Avoid extra DenseMap lookups in StackColoring::calculateLocalLiveness.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175487 91177308-0d34-0410-b5e6-96231b3b80d8

Expand pseudos/macros BteqzT8SltiX16, BteqzT8SltiuX16,
BtnezT8SltiX16, BtnezT8SltiuX16 .

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175486 91177308-0d34-0410-b5e6-96231b3b80d8

Make the dump() function const and reduce the number of hash lookups it performs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175485 91177308-0d34-0410-b5e6-96231b3b80d8

Use a reference into the BlockLiveness DenseMap to avoid repeated hash lookups in collectMarkers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175484 91177308-0d34-0410-b5e6-96231b3b80d8

Remove some unused private fields from the AArch64MCCodeEmitter. These
fields were only ever set in the constructor. The create method retains
its consistent interface so that these bits can be re-threaded through
the emitter if they're ever needed.

This was found by the -Wunused-private-field Clang warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175482 91177308-0d34-0410-b5e6-96231b3b80d8

Temporarily revert r175470 for more review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175476 91177308-0d34-0410-b5e6-96231b3b80d8

Expand pseudos BteqzT8CmpiX16 and BtnezT8CmpiX16.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175474 91177308-0d34-0410-b5e6-96231b3b80d8

Use array_pod_sort instead of std::sort.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175472 91177308-0d34-0410-b5e6-96231b3b80d8

Check to see if the 'no-builtin' attribute is set before simplifying a library call.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175470 91177308-0d34-0410-b5e6-96231b3b80d8

Add some convenience methods for querying function attributes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175469 91177308-0d34-0410-b5e6-96231b3b80d8

X86FrameLowering.cpp: Fixup. Sorry for the breakage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175467 91177308-0d34-0410-b5e6-96231b3b80d8

Use LLVM_DELETED_FUNCTION rather than '// do not implement' comments.

Also removes some redundant DNI comments on function declarations already
using the macro.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175466 91177308-0d34-0410-b5e6-96231b3b80d8

X86FrameLowering.cpp: Fix a warning in -Asserts. [-Wunused-variable]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175464 91177308-0d34-0410-b5e6-96231b3b80d8

Remove a useless assert.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175463 91177308-0d34-0410-b5e6-96231b3b80d8

Comment out the rdar number.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175460 91177308-0d34-0410-b5e6-96231b3b80d8

[fast-isel] Remove an invalid assert.

If the memcpy has an odd length with an alignment of 2, this would incorrectly
assert on the last 1 byte copy.
rdar://13202135

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175459 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a 32/64 bit incompatibility in the HiPE prologue generation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175458 91177308-0d34-0410-b5e6-96231b3b80d8

Support for HiPE-compatible code emission, patch by Yiannis Tsiouris.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175457 91177308-0d34-0410-b5e6-96231b3b80d8

[ms-inline asm] Remove a redundant call to the setHasMSInlineAsm function.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175456 91177308-0d34-0410-b5e6-96231b3b80d8

Use llvm::cast instead of reinterpret_cast.
Also, GetElementPtrInst::getType() method returns SequentialType now, instead of
PointerType. There wasn't any issue yet, so no testcase attached.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175452 91177308-0d34-0410-b5e6-96231b3b80d8

Add front/back/erase to MapVector.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175449 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: Use MULADD_IEEE/V_MAD_F32 instruction for mad pattern

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175446 91177308-0d34-0410-b5e6-96231b3b80d8

R600: Support for TBO

NOTE: This is a candidate for the Mesa stable branch.

Reviewed-by: Tom Stellard <thomas.stellard at amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175445 91177308-0d34-0410-b5e6-96231b3b80d8

R600: Increase number of ArrayBase Reg to 32

Reviewed-by: Tom Stellard <thomas.stellard at amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175443 91177308-0d34-0410-b5e6-96231b3b80d8

[asan] revert r175266 as it breaks code with packed structures. supporting long double will require a more general solution

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175442 91177308-0d34-0410-b5e6-96231b3b80d8

Futureproof AttrBuild if we ever have more than 64 attr enum values.

Currently we're at 34. Bitset should compile into virtually the same code as
uint64_t here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175437 91177308-0d34-0410-b5e6-96231b3b80d8

Revert CMake version bump since we don't actually need 2.8.6 any more.

The problem call was removed in the interim. Apologies.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175435 91177308-0d34-0410-b5e6-96231b3b80d8

Note minimum required version of CMake

We already use features from 2.8.6, this just gives a slightly more friendly
message when the dependency isn't met.

