platform/upstream/llvm.git
7 years agoP0620 follow-up: deducing `auto` from braced-init-list in new expr
Zhihao Yuan [Mon, 11 Dec 2017 18:29:54 +0000 (18:29 +0000)]
P0620 follow-up: deducing `auto` from braced-init-list in new expr

Summary:
This is a side-effect brought in by p0620r0, which allows other placeholder types (derived from `auto` and `decltype(auto)`) to be usable in a `new` expression with a single-clause //braced-init-list// as its initializer (8.3.4 [expr.new]/2).  N3922 defined its semantics.

References:
 http://wg21.link/p0620r0
 http://wg21.link/n3922

Reviewers: rsmith, aaron.ballman

Reviewed By: rsmith

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D39451

llvm-svn: 320401

7 years ago[X86] Add AND/OR/XOR schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 18:23:24 +0000 (18:23 +0000)]
[X86] Add AND/OR/XOR schedule tests

llvm-svn: 320400

7 years ago[dwarfdump] Fix off-by-one bug in accelerator table extractor.
Jonas Devlieghere [Mon, 11 Dec 2017 18:22:47 +0000 (18:22 +0000)]
[dwarfdump] Fix off-by-one bug in accelerator table extractor.

This fixes a bug where the verifier was complaining about empty
accelerator tables. When the table is empty, its size is not a valid
offset as it points after the end of the section.

This patch also makes the extractor return llvm:Error instead of bool
for better error reporting in the verifier.

Differential revision: https://reviews.llvm.org/D41063

rdar://35932007

llvm-svn: 320399

7 years agoRevert 320391: Certain targets are failing, pulling back to diagnose.
Erich Keane [Mon, 11 Dec 2017 18:14:51 +0000 (18:14 +0000)]
Revert 320391: Certain targets are failing, pulling back to diagnose.

llvm-svn: 320398

7 years ago[X86] Add ADD/SUB schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 18:13:40 +0000 (18:13 +0000)]
[X86] Add ADD/SUB schedule tests

llvm-svn: 320397

7 years ago[Sema] Fix crash in unused-lambda-capture warning for VLAs
Malcolm Parsons [Mon, 11 Dec 2017 18:00:36 +0000 (18:00 +0000)]
[Sema] Fix crash in unused-lambda-capture warning for VLAs

Summary:
Clang was crashing when diagnosing an unused-lambda-capture for a VLA because
From.getVariable() is null for the capture of a VLA bound.
Warning about the VLA bound capture is not helpful, so only warn for the VLA
itself.

Fixes: PR35555
Reviewers: aaron.ballman, dim, rsmith

Reviewed By: aaron.ballman, dim

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D41016

llvm-svn: 320396

7 years ago[X86] Add ADC/SBB schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 17:59:05 +0000 (17:59 +0000)]
[X86] Add ADC/SBB schedule tests

llvm-svn: 320395

7 years agoSort.
Rui Ueyama [Mon, 11 Dec 2017 17:52:43 +0000 (17:52 +0000)]
Sort.

llvm-svn: 320394

7 years agoChange function type more LLVM-ish.
Rui Ueyama [Mon, 11 Dec 2017 17:52:28 +0000 (17:52 +0000)]
Change function type more LLVM-ish.

Also fix variable naming style.

llvm-svn: 320393

7 years ago[X86] Add MOVSLQ schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 17:37:08 +0000 (17:37 +0000)]
[X86] Add MOVSLQ schedule tests

llvm-svn: 320392

7 years agoFor Linux/gnu compatibility, preinclude <stdc-predef.h> if the file is available
Erich Keane [Mon, 11 Dec 2017 17:36:42 +0000 (17:36 +0000)]
For Linux/gnu compatibility, preinclude <stdc-predef.h> if the file is available

As reported in llvm bugzilla 32377.
Here’s a patch to add preinclude of stdc-predef.h.

The gcc documentation says “On GNU/Linux, <stdc-predef.h> is pre-included.” See https://gcc.gnu.org/gcc-4.8/porting_to.html;

The preinclude is inhibited with –ffreestanding.

Basically I fixed the failing test cases by adding –ffreestanding which inhibits this behavior.

I fixed all the failing tests, including some in extra/test, there's a separate patch for that which is linked here

Note: this is a recommit after a test failure took down the original (r318669)

Patch By: mibintc
Differential Revision: https://reviews.llvm.org/D34158

llvm-svn: 320391

7 years agoCreate reserved symbols early so they can be versioned.
Rafael Espindola [Mon, 11 Dec 2017 17:23:28 +0000 (17:23 +0000)]
Create reserved symbols early so they can be versioned.

This fixes pr35570.

We were creating these symbols after parsing version scripts, so they
could not be versioned.

We cannot move the version script parsing later because we need it for
lto.

