LLVM GN Syncbot [Thu, 17 Dec 2020 00:07:28 +0000 (00:07 +0000)]
[gn build] Port
ac068e014b2
Atmn [Sun, 6 Dec 2020 02:41:09 +0000 (21:41 -0500)]
[OpenMP][Libomptarget][NFC] Use CMake Variables
This patchs adds CMake variables to add subdirectories and include
directories for libomptarget and explicitly gives the location of source
files.
Differential Revision: https://reviews.llvm.org/D93290
Hongtao Yu [Wed, 16 Dec 2020 20:54:50 +0000 (12:54 -0800)]
[CSSPGO] Consume pseudo-probe-based AutoFDO profile
This change enables pseudo-probe-based sample counts to be consumed by the sample profile loader under the regular `-fprofile-sample-use` switch with minimal adjustments to the existing sample file formats. After the counts are imported, a probe helper, aka, a `PseudoProbeManager` object, is automatically launched to verify the CFG checksum of every function in the current compilation against the corresponding checksum from the profile. Mismatched checksums will cause a function profile to be slipped. A `SampleProfileProber` pass is scheduled before any of the `SampleProfileLoader` instances so that the CFG checksums as well as probe mappings are available during the profile loading time. The `PseudoProbeManager` object is set up right after the profile reading is done. In the future a CFG-based fuzzy matching could be done in `PseudoProbeManager`.
Samples will be applied only to pseudo probe instructions as well as probed callsites once the checksum verification goes through. Those instructions are processed in the same way that regular instructions would be processed in the line-number-based scenario. In other words, a function is processed in a regular way as if it was reduced to just containing pseudo probes (block probes and callsites).
**Adjustment to profile format **
A CFG checksum field is being added to the existing AutoFDO profile formats. So far only the text format and the extended binary format are supported. For the text format, a new line like
```
!CFGChecksum: 12345
```
is added to the end of the body sample lines. For the extended binary profile format, we introduce a metadata section to store the checksum map from function names to their CFG checksums.
Differential Revision: https://reviews.llvm.org/D92347
Guozhi Wei [Wed, 16 Dec 2020 23:42:08 +0000 (15:42 -0800)]
[MBP] Add whole chain to BlockFilterSet instead of individual BB
Currently we add individual BB to BlockFilterSet if its frequency satisfies
LoopFreq / Freq <= LoopToColdBlockRatio
LoopFreq is edge frequency from outside to loop header.
LoopToColdBlockRatio is a command line parameter.
It doesn't make sense since we always layout whole chain, not individual BBs.
It may also cause a tricky problem. Sometimes it is possible that the LoopFreq
of an inner loop is smaller than LoopFreq of outer loop. So a BB can be in
BlockFilterSet of inner loop, but not in BlockFilterSet of outer loop,
like .cold in the test case. So it is added to the chain of inner loop. When
work on the outer loop, .cold is not added to BlockFilterSet, so the edge to
successor .problem is not counted in UnscheduledPredecessors of .problem chain.
But other blocks in the inner loop are added BlockFilterSet, so the whole inner
loop chain can be layout, and markChainSuccessors is called to decrease
UnscheduledPredecessors of following chains. markChainSuccessors calls
markBlockSuccessors for every BB, even it is not in BlockFilterSet, like .cold,
so .problem chain's UnscheduledPredecessors is decreased, but this edge was not
counted on in fillWorkLists, so .problem chain's UnscheduledPredecessors
becomes 0 when it still has an unscheduled predecessor .pred! And it causes
problems in following various successor BB selection algorithms.
Differential Revision: https://reviews.llvm.org/D89088
alex-t [Tue, 15 Dec 2020 16:05:56 +0000 (19:05 +0300)]
Disable Jump Threading for the targets with divergent control flow
Details: Jump Threading does not make sense for the targets with divergent CF
since they do not use branch prediction for speculative execution.
Also in the high level IR there is no enough information to conclude that the branch is divergent or uniform.
This may cause errors in further CF lowering.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D93302
Tom Roeder [Wed, 2 Dec 2020 01:23:55 +0000 (17:23 -0800)]
[ASTImporter] Add support for importing GenericSelectionExpr AST nodes.
This allows ASTs to be merged when they contain GenericSelectionExpr
nodes (this is _Generic from C11). This is needed, for example, for
CTU analysis of C code that makes use of _Generic, like the Linux
kernel.
The node is already supported in the AST, but it didn't have a matcher
in ASTMatchers. So, this change adds the matcher and adds support to
ASTImporter. Additionally, this change adds support for structural
equivalence of _Generic in the AST.
Reviewed By: martong, aaron.ballman
Differential Revision: https://reviews.llvm.org/D92600
Tom Roeder [Wed, 16 Dec 2020 22:58:22 +0000 (14:58 -0800)]
Test commit: add valid punctuation to a comment. NFC.
Gulfem Savrun Yeniceri [Wed, 16 Dec 2020 23:05:35 +0000 (15:05 -0800)]
[IR] Fixed the typo in attributes test
Fixed the typo introduced in D90275.
Differential Revision: https://reviews.llvm.org/D93420
Harald van Dijk [Wed, 16 Dec 2020 22:39:57 +0000 (22:39 +0000)]
[X86] Avoid %fs:(%eax) references in x32 mode
The ABI explains that %fs:(%eax) zero-extends %eax to 64 bits, and adds
that the TLS base address, but that the TLS base address need not be
at the start of the TLS block, TLS references may use negative offsets.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D93158
Thomas Preud'homme [Tue, 15 Dec 2020 23:05:45 +0000 (23:05 +0000)]
[Test] Fix undef var in catch-undef-behavior.c
Commit
9e52c43090f8cd980167bbd2719878ae36bcf6b5 removed the directive
defining LINE_1600 but left a string substitution to that variable in a
CHECK-NOT directive. This will make that CHECK-NOT directive always fail
to match, no matter the string.
