Summary:
On MIPS32r6/MIPS64r6, floating point comparisons return 0 or -1 but integer
comparisons return 0 or 1.
Updated the various uses of getBooleanContents. Two simplifications had to be
disabled when float and int boolean contents differ:
- ScalarizeVecRes_VSELECT except when the kind of boolean contents is trivially
discoverable (i.e. when the condition of the VSELECT is a SETCC node).
- visitVSELECT (select C, 0, 1) -> (xor C, 1).
Come to think of it, this one could test for the common case of 'C'
being a SETCC too.
Preserved existing behaviour for all other targets and updated the affected
MIPS32r6/MIPS64r6 tests. This also fixes the pi benchmark where the 'low'
variable was counting in the wrong direction because it thought it could simply
add the result of the comparison.
Reviewers: hfinkel
Reviewed By: hfinkel
Subscribers: hfinkel, jholewinski, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D4389
llvm-svn: 212697
Without some mention of armv6m in a subdirectory of builtins, the make code
doesn't even know that armv6m exists and is something it should be looking for
in the platform-specific Makefiles. This means that none of the functions
listed actually get built and we end up with an almost entirely empty
libclang_rt.a for armv6m.
Unfortunately, the assembly code in the usual arm directory has no hope of
running on armv6m, which only supports Thumb-1 (not even ARM mode), so adding
it there won't work. Realistically, we probably *will* want to put any
optimised versions in a separate directory, so creating it now is harmless.
rdar://problem/17613576
llvm-svn: 212696
combine into half-shuffles through unpack instructions that expand the
half to a whole vector without messing with the dword lanes.
This fixes some redundant instructions in splat-like lowerings for
v16i8, which are now getting to be *really* nice.
llvm-svn: 212695
that splat i8s into i16s.
Previously, we would try much too hard to arrange a sequence of i8s in
one half of the input such that we could unpack them into i16s and
shuffle those into place. This isn't always going to be a cheaper i8
shuffle than our other strategies. The case where it is always going to
be cheaper is when we can arrange all the necessary inputs into one half
using just i16 shuffles. It happens that viewing the problem this way
also makes it much easier to produce an efficient set of shuffles to
move the inputs into one half and then unpack them.
With this, our splat code gets one step closer to being not terrible
with the new experimental lowering strategy. It also exposes two
combines missing which I will add next.
llvm-svn: 212692
isDereferenceablePointer should not give up upon encountering any bitcast. If
we're casting from a pointer to a larger type to a pointer to a small type, we
can continue by examining the bitcast's operand. This missing capability
was noted in a comment in the function.
In order for this to work, isDereferenceablePointer now takes an optional
DataLayout pointer (essentially all callers already had such a pointer
available). Most code uses isDereferenceablePointer though
isSafeToSpeculativelyExecute (which already took an optional DataLayout
pointer), and to enable the LICM test case, LICM needs to actually provide its DL
pointer to isSafeToSpeculativelyExecute (which it was not doing previously).
llvm-svn: 212686
This adds a utility method to access the WinCFI information in bulk and uses
that to iterate rather than requesting the count and individually iterating
them. This is in preparation for restructuring WinCFI handling to enable more
clear sharing across architectures to enable unwind information emission for
Windows on ARM.
llvm-svn: 212683
These fix the broken debian lldb build, which is using g++ 4.7.2.
TypeFormat changes:
1. stopped using the C++11 "dtor = default;" construct.
The generated default destructor in the two derived classes wanted
them to have a different throws() semantic that was causing 4.7 to
fail to generate it. I switched these to empty destructors defined
in the .cpp file.
2. Switched the m_types map from an ordered map to an unordered_map.
g++ 4.7's c++ library supports the C++11 emplace() used by TypeFormat
but the same c++ library's map impl does not. Since TypeFormat didn't
look like it depended on ordering in the map, I just switched it to
a std::unordered_map.
NativeProcessLinux - g++ 4.7 chokes on lexing the "<::" in
static_cast<::pid_t>(wpid). g++ 4.8+ and clang are fine with it.
I just put a space in between the "<" and the "::" and that cleared
it up.
llvm-svn: 212681
shuffles specifically for cases where a small subset of the elements in
the input vector are actually used.
This is specifically targetted at improving the shuffles generated for
trunc operations, but also helps out splat-like operations.
There is still some really low-hanging fruit here that I want to address
but this is a huge step in the right direction.
llvm-svn: 212680
This makes running libcxxabi tests on Linux _much_ easier.
