Summary:
If LAA expands a bound that is loop invariant, but not hoisted out
of the loop body, it used to use that value anyway, causing a
non-domination error, because the memcheck block is of course not
dominated by the scalar loop body. Detect this situation and expand
the SCEV expression instead.
Fixes PR31251
Reviewers: anemet
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D27397
llvm-svn: 288705
This changes the scalar non-intrinsic non-avx roundss/sd instruction
definitions not to read their destination register - allowing partial dependency
breaking.
This fixes PR31143.
Differential Revision: https://reviews.llvm.org/D27323
llvm-svn: 288703
Structure the definitions a bit more like the other classes.
The main change here is to split EXP with the done bit set
to a separate opcode, so we can set mayLoad = 1 so that it won't
be reordered before the other exp stores, since this has the special
constraint that if the done bit is set then this should be the last
exp in she shader.
Previously all exp instructions were inferred to have unmodeled
side effects.
llvm-svn: 288695
Summary:
Libc++ frequently has the need to parse more than just the builtin *test keywords* (`RUN`, `REQUIRES`, `XFAIL`, ect). For example libc++ currently needs a new keyword `MODULES-DEFINES: macro list...`. Instead of re-implementing the script parsing in libc++ this patch allows `parseIntegratedTestScript` to take custom parsers.
This patch introduces a new class `IntegratedTestKeywordParser` which implements the logic to parse/process a test keyword. Parsing of various keyword "kinds" are supported out of the box, including 'TAG', 'COMMAND', and 'LIST', which parse keywords such as `END.`, `RUN:` and `XFAIL:` respectively.
As an example after this change libc++ can implement the `MODULES-DEFINES` simply using:
```
mparser = IntegratedTestKeywordParser('MODULES-DEFINES:', ParserKind.LIST)
parseIntegratedTestScript(test, additional_parsers=[mparser])
macro_list = mparser.getValue()
```
Reviewers: ddunbar, modocache, rnk, danalbert, jroelofs
Subscribers: mgrang, llvm-commits, cfe-commits
Differential Revision: https://reviews.llvm.org/D27005
llvm-svn: 288694
Bring the sorting check back that I removed in r288655 but put it under
EXPENSIVE_CHECKS this time. Also document that this the check isn't
purely about having a sorted list but also about operator < having the
correct transitive behavior.
Apply the same to the other check in the file.
llvm-svn: 288693
It's useful to be able to disable visibility annotations entirely; for
example, if we're building libc++abi static to include in another library,
and we don't want any libc++abi functions getting exported out of that
library. This is a generalization of _LIBCXXABI_DISABLE_DLL_IMPORT_EXPORT.
Differential Revision: https://reviews.llvm.org/D26950
llvm-svn: 288692
LLVM build trees export the gtest library through a special export set. If you're building against a build tree you shouldn't need to re-add gtest, but if you're building against an installed LLVM you do.
llvm-svn: 288691
It's useful to be able to disable visibility annotations entirely; for
example, if we're building libc++ static to include in another library,
and we don't want any libc++ functions getting exported out of that
library. This is a generalization of _LIBCPP_DISABLE_DLL_IMPORT_EXPORT.
Differential Revision: https://reviews.llvm.org/D26934
llvm-svn: 288690
Summary:
Similar to r288685.
getExceptionSpec returned structure with pointers to temporarily object created
by computeImplicitExceptionSpec.
Reviewers: rsmith
Subscribers: aizatsky, cfe-commits
Differential Revision: https://reviews.llvm.org/D27422
llvm-svn: 288689
This diff
1. Adds a comment to ObjectFileELF.cpp about the current
approach to determining the OS.
2. Replaces the check in SymbolFileDWARF.cpp with a more robust one.
Test plan:
Built (on Linux) a test binary linked to a c++ shared library
which contains just an implementation of a function TestFunction,
the library (the binary itself) doesn't contain ELF notes
and EI_OSABI is set to System V.
Checked in lldb that now "p TestFunction()" works fine
(and doesn't work without this patch).
Differential revision: https://reviews.llvm.org/D27380
llvm-svn: 288687
so we can stop using DW_OP_bit_piece with the wrong semantics.
The entire back story can be found here:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20161114/405934.html
The gist is that in LLVM we've been misinterpreting DW_OP_bit_piece's
offset field to mean the offset into the source variable rather than
the offset into the location at the top the DWARF expression stack. In
order to be able to fix this in a subsequent patch, this patch
introduces a dedicated DW_OP_LLVM_fragment operation with the
semantics that we used to apply to DW_OP_bit_piece, which is what we
actually need while inside of LLVM. This patch is complete with a
bitcode upgrade for expressions using the old format. It does not yet
fix the DWARF backend to use DW_OP_bit_piece correctly.
Implementation note: We discussed several options for implementing
this, including reserving a dedicated field in DIExpression for the
fragment size and offset, but using an custom operator at the end of
the expression works just fine and is more efficient because we then
only pay for it when we need it.
