platform/upstream/llvm.git
2 years ago[LegacyPM] Remove CGProfileLegacyPass
Fangrui Song [Sat, 16 Jul 2022 07:39:55 +0000 (00:39 -0700)]
[LegacyPM] Remove CGProfileLegacyPass

It's mostly a no-op after I removed legacy PGO passes in D123834.

2 years ago[SDAG] Add test for DAGCombiner multiple result commutative CSE
Itay Bookstein [Fri, 15 Jul 2022 20:30:57 +0000 (23:30 +0300)]
[SDAG] Add test for DAGCombiner multiple result commutative CSE

This commit adds a test for DAGCombiner commutative CSE on
nodes with multiple results (UMUL_LOHI). In this commit it
asserts the lack of CSE, a later commit will demonstrate
the CSE in the changed assertions.

Signed-off-by: Itay Bookstein <ibookstein@gmail.com>
Reviewed By: barannikov88

Differential Revision: https://reviews.llvm.org/D129905

2 years ago[mli][sparse] Remove manual dense dealloc from test case
Matthias Springer [Fri, 15 Jul 2022 06:17:50 +0000 (08:17 +0200)]
[mli][sparse] Remove manual dense dealloc from test case

Differential Revision: https://reviews.llvm.org/D129834

2 years ago[MLIR] Clean up checks for alloc-like ops in analysis
Uday Bondhugula [Sat, 16 Jul 2022 07:21:55 +0000 (12:51 +0530)]
[MLIR] Clean up checks for alloc-like ops in analysis

Clean up checks for alloc-like ops in analysis. Use the analysis
utility to properly check for the desired kind of effects. The previous
locality utility worked for all practical purposes but wasn't sound and
was locally duplicate code. Instead, use mlir::hasSingleEffect.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D129439

2 years ago[flang] Use *X instead of X.getValue() (NFC)
Kazu Hirata [Sat, 16 Jul 2022 07:24:02 +0000 (00:24 -0700)]
[flang] Use *X instead of X.getValue() (NFC)

Per Flang C++ Style Guide, this patch replaces X.getValue() with *X
where *X is protected by a presence test.

2 years ago[MLIR] NFC. Clean up logic of hasSingleEffect
Uday Bondhugula [Sat, 16 Jul 2022 07:09:24 +0000 (12:39 +0530)]
[MLIR] NFC. Clean up logic of hasSingleEffect

Clean up conditional logic of hasSingleEffect. NFC.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D129438

2 years agoUse drop_begin (NFC)
Kazu Hirata [Sat, 16 Jul 2022 06:58:11 +0000 (23:58 -0700)]
Use drop_begin (NFC)

2 years ago[flang][OpenMP] Fix warning due to uninitialized pointer dereference during atomic...
Nimish Mishra [Sat, 16 Jul 2022 07:20:19 +0000 (12:50 +0530)]
[flang][OpenMP] Fix warning due to uninitialized pointer dereference during atomic update lowering

Reviewed By: kiranchandramohan, kazu

Differential Revision: https://reviews.llvm.org/D129914

2 years agoExample:
Stella Laurenzo [Sat, 16 Jul 2022 04:03:04 +0000 (21:03 -0700)]
Example:

```
// -----// IR Dump Before LowerLinalgMicrokernels (iree-vmvx-lower-linalg-microkernels) //----- //
```

I've been meaning to suggest this for a long time, and I think the only reason we don't have it is because we didn't used to have the `getArgument()` handy when printing these comments. When debugging or putting a pipeline together based on such dumps, I often find myself grepping for the argument name of the pass (which is often related but not universally).

2 years agoUse drop_begin (NFC)
Kazu Hirata [Sat, 16 Jul 2022 06:24:59 +0000 (23:24 -0700)]
Use drop_begin (NFC)

2 years ago[MLIR][Liveness] Add `currentlyLiveValues`, a way to get a set of values that are...
bzcheeseman [Sun, 10 Jul 2022 15:27:45 +0000 (08:27 -0700)]
[MLIR][Liveness] Add `currentlyLiveValues`, a way to get a set of values that are live as of a given operation.

This change allows the user of LivenessBlockInfo to specify an op within the block and get a set of all values that are live as of that op. Semantically it relies on having a dominance-based region that has ordered operations. For DFG regions, computing liveness statically this way doesn't really make sense, it likely needs to be done at runtime.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D129447

2 years ago[sanitizer_common] Several Solaris procmaps fixes
Rainer Orth [Sat, 16 Jul 2022 04:59:00 +0000 (06:59 +0200)]
[sanitizer_common] Several Solaris procmaps fixes

Since the introduction of GoogleTest sharding in D122251
<https://reviews.llvm.org/D122251>, some of the Solaris sanitizer tests
have been running extremly long (up to an hour) while they took mere
seconds before.  Initial investigation suggests that massive lock
contention in Solaris procfs is involved here.

However, there's an easy way to somewhat reduce the impact: while the
current `ReadProcMaps` uses `ReadFileToBuffer` to read `/proc/self/xmap`,
that function primarily caters to Linux procfs reporting file sizes of 0
while the size on Solaris is accurate.  This patch makes use of that,
reducing the number of syscalls involved and reducing the runtime of
affected tests by a factor of 4.

Besides, it handles shared mappings and doesn't call `readlink` for unnamed
map entries.

Tested on `sparcv9-sun-solaris2.11` and `amd64-pc-solaris2.11`.

Differential Revision: https://reviews.llvm.org/D129837

2 years ago[lld-macho][nfc] Add more tests + comments around ICF + unwind info interaction
Jez Ng [Sat, 16 Jul 2022 03:38:48 +0000 (23:38 -0400)]
[lld-macho][nfc] Add more tests + comments around ICF + unwind info interaction

While working on {D129830}, I realized that our handling of ICF +
eh_frame combined was untested. Additionally I realized that the comment
explaining why we were safely slicing away the functionAddress reloc
from our compact unwind entries was... insufficient and slightly
misleading. I've tried to clarify it.

