trivial.obj.elf-amdhsa-gfx803 binary back.
It was removed in rGcbedbd12e9837e049f0a936636a82ff39b75692b by mistake.
Georgii Rymar [Fri, 21 Aug 2020 11:24:12 +0000 (14:24 +0300)]
[llvm-readobj/elf][test] - Add testing for EM_* specific OS/ABI values.
We have no tests for OS/ABI values specific to
EM_TI_C6000, ELFOSABI_AMDGPU_MESA3D and ELFOSABI_ARM machines.
Also, related arrays in the code are not grouped together.
(That is why such testing was missed I guess).
The patch fixes that all.
Differential revision: https://reviews.llvm.org/D86341
Sam Tebbs [Wed, 26 Aug 2020 10:11:15 +0000 (11:11 +0100)]
[RDA] Don't visit the BB of the instruction in getReachingUniqueMIDef
If the basic block of the instruction passed to getUniqueReachingMIDef
is a transitive predecessor of itself and has a definition of the
register, the function will return that definition even if it is after
the instruction given to the function. This patch stops the function
from scanning the instruction's basic block to prevent this.
Differential Revision: https://reviews.llvm.org/D86607
LLVM GN Syncbot [Wed, 26 Aug 2020 11:33:42 +0000 (11:33 +0000)]
[gn build] Port
357dc1ed125
Martin Storsjö [Sun, 16 Aug 2020 12:12:55 +0000 (15:12 +0300)]
[libunwind] Convert x86, x86_64, arm64 register restore functions to C calling convention and name mangling
Currently, the assembly functions for restoring register state have
been direct implementations of the Registers_*::jumpto() method
(contrary to the functions for saving register state, which are
implementations of the extern C function __unw_getcontext). This has
included having the assembly function name match the C++ mangling of
that method name (and having the function match the C++ member
function calling convention). To simplify the interface of the assembly
implementations, make the functions have C calling conventions and
name mangling.
This fixes building the library in with a MSVC C++ ABI with clang-cl,
which uses a significantly different method name mangling scheme.
(The library might not be of much use as C++ exception unwinder in such
an environment, but the libunwind.h interface for stepwise unwinding
still is usable, as is the _Unwind_Backtrace function.)
Differential Revision: https://reviews.llvm.org/D86041
Benjamin Kramer [Wed, 26 Aug 2020 11:25:41 +0000 (13:25 +0200)]
Copy m_plan_is_for_signal_trap member.
Otherwise it would stay uninitialized. Found by msan.
Georgii Rymar [Tue, 25 Aug 2020 11:55:25 +0000 (14:55 +0300)]
[obj2yaml] - Cleanup error reporting (remove Error.cpp/.h files)
This removes Error.cpp/.h files from obj2yaml.
These files are not needed because we are
using `Error`s instead of error codes widely and do
not need a logic related to obj2yaml specific
error codes anymore.
I had to adjust just a few lines of tool's code
to remove remaining dependencies.
Differential revision: https://reviews.llvm.org/D86536
Pavel Labath [Fri, 21 Aug 2020 13:13:29 +0000 (15:13 +0200)]
[lldb/DWARF] More DW_AT_const_value fixes
This fixes several issues in handling of DW_AT_const_value attributes:
- the first is that the size of the data given by data forms does not
need to match the size of the underlying variable. We already had the
case to handle this for DW_FORM_(us)data -- this extends the handling
to other data forms. The main reason this was not picked up is because
clang uses leb forms in these cases while gcc prefers the fixed-size
ones.
- The handling of DW_AT_strp form was completely broken -- we would end
up using the pointer value as the result. I've reorganized this code
so that it handles all string forms uniformly.
- In case of a completely bogus form we would crash due to
strlen(nullptr).
Depends on D86311.
Differential Revision: https://reviews.llvm.org/D86348
Georgii Rymar [Tue, 25 Aug 2020 10:37:32 +0000 (13:37 +0300)]
[llvm-readobj] - Don`t crash when --section-symbols is requested for an object w/o .symtab.
llvm-readobj crashes when `-S --section-symbols` is used
on an object that has no symbol table.
The patch fixes it.
Differential revision: https://reviews.llvm.org/D86520
Georgii Rymar [Mon, 24 Aug 2020 13:24:36 +0000 (16:24 +0300)]
[llvm-readelf][test] - Refine the sections-ext.test
The `sections-ext.test` is a test for ELF that is used
to test `--st`, `--sr` and `--sd` extension options for `-S`.
There are 2 problems with it:
1) It is broken, because for CHECK lines it contains there is
no corresponding `FileCheck` call.
2) It uses the precompiled object: `trivial.obj.elf-i386`.
This is the last ELF test where `trivial.obj.elf-i386` is used so we can get
rid of the binary and use an YAML description.