One option is to move both addReservedSymbols and
createSyntheticSections earlier. The disadvantage is that some
sections created by createSyntheticSections replace other input
sections. For example, gdb index replaces .debug_gnu_pubnames, so it
wants to run after gc sections so that it can set S->Live to false.

What this patch does instead is to move just the ElfHeader creation
early.

llvm-svn: 320390

7 years agoNormalize line endings. NFCI.
Simon Pilgrim [Mon, 11 Dec 2017 17:01:21 +0000 (17:01 +0000)]
Normalize line endings. NFCI.

llvm-svn: 320389

7 years ago[GlobalISel] Disable GISel for big endian.
Amara Emerson [Mon, 11 Dec 2017 16:58:29 +0000 (16:58 +0000)]
[GlobalISel] Disable GISel for big endian.

This is due to PR26161 needing to be resolved before we can fix
big endian bugs like PR35359. The work to split aggregates into smaller LLTs
instead of using one large scalar will take some time, so in the mean time
we'll fall back to SDAG.

Some ARM BE tests xfailed for now as a result.

Differential Revision: https://reviews.llvm.org/D40789

llvm-svn: 320388

7 years ago[X86] Add LWP schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 16:47:21 +0000 (16:47 +0000)]
[X86] Add LWP schedule tests

Tag LWP instructions as WriteSystem

llvm-svn: 320387

7 years ago[X86] Add INT/INTO schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 16:32:58 +0000 (16:32 +0000)]
[X86] Add INT/INTO schedule tests

llvm-svn: 320386

7 years ago[X86] Add IN/OUT schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 16:16:40 +0000 (16:16 +0000)]
[X86] Add IN/OUT schedule tests

llvm-svn: 320385

7 years ago[X86] Add IDIV schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 16:08:21 +0000 (16:08 +0000)]
[X86] Add IDIV schedule tests

llvm-svn: 320384

7 years ago[X86] Add CMPXCHG schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 16:04:08 +0000 (16:04 +0000)]
[X86] Add CMPXCHG schedule tests

llvm-svn: 320383

7 years ago[X86] Add CLZERO schedule test
Simon Pilgrim [Mon, 11 Dec 2017 15:53:12 +0000 (15:53 +0000)]
[X86] Add CLZERO schedule test

llvm-svn: 320382

7 years ago[MSan] Hotfix compilation
Alexander Potapenko [Mon, 11 Dec 2017 15:48:56 +0000 (15:48 +0000)]
[MSan] Hotfix compilation

For some reason the override directives got removed in r320373.
I suspect this to be an unwanted effect of clang-format.

llvm-svn: 320381

7 years ago[X86] Add ADCX/ADOX/XADD/XLAT schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 15:41:52 +0000 (15:41 +0000)]
[X86] Add ADCX/ADOX/XADD/XLAT schedule tests

llvm-svn: 320380

7 years ago[X86] Modify Nontemporal tests to avoid deadstore optimization.
Nirav Dave [Mon, 11 Dec 2017 15:35:40 +0000 (15:35 +0000)]
[X86] Modify Nontemporal tests to avoid deadstore optimization.

llvm-svn: 320379

7 years ago[AMDGPU] Rename Bonaire target to be gfx704; update target feature handling
Tony Tye [Mon, 11 Dec 2017 15:35:27 +0000 (15:35 +0000)]
[AMDGPU] Rename Bonaire target to be gfx704; update target feature handling

- Rename Bonaire target to be gfx704.
- Eliminate gfx800 and make Iceland and Tonga both use gfx802 as they use the same code.
- List target features supported by each processor in the processor table together with the default value.
- Add xnack flag to e_flags.
- Remove xnack from kernel metadata and kernel descriptor since it is now a whole code object property.

Differential Revision: https://reviews.llvm.org/D40051

llvm-svn: 320378

7 years agodotest.py: Correctly annotate lldbinline tests with debug info categories
Pavel Labath [Mon, 11 Dec 2017 15:28:21 +0000 (15:28 +0000)]
dotest.py: Correctly annotate lldbinline tests with debug info categories

This enables one to run all dwo tests with dotest.py --category dwo, or
skip them with --skip-category.

llvm-svn: 320377

7 years ago[X86] Add SETCC/STC/STD/UD2 schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 15:25:31 +0000 (15:25 +0000)]
[X86] Add SETCC/STC/STD/UD2 schedule tests

llvm-svn: 320376

7 years ago[AMDGPU][MC][GFX9] Corrected encoding of ttmp registers, disabled tba/tma
Dmitry Preobrazhensky [Mon, 11 Dec 2017 15:23:20 +0000 (15:23 +0000)]
[AMDGPU][MC][GFX9] Corrected encoding of ttmp registers, disabled tba/tma

See bugs 35494 and 35559:
https://bugs.llvm.org/show_bug.cgi?id=35494
https://bugs.llvm.org/show_bug.cgi?id=35559

Reviewers: vpykhtin, artem.tamazov, arsenm

Differential Revision: https://reviews.llvm.org/D41007

llvm-svn: 320375

7 years ago[DAGCombiner] protect against an infinite loop between shl <--> mul (PR35579)
Sanjay Patel [Mon, 11 Dec 2017 15:19:31 +0000 (15:19 +0000)]
[DAGCombiner] protect against an infinite loop between shl <--> mul (PR35579)

At first, I tried to thread the x86 needle and use a target hook (isVectorShiftByScalarCheap())
to disable the transform only for non-splat pow-of-2 constants, but not AVX2, but only some
element types, but...it's difficult.