Reviewed By: #lld-macho, thevinster

Differential Revision: https://reviews.llvm.org/D129894

2 years agoUse has_value/value instead of hasValue/getValue (NFC)
Kazu Hirata [Sat, 16 Jul 2022 04:48:17 +0000 (21:48 -0700)]
Use has_value/value instead of hasValue/getValue (NFC)

2 years ago[BOLT][TEST] Add icp-inline.s test
Amir Ayupov [Sat, 16 Jul 2022 03:41:03 +0000 (20:41 -0700)]
[BOLT][TEST] Add icp-inline.s test

Add a test for `-icp-inline` knob, which ensures that ICP is only performed for
functions that can be subsequently inlined.

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D129803

2 years agoUse value instead of getValue (NFC)
Kazu Hirata [Sat, 16 Jul 2022 03:03:13 +0000 (20:03 -0700)]
Use value instead of getValue (NFC)

2 years agoUse value_or (NFC)
Kazu Hirata [Sat, 16 Jul 2022 02:46:29 +0000 (19:46 -0700)]
Use value_or (NFC)

2 years ago[LegacyPM] Remove ModuleSanitizerCoverageLegacyPass
Fangrui Song [Sat, 16 Jul 2022 02:01:20 +0000 (19:01 -0700)]
[LegacyPM] Remove ModuleSanitizerCoverageLegacyPass

Follow the steps of various other legacy instrumentation passes removed for
15.0.0.

2 years ago[mlir] remove unnecessary statements in test
Ashay Rane [Sat, 16 Jul 2022 00:53:06 +0000 (17:53 -0700)]
[mlir] remove unnecessary statements in test

A previous commit (f2b94bd) added some unnecessary statements that
dereferenced operations only to get the operations back.  This patch
removes the unnecessary statements.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D129913

2 years ago[X86][FP16] Enable vector support for FP16 emulation
Phoebe Wang [Sat, 16 Jul 2022 01:00:54 +0000 (09:00 +0800)]
[X86][FP16] Enable vector support for FP16 emulation

This is follow up of D107082, which enable vector support according to psABI.

Reviewed By: skan

Differential Revision: https://reviews.llvm.org/D127982

2 years ago[lit][unit] add a test for sanitizer-only test failures
Yuanfang Chen [Sat, 16 Jul 2022 01:21:35 +0000 (18:21 -0700)]
[lit][unit] add a test for sanitizer-only test failures

Follow-up for 6b02c53936b9e77fee.

2 years ago[mlir][ods] Allow specifying return types of builders
Jeff Niu [Thu, 14 Jul 2022 17:31:38 +0000 (10:31 -0700)]
[mlir][ods] Allow specifying return types of builders

This patch allows custom attribute and type builders to return
something other than the C++ type of the attribute or type.

This is useful for attributes or types that may perform extra work during
construction (e.g. canonicalization) that could result in a different
kind of attribute or type being returned.

Reviewed By: rriddle, lattner

Differential Revision: https://reviews.llvm.org/D129792

2 years ago[clang-format][NFC] Replace SmallVector parameter with ArrayRef
owenca [Sat, 16 Jul 2022 00:25:58 +0000 (17:25 -0700)]
[clang-format][NFC] Replace SmallVector parameter with ArrayRef

Addresses https://reviews.llvm.org/D129466#3654410.

2 years ago[gn build] Port 4162aefad125
LLVM GN Syncbot [Sat, 16 Jul 2022 00:15:08 +0000 (00:15 +0000)]
[gn build] Port 4162aefad125

2 years agoRevert "Re-apply 5acd47169884, Add a shared-memory based orc::MemoryMapper, with...
Mitch Phillips [Sat, 16 Jul 2022 00:11:55 +0000 (17:11 -0700)]
Revert "Re-apply 5acd47169884, Add a shared-memory based orc::MemoryMapper, with fixes."

This reverts commit 32d8d23cd0b2d4d010eb112dfe5216f11b2681f9.

Reason: Broke the UBSan buildbots. See more details on Phabricator:
https://reviews.llvm.org/D128544

2 years ago[test] Fix old memory leak in a test
Vitaly Buka [Fri, 15 Jul 2022 23:55:13 +0000 (16:55 -0700)]
[test] Fix old memory leak in a test

2 years ago[lit][test] explicitly use utf-8 encoding to write testing json file
Yuanfang Chen [Fri, 15 Jul 2022 23:40:43 +0000 (16:40 -0700)]
[lit][test] explicitly use utf-8 encoding to write testing json file

Related test failure:
https://lab.llvm.org/buildbot/#/builders/178/builds/2527/steps/13/logs/FAIL__lit___googletest-timeout_py

2 years ago[Sanitizer][Darwin][NFC] Small SANITIZER_APPLE cleanup
Julian Lettner [Fri, 15 Jul 2022 23:12:31 +0000 (16:12 -0700)]
[Sanitizer][Darwin][NFC] Small SANITIZER_APPLE cleanup

We introduced `SANITIZER_APPLE` here: https://reviews.llvm.org/D126263

2 years ago[unittests/Tooling/DependencyScannerTest.cpp] Use `using namespace` instead of wrappi...
Argyrios Kyrtzidis [Fri, 15 Jul 2022 23:08:48 +0000 (16:08 -0700)]
[unittests/Tooling/DependencyScannerTest.cpp] Use `using namespace` instead of wrapping the `.cpp` file contents in namespaces, NFC

This makes the file consistent with the coding style of the rest of LLVM.