This commit follows the pattern done in
9e52c43090f8cd980167bbd2719878ae36bcf6b5 of simplifying the CHECK-NOT to
only look for the function name and the opening parenthesis, thereby not
requiring the LINE_1600 variable.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D93350
Alexander Belyaev [Wed, 16 Dec 2020 22:09:04 +0000 (23:09 +0100)]
Revert "[mlir] Lookup the latest value with a legal type when remapping values."
This reverts commit
f8184d4c44dff1fab13122221f0c23ab50936647.
Mehdi Amini [Wed, 16 Dec 2020 04:18:41 +0000 (04:18 +0000)]
Improve the verifier diagnostic on dominance error
Address PR47937
Differential Revision: https://reviews.llvm.org/D93361
Roman Lebedev [Wed, 16 Dec 2020 20:04:41 +0000 (23:04 +0300)]
[SimplifyCFG] Teach mergeEmptyReturnBlocks() to preserve DomTree
A first real transformation that didn't already knew how to do that,
but it's pretty tame - either change successor of all the predecessors
of a block and carefully delay deletion of the block until afterwards
the DomTree updates are appled, or add a successor to the block.
There wasn't a great test coverage for this, so i added extra, to be sure.
Roman Lebedev [Wed, 16 Dec 2020 21:11:42 +0000 (00:11 +0300)]
[SimplifyCFG] TryToSimplifyUncondBranchFromEmptyBlock() already knows how to preserve DomTree
... so just ensure that we pass DomTreeUpdater it into it.
Fixes DomTree preservation for a large number of tests,
all of which are marked as such so that they do not regress.
Roman Lebedev [Wed, 16 Dec 2020 19:34:05 +0000 (22:34 +0300)]
[SimplifyCFG] MergeBlockIntoPredecessor() already knows how to preserve DomTree
... so just ensure that we pass DomTreeUpdater it into it.
Fixes DomTree preservation for a large number of tests,
all of which are marked as such so that they do not regress.
Roman Lebedev [Wed, 16 Dec 2020 18:48:29 +0000 (21:48 +0300)]
[SimplifyCFG] removeUnreachableBlocks() already knows how to preserve DomTree
... so just ensure that we pass DomTreeUpdater it into it.
Apparently, there were no dedicated tests just for that functionality,
so i'm adding one here.
Roman Lebedev [Wed, 16 Dec 2020 18:26:43 +0000 (21:26 +0300)]
[NFCI][SimplifyCFG] Mark all the SimplifyCFG tests that already don't invalidate DomTree as such
First step after
e1133179587dd895962a2fe4d6eb0cb1e63b5ee2,
in these tests, DomTree is valid afterwards, so mark them as such,
so that they don't regress.
In further steps, SimplifyCFG transforms shall taught to preserve DomTree,
in as small steps as possible.
Fangrui Song [Wed, 16 Dec 2020 22:02:25 +0000 (14:02 -0800)]
[AArch64InstPrinter] Use * 4096 instead of << 12
Left shirting a negative integer is undefined before C++20.
Louis Dionne [Wed, 16 Dec 2020 22:00:09 +0000 (17:00 -0500)]
[libc++] Fix CI Dockerfile
Installing clang-format-11 doesn't seem to work if it's done before
we've installed LLVM. I must admit I didn't try to get to the bottom
of the issue, since installing it after seems to work.
Reid Kleckner [Wed, 16 Dec 2020 21:48:47 +0000 (13:48 -0800)]
Fix split-debug.c test on Windows
Rong Xu [Wed, 16 Dec 2020 21:11:59 +0000 (13:11 -0800)]
[PGO] Use the sum of profile counts to fix the function entry count
Raw profile count values for each BB are not kept after profile
annotation. We record function entry count and branch weights
and use them to compute the count when needed. This mechanism
works well in a perfect world, but often breaks in real programs,
because of number prevision, inconsistent profile, or bugs in
BFI). This patch uses sum of profile count values to fix
function entry count to make the BFI count close to real profile
counts.
Differential Revision: https://reviews.llvm.org/D61540
Marek Kurdej [Wed, 16 Dec 2020 21:33:10 +0000 (16:33 -0500)]
[libc++] Install git-clang-format on CI nodes.
Two problems fixed:
* an old version of clang-format get installed by default (6.0).
* git-clang-format is not present, only git-clang-format-<version> (e.g. git-clang-format-6.0).
Solution:
* install clang-format-11 with explicit version
* make symlink git-clang-format to the latest version of git-clang-format-<version>
Differential Revision: https://reviews.llvm.org/D93201
Reid Kleckner [Wed, 16 Dec 2020 21:33:06 +0000 (13:33 -0800)]
Fix XCore test on Windows, the register order is reversed, for reasons unknown
Sanjay Patel [Wed, 16 Dec 2020 21:13:55 +0000 (16:13 -0500)]
[VectorCombine] adjust test alignments for better coverage; NFC
Fangrui Song [Wed, 16 Dec 2020 21:29:40 +0000 (13:29 -0800)]
[libc++abi] Remove redundant null pointer check in operator delete
Similar to D52401. Normally operator delete is defined in libc++abi
(LIBCPP_DISABLE_NEW_DELETE_DEFINITIONS is off by default).
C89 4.10.3.2 The free function
C99 7.20.3.2 The free function
C11 7.22.3.3 The free function
If ptr is a null pointer, no action shall occur.
free on MSDN:
If memblock is NULL, the pointer is ignored and free immediately returns.
