While the memmove workaround fixed it for GCC 6.3. GCC 4.8 and GCC 7.1
are still broken. I have no clue what's going on, just blacklist GCC for
now.
Needless to say this code is ubsan, asan and msan-clean.
llvm-svn: 322862
We need to decompose relocation type for N32 / N64 ABI. Let's do it
before any other manipulations with relocation type in the `relocateOne`
routine.
llvm-svn: 322860
I've seen random crashes with GCC 4.8, GCC 6.3 and GCC 7.3, triggered by
my Optional change. All of them affect a different set of targets. This
change fixes the instance of the problem I'm seeing on my local machine,
let's hope it's good enough for the other instances too.
llvm-svn: 322859
Firstly, each offloading entry must have a unique name or the
linker will complain if there are multiple files with target
regions. Secondly, the compiler must not introduce padding so
mark the struct with a PackedAttr.
Differential Revision: https://reviews.llvm.org/D42168
llvm-svn: 322858
Summary:
This improves performance of code completion, because we avoid stating
the files from the preamble and never attempt to parse the files
without using the preamble if it's provided.
However, the change comes at a cost of sometimes providing incorrect
results when doing code completion after making actually considerable
changes to the files used in the preamble or the preamble itself.
Eventually the preamble will get rebuilt and code completion will
be providing the correct results.
Reviewers: bkramer, sammccall, jkorous-apple
Reviewed By: sammccall
Subscribers: klimek, cfe-commits
Differential Revision: https://reviews.llvm.org/D41991
llvm-svn: 322854
Summary:
The new method 'OverridePreamble' allows to override the preamble of
any source file without checking if preamble bounds or dependencies
were changed.
This is used for completion in clangd.
Reviewers: bkramer, sammccall
Reviewed By: sammccall
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D41990
llvm-svn: 322853
The goal is to have -polly-stmt-granularity=bb and
-polly-stmt-granularity=scalar-indep to have the same names if there is
just one statement per basic block.
This fixes a fluke when Polybench's jacobi-2d is optimized differently
depending on the -polly-stmt-granularity option, although both options
create the same SCoP, just with different statement names.
The new naming scheme is:
With -polly-use-llvm-names=0:
Stmt<BBIdx as decimal><Idx within BB as letter>
With -polly-use-llvm-names=1:
Stmt_BBName_<Idx within BB as letter>
The <Idx within BB> suffix is omitted for the main statement of a BB. The
main statement is either the one containing the first store or call
(those cannot be removed by the simplifyer), or if there is no such
instruction, the first. If after simplification there is just a single
statement left, it should be the main statement and have the same names as
with -polly-stmt-granularity=bb.
Differential Revision: https://reviews.llvm.org/D42136
llvm-svn: 322852
This will give control of the statement's name to the caller.
Required to give -polly-stmt-granularity=scalar-indep more control
over the name of the generated statement in a follow-up commit.
llvm-svn: 322851
This makes uses of Optional more transparent to the compiler (and
clang-tidy) and generates slightly smaller code.
This is a re-land of r317019, which had issues with GCC 4.8 back then.
Those issues don't reproduce anymore, but I'll watch the buildbots
closely in case anything goes wrong.
llvm-svn: 322838
Refactor in r322653 broke the logging test on windows because MSVC uses
a fully qualified name as the value of __FUNCTION__, which is not what
we were expecting. I'm changing the code to use the more portable
__func__, which behaves the same as on clang/gcc.
The standard still says that the value of __func__ is implementation
defined, so if we run into problems with this test again we may just
need to drop it.
llvm-svn: 322834
Follow-up to r322120 which can cause assertions for AArch64 because
v1f64 and v1i64 are legal types.
Differential Revision: https://reviews.llvm.org/D42097
llvm-svn: 322823
Summary:
A follow-up fix of rL311652.
The previous `vector` in our test is different with `std::vector`, so
The check still generates fixes for std::vector (`auto p =
std::unique_ptr<Foo>(new Foo({1,2,3}))`) in real world, the patch makes the
vector behavior in test align with std::vector (both AST nodes are the same now).
Reviewers: ilya-biryukov, alexfh
Reviewed By: ilya-biryukov
Subscribers: klimek, xazax.hun, cfe-commits
Differential Revision: https://reviews.llvm.org/D41852
llvm-svn: 322822
Summary:
This makes performance slower but more predictable (it always processes
every symbol). We need to find ways to make this fast, possibly by precomputing
short queries or capping the number of scored results. But our current approach
is too naive.
It also no longer returns results in a "good" order. In fact it's pathological:
the top N results are ranked from worst to best. Indexes aren't responsible for
ranking and MergedIndex can't do a good job, so I'm pleased that this will make
any hidden assumptions we have more noticeable :-)
Reviewers: hokein
Subscribers: klimek, ilya-biryukov, cfe-commits
Differential Revision: https://reviews.llvm.org/D42060
llvm-svn: 322821
LTO sets dso_local as an optimization, so don't clear it.
This avoid clearing it from undefined hidden symbols, which would then
fail the verifier.
llvm-svn: 322814
The dump function for Scope only has 20 out of the 24 flags. Since it looped
until no flags were left, having an unknown flag lead to an infinite loop.
That loop has been changed to a single pass for each flag, plus an assert to
alert if new flags are added.
llvm-svn: 322813
For example, a build_vector of i64 bitcasted from v2i32 can be turned into a concat_vectors of the v2i32 vectors with a bitcast to a vXi64 type
Differential Revision: https://reviews.llvm.org/D42090
llvm-svn: 322811
This patch fixes almost all currently failing tests when
using GCC ToT.
The specific changes are:
(A) Workaround gcc.gnu.org/PR83921 which rejects variables w/o initializers
in constexpr contexts -- even when the variable is an empty class. This
bug has been worked around at all callsites by adding an initializer.
Additionally a new test, constexpr_init.pass.cpp, has been added to
test that Clang doesn't suffer from these bugs.
(B) Fix streambuf.assign/swap.pass.cpp. This test was never actually
calling the swap method as intended. In fact, the swap function it
intended to call was ill-formed when instantiated. GCC diagnosed
this ill-formedness w/o needing an instantiation.
(C) size_delete11.pass.cpp was fixed by adding c++2a to the list of
unsupported dialects.
llvm-svn: 322810
Benjamin Kramer