2 years ago[flang] Add semantics test for EVENT POST statement
Naje George [Mon, 11 Jul 2022 20:17:04 +0000 (13:17 -0700)]
[flang] Add semantics test for EVENT POST statement

Reviewed By: ktras

Differential Revision: https://reviews.llvm.org/D129515

2 years ago[mlir][Linalg] Add a pattern to decompose `linalg.generic` ops.
Mahesh Ravishankar [Wed, 13 Jul 2022 23:22:47 +0000 (23:22 +0000)]
[mlir][Linalg] Add a pattern to decompose `linalg.generic` ops.

This patch adds a pattern to decompose a `linalg.generic` operations
that
- has only parallel iterator types
- has more than 2 statements (including the yield)

into multiple `linalg.generic` operation such that each operation has
a single statement and a yield.
The pattern added here just splits the matching `linalg.generic` into
two `linalg.generic`s, one containing the first statement, and the
other containing the remaining. The same pattern can be applied
repeatedly on the second op to ultimately fully decompose the generic
op.

Differential Revision: https://reviews.llvm.org/D129704

2 years ago[nfc][amdgpu] Remove dead variable and function
Jon Chesterfield [Fri, 15 Jul 2022 22:55:42 +0000 (23:55 +0100)]
[nfc][amdgpu] Remove dead variable and function

2 years ago[lit][unit] make sure to include failed shard in test summary report
Yuanfang Chen [Fri, 15 Jul 2022 22:33:57 +0000 (15:33 -0700)]
[lit][unit] make sure to include failed shard in test summary report

For sanitizer only failures, the tests all pass but still need to
put the shard in failed test summary.

2 years ago[PGO] Report number of counts being dropped when a hash-mismatch happens
Rong Xu [Fri, 15 Jul 2022 21:51:49 +0000 (14:51 -0700)]
[PGO] Report number of counts being dropped when a hash-mismatch happens

This patch reports number of counts being dropped when a hash-mismatch
happens. This information will be helpful to the users -- if the dropped
counts are large, the user should redo the instrumentation build and
recollect the profile.

Differential Revision: https://reviews.llvm.org/D129001

2 years ago[mlir] check whether region and block visitors are interrupted
Ashay Rane [Thu, 14 Jul 2022 01:20:56 +0000 (18:20 -0700)]
[mlir] check whether region and block visitors are interrupted

The visitor functions for `Region` and `Block` types did not always
check the value returned by recursive calls.  This caused the top-level
visitor invocation to return `WalkResult::advance()` even if one or more
recursive invocations returned `WalkResult::interrupt()`.  This patch
fixes the problem by check if any recursive call is interrupted, and if
so, return `WalkResult::interrupt()`.

Reviewed By: dcaballe

Differential Revision: https://reviews.llvm.org/D129718

2 years ago[CUDA] Make the new driver properly ignore non-CUDA inputs
Joseph Huber [Fri, 15 Jul 2022 17:52:27 +0000 (13:52 -0400)]
[CUDA] Make the new driver properly ignore non-CUDA inputs

The new driver generated offloadinga actions for each active toolchain.
However, for CUDA and HIP it is possible for the toolchain to be active
but one of the files is not a valid input. This can occur if the user
compiles both a CUDA and C source file in the same compiler invocation.
This patch adds some simple logic to quit if the input is not valid as
well.

Reviewed By: tra, MaskRay

Differential Revision: https://reviews.llvm.org/D129885

2 years ago[lit] fix a bug in 4cd1c96d375aa
Yuanfang Chen [Fri, 15 Jul 2022 21:35:53 +0000 (14:35 -0700)]
[lit] fix a bug in 4cd1c96d375aa

Only report failure for tests that actually runs.

2 years ago[lsan][Darwin][nfc] Fix thread vector size
Leonard Grey [Fri, 15 Jul 2022 19:36:01 +0000 (15:36 -0400)]
[lsan][Darwin][nfc] Fix thread vector size

The reserve constructor was removed in 44f55509d75d8c67077810bb6d9f3bedaea05831
but this one was missed. As a result, we attempt to iterate through 1024 threads
each time, most of which are 0.

Differential Revision: https://reviews.llvm.org/D129897

2 years ago[bolt] Include <atomic>
Fangrui Song [Fri, 15 Jul 2022 21:27:01 +0000 (14:27 -0700)]
[bolt] Include <atomic>

2 years ago[Libomptarget] Fix warnings on address space attributes
Joseph Huber [Fri, 15 Jul 2022 19:34:50 +0000 (15:34 -0400)]
[Libomptarget] Fix warnings on address space attributes

The device runtime uses the address space attribute to control the
placement of important constants on the GPU. The changes made in D126061
caused these to start emitting errors as they were not applied to the
type. This patch fixes the issues to make the warnings go away.

Reviewed By: ye-luo

Differential Revision: https://reviews.llvm.org/D129896

2 years ago[PGO] Don't cross reference CSFDO profile and non-CSFDO profile
Rong Xu [Fri, 15 Jul 2022 20:57:23 +0000 (13:57 -0700)]
[PGO] Don't cross reference CSFDO profile and non-CSFDO profile

Don't cross reference CSFDO profile and non-CSFDO profile when
checking the function hash. Only return hash_mismatch when
CS bits match, and return unknown_function otherwise.

Differential Revision: https://reviews.llvm.org/D129000

2 years ago[PGO] Improve hash-mismatch warning message
Rong Xu [Fri, 15 Jul 2022 20:44:55 +0000 (13:44 -0700)]
[PGO] Improve hash-mismatch warning message

This patch improves FDO hash-mismatch handling:
(1) filter out warnings to weak functions.
Weak functions definition will be overridden by a strong definition by linker.
The hash mismatch in profile use compilation is expected.
Make the profile hash mismatch warning under the existing option (default true).

(2) add an option to trace the hash of functions with the specific string.
Note that an empty string parameter will trace all functions.

