[X86] Fix indentation. NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283065 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r283057 and r283058

They've broken the sanitizer-bootstrap bots.  Reverting while I investigate.

Original commit messages:

r283057: "[ConstantRange] Make getEquivalentICmp smarter"

r283058: "[SCEV] Rely on ConstantRange instead of custom logic; NFCI"

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283062 91177308-0d34-0410-b5e6-96231b3b80d8

[PowerPC] Refactor soft-float support, and enable PPC64 soft float

This change enables soft-float for PowerPC64, and also makes soft-float disable
all vector instruction sets for both 32-bit and 64-bit modes. This latter part
is necessary because the PPC backend canonicalizes many Altivec vector types to
floating-point types, and so soft-float breaks scalarization support for many
operations. Both for embedded targets and for operating-system kernels desiring
soft-float support, it seems reasonable that disabling hardware floating-point
also disables vector instructions (embedded targets without hardware floating
point support are unlikely to have Altivec, etc. and operating system kernels
desiring not to use floating-point registers to lower syscall cost are unlikely
to want to use vector registers either). If someone needs this to work, we'll
need to change the fact that we promote many Altivec operations to act on
v4f32. To make it possible to disable Altivec when soft-float is enabled,
hardware floating-point support needs to be expressed as a positive feature,
like the others, and not a negative feature, because target features cannot
have dependencies on the disabling of some other feature. So +soft-float has
now become -hard-float.

Fixes PR26970.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283060 91177308-0d34-0410-b5e6-96231b3b80d8

Remove duplicated code; NFC

ICmpInst::makeConstantRange does exactly the same thing as
ConstantRange::makeExactICmpRegion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283059 91177308-0d34-0410-b5e6-96231b3b80d8

[SCEV] Rely on ConstantRange instead of custom logic; NFCI

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283058 91177308-0d34-0410-b5e6-96231b3b80d8

[ConstantRange] Make getEquivalentICmp smarter

This change teaches getEquivalentICmp to be smarter about generating
ICMP_NE and ICMP_EQ predicates.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283057 91177308-0d34-0410-b5e6-96231b3b80d8

[SCEV] Remove commented out code; NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283056 91177308-0d34-0410-b5e6-96231b3b80d8

COFF: Fix short import lib import name type bitshift

As per the PE COFF spec (section 8.3, Import Name Type)
Offset: 18 Size 2 bits Name: Type
Offset: 20 Size 3 bits Name: Name Type

Offset: 20 added based on 18+2

Partially commited as rL279069

Differential Revision: https://reviews.llvm.org/D23540

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283055 91177308-0d34-0410-b5e6-96231b3b80d8

[X86][SSE] Cleaned up shuffle decode assertion messages

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283050 91177308-0d34-0410-b5e6-96231b3b80d8

[SLPVectorizer][X86] Added fptosi/fptoui tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283048 91177308-0d34-0410-b5e6-96231b3b80d8

[CostModel][X86] Added tests for current fptosi/fptoui costs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283047 91177308-0d34-0410-b5e6-96231b3b80d8

[SLPVectorizer][X86] Added fcopysign tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283046 91177308-0d34-0410-b5e6-96231b3b80d8

[SLPVectorizer][X86] Added fabs tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283045 91177308-0d34-0410-b5e6-96231b3b80d8

[CostModel][X86] Added fcopysign costs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283044 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef for MemoryBuffer identifier API (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283043 91177308-0d34-0410-b5e6-96231b3b80d8

[CostModel][X86] Added fabs costs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283042 91177308-0d34-0410-b5e6-96231b3b80d8

Fix signed/unsigned warning

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283041 91177308-0d34-0410-b5e6-96231b3b80d8

[X86][SSE] Add support for combining target shuffles to binary BLEND

We already had support for 1-input BLEND with zero - this adds support for 2-input BLEND as well.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283040 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in Registry API (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283039 91177308-0d34-0410-b5e6-96231b3b80d8

[X86][SSE] Always combine target shuffles to MOVSD/MOVSS

Now we can commute to BLENDPD/BLENDPS on SSE41+ targets if necessary, so simplify the combine matching where we can.