Also, there is a `Inputs/trivial.ll` file that describes how `trivial*` objects
in `Inputs` folders are created. I've removed it from `ELF`, because it is not
actual anymore (we have no more input binaries created with the use of trivial.ll there)
and copied the refined versions of it to `COFF`, `MachO` and `wasm` Input folders.
Differential revision: https://reviews.llvm.org/D86462
Kai Nacke [Thu, 9 Jul 2020 12:17:33 +0000 (14:17 +0200)]
[SystemZ/ZOS] Additions to the build system.
This change extend the CMake files with the necessary additions
to build LLVM for z/OS.
Reviewed By: hubert.reinterpretcast
Differential Revision: https://reviews.llvm.org/D83866
David Spickett [Tue, 25 Aug 2020 15:50:03 +0000 (16:50 +0100)]
[lldb] Correct wording of EXP_MSG
EXP_MSG generates a message to show on assert
failure. Currently it looks like:
AssertionError: False is not True : '<cmd>'
returns expected result, got '<actual output>'
Which seems to say that the test failed but
also got the expected result.
It should say:
AssertionError: False is not True : '<cmd>'
returned unexpected result, got '<actual output>'
Reviewed By: teemperor, #lldb
Differential Revision: https://reviews.llvm.org/D86603
Pierre Gousseau [Wed, 15 Jan 2020 18:33:28 +0000 (18:33 +0000)]
[X86] Make sure we do not clobber RBX with mwaitx when used as a base
pointer.
mwaitx uses EBX as one of its argument.
Using this instruction clobbers RBX as it is defined to hold one of the
input. When the backend uses dynamically allocated stack, RBX is used as
a reserved register for the base pointer.
This patch is adapted from @qcolombet patch for cmpxchg at r263325.
This fixes PR43528.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D73475
Gabriel Hjort Åkerlund [Wed, 26 Aug 2020 08:53:02 +0000 (10:53 +0200)]
[GlobalISel] Fix and tidy up documentation for ValueMapping class (NFC)
The documentation was missing a '*/' in '/*<2x32-bit> vadd {0, 64, VPR}',
and the example code are now aligned to improve readability.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D86201
Gabriel Hjort Åkerlund [Wed, 26 Aug 2020 08:48:15 +0000 (10:48 +0200)]
[TableGen][GlobalISel] Fix tblgen optimization bug
When optimizing the table, PointerToAnyOperandMatchers would be
incorrectly reported as identical even though they have different
SizeInBits values. This bug was due to failing to overload the
isIdentical() method, which this patch addresses.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D86199
sstefan1 [Wed, 26 Aug 2020 08:37:48 +0000 (10:37 +0200)]
Reland [IR] Intrinsics default attributes and opt-out flag
Intrinsic properties can now be set to default and applied to all
intrinsics. If the attributes are not needed, the user can opt-out by
setting the DisableDefaultAttributes flag to true.
Differential Revision: https://reviews.llvm.org/D70365
Cullen Rhodes [Tue, 25 Aug 2020 12:27:15 +0000 (12:27 +0000)]
[AArch64][AsmParser] Fix bug in operand printer
The switch in AArch64Operand::print was changed in D45688 so the shift
can be printed after printing the register. This is implemented with
LLVM_FALLTHROUGH and was broken in D52485 when BTIHint was put between
the register and shift operands.
Reviewed By: ostannard
Differential Revision: https://reviews.llvm.org/D86535
Nithin Vadukkumchery Rajendrakumar [Wed, 26 Aug 2020 09:22:55 +0000 (11:22 +0200)]
[analyzer] Add modeling of assignment operator in smart ptr
Summary: Support for 'std::unique_ptr>::operator=' in SmartPtrModeling
Reviewers: NoQ, Szelethus, vsavchenko, xazax.hun
Reviewed By: NoQ, vsavchenko, xazax.hun
Subscribers: martong, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D86293
Sander de Smalen [Wed, 19 Aug 2020 10:06:51 +0000 (11:06 +0100)]
[AArch64][SVE] Fix calculation restore point for SVE callee saves.
This fixes an issue where the restore point of callee-saves in the
function epilogues was incorrectly calculated when the basic block
consisted of only a RET instruction. This caused dealloc instructions
to be inserted in between the block of callee-save restore instructions,
rather than before it.
Reviewed By: paulwalker-arm
Differential Revision: https://reviews.llvm.org/D86099
Yang Zhihui [Wed, 26 Aug 2020 08:28:50 +0000 (04:28 -0400)]
[NFC] Fix some spelling errors in clang Driver Options.td
Differential Revision: https://reviews.llvm.org/D86427
Haojian Wu [Wed, 26 Aug 2020 08:50:31 +0000 (10:50 +0200)]
[clangd] Compute the inactive code range for semantic highlighting.