Differential Revision: https://reviews.llvm.org/D129002

2 years ago[mlir][sparse] Introduce new reduce op
Jim Kitchen [Fri, 15 Jul 2022 20:26:41 +0000 (15:26 -0500)]
[mlir][sparse] Introduce new reduce op

A new sparse_tensor operation allows for
custom reduction code to be injected during
linalg.generic lowering for sparse tensors.
An identity value is provided to indicate
the starting value of the reduction. A single
block region is required to contain the
custom reduce computation.

Reviewed by: aartbik

Differential Revision: https://reviews.llvm.org/D128004

2 years ago[LSR] Allow already invariant operand for ICmpZero matching [try 2]
Philip Reames [Fri, 15 Jul 2022 20:23:10 +0000 (13:23 -0700)]
[LSR] Allow already invariant operand for ICmpZero matching [try 2]

Changes since initial commit:

* Wrapping a pointer in an SCEV unknown hides the base, and SCEV is only able to compute a subtraction when the bases are known to be equal. This results in a SCEVCouldNotCompute flowing forward and triggering asserts. Test case added in d767b392.
* isLoopInvariant returns true for instructions outside the loop, but not necessarily *above* the loop. Since this code is allowed to visit uses of an IV outside of a loop, we have to make sure the operands of the compare are both invariant and dominating the header. Test case added in 2aed3cdb.

Original commit message follows...

The ICmpZero matching is checking to see if the expression is loop invariant per SCEV and expandable. This allows expressions inside the loop which can be made loop invariant to be seamlessly expanded, but is overly conservative for expressions which already *are* loop invariant.

As a simple justification for why this is correct, consider a loop invariant urem as RHS vs an alternate function with that same urem wrapped inside a helper call. Why would it be legal to match the later, but not the former?

Differential Revision: https://reviews.llvm.org/D129793

2 years ago[reland][lit][unit] makes sure to print stderr when no test failures are found
Yuanfang Chen [Fri, 15 Jul 2022 20:23:22 +0000 (13:23 -0700)]
[reland][lit][unit] makes sure to print stderr when no test failures are found

In some situations, like running tests with sanitizers, all test
passes but the shard could still fail due to memory issues.

Relands 65769429c0644c81d271e9d971f5b64191503

2 years agoRevert "[lit][unit] makes sure to print stderr when no test failures are found"
Yuanfang Chen [Fri, 15 Jul 2022 20:21:01 +0000 (13:21 -0700)]
Revert "[lit][unit] makes sure to print stderr when no test failures are found"

This reverts commit 65769429c0644c81d271e9d971f5b64191503f6e.

This patch is incomplete.

2 years ago[lldb] [llgs] Improve stdio forwarding in multiprocess+nonstop
Michał Górny [Thu, 30 Jun 2022 18:24:04 +0000 (20:24 +0200)]
[lldb] [llgs] Improve stdio forwarding in multiprocess+nonstop

Enable stdio forwarding when nonstop mode is enabled, and disable it
once it is disabled.  This makes it possible to cleanly handle stdio
forwarding while running multiple processes in non-stop mode.

Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.llvm.org/D128932

2 years ago[lldb] [test] Skip test_leave_nonstop on Windows
Michał Górny [Fri, 15 Jul 2022 19:47:37 +0000 (21:47 +0200)]
[lldb] [test] Skip test_leave_nonstop on Windows

2 years ago[mlir][Linalg] Extend Generic op dedup canonicalization to drop redundant result...
Mahesh Ravishankar [Fri, 15 Jul 2022 04:29:08 +0000 (04:29 +0000)]
[mlir][Linalg] Extend Generic op dedup canonicalization to drop redundant result values.

2 years ago[mlir][Linalg] NFC: Refactor canonicalization for deduping generic op operands.
Mahesh Ravishankar [Fri, 15 Jul 2022 03:56:48 +0000 (03:56 +0000)]
[mlir][Linalg] NFC: Refactor canonicalization for deduping generic op operands.

This is a NFC change to make it easier to update this canonicalization
for more use cases. The refactoring makes things easier to
understand/adapt.

Differential Revision: https://reviews.llvm.org/D129829

2 years ago[lsan][nfc] Add missing newline to VReport
Leonard Grey [Fri, 15 Jul 2022 19:44:23 +0000 (15:44 -0400)]
[lsan][nfc] Add missing newline to VReport

2 years ago[Flang] Set constructExit for Where and Forall constructs
Jonathon Penix [Fri, 15 Jul 2022 19:32:36 +0000 (12:32 -0700)]
[Flang] Set constructExit for Where and Forall constructs

Evaluations for the Where and Forall constructs previously did
not have their constructExit field fixed up. This could lead to
falling through to subsequent case blocks in select case
statements if either a Where or Forall construct was the final part
of one case block. Setting the constructExit field results in the
proper branching behavior.

Fixes issue: https://github.com/llvm/llvm-project/issues/56500

Differential Revision: https://reviews.llvm.org/D129879

Change-Id: Ia868df12084520a935f087524e118bcdf47f6d7a

2 years ago[lit][unit] makes sure to print stderr when no test failures are found
Yuanfang Chen [Fri, 15 Jul 2022 19:16:18 +0000 (12:16 -0700)]
[lit][unit] makes sure to print stderr when no test failures are found

In some situations, like running tests with sanitizers, all test passes
but the shard could still fail due to memory issues.

2 years ago[lld-macho][nfc] Fix numeric substitutions in icf.s test
Jez Ng [Fri, 15 Jul 2022 19:28:14 +0000 (15:28 -0400)]
[lld-macho][nfc] Fix numeric substitutions in icf.s test

We were re-defining the various numeric variables when we actually
intended to check already-defined variables against the value on the
current CHECK line.