This required me to add a couple of scalar math movsd/moss fold patterns that hadn't been needed in the past.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283038 91177308-0d34-0410-b5e6-96231b3b80d8

[X86][SSE] Enable commutation from MOVSD/MOVSS to BLENDPD/BLENDPS on SSE41+ targets

Instead of selecting between MOVSD/MOVSS and BLENDPD/BLENDPS at shuffle lowering by subtarget this will help us select the instruction based on actual commutation requirements.

We could possibly add BLENDPD/BLENDPS -> MOVSD/MOVSS commutation and MOVSD/MOVSS memory folding using a similar approach if it proves useful

I avoided adding AVX512 handling as I'm not sure when we should be making use of VBLENDPD/VBLENDPS on EVEX targets

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283037 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r283029 - [cmake] Make LIT_COMMAND configurable and improve fallback support

Revert the change in r283029 (and the fixup in r283033) due to buildbot
breakage. The fixup is ineffective for the bots that do not force clean
build since the wrong value is already cached in CMakeCache.txt.

Reverting it should result in the cache variable being removed
and therefore it should be possible to re-introduce it after all
buildbots build this revision.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283036 91177308-0d34-0410-b5e6-96231b3b80d8

[X86][SSE] Regenerate vselect tests and improve AVX1/AVX2 coverage

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283035 91177308-0d34-0410-b5e6-96231b3b80d8

Revert "[MC] Prevent out of order HashDirective lexing in AsmLexer."

This reverts commit r282992 which appears to be causing an LTO test failure.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283034 91177308-0d34-0410-b5e6-96231b3b80d8

[cmake] Fix incorrect default for LIT_COMMAND, from r283029

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283033 91177308-0d34-0410-b5e6-96231b3b80d8

[cmake] Make LIT_COMMAND configurable and improve fallback support

Make LIT_COMMAND configurable, use source tree only when actually
available and extend the default search to other common executable names
'lit.py' and 'lit', in order to increase uniformity between all LLVM
projects and support using installed lit.

Changing the conditional used to determine whether in-tree or external
lit is being used covers the case when LLVM_MAIN_SRC_DIR is defined but
does not exist (anymore). In this case, the functions falls back to
looking for installed lit rather than attempting to use a non-existing
path. The same conditional is used in clang already.

Making LIT_COMMAND a cache variable in case the source tree variant is
used serves two purposes. Firstly, it increases uniformity between
the two branches since find_program() implicitly makes LIT_COMMAND
a cache variable. Secondly, it allows overriding the lit executable used
to run the tests when the LLVM source tree is provided. Gentoo is
planning to use this to use installed (and byte-compiled) lit instead of
re-compiling it in every LLVM project.

Extending default search is meant to increase uniformity between
different LLVM projects. The 'lit.py' name is already used by a few of
them, and 'lit' is the name used by utils/lit/setup.py when installing.

Differential Revision: https://reviews.llvm.org/D25076

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283029 91177308-0d34-0410-b5e6-96231b3b80d8

[OCaml] Install .mli (interface) files

Install the OCaml interface .mli files. Those files were most likely
omitted because they are input files for the compiled .cmi files.
However, installing them is reasonable since -- unlike .cmi files --
they are human-readable.

The issue was originally spotted by @jpdeplaix.

Differential Revision: https://reviews.llvm.org/D25128

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283028 91177308-0d34-0410-b5e6-96231b3b80d8

Split a comment into generic description and note about the specific
cmake use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283027 91177308-0d34-0410-b5e6-96231b3b80d8

Retire LLVM_BINDIR and friends. They haven't been provided with actual
values since the switch to cmake.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283026 91177308-0d34-0410-b5e6-96231b3b80d8

[libFuzzer] add fuzzer test for libxml2, finds https://bugzilla.gnome.org/show_bug.cgi?id=751631

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283024 91177308-0d34-0410-b5e6-96231b3b80d8

GC HAVE_STRTOQ

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283023 91177308-0d34-0410-b5e6-96231b3b80d8

Retire bugpoint's -R. hack.

