Dimitry Andric [Mon, 22 Apr 2019 19:20:46 +0000 (19:20 +0000)]
Use correct way to test for MIPS arch after rOMP355687
Summary:
I ran into some issues after rOMP355687, where __atomic_fetch_add was
being used incorrectly on x86, and this turns out to be caused by the
following added conditionals:
```
#if defined(KMP_ARCH_MIPS)
```
The problem is, these macros are always defined, and are either 0 or 1
depending on the architecture. E.g. the correct way to test for MIPS
is:
```
#if KMP_ARCH_MIPS
```
Reviewers: petarj, jlpeyton, Hahnfeld, AndreyChurbanov
Reviewed By: petarj, AndreyChurbanov
Subscribers: AndreyChurbanov, sdardis, arichardson, atanasyan, jfb, jdoerfert, openmp-commits, llvm-commits
Tags: #openmp
Differential Revision: https://reviews.llvm.org/D60938
llvm-svn: 358911
Matt Arsenault [Mon, 22 Apr 2019 19:14:27 +0000 (19:14 +0000)]
Use const DebugLoc&
llvm-svn: 358910
Matt Arsenault [Mon, 22 Apr 2019 19:14:26 +0000 (19:14 +0000)]
AMDGPU: Skip debug instructions in assert
These are inserted after branch relaxation, and for some reason it's
decided to put them in the long branch expansion block. It's probably
not great to rely on the source block address, so this should probably
be switched to being PC relative instead of relying on the block
address
llvm-svn: 358909
Casey Carter [Mon, 22 Apr 2019 19:08:13 +0000 (19:08 +0000)]
[libc++][test] Update some wstring_convert tests for MSVC quirks
Due to MSVC's decision to encode `wchar_t` as UTF-16, it rejects wide
character/string literals that expect a character value greater than
`\xffff`. UTF-16 `wchar_t` is clearly non-conforming, given that the
standard requires wchar_t to be capable of representing all characters
in the supported wide character execution sets, but rejecting e.g.
`\x40003` is a reasonably sane compromise given that encoding choice:
there's an expectation that `\xFOO` produces a single character in the
resulting literal. Consequently `L'\x40003'`/`L"\x40003"` are ill-formed
literals on MSVC. `L'\U00040003'` is a high surrogate (and produces a
warning about ignoring the "second character" in a multi-character
literal), and `L"\U00040003"` is a perfectly-valid `const wchar_t[3]`.
This change updates these tests to use universal-character-names instead
of raw values for the intended character values, which technically makes
them portable even to implementations that don't use a unicode
transformation format encoding for their wide character execution
character set. The two-character literal `L"\u1005e"` is awkward - the
`e` looks like part of the UCN's hex encoding - but necessary to compile
in '03 mode since '03 didn't allow UCNs to be used for members of the
basic execution character set even in character/string literals.
I've also eliminated the extraneous `\x00` "bonus null-terminator" in
some of the string literals which doesn't affect the tested behavior.
I'm sorry about using `*L"\U00040003"` in `conversions.string/to_bytes.pass.cpp`,
but it's correct for platforms with 32-bit wchar_t, *and* doesn't
trigger narrowing warnings as did the prior `CharT(0x40003)`.
Differential Revision: https://reviews.llvm.org/D60950
llvm-svn: 358908
Philip Reames [Mon, 22 Apr 2019 18:51:58 +0000 (18:51 +0000)]
[Tests] Revise a test as requested by reviewer in D59703
llvm-svn: 358907
Philip Reames [Mon, 22 Apr 2019 18:28:44 +0000 (18:28 +0000)]
[Tests] Add a negative test for masked.gather part of D59703
llvm-svn: 358906
Ben Hamilton [Mon, 22 Apr 2019 17:54:11 +0000 (17:54 +0000)]
[ASTMatchers] Introduce Objective-C matchers `isClassMessage`, `isClassMethod`, and `isInstanceMethod`
Summary:
isClassMessage is an equivalent to isInstanceMessage for ObjCMessageExpr, but matches message expressions to classes.
isClassMethod and isInstanceMethod check whether a method declaration (or definition) is for a class method or instance method (respectively).
Contributed by @mywman!
Reviewers: benhamilton, klimek, mwyman
Reviewed By: benhamilton, mwyman
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D60920
llvm-svn: 358904
Justin Bogner [Mon, 22 Apr 2019 17:38:29 +0000 (17:38 +0000)]
[IPSCCP] Add missing `AssumptionCacheTracker` dependency
Back in August, r340525 introduced a dependency on the assumption
cache tracker in the ipsccp pass, but that commit missed a call to
INITIALIZE_PASS_DEPENDENCY, which leaves the assumption cache
improperly registered if SCCP is the only thing that pulls it in.
llvm-svn: 358903
Alexey Bataev [Mon, 22 Apr 2019 17:25:31 +0000 (17:25 +0000)]
[OPENMP][NVPTX] Fix the test, NFC.
Fix the test to run it really in SPMD mode without runtime. Previously
it was run in SPMD + full runtime mode and does not allow to cehck the
functionality correctly.
llvm-svn: 358902
Philip Reames [Mon, 22 Apr 2019 17:13:43 +0000 (17:13 +0000)]
[LPM/BPI] Preserve BPI through trivial loop pass pipeline (e.g. LCSSA, LoopSimplify)
Currently, we do not expose BPI to loop passes at all. In the old pass manager, we appear to have been ignoring the fact that LCSSA and/or LoopSimplify didn't preserve BPI, and making it available to the following loop passes anyways. In the new one, it's invalidated before running any loop pass if either LCSSA or LoopSimplify actually make changes. If they don't make changes, then BPI is valid and available. So, we go ahead and teach LCSSA and LoopSimplify how to preserve BPI for consistency between old and new pass managers.