Reviewed By: #libc_abi, ldionne
Differential Revision: https://reviews.llvm.org/D93339
Nikita Popov [Wed, 16 Dec 2020 20:34:47 +0000 (21:34 +0100)]
[DSE] Pass MemoryLocation by const ref (NFC)
Mircea Trofin [Tue, 15 Dec 2020 17:00:05 +0000 (09:00 -0800)]
[NFC] Remove unused prefixes in llvm/test/CodeGen/X86
Differential Revision: https://reviews.llvm.org/D92965
Sanjay Patel [Wed, 16 Dec 2020 19:22:46 +0000 (14:22 -0500)]
[VectorCombine] optimize alignment for load transform
Here's another minimal step suggested by D93229 / D93397 .
(I'm trying to be extra careful in these changes because
load transforms are easy to get wrong.)
We can optimistically choose the greater alignment of a
load and its pointer operand. As the test diffs show, this
can improve what would have been unaligned vector loads
into aligned loads.
When we enhance with gep offsets, we will need to adjust
the alignment calculation to include that offset.
Differential Revision: https://reviews.llvm.org/D93406
Shivanshu Goyal [Wed, 16 Dec 2020 20:01:54 +0000 (12:01 -0800)]
Print source location in the error message when parens are missing around sizeof typename and the expression is inside macro expansion
Given the following code:
```
void Foo(int);
void Baz()
{
Bar(sizeof int);
}
```
The error message which is printed today is this:
```
error: expected parentheses around type name in sizeof expression
```
There is no source location printed whatsoever, so fixing a compile break like this becomes extremely hard in a large codebase.
My change improves the error message. But it doesn't output a FixItHint because I wasn't able to figure out how to get the locations for left and right parens. So any tips would be appreciated.
```
<source>:7:6: error: expected parentheses around type name in sizeof expression
Bar(sizeof int);
^
```
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D91129
Richard Smith [Wed, 16 Dec 2020 19:59:57 +0000 (11:59 -0800)]
PR47474: Add test for Clang's current behavior.
Our current behavior rejects the example, following the current language
rules, but it's likely the rules will be revised to allow this example.
Mitch Phillips [Wed, 16 Dec 2020 19:57:55 +0000 (11:57 -0800)]
[sanitizer-common] Pickup llvm-symbolizer from $OUT/bin IFF exists.
Fix-forward for D93352.
Slight rework of the same idea, pickup the external symbolizer from the
binary directory iff it exists.
Yaxun (Sam) Liu [Thu, 10 Dec 2020 21:30:24 +0000 (16:30 -0500)]
[clang-offload-bundler] Add option -allow-missing-bundles
There are out-of-tree tools using clang-offload-bundler to extract
bundles from bundled files. When a bundle is not in the bundled
file, clang-offload-bundler is expected to emit an error message
and return non-zero value. However currently clang-offload-bundler
silently generates empty file for the missing bundles.
Since OpenMP/HIP toolchains expect the current behavior, an option
-allow-missing-bundles is added to let clang-offload-bundler
create empty file when a bundle is missing when unbundling.
The unbundling job action is updated to use this option by
default.
clang-offload-bundler itself will emit error when a bundle
is missing when unbundling by default.
Changes are also made to check duplicate targets in -targets
option and emit error.
Differential Revision: https://reviews.llvm.org/D93068
Emma Blink [Wed, 16 Dec 2020 19:42:07 +0000 (14:42 -0500)]
Mark implicit coroutine variables as being implicit
This prevents the clang-tidy readability-identifier-naming check from
triggering on implicit __coro_gro and __promise variables in coroutines.
Eugene Zhulenev [Wed, 16 Dec 2020 10:17:53 +0000 (02:17 -0800)]
[mlir] Async: re-enable tests after fixing fkakines
Test flakiness was fixed by: https://github.com/llvm/llvm-project/commit/
9edcedf7f222ce7c893d1e3bf19b3a7a1f0f2218
Runs these tests to verify that all parts of the lowering work correctly.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D93384
Eugene Zhulenev [Wed, 16 Dec 2020 13:14:12 +0000 (05:14 -0800)]
[mlir] AsyncRuntime: mode runtime declarations to mlir::runtime namespace
Define Async runtime related typedefs in the `mlir::runtime` namespace.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D93391
Peter Steinfeld [Tue, 15 Dec 2020 22:44:22 +0000 (14:44 -0800)]
[flang] Handle undeclared names in EQUIVALENCE statements
Names in EQUIVALENCE statements are only allowed to indicate local
objects as per 19.5.1.4, paragraph 2, item (10). Thus, a name appearing
in an EQUIVALENCE statement with no corresponding declaration in the
same scope is an implicit declaration of the name. If that scope
contains an IMPLICIT NONE, it's an error.
I implemented this by adding a state variable to ScopeHandler to
indicate if we're resolving the names in an EQUIVALENCE statement and
then checked this state when resolving names. I also added a test to
the existing tests for EQUIVALENCE statements.
Differential Revision: https://reviews.llvm.org/D93345
Jordan Rupprecht [Wed, 16 Dec 2020 18:53:18 +0000 (10:53 -0800)]
[lldb][NFC] Apply performance-faster-string-find (`str.find("X")` -> `str.find('x')`)
Florian Hahn [Wed, 16 Dec 2020 16:51:14 +0000 (16:51 +0000)]
[SimplifyCFG] Precommit test for preserving !annotation.
Fangrui Song [Wed, 16 Dec 2020 18:09:58 +0000 (10:09 -0800)]
[Go] Fix bindings/go/llvm/IRBindings.cpp
Christian Sigg [Tue, 15 Dec 2020 20:15:28 +0000 (21:15 +0100)]
[mlir] Fix for gpu-async-region pass.
- the !gpu.async.token is the second result of 'gpu.alloc async', not the first.
- async.execute construction takes operand types not yet wrapped in !async.value.