Differential Revision: https://reviews.llvm.org/D27361
rdar://problem/29335809
llvm-svn: 288683
On MSVC, if an implicit instantiation already exists and an explicit
instantiation definition with a DLL attribute is created, the DLL
attribute still takes effect. Make clang match this behavior for
exporting.
Differential Revision: https://reviews.llvm.org/D26657
llvm-svn: 288682
Since we’re adding an entry into COMPILER_RT_SUPPORTED_ARCH for all architectures of all Darwin platforms, COMPILER_RT_SUPPORTED_ARCH often ends up having duplicate items. Let’s remove them.
Differential Revision: https://reviews.llvm.org/D25996
llvm-svn: 288681
On Linux (and probably on other Unix-like systems), unlink(2) is
noticeably slow. It takes 250 milliseconds to remove a 1 GB file
on ext4 filesystem on my machine, whether the file is on SSD or
on a spinning disk.
To create a new result file, we remove existing file first. So, if
you repeatedly link a 1 GB program in a regular compile-link-debug
cycle, every cycle wastes 250 milliseconds only to remove a file.
Since LLD can link a 1 GB in about 5 seconds, that waste actually
matters.
This patch defines `unlinkAsync` function. The function spawns a
background thread to call unlink. The calling thread returns
almost immediately.
Differential Revision: https://reviews.llvm.org/D27295
llvm-svn: 288680
This fix, while a bit complicated, preserves the reusability while fixing the issues reported on llvm-commits with visual studio generators.
llvm-svn: 288679
On a method call, the ObjCGenerics checker uses the type tracked by
DynamicTypePropagation for the receiver to to infer substituted parmeter types
for the called methods and warns when the argument type does not match the
parameter.
Unfortunately, using the tracked type can result in false positives when the
receiver has a non-invariant type parameter and has been intentionally upcast.
For example, becaue NSArray's type parameter is covaraint, the following code
is perfectly safe:
NSArray<NSString *> *allStrings = ...
NSDate *date = ...;
NSArray<NSObject *> *allObjects = allStrings;
NSArray<NSObject *> *moreObjects = [allObjects arrayByAddingObject:date];
but the checker currently warns that the date parameter is not an NSString *.
To avoid this kind of false positive, the checker will now only warn when
the class defining the called method has only invariant type parameters.
rdar://problem/28803951
llvm-svn: 288677
We treat bitwise 'not' as a special operation and try not to reduce its all-ones mask.
Presumably, this is because a 'not' may be cheaper than a generic 'xor' or it may get
folded into another logic op if the target has those. However, if we can remove a logic
instruction by changing the xor's constant mask value, that should always be a win.
Note that the IR version of SimplifyDemandedBits() does not treat 'not' as a special-case
currently (although that's marked with a FIXME). So if you run this IR through -instcombine,
you should get the same end result. I'm hoping to add a different backend transform that
will expose this problem though, so I need to solve this first.
Differential Revision: https://reviews.llvm.org/D27356
llvm-svn: 288676
TSan runtime shouldn't contain memset, so internal_memset is used
instead and syntax that emits memset is avoided.
This doesn't fail in-tree due to TSan being build with -03, but it fails
when TSan is built with -O0, and is (I think) a true positive.
Patch by Sam McCall, review: https://reviews.llvm.org/D27407
llvm-svn: 288672
The relocation R_AARCH64_LDST16_ABS_LO12_NC should set a ld/st
immediate value to bits [11:1] not [11:2]. This patches fixes it
and adds a testcase for regression.
With this fix all the faulty tests on test-suite (clavm, lencod,
and trimaran) pass.
llvm-svn: 288670
Doing so changes the evaluation order for relocation composition.
Patch By: Daniel Sanders
Reviewers: vkalintiris, atanasyan
Differential Revision: https://reviews.llvm.org/D26401
llvm-svn: 288666
Summary:
CXXDeleteExpr::getDestroyedType() can return a null QualType if the destroyed
type is a dependent type. This patch protects against this.
Reviewers: klimek
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D27350
llvm-svn: 288665
Rationale:
scripts/Python/modules: android is excluded at a higher level, so no point in
checking here
tools/lldb-mi: lldb-mi builds fine (with some cosmetic tweaks) on android, and
there is no reason it shouldn't.
tools/lldb-server: LLDB_DISABLE_LIBEDIT/CURSES already take the platform into
account, so there is no point in checking again.
I am reasonably confident this should not break the build on any platform, but
I'll keep an eye out on the bots.
llvm-svn: 288661
Under libcpp-no-exceptions, noexcept is trivially true. Some tests expect in
the usual setting to return false, so adjust them to expect true under
libcpp-no-exceptions.
Differential Revision: https://reviews.llvm.org/D27310
llvm-svn: 288660
This function seems target-independent so far: all the target-specific behaviour
is isolated in the CCAssignFn and the ValueHandler (which we're also extracting
into the generic CallLowering).
The intention is to use this in the ARM backend.
Differential Revision: https://reviews.llvm.org/D27045
llvm-svn: 288658
Rui Ueyama