This patch avoids an invalidation between the two requires in the following trivial pass pipeline:
opt -passes="requires<branch-prob>,loop(no-op-loop),requires<branch-prob>"
(when the input file is one which requires either LCSSA or LoopSimplify to canonicalize the loops)
Differential Revision: https://reviews.llvm.org/D60790
llvm-svn: 358901
Wei Mi [Mon, 22 Apr 2019 17:04:51 +0000 (17:04 +0000)]
[PGO/SamplePGO][NFC] Move the function updateProfWeight from Instruction
to CallInst.
The issue was raised here: https://reviews.llvm.org/D60903#1472783
The function Instruction::updateProfWeight is only used for CallInst in
profile update. From the current interface, it is very easy to think that
the function can also be used for branch instruction. However, Branch
instruction does't need the scaling the function provides for
branch_weights and VP (value profile), in addition, scaling may introduce
inaccuracy for branch probablity.
The patch moves the function updateProfWeight from Instruction class to
CallInst to remove the confusion. The patch also changes the scaling of
branch_weights from a loop to a block because we know that ProfileData
for branch_weights of CallInst will only have two operands at most.
Differential Revision: https://reviews.llvm.org/D60911
llvm-svn: 358900
Sam Clegg [Mon, 22 Apr 2019 16:12:54 +0000 (16:12 +0000)]
[WebAssembly] Error on relocations against undefined data symbols.
We can't (currently) meaningfully resolve certain types of relocations
against undefined data symbols. Previously when `--allow-undefined` was
used we were treating such relocation much like weak data symbols and
simply inserting zeros. This change turns such use cases in to an
error.
This means that `--allow-undefined` is no longer effective for data
symbols.
Fixes https://bugs.llvm.org/show_bug.cgi?id=40364
Differential Revision: https://reviews.llvm.org/D60882
llvm-svn: 358899
Matt Davis [Mon, 22 Apr 2019 16:04:44 +0000 (16:04 +0000)]
[sema][objc] Minor refactor to OverrideSearch. NFCI.
Summary:
* Removed a member that was only used during construction.
* Use range-based for iteration when accessing the result of the search.
* Require an `ObjCMethodDecl` reference upon construction of an
* Constify.
Reviewers: rjmccall
Reviewed By: rjmccall
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D60850
llvm-svn: 358898
Fangrui Song [Mon, 22 Apr 2019 15:53:43 +0000 (15:53 +0000)]
Use llvm::stable_sort. NFC
llvm-svn: 358897
Louis Dionne [Mon, 22 Apr 2019 15:40:50 +0000 (15:40 +0000)]
[NFC] Fix typo in debug log
llvm-svn: 358896
Aaron Ballman [Mon, 22 Apr 2019 15:31:09 +0000 (15:31 +0000)]
Remove spurious semicolons; NFC.
llvm-svn: 358895
Matt Arsenault [Mon, 22 Apr 2019 15:22:46 +0000 (15:22 +0000)]
AMDGPU/GlobalISel: Fix non-power-of-2 G_EXTRACT sources
llvm-svn: 358894
Fangrui Song [Mon, 22 Apr 2019 15:19:13 +0000 (15:19 +0000)]
STLExtras: add stable_sort wrappers
llvm-svn: 358893
Matt Arsenault [Mon, 22 Apr 2019 15:10:42 +0000 (15:10 +0000)]
GlobalISel: Legalize scalar G_EXTRACT sources
llvm-svn: 358892
Nico Weber [Mon, 22 Apr 2019 15:05:18 +0000 (15:05 +0000)]
llvm-undname: Fix an assert-on-invalid, found by oss-fuzz
llvm-svn: 358891
Matt Arsenault [Mon, 22 Apr 2019 14:54:39 +0000 (14:54 +0000)]
AMDGPU: Fix not checking for copy when looking at copy src
Effectively reverts r356956. The check for isFullCopy was excessive,
but there still needs to be a check that this is a copy.
llvm-svn: 358890
George Rimar [Mon, 22 Apr 2019 14:53:32 +0000 (14:53 +0000)]
[LLD][ELF] - Remove a binary from the inputs. NFCI.
section-index.elf was removed and the corresponding test
was replaced with a yaml2obj based test.
llvm-svn: 358889
Dmitry Preobrazhensky [Mon, 22 Apr 2019 14:35:47 +0000 (14:35 +0000)]
[AMDGPU][MC] Corrected parsing of SP3 'neg' modifier
See bug 41156: https://bugs.llvm.org/show_bug.cgi?id=41156
Reviewers: artem.tamazov, arsenm
Differential Revision: https://reviews.llvm.org/D60624
llvm-svn: 358888
Simon Pilgrim [Mon, 22 Apr 2019 14:04:35 +0000 (14:04 +0000)]
[TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling
This patch adds support for BigBitWidth -> SmallBitWidth bitcasts, splitting the DemandedBits/Elts accordingly.
The AMDGPU backend needed an extra (srl (and x, c1 << c2), c2) -> (and (srl(x, c2), c1) combine to encourage BFE creation, I investigated putting this in DAGCombine but it caused a lot of noise on other targets - some improvements, some regressions.
The X86 changes are all definite wins.