- fix typo
Reviewed By: herhut
Differential Revision: https://reviews.llvm.org/D93156
LLVM GN Syncbot [Wed, 16 Dec 2020 17:38:06 +0000 (17:38 +0000)]
[gn build] Port
6eff12788ee
Bardia Mahjour [Wed, 16 Dec 2020 17:34:38 +0000 (12:34 -0500)]
[DDG] Data Dependence Graph - DOT printer - recommit
This is being recommitted to try and address the MSVC complaint.
This patch implements a DDG printer pass that generates a graph in
the DOT description language, providing a more visually appealing
representation of the DDG. Similar to the CFG DOT printer, this
functionality is provided under an option called -dot-ddg and can
be generated in a less verbose mode under -dot-ddg-only option.
Reviewed By: Meinersbur
Differential Revision: https://reviews.llvm.org/D90159
Mehdi Amini [Wed, 16 Dec 2020 17:31:22 +0000 (17:31 +0000)]
Remove spurious MLIRLLVMConversionsIncGen dependency from LLVM Dialect (NFC)
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D93335
Sanjay Patel [Wed, 16 Dec 2020 17:11:57 +0000 (12:11 -0500)]
[VectorCombine] loosen alignment constraint for load transform
As discussed in D93229, we only need a minimal alignment constraint
when querying whether a hypothetical vector load is safe. We still
pass/use the potentially stronger alignment attribute when checking
costs and creating the new load.
There's already a test that changes with the minimum code change,
so splitting this off as a preliminary commit independent of any
gep/offset enhancements.
Differential Revision: https://reviews.llvm.org/D93397
Fangrui Song [Wed, 16 Dec 2020 17:20:55 +0000 (09:20 -0800)]
[AArch64InstPrinter] Change printADRPLabel to print the target address in hexadecimal form
Similar to D77853. Change ADRP to print the target address in hex, instead of the raw immediate.
The behavior is similar to GNU objdump but we also include `0x`.
Note: GNU objdump is not consistent whether or not to emit `0x` for different architectures. We try emitting 0x consistently for all targets.
```
GNU objdump: adrp x16,
10000000
Old llvm-objdump: adrp x16, #0
New llvm-objdump: adrp x16, 0x10000000
```
`adrp Xd, 0x...` assembles to a relocation referencing `*ABS*+0x10000` which is not intended. We need to use a linker or use yaml2obj.
The main test is `test/tools/llvm-objdump/ELF/AArch64/pcrel-address.yaml`
Differential Revision: https://reviews.llvm.org/D93241
Esme-Yi [Wed, 16 Dec 2020 17:12:24 +0000 (17:12 +0000)]
Revert "[PowerPC] Extend folding RLWINM + RLWINM to post-RA."
This reverts commit
1c0941e1524f499e3fbde48fc3bdd0e70fc8f2e4.
Whitney Tsang [Wed, 16 Dec 2020 17:05:44 +0000 (17:05 +0000)]
[LoopNest] Handle loop-nest passes in LoopPassManager
Per http://llvm.org/OpenProjects.html#llvm_loopnest, the goal of this
patch (and other following patches) is to create facilities that allow
implementing loop nest passes that run on top-level loop nests for the
New Pass Manager.
This patch extends the functionality of LoopPassManager to handle
loop-nest passes by specializing the definition of LoopPassManager that
accepts both kinds of passes in addPass.
Only loop passes are executed if L is not a top-level one, and both
kinds of passes are executed if L is top-level. Currently, loop nest
passes should have the following run method:
PreservedAnalyses run(LoopNest &, LoopAnalysisManager &,
LoopStandardAnalysisResults &, LPMUpdater &);
Reviewed By: Whitney, ychen
Differential Revision: https://reviews.llvm.org/D87045
Jon Chesterfield [Wed, 16 Dec 2020 17:00:18 +0000 (17:00 +0000)]
[NFC] Use regex for code object version in hip tests
[NFC] Use regex for code object version in hip tests
Extracted from D93258. Makes tests robust to changes in default
code object version.
Reviewed By: t-tye
Differential Revision: https://reviews.llvm.org/D93398
Fangrui Song [Wed, 16 Dec 2020 16:59:38 +0000 (08:59 -0800)]
[ELF] --emit-relocs: fix a crash if .rela.dyn is an empty output section
Fix PR48357: If .rela.dyn appears as an output section description, its type may
be SHT_RELA (due to the empty synthetic .rela.plt) while there is no input
section. The empty .rela.dyn may be retained due to a reference in a linker
script. Don't crash.
Reviewed By: grimar
Differential Revision: https://reviews.llvm.org/D93367
Jon Chesterfield [Wed, 16 Dec 2020 16:40:21 +0000 (16:40 +0000)]
[libomptarget][nfc] Replace static const with enum
[libomptarget][nfc] Replace static const with enum
Semantically identical. Replaces 0xff... with ~0 to spare counting the f.
Has the advantage that the compiler doesn't need to prove the 4/8 byte
value dead before discarding it, and sidesteps the compilation question
associated with what static means for a single source language.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D93328
Jonas Hahnfeld [Sat, 5 Dec 2020 11:52:38 +0000 (12:52 +0100)]
[CMake] Avoid __FakeVCSRevision.h with no git repository
Set the return variable to "" in find_first_existing_vc_file to
say that there is a repository, but no file to depend on. This works
transparently for all other callers that handle undefinedness and
equality to an empty string the same way.
Use the knowledge to avoid depending on __FakeVCSRevision.h if there
is no git repository at all (for example when building a release) as
there is no point in regenerating an empty VCSRevision.h.
Differential Revision: https://reviews.llvm.org/D92718
peter klausler [Tue, 15 Dec 2020 18:59:26 +0000 (10:59 -0800)]
[flang] Fix crash in folding (#48437)
Elemental intrinsic function folding was not taking the lower
bounds of constant array arguments into account; these lower bounds
can be distinct from 1 when named constants appear as arguments.