Here we just avoid the loop with the x86 vector transform that conflicts with the general DAG
combine and preserve all of the existing behavior AFAICT otherwise.

Some tests that will probably fail if someone does try to restrict this in a more targeted way
for x86-only may be found in:

test/CodeGen/X86/combine-mul.ll
test/CodeGen/X86/vector-mul.ll
test/CodeGen/X86/widen_arith-5.ll

This should prevent the infinite looping seen with:
https://bugs.llvm.org/show_bug.cgi?id=35579

Differential Revision: https://reviews.llvm.org/D41040

llvm-svn: 320374

7 years ago[MSan] introduce getShadowOriginPtr(). NFC.
Alexander Potapenko [Mon, 11 Dec 2017 15:05:22 +0000 (15:05 +0000)]
[MSan] introduce getShadowOriginPtr(). NFC.

This patch introduces getShadowOriginPtr(), a method that obtains both the shadow and origin pointers for an address as a Value pair.
The existing callers of getShadowPtr() and getOriginPtr() are updated to use getShadowOriginPtr().

The rationale for this change is to simplify KMSAN instrumentation implementation.
In KMSAN origins tracking is always enabled, and there's no direct mapping between the app memory and the shadow/origin pages.
Both the shadow and the origin pointer for a given address are obtained by calling a single runtime hook from the instrumentation,
therefore it's easier to work with those pointers together.

Reviewed at https://reviews.llvm.org/D40835.

llvm-svn: 320373

7 years ago[ELF] Improve comments in aarch64 errata fix test [NFC]
Peter Smith [Mon, 11 Dec 2017 15:00:58 +0000 (15:00 +0000)]
[ELF] Improve comments in aarch64 errata fix test [NFC]

Comment improvements split out from review D36749. No changes to any non
comment line.

llvm-svn: 320372

7 years ago[X86] Add SAR/SHL/SHR schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 14:56:44 +0000 (14:56 +0000)]
[X86] Add SAR/SHL/SHR schedule tests

llvm-svn: 320371

7 years ago[X86] Add RCL/RCR schedule tests
Simon Pilgrim [Mon, 11 Dec 2017 14:46:42 +0000 (14:46 +0000)]
[X86] Add RCL/RCR schedule tests

llvm-svn: 320370

7 years ago[Hexagon] Crash in instruction selection for insert_vector_elt for HVX
Krzysztof Parzyszek [Mon, 11 Dec 2017 14:46:06 +0000 (14:46 +0000)]
[Hexagon] Crash in instruction selection for insert_vector_elt for HVX

A wrong type was passed to insertVector, causing an out-of-bounds value
to be added an an operand to HexagonISD::INSERT. This later failed in
instruction selection.

llvm-svn: 320369

7 years ago[PowerPC] Sign-extend negative constant stores
Nemanja Ivanovic [Mon, 11 Dec 2017 14:35:48 +0000 (14:35 +0000)]
[PowerPC] Sign-extend negative constant stores

Second part of https://reviews.llvm.org/D40348.
Revision r318436 has extended all constants feeding a store to 64 bits
to allow for CSE on the SDAG. However, negative constants were zero extended
which made the constant being loaded appear to be a positive value larger than
16 bits. This resulted in long sequences to materialize such constants
rather than simply a "load immediate". This patch just sign-extends those
updated constants so that they remain 16-bit signed immediates if they started
out that way.

llvm-svn: 320368

7 years agoAdd a StringList constructor to Args class
Pavel Labath [Mon, 11 Dec 2017 14:22:30 +0000 (14:22 +0000)]
Add a StringList constructor to Args class

Host::GetEnvironment returns a StringList, but the interface for
launching a process takes Args. The fact that we use two classes for
representing an environment is not ideal, but for now we should at least
have an easy way to convert between the two.

llvm-svn: 320366

7 years ago[DAGCombiner] Add combined indexed load to the work list
Nemanja Ivanovic [Mon, 11 Dec 2017 14:16:02 +0000 (14:16 +0000)]
[DAGCombiner] Add combined indexed load to the work list

This commit is the first part of https://reviews.llvm.org/D40348.
In order to allow target combines to be performed on newly combined
indexed loads, add them back to the worklist. The remainder of the
above patch will be committed in subsequent revisions and will use
this. Test cases will be included with those follow-up commits.

llvm-svn: 320365

7 years ago[libcxx] Define istream_iterator equality comparison operators out-of-line
Roger Ferrer Ibanez [Mon, 11 Dec 2017 13:54:58 +0000 (13:54 +0000)]
[libcxx] Define istream_iterator equality comparison operators out-of-line

Currently libc++ defines operator== and operator!= as friend functions in the
definition of the istream_iterator class template. Such definition has a subtle
difference from an out-of-line definition required by the C++ Standard: these
functions can only be found by argument-dependent lookup, but not by qualified
lookup.