Differential Revision: https://reviews.llvm.org/D60462
llvm-svn: 358887
Sanjay Patel [Mon, 22 Apr 2019 13:42:50 +0000 (13:42 +0000)]
[DAGCombiner] make variable name less ambiguous; NFC
llvm-svn: 358886
George Rimar [Mon, 22 Apr 2019 13:40:42 +0000 (13:40 +0000)]
[LLD][ELF] - Do not forget to use ch_addralign field after decompressing the sections.
LLD did not use ELF::Chdr::ch_addralign for decompressed sections.
This resulted in a broken output.
Fixes https://bugs.llvm.org/show_bug.cgi?id=40482.
Differential revision: https://reviews.llvm.org/D60959
llvm-svn: 358885
Sanjay Patel [Mon, 22 Apr 2019 13:36:07 +0000 (13:36 +0000)]
[DAGCombiner] prepare shuffle-of-splat to handle more patterns; NFC
llvm-svn: 358884
Robert Widmann [Mon, 22 Apr 2019 13:13:22 +0000 (13:13 +0000)]
[LLVM-C] Add accessors to the default floating-point metadata node
Summary: Add a getter and setter pair for floating-point accuracy metadata.
Reviewers: whitequark, deadalnix
Reviewed By: whitequark
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D60527
llvm-svn: 358883
Bruno Ricci [Mon, 22 Apr 2019 12:19:00 +0000 (12:19 +0000)]
[Sema] ADL: Associated namespaces for class types and enumeration types (CWG 1691)
CWG 1691 changed the definition of the namespaces associated with a class
type or enumeration type.
For a class type, the associated namespaces are the innermost enclosing
namespaces of the associated classes. For an enumeration type, the associated
namespace is the innermost enclosing namespace of its declaration.
This also fixes CWG 1690 and CWG 1692.
Differential Revision: https://reviews.llvm.org/D60573
Reviewed By: rjmccall, rsmith
llvm-svn: 358882
Bruno Ricci [Mon, 22 Apr 2019 11:40:31 +0000 (11:40 +0000)]
[Sema][NFC] Add more tests for the behavior of argument-dependent name lookup
The goal here is to exercise each rule in [basic.lookup.argdep] at least once.
These new tests expose what I believe are 2 issues:
1. CWG 1691 needs to be implemented (p2: [...] Its associated namespaces are
the innermost enclosing namespaces of its associated classes [...]) The
corresponding tests are adl_class_type::X2 and adl_class_type::X5.
2. The end of paragraph 2 ([...] Additionally, if the aforementioned set of
overloaded functions is named with a template-id, its associated classes
and namespaces also include those of its type template-arguments and its
template template-arguments.) is not implemented. Closely related, the
restriction on non-dependent parameter types in this same paragraph needs
to be removed. The corresponding tests are in adl_overload_set (both issues
are from CWG 997).
Differential Revision: https://reviews.llvm.org/D60570
Reviewed By: riccibruno, Quuxplusone
llvm-svn: 358881
Serguei Katkov [Mon, 22 Apr 2019 10:35:07 +0000 (10:35 +0000)]
[NewPM] Add Option handling for SimpleLoopUnswitch
This patch enables passing options to SimpleLoopUnswitch via the passes pipeline.
Reviewers: chandlerc, fedor.sergeev, leonardchan, philip.pfaffe
Reviewed By: fedor.sergeev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D60676
llvm-svn: 358880
Simon Pilgrim [Mon, 22 Apr 2019 10:19:09 +0000 (10:19 +0000)]
[AMDGPU] Regenerate uitofp i8 to float conversion tests.
Prep work for D60462
llvm-svn: 358879
Serguei Katkov [Mon, 22 Apr 2019 09:53:26 +0000 (09:53 +0000)]
[NewPM] Add dummy Test for LoopVectorize option parsing.
llvm-svn: 358878
Kristof Umann [Mon, 22 Apr 2019 09:20:23 +0000 (09:20 +0000)]
[analyzer][www] Moving MoveChecker out of alpha is no longer an open project.
llvm-svn: 358877
Nikita Popov [Mon, 22 Apr 2019 09:01:38 +0000 (09:01 +0000)]
Revert "[ConstantRange] Rename make{Guaranteed -> Exact}NoWrapRegion() NFC"
This reverts commit
7bf4d7c07f2fac862ef34c82ad0fef6513452445.
After thinking about this more, this isn't right, the range is not exact
in the same sense as makeExactICmpRegion(). This needs a separate
function.
llvm-svn: 358876
Nikita Popov [Mon, 22 Apr 2019 08:36:05 +0000 (08:36 +0000)]
[ConstantRange] Rename make{Guaranteed -> Exact}NoWrapRegion() NFC
Following D60632 makeGuaranteedNoWrapRegion() always returns an
exact nowrap region. Rename the function accordingly. This is in
line with the naming of makeExactICmpRegion().
llvm-svn: 358875
George Rimar [Mon, 22 Apr 2019 07:57:07 +0000 (07:57 +0000)]
[LLD][ELF] - Handle quoted strings in the linker scripts correctly.
This is the https://bugs.llvm.org/show_bug.cgi?id=41356,
Seems it is kind of unusual case but it is possible to
have sections that require quotes for their namings.
Like "aaa bbb".
This patch adds support for those.
Differential revision: https://reviews.llvm.org/D60901
llvm-svn: 358874
Aleksandr Urakov [Mon, 22 Apr 2019 07:14:40 +0000 (07:14 +0000)]
[NativePDB] Add anonymous namespaces support
Summary:
This patch adds anonymous namespaces support to the native PDB plugin.