Reviewed By: #lld-macho, thakis

Differential Revision: https://reviews.llvm.org/D129831

2 years ago[test] Reduced test for second distinct issue triggering revert of 9153515
Philip Reames [Fri, 15 Jul 2022 19:08:24 +0000 (12:08 -0700)]
[test] Reduced test for second distinct issue triggering revert of 9153515

2 years ago[mlir][sparse] Add 3-dimensional sparse tensor multiplication integration test
Rajas Vanjape [Fri, 15 Jul 2022 18:09:17 +0000 (11:09 -0700)]
[mlir][sparse] Add 3-dimensional sparse tensor multiplication integration test

This diff adds an integration test which does element wise multiplication for two sparse 3-d tensors of size 3x3x5

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D129638

2 years ago[syntax] Some #includes cleanup, NFC.
Haojian Wu [Fri, 15 Jul 2022 18:49:12 +0000 (20:49 +0200)]
[syntax] Some #includes cleanup, NFC.

2 years ago[llvm-cov] Add error message for missing profdata on report and export subcommands.
Zequan Wu [Thu, 14 Jul 2022 18:29:10 +0000 (11:29 -0700)]
[llvm-cov] Add error message for missing profdata on report and export subcommands.

When profdata is missing on report and export commands, the error message is
indistinguishable from missing instrumented binary file. This adds the error
message for report and export commands.

Differential Revision: https://reviews.llvm.org/D129791

2 years ago[Reassociate] Enable FP reassociation via 'reassoc' and 'nsz'
Warren Ristow [Fri, 15 Jul 2022 18:44:35 +0000 (11:44 -0700)]
[Reassociate] Enable FP reassociation via 'reassoc' and 'nsz'

Compiling with '-ffast-math' tuns on all the FastMathFlags (FMF), as
expected, and that enables FP reassociation. Only the two FMF flags
'reassoc' and 'nsz' are technically required to perform reassociation,
but disabling other unrelated FMF bits is needlessly suppressing the
optimization.

This patch fixes that needless suppression, and makes appropriate
adjustments to test-cases, fixing some outstanding TODOs in the process.

Fixes: #56483

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D129523

2 years ago[test] Reduced test which triggered revert of 9153515
Philip Reames [Fri, 15 Jul 2022 18:31:29 +0000 (11:31 -0700)]
[test] Reduced test which triggered revert of 9153515

2 years ago[lldb] [llgs] Fix disabling non-stop mode
Michał Górny [Thu, 30 Jun 2022 08:47:15 +0000 (10:47 +0200)]
[lldb] [llgs] Fix disabling non-stop mode

Stop all processes and clear notification queues when disabling non-stop
mode.  Ensure that no stop notifications are sent for processes stopped
due to the mode switch.

Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.llvm.org/D128893

2 years ago[lldb] [test] Skip test_stop_reason_while_running on Windows
Michał Górny [Fri, 15 Jul 2022 18:14:26 +0000 (20:14 +0200)]
[lldb] [test] Skip test_stop_reason_while_running on Windows

2 years ago[libc] add rounding modes to printf float conv
Michael Jones [Wed, 13 Jul 2022 18:47:03 +0000 (11:47 -0700)]
[libc] add rounding modes to printf float conv

This adds functionality for rounding towards negative inf, positive inf,
and zero to the float hex conversion (%a).

Reviewed By: lntue

Differential Revision: https://reviews.llvm.org/D129702

2 years ago[AMDGPU] Add remarks to output some resource usage
Vang Thao [Fri, 11 Mar 2022 00:46:50 +0000 (00:46 +0000)]
[AMDGPU] Add remarks to output some resource usage

Add analyis remarks to output kernel name, register usage, occupancy,
scratch usage, spills, and LDS information.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D123878

2 years ago[mlir][linalg] Fix tensor tiling together with interchange
Lei Zhang [Fri, 15 Jul 2022 17:43:57 +0000 (13:43 -0400)]
[mlir][linalg] Fix tensor tiling together with interchange

In `linalg::tileConsumerAndFuseProducers`, there are two levels of
tiling and fusion; we partition the tile sizes and only use one
half for each of them. The partition is using the first non-parallel
dimension *after* interchange as the boundary. However, concrete
tiling happens *together with* loop interchange, so we still need
to provide the partial tile sizes *before* the interchange.
Otherwise, there will be inconsistency, which is what this patch
is to fix.

Reviewed By: ThomasRaoux

Differential Revision: https://reviews.llvm.org/D129804

2 years agoRevert "[LSR] Allow already invariant operand for ICmpZero matching"
Philip Reames [Fri, 15 Jul 2022 17:47:32 +0000 (10:47 -0700)]
Revert "[LSR] Allow already invariant operand for ICmpZero matching"

This reverts commit 9153515a7bea9fb9dd4c76f70053a170bf825f35.  Builtbot crash was reported in the commit thread, reverting while investigating.

2 years ago[mlir] Enable perfect forwarding in RewritePatternSet::add
Laszlo Kindrat [Fri, 15 Jul 2022 17:43:41 +0000 (19:43 +0200)]
[mlir] Enable perfect forwarding in RewritePatternSet::add

This patch modifies the implementation of `RewritePatternSet::add` to perfectly forward its arguments to pattern constructors. Without this, code like the following compiles but, due to the limited lifetime of the temporary TypeConverter, can produce unexpected behavior:
```
RewritePatternSet patterns(context);
patterns.add<SomeOpConversion, OtherOpConversion>(TypeConverter(), context);

if (failed(applyPartialConversion(getOperation(), target, std::move(patterns))))
  return signalPassFailure();
```

The patch also changes the linalg fusion pattern implementation to correctly fill the test pattern set given the new behavior.

Author: Laszlo Kindrat <laszlokindrat@gmail.com>

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D129601

2 years ago[sanitizer] Add "mainfile" prefix to sanitizer special case list
Fangrui Song [Fri, 15 Jul 2022 17:39:26 +0000 (10:39 -0700)]
[sanitizer] Add "mainfile" prefix to sanitizer special case list

When an issue exists in the main file (caller) instead of an included file
(callee), using a `src` pattern applying to the included file may be
inappropriate if it's the caller's responsibility. Add `mainfile` prefix to check
the main filename.