LLVM bugzilla #48437.
Differential Revision: https://reviews.llvm.org/D93321
Kai Nacke [Wed, 16 Dec 2020 11:01:08 +0000 (06:01 -0500)]
[Doc][SystemZ] Add Linux/SystemZ to Getting Started guide.
The Linux/SystemZ platform is missing in the Getting Started guide
as platform on which LLVM is known to work.
Reviewed by: uweigand
Differential Revision: https://reviews.llvm.org/D93388
Florian Hahn [Wed, 16 Dec 2020 11:30:01 +0000 (11:30 +0000)]
[InstCombine] Precommit tests for !annotation metadata handling.
Matt Arsenault [Wed, 16 Dec 2020 15:25:07 +0000 (10:25 -0500)]
LangRef: Update byval/sret description for required types
Tim Keith [Wed, 16 Dec 2020 15:06:53 +0000 (07:06 -0800)]
[flang] Handle multiple names for same operator
Some operators have more than one name, e.g. operator(==), operator(.eq).
That was working correctly in generic definitions but they can also
appear in other contexts: USE statements and access statements, for
example.
This changes FindInScope to always look for each of the names for
a symbol. So an operator may be use-associated under one name but
declared private under another name and it will be the same symbol.
This replaces GenericSpecInfo::FindInScope which was only usable in
some cases.
Add a version of FindInScope() that looks in the current scope to
simplify many of the calls.
Differential Revision: https://reviews.llvm.org/D93344
Erik Pilkington [Tue, 15 Dec 2020 22:01:20 +0000 (17:01 -0500)]
NFC: balance a quote in AttrDocs.td
This was confusing my editor.
Erik Pilkington [Tue, 15 Dec 2020 22:00:20 +0000 (17:00 -0500)]
[Sema] Fix a miscompile by retaining array qualifiers when folding VLAs to constant arrays
rdar://
72243125
Differential revision: https://reviews.llvm.org/D93247
Paul C. Anagnostopoulos [Fri, 11 Dec 2020 21:19:52 +0000 (16:19 -0500)]
[TableGen] Slim down the data structures in xxxGenInstrInfo.inc, step 1
Simon Pilgrim [Wed, 16 Dec 2020 14:42:08 +0000 (14:42 +0000)]
[X86] Regenerate bit extraction tests, cleaning up check-prefixes.
As noticed on D92965, we needed to simplify the prefixes to ensure all RUNs were properly covered.
We should never have a target with BMI2 without BMI1, so use that as the 'BMI level' and then check with/without TBM (all TBM targets have at least BMI1).
Simon Pilgrim [Wed, 16 Dec 2020 13:31:10 +0000 (13:31 +0000)]
[X86] Rename reduction combiners to make it clearer whats happening. NFCI.
Since these are all working on reduction patterns, actually use that term in the function name to make them easier to search for.
At some point we're likely to start working with the ISD::VECREDUCE_* opcodes directly in the x86 backend, but that is still some way off.
Mitch Phillips [Wed, 16 Dec 2020 14:36:22 +0000 (06:36 -0800)]
[sanitizer-common] Force pickup of llvm-symbolizer from new binaries.
It's possible currently that the sanitizer runtimes when testing grab
the path to the symbolizer through *SAN_SYMBOLIZER_PATH=...
This can be polluted by things like Android's setup script. This patch
forces external_symbolizer_path=$new_build_out_dir/llvm-symbolizer when
%env_tool_options is used.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D93352
diggerlin [Wed, 16 Dec 2020 14:34:59 +0000 (09:34 -0500)]
[XCOFF][AIX] Emit EH information in traceback table
SUMMARY:
In order for the runtime on AIX to find the compact unwind section(EHInfo table),
we would need to set the following on the traceback table:
The 6th byte's longtbtable field to true to signal there is an Extended TB Table Flag.
The Extended TB Table Flag to be 0x08 to signal there is an exception handling info presents.
Emit the offset between ehinfo TC entry and TOC base after all other optional portions of traceback table.
The patch is authored by Jason Liu.
Reviewers: David Tenty, Digger Lin
Differential Revision: https://reviews.llvm.org/D92766
Caroline Concatto [Thu, 10 Dec 2020 17:23:46 +0000 (17:23 +0000)]
[SLPVectorizer]Migrate getEntryCost to return InstructionCost
This patch also changes:
the return type of getGatherCost and
the signature of the debug function dumpTreeCosts
to use InstructionCost.
This patch is part of a series of patches to use InstructionCost instead of
unsigned/int for the cost model functions.
See this thread for context:
http://lists.llvm.org/pipermail/llvm-dev/2020-November/146408.html
See this patch for the introduction of the type:
https://reviews.llvm.org/D91174
Depends on D93049
Differential Revision: https://reviews.llvm.org/D93127
Denis Antrushin [Tue, 15 Dec 2020 12:14:39 +0000 (19:14 +0700)]
[Statepoints] Extract invoke tests into separate file. NFC.
Extract VReg lowering tests with invokes into separate file
for easier maintenance/modification.
Check MIR after register allocation - at this point all
transformations we're interested in has been applied and verifying
of MIR is simpler than that of assembly.
Stephan Bergmann [Wed, 16 Dec 2020 11:46:10 +0000 (12:46 +0100)]
[PATCH] [compiler-rt] [builtins] Fix name of __aarch64_have_lse_atomics on Darwin
...where the name of that variable defined in
compiler-rt/lib/builtins/cpu_model.c is decorated with a leading underscore
Differential Revision: https://reviews.llvm.org/D93390
Bradley Smith [Wed, 9 Dec 2020 11:03:54 +0000 (11:03 +0000)]
[CostModel] Add costs for llvm.experimental.vector.{extract,insert} intrinsics
Adds cost model support for the new llvm.experimental.vector.{extract,insert}
intrinsics, using the existing getExtractSubvectorOverhead and
getInsertSubvectorOverhead functions for shuffles.