This patch changes the definition, so that it conforms to the C++ Standard and
adds a check involving qualified lookup to the test suite.

Patch contributed by Mikhail Maltsev.

Differential Revision: https://reviews.llvm.org/D40415

llvm-svn: 320363

7 years ago[ARM GlobalISel] Add test for a MOVTi16 pattern. NFC
Diana Picus [Mon, 11 Dec 2017 13:28:45 +0000 (13:28 +0000)]
[ARM GlobalISel] Add test for a MOVTi16 pattern. NFC

Add test for matching an OR with 0xFFFF0000 to a MOVTi16.

llvm-svn: 320362

7 years ago[X86] Add fsgsbase schedule tests.
Simon Pilgrim [Mon, 11 Dec 2017 13:25:02 +0000 (13:25 +0000)]
[X86] Add fsgsbase schedule tests.

llvm-svn: 320361

7 years ago[ScopBuilder] Split statements on encountering store instructions.
Michael Kruse [Mon, 11 Dec 2017 12:51:24 +0000 (12:51 +0000)]
[ScopBuilder] Split statements on encountering store instructions.

Introduce -polly-stmt-granularity=store option.

Contributed-by: Nandini Singhal <cs15mtech01004@iith.ac.in>
Differential Revision: https://reviews.llvm.org/D37337

llvm-svn: 320360

7 years ago[RISCV] Add custom CC_RISCV calling convention and improved call support
Alex Bradbury [Mon, 11 Dec 2017 12:49:02 +0000 (12:49 +0000)]
[RISCV] Add custom CC_RISCV calling convention and improved call support

The TableGen-based calling convention definitions are inflexible, while
writing a function to implement the calling convention is very
straight-forward, and allows difficult cases to be handled more easily. With
this patch adds support for:
* Passing large scalars according to the RV32I calling convention
* Byval arguments
* Passing values on the stack when the argument registers are exhausted

The custom CC_RISCV calling convention is also used for returns.

This patch also documents the ABI lowering that a language frontend is
expected to perform. I would like to work to simplify these requirements over
time, but this will require further discussion within the LLVM community.

We add PendingArgFlags CCState, as a companion to PendingLocs.

The PendingLocs vector is used by a number of backends to handle arguments
that are split during legalisation. However CCValAssign doesn't keep track of
the original argument alignment. Therefore, add a PendingArgFlags vector which
can be used to keep track of the ISD::ArgFlagsTy for every value added to
PendingLocs.

Differential Revision: https://reviews.llvm.org/D39898

llvm-svn: 320359

7 years ago[RISCV] Allow lowering of dynamic_stackalloc, stacksave, stackrestore
Alex Bradbury [Mon, 11 Dec 2017 12:38:17 +0000 (12:38 +0000)]
[RISCV] Allow lowering of dynamic_stackalloc, stacksave, stackrestore

llvm-svn: 320358

7 years ago[RISCV] Implement prolog and epilog insertion
Alex Bradbury [Mon, 11 Dec 2017 12:34:11 +0000 (12:34 +0000)]
[RISCV] Implement prolog and epilog insertion

As frame pointer elimination isn't implemented until a later patch and we make
extensive use of update_llc_test_checks.py, this changes touches a lot of the
RISC-V tests.

Differential Revision: https://reviews.llvm.org/D39849

llvm-svn: 320357

7 years ago[X86] Regenerate fsgsbase intrinsic tests. NFCI.
Simon Pilgrim [Mon, 11 Dec 2017 12:22:15 +0000 (12:22 +0000)]
[X86] Regenerate fsgsbase intrinsic tests. NFCI.

llvm-svn: 320356

7 years ago[ARM] Use ADDCARRY / SUBCARRY
Roger Ferrer Ibanez [Mon, 11 Dec 2017 12:13:45 +0000 (12:13 +0000)]
[ARM] Use ADDCARRY / SUBCARRY

This is a preparatory step for D34515.