For the example below, the issue may reside in a.c (foo should not be called
with a misaligned pointer or foo should switch to an unaligned load), but with
`src` we can only apply to the innocent callee a.h. With this patch we can use
the more appropriate `mainfile:a.c`.
```
//--- a.h
// internal linkage
static inline int load(int *x) { return *x; }

//--- a.c, -fsanitize=alignment
#include "a.h"
int foo(void *x) { return load(x); }
```

See the updated clang/docs/SanitizerSpecialCaseList.rst for a caveat due
to C++ vague linkage functions.

Reviewed By: #sanitizers, kstoimenov, vitalybuka

Differential Revision: https://reviews.llvm.org/D129832

2 years ago[LV] Use umax(VF * UF, MinProfTC) for scalable vectors.
Florian Hahn [Fri, 15 Jul 2022 17:07:39 +0000 (10:07 -0700)]
[LV] Use umax(VF * UF, MinProfTC) for scalable vectors.

For scalable vectors, it is not sufficient to only check
MinProfitableTripCount if it is >= VF.getKnownMinValue() * UF, because
this property may not holder for larger values of vscale. In those
cases, compute umax(VF * UF, MinProfTC) instead.

This should fix
https://lab.llvm.org/buildbot/#/builders/197/builds/2262

2 years ago[Libomptarget][NFC] Fix signed comparison warnings
Joseph Huber [Fri, 15 Jul 2022 17:22:00 +0000 (13:22 -0400)]
[Libomptarget][NFC] Fix signed comparison warnings

Summary:
Non-functional change, just fixing some sign comparison warnings by
making both match.

2 years agoRevert "[mlir] Enable perfect forwarding in RewritePatternSet::add"
lorenzo chelini [Fri, 15 Jul 2022 17:16:09 +0000 (19:16 +0200)]
Revert "[mlir] Enable perfect forwarding in RewritePatternSet::add"

Did not preserve author information.

This reverts commit b0afda78f007740371307bfacbe4a486a4b77a3e.

2 years agogithub: Add a helpful message for issues without milestones
Tom Stellard [Fri, 15 Jul 2022 17:10:27 +0000 (10:10 -0700)]
github: Add a helpful message for issues without milestones

The /branch and /cherry-pick commands only work when an isssue has
a milestone, so give the user a helpful error message when they
try these commands on issue without a milestone.

Reviewed By: thieta, kwk

Differential Revision: https://reviews.llvm.org/D125744

2 years ago[libc][arm32] add string stdlib & math entrypoints
Michael Jones [Fri, 15 Jul 2022 00:02:43 +0000 (17:02 -0700)]
[libc][arm32] add string stdlib & math entrypoints

This patch adds all the string and stdlib entrypoints, as well as a few
math entrypoints to the arm32 build.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D129821

2 years ago[mlir] Enable perfect forwarding in RewritePatternSet::add
lorenzo chelini [Fri, 15 Jul 2022 17:06:10 +0000 (19:06 +0200)]
[mlir] Enable perfect forwarding in RewritePatternSet::add

This patch modifies the implementation of `RewritePatternSet::add` to perfectly forward its arguments to pattern constructors. Without this, code like the following compiles but, due to the limited lifetime of the temporary TypeConverter, can produce unexpected behavior:
```
RewritePatternSet patterns(context);
patterns.add<SomeOpConversion, OtherOpConversion>(TypeConverter(), context);

if (failed(applyPartialConversion(getOperation(), target, std::move(patterns))))
  return signalPassFailure();
```

The patch also changes the linalg fusion pattern implementation to correctly fill the test pattern set given the new behavior.

Author: Laszlo Kindrat <laszlokindrat@gmail.com>

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D129601

2 years ago[ELF][RISCV] Relax local-exec TLS model
Fangrui Song [Fri, 15 Jul 2022 17:08:08 +0000 (10:08 -0700)]
[ELF][RISCV] Relax local-exec TLS model

In -mrelax mode, GCC/Clang may generate a local-exec TLS code sequence like:
```
# R_RISCV_TPREL_HI20, R_RISCV_RELAX
lui rd, %tprel_hi(x)
# R_RISCV_TPREL_ADD, R_RISCV_RELAX
add rd, rd, tp, %tprel_add(x)
# (R_RISCV_TPREL_LO12_I || R_RISCV_TPREL_LO12_S), R_RISCV_RELAX
addi rd, rd, %tprel_lo(x) || sw rs, %tprel(x)(rd)
```

Note: st_value(x) for TLS should be in the range [0,p_memsz(PT_TLS)).
When st_value(x) < 2048 (i.e. hi20(x) == 0), the linker can relax the code
sequence to:
```
addi rd, tp, st_value(x) || sw rs, st_value(x)(rd)
```

Differential Revision: https://reviews.llvm.org/D129425

2 years ago[analyzer] Add new function `clang_analyzer_value` to ExprInspectionChecker
Denys Petrov [Sat, 9 Jul 2022 18:21:10 +0000 (21:21 +0300)]
[analyzer] Add new function `clang_analyzer_value` to ExprInspectionChecker

Summary: Introduce a new function 'clang_analyzer_value'. It emits a report that in turn prints a RangeSet or APSInt associated with SVal. If there is no associated value, prints "n/a".

2 years ago[LV] Add scalable vector test showing incorrect min-trip count check.
Florian Hahn [Fri, 15 Jul 2022 16:56:52 +0000 (09:56 -0700)]
[LV] Add scalable vector test showing incorrect min-trip count check.

The test shows a case where the minimum trip count check incorrectly
only checks the minimum profitable trip count computed due to runtime
checks. This is incorrect for scalable VFs, because the VF * UF may
exceed the minimum profitable trip count for vscale > 1.

This is the likely reason for
https://lab.llvm.org/buildbot/#/builders/197/builds/2262 failing.