Previously this case would throw an assertion.
Differential Revision: https://reviews.llvm.org/D93043
ergawy [Wed, 16 Dec 2020 13:20:24 +0000 (08:20 -0500)]
[mlir][spirv] Add parsing and printing support for SpecConstantOperation
Adds more support for `SpecConstantOperation` by defining a custom
syntax for the op and implementing its parsing and printing.
Reviewed By: mravishankar, antiagainst
Differential Revision: https://reviews.llvm.org/D92919
Caroline Concatto [Thu, 10 Dec 2020 13:52:35 +0000 (13:52 +0000)]
[CostModel]Migrate getTreeCost() to use InstructionCost
This patch changes the type of cost variables (for instance: Cost, ExtractCost,
SpillCost) to use InstructionCost.
This patch also changes the type of cost variables to InstructionCost in other
functions that use the result of getTreeCost()
This patch is part of a series of patches to use InstructionCost instead of
unsigned/int for the cost model functions.
See this thread for context:
http://lists.llvm.org/pipermail/llvm-dev/2020-November/146408.html
Depends on D91174
Differential Revision: https://reviews.llvm.org/D93049
Caroline Concatto [Wed, 25 Nov 2020 12:06:27 +0000 (12:06 +0000)]
[CostModel]Replace FixedVectorType by VectorType in costgetIntrinsicInstrCost
This patch replaces FixedVectorType by VectorType in getIntrinsicInstrCost
in BasicTTIImpl.h. It re-arranges the scalable type test earlier return
and add tests for scalable types.
Depends on D91532
Differential Revision: https://reviews.llvm.org/D92094
Joe Ellis [Wed, 16 Dec 2020 09:01:28 +0000 (09:01 +0000)]
[clang][AArch64][SVE] Avoid going through memory for VLAT <-> VLST casts
This change makes use of the llvm.vector.extract intrinsic to avoid
going through memory when performing bitcasts between vector-length
agnostic types and vector-length specific types.
Depends on D91362
Reviewed By: c-rhodes
Differential Revision: https://reviews.llvm.org/D92761
Muhammad Omair Javaid [Wed, 16 Dec 2020 11:09:24 +0000 (16:09 +0500)]
[LLDB] Skip TestGuiBasicDebug.py on Arm and AArch64/Linux
TestGuiBasicDebug.py is intermittenly timing out on LLDB AArch64/Linux
buildbot. Putting SkipIf decorator untill root cuase is identified.
Bangtian Liu [Wed, 16 Dec 2020 11:51:26 +0000 (11:51 +0000)]
Revert "Ensure SplitEdge to return the new block between the two given blocks"
This reverts commit
cf638d793c489632bbcf0ee0fbf9d0f8c76e1f48.
Simon Pilgrim [Wed, 16 Dec 2020 11:24:19 +0000 (11:24 +0000)]
[X86][SSE] combineReductionToHorizontal - don't rely on widenSubVector to handle illegal vector types.
Thanks to @asbirlea for reporting the bug.
Muhammad Omair Javaid [Wed, 16 Dec 2020 11:02:54 +0000 (16:02 +0500)]
Revert "[LLDB] Remove AArch64/Linux xfail decorator from TestGuiBasicDebug"
This reverts commit
3d27a99b2ed24e1951483cf13357ec188ad44bb0.
Georgii Rymar [Wed, 9 Dec 2020 13:46:42 +0000 (16:46 +0300)]
[yaml2obj/obj2yaml] - Make Value/Size fields of Symbol optional.
When a field is optional we can use the `=<none>` syntax in macros.
This patch makes `Value`/`Size` fields of `Symbol` optional
and adds test cases for them.
Differential revision: https://reviews.llvm.org/D93010
Georgii Rymar [Tue, 15 Dec 2020 12:45:15 +0000 (15:45 +0300)]
[lib/Object] - Make ELFObjectFile::getSymbol() return Expected<>.
This was requested in comments for D93209:
https://reviews.llvm.org/D93209#inline-871192
D93209 fixes an issue with `ELFFile<ELFT>::getEntry`,
after what `getSymbol` starts calling `report_fatal_error` for previously
missed invalid cases.
This patch makes it return `Expected<>` and updates callers.
For few of them I had to add new `report_fatal_error` calls. But I see no
way to avoid it currently. The change would affects too many places, e.g:
`getSymbolBinding` and other methods are used from `ELFSymbolRef`
which is used in too many places across LLVM.
Differential revision: https://reviews.llvm.org/D93297
Georgii Rymar [Fri, 4 Dec 2020 08:54:03 +0000 (11:54 +0300)]
[llvm-readelf/obj] - Handle out-of-order PT_LOADs better.
This is https://bugs.llvm.org/show_bug.cgi?id=45698.
Specification says that
"Loadable segment entries in the program header table appear
in ascending order, sorted on the p_vaddr member."
Our `toMappedAddr()` relies on this condition. This patch
adds a warning when the sorting order of loadable segments is wrong.
In this case we force segments sorting and that allows
`toMappedAddr()` to work as expected.
Differential revision: https://reviews.llvm.org/D92641
Jan Svoboda [Tue, 15 Dec 2020 12:08:04 +0000 (13:08 +0100)]
[clang][cli] Port DependencyOutput string based options to new option parsing system
Depends on D84186
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D84187
Piotr Sobczak [Wed, 16 Dec 2020 09:12:54 +0000 (10:12 +0100)]
[AMDGPU] Avoid calling copyFastMathFlags in wrong context
Calling Instruction::copyFastMathFlags() assumes the caller is
FPMathOperator. Avoid calling the function for instructions
that are not instances of FPMathOperator.