This change:
 - makes nodes ISD::ADDCARRY and ISD::SUBCARRY legal for i32
 - lowering is done by first converting the boolean value into the carry flag
   using (_, C) ← (ARMISD::ADDC R, -1) and converted back to an integer value
   using (R, _) ← (ARMISD::ADDE 0, 0, C). An ARMISD::ADDE between the two
   operations does the actual addition.
 - for subtraction, given that ISD::SUBCARRY second result is actually a
   borrow, we need to invert the value of the second operand and result before
   and after using ARMISD::SUBE. We need to invert the carry result of
   ARMISD::SUBE to preserve the semantics.
 - given that the generic combiner may lower ISD::ADDCARRY and
   ISD::SUBCARRYinto ISD::UADDO and ISD::USUBO we need to update their lowering
   as well otherwise i64 operations now would require branches. This implies
   updating the corresponding test for unsigned.
 - add new combiner to remove the redundant conversions from/to carry flags
   to/from boolean values (ARMISD::ADDC (ARMISD::ADDE 0, 0, C), -1) → C
 - fixes PR34045
 - fixes PR34564
 - fixes PR35103

Differential Revision: https://reviews.llvm.org/D35192

llvm-svn: 320355

7 years ago[mips] Minor update to the comment (NFC)
Aleksandar Beserminji [Mon, 11 Dec 2017 12:12:16 +0000 (12:12 +0000)]
[mips] Minor update to the comment (NFC)

llvm-svn: 320354

7 years ago[RISCV] Support lowering FrameIndex
Alex Bradbury [Mon, 11 Dec 2017 11:53:54 +0000 (11:53 +0000)]
[RISCV] Support lowering FrameIndex

Introduces the AddrFI "addressing mode", which is necessary simply because
it's not possible to write a pattern that directly matches a frameindex.

Ensure callee-saved registers are accessed relative to the stackpointer. This
is necessary as callee-saved register spills are performed before the frame
pointer is set.

Move HexagonDAGToDAGISel::isOrEquivalentToAdd to SelectionDAGISel, so we can
make use of it in the RISC-V backend.

Differential Revision: https://reviews.llvm.org/D39848

llvm-svn: 320353

7 years ago[ARM GlobalISel] Add tests for PKHBT and PKHTB
Diana Picus [Mon, 11 Dec 2017 11:44:23 +0000 (11:44 +0000)]
[ARM GlobalISel] Add tests for PKHBT and PKHTB

Test (some of) the patterns for selecting PKHBT and PKHTB. The others
are just very similar to the ones we're testing and there would be
little value in covering them as well.

llvm-svn: 320352

7 years ago[mips] Removal of microMIPS64R6
Aleksandar Beserminji [Mon, 11 Dec 2017 11:29:17 +0000 (11:29 +0000)]
[mips] Removal of microMIPS64R6

microMIPS64R6 is removed from backend, and therefore frontend
will show an error when target is microMIPS64R6.

This is Clang part of patch.

Differential Revision: https://reviews.llvm.org/D35624

llvm-svn: 320351

7 years ago[mips] Removal of microMIPS64R6
Aleksandar Beserminji [Mon, 11 Dec 2017 11:21:40 +0000 (11:21 +0000)]
[mips] Removal of microMIPS64R6

All files and parts of files related to microMIPS4R6 are removed.
When target is microMIPS4R6, errors are printed.

This is LLVM part of patch.

Differential Revision: https://reviews.llvm.org/D35625

llvm-svn: 320350

7 years agoFix osx build broken in r320346
Pavel Labath [Mon, 11 Dec 2017 11:12:22 +0000 (11:12 +0000)]
Fix osx build broken in r320346

llvm-svn: 320349

7 years ago[AVR] Implement some missing code paths
Dylan McKay [Mon, 11 Dec 2017 11:01:27 +0000 (11:01 +0000)]
[AVR] Implement some missing code paths

This has been broken since r320009.

llvm-svn: 320348

7 years ago[AVR] Fix incorrectly-calculated AVRMCExpr evaluations
Dylan McKay [Mon, 11 Dec 2017 11:01:19 +0000 (11:01 +0000)]
[AVR] Fix incorrectly-calculated AVRMCExpr evaluations

This has been broken since r320009.

llvm-svn: 320347

7 years agoMove PseudoTerminal to the lldb_private namespace
Pavel Labath [Mon, 11 Dec 2017 10:09:14 +0000 (10:09 +0000)]
Move PseudoTerminal to the lldb_private namespace

lldb_utility doesn't make sense, as it is no longer even living in the
"utility" module.

llvm-svn: 320346

7 years agoMainLoop: avoid infinite loop when pty slave gets closed
Pavel Labath [Mon, 11 Dec 2017 09:33:18 +0000 (09:33 +0000)]
MainLoop: avoid infinite loop when pty slave gets closed

Summary:
For ptys (at least on Linux), the end-of-file (closing of the slave FD)
is signalled by the POLLHUP flag. We were ignoring this flag, which
meant that when this happened, we would spin in a loop, continuously
calling poll(2) and not making any progress.

This makes sure we treat POLLHUP as a read event (reading will return
0), and we call the registered callback when it happens. This is the
behavior our clients expect (and is consistent with how select(2)
works).

