[CMake] Introduce LLVM_TARGET_TRIPLE_ENV as an option to override LLVM_DEFAULT_TARGET_TRIPLE at runtime.

No behavior is changed if LLVM_TARGET_TRIPLE_ENV is blank or undefined.

If LLVM_TARGET_TRIPLE_ENV is "TEST_TARGET_TRIPLE" and $TEST_TARGET_TRIPLE is not blank,
llvm::sys::getDefaultTargetTriple() returns $TEST_TARGET_TRIPLE.
Lit resets config.target_triple and config.environment[LLVM_TARGET_TRIPLE_ENV] to change the default target.

Without changing LLVM_DEFAULT_TARGET_TRIPLE nor rebuilding, lit can be run;

  TEST_TARGET_TRIPLE=i686-pc-win32 bin/llvm-lit -sv path/to/test/
  TEST_TARGET_TRIPLE=i686-pc-win32 ninja check-clang-tools

Differential Revision: https://reviews.llvm.org/D33662

llvm-svn: 305632

llvm-svn: 305630

llvm-svn: 305629

llvm-svn: 305628

llvm-svn: 305627

llvm-svn: 305626

Re-apply r276044/r279124/r305516. Fixed a problem where we would refuse
to place spills as the very first instruciton of a basic block and thus
artifically increase pressure (test in
test/CodeGen/PowerPC/scavenging.mir:spill_at_begin)

This is a variant of scavengeRegister() that works for
enterBasicBlockEnd()/backward(). The benefit of the backward mode is
that it is not affected by incomplete kill flags.

This patch also changes
PrologEpilogInserter::doScavengeFrameVirtualRegs() to use the register
scavenger in backwards mode.

Differential Revision: http://reviews.llvm.org/D21885

llvm-svn: 305625

Summary:
This is my misunderstanding on isBarrier. It's not for memory barriers,
but for other control flow purposes. lwsync doesn't have it either.

This fixes a simple crash with -verify-machineinstrs like below:

  define void @Foo() {
  entry:
    %tmp = load atomic i64, i64* undef acquire, align 8
    unreachable
  }

I deliberately don't want to check in the test, since there is little
chance to regress on such a mistake. Such a test adds noise to the code
base.

I plan to check in first, since it fixes a crash, and the fix is obvious.

Reviewers: kbarton, echristo

Subscribers: sanjoy, nemanjai, hiraditya, llvm-commits

Differential Revision: https://reviews.llvm.org/D34314

llvm-svn: 305624

The analysis is expected to be preserved by SelectionDAG.

llvm-svn: 305621

Fixes PR33453.

Differential Revision:  https://reviews.llvm.org/D34303

llvm-svn: 305618

This was a regression introduced in a previous patch.  Adding
back the code that handles this case.

llvm-svn: 305617

This ensures that symbolic relocations are generated for stack
pointer manipulations.

These relocations are of type R_WEBASSEMBLY_GLOBAL_INDEX_LEB.
This change also adds support for reading relocations of this
type in WasmObjectFile.cpp.

Since its a globally imported symbol this does mean that
the get_global/set_global instruction won't be valid until
the objects are linked that global used in no longer an
imported global.

Differential Revision: https://reviews.llvm.org/D34172

llvm-svn: 305616

Suppose we had a type index offsets array with a boundary at type index
N. Then you request the name of the type with index N+1, and that name
requires the name of index N-1 (think a parameter list, for example). We
didn't handle this, and we would print something like (<unknown UDT>,
<unknown UDT>).

The fix for this is not entirely trivial, and speaks to a larger
problem. I think we need to kill TypeDatabase, or at the very least kill
TypeDatabaseVisitor. We need a thing that doesn't do any caching
whatsoever, just given a type index it can compute the type name "the
slow way". The reason for the bug is that we don't have anything like
that. Everything goes through the type database, and if we've visited a
record, then we're "done". It doesn't know how to do the expensive thing
of re-visiting dependent records if they've not yet been visited.

What I've done here is more or less copied the code (albeit greatly
simplified) from TypeDatabaseVisitor, but wrapped it in an interface
that just returns a std::string. The logic of caching the name is now in
LazyRandomTypeCollection. Eventually I'd like to move the record
database here as well and the visited record bitfield here as well, at
which point we can actually just delete TypeDatabase. I don't see any
reason for it if a "sequential" collection is just a special case of a
random access collection with an empty partial offsets array.

Differential Revision: https://reviews.llvm.org/D34297

llvm-svn: 305612

I'm trying to get rid of the TypeDatabase class, so the first
step is to minimize its footprint.

llvm-svn: 305611

Previously only the error codes were reported which
meant that useful information about malformed inputs
was not shown.

Differential Revision: https://reviews.llvm.org/D34008

llvm-svn: 305609

Davide Italiano reported the following issue if llvm
is compiled with gcc -Wstrict-aliasing -Werror:
.....
lib/Target/BPF/CMakeFiles/LLVMBPFCodeGen.dir/BPFISelDAGToDAG.cpp.o
../lib/Target/BPF/BPFISelDAGToDAG.cpp: In member function ‘virtual
void {anonymous}::BPFDAGToDAGISel::PreprocessISelDAG()’:
../lib/Target/BPF/BPFISelDAGToDAG.cpp:264:26: warning: dereferencing
type-punned pointer will break strict-aliasing rules
[-Wstrict-aliasing]
       val = *(uint16_t *)new_val;
.....

The error is caused by my previous commit (revision 305560).

This patch fixed the issue by introducing an union to avoid
type casting.

Signed-off-by: Yonghong Song <yhs@fb.com>
llvm-svn: 305608

Summary: As far as I can tell we should be able to implement these almost the same way we do unsigned, but using signed comparisons and checks for min signed value instead of min unsigned value.