Adds a check-libcxxabi target to cmake.
Also defaults to building a dynamic libc++abi. This is so that the
default options still test the libc++abi that is being built. There are
two problems with testing a static libc++abi. In the case of a
standalone build, the tests will link the system's libc++, which might
not have been built against our libc++abi. In the case of an in tree
build, libc++ will prefer a dynamic libc++abi from the system over a
static libc++abi from the output directory.
llvm-svn: 212672
debug sessions simultaneously to expose race conditoin/locking
issues.
This directory has an inferior program, testprog.cpp that has a
couple of functions we can put breakpoints on.
It has a driver program, multi-process-driver.cpp, which links
against the LLDB solib and uses the SB APIs. It creates 50 pthreads,
creates a debugger on all of them, launches a debug session of the
inferior testprog, hits a couple breakpoints, walks the stack,
continues, etc., and then kills the inferior and ends the debug
session.
A pass is if all fifty debug sessions complete successfully
in the alloted time (~60 seconds).
We may need to tweak this one to work correctly on different
platforms/targets but I wanted to get it checked in to start.
llvm-svn: 212671
Although this is nominally a -W option, we actually handle it in the driver
exactly as an f-group flag that's translated directly to -mllvm.
That means f_Group (and unintuitively, not W_Group) has the semantics we want
to make it behave like a standard warning flag: no automatic forwarding, no
warning for link invocations and compile-only.
Silences diagnostics like:
[691/1545] Linking CXX executable bin/llvm-diff
clang-3.5: warning: argument unused during compilation: '-Wframe-larger-than=2048'
(Hopefully we can move towards handling these in the frontend but that'll
require some infrastructure work.)
llvm-svn: 212670
The sret paramater consumes the register after the implicit 'this'
parameter, as with other calling conventions.
Fixes PR20278, which turned out to be very easy.
llvm-svn: 212669
Otherwise, it can be accidentally redefined when we build specific sanitizer
runtime. This definition should be provided only once - when we build
sanitizer_common library.
llvm-svn: 212663
Marked skipped for Linux:
TestCallStopAndContinue
TestConvenienceVariables
TestStopHookMultipleThreads
Fixed up gdb-remote port-grabbing code to use a random port in a wide range,
and to allow that to fail more gracefully. This appears to have solved some
gdb-remote intermittent failing behavior.
llvm-svn: 212662
don't need to set it manually.
This is based on feedback from Tom who pointed out that if every target
needs to handle this we need to reach out to those maintainers. In fact,
it doesn't make sense to duplicate everything when anything other than
expand seems unlikely at this stage.
llvm-svn: 212661
All tests matching '-p TestObjCMethods' now are marked correctly for MacOSX, and some
error classes have been removed in cleanup code looking for files that
might not exist due to previous failure.
See http://llvm.org/bugs/show_bug.cgi?id=20267
llvm-svn: 212650
Reverted by Eric Christopher (Thanks!) in r212203 after Bob Wilson
reported LTO issues. Duncan Exon Smith and Aditya Nandakumar helped
provide a reduced reproduction, though the failure wasn't too hard to
guess, and even easier with the example to confirm.
The assertion that the subprogram metadata associated with an
llvm::Function matches the scope data referenced by the DbgLocs on the
instructions in that function is not valid under LTO. In LTO, a C++
inline function might exist in multiple CUs and the subprogram metadata
nodes will refer to the same llvm::Function. In this case, depending on
the order of the CUs, the first intance of the subprogram metadata may
not be the one referenced by the instructions in that function and the
assertion will fail.
A test case (test/DebugInfo/cross-cu-linkonce-distinct.ll) is added, the
assertion removed and a comment added to explain this situation.
Original commit message:
If a function isn't actually in a CU's subprogram list in the debug info
metadata, ignore all the DebugLocs and don't try to build scopes, track
variables, etc.
While this is possibly a minor optimization, it's also a correctness fix
for an incoming patch that will add assertions to LexicalScopes and the
debug info verifier to ensure that all scope chains lead to debug info
for the current function.
Fix up a few test cases that had broken/incomplete debug info that could
violate this constraint.
Add a test case where this occurs by design (inlining a
debug-info-having function in an attribute nodebug function - we want
this to work because /if/ the nodebug function is then inlined into a
debug-info-having function, it should be fine (and will work fine - we
just stitch the scopes up as usual), but should the inlining not happen
we need to not assert fail either).
llvm-svn: 212649
Daniel Sanders