Reviewers: eugene, beanz

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D41008

llvm-svn: 320345

7 years ago[DAGCombiner] Support folding (mulhs/u X, 0)->0 for vectors.
Craig Topper [Mon, 11 Dec 2017 08:33:20 +0000 (08:33 +0000)]
[DAGCombiner] Support folding (mulhs/u X, 0)->0 for vectors.

We should probably also fold (mulhs/u X, 1) for vectors, but that's harder.

llvm-svn: 320344

7 years ago[DAGCombiner] Reuse existing SDLoc variable instead of creating a new one. NFC
Craig Topper [Mon, 11 Dec 2017 08:33:19 +0000 (08:33 +0000)]
[DAGCombiner] Reuse existing SDLoc variable instead of creating a new one. NFC

llvm-svn: 320343

7 years ago[X86] Regenerate test with update_llc_test_checks.py
Craig Topper [Mon, 11 Dec 2017 06:16:26 +0000 (06:16 +0000)]
[X86] Regenerate test with update_llc_test_checks.py

llvm-svn: 320342

7 years agoRevert "[DataEncoder] Replace buggy versions of write functions."
Davide Italiano [Mon, 11 Dec 2017 05:09:35 +0000 (05:09 +0000)]
Revert "[DataEncoder] Replace buggy versions of write functions."

The commit exposes a bunch of failures in the LLDB testsuite that
I need to analyze more carefully. Reverting for now.

llvm-svn: 320341

7 years ago[X86] Add a test case for masked scatter where the index needs to be legalized from...
Craig Topper [Mon, 11 Dec 2017 01:48:10 +0000 (01:48 +0000)]
[X86] Add a test case for masked scatter where the index needs to be legalized from v2i32 while other types are legal.

llvm-svn: 320340

7 years agoFix test to test what it intended to test.
Rafael Espindola [Mon, 11 Dec 2017 01:06:39 +0000 (01:06 +0000)]
Fix test to test what it intended to test.

The test was added to test that we don't merge read only data
sections, but the sections in the test were rw.

llvm-svn: 320339

7 years ago[DarwinLog] Remove other dead code. Found while reading this file.
Davide Italiano [Sun, 10 Dec 2017 23:49:23 +0000 (23:49 +0000)]
[DarwinLog] Remove other dead code. Found while reading this file.

llvm-svn: 320338

7 years ago[MachException] Garbage collect unused and dead code.
Davide Italiano [Sun, 10 Dec 2017 23:33:33 +0000 (23:33 +0000)]
[MachException] Garbage collect unused and dead code.

llvm-svn: 320337

7 years ago[ScopBuilder] Fix typo. NFC.
Michael Kruse [Sun, 10 Dec 2017 22:56:32 +0000 (22:56 +0000)]
[ScopBuilder] Fix typo. NFC.

Contributed-by: Nandini Singhal <cs15mtech01004@iith.ac.in>
Differential Revision: https://reviews.llvm.org/D41047

llvm-svn: 320336

7 years ago[DataEncoder] Replace buggy versions of write functions.
Davide Italiano [Sun, 10 Dec 2017 22:54:07 +0000 (22:54 +0000)]
[DataEncoder] Replace buggy versions of write functions.

They cause an ubsan error when ran through the testsuite (store
to misaligned address is UB). This commit kills two birds with
one stone, as we also remove some code while fixing it.

<rdar://problem/35941757>

llvm-svn: 320335

7 years ago[X86] Add ROL/ROR schedule tests
Simon Pilgrim [Sun, 10 Dec 2017 22:11:56 +0000 (22:11 +0000)]
[X86] Add ROL/ROR schedule tests

llvm-svn: 320334

7 years ago[X86] Add DIV/MUL/NEG/NOP/NOT/PAUSE schedule tests
Simon Pilgrim [Sun, 10 Dec 2017 21:56:24 +0000 (21:56 +0000)]
[X86] Add DIV/MUL/NEG/NOP/NOT/PAUSE schedule tests

llvm-svn: 320333

7 years ago[X86] Add DEC/INC schedule tests
Simon Pilgrim [Sun, 10 Dec 2017 21:28:00 +0000 (21:28 +0000)]
[X86] Add DEC/INC schedule tests

