This is a follow-on to r247308.

llvm-svn: 247443

We used different conditions to determine if we should emit startproc vs
endproc.  Use the same condition to ensure that they will always be
paired.

This fixes PR24374.

llvm-svn: 247435

The rest of the EH pads are fine, since they have at most one label and
take fewer operands for the personality.

Old catchpad vs. new:
  %5 = catchpad [i8* bitcast (i32 ()* @"\01?filt$0@0@main@@" to i8*)] to label %__except.ret.10 unwind label %catchendblock.9
-----
  %5 = catchpad [i8* bitcast (i32 ()* @"\01?filt$0@0@main@@" to i8*)]
          to label %__except.ret.10 unwind label %catchendblock.9

llvm-svn: 247433

We used to have this magic "hasLoadLinkedStoreConditional()" callback,
which really meant two things:
- expand cmpxchg (to ll/sc).
- expand atomic loads using ll/sc (rather than cmpxchg).

Remove it, and, instead, introduce explicit callbacks:
- bool shouldExpandAtomicCmpXchgInIR(inst)
- AtomicExpansionKind shouldExpandAtomicLoadInIR(inst)

Differential Revision: http://reviews.llvm.org/D12557

llvm-svn: 247429

This lets us generalize its usage to the other atomic instructions.

llvm-svn: 247428

This brings a warning.

  cl : Command line warning D9035: option 'Og-' has been deprecated and will be removed in a future release

We should resolve PR11951 to remove this tweak.

llvm-svn: 247427

llvm-svn: 247425

llvm-svn: 247422

llvm-svn: 247420

It doesn't seem to like the '|&' in the test command.

llvm-svn: 247418

llvm-svn: 247417

llvm-svn: 247416

llvm-svn: 247415

These tests were found by llvm-mc-fuzzer (see http://reviews.llvm.org/D12723)
and were verified by checking the disassembler output is accepted by GAS.

llvm-svn: 247414

Source code was assuming that llvm-config.h would be included somehow but
up to r247253 that added #include "llvm/Support/Compiler.h" to StringRef.h
the config file was not actually included. The inclusion of llvm-config.h
caused a change of behaviour in tools/clang/test/Frontend/source-col-map.c:
previously it would output the original UTF-8 but now it outputs <U+03B1>.

llvm-svn: 247409

These tests were found by llvm-mc-fuzzer (see http://reviews.llvm.org/D12723)
and verified by checking the disassembler output is accepted by GAS.

The problematic tests from the previous commit have been moved to
valid-xfail.txt for now.

Also, give invalid instructions some coverage. invalid-xfail.txt contains
instructions that should be invalid but successfully disassemble.

llvm-svn: 247407

A small number of the added tests have operands that change on each round trip.

llvm-svn: 247406

These tests were found by llvm-mc-fuzzer (see http://reviews.llvm.org/D12723)
and verified by checking the disassembler output is accepted by GAS.

llvm-svn: 247405

It caused crash in MachineInstr::hasPropertyInBundle() since r247237.

llvm-svn: 247395

llvm-svn: 247391

When cloning the debug info for a function that hasn't been linked,
strip the DIEs from all location attributes that wouldn't contain any
meaningful information anyway.

This kind of situation can happen when a function got discarded by the
linker, but its debug information is still wanted in the final link
because it was marked as required as some other DIE dependency. The easiest
way to get into that situation is to have using directives. They get
linked unconditionally, but their targets might not always be present.

llvm-svn: 247386

lldb doesn't like having variables named as an existing type. In order to
ease debugging, rename those variables to avoid that conflict.

llvm-svn: 247385

llvm-svn: 247381

update.py:
import fileinput
import sys
import re

alias_match_prefix = r"(.*(?:=|:|^)\s*(?:external |)(?:(?:private|internal|linkonce|linkonce_odr|weak|weak_odr|common|appending|extern_weak|available_externally) )?(?:default |hidden |protected )?(?:dllimport |dllexport )?(?:unnamed_addr |)(?:thread_local(?:\([a-z]*\))? )?alias"
plain = re.compile(alias_match_prefix + r" (.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|addrspacecast|\[\[[a-zA-Z]|\{\{).*$)")
cast  = re.compile(alias_match_prefix + r") ((?:bitcast|inttoptr|addrspacecast)\s*\(.* to (.*?)(| addrspace\(\d+\) *)\*\)\s*(?:;.*)?$)")
gep   = re.compile(alias_match_prefix + r") ((?:getelementptr)\s*(?:inbounds)?\s*\((?P<type>.*), (?P=type)(?:\s*addrspace\(\d+\)\s*)?\* .*\)\s*(?:;.*)?$)")

