llvm-svn: 287891

No functionality change intended. Fix by clang-tidy's
performance-unnecessary-value-param check.

llvm-svn: 287890

llvm-svn: 287889

No functional change.

llvm-svn: 287888

Test passes consistently, at least on linux.

llvm-svn: 287887

Vectorize UINT_TO_FP v2i32 -> v2f64 instead of scalarization (albeit still on the SIMD unit).

The codegen matches that generated by legalization (and is in fact used by AVX for UINT_TO_FP v4i32 -> v4f64), but has to be done in the x86 backend to account for legalization via 4i32.

Differential Revision: https://reviews.llvm.org/D26938

llvm-svn: 287886

Passing consistently now.

llvm-svn: 287885

The test has been passing for a while now.

llvm-svn: 287884

It consistently passes for linux-clang-i386, and linux-gcc-x86_64.

llvm-svn: 287883

Use 512-bit instructions with subvector insertion/extraction like we do in a number of similar circumstances

llvm-svn: 287882

This test passes consistently on linux, so I am removing the overall XFAIL. If it
fails on your configuration, please put a targeted xfail instead (i'll add them
my self if I get any breakage emails).

llvm-svn: 287881

This refactors AdbClient interface in terms of std::chrono.

llvm-svn: 287880

This refactors the class implementations to use chrono. I'll follow this up with
a refactor of the class interface.

llvm-svn: 287879

[X86][AVX512DQVL] Add awareness of vcvtqq2ps and vcvtuqq2ps implicit zeroing of upper 64-bits of xmm result

llvm-svn: 287878

llvm-svn: 287877

llvm-svn: 287876

The bug arises during register allocation on i686 for
CMPXCHG8B instruction when base pointer is needed. CMPXCHG8B
needs 4 implicit registers (EAX, EBX, ECX, EDX) and a memory address,
plus ESI is reserved as the base pointer. With such constraints the only
way register allocator would do its job successfully is when the addressing
mode of the instruction requires only one register. If that is not the case
- we are emitting additional LEA instruction to compute the address.

It fixes PR28755.

Patch by Alexander Ivchenko <alexander.ivchenko@intel.com>

Differential Revision: https://reviews.llvm.org/D25088

llvm-svn: 287875

Move the definitions of three variables out of the switch.

Patch by Alexander Ivchenko <alexander.ivchenko@intel.com>

Differential Revision: https://reviews.llvm.org/D25192

llvm-svn: 287874

- It does not modify the input instruction
- Second operand of any address is always an Index Register,
  make sure we actually check for that, instead of a check for
  an immediate value

Patch by Alexander Ivchenko <alexander.ivchenko@intel.com>

Differential Revision: https://reviews.llvm.org/D24938

llvm-svn: 287873

(commit again after fixing the buildbot failures)
This adds various overloads of the following builtins to altivec.h:

    vec_neg
    vec_nabs
    vec_adde
    vec_addec
    vec_sube
    vec_subec
    vec_subc

Note that for vec_sub builtins on 32 bit integers, the semantics is similar to
what ISA describes for instructions like vsubecuq that work on quadwords: the
first operand is added to the one's complement of the second operand. (As
opposed to two's complement which I expected).

llvm-svn: 287872

llvm-svn: 287871

Replace the CVTTPD2DQ/CVTTPD2UDQ and CVTDQ2PD/CVTUDQ2PD opcodes with general versions.

This is an initial step towards similar FP_TO_SINT/FP_TO_UINT and SINT_TO_FP/UINT_TO_FP lowering to AVX512 CVTTPS2QQ/CVTTPS2UQQ and CVTQQ2PS/CVTUQQ2PS with illegal types.

Differential Revision: https://reviews.llvm.org/D27072

llvm-svn: 287870

    
The .ARM.exidx table has an entry for each function with the first entry
giving the start address of the function, the table is sorted in ascending
order of function address. Given a PC value, the unwinder will search the
table for the entry that contains the PC value.
    
If the table entry happens to be the last, the range of the addresses that
the final unwinding table describes will extend to the end of the address
space. To prevent an incorrect address outside the address range of the
program matching the last entry we follow ld.bfd's example and add a
sentinel EXIDX_CANTUNWIND entry at the end of the table. This gives the
final real table entry an upper bound.
    
In addition the llvm libunwind unwinder currently depends on the presence
of a sentinel entry (PR31091).

Differential revision: https://reviews.llvm.org/D26977

llvm-svn: 287869

I managed to confuse me with two reviews of the same thing and ended commiting the wrong one.

llvm-svn: 287868

the chrono library there uses long long as the underlying chrono type, but
defines int64_t as long (or the other way around, I am not sure). In any case,
this caused the implicit conversion to not trigger. This should address that.

