FIXME: llvm/test may be aware of LLVM_PLUGIN_EXT, like as clang/test does.
FIXME: CMAKE_*_SUFFIX may be set in HandleLLVMOptions if those variables could be writable, rather than to set one as target properties.
llvm-svn: 201316
This changes ARM to use @llvm.fabs for floating-point vabs. Patterns
already existed in the backend, and it might help mid-end phases since
it's more likely to be understood than @llvm.arm.neon.vabs.
llvm-svn: 201313
Summary:
I'm not absolutely sure this is 100% correct solution, but it seems to
do what I expect.
Reviewers: djasper, klimek
Reviewed By: djasper
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D2756
llvm-svn: 201308
This commit is not strictly correct nor accounts for all uses (shared
objects, for example), but it allows one to test the compiler-rt library
on GNU targets.
Using this patch to run the test-suite has already shown me problems
on ARM. Since this is a Darwin-only flag, nobody is using it, so it
shouldn't be a problem.
I will need extension to deal with the shared cases, but since we're
not compiling libclang_rt.so, that's not yet applicable. Many other
problems will have to be fixed first in compiler-rt (such as removing
the 'arch' name from it and making it trully multi-arch, moving it to
the default lib directory, make both .a and .so variants, etc).
llvm-svn: 201307
aka asynchronous signal handlers, which subclasses should fill
in as appropriate. For most Unix user process environments,
the one entry in this list is _sigtramp. For bare-board and
kernel environments, there will be different sets of trap
handlers.
The unwinder needs to know when a frame is a trap handler
because the rules it enforces for the frame "above" the
trap handler is different from most middle-of-the-stack frames.
<rdar://problem/15835846>
llvm-svn: 201300
They were enforcing 16-byte alignment on stack frames for Darwin x86 programs.
But we've found that trap handlers typically don't have the stack pointer
aligned correctly when a trap happens and lldb wasn't backtracing all
the way through. This method is only used as a safety guard to prevent
lldb's unwinder from using a bogus address as a stack frame - we'll still
enforce word-size alignment on stack frames so that should be fine.
Also rolled back akaylor's changes from August 2013 in r188952 which changed
the i386 ABI plugin to relax the CallFrameAddressIsValid offsets for non-Darwin
targets where only 4-byte alignment is enforced. Now Darwin is the same as
those environments.
<rdar://problem/15982682>
llvm-svn: 201292
This fix checks the original LLVM IR node to identify opaque constants by
looking for the bitcast-constant pattern. Originally we looked at the generated
SDNode, but this might lead to incorrect results. The SDNode could have been
generated by an constant expression that was folded to a constant.
This fixes <rdar://problem/16050719>
llvm-svn: 201291
According to the GNU docs, zero-sized bitfields should not be affected by the
packed attribute.
Differential Revision: http://llvm-reviews.chandlerc.com/D2693
llvm-svn: 201288
As defined in LangRef, aliases do not have sections. However, LLVM's
GlobalAlias class inherits from GlobalValue, which means we can read and
set its section. We should probably ban that as a separate change,
since it doesn't make much sense for an alias to have a section that
differs from its aliasee.
Fixes PR18757, where the section was being lost on the global in code
from Clang like:
extern "C" {
__attribute__((used, section("CUSTOM"))) static int in_custom_section;
}
Reviewers: rafael.espindola
Differential Revision: http://llvm-reviews.chandlerc.com/D2758
llvm-svn: 201286
I was insightless then about unknown optional parameters.
(Consider that LINK_LIBS foo bar ADDITIONAL_HEADERS qux quux)
Suggested by Michael Kruse. Thanks!
llvm-svn: 201283
useBitFieldTypeAlignment() and appears to ignore the special
bit-packing semantics of __attribute__((packed)).
Further flesh out an already-extensive comment.
llvm-svn: 201282
logical operations on the i1's driving them. This is a bad idea for every
target I can think of (confirmed with micro tests on all of: x86-64, ARM,
AArch64, Mips, and PowerPC) because it forces the i1 to be materialized into
a general purpose register, whereas consuming it directly into a select generally
allows it to exist only transiently in a predicate or flags register.
Chandler ran a set of performance tests with this change, and reported no
measurable change on x86-64.
llvm-svn: 201275
These features are new in VS 2013 and are necessary in order to layout
std::ostream correctly. Currently we have an ABI incompatibility when
self-hosting with the 2013 stdlib in our convertible_fwd_ostream wrapper
in gtest.
This change adds another implicit attribute, MSVtorDispAttr, because
implicit attributes are currently the best way to make sure the
information stays on class templates through instantiation.
Reviewers: majnemer
Differential Revision: http://llvm-reviews.chandlerc.com/D2746
llvm-svn: 201274
'OK_NonUniformConstValue' to identify operands which are constants but
not constant splats.
The cost model now allows returning 'OK_NonUniformConstValue'
for non splat operands that are instances of ConstantVector or
ConstantDataVector.
With this change, targets are now able to compute different costs
for instructions with non-uniform constant operands.
For example, On X86 the cost of a vector shift may vary depending on whether
the second operand is a uniform or non-uniform constant.
This patch applies the following changes:
- The cost model computation now takes into account non-uniform constants;
- The cost of vector shift instructions has been improved in
X86TargetTransformInfo analysis pass;
- BBVectorize, SLPVectorizer and LoopVectorize now know how to distinguish
between non-uniform and uniform constant operands.
Added a new test to verify that the output of opt
'-cost-model -analyze' is valid in the following configurations: SSE2,
SSE4.1, AVX, AVX2.
llvm-svn: 201272
Instead of expanding a packed shift into a sequence of scalar shifts,
the backend now tries (when possible) to convert the vector shift into a
vector multiply.
Before this change, a shift of a MVT::v8i16 vector by a
build_vector of constants was always scalarized into a long sequence of "vector
extracts + scalar shifts + vector insert".
With this change, if there is SSE2 support, we emit a single vector multiply.
This change also affects SSE4.1, AVX, AVX2 shifts:
- A shift of a MVT::v4i32 vector by a build_vector of non uniform constants
is now lowered when possible into a single SSE4.1 vector multiply.
- Packed v16i16 shift left by constant build_vector are now expanded when
possible into a single AVX2 vpmullw.
This change also improves the lowering of AVX512f vector shifts.
Added test CodeGen/X86/vec_shift6.ll with some code examples that are affected
by this change.
llvm-svn: 201271
This does exactly the same thing as for Win32, except it passes -m64 to
the compiler and the files go in a different directory.
Differential Revision: http://llvm-reviews.chandlerc.com/D2749
llvm-svn: 201269
NAKAMURA Takumi