def conv(line):
  m = re.match(cast, line)
  if m:
    return m.group(1) + " " + m.group(3) + ", " + m.group(2)
  m = re.match(gep, line)
  if m:
    return m.group(1) + " " + m.group(3) + ", " + m.group(2)
  m = re.match(plain, line)
  if m:
    return m.group(1) + ", " + m.group(2) + m.group(3) + "*" + m.group(4) + "\n"
  return line

for line in sys.stdin:
  sys.stdout.write(conv(line))

apply.sh:
for name in "$@"
do
  python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
  rm -f "$name.tmp"
done

The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh

llvm-svn: 247378

The former setup once resulted in us ignoring the module for C compilations,
but Clang now errors on this if the header is included from C code (which it is).

llvm-svn: 247377

llvm-svn: 247376

Fix embarrassing bugs I introduced to the `SlotTracker` in or around
r235785.  I had us iterating through every instruction in a function
(and hitting a map in the LLVMContext) for every basic block in the
function.

While there, completely avoid the call to
`SlotTracker::processFunctionMetadata()` from
`SlotTracker::processFunction()` if we've speculatively done this
already in `SlotTracker::processModule()` by checking
`ShouldInitializeAllMetadata` (this wasn't an algorithmic problem, but
it's touching the same line of code).

Fixes PR24699.

llvm-svn: 247372

Revert "[InstCombineCalls] Use isKnownNonNullAt() to check nullness of passing arguments at callsite"

This reverts commit r247356.

Breaks test/Transforms/InstCombine/pr8547.ll with:

Wrong types for attribute: byval inalloca nest noalias nocapture nonnull readnone readonly sret dereferenceable(1) dereferenceable_or_null(1)
  %call = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([10 x i8], [10 x i8]* @.str, i64 0, i64 0), i32 nonnull %conv2) #0
LLVM ERROR: Broken function found, compilation aborted!

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 247371

llvm-svn: 247364

llvm-svn: 247361

llvm-svn: 247359

When the driver tries to locate a program by its name, e.g. a linker, it
scans the paths provided by the toolchain using the ScanDirForExecutable
function. If the lookup fails, the driver uses
llvm::sys::findProgramByName. Unlike llvm::sys::findProgramByName,
ScanDirForExecutable is not aware of file extensions. If the program has
the "exe" extension in its name, which is very common on Windows,
ScanDirForExecutable won't find it under the toolchain-provided paths.

This patch changes the Windows version of the "`can_execute`" function
called by ScanDirForExecutable to respect file extensions, similarly to
llvm::sys::findProgramByName.

Patch by Oleg Ranevskyy

Reviewers: rnk

Differential Revision: http://reviews.llvm.org/D12711

llvm-svn: 247358

llvm-svn: 247357

Summary: This patch replaces isKnownNonNull() with isKnownNonNullAt() when checking nullness of passing arguments at callsite. In this way it can handle cases where the argument does not have nonnull attribute but has a dominating null check from the CFG.

Reviewers: reames

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D12779

llvm-svn: 247356

llvm-svn: 247354

Summary: This patch replaces isKnownNonNull() with isKnownNonNullAt() when checking nullness of gc.relocate return value. In this way it can handle cases where the relocated value does not have nonnull attribute but has a dominating null check from the CFG.

Reviewers: reames

Subscribers: llvm-commits, sanjoy

Differential Revision: http://reviews.llvm.org/D12772

llvm-svn: 247353

llvm-svn: 247352

The Win32 EH runtime caller does not preserve EBP, even though it does
preserve the CSRs (EBX, ESI, EDI) for us. The result was that each
finally funclet call would leave the frame pointer off by 12 bytes.

llvm-svn: 247348

llvm-svn: 247345

llvm-svn: 247344