2 years ago[LSR] Allow already invariant operand for ICmpZero matching
Philip Reames [Fri, 15 Jul 2022 16:31:56 +0000 (09:31 -0700)]
[LSR] Allow already invariant operand for ICmpZero matching

The ICmpZero matching is checking to see if the expression is loop invariant per SCEV and expandable. This allows expressions inside the loop which can be made loop invariant to be seamlessly expanded, but is overly conservative for expressions which already *are* loop invariant.

As a simple justification for why this is correct, consider a loop invariant urem as RHS vs an alternate function with that same urem wrapped inside a helper call. Why would it be legal to match the later, but not the former?

Differential Revision: https://reviews.llvm.org/D129793

2 years ago[gn build] Port 32d8d23cd0b2
LLVM GN Syncbot [Fri, 15 Jul 2022 16:46:00 +0000 (16:46 +0000)]
[gn build] Port 32d8d23cd0b2

2 years ago[X86] test-vs-bittest.ll - add test coverage for (and (srl (not x), c), 1) patterns
Simon Pilgrim [Fri, 15 Jul 2022 15:41:08 +0000 (16:41 +0100)]
[X86] test-vs-bittest.ll - add test coverage for (and (srl (not x), c), 1) patterns

These can appear as well as the more common (and (not (srl x, c)), 1) patterns

2 years agoRe-apply 5acd47169884, Add a shared-memory based orc::MemoryMapper, with fixes.
Anubhab Ghosh [Fri, 15 Jul 2022 16:34:53 +0000 (09:34 -0700)]
Re-apply 5acd47169884, Add a shared-memory based orc::MemoryMapper, with fixes.

The original commit was reverted in 3e9cc543f223 due to buildbot failures, which
should be fixed by the addition of dependencies on librt.

Differential Revision: https://reviews.llvm.org/D128544

2 years ago[lldb] [llgs] Fix `?` packet response for running threads
Michał Górny [Thu, 30 Jun 2022 05:58:12 +0000 (07:58 +0200)]
[lldb] [llgs] Fix `?` packet response for running threads

Fix the response to `?` packet for threads that are running at the time
(in non-stop mode).  The previous code would wrongly send or queue
an empty response for them.

Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.llvm.org/D128879

2 years ago[lldb] [test] Skip TestNonStop → test_stdio on Windows
Michał Górny [Fri, 15 Jul 2022 16:32:59 +0000 (18:32 +0200)]
[lldb] [test] Skip TestNonStop → test_stdio on Windows

2 years ago[MLIR][Presburger] reintroduce int64_t versions of floorDiv, ceilDiv in mlir::presbur...
Arjun P [Fri, 15 Jul 2022 16:27:32 +0000 (17:27 +0100)]
[MLIR][Presburger] reintroduce int64_t versions of floorDiv, ceilDiv in mlir::presburger namespace

This is useful because MPInt.h defines identically-named functions that
operate on MPInts, which would otherwie become the only candidates of
overload resolution when calling e.g. ceilDiv from the mlir::presburger
namespace (iff MPInt.h is included). So to access the 64-bit overloads, an
explict call to mlir::ceilDiv would be required. This patch adds `using`
declarations allowing overload resolution to transparently call the right
function.

Reviewed By: Groverkss

Differential Revision: https://reviews.llvm.org/D129820

2 years ago[MLIR][Presburger] MPInt: add missing fastpath in ceilDiv
Arjun P [Fri, 15 Jul 2022 16:25:40 +0000 (17:25 +0100)]
[MLIR][Presburger] MPInt: add missing fastpath in ceilDiv

This is not a bug in functionality, just a missed optimization.

Reviewed By: Groverkss

Differential Revision: https://reviews.llvm.org/D129815

2 years ago[analyzer][NFC] Tidy up handler-functions in SymbolicRangeInferrer
Denys Petrov [Tue, 12 Jul 2022 20:18:37 +0000 (23:18 +0300)]
[analyzer][NFC] Tidy up handler-functions in SymbolicRangeInferrer

Summary: Sorted some handler-functions into more appropriate visitor functions of the SymbolicRangeInferrer.
- Spread `getRangeForNegatedSub` body over several visitor functions: `VisitSymExpr`, `VisitSymIntExpr`, `VisitSymSymExpr`.
- Moved `getRangeForComparisonSymbol` from `infer` to `VisitSymSymExpr`.

Differential Revision: https://reviews.llvm.org/D129678

2 years ago[mlir][math] Added math::tan operation.
Slava Zakharin [Tue, 12 Jul 2022 04:01:04 +0000 (21:01 -0700)]
[mlir][math] Added math::tan operation.

Differential Revision: https://reviews.llvm.org/D129539

2 years ago[Clang] Add a new clang option "-ftime-trace=<value>"
dongjunduo [Fri, 17 Jun 2022 11:10:02 +0000 (11:10 +0000)]
[Clang] Add a new clang option "-ftime-trace=<value>"

The time profiler traces the stages during the clang compile
process. Each compiling stage of a single source file
corresponds to a separately .json file which holds its
time tracing data. However, the .json files are stored in the
same path/directory as its corresponding stage's '-o' option.
For example, if we compile the "demo.cc" to "demo.o" with option
"-o /tmp/demo.o", the time trace data file path is "/tmp/demo.json".

A typical c++ project can contain multiple source files in different
path, but all the json files' paths can be a mess.

The option "-ftime-trace=<value>" allows you to specify where the json
files should be stored. This allows the users to place the time trace
data files of interest in the desired location for further data analysis.

Usage:
    - clang/clang++ -ftime-trace ...
    - clang/clang++ -ftime-trace=the-directory-you-want ...
    - clang/clang++ -ftime-trace=the-directory-you-want/ ...
    - clang/clang++ -ftime-trace=the-full-file-path-you-want ...