Patch from Keith Walker.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175434 91177308-0d34-0410-b5e6-96231b3b80d8

AArch64: adjust tests which rely on a default JIT

Profiling tests *do* need a JIT. They'll pass if a cross-compiler targetting
AArch64 by default has been built, but fail if a native AArch64 compiler has
been build. Therefore XFAIL is inappropriate and we mark them unsupported.

ExecutionEngine tests are JIT by definition, they should also be unsupported.

Transforms/LICM only uses the interpreter to check the output is still sane
after optimisation. It can be switched to use an interpreter.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175433 91177308-0d34-0410-b5e6-96231b3b80d8

Add `unsigned DISubprogram::getFlags() const` for DragonEgg.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175430 91177308-0d34-0410-b5e6-96231b3b80d8

Add `DIFile DISubprogram::getFile() const` for DragonEgg.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175428 91177308-0d34-0410-b5e6-96231b3b80d8

DIBuilder: Correct the null/0 type of trailing fields in struct debug info.

Paired with an Clang commit so this may cause temporary build failures.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175426 91177308-0d34-0410-b5e6-96231b3b80d8

DIBuilder: Add function and method definitions to the list of all subprograms

Previously we seemed to be assuming that all functions were definitions and all
methods were declarations. This may be consistent with how Clang uses DIBuilder
but doesn't have to be true of all clients (such as DragonEgg).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175423 91177308-0d34-0410-b5e6-96231b3b80d8

[ms-inline asm] Fix undefined behavior to reset hasMSInlineAsm in advance of SelectAllBasicBlocks().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175422 91177308-0d34-0410-b5e6-96231b3b80d8

Narrow the return types of a few DIBuilder utility functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175421 91177308-0d34-0410-b5e6-96231b3b80d8

Expand macro/pseudo instructions BtnezT8SltX16 and BtnezT8SltuX16.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175420 91177308-0d34-0410-b5e6-96231b3b80d8

Expand pseudo/macro BteqzT8SltuX16 . There is no test case because
at this time, llvm is generating a different but equivalent pattern
that would lead to this instruction. I am trying to think of a way
to get it to generate this. If I can't, I may just remove the pseudo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175419 91177308-0d34-0410-b5e6-96231b3b80d8

Expand pseudo/macro BteqzT8SltX16.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175417 91177308-0d34-0410-b5e6-96231b3b80d8

Expand macro/pseudo BteqzT8CmpX16.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175416 91177308-0d34-0410-b5e6-96231b3b80d8

Spelling corrections

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175415 91177308-0d34-0410-b5e6-96231b3b80d8

Beginning of expanding all current mips16 macro/pseudo instruction sequences.
This expansion will be moved to expandISelPseudos as soon as I can figure
out how to do that. There are other instructions which use this
ExpandFEXT_T8I816_ins and as soon as I have finished expanding them all,
I will delete the macro asm string text so it has no way to be used
in the future.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175413 91177308-0d34-0410-b5e6-96231b3b80d8

X86: Add a note.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175408 91177308-0d34-0410-b5e6-96231b3b80d8

[XCore] Add missing 2r instructions.

These instructions are not targeted by the compiler but it is needed for
the MC layer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175407 91177308-0d34-0410-b5e6-96231b3b80d8

[XCore] Add TSETR instruction.

This instruction is not targeted by the compiler but it is needed for the
MC layer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175406 91177308-0d34-0410-b5e6-96231b3b80d8

[XCore] Add missing u10 / lu10 instructions.

These instructions are not targeted by the compiler but they are
needed for the MC layer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175404 91177308-0d34-0410-b5e6-96231b3b80d8

[XCore] Add missing u6 / lu6 instructions.

These instructions are not targeted by the compiler but they are
needed for the MC layer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175403 91177308-0d34-0410-b5e6-96231b3b80d8

Return false instead of 0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175402 91177308-0d34-0410-b5e6-96231b3b80d8

Force a cpu for test. It failed on atom due to different scheduling decisions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175401 91177308-0d34-0410-b5e6-96231b3b80d8

AArch64: Avoid shifts by 64, that's undefined behavior.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175400 91177308-0d34-0410-b5e6-96231b3b80d8

Add multithreading functions and shutdown to the C API.  Patch by Moritz
Maxeiner.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175398 91177308-0d34-0410-b5e6-96231b3b80d8

BBVectorize: Fix an invalid reference bug

This fixes PR15289. This bug was introduced (recently) in r175215; collecting
all std::vector references for candidate pairs to delete at once is invalid
because subsequent lookups in the owning DenseMap could invalidate the
references.

bugpoint was able to reduce a useful test case. Unfortunately, because whether
or not this asserts depends on memory layout, this test case will sometimes
appear to produce valid output. Nevertheless, running under valgrind will
reveal the error.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175397 91177308-0d34-0410-b5e6-96231b3b80d8

Make the visibility of LLVMPPCCompilationCallback work with GCC.