I had to reference from the main function variables of the types that are inside
of the anonymous namespace to include them in debug info. Without the references
they are not included. I think it's because they are static, then are visible
only in the current translation unit, so they are not needed without any
references to them.
There is also the problem case with variables of types that are nested in
template structs. For now I've left FIXME in the test because this case is not
related to the change.
Reviewers: zturner, asmith, labath, stella.stamenova, amccarth
Reviewed By: amccarth
Subscribers: zloyrobot, aprantl, teemperor, lldb-commits, leonid.mashinskiy
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D60817
llvm-svn: 358873
Craig Topper [Mon, 22 Apr 2019 06:12:02 +0000 (06:12 +0000)]
[X86] Reject 512-bit types in getRegForInlineAsmConstraint when AVX512 is not enabled. Same for 256 bit and AVX.
llvm-svn: 358872
Sam Clegg [Mon, 22 Apr 2019 05:26:44 +0000 (05:26 +0000)]
[WebAssembly] Fix R_WASM_FUNCTION_OFFSET_I32 relocation warnings
We were incorrectly used the symbol table version of the function rather
than the object-local version when checking the existing relocation
value.
This was causing erroneous warnings for comat symbols defined in
multiple object.s
Fixes: https://bugs.llvm.org/show_bug.cgi?id=40503
Differential Revision: https://reviews.llvm.org/D60928
llvm-svn: 358871
Fangrui Song [Mon, 22 Apr 2019 03:10:40 +0000 (03:10 +0000)]
[ELF][X86] Allow R_386_TLS_LDO_32 and R_X86_64_DTPOFF{32,64} to preemptable local-dynamic symbols
Summary:
Fixes PR35242. A simplified reproduce:
thread_local int i; int f() { return i; }
% {g++,clang++} -fPIC -shared -ftls-model=local-dynamic -fuse-ld=lld a.cc
ld.lld: error: can't create dynamic relocation R_X86_64_DTPOFF32 against symbol: i in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output
In isStaticLinkTimeConstant(), Syn.IsPreemptible is true, so it is not
seen as a constant. The error is then issued in processRelocAux().
A symbol of the local-dynamic TLS model cannot be preempted but it can
preempt symbols of the global-dynamic TLS model in other DSOs.
So it makes some sense that the variable is not static.
This patch fixes the linking error by changing getRelExpr() on
R_386_TLS_LDO_32 and R_X86_64_DTPOFF{32,64} from R_ABS to R_DTPREL.
R_PPC64_DTPREL_* and R_MIPS_TLS_DTPREL_* need similar fixes, but they are not handled in this patch.
As a bonus, we use `if (Expr == R_ABS && !Config->Shared)` to find
ld-to-le opportunities. R_ABS is overloaded here for such STT_TLS symbols.
A dedicated R_DTPREL is clearer.
Differential Revision: https://reviews.llvm.org/D60945
llvm-svn: 358870
Lang Hames [Mon, 22 Apr 2019 03:03:09 +0000 (03:03 +0000)]
[JITLink] Remove a lot of reduntant 'JITLink_' prefixes. NFC.
llvm-svn: 358869
Fangrui Song [Mon, 22 Apr 2019 02:48:37 +0000 (02:48 +0000)]
[ELF][X86] Rename R_RELAX_TLS_GD_TO_IE_END to R_RELAX_TLS_GD_TO_IE_GOTPLT
Summary:
This relocation type is used by R_386_TLS_GD. Its formula is the same as
R_GOTPLT (e.g R_X86_64_GOT{32,64} R_386_TLS_GOTIE). Rename it to be clearer.
Differential Revision: https://reviews.llvm.org/D60941
llvm-svn: 358868
Fangrui Song [Mon, 22 Apr 2019 02:23:09 +0000 (02:23 +0000)]
[cmake] Add llvm-jit to LLVM_TEST_DEPENDS
Otherwise llvm-jit would say "utils/lit/lit/llvm/subst.py:127: note: Did not find llvm-jitlink in ..."
llvm-svn: 358867
Fangrui Song [Mon, 22 Apr 2019 01:38:53 +0000 (01:38 +0000)]
[clangd] Support dependent bases in type hierarchy
Patch by Nathan Ridge!
Dependent bases are handled heuristically, by replacing them with the
class template that they are a specialization of, where possible. Care
is taken to avoid infinite recursion.
Differential Revision: https://reviews.llvm.org/D59756
llvm-svn: 358866
Lang Hames [Mon, 22 Apr 2019 01:35:16 +0000 (01:35 +0000)]
[JITLink] Fix section start address calculation in eh-frame recorder.
Section atoms are not sorted, so we need to scan the whole section to find the
start address.
No test case: Found by inspection, and any reproduction would depend on pointer
ordering.
llvm-svn: 358865
Nico Weber [Sun, 21 Apr 2019 23:50:24 +0000 (23:50 +0000)]
Attemp get llvm-jitlink building on Windows
By removing an include of dlfcn.h that looks unused.
And clang-format a too-long line while here.
llvm-svn: 358864
Lang Hames [Sun, 21 Apr 2019 20:34:19 +0000 (20:34 +0000)]
[JITLink] Add an option to dump relocated section content.