Include i686 (non-REX) variant tests as well

llvm-svn: 320332

7 years ago[X86] Add INS/OUTS schedule tests
Simon Pilgrim [Sun, 10 Dec 2017 21:10:28 +0000 (21:10 +0000)]
[X86] Add INS/OUTS schedule tests

llvm-svn: 320331

7 years ago[X86] Add CMPS/MOVS/SCAS/STOS schedule tests
Simon Pilgrim [Sun, 10 Dec 2017 20:58:22 +0000 (20:58 +0000)]
[X86] Add CMPS/MOVS/SCAS/STOS schedule tests

llvm-svn: 320330

7 years ago[X86] Add CMOV schedule tests
Simon Pilgrim [Sun, 10 Dec 2017 20:46:57 +0000 (20:46 +0000)]
[X86] Add CMOV schedule tests

llvm-svn: 320329

7 years ago[X86] Add BT/BTC/BTR/BTS schedule tests
Simon Pilgrim [Sun, 10 Dec 2017 20:22:47 +0000 (20:22 +0000)]
[X86] Add BT/BTC/BTR/BTS schedule tests

llvm-svn: 320328

7 years agoMove RelaPlt and RelaIplt to InX. NFC.
Rafael Espindola [Sun, 10 Dec 2017 20:07:03 +0000 (20:07 +0000)]
Move RelaPlt and RelaIplt to InX. NFC.

llvm-svn: 320327

7 years ago[X86] Add VCOMISDZrr, VCOMISSZrr, VUCOMISDZrr, and VUCOMISSZrr to the skylake server...
Craig Topper [Sun, 10 Dec 2017 19:47:57 +0000 (19:47 +0000)]
[X86] Add VCOMISDZrr, VCOMISSZrr, VUCOMISDZrr, and VUCOMISSZrr to the skylake server sheduler model

llvm-svn: 320326

7 years ago[X86] Rename some instructions that start with Int_ to have the _Int at the end.
Craig Topper [Sun, 10 Dec 2017 19:47:56 +0000 (19:47 +0000)]
[X86] Rename some instructions that start with Int_ to have the _Int at the end.

This matches AVX512 version and is more consistent overall. And improves our scheduler models.

In some cases this adds _Int to instructions that didn't have any Int_ before. It's a side effect of the adjustments made to some of the multiclasses.

llvm-svn: 320325

7 years agoMove RelaDyn to InX. NFC.
Rafael Espindola [Sun, 10 Dec 2017 19:44:42 +0000 (19:44 +0000)]
Move RelaDyn to InX. NFC.

llvm-svn: 320324

7 years agoRemove this->. NFC.
Rafael Espindola [Sun, 10 Dec 2017 19:28:32 +0000 (19:28 +0000)]
Remove this->. NFC.

llvm-svn: 320323

7 years ago[X86][X87] Fix typo in znver1 FIST/FISTT schedule patterns
Simon Pilgrim [Sun, 10 Dec 2017 19:19:22 +0000 (19:19 +0000)]
[X86][X87] Fix typo in znver1 FIST/FISTT schedule patterns

llvm-svn: 320322

7 years ago[X86][X87] Add missing x87 scheduler tests
Simon Pilgrim [Sun, 10 Dec 2017 18:53:15 +0000 (18:53 +0000)]
[X86][X87] Add missing x87 scheduler tests

Split off some 'n' instruction versions to make it clearer when WAIT is being inserted

llvm-svn: 320321

7 years ago[X86] Rename some instructions from 'rb' to 'rrb' to make 'b' a proper suffix. Fix...
Craig Topper [Sun, 10 Dec 2017 17:42:44 +0000 (17:42 +0000)]
[X86] Rename some instructions from 'rb' to 'rrb' to make 'b' a proper suffix. Fix the scheduling information for some of them.

Some of the scheduling information was only present for the 'rb' version' and not the 'rr' version. Now we match 'rr(b?)'

llvm-svn: 320320

7 years ago[X86] Add VCVTQQ2PS to the skylake server scheduler models.
Craig Topper [Sun, 10 Dec 2017 17:42:43 +0000 (17:42 +0000)]
[X86] Add VCVTQQ2PS to the skylake server scheduler models.

llvm-svn: 320319

7 years ago[X86] Add VPMULLWZ256 to the skylake server scheduler model
Craig Topper [Sun, 10 Dec 2017 17:42:42 +0000 (17:42 +0000)]
[X86] Add VPMULLWZ256 to the skylake server scheduler model

llvm-svn: 320318

7 years ago[X86] Add 256/512-bit EVEX VPSADBW instructions to skylake server scheduler model.
Craig Topper [Sun, 10 Dec 2017 17:42:41 +0000 (17:42 +0000)]
[X86] Add 256/512-bit EVEX VPSADBW instructions to skylake server scheduler model.

llvm-svn: 320317

7 years ago[X86] Fix a few instructions that were named Z512 instead of just Z.
Craig Topper [Sun, 10 Dec 2017 17:42:39 +0000 (17:42 +0000)]
[X86] Fix a few instructions that were named Z512 instead of just Z.