GCC warns about the attribute being ignored if it occurs after void*.
There seems to be some kind of incompatibility between clang and gcc here, but
I can't fathom who's right.

void* LLVM_LIBRARY_VISIBILITY foo(); // clang: hidden, gcc: default
LLVM_LIBRARY_VISIBILITY void *bar(); // clang: hidden, gcc: hidden
void LLVM_LIBRARY_VISIBILITY qux();  // clang: hidden, gcc: hidden

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175394 91177308-0d34-0410-b5e6-96231b3b80d8

Remove use of reverse iterators in repairIntervalsInRange(). While they were
arguably better than forward iterators for this use case, they are confusing and
there are some implementation problems with reverse iterators and MI bundles.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175393 91177308-0d34-0410-b5e6-96231b3b80d8

Use ArrayRef instead of a reference to a SmallVectorImpl.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175385 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a conversion from a forward iterator to a reverse iterator in
MachineBasicBlock::SplitCriticalEdge. Since this is an iterator rather than
an instr_iterator, the isBundled() check only passes if getFirstTerminator()
returned end() and the garbage memory happens to lean that way.

Multiple successors can be present without any terminator instructions in the
case of exception handling with a fallthrough.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175383 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for updating the LiveIntervals of registers used by 'exotic'
terminators that actually have register uses when splitting critical edges.

This commit also introduces a method repairIntervalsInRange() on LiveIntervals,
which allows for repairing LiveIntervals in a small range after an arbitrary
target hook modifies, inserts, and removes instructions. It's pretty limited
right now, but I hope to extend it to support all of the things that are done
by the convertToThreeAddress() target hooks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175382 91177308-0d34-0410-b5e6-96231b3b80d8

The transform is:

    (or (bool?A:B),(bool?C:D)) --> (bool?(or A,C):(or B,D))

By the time the OR is visited, both the SELECTs have been visited and not
optimized and the OR itself hasn't been transformed so we do this transform in
the hopes that the new ORs will be optimized.

The transform is explicitly disabled for vector-selects until "codegen matures
to handle them better".

Patch by Muhammad Tauqir!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175380 91177308-0d34-0410-b5e6-96231b3b80d8

Clean up mips16 td file in preparation for massive pseudo lowering work.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175379 91177308-0d34-0410-b5e6-96231b3b80d8

Documentation: correct syntax (one missing comma, one extra comma)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175375 91177308-0d34-0410-b5e6-96231b3b80d8

GCC doesn't like ++ on enums.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175373 91177308-0d34-0410-b5e6-96231b3b80d8

Remove comma at the end of enum. Still my favourite C++11 feature.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175372 91177308-0d34-0410-b5e6-96231b3b80d8

Typo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175371 91177308-0d34-0410-b5e6-96231b3b80d8

Turn the enum attributes DenseSet in AttrBuilder into a set of bits.

Avoids malloc and is a lot denser. We lose iteration over target independent
attributes, but that's a strange interface anyways and didn't have any users
outside of AttrBuilder.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175370 91177308-0d34-0410-b5e6-96231b3b80d8

One more try to make this look nice. I have lots of pseudo lowering
as well as 16/32 bit variants to do and so I want this to look nice
when I do it. I've been experimenting with this. No new test cases
are needed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175369 91177308-0d34-0410-b5e6-96231b3b80d8

Replace erase loop with std::remove_if.

This avoids unnecessary copies. No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175367 91177308-0d34-0410-b5e6-96231b3b80d8

Replace loop with std::find.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175366 91177308-0d34-0410-b5e6-96231b3b80d8

LegalizeDAG.cpp doesn't need DenseMap.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175365 91177308-0d34-0410-b5e6-96231b3b80d8

Reduce indents in LSRInstance::NarrowSearchSpaceByCollapsingUnrolledCode method.
No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175364 91177308-0d34-0410-b5e6-96231b3b80d8

[msvc x64] Update X86CompilationCallback_Win64.asm corresponding to r175267.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175363 91177308-0d34-0410-b5e6-96231b3b80d8

Use const reference instead of vector object when passing an argument to
updateScheduledPressure method.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175362 91177308-0d34-0410-b5e6-96231b3b80d8

Target/R600/CMakeLists.txt: Prune SILowerLiteralConstants.cpp corresponding to r175354.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175361 91177308-0d34-0410-b5e6-96231b3b80d8

MCParser: Reject .balign with non-pow2 alignments.