Reviewers: pete, davide, sanjoy

Reviewed By: davide

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D33815

llvm-svn: 305607

llvm-svn: 305606

llvm-svn: 305605

[TargetLowering] Use ConstantSDNode::isOne and getSExtValue instead of getting the underlying APInt first. NFC

llvm-svn: 305604

llvm-svn: 305603

llvm-svn: 305602

For the following motivating example
  bool c();
  void f();
  bool start() {
    bool result = c();
    if (!c()) {
      result = false;
      goto exit;
    }
    f();
    result = true;
  exit:
    return result;
  }

we would previously generate a single DW_AT_const_value(1) because
only the DBG_VALUE in the second-to-last basic block survived
codegen. This patch improves the heuristic used to determine when a
DBG_VALUE is available at the beginning of its variable's enclosing
lexical scope:

- Stop giving singular constants blanket permission to take over the
  entire scope. There is still a special case for constants in the
  function prologue that we also miight want to retire later.

- Use the lexical scope information to determine available-at-entry
  instead of proximity to the function prologue.

After this patch we generate a location list with a more accurate
narrower availability for the constant true value. As a pleasant side
effect, we also generate inline locations instead of location lists
where a loacation covers the entire range of the enclosing lexical
scope.

Measured on compiling llc with four targets this doesn't have an
effect on compile time and reduces the size of the debug info for llc
by ~600K.

rdar://problem/30286912

llvm-svn: 305599

llvm-svn: 305598

The verifier should not output any message in such a case.
Added test case with no .apple_name section in the file to verify new functionality.
Made existing test case more specific.

llvm-svn: 305597

llvm-svn: 305596

Found by -Wpessimizing-move.

llvm-svn: 305595

In this patch, I flip the switch in DriverUtils from using the external
cvtres.exe tool to using the Windows Resource library in llvm.

I also fixed a bug where .rsrc sections were marked as discardable
memory and therefore were placed in the wrong order in the final PE.

Furthermore, I modified WindowsResource to write the coff directly to a
memory buffer instead of to file, also had it use the machine types
already declared in COFF.h instead creating my own enum.

Finally, I flipped the switch to allow all unit tests that had
previously run only on windows due to a winres dependency to run
cross-platform.

Reviewers: zturner, ruiu

Subscribers: llvm-commits, hiraditya

Differential Revision: https://reviews.llvm.org/D34265

llvm-svn: 305592

Summary:
When we fold vector constants that are operands of phi's that feed into select,
we need to set the correct insertion point for the *new* selects that get generated.
The correct insertion point is the incoming block for the phi.
Such cases can occur with patch r298845, which fixed folding of
vector constants, but the new selects could be inserted incorrectly (as the added
test case shows).

Reviewers: majnemer, spatel, sanjoy

Reviewed by: spatel

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D34162

llvm-svn: 305591

This is a workaround for an ODR conflict with the definition in
AMDGPUCodeObjectMetadata.cpp.

llvm-svn: 305584

llvm-svn: 305583

llvm-svn: 305579

The recommit fixes two bugs: The first one is to use CurrentBlock instead of
PREInstr's Parent as param of performScalarPREInsertion because the Parent
of a clone instruction may be uninitialized. The second one is stop PRE when
CurrentBlock to its predecessor is a backedge and an operand of CurInst is
defined inside of CurrentBlock. The same value defined inside of loop in last
iteration can not be regarded as available.

Right now scalarpre doesn't have phi-translate support, so it will miss some
simple pre opportunities. Like the following testcase, current scalarpre cannot
recognize the last "a * b" is fully redundent because a and b used by the last
"a * b" expr are both defined by phis.

long a[100], b[100], g1, g2, g3;
__attribute__((pure)) long goo();

void foo(long a, long b, long c, long d) {

  g1 = a * b;
  if (__builtin_expect(g2 > 3, 0)) {
    a = c;
    b = d;
    g2 = a * b;
  }
  g3 = a * b;      // fully redundant.

}
The patch adds phi-translate support in scalarpre. This is only a temporary
solution before the newpre based on newgvn is available.

Differential Revision: https://reviews.llvm.org/D32252

llvm-svn: 305578

Whenever we don't know what to do with an instruction, we send
it to overdefined anyway.

llvm-svn: 305575

llvm-svn: 305570

Revert because of reports of some PPC input starting to spill when it
was predicted that it wouldn't and no spillslot was reserved.

This reverts commit r305516.

llvm-svn: 305566

llvm-svn: 305564

[InstCombine] Add test cases to show missed opportunities due to overly conservative single use checks. NFC

llvm-svn: 305562

If users tried to have a structure decl/init code like below
   struct test_t t = { .memeber1 = 45 };
It is very likely that compiler will generate a readonly section
to hold up the init values for variable t. Later load of t members,
e.g., t.member1 will result in a read from readonly section.

BPF program cannot handle relocation. This will force users to
write:
  struct test_t t = {};
  t.member1 = 45;
This is just inconvenient and unintuitive.

This patch addresses this issue by implementing BPF PreprocessISelDAG.
For any load from a global constant structure or an global array of
constant struct, it attempts to
translate it into a constant directly. The traversal of the
constant struct and other constant data structures are similar
to where the assembler emits read-only sections.

Four different unit test cases are also added to cover
different scenarios.

Signed-off-by: Yonghong Song <yhs@fb.com>
llvm-svn: 305560

o This is discovered during my study of 32-bit subregister
  support.
o This is no impact on current functionality since we
  only support 64-bit registers.
o Searching the web, looks like the issue has been discovered
  before, so fix it now.

Signed-off-by: Yonghong Song <yhs@fb.com>
llvm-svn: 305559