Differential Revision: https://reviews.llvm.org/D85635
Jan Kratochvil [Wed, 26 Aug 2020 07:57:53 +0000 (09:57 +0200)]
[Support] Speedup llvm-dwarfdump 3.9x
Currently `strace llvm-dwarfdump x.debug >/tmp/file`:
ioctl(1, TCGETS, 0x7ffd64d7f340) = -1 ENOTTY (Inappropriate ioctl for device)
write(1, " DW_AT_decl_line\t(89)\n"..., 4096) = 4096
ioctl(1, TCGETS, 0x7ffd64d7f400) = -1 ENOTTY (Inappropriate ioctl for device)
ioctl(1, TCGETS, 0x7ffd64d7f410) = -1 ENOTTY (Inappropriate ioctl for device)
ioctl(1, TCGETS, 0x7ffd64d7f400) = -1 ENOTTY (Inappropriate ioctl for device)
After this patch:
write(1, "
0000000000001102 \"strlen\")\n "..., 4096) = 4096
write(1, "site\n DW_AT_low"..., 4096) = 4096
write(1, "d53)\n\n0x000e4d4d: DW_TAG_G"..., 4096) = 4096
The same speedup can be achieved by `--color=0` but that is not much convenient.
This implementation has been suggested by Joerg Sonnenberger.
Differential Revision: https://reviews.llvm.org/D86406
Raphael Isemann [Wed, 26 Aug 2020 08:22:04 +0000 (10:22 +0200)]
[lldb] XFAIL TestMemoryHistory on Linux
This test appears to have never worked on Linux but it seems none of the current
bots ever ran this test as it required enabling compiler-rt (otherwise it
would have just been skipped).
This just copies over the XFAIL decorator that are already on all other sanitizer
tests.
Jay Foad [Mon, 24 Aug 2020 12:24:21 +0000 (13:24 +0100)]
[SelectionDAG] Handle non-power-of-2 bitwidths in expandROT
Differential Revision: https://reviews.llvm.org/D86449
River Riddle [Wed, 26 Aug 2020 08:13:01 +0000 (01:13 -0700)]
[mlir] Fix bug in block merging when the types of the operands differ
The merging algorithm was previously not checking for type equivalence.
Fixes PR47314
Differential Revision: https://reviews.llvm.org/D86594
Shinji Okumura [Wed, 26 Aug 2020 07:57:52 +0000 (16:57 +0900)]
[Attributor] Provide an edge-based interface in AAIsDead
This patch produces an edge-based interface in AAIsDead.
By this, we can query a set of basic blocks that are directly reachable from a given basic block.
This is specifically useful for implementation of AAReachability.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D85547
Eduardo Caldas [Wed, 26 Aug 2020 07:18:01 +0000 (07:18 +0000)]
[SyntaxTree] Fix C++ versions on tests of `BuildTreeTest.cpp`
Differential Revision: https://reviews.llvm.org/D86591
Eduardo Caldas [Tue, 25 Aug 2020 07:47:52 +0000 (07:47 +0000)]
[SyntaxTree] Add support for `CallExpression`
* Generate `CallExpression` syntax node for all semantic nodes inheriting from
`CallExpr` with call-expression syntax - except `CUDAKernelCallExpr`.
* Implement all the accessors
* Arguments of `CallExpression` have their own syntax node which is based on
the `List` base API
Differential Revision: https://reviews.llvm.org/D86544
Roman Lebedev [Wed, 26 Aug 2020 06:08:24 +0000 (09:08 +0300)]
[InstCombine] PHI-of-extractvalues -> extractvalue-of-PHI, aka invokes are bad
While since D86306 we do it's sibling fold for `insertvalue`,
we should also do this for `extractvalue`'s.