It got disconnected during the cmake conversion. For Miscompilation.cpp,
it was purely advisory for the user and the ToolRunner.cpp version was
trying to compensate for libs and bins in the same directory, which
hasn't been the case for a very long time.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283022 91177308-0d34-0410-b5e6-96231b3b80d8

[libFuzzer] fix a recent bugs (buffer overflow)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283021 91177308-0d34-0410-b5e6-96231b3b80d8

[X86] Cleanup patterns for using VMOVDDUP for broadcasts.

-Remove OptForSize. Not all of the backend follows the same rules for creating broadcasts and there is no conflicting pattern.
-Don't stop selecting VEX VMOVDDUP when AVX512 is supported. We need VLX for EVEX VMOVDDUP.
-Only use VMOVDDUP for v2i64 broadcasts if AVX2 is not supported.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283020 91177308-0d34-0410-b5e6-96231b3b80d8

Revert "Use StringRef instead of raw pointer in TargetRegistry API (NFC)"

This reverts commit r283017. Creates an infinite loop somehow.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283019 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef instead of raw pointers in MCAsmInfo/MCInstrInfo APIs (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283018 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef instead of raw pointer in TargetRegistry API (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283017 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef instead of raw pointer in ExecutionEngine

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283016 91177308-0d34-0410-b5e6-96231b3b80d8

[AVX-512] Add EVEX versions of VPBROADCASTW patterns with truncated i32 loads.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283015 91177308-0d34-0410-b5e6-96231b3b80d8

[AVX-512] Add VLX command lines to 128 and 256-bit shufffle tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283014 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in Datalayout API (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283013 91177308-0d34-0410-b5e6-96231b3b80d8

DIFlags: use StringRef instead of raw pointer (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283012 91177308-0d34-0410-b5e6-96231b3b80d8

Revert "Use StringRef in Datalayout API (NFC)"

This reverts commit r283009. Bots are broken.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283011 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in Datalayout API (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283009 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in Pass Info/Support API (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283008 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in CommandLine Options handling (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283007 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef instead of raw pointer in MachinePassRegistry (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283006 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in TLI instead of raw pointer (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283005 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in Pass/PassManager APIs (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283004 91177308-0d34-0410-b5e6-96231b3b80d8

Revert "AMDGPU: Don't use offen if it is 0"

This reverts commit r282999.
Tests are not passing: http://lab.llvm.org:8011/builders/clang-x86_64-linux-selfhost-modules/builds/20038

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283003 91177308-0d34-0410-b5e6-96231b3b80d8

Remove getTargetTriple and update all uses to use the Triple off
of the TargetMachine. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283002 91177308-0d34-0410-b5e6-96231b3b80d8

Stop calling getTargetTriple off of the AsmPrinter and constructing a
TargetTriple, just grab it off of the TargetMachine. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283001 91177308-0d34-0410-b5e6-96231b3b80d8

Remove TargetTriple from AArch64MCInstLower as it's used in few places
and can be pulled from the TargetMachine. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@283000 91177308-0d34-0410-b5e6-96231b3b80d8

AMDGPU: Don't use offen if it is 0

This removes many re-initializations of a base register to 0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282999 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in LTOCodegenerator (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282998 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in LTOModule implementation (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282997 91177308-0d34-0410-b5e6-96231b3b80d8

Use StringRef in Triple API (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282996 91177308-0d34-0410-b5e6-96231b3b80d8

[libFuzzer] implement the -shrink=1 option that tires to make elements of the corpus smaller, off by default

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282995 91177308-0d34-0410-b5e6-96231b3b80d8

[MC] Prevent out of order HashDirective lexing in AsmLexer.

To lex hash directives we peek ahead to find component tokens, create a
unified token, and unlex the peeked tokens so the parser does not need
to parse the tokens then. Make sure we do not to lex another hash
directive during peek operation.