The -dump-relocated-section-content option will dump the contents of each
section after relocations are applied, and before any checks are run or
code executed.
llvm-svn: 358863
Nico Weber [Sun, 21 Apr 2019 20:14:21 +0000 (20:14 +0000)]
gn build: Re-run `git ls-files '*.gn' '*.gni' | xargs llvm/utils/gn/gn.py format`
llvm-svn: 358862
Nico Weber [Sun, 21 Apr 2019 20:08:45 +0000 (20:08 +0000)]
gn build: Merge r358749
Since the symlinks list for llvm-symbolizer is now never empty,
the :symlinks target no longer needs an explicit dep on :llvm-symbolizer
-- there will be at least one dep on a symlink, and each symlink depends
on :llvm-symbolizer already.
Since llvm-symbolizer:symlinks now produces symlinks that check-llvm
uses, make llvm/test depend on the symlink target.
llvm-svn: 358861
Nico Weber [Sun, 21 Apr 2019 19:45:37 +0000 (19:45 +0000)]
gn build: Merge r358818 (JITLink)
llvm-svn: 358860
Don Hinton [Sun, 21 Apr 2019 19:20:13 +0000 (19:20 +0000)]
[cmake] Fix bug in r358779 - [CMake] Pass monorepo build settings in cross compile
Escape semicolons in the targets list so that cmake doesn't expand
them to spaces.
llvm-svn: 358859
Nico Weber [Sun, 21 Apr 2019 19:18:41 +0000 (19:18 +0000)]
Attempt to fix LLVM_ENABLE_THREADS=OFF build after r358665
llvm-svn: 358858
Nico Weber [Sun, 21 Apr 2019 17:19:27 +0000 (17:19 +0000)]
llvm-undname: Fix hex escapes in wchar_t, char16_t, char32_t strings
llvm-undname used to put '\x' in front of every pair of nibbles, but
u"\xD7\xFF" produces a string with 6 bytes: \xD7 \0 \xFF \0 (and \0\0). Correct
for a single character (plus terminating \0) is u\xD7FF instead.
Now, wchar_t, char16_t, and char32_t strings roundtrip from source to
clang-cl (and cl.exe) and then llvm-undname.
(...at least as long as it's not a string like L"\xD7FF" L"foo" which
gets demangled as L"\xD7FFfoo", where the compiler then considers the
"f" as part of the hex escape. That seems ok.)
Also add a comment saying that the "almost-valid" char32_t string I
added in my last commit is actually produced by compilers.
llvm-svn: 358857
Nico Weber [Sun, 21 Apr 2019 16:58:25 +0000 (16:58 +0000)]
llvm-undname: Fix stack overflow on almost-valid
If a unsigned with all 4 bytes non-0 was passed to outputHex(), there
were two off-by-ones in it:
- Both MaxPos and Pos left space for the final \0, which left the buffer
one byte to small. Set MaxPos to 16 instead of 15 to fix.
- The `assert(Pos >= 0);` was after a `Pos--`, move it up one line.
Since valid Unicode codepoints are <= 0x10ffff, this could never really
happen in practice.
Found by oss-fuzz.
llvm-svn: 358856
Nikita Popov [Sun, 21 Apr 2019 15:23:05 +0000 (15:23 +0000)]
[ConstantRange] Add saturating add/sub methods
Add support for uadd_sat and friends to ConstantRange, so we can
handle uadd.sat and friends in LVI. The implementation is forwarding
to the corresponding APInt methods with appropriate bounds.
One thing worth pointing out here is that the handling of wrapping
ranges is not maximally accurate. A simple example is that adding 0
to a wrapped range will return a full range, rather than the original
wrapped range. The tests also only check that the non-wrapping
envelope is correct and minimal.
Differential Revision: https://reviews.llvm.org/D60946
llvm-svn: 358855
Nikita Popov [Sun, 21 Apr 2019 15:22:54 +0000 (15:22 +0000)]
[ConstantRange] Add getNonEmpty() constructor
ConstantRanges have an annoying special case: If upper and lower are
the same, it can be either an empty or a full set. When constructing
constant ranges nearly always a full set is intended, but this still
requires an explicit check in many places.
This revision adds a getNonEmpty() constructor that disambiguates this
case: If upper and lower are the same, a full set is created.
Differential Revision: https://reviews.llvm.org/D60947
llvm-svn: 358854
Sanjay Patel [Sun, 21 Apr 2019 15:01:19 +0000 (15:01 +0000)]
[AArch64] add tests with multiple binop+splat vals; NFC
See D60890 for context.
llvm-svn: 358853
Nico Weber [Sun, 21 Apr 2019 14:25:07 +0000 (14:25 +0000)]
llvm-undname: Fix stack overflow on invalid found by oss-fuzz
llvm-svn: 358852
Nico Weber [Sun, 21 Apr 2019 14:07:13 +0000 (14:07 +0000)]
gn build: Fix build after r358837
llvm-svn: 358851
Pavel Labath [Sun, 21 Apr 2019 13:12:40 +0000 (13:12 +0000)]
Minidump: yamlify module-related unit tests
The tests reading the untouched module list are now not using any lldb
code (as module list loading lives in llvm now), so they can be removed.
The "filtering" of the module list remains (and probably will remain) an
lldb concept, so I keep those tests, but replace the checked-in binaries
with their yaml equivalents.
The binaries which are no longer referenced by any tests have been
removed.
llvm-svn: 358850
Bruno Ricci [Sun, 21 Apr 2019 13:12:10 +0000 (13:12 +0000)]
[Sema][MSVC] Fix bogus microsoft-pure-definition warning on member function of class template
Clang emits a warning when using a pure specifier =0 in a function definition
at class scope (a MS-specific construct), when using -fms-extensions.