GNU as rejects them and there are configure scripts in the wild that check if
the assembler rejects ".align 3" to determine whether the alignment is in bytes
or powers of two.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175360 91177308-0d34-0410-b5e6-96231b3b80d8

Minor cleanups. No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175359 91177308-0d34-0410-b5e6-96231b3b80d8

Add even more constatness in MachineDominators.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175358 91177308-0d34-0410-b5e6-96231b3b80d8

Replace "check:" wth "CHECK:".
Also fix one test by changing "vpermilps" to "vpshufd".

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175357 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: Add pattern to simplify i64 loading

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175356 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: nuke SReg_1 v3

It's completely unnecessary and can be replace with proper
SReg_64 handling instead.

This actually fixes a piglit test on SI.

v2: use correct register class in addRegisterClass,
    set special classes as not allocatable
v3: revert setting special classes as not allocateable

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175355 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: cleanup literal handling v3

Seems to be allot simpler, and also paves the
way for further improvements.

v2: rebased on master, use 0 in BUFFER_LOAD_FORMAT_XYZW,
    use VGPR0 in dummy EXP, avoid compiler warning, break
    after encoding the first literal.
v3: correctly use V_ADD_F32_e64

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175354 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: replace AllReg_* with [SV]Src_* v2

Mark all the operands that can also have an immediate.

v2: SOFFSET is also an SSrc_32 operand

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175353 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: fix VOPC encoding v2

Previously it only worked because of coincident.

v2: fix 64bit versions, use 0x80 (inline 0) instead of SGPR0
    for the unused SRC2

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175352 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: move *_Helper definitions to SIInstrFormat.td

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175351 91177308-0d34-0410-b5e6-96231b3b80d8

R600/SI: remove some more unused code

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175350 91177308-0d34-0410-b5e6-96231b3b80d8

R600/structurizer: improve inverting conditions

Stop adding more instructions than necessary.

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175349 91177308-0d34-0410-b5e6-96231b3b80d8

R600/structurizer: improve loop handling

Generate more than one loop if it seems to make sense.

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175348 91177308-0d34-0410-b5e6-96231b3b80d8

R600/structurizer: improve finding condition values

Using the new NearestCommonDominator class.

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175347 91177308-0d34-0410-b5e6-96231b3b80d8

R600/structurizer: improve PHI value finding

Using the new NearestCommonDominator class.

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175346 91177308-0d34-0410-b5e6-96231b3b80d8

R600/structurizer: add class to find the Nearest Common Dominator

This is a candidate for the stable branch.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175345 91177308-0d34-0410-b5e6-96231b3b80d8

Add some constantness in Machine Dominators.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175344 91177308-0d34-0410-b5e6-96231b3b80d8

Remove trailing spaces.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175343 91177308-0d34-0410-b5e6-96231b3b80d8

Use a different scheme to chose 16/32 variants. This scheme is more
consistent with how BuildMI works. No new tests needed. All should work
the same as before.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175342 91177308-0d34-0410-b5e6-96231b3b80d8

[ELF] Add STN_UNDEF.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175338 91177308-0d34-0410-b5e6-96231b3b80d8

Reinitialize the ivars in the subtarget so that they can be reset with the new features.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175336 91177308-0d34-0410-b5e6-96231b3b80d8

[ms-inline asm] Do not omit the frame pointer if we have ms-inline assembly.

If the frame pointer is omitted, and any stack changes occur in the inline
assembly, e.g.: "pusha", then any C local variable or C argument references
will be incorrect.

I pass no judgement on anyone who would do such a thing. ;)
rdar://13218191

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175334 91177308-0d34-0410-b5e6-96231b3b80d8

Derive ELF section type from the name in some cases where GNU as does
so.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175327 91177308-0d34-0410-b5e6-96231b3b80d8

[mips] Remove SDNPWantParent from the list of SDNodeProperties.

No functionality change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175325 91177308-0d34-0410-b5e6-96231b3b80d8

Pacify -Wnon-virtual-dtor

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175324 91177308-0d34-0410-b5e6-96231b3b80d8

Temporary revert of 175320.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175322 91177308-0d34-0410-b5e6-96231b3b80d8

Reinitialize the ivars in the subtarget.

When we're recalculating the feature set of the subtarget, we need to have the
ivars in their initial state.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175320 91177308-0d34-0410-b5e6-96231b3b80d8

Suppress a GCC -Wunused-variable warning in -Asserts builds

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175319 91177308-0d34-0410-b5e6-96231b3b80d8

c[ms-inline asm] It's possible to have a SizeDirective rewrite and an
Input/Output rewrite to the same location.  Make sure the SizeDirective rewrite
is performed first.  This also ensure the sort algorithm is stable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175317 91177308-0d34-0410-b5e6-96231b3b80d8