Qiu Chaofan [Wed, 16 Dec 2020 09:19:54 +0000 (17:19 +0800)]
[NFC] Pre-commit test for long-double builtins
This test reflects clang behavior on long-double type math library
builtins under default or explicit 128-bit long-double options.
Jan Svoboda [Wed, 16 Dec 2020 09:05:09 +0000 (10:05 +0100)]
[clang][cli] Port LangOpts option flags to new option parsing system
Depends on D83892 & D83694 & D93104.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D83979
Sebastian Neubauer [Tue, 15 Dec 2020 10:15:53 +0000 (11:15 +0100)]
[AMDGPU] Allow no saddr for global addtid insts
I think the global_load/store_dword_addtid instructions support
switching off the scalar address.
Add assembler and disassembler support for this.
Differential Revision: https://reviews.llvm.org/D93288
Jan Svoboda [Fri, 11 Dec 2020 14:41:35 +0000 (15:41 +0100)]
[clang][cli] Port CodeGen option flags to new option parsing system
Depends on D83697.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D83892
Alexander Belyaev [Wed, 16 Dec 2020 07:38:12 +0000 (08:38 +0100)]
[mlir] Lookup the latest value with a legal type when remapping values.
The current condition implies that the target materialization will be
called even if the type is the new operand type is legal, but slightly
different. For example, if there is a bufferization pattern that changes
memref layout, then target materialization for an illegal type
(TensorType) would be called.
Differential Revision: https://reviews.llvm.org/D93126
Jan Svoboda [Wed, 9 Dec 2020 12:26:39 +0000 (13:26 +0100)]
[clang][cli] Prevent double denormalization
If both flags created through BoolOption are CC1Option and the keypath has a non-default or non-implied value, the denormalizer gets called twice. If the denormalizer has the ability to generate both flags, we can end up generating the same flag twice.
Reviewed By: dexonsmith, Bigcheese
Differential Revision: https://reviews.llvm.org/D93094
Jan Svoboda [Tue, 8 Dec 2020 17:15:21 +0000 (18:15 +0100)]
[clang][cli] Do not marshall only CC1Option flags in BoolOption
We cannot be sure whether a flag is CC1Option inside the definition of `BoolOption`. Take the example below:
```
let Flags = [CC1Option] in {
defm xxx : BoolOption<...>;
}
```
where TableGen applies `Flags = [CC1Option]` to the `xxx` and `no_xxx` records **after** they have been is fully declared by `BoolOption`.
For the refactored `-f[no-]debug-pass-manager` flags (see the diff), this means `BoolOption` never adds any marshalling info, as it doesn't see either of the flags as `CC1Option`.
For that reason, we should defensively append the marshalling information to both flags inside `BoolOption`. Now the check for `CC1Option` needs to happen later, in the parsing macro, when all TableGen logic has been resolved.
However, for some flags defined through `BoolOption`, we can run into issues:
```
// Options.td
def fenable_xxx : /* ... */;
// Both flags are CC1Option, the first is implied.
defm xxx : BoolOption<"xxx,
"Opts.Xxx", DefaultsToFalse,
ChangedBy<PosFlag, [CC1Option], "", [fenable_xxx]>,
ResetBy<NegFlag, [CC1Option]>>;
```
When parsing `clang -cc1 -fenable-xxx`:
* we run parsing for `PosFlag`:
* set `Opts.Xxx` to default `false`,
* discover `PosFlag` is implied by `-fenable-xxx`: set `Opts.Xxx` to `true`,
* don't see `-fxxx` on command line: do nothing,
* we run parsing for `NegFlag`:
* set `Opts.Xxx` to default `false`,
* discover `NegFlag` cannot be implied: do nothing,
* don't see `-fno-xxx` on command line: do nothing.
Now we ended up with `Opts.Xxx` set to `false` instead of `true`. For this reason, we need to ensure to append the same `ImpliedByAnyOf` instance to both flags.
This means both parsing runs now behave identically (they set the same default value, run the same "implied by" check, and call `makeBooleanOptionNormalizer` that already has information on both flags, so it returns the same value in both calls).
The solution works well, but what might be confusing is this: you have defined a flag **A** that is not `CC1Option`, but can be implied by another flag **B** that is `CC1Option`:
* if **A** is defined manually, it will never get implied, as the code never runs
```
def no_signed_zeros : Flag<["-"], "fno-signed-zeros">, Group<f_Group>, Flags<[]>,
MarshallingInfoFlag<"LangOpts->NoSignedZero">, ImpliedByAnyOf<[menable_unsafe_fp_math]>;
```
* if **A** is defined by `BoolOption`, it could get implied, as the code is run by its `CC1Option` counterpart:
```
defm signed_zeros : BoolOption<"signed-zeros",
"LangOpts->NoSignedZero", DefaultsToFalse,
ChangedBy<NegFlag, [], "Allow optimizations that ignore the sign of floating point zeros",
[cl_no_signed_zeros, menable_unsafe_fp_math]>,
ResetBy<PosFlag, [CC1Option]>, "f">, Group<f_Group>;
```
This is a surprising inconsistency.
One solution might be to somehow propagate the final `Flags` of the implied flag in `ImpliedByAnyOf` and check whether it has `CC1Option` in the parsing macro. However, I think it doesn't make sense to spend time thinking about a corner case that won't come up in real code.
An observation: it is unfortunate that the marshalling information is a part of the flag definition. If we represented it in a separate structure, we could avoid the "double parsing" problem by having a single source of truth. This would require a lot of additional work though.