However, to detect this, it was using FD->isCanonicalDecl() on function
declaration, which was also detecting out-of-class definition of member
functions of class templates. Fix this by using !FD->isOutOfLine() instead.
Fixes PR21334.
Differential Revision: https://reviews.llvm.org/D29707
Reviewed By: riccibruno
Reviewers: rnk, riccibruno
Patch By: Rudy Pons
llvm-svn: 358849
Pavel Labath [Sun, 21 Apr 2019 13:02:49 +0000 (13:02 +0000)]
@skipIfLinux flaky lldb-mi tests
llvm-svn: 358848
Pavel Labath [Sun, 21 Apr 2019 13:02:41 +0000 (13:02 +0000)]
Make TestVSCode_step pass reliably
Summary:
The test was failing occasionally (1% of runs or so), because of
unpredictable timings between the two threads spawned by the test. If
the second thread hit the breakpoint right as we were stepping out of
the function on the first thread, we would still be stuck at the inner
frame when the process stopped.
This would cause errors like:
File "/home/worker/lldb-x86_64-debian/lldb-x86_64-debian/llvm/tools/lldb/packages/Python/lldbsuite/test/tools/lldb-vscode/step/TestVSCode_step.py", line 67, in test_step
self.assertEqual(x1, x3, 'verify step out variable')
AssertionError: 2 != 1 : verify step out variable
AFAICT, lldb-vscode is doing the right thing here, and the problem is
that the test is not taking this sequence of events into account. Since
the test is about testing stepping, it does not seem necessary to have
threads in the inferior at all, so I just rewrite the test to execute
the code we're supposed to step through directly on the main thread.
Reviewers: clayborg, jgorbe
Subscribers: jfb, lldb-commits
Differential Revision: https://reviews.llvm.org/D60608
llvm-svn: 358847
Pavel Labath [Sun, 21 Apr 2019 12:48:53 +0000 (12:48 +0000)]
modify-python-lldb.py: Remove docstring formatting code
The strings have been already cleaned up in r358683, so this code is not
doing anything anymore.
While comparing the outputs before and after removing the formatting
code, I've found a couple of docstrings that managed to escape my perl
script in r358683, so I format them manually with this patch.
llvm-svn: 358846
David Green [Sun, 21 Apr 2019 09:54:29 +0000 (09:54 +0000)]
[ARM] Rewrite isLegalT2AddressImmediate
This does two main things, firstly adding some at least basic addressing modes
for i64 types, and secondly treats floats and doubles sensibly when there is no
fpu. The floating point change can help codesize in some cases, especially with
D60294.
Most backends seems to not consider the exact VT in isLegalAddressingMode,
instead switching on type size. That is now what this does when the target does
not have an fpu (as the float data will be loaded using LDR's). i64's currently
use the address range of an LDRD (even though they may be legalised and loaded
with an LDR). This is at least better than marking them all as illegal
addressing modes.
I have not attempted to do much with vectors yet. That will need changing once
MVE is added.
Differential Revision: https://reviews.llvm.org/D60677
llvm-svn: 358845
Craig Topper [Sun, 21 Apr 2019 07:12:56 +0000 (07:12 +0000)]
[X86] Add the rounding control operand to the printing for some scalar FMA instructions.
llvm-svn: 358844
Fangrui Song [Sun, 21 Apr 2019 06:17:40 +0000 (06:17 +0000)]
[CachePruning] Simplify comparator
llvm-svn: 358843
Fangrui Song [Sun, 21 Apr 2019 06:12:00 +0000 (06:12 +0000)]
[JITLink] Add dependency on MCParser to unit test after rL358818
This is required by -DBUILD_SHARED_LIBS=on builds for createMCAsmParser.
llvm-svn: 358842
Craig Topper [Sun, 21 Apr 2019 05:18:04 +0000 (05:18 +0000)]
[X86] Don't form masked vfpclass instruction from and+vfpclass unless the fpclass only has a single use.
llvm-svn: 358841
Lang Hames [Sun, 21 Apr 2019 04:48:32 +0000 (04:48 +0000)]
[JITLink] Remove an overly strict error check in JITLink's eh-frame parser.
The error check required FDEs to refer to the most recent CIE, but the eh-frame
spec allows them to refer to any previously seen CIE. This patch removes the
offending check.
llvm-svn: 358840
Lang Hames [Sun, 21 Apr 2019 03:14:43 +0000 (03:14 +0000)]
[BinaryFormat] Fix bitfield-ordering of MachO::relocation_info on big-endian.
Hopefully this will fix the JITLink regression test failures on big-endian
testers (e.g.
http://lab.llvm.org:8011/builders/clang-s390x-linux-lnt/builds/12702)
llvm-svn: 358839
Lang Hames [Sun, 21 Apr 2019 03:14:42 +0000 (03:14 +0000)]
[JITLink] Factor basic common GOT and stub creation code into its own class.
llvm-svn: 358838
Petr Hosek [Sun, 21 Apr 2019 01:09:15 +0000 (01:09 +0000)]
[gn] Move Features.inc to clangd, create a config for it
ClangdLSPServer and clangd unittests now include Features.inc so we
need to append the target_gen_dir that contains it to their
include_dirs. To do so, we use a public config that's applied to
any target that depends on the features one.