Differential Revision: https://reviews.llvm.org/D128048

2 years ago[NFC][SCEV] Rename variable to correspond its current meaning
Max Kazantsev [Fri, 15 Jul 2022 15:33:32 +0000 (22:33 +0700)]
[NFC][SCEV] Rename variable to correspond its current meaning

2 years ago[lldb] [llgs] Send process output asynchronously in non-stop mode
Michał Górny [Wed, 29 Jun 2022 19:48:10 +0000 (21:48 +0200)]
[lldb] [llgs] Send process output asynchronously in non-stop mode

Introduce a new %Stdio notification category and use it to send process
output asynchronously when running in non-stop mode.  This is an LLDB
extension since GDB does not use the 'O' packet for process output,
just for replies to 'qRcmd' packets.

Using the async notification mechanism implies that only the first
output packet is sent immediately to the client.  The client needs
to request subsequent notifications (if any) using the new vStdio packet
(that works pretty much like vStopped for the Stop notification queue).

The packet handler in lldb-server tests is updated to handle the async
stdio packets in addition to the regular O packets.  However, due
to the implications noted above, it can only handle the first output
packet sent by the server.  Subsequent packets need to be explicitly
requested via vStdio.

Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.llvm.org/D128849

2 years ago[mlir][mbr] Improve diagnostics on error with `raise from`.
Ingo Müller [Thu, 14 Jul 2022 12:17:29 +0000 (12:17 +0000)]
[mlir][mbr] Improve diagnostics on error with `raise from`.

This commit extends the `raise` statements on errors in user-provided
code with `from e` clauses that attach the original exception to the one
being raised. This allows to debug the root cause of the error more
easily.

Reviewed By: SaurabhJha

Differential Revision: https://reviews.llvm.org/D129762

2 years agoRe-run update_test_checks for some tests
Nicolai Hähnle [Fri, 15 Jul 2022 15:10:12 +0000 (17:10 +0200)]
Re-run update_test_checks for some tests

Motivation is to simplify the test diff for an upcoming real change.

2 years ago[SCEV] List all binops in getOperandsToCreate()
Nikita Popov [Fri, 15 Jul 2022 15:06:07 +0000 (17:06 +0200)]
[SCEV] List all binops in getOperandsToCreate()

Explicitly list all binops rather than having a default case. There
were two bugs here:
1. U->getOpcode() was used instead of BO->Opcode, which means we
   used the logic for the wrong opcode in some cases.
2. SCEV construction does not support LShr. We should return
   unknown for it rather than recursing into the operands.

2 years agoClarify the behavior of the llvm.vector.insert/extract intrinsics when the index
David Kreitzer [Wed, 6 Jul 2022 18:15:16 +0000 (11:15 -0700)]
Clarify the behavior of the llvm.vector.insert/extract intrinsics when the index
is out of range. Both intrinsics return a poison value.

Consequently, mark the intrinsics speculatable.
Differential Revision: https://reviews.llvm.org/D129656

2 years ago[MLIR][Presburger] SlowMPInt: gcd: assert that operands are non-negative
Arjun P [Fri, 15 Jul 2022 14:32:15 +0000 (15:32 +0100)]
[MLIR][Presburger] SlowMPInt: gcd: assert that operands are non-negative

2 years ago[docs] Note about how to handle 'llvm-mt: error: no libxml2'
Hans Wennborg [Thu, 14 Jul 2022 13:31:53 +0000 (15:31 +0200)]
[docs] Note about how to handle 'llvm-mt: error: no libxml2'

See https://github.com/llvm/llvm-project/issues/55817 and
https://discourse.llvm.org/t/cannot-cmake-self-hosted-clang-on-windows-for-lack-of-libxml2/58793

Not sure where is the best place to put this, but hopefully this will be
found by those searching for the error message.

Differential revision: https://reviews.llvm.org/D129770

2 years ago[lldb] [test] Skip new NonStop tests on Windows
Michał Górny [Fri, 15 Jul 2022 14:02:36 +0000 (16:02 +0200)]
[lldb] [test] Skip new NonStop tests on Windows

2 years ago[lldb][AArch64] Enable int128 test
David Spickett [Fri, 15 Jul 2022 13:35:48 +0000 (13:35 +0000)]
[lldb][AArch64] Enable int128 test

Other 128 bit tests run on AArch64 fine, it's
Arm (as in 32 bit Arm) that needs to be skipped.

2 years agoReland "[lldb] [llgs] Fix multi-resume bugs with nonstop mode"
Michał Górny [Tue, 28 Jun 2022 04:00:46 +0000 (06:00 +0200)]
Reland "[lldb] [llgs] Fix multi-resume bugs with nonstop mode"

Improve handling of multiple successive continue packets in non-stop
mode.  More specifically:

1. Explicitly send error response (instead of crashing on assertion)
   if the user attempts to resume the same process twice.  Since we
   do not support thread-level non-stop mode, one needs to always stop
   the process explicitly before resuming another thread set.

2. Actually stop the process if "vCont;t" is delivered to a running
   process.  Similarly, we only support stopping all the running threads
   simultaneously (via -1) and return an error in any other case.

With this patch, running multiple processes simultaneously is still
unsupported.  The patch also employs a hack to avoid enabling stdio
forwarding on "vCont;t" packet.  Both of these issues are addressed
by followup patches.

Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.llvm.org/D128710

2 years ago[pseudo] Share the underly payload when stripping comments for a token stream
Haojian Wu [Fri, 15 Jul 2022 13:12:46 +0000 (15:12 +0200)]
[pseudo] Share the underly payload when stripping comments for a token stream

`stripComments(cook(...))` is a common pattern being written.
Without this patch, this has a use-after-free issue (cook returns a temporary
TokenStream object which has its own payload, but the payload is not
shared with the one returned by stripComments).

Reviewed By: sammccall

Differential Revision: https://reviews.llvm.org/D125311