This fixes PR28921.

Reviewers: rnk, loladiro

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D24839

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282992 91177308-0d34-0410-b5e6-96231b3b80d8

[ASAN] Add the binder globals on Darwin to llvm.compiler.used to avoid LTO dead-stripping

The binder is in a specific section that "reverse" the edges in a
regular dead-stripping: the binder is live as long as a global it
references is live.

This is a big hammer that prevents LLVM from dead-stripping these,
while still allowing linker dead-stripping (with special knowledge
of the section).

Differential Revision: https://reviews.llvm.org/D24673

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282988 91177308-0d34-0410-b5e6-96231b3b80d8

[libFuzzer] remove some experimental code

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282983 91177308-0d34-0410-b5e6-96231b3b80d8

[llvm-objdump] Switch to a range loop. NFCI.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282982 91177308-0d34-0410-b5e6-96231b3b80d8

ScheduleDAGInstrs: Cleanup, use range based for; NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282979 91177308-0d34-0410-b5e6-96231b3b80d8

[libFuzzer] fix openssl fuzzer tests when running on a machine w/o openssl installed

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282972 91177308-0d34-0410-b5e6-96231b3b80d8

[libFuzzer] remove unused option

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282971 91177308-0d34-0410-b5e6-96231b3b80d8

[SEH] Emit the parent frame offset label even if there are no funclets

This avoids errors about references to undefined local labels from
unreferenced filter functions.

Fixes (sort of) PR30431

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282967 91177308-0d34-0410-b5e6-96231b3b80d8

[AArch64][RegisterBankInfo] Use the helper functions for the checks

This makes sure the helper functions work as expected.

NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282961 91177308-0d34-0410-b5e6-96231b3b80d8

[AArch64][RegisterBankInfo] Rename getValueMappingIdx to getValueMapping

We don't return index, we return the actual ValueMapping.

NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282960 91177308-0d34-0410-b5e6-96231b3b80d8

[AArch64][RegisterBankInfo] Compress the ValueMapping table a bit.

We don't need to have singleton ValueMapping on their own, we can just
reuse one of the elements of the 3-ops mapping.
This allows even more code sharing.

NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282959 91177308-0d34-0410-b5e6-96231b3b80d8

[AArch64][RegisterBankInfo] Refactor the code to access AArch64::ValMapping

Use a helper function to access ValMapping. This should make the code
easier to understand and maintain.

NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282958 91177308-0d34-0410-b5e6-96231b3b80d8

[AArch64][RegisterBankInfo] Rename getRegBankIdx to getRegBankIdxOffset

The function name did not make it clear that the returned value was an
offset to apply to a register bank index.

NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282957 91177308-0d34-0410-b5e6-96231b3b80d8

[AArch64][RegisterBankInfo] Use the static opds mapping for alt mappings

Avoid to rely on the dynamically allocated operands mapping for the
alternative mapping.
NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282956 91177308-0d34-0410-b5e6-96231b3b80d8

[libFuzzer] move common parts of shell scripts into a separate file

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282954 91177308-0d34-0410-b5e6-96231b3b80d8

[libfuzzer] sancov documentation update

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282953 91177308-0d34-0410-b5e6-96231b3b80d8

NFC Add const

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282952 91177308-0d34-0410-b5e6-96231b3b80d8

NFC fix class members initialization

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282951 91177308-0d34-0410-b5e6-96231b3b80d8

NFC fix doxygen comments

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282950 91177308-0d34-0410-b5e6-96231b3b80d8

Do not pass a superblock to PDBFileBuilder.

When we create a PDB file using PDBFileBuilder, the information
in the superblock, such as the size of the resulting file, is not
available.

Previously, PDBFileBuilder::initialize took a superblock assuming
that all the members of the struct are correct. That is useful when
you want to restore the exact information from a YAML file, but
that's probably the only use case in which that is useful.
When we are creating a PDB file on the fly, we have to backfill the
members.