Differential Revision: https://reviews.llvm.org/D60919
llvm-svn: 358837
Lang Hames [Sun, 21 Apr 2019 00:35:58 +0000 (00:35 +0000)]
[JITLink] Add dependencies on MCDissassembler and Target to unit test.
llvm-svn: 358836
Nico Weber [Sat, 20 Apr 2019 23:59:06 +0000 (23:59 +0000)]
llvm-undname: Improve string literal demangling with embedded \0 chars
- Don't assert when a string looks like a u32 string to the heuristic
but doesn't have a length that's 0 mod 4. Instead, classify those
as u16 with embedded \0 chars. Found by oss-fuzz.
- Print embedded nul bytes as \0 instead of \x00.
llvm-svn: 358835
Nico Weber [Sat, 20 Apr 2019 23:22:45 +0000 (23:22 +0000)]
ftime-trace: Trace the name of the currently active pass as well.
Differential Revision: https://reviews.llvm.org/D60782
llvm-svn: 358834
Lang Hames [Sat, 20 Apr 2019 22:59:43 +0000 (22:59 +0000)]
[JITLink] Add yet more detail to MachO/x86-64 unsupported relocation errors.
Knowing the address/symbolnum field values makes it easier to identify the
unsupported relocation, and provides enough information for the full bit
pattern of the relocation to be reconstructed.
llvm-svn: 358833
Lang Hames [Sat, 20 Apr 2019 22:15:57 +0000 (22:15 +0000)]
[JITLink][ORC] Add JITLink to the list of dependencies for ORC.
The new ObjectLinkingLayer in ORC depends on JITLink.
This should fix the build error at
http://lab.llvm.org:8011/builders/clang-ppc64le-linux-multistage/builds/9621
llvm-svn: 358832
Lang Hames [Sat, 20 Apr 2019 22:06:12 +0000 (22:06 +0000)]
[JITLink] Fix a bad formatv format string.
llvm-svn: 358831
Lang Hames [Sat, 20 Apr 2019 21:32:49 +0000 (21:32 +0000)]
[JITLink] Disable MachO/x86-64 regression test if the X86 target is not built.
llvm-svn: 358830
Amara Emerson [Sat, 20 Apr 2019 21:25:00 +0000 (21:25 +0000)]
Revert r358800. Breaks Obsequi from the test suite.
The last attempt fixed gcc and consumer-typeset, but Obsequi seems to fail with
a different issue.
llvm-svn: 358829
Lang Hames [Sat, 20 Apr 2019 20:05:30 +0000 (20:05 +0000)]
[JITLink] Add llvm-jitlink to the list of available tools in lit.
Should fix the 'llvm-jitlink command not found' errors that are appearing on
some builders.
llvm-svn: 358828
Lang Hames [Sat, 20 Apr 2019 19:48:45 +0000 (19:48 +0000)]
[JITLink] Add BinaryFormat to JITLink's dependencies.
Hopefully this will fix the missing dependence on llvm::identify_magic that is
showing up on some PPC bots. E.g.
http://lab.llvm.org:8011/builders/clang-ppc64le-linux-multistage/builds/9617
llvm-svn: 358827
Lang Hames [Sat, 20 Apr 2019 18:50:13 +0000 (18:50 +0000)]
[JITLink] Add more detail to MachO/x86-64 "unsupported relocation" errors.
The extra information here will be helpful in diagnosing errors, like the
ones currently occuring on the PPC big-endian bots. :)
llvm-svn: 358826
Lang Hames [Sat, 20 Apr 2019 18:30:17 +0000 (18:30 +0000)]
[JITLink] Add check to JITLink unit test to bail out for unsupported targets.
This should prevent spurious JITLink unit test failures for builds that do not
support the target(s) required by the tests.
llvm-svn: 358825
Lang Hames [Sat, 20 Apr 2019 18:30:16 +0000 (18:30 +0000)]
[JITLink] Silence some MSVC implicit cast warnings.
llvm-svn: 358824
Lang Hames [Sat, 20 Apr 2019 17:58:29 +0000 (17:58 +0000)]
[JITLink] Add llvm-jitlink subdirectory to tools/LLVMBuild.txt
llvm-svn: 358823
Lang Hames [Sat, 20 Apr 2019 17:49:58 +0000 (17:49 +0000)]
[JITLink] Use memset instead of bzero.
llvm-svn: 358822
Lang Hames [Sat, 20 Apr 2019 17:37:09 +0000 (17:37 +0000)]
[JITLink] Silence a narrowing conversion warning.
llvm-svn: 358821
Lang Hames [Sat, 20 Apr 2019 17:35:28 +0000 (17:35 +0000)]
[JITLink] Update BuildingAJIT tutorials to account for API changes in r358818.
DynamicLibrarySearchGenerator::GetForCurrentProcess now takes a char (the global
prefix) rather than a DataLayout reference.
llvm-svn: 358820
Lang Hames [Sat, 20 Apr 2019 17:29:57 +0000 (17:29 +0000)]
[JITLink] Fix a missing header and bad prototype.
llvm-svn: 358819
Lang Hames [Sat, 20 Apr 2019 17:10:34 +0000 (17:10 +0000)]
Initial implementation of JITLink - A replacement for RuntimeDyld.
Summary:
JITLink is a jit-linker that performs the same high-level task as RuntimeDyld:
it parses relocatable object files and makes their contents runnable in a target
process.
JITLink aims to improve on RuntimeDyld in several ways:
(1) A clear design intended to maximize code-sharing while minimizing coupling.