This makes things consistent with our normal instruction naming.

llvm-svn: 320316

7 years ago[X86] Add VPSRLWZrr to skylake server scheduler model.
Craig Topper [Sun, 10 Dec 2017 17:42:38 +0000 (17:42 +0000)]
[X86] Add VPSRLWZrr to skylake server scheduler model.

llvm-svn: 320315

7 years ago[X86] Add VPUNPCKLWDZrr to skylake server scheduler model.
Craig Topper [Sun, 10 Dec 2017 17:42:37 +0000 (17:42 +0000)]
[X86] Add VPUNPCKLWDZrr to skylake server scheduler model.

llvm-svn: 320314

7 years ago[X86] Adjust tablegen includes so we can use Instructions in scheduler models instead...
Craig Topper [Sun, 10 Dec 2017 17:42:36 +0000 (17:42 +0000)]
[X86] Adjust tablegen includes so we can use Instructions in scheduler models instead of just instregexs.

This separates the CPU specific scheduler model includes to occur after the instructions. Moves the instruction includes between the basic scheduler information and the CPU specific scheduler models.

llvm-svn: 320313

7 years ago[SimplifyLibCalls] propagate FMF when folding pow(x, -1.0) call
Sanjay Patel [Sun, 10 Dec 2017 17:25:54 +0000 (17:25 +0000)]
[SimplifyLibCalls] propagate FMF when folding pow(x, -1.0) call

Follow-up for a bug that's similar to:
https://bugs.llvm.org/show_bug.cgi?id=35601

llvm-svn: 320312

7 years ago[InstCombine] add test for pow(x, -1.0) with FMF; NFC
Sanjay Patel [Sun, 10 Dec 2017 17:21:51 +0000 (17:21 +0000)]
[InstCombine] add test for pow(x, -1.0) with FMF; NFC

llvm-svn: 320311

7 years ago[SimplifyLibCalls] propagate FMF when folding pow(x, 2.0) call (PR35601)
Sanjay Patel [Sun, 10 Dec 2017 16:52:26 +0000 (16:52 +0000)]
[SimplifyLibCalls] propagate FMF when folding pow(x, 2.0) call (PR35601)

This should fix the larger problem with sqrt shown in:
https://bugs.llvm.org/show_bug.cgi?id=35601

llvm-svn: 320310

7 years ago[InstCombine] add test for pow(x, 2.0) with FMF; NFC
Sanjay Patel [Sun, 10 Dec 2017 16:43:34 +0000 (16:43 +0000)]
[InstCombine] add test for pow(x, 2.0) with FMF; NFC

llvm-svn: 320309

7 years ago[X86] Flag BroadWell scheduler model as complete
Simon Pilgrim [Sun, 10 Dec 2017 13:49:51 +0000 (13:49 +0000)]
[X86] Flag BroadWell scheduler model as complete

Locally tag COPY as WriteMove, which has caused some reg-reg + reg-mem instruction tests to reorder.

llvm-svn: 320308

7 years agoRegenerate some AVX2+ scheduling tests that got missed
Simon Pilgrim [Sun, 10 Dec 2017 13:41:29 +0000 (13:41 +0000)]
Regenerate some AVX2+ scheduling tests that got missed

llvm-svn: 320307

7 years agoStrip trailing whitespace. NFCI.
Simon Pilgrim [Sun, 10 Dec 2017 13:00:37 +0000 (13:00 +0000)]
Strip trailing whitespace. NFCI.

llvm-svn: 320306

7 years agoRegenerate some scheduling tests that got missed
Simon Pilgrim [Sun, 10 Dec 2017 12:59:55 +0000 (12:59 +0000)]
Regenerate some scheduling tests that got missed

llvm-svn: 320305

7 years ago[X86] Flag ZNVER1 scheduler model as complete
Simon Pilgrim [Sun, 10 Dec 2017 12:43:53 +0000 (12:43 +0000)]
[X86] Flag ZNVER1 scheduler model as complete

We just have to locally tag COPY as WriteMove

llvm-svn: 320304

7 years ago[X86] Flag SLM scheduler model as complete
Simon Pilgrim [Sun, 10 Dec 2017 12:36:29 +0000 (12:36 +0000)]
[X86] Flag SLM scheduler model as complete

We just have to locally tag COPY as WriteMove

llvm-svn: 320303

7 years ago[X86][AVX[ Tag VZEROALL/VZEROUPPER instructions scheduler classes
Simon Pilgrim [Sun, 10 Dec 2017 12:26:35 +0000 (12:26 +0000)]
[X86][AVX[ Tag VZEROALL/VZEROUPPER instructions scheduler classes

llvm-svn: 320302

7 years ago[X86] Tag SSE4A instructions as SSE INTALU scheduler classes
Simon Pilgrim [Sun, 10 Dec 2017 12:08:04 +0000 (12:08 +0000)]
[X86] Tag SSE4A instructions as SSE INTALU scheduler classes

llvm-svn: 320301

7 years ago[X86] Flag BTVER2 scheduler model as complete
Simon Pilgrim [Sun, 10 Dec 2017 11:51:29 +0000 (11:51 +0000)]
[X86] Flag BTVER2 scheduler model as complete

We just have to locally tag COPY as WriteMove

llvm-svn: 320300