And unlike that one, the results here are, quite honestly, shocking,
as it can be observed here on vanilla llvm test-suite + RawSpeed results:
```
| statistic name | baseline | proposed | Δ | % | |%| |
|----------------------------------------------------|-----------|-----------|--------:|--------:|-------:|
| asm-printer.EmittedInsts | 7945095 | 7942507 | -2588 | -0.03% | 0.03% |
| assembler.ObjectBytes |
273209920 |
273069800 | -140120 | -0.05% | 0.05% |
| early-cse.NumCSE | 2183363 | 2183398 | 35 | 0.00% | 0.00% |
| early-cse.NumSimplify | 541847 | 550017 | 8170 | 1.51% | 1.51% |
| instcombine.NumAggregateReconstructionsSimplified | 2139 | 108 | -2031 | -94.95% | 94.95% |
| instcombine.NumCombined | 3601364 | 3635448 | 34084 | 0.95% | 0.95% |
| instcombine.NumConstProp | 27153 | 27157 | 4 | 0.01% | 0.01% |
| instcombine.NumDeadInst | 1694521 | 1765022 | 70501 | 4.16% | 4.16% |
| instcombine.NumPHIsOfExtractValues | 0 | 37546 | 37546 | 0.00% | 0.00% |
| instcombine.NumSunkInst | 63158 | 63686 | 528 | 0.84% | 0.84% |
| instcount.NumBrInst | 874304 | 871857 | -2447 | -0.28% | 0.28% |
| instcount.NumCallInst | 1757657 | 1758402 | 745 | 0.04% | 0.04% |
| instcount.NumExtractValueInst | 45623 | 11483 | -34140 | -74.83% | 74.83% |
| instcount.NumInsertValueInst | 4983 | 580 | -4403 | -88.36% | 88.36% |
| instcount.NumInvokeInst | 61018 | 59478 | -1540 | -2.52% | 2.52% |
| instcount.NumLandingPadInst | 35334 | 34215 | -1119 | -3.17% | 3.17% |
| instcount.NumPHIInst | 344428 | 331116 | -13312 | -3.86% | 3.86% |
| instcount.NumRetInst | 100773 | 100772 | -1 | 0.00% | 0.00% |
| instcount.TotalBlocks | 1081154 | 1077166 | -3988 | -0.37% | 0.37% |
| instcount.TotalFuncs | 101443 | 101442 | -1 | 0.00% | 0.00% |
| instcount.TotalInsts | 8890201 | 8833747 | -56454 | -0.64% | 0.64% |
| instsimplify.NumSimplified | 75822 | 75707 | -115 | -0.15% | 0.15% |
| simplifycfg.NumHoistCommonCode | 24203 | 24197 | -6 | -0.02% | 0.02% |
| simplifycfg.NumHoistCommonInstrs | 48201 | 48195 | -6 | -0.01% | 0.01% |
| simplifycfg.NumInvokes | 2785 | 4298 | 1513 | 54.33% | 54.33% |
| simplifycfg.NumSimpl | 997332 | 1018189 | 20857 | 2.09% | 2.09% |
| simplifycfg.NumSinkCommonCode | 7088 | 6464 | -624 | -8.80% | 8.80% |
| simplifycfg.NumSinkCommonInstrs | 15117 | 14021 | -1096 | -7.25% | 7.25% |
```
... which tells us that this new fold fires whopping 38k times,
increasing the amount of SimplifyCFG's `invoke`->`call` transforms by +54% (+1513) (again, D85787 did that last time),
decreasing total instruction count by -0.64% (-56454),
and sharply decreasing count of `insertvalue`'s (-88.36%, i.e. 9 times less)
and `extractvalue`'s (-74.83%, i.e. four times less).
This causes geomean -0.01% binary size decrease
http://llvm-compile-time-tracker.com/compare.php?from=
4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=
97dacca0111cb2ae678204e52a3cee00e3a69208&stat=size-text
and, ignoring `O0-g`, is a geomean -0.01%..-0.05% compile-time improvement
http://llvm-compile-time-tracker.com/compare.php?from=
4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=
97dacca0111cb2ae678204e52a3cee00e3a69208&stat=instructions
The other thing that tells is, is that while this is a massive win for `invoke`->`call` transform
`InstCombinerImpl::foldAggregateConstructionIntoAggregateReuse()` fold,
which is supposed to be dealing with such aggregate reconstructions,
fires a lot less now. There are two reasons why:
1. After this fold, as it can be seen in tests, we may (will) end up with trivially redundant PHI nodes.
We don't CSE them in InstCombine presently, which means that EarlyCSE needs to run and then InstCombine rerun.
2. But then, EarlyCSE not only manages to fold such redundant PHI's,
it also sees that the extract-insert chain recreates the original aggregate,
and replaces it with the original aggregate.
The take-aways are
1. We maybe should do most trivial, same-BB PHI CSE in InstCombine
2. I need to check if what other patterns remain, and how they can be resolved.
(i.e. i wonder if `foldAggregateConstructionIntoAggregateReuse()` might go away)
This is a reland of the original commit
fcb51d8c2460faa23b71e06abb7e826243887dd6,
because originally i forgot to ensure that the base aggregate types match.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D86530
Roman Lebedev [Wed, 26 Aug 2020 06:28:22 +0000 (09:28 +0300)]
[NFC][InstCombine] Add a PHI-of-insertvalues test with different base aggregate types
Mehdi Amini [Wed, 26 Aug 2020 05:43:43 +0000 (05:43 +0000)]
Add assertion in PatternRewriter::create<> to defend the same way as OpBuilder::create<> against missing dialect registration (NFC)
The code would have failed a few line later, but that way the error
message is more clear/friendly to debug.
Mehdi Amini [Wed, 26 Aug 2020 05:08:59 +0000 (05:08 +0000)]
Adjust assertion when casting to an unregistered operation
This assertion does not achieve what it meant to do originally, as it
would fire only when applied to an unregistered operation, which is a
fairly rare circumstance (it needs a dialect or context allowing
unregistered operation in the input in the first place).
Instead we relax it to only fire when it should have matched but didn't
because of the misconfiguration.