RuntimeDyld has been developed in an ad-hoc fashion for a number of years and
this had led to intermingling of code for multiple architectures (e.g. in
RuntimeDyldELF::processRelocationRef) in a way that makes the code more
difficult to read, reason about, extend. JITLink is designed to isolate
format and architecture specific code, while still sharing generic code.
(2) Support for native code models.
RuntimeDyld required the use of large code models (where calls to external
functions are made indirectly via registers) for many of platforms due to its
restrictive model for stub generation (one "stub" per symbol). JITLink allows
arbitrary mutation of the atom graph, allowing both GOT and PLT atoms to be
added naturally.
(3) Native support for asynchronous linking.
JITLink uses asynchronous calls for symbol resolution and finalization: these
callbacks are passed a continuation function that they must call to complete the
linker's work. This allows for cleaner interoperation with the new concurrent
ORC JIT APIs, while still being easily implementable in synchronous style if
asynchrony is not needed.
To maximise sharing, the design has a hierarchy of common code:
(1) Generic atom-graph data structure and algorithms (e.g. dead stripping and
| memory allocation) that are intended to be shared by all architectures.
|
+ -- (2) Shared per-format code that utilizes (1), e.g. Generic MachO to
| atom-graph parsing.
|
+ -- (3) Architecture specific code that uses (1) and (2). E.g.
JITLinkerMachO_x86_64, which adds x86-64 specific relocation
support to (2) to build and patch up the atom graph.
To support asynchronous symbol resolution and finalization, the callbacks for
these operations take continuations as arguments:
using JITLinkAsyncLookupContinuation =
std::function<void(Expected<AsyncLookupResult> LR)>;
using JITLinkAsyncLookupFunction =
std::function<void(const DenseSet<StringRef> &Symbols,
JITLinkAsyncLookupContinuation LookupContinuation)>;
using FinalizeContinuation = std::function<void(Error)>;
virtual void finalizeAsync(FinalizeContinuation OnFinalize);
In addition to its headline features, JITLink also makes other improvements:
- Dead stripping support: symbols that are not used (e.g. redundant ODR
definitions) are discarded, and take up no memory in the target process
(In contrast, RuntimeDyld supported pointer equality for weak definitions,
but the redundant definitions stayed resident in memory).
- Improved exception handling support. JITLink provides a much more extensive
eh-frame parser than RuntimeDyld, and is able to correctly fix up many
eh-frame sections that RuntimeDyld currently (silently) fails on.
- More extensive validation and error handling throughout.
This initial patch supports linking MachO/x86-64 only. Work on support for
other architectures and formats will happen in-tree.
Differential Revision: https://reviews.llvm.org/D58704
llvm-svn: 358818
Craig Topper [Sat, 20 Apr 2019 15:26:44 +0000 (15:26 +0000)]
[X86] Disable argument copy elision for arguments passed via pointers
Summary:
If you pass two 1024 bit vectors in IR with AVX2 on Windows 64. Both vectors will be split in four 256 bit pieces. The four pieces of the first argument will be passed indirectly using 4 gprs. The second argument will get passed via pointers in memory.
The PartOffsets stored for the second argument are all in terms of its original 1024 bit size. So the PartOffsets for each piece are 32 bytes apart. So if we consider it for copy elision we'll only load an 8 byte pointer, but we'll move the address 32 bytes. The stack object size we create for the first part is probably wrong too.
This issue was encountered by ISPC. I'm working on getting a reduce test case, but wanted to go ahead and get feedback on the fix.
Reviewers: rnk
Reviewed By: rnk
Subscribers: dbabokin, llvm-commits, hiraditya
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D60801
llvm-svn: 358817
Luqman Aden [Sat, 20 Apr 2019 13:14:18 +0000 (13:14 +0000)]
[CorrelatedValuePropagation] Mark subs that we know not to wrap with nuw/nsw.
Summary:
Teach CorrelatedValuePropagation to also handle sub instructions in addition to add. Relatively simple since makeGuaranteedNoWrapRegion already understood sub instructions. Only subtle change is which range is passed as "Other" to that function, since sub isn't commutative.
Note that CorrelatedValuePropagation::processAddSub is still hidden behind a default-off flag as IndVarSimplify hasn't yet been fixed to strip the added nsw/nuw flags and causes a miscompile. (PR31181)
Reviewers: sanjoy, apilipenko, nikic
Reviewed By: nikic
Subscribers: hiraditya, jfb, jdoerfert, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D60036
llvm-svn: 358816
Fangrui Song [Sat, 20 Apr 2019 13:00:50 +0000 (13:00 +0000)]
[ExecutionDomainFix] Optimize a binary search insertion
llvm-svn: 358815
Fangrui Song [Sat, 20 Apr 2019 13:00:09 +0000 (13:00 +0000)]
[llvm-symbolizer] Fix section index at the end of a section
This is very minor issue. The returned section index is only used by
DWARFDebugLine as an llvm::upper_bound input and the use case shouldn't
cause any behavioral change.
llvm-svn: 358814
Nikita Popov [Sat, 20 Apr 2019 12:05:53 +0000 (12:05 +0000)]
[IndVarSimplify] Generate full checks for some LFTR tests; NFC
llvm-svn: 358813
Nikita Popov [Sat, 20 Apr 2019 12:05:43 +0000 (12:05 +0000)]
[IndVarSimplify] Add tests for PR31181; NFC
llvm-svn: 358812
Sam McCall [Sat, 20 Apr 2019 11:48:11 +0000 (11:48 +0000)]
[ADT] Avoid warning in bsearch testcase
llvm-svn: 358811