Also fix up the relevant unit test.

llvm-svn: 287867

Skip tests that expect an exception be thrown.

Differential Revision: https://reviews.llvm.org/D26606

llvm-svn: 287866

Skip tests that expect an exception be thrown and/or disable
unreachable catch handlers.

Differential Revision: https://reviews.llvm.org/D26608

llvm-svn: 287865

Summary:
This replaces the usage of raw integers with duration classes in the gdb-remote
packet management functions. The values are still converted back to integers once
they go into the generic Communication class -- that I am leaving to a separate
change.

The changes are mostly straight-forward (*), the only tricky part was
representation of infinite timeouts.

Currently, we use UINT32_MAX to denote infinite timeout. This is not well suited
for duration classes, as they tend to do arithmetic on the values, and the
identity of the MAX value can easily get lost (e.g.
microseconds(seconds(UINT32_MAX)).count() != UINT32_MAX). We cannot use zero to
represent infinity (as Listener classes do) because we already use it to do
non-blocking polling reads. For this reason, I chose to have an explicit value
for infinity.

The way I achieved that is via llvm::Optional, and I think it reads quite
natural. Passing llvm::None as "timeout" means "no timeout", while passing zero
means "poll". The only tricky part is this breaks implicit conversions (seconds
are implicitly convertible to microseconds, but Optional<seconds> cannot be
easily converted into Optional<microseconds>). For this reason I added a special
class Timeout, inheriting from Optional, and enabling the necessary conversions
one would normally expect.

(*) The other tricky part was GDBRemoteCommunication::PopPacketFromQueue, which
was needlessly complicated. I've simplified it, but that one is only used in
non-stop mode, and so is untested.

Reviewers: clayborg, zturner, jingham

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D26971

llvm-svn: 287864

Summary:
* Add -dump_dels option to dump all declarations from old header. It
  will allow clang-move used as a frontend to get declarations from
  header. Further more, this will make debugging stuff easier. Currently only
  class/function types are supported.
* Refactoring code a little bit by creating a ClangMoveContext which
  holds all options for ClangMoveTool, which can simplify the code in
  some degree.

Reviewers: ioeric

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D27059

llvm-svn: 287863

r287386 added a \x13 character inside a string literal. Most likely this
was by mistake, so remove it.

Differential Revision: https://reviews.llvm.org/D26973

llvm-svn: 287862

Correct 0 byte core files uploaded by arc.

llvm-svn: 287861

llvm-svn: 287860

llvm-svn: 287859

Summary:
This patch changes the way ProcessElfCore.cpp handles signal information.
The patch changes ProcessElfCore.cpp to use the signal from si_signo in SIGINFO notes in preference to the value of cursig in PRSTATUS notes. The value from SIGINFO seems to be more thread specific. The value from PRSTATUS is usually the same for all threads even if only one thread received a signal.
If it cannot find any SIGINFO blocks it reverts to the old behaviour and uses the value from cursig in PRSTATUS. If after that no thread appears to have been stopped it forces the status of the first thread to be SIGSTOP to prevent lldb hanging waiting for any thread from the core file to change state.

The order is:
- If one or more threads have a non-zero si_signo in SIGINFO that will be used.
- If no threads had a SIGINFO block with a non-zero si_signo set all threads signals to the value in cursig in their PRSTATUS notes.
- If no thread has a signal set to a non-zero value set the signal for only the first thread to SIGSTOP.


This resolves two issues. The first was identified in bug 26322, the second became apparent while investigating this problem and looking at the signal values reported for each thread via “thread list”.

Firstly lldb is able to load core dumps generated by gcore where each thread has a SIGINFO note containing a signal number but cursig in the PRSTATUS block for each thread is 0.

Secondly if a SIGINFO note was found the “thread list” command will no longer show the same signal number for all threads. At the moment if a process crashes, for example with SIGILL, all threads will show “stop reason = signal SIGILL”. With this patch only the thread that executed the illegal instruction shows that stop reason. The other threads show “stop reason = signal 0”.

Reviewers: jingham, clayborg

Subscribers: sas, labath, lldb-commits

Differential Revision: https://reviews.llvm.org/D26676

llvm-svn: 287858

Summary: The default constructor for a StringRef stores an empty string.

Reviewers: beanz, zturner

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D27067

llvm-svn: 287857

Patch by Daniel Cederman.

Reviewers: stoklund, arsenm

Subscribers: arsenm, llvm-commits

Differential Revision: https://reviews.llvm.org/D27046

llvm-svn: 287856

llvm-svn: 287854

llvm-svn: 287853

Differential Revision: https://reviews.llvm.org/D27006

llvm-svn: 287852

Differential Revision: https://reviews.llvm.org/D26827

llvm-svn: 287851