Differential Revision: https://reviews.llvm.org/D86588
Siva Chandra Reddy [Wed, 26 Aug 2020 06:32:55 +0000 (23:32 -0700)]
[libc][NFC] For remquo quotient, compare only 3 bits of MPFR and libc results.
Mateusz Mikuła [Wed, 26 Aug 2020 06:25:52 +0000 (09:25 +0300)]
[LLD][MinGW] Handle allow-multiple-definition flag
Basically copied from ELF driver.
Differential Revision: https://reviews.llvm.org/D86512
Mateusz Mikuła [Wed, 26 Aug 2020 06:24:37 +0000 (09:24 +0300)]
[LLD][MinGW] Cleanup Options.td file. NFC.
Based on ELF driver Options.td.
Differential Revision: https://reviews.llvm.org/D86509
Martin Storsjö [Thu, 20 Aug 2020 08:53:56 +0000 (11:53 +0300)]
[MC] [Win64EH] Update the AArch64/seh.s test slightly. NFC.
Update the comment stating the aim of the test - this is currently
only checking that these assembler directives doesn't cause the
assembler to fail, but the results of the testcase aren't particularly
correct yet.
Remove bits of the testcase that are even less likely to be found in
the wild (the .seh_startchained/.seh_endchained block), where the
testcase currently doesn't really generate anything interesting
anyway.
Differential Revision: https://reviews.llvm.org/D86524
Martin Storsjö [Mon, 24 Aug 2020 09:29:56 +0000 (12:29 +0300)]
[llvm-readobj] Fix arm64 unwind opcode disassembly printing
Add a missing minus, fix vertical alignment of instructions for one opcode.
Differential Revision: https://reviews.llvm.org/D86523
Thomas Raoux [Wed, 26 Aug 2020 06:19:09 +0000 (23:19 -0700)]
[mlir][spirv] Infer converted type of scf.for from the init value
Instead of using the TypeConverter infer the value of the alloca created based
on the init value. This will allow some ambiguous types like multidimensional
vectors to be converted correctly.
Differential Revision: https://reviews.llvm.org/D86582
Roman Lebedev [Wed, 26 Aug 2020 06:23:22 +0000 (09:23 +0300)]
Revert "[InstCombine] PHI-of-extractvalues -> extractvalue-of-PHI, aka invokes are bad"
This reverts commit
fcb51d8c2460faa23b71e06abb7e826243887dd6.
As buildbots report, there's apparently some missing check to ensure
that the types of incoming values match the type of PHI.
Let's revert for a moment.
Roman Lebedev [Wed, 26 Aug 2020 06:08:24 +0000 (09:08 +0300)]
[InstCombine] PHI-of-extractvalues -> extractvalue-of-PHI, aka invokes are bad
While since D86306 we do it's sibling fold for `insertvalue`,
we should also do this for `extractvalue`'s.
And unlike that one, the results here are, quite honestly, shocking,
as it can be observed here on vanilla llvm test-suite + RawSpeed results:
```
| statistic name | baseline | proposed | Δ | % | |%| |
|----------------------------------------------------|-----------|-----------|--------:|--------:|-------:|
| asm-printer.EmittedInsts | 7945095 | 7942507 | -2588 | -0.03% | 0.03% |
| assembler.ObjectBytes |
273209920 |
273069800 | -140120 | -0.05% | 0.05% |
| early-cse.NumCSE | 2183363 | 2183398 | 35 | 0.00% | 0.00% |
| early-cse.NumSimplify | 541847 | 550017 | 8170 | 1.51% | 1.51% |
| instcombine.NumAggregateReconstructionsSimplified | 2139 | 108 | -2031 | -94.95% | 94.95% |
| instcombine.NumCombined | 3601364 | 3635448 | 34084 | 0.95% | 0.95% |
| instcombine.NumConstProp | 27153 | 27157 | 4 | 0.01% | 0.01% |
| instcombine.NumDeadInst | 1694521 | 1765022 | 70501 | 4.16% | 4.16% |
| instcombine.NumPHIsOfExtractValues | 0 | 37546 | 37546 | 0.00% | 0.00% |
| instcombine.NumSunkInst | 63158 | 63686 | 528 | 0.84% | 0.84% |
| instcount.NumBrInst | 874304 | 871857 | -2447 | -0.28% | 0.28% |
| instcount.NumCallInst | 1757657 | 1758402 | 745 | 0.04% | 0.04% |
| instcount.NumExtractValueInst | 45623 | 11483 | -34140 | -74.83% | 74.83% |
| instcount.NumInsertValueInst | 4983 | 580 | -4403 | -88.36% | 88.36% |
| instcount.NumInvokeInst | 61018 | 59478 | -1540 | -2.52% | 2.52% |
| instcount.NumLandingPadInst | 35334 | 34215 | -1119 | -3.17% | 3.17% |
| instcount.NumPHIInst | 344428 | 331116 | -13312 | -3.86% | 3.86% |
| instcount.NumRetInst | 100773 | 100772 | -1 | 0.00% | 0.00% |
| instcount.TotalBlocks | 1081154 | 1077166 | -3988 | -0.37% | 0.37% |