This patch redefines PDBFileBuilder::initialize to take only a
block size. Now all the other members are left as default values,
so that they'll be updated when commit() is called.

Differential Revision: https://reviews.llvm.org/D25108

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282944 91177308-0d34-0410-b5e6-96231b3b80d8

Pass a filename instead of a msf::WritableStream to PDBFileBuilder::commit.

WritableStream needs the exact file size to open a file, but
until we fix the final layout of a PDB file, we don't know the
size of the file.

This patch changes the parameter type of PDBFileBuilder::commit
to solve that chiecken-and-egg problem. Now the function opens
a file after fixing the layout, so it can create a file with the
exact size.

Differential Revision: https://reviews.llvm.org/D25107

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282940 91177308-0d34-0410-b5e6-96231b3b80d8

Sort LLVM_VERSION_INFO

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282939 91177308-0d34-0410-b5e6-96231b3b80d8

GC left-over from workarounds for missing pid_t and size_t

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282938 91177308-0d34-0410-b5e6-96231b3b80d8

GC ENABLE_PIC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282936 91177308-0d34-0410-b5e6-96231b3b80d8

GC HAVE___DSO_HANDLE

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282935 91177308-0d34-0410-b5e6-96231b3b80d8

Correctly expand HOST_LINK_VERSION.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282934 91177308-0d34-0410-b5e6-96231b3b80d8

Fix expansion of HAVE_SYS_MMAN_H

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282933 91177308-0d34-0410-b5e6-96231b3b80d8

GC HAVE_LINK_EXPORT_DYNAMIC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282932 91177308-0d34-0410-b5e6-96231b3b80d8

GC HAVE_MMAP and HAVE_MMAP_FILE

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282931 91177308-0d34-0410-b5e6-96231b3b80d8

Spell comment consistently with other library comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282930 91177308-0d34-0410-b5e6-96231b3b80d8

Sort LINK_POLLY_INTO_TOOLS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282929 91177308-0d34-0410-b5e6-96231b3b80d8

GC STDC_HEADERS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282928 91177308-0d34-0410-b5e6-96231b3b80d8

Deal with the (historic) MAP_ANONYMOUS vs MAP_ANON directly by using CPP
to check for the former, don't depend on (dangling) HAVE_MMAP_ANONYMOUS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282925 91177308-0d34-0410-b5e6-96231b3b80d8

Retire NEED_DEV_ZERO_FOR_MMAP. It should be needed only on outdated
systems. It wasn't even hooked up in cmake, so problems on such systems
would be visible with 3.9 release already.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282924 91177308-0d34-0410-b5e6-96231b3b80d8

HAVE_LINK_R is not the only reason why this needs config.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282923 91177308-0d34-0410-b5e6-96231b3b80d8

GC HAVE_LIBDL, HAVE_LIBM and HAVE_LIBOLE32

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282922 91177308-0d34-0410-b5e6-96231b3b80d8

Sort HAVE_LIBEDIT.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282921 91177308-0d34-0410-b5e6-96231b3b80d8

X86: Allow conditional tail calls in Win64 "leaf" functions (PR26302)

We can't use Jcc to leave a Win64 function in general, because that
confuses the unwinder. However, for "leaf" functions, that is, functions
where the return address is always on top of the stack and which don't
have unwind info, it's OK.

Differential Revision: https://reviews.llvm.org/D24836

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282920 91177308-0d34-0410-b5e6-96231b3b80d8

Turn ENABLE_CRASH_OVERRIDES into a 0/1 definition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282919 91177308-0d34-0410-b5e6-96231b3b80d8

Convert ENABLE_BACKTRACES into a 0/1 definition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282918 91177308-0d34-0410-b5e6-96231b3b80d8

GC TIME_WITH_SYS_TIME and TM_IN_SYS_TIME

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282917 91177308-0d34-0410-b5e6-96231b3b80d8

GC STAT_MACROS_BROKEN.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@282915 91177308-0d34-0410-b5e6-96231b3b80d8