Note: the original patch missed the `CC1Option` check in the parsing macro, making my reasoning here incomplete. Moreover, it contained a change to denormalization that wasn't necessarily related to these changes, so I've extracted that to a follow-up patch: D93094.
Reviewed By: dexonsmith, Bigcheese
Differential Revision: https://reviews.llvm.org/D93008
Raul Tambre [Wed, 16 Dec 2020 08:14:41 +0000 (10:14 +0200)]
Revert "[Compiler-rt][AArch64] Workaround for .cfi_startproc assembler parser bug."
039cb03dd0dcff4daa17a062f7414ac22bf5f2eb (D93378) fixed the assembly separator, so the workaround is no longer necessary.
This reverts commit
3000c19df64f89ff319590f3a6e4d6b93d20983d.
Differential Revision: https://reviews.llvm.org/D93379
Fangrui Song [Wed, 16 Dec 2020 08:13:22 +0000 (00:13 -0800)]
[builtins] Use %% as the separator for aarch64-*-darwin and ; for powerpc{32,64}
`;` is the default comment marker, which is also used by powerpc*-*-elf target triples.
`@` is the comment marker of powerpc*-*-darwin but the Darwin support has been deleted for PowerPC (D72063).
`%%` is the statement separator used by aarch64-*-darwin (see AArch64MCAsmInfoDarwin, it uses `;` as the comment marker, which is different from most other targets)
Reviewed By: tambre
Differential Revision: https://reviews.llvm.org/D93378
Siva Chandra Reddy [Wed, 16 Dec 2020 07:44:44 +0000 (23:44 -0800)]
[libc][NFC] Use ASSERT_FP_EQ to comapre NaN values in tests.
Stanislav Mekhanoshin [Tue, 15 Dec 2020 23:31:29 +0000 (15:31 -0800)]
[AMDGPU] Print SCRATCH_EN field after the kernel
Differential Revision: https://reviews.llvm.org/D93353
Siva Chandra Reddy [Wed, 16 Dec 2020 06:23:22 +0000 (22:23 -0800)]
[libc][NFC] Rename global `nan` in tests to `aNaN`.
The name `nan` conflicts with the function `nan` defined declared in
math.h.
Siva Chandra Reddy [Wed, 16 Dec 2020 05:40:40 +0000 (21:40 -0800)]
[libc][NFC][Obvious] Remove few unnecessary #include directives in tests.
Yaxun (Sam) Liu [Fri, 4 Dec 2020 22:38:57 +0000 (17:38 -0500)]
[HIP] unbundle bundled preprocessor output
There is a use case that users want to emit preprocessor
output as file and compile the preprocessor output later
with -x hip-cpp-output.
Clang emits bundled preprocessor output when users
compile with -E for combined host/device compilations.
Clang should be able to compile the bundled preprocessor
output with -x hip-cpp-output. Basically clang should
unbundle the bundled preprocessor output and launch
device and host compilation actions.
Currently there is a bug in clang driver causing bundled
preprocessor output not unbundled.
This patch fixes that.
Differential Revision: https://reviews.llvm.org/D92720
Philip Reames [Wed, 16 Dec 2020 03:06:12 +0000 (19:06 -0800)]
[LV] Weaken a unnecessarily strong assert [NFC]
Account for the fact that (in the future) the latch might be a switch not a branch. The existing code is correct, minus the assert.
Lang Hames [Wed, 16 Dec 2020 01:18:30 +0000 (12:18 +1100)]
[JITLink][ORC] Enable creation / linking of raw jitlink::LinkGraphs.
Separates link graph creation from linking. This allows raw LinkGraphs to be
created and passed to a link. ObjectLinkingLayer is updated to support emission
of raw LinkGraphs in addition to object buffers.
Raw LinkGraphs can be created by in-memory compilers to bypass object encoding /
decoding (though this prevents caching, as LinkGraphs have do not have an
on-disk representation), and by utility code to add programatically generated
data structures to the JIT target process.
Zakk Chen [Wed, 16 Dec 2020 01:27:38 +0000 (17:27 -0800)]
[RISCV] Refine vector load/store tablegen pattern, NFC.
Refine tablegen pattern for vector load/store, and follow
D93012 to separate masked and unmasked definitions for
pseudo load/store instructions.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D93284
Philip Reames [Wed, 16 Dec 2020 02:46:32 +0000 (18:46 -0800)]
[LV] Extend dead instruction detection to multiple exiting blocks
Given we haven't yet enabled multiple exiting blocks, this is currently non functional, but it's an obvious extension which cleans up a later patch.
I don't think this is worth review (as it's pretty obvious), if anyone disagrees, feel feel to revert or comment and I will.
Nathan James [Wed, 16 Dec 2020 01:56:30 +0000 (01:56 +0000)]
[ASTMatchers][NFC] Strip prefixes eagerly in dynamic matcher arg traits.
peter klausler [Tue, 15 Dec 2020 19:06:44 +0000 (11:06 -0800)]
[flang] Implement STORAGE_SIZE(), SIZEOF(), C_SIZEOF()
STORAGE_SIZE() is a standard inquiry intrinsic (size in bits
of an array element of the same type as the argument); SIZEOF()
is a common extension that returns the size in bytes of its
argument; C_SIZEOF() is a renaming of SIZEOF() in module ISO_C_BINDING.
STORAGE_SIZE() and SIZEOF() are implemented via rewrites to
expressions; these expressions will be constant when the necessary
type parameters and bounds are also constant.
Code to calculate the sizes of types (with and without alignment)
was isolated into Evaluate/type.* and /characteristics.*.
Code in Semantics/compute-offsets.* to calculate sizes and alignments
of derived types' scopes was exposed so that it can be called at type
instantiation time (earlier than before) so that these inquiry intrinsics
could be called from specification expressions.
Differential Revision: https://reviews.llvm.org/D93322