| instcount.TotalFuncs | 101443 | 101442 | -1 | 0.00% | 0.00% |
| instcount.TotalInsts | 8890201 | 8833747 | -56454 | -0.64% | 0.64% |
| instsimplify.NumSimplified | 75822 | 75707 | -115 | -0.15% | 0.15% |
| simplifycfg.NumHoistCommonCode | 24203 | 24197 | -6 | -0.02% | 0.02% |
| simplifycfg.NumHoistCommonInstrs | 48201 | 48195 | -6 | -0.01% | 0.01% |
| simplifycfg.NumInvokes | 2785 | 4298 | 1513 | 54.33% | 54.33% |
| simplifycfg.NumSimpl | 997332 | 1018189 | 20857 | 2.09% | 2.09% |
| simplifycfg.NumSinkCommonCode | 7088 | 6464 | -624 | -8.80% | 8.80% |
| simplifycfg.NumSinkCommonInstrs | 15117 | 14021 | -1096 | -7.25% | 7.25% |
```
... which tells us that this new fold fires whopping 38k times,
increasing the amount of SimplifyCFG's `invoke`->`call` transforms by +54% (+1513) (again, D85787 did that last time),
decreasing total instruction count by -0.64% (-56454),
and sharply decreasing count of `insertvalue`'s (-88.36%, i.e. 9 times less)
and `extractvalue`'s (-74.83%, i.e. four times less).
This causes geomean -0.01% binary size decrease
http://llvm-compile-time-tracker.com/compare.php?from=
4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=
97dacca0111cb2ae678204e52a3cee00e3a69208&stat=size-text
and, ignoring `O0-g`, is a geomean -0.01%..-0.05% compile-time improvement
http://llvm-compile-time-tracker.com/compare.php?from=
4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=
97dacca0111cb2ae678204e52a3cee00e3a69208&stat=instructions
The other thing that tells is, is that while this is a massive win for `invoke`->`call` transform
`InstCombinerImpl::foldAggregateConstructionIntoAggregateReuse()` fold,
which is supposed to be dealing with such aggregate reconstructions,
fires a lot less now. There are two reasons why:
1. After this fold, as it can be seen in tests, we may (will) end up with trivially redundant PHI nodes.
We don't CSE them in InstCombine presently, which means that EarlyCSE needs to run and then InstCombine rerun.
2. But then, EarlyCSE not only manages to fold such redundant PHI's,
it also sees that the extract-insert chain recreates the original aggregate,
and replaces it with the original aggregate.
The take-aways are
1. We maybe should do most trivial, same-BB PHI CSE in InstCombine
2. I need to check if what other patterns remain, and how they can be resolved.
(i.e. i wonder if `foldAggregateConstructionIntoAggregateReuse()` might go away)
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D86530
Jianzhou Zhao [Tue, 25 Aug 2020 00:36:24 +0000 (00:36 +0000)]
Fix a 32-bit overflow issue when reading LTO-generated bitcode files whose strtab are of size > 2^29
This happens when using -flto and -Wl,--plugin-opt=emit-llvm to create a linked LTO bitcode file, and the bitcode file has a strtab with size > 2^29.
All the issues relate to a pattern like this
size_t x64 = y64 + z32 * C
When z32 is >= (2^32)/C, z32 * C overflows.
Reviewed-by: MaskRay
Differential Revision: https://reviews.llvm.org/D86500
Mehdi Amini [Wed, 26 Aug 2020 05:08:32 +0000 (05:08 +0000)]
Remove the use of global dialect registration from the standalone-translate.cpp example (NFC)
Siva Chandra Reddy [Wed, 26 Aug 2020 04:57:46 +0000 (21:57 -0700)]
[libc][obvious] Add back the accidentally removed MPFRNumber destructor.
Siva Chandra Reddy [Fri, 21 Aug 2020 05:36:53 +0000 (22:36 -0700)]
[libc] Extend MPFRMatcher to handle multiple-input-multiple-output functions.
Tests for frexp[f|l] now use the new capability. Not all input-output
combinations have been addressed by this change. Support for newer combinations
can be added in future as needed.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D86506
Xing GUO [Wed, 26 Aug 2020 04:30:33 +0000 (12:30 +0800)]
[DWARFYAML] Use writeDWARFOffset() to write the prologue_length field. NFC.
Use writeDWARFOffset() to simplify the logic. NFC.
Adrien Guinet [Wed, 26 Aug 2020 01:55:50 +0000 (18:55 -0700)]
[llvm-lipo] Add support for bitcode files
A Mach-O universal binary may contain bitcode as a slice.
This diff adds proper handling of such binaries to llvm-lipo.
