Patch by Clement Courbet

llvm-svn: 266986

Differential Revision: http://reviews.llvm.org/D14822

llvm-svn: 266985

Reviewers: nhaustov, tstellarAMD

Subscribers: arsenm

Differential Revision: http://reviews.llvm.org/D19317

llvm-svn: 266984

[Clang][AVX512][BuiltIn] Adding intrinsics of VGATHER{DPS|DPD} , VPGATHER{QD|QQ|DD|DQ} and VGATHERPF{0|1}{DPS|QPS|DPD|QPD} instruction set .

Differential Revision: http://reviews.llvm.org/D19224

llvm-svn: 266983

llvm-svn: 266982

A DenseMap doesn't store the hashes, so it needs to recompute them when
the table is resized.

In some applications the hashing cost is noticeable. That is the case
for example in lld for symbol names (StringRef).

This patch adds a templated structure that can wraps any value that can
go in a DenseMap and caches the hash.

llvm-svn: 266981

Differential Revision: http://reviews.llvm.org/D18855

llvm-svn: 266980

llvm-svn: 266979

llvm-svn: 266978

[mips][microMIPS] Implement LL, SC, MOVEP, ROTR, ROTRV and SYSCALL instructions and add tests for LWM32 and SWM32
Differential Revision: http://reviews.llvm.org/D19150

llvm-svn: 266977

Quit parsing MS-style inline assembly if the following statement has GCC style.
Enables compilation of code like

void f() {
  __asm mov ebx, ecx
  __asm__("movl %ecx, %edx");
}

Differential Revision: http://reviews.llvm.org/D18652

llvm-svn: 266976

It breaks on windows, need to investigate. It's not testing the
important part of that change anyways.

llvm-svn: 266975

SectionOrder vector was a part of LinkerScript class.

It can be removed because Commands vector contains the 
same information and SectiorOrder is just a subset.

Differential revision: http://reviews.llvm.org/D19171

llvm-svn: 266974

This allows using a different standard library (the one from argv[0] in
the compilation database) with the correct builtins.

Differential Revision: http://reviews.llvm.org/D19356

llvm-svn: 266973

Add -miamcu option which:
  * Sets IAMCU triple
  * Sets IAMCU ABI
  * Enforces static compilation

Differential Revision: http://reviews.llvm.org/D18398

llvm-svn: 266972

llvm-svn: 266971

Reviewers: bkramer

Subscribers: cfe-commits

Differential Revision: http://reviews.llvm.org/D19323

llvm-svn: 266970

[AArch64][CodeGen] Fix of PR27158: incorrect peephole optimization in AArch64InstrInfo::optimizeCompareInstr

AArch64InstrInfo::optimizeCompareInstr has bug PR27158 which causes generation of incorrect code.
A compare instruction is substituted with another instruction which does not
produce the same flags as the original compare instruction.
This patch contains:
1. Fix of the bug.
2. A regression test in MIR.
3. A new test to check that SUBS is replaced by SUB.

Differential Revision: http://reviews.llvm.org/D18838

llvm-svn: 266969

llvm-svn: 266968

[X86] Fix vector-tzcnt-512 test to disable CDI while enabling BWI for one of the runs. Update check patterns accordingly.

llvm-svn: 266967

llvm-svn: 266966

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266965

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266964

llvm-svn: 266963

[X86] Remove redundant calls to setOperationAction for EXTRACT_VECTOR_ELT/INSERT_VECTOR_ELT from SSE41 block. They were already done in an earlier block. NFC

llvm-svn: 266962

[X86] Remove some operations from the default Expand all vector ops loop. Instead let them stay Legal and mark them Expand for specific types where needed. Reduces overall number of calls to setOperationAction. NFC

llvm-svn: 266961

[X86] Remove old leftover MMX code that sets various 64-bit vector operations to Expand. These vector types aren't legal so these operations would never make it far enough to need to expand. NFC

llvm-svn: 266960

Differential Revision: http://reviews.llvm.org/D18494

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266959

This help to streamline the process of handling importing since
we don't need to special case alias everywhere: just like
linkonce_odr function, make sure at least one alias is emitted
by turning it weak.

Differential Revision: http://reviews.llvm.org/D19308

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266958

llvm-svn: 266957

Summary:
 when setting LIBCXX_ENABLE_EXCEPTIONS=false, _LIBCPP_NO_EXCEPTIONS wil be defined in both commandline and _config

Reviewers: bcraig, EricWF

Subscribers: cfe-commits

Differential Revision: http://reviews.llvm.org/D19344

llvm-svn: 266956

Summary:
`llvm.guard(false)` always bails out of the current compilation unit, so
we can prune any control flow following it.

Reviewers: hfinkel, pcc, reames

Subscribers: majnemer, reames, mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D19245

llvm-svn: 266955

-> Drop linkage when converting to decl
-> Handling of unnamed_addr.

llvm-svn: 266954

This is not on by default (but it might be in the future).
The knob to enable the optimization is -lto-discard-value-names.

llvm-svn: 266953

[X86] Remove unnecessary setting of CTTZ_ZERO_UNDEF to Custom for vector types where we can't do any better than the Custom lowering of CTTZ. LegalizeVectorOps will expand to CTTZ since its marked Custom.

CTTZ_ZERO_UNDEF can be custom lowered specially if CTLZ is supported. Otherwise CTTZ and CTTZ_ZERO_UNDEF are handled the same way by using CTPOP and bitmath.

llvm-svn: 266952

[SelectionDAG] Teach LegalizeVectorOps to directly Expand CTTZ_ZERO_UNDEF/CTLZ_ZERO_UNDEF to CTTZ/CTLZ directly if those ops are Legal/Custom instead of deferring it to LegalizeOps.

This is needed to support CTTZ/CTLZ Custom correctly since LegalizeOps would be too late to do the custom lowering.

llvm-svn: 266951

Without BWI we have to split the vectors into 256-bit vectors so we can use AVX2 pshufb and then concatenate the results.

llvm-svn: 266950

The iteratitive algorithm from r265456 claimed but failed to create a
post-order traversal.  It had the same error that was fixed in the
ValueEnumerator in r266947: now, instead of pushing all operands on the
worklist at once, we pause whenever an operand gets pushed in order to
go depth-first (I know, it sounds obvious).

Sadly, I have no idea how to observe this from outside the algorithm and
so I haven't written a test.  The output should be the same; it should
just use fewer temporary nodes now.  I've added some comments that I
hope make the current logic clear enough it's unlikely to regress.

llvm-svn: 266949

Summary:
The function importer already decided what symbols need to be pulled
in. Also these magically added ones will not be in the export list
for the source module, which can confuse the internalizer for
instance.

Reviewers: tejohnson, rafael

Subscribers: joker.eph, llvm-commits

Differential Revision: http://reviews.llvm.org/D19096

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266948

Emit metadata nodes in post-order.  The iterative algorithm from r266709
failed to maintain this property.  After understanding my mistake, it
wasn't too hard to write a test with llvm-bcanalyzer (and I've actually
made this change once before: see r220340).

This also reverts the "noisy" testcase change from r266709.  That should
have been more of a red flag :/.

Note: The same bug crept into the ValueMapper in r265456.  I'm still
working on the fix.

llvm-svn: 266947