Test plan: make check-all
Differential revision: https://reviews.llvm.org/D85740
Jason Molenda [Wed, 26 Aug 2020 04:03:39 +0000 (21:03 -0700)]
Ah, one test too many updated. This one should be unmodified.
Jason Molenda [Wed, 26 Aug 2020 03:53:01 +0000 (20:53 -0700)]
Update UnwindPlan dump to list if it is a trap handler func; also Command
Update the "image show-unwind" command output to show if the function
being shown is listed as a user-setting or platform trap handler.
Update the individual UnwindPlan dumps to show whether the unwind plan
is registered as a trap handler.
Teresa Johnson [Wed, 4 Mar 2020 23:38:45 +0000 (15:38 -0800)]
[Docs] Document --lto-whole-program-visibility
Summary:
Documents interaction of linker option added in D71913 with LTO
visibility.
Reviewers: pcc
Subscribers: inglorion, hiraditya, steven_wu, dexonsmith, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D75655
Mikhail R. Gadelha [Tue, 25 Aug 2020 23:19:58 +0000 (19:19 -0400)]
Add Z3 to system libraries list if enabled
Without this trying to link static LLVM libraries (built with Z3 enabled) fails because `llvm-config` doesn't print `-lz3`.
We are already using this patch at MSYS2: https://github.com/msys2/MINGW-packages/blob/master/mingw-w64-clang/0013-Add-Z3-to-system-libraries-list-if-enabled.patch
Reviewed By: mikhail.ramalho
Differential Revision: https://reviews.llvm.org/D85195
Craig Topper [Wed, 26 Aug 2020 01:20:41 +0000 (18:20 -0700)]
[X86] Add an isel pattern for (i8 (trunc (i16 (bitconvert (v16i1 X))))) to avoid an extra EXTRACT_SUBREG
Since we can only copy to GR32 we had to EXTRACT from GR32, but
we would first go to GR16 and then the truncate would extra again
to GR8. This adds a special case to go directly from GR32 to GR8.
This would eventually get cleaned up, but though maybe we should
avoid doing it in the first place. Our k-register handling is weird
and we could probably stand to have some more special ISD nodes
for the conversions so the i32 type would be explicit.
Volodymyr Sapsai [Thu, 23 Jul 2020 19:47:16 +0000 (12:47 -0700)]
[Modules] Improve error message when cannot find parent module for submodule definition.
Before the change the diagnostic for
module unknown.submodule {}
was "error: expected module name" which is incorrect and misleading
because both "unknown" and "submodule" are valid module names.
We already have a better error message when a parent module is a
submodule itself and is missing. Make the error for a missing top-level
module more like the one for a submodule.
rdar://problem/
64424407
Reviewed By: bruno
Differential Revision: https://reviews.llvm.org/D84458
Mehdi Amini [Tue, 25 Aug 2020 23:21:15 +0000 (23:21 +0000)]
Remove global registration from the test dialect in MLIR (NFC)
Craig Topper [Tue, 25 Aug 2020 23:00:12 +0000 (16:00 -0700)]
[X86] Remove extra getOperand(0) call from recently introduced store(extract_element(vtrunc)) to truncated store combine.
The IsExtractedElement already called getOperand(0) so Extract
here is the source vector. We shouldn't call getOperand(0). This
worked for the original test cases because the result was a
bitcast so the getOperand(0) accidently peeked through the bitcast
which is what we wanted.
In the failing case here, the operand turns out to be undef so
the getOperand(0) asserts because undef has no operands.
Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=25184
Differential Revision: https://reviews.llvm.org/D86428
Mehdi Amini [Tue, 25 Aug 2020 23:08:43 +0000 (23:08 +0000)]
Add llvm_unreachable after fully covered switch to silence some warnings from GCC (NFC)
Zequan Wu [Tue, 25 Aug 2020 22:04:33 +0000 (15:04 -0700)]
Revert "[Coverage] Enable emitting gap area between macros"
This reverts commit
a31c89c1b7a0a2fd3e2c0b8a587a60921abf4abd.
Craig Topper [Tue, 25 Aug 2020 22:16:50 +0000 (15:16 -0700)]
[X86] Remove a redundant COPY_TO_REGCLASS for VK16 after a KMOVWkr in an isel output pattern.
KMOVWkr produces VK16, there's no reason to copy it to VK16 again.
Test changes are presumably because we were scheduling based on
the COPY that is no longer there.
Shoaib Meenai [Tue, 25 Aug 2020 22:03:37 +0000 (15:03 -0700)]
[llvm-libtool-darwin] Address post-commit feedback
Address James Henderson's comments on https://reviews.llvm.org/D86359.
Dave Lee [Mon, 24 Aug 2020 20:46:23 +0000 (13:46 -0700)]
Remove unused/misnamed SetObjectModificationTime
Remove `SetObjectModificationTime` which is not currently used, and assigns to the wrong member.
Differential Revision: https://reviews.llvm.org/D86493
Mircea Trofin [Tue, 25 Aug 2020 16:58:49 +0000 (09:58 -0700)]
[MLInliner] Simplify TFUTILS_SUPPORTED_TYPES
We only need the C++ type and the corresponding TF Enum. The other
parameter was used for the output spec json file, but we can just
standardize on the C++ type name there.
Differential Revision: https://reviews.llvm.org/D86549
Stanislav Mekhanoshin [Tue, 25 Aug 2020 18:58:23 +0000 (11:58 -0700)]
[AMDGPU] Remove unsound dependency on ISA version in waitcnt
Differential Revision: https://reviews.llvm.org/D86566
Fangrui Song [Tue, 25 Aug 2020 20:37:29 +0000 (13:37 -0700)]
[TargetLoweringObjectFileImpl] Make .llvmbc and .llvmcmd non-SHF_ALLOC
There are two ways .llvmbc can be produced:
* clang -c -fembed-bitcode=all (which also produces .llvmcmd)
* LTO backend: ld.lld -mllvm -lto-embed-bitcode or -plugin-opt=-lto-embed-bitcode
.llvmbc and .llvmcmd have the SHF_ALLOC flag, so they can be dropped by
--gc-sections.
This patch sets SectionKind::Metadata to drop the SHF_ALLOC flag. This
is conceptually correct: the two sections are not part of the process
image, so SHF_ALLOC is not appropriate.
`test/LTO/X86/embed-bitcode.ll`: changed `llvm-objcopy -O binary --only-section` to
`llvm-objcopy --dump-section`. `-O binary` does not dump non-SHF_ALLOC sections.
Reviewed By: tejohnson
Differential Revision: https://reviews.llvm.org/D86374
Krzysztof Parzyszek [Mon, 24 Aug 2020 18:25:06 +0000 (13:25 -0500)]
[SDAG] Improve MemSDNode::getBasePtr
It returned getOperand(1), except for STORE for which it returned
getOperand(2). Handle MSTORE, MGATHER, and MSCATTER as well.
aartbik [Tue, 25 Aug 2020 19:26:28 +0000 (12:26 -0700)]
[mlir] [LLVMIR] Mark reductions as side-effect free
Attribute was missing from original base class.
Reviewed By: bkramer
Differential Revision: https://reviews.llvm.org/D86569
Shilei Tian [Tue, 25 Aug 2020 20:04:59 +0000 (16:04 -0400)]
[OpenMP] Pack first-private arguments to improve efficiency of data transfer
In this patch, we pack all small first-private arguments, allocate and transfer them all at once to reduce the number of data transfer which is very expensive.
Let's take the test case as example.
```
int main() {
int data1[3] = {1}, data2[3] = {2}, data3[3] = {3};
int sum[16] = {0};
#pragma omp target teams distribute parallel for map(tofrom: sum) firstprivate(data1, data2, data3)
for (int i = 0; i < 16; ++i) {
for (int j = 0; j < 3; ++j) {
sum[i] += data1[j];
sum[i] += data2[j];
sum[i] += data3[j];
}
}
}
```
Here `data1`, `data2`, and `data3` are three first-private arguments of the target region. In the previous `libomptarget`, it called data allocation and data transfer three times, each of which allocated and transferred 12 bytes. With this patch, it only calls allocation and transfer once. The size is `(12+4)*3=48` where 12 is the size of each array and 4 is the padding to keep the address aligned with 8. It is implemented in this way:
1. First collect all information for those *first*-private arguments. _private_ arguments are not the case because private arguments don't need to be mapped to target device. It just needs a data allocation. With the patch for memory manager, the data allocation could be very cheap, especially for the small size. For each qualified argument, push a place holder pointer `nullptr` to the `vector` for kernel arguments, and we will update them later.
2. After we have all information, create a buffer that can accommodate all arguments plus their paddings. Copy the arguments to the buffer at the right place, i.e. aligned address.
3. Allocate a target memory with the same size as the host buffer, transfer the host buffer to target device, and finally update all place holder pointers in the arguments `vector`.
The reason we only consider small arguments is, the data transfer is asynchronous. Therefore, for the large argument, we could continue to do things on the host side meanwhile, hopefully, the data is also being transferred. The "small" is defined by that the argument size is less than a predefined value. Currently it is 1024. I'm not sure whether it is a good one, and that is an open question. Another question is, do we need to make it configurable via an environment variable?
Reviewed By: ye-luo
Differential Revision: https://reviews.llvm.org/D86307
Stanislav Mekhanoshin [Tue, 25 Aug 2020 19:13:24 +0000 (12:13 -0700)]
[AMDGPU] Switch to named simm16 in vscnt insertion
Differential Revision: https://reviews.llvm.org/D86568
Ankit Aggarwal [Tue, 25 Aug 2020 19:47:44 +0000 (14:47 -0500)]
[Hexagon] Check if EVT is simple type in HVX lowering