platform/upstream/llvm.git
5 years ago[NFC][InstCombine] Better commutative tests for "shift amount reassociation in bittes...
Roman Lebedev [Mon, 1 Jul 2019 14:28:24 +0000 (14:28 +0000)]
[NFC][InstCombine] Better commutative tests for "shift amount reassociation in bittest" pattern.

As discussed in https://reviews.llvm.org/D63829
*if* *both* shifts are one-use, we'd most likely want to produce `lshr`,
and not rely on ordering.

Also, there should likely be a *separate* fold to do this reordering.

llvm-svn: 364772

5 years ago[ASTImporter] Mark erroneous nodes in from ctx
Gabor Marton [Mon, 1 Jul 2019 14:19:53 +0000 (14:19 +0000)]
[ASTImporter] Mark erroneous nodes in from ctx

Summary:
During import of a specific Decl D, it may happen that some AST nodes
had already been created before we recognize an error. In this case we
signal back the error to the caller, but the "to" context remains
polluted with those nodes which had been created. Ideally, those nodes
should not had been created, but that time we did not know about the
error, the error happened later.  Since the AST is immutable (most of
the cases we can't remove existing nodes) we choose to mark these nodes
as erroneous.
Here are the steps of the algorithm:
1) We keep track of the nodes which we visit during the import of D: See
ImportPathTy.
2) If a Decl is already imported and it is already on the import path
(we have a cycle) then we copy/store the relevant part of the import
path. We store these cycles for each Decl.
3) When we recognize an error during the import of D then we set up this
error to all Decls in the stored cycles for D and we clear the stored
cycles.

Reviewers: a_sidorin, a.sidorin, shafik

Subscribers: rnkovacs, dkrupp, Szelethus, gamesh411, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D62375

llvm-svn: 364771

5 years ago[Hexagon] Rework VLCR algorithm
Krzysztof Parzyszek [Mon, 1 Jul 2019 13:50:47 +0000 (13:50 +0000)]
[Hexagon] Rework VLCR algorithm

Add code to catch pattern for commutative instructions for VLCR.

Patch by Suyog Sarda.

llvm-svn: 364770

5 years agoAMDGPU: Convert some places to Register
Matt Arsenault [Mon, 1 Jul 2019 13:44:46 +0000 (13:44 +0000)]
AMDGPU: Convert some places to Register

llvm-svn: 364769

5 years agoAMDGPU/GlobalISel: Fix RegBankSelect for G_FCANONICALIZE
Matt Arsenault [Mon, 1 Jul 2019 13:40:18 +0000 (13:40 +0000)]
AMDGPU/GlobalISel: Fix RegBankSelect for G_FCANONICALIZE

llvm-svn: 364768

5 years agoAMDGPU/GlobalISel: Fix RegBankSelect for G_BUILD_VECTOR
Matt Arsenault [Mon, 1 Jul 2019 13:40:17 +0000 (13:40 +0000)]
AMDGPU/GlobalISel: Fix RegBankSelect for G_BUILD_VECTOR

llvm-svn: 364767

5 years agoAMDGPU/GlobalISel: Fail on store to 32-bit address space
Matt Arsenault [Mon, 1 Jul 2019 13:37:39 +0000 (13:37 +0000)]
AMDGPU/GlobalISel: Fail on store to 32-bit address space

llvm-svn: 364766

5 years agoAMDGPU/GlobalISel: Improve icmp selection coverage.
Matt Arsenault [Mon, 1 Jul 2019 13:34:26 +0000 (13:34 +0000)]
AMDGPU/GlobalISel: Improve icmp selection coverage.

Select s64 eq/ne scalar icmp.

llvm-svn: 364765

5 years ago[NFC][InstCombine] Improve test coverage for ((~x) + y) + 1 -> y - x fold fold ...
Roman Lebedev [Mon, 1 Jul 2019 13:31:06 +0000 (13:31 +0000)]
[NFC][InstCombine] Improve test coverage for  ((~x) + y) + 1 -> y - x fold fold (PR42459)

So we indeed to have this fold, but only if +1 is not the last operation..

llvm-svn: 364764

5 years agoAMDGPU/GlobalISel: RegBankSelect for WWM/WQM
Matt Arsenault [Mon, 1 Jul 2019 13:30:12 +0000 (13:30 +0000)]
AMDGPU/GlobalISel: RegBankSelect for WWM/WQM

llvm-svn: 364763

5 years agoAMDGPU/GlobalISel: Use vcc reg bank for amdgcn.wqm.vote
Matt Arsenault [Mon, 1 Jul 2019 13:30:09 +0000 (13:30 +0000)]
AMDGPU/GlobalISel: Use vcc reg bank for amdgcn.wqm.vote

llvm-svn: 364762

5 years agoAMDGPU/GlobalISel: Fix scc->vcc copy handling
Matt Arsenault [Mon, 1 Jul 2019 13:22:07 +0000 (13:22 +0000)]
AMDGPU/GlobalISel: Fix scc->vcc copy handling

This was checking the size of the register with the value of the size,
which happens to be exec. Also fix assuming VCC is 64-bit to fix
wave32.

Also remove some untested handling for physical registers which is
skipped. This doesn't insert the V_CNDMASK_B32 if SCC is the physical
copy source. I'm not sure if this should be trying to handle this
special case instead of dealing with this in copyPhysReg.

llvm-svn: 364761

5 years agoAMDGPU/GlobalISel: Use and instead of BFE with inline immediate
Matt Arsenault [Mon, 1 Jul 2019 13:22:06 +0000 (13:22 +0000)]
AMDGPU/GlobalISel: Use and instead of BFE with inline immediate

Zext from s1 is the only case where this should do anything with the
current legal extensions.

llvm-svn: 364760

5 years agoGlobalISel: Add GINodeEquiv for min/max
Matt Arsenault [Mon, 1 Jul 2019 13:22:04 +0000 (13:22 +0000)]
GlobalISel: Add GINodeEquiv for min/max

llvm-svn: 364759

5 years agoGlobalISel: Add DAG compat for G_FCANONICALIZE
Matt Arsenault [Mon, 1 Jul 2019 13:22:00 +0000 (13:22 +0000)]
GlobalISel: Add DAG compat for G_FCANONICALIZE

llvm-svn: 364758

5 years ago[mips] Add missing schedinfo for MSA and ASE instructions
Simon Atanasyan [Mon, 1 Jul 2019 13:21:05 +0000 (13:21 +0000)]
[mips] Add missing schedinfo for MSA and ASE instructions

llvm-svn: 364757

5 years ago[mips] Add missing schedinfo for atomic instructions
Simon Atanasyan [Mon, 1 Jul 2019 13:20:56 +0000 (13:20 +0000)]
[mips] Add missing schedinfo for atomic instructions

llvm-svn: 364756

5 years ago[mips] Add missing schedinfo for ADJCALLSTACKDOWN, ADJCALLSTACKUP
Simon Atanasyan [Mon, 1 Jul 2019 13:20:48 +0000 (13:20 +0000)]
[mips] Add missing schedinfo for ADJCALLSTACKDOWN, ADJCALLSTACKUP

llvm-svn: 364755

5 years agoDon't check the validity of newly contructed data buffers
Pavel Labath [Mon, 1 Jul 2019 13:18:19 +0000 (13:18 +0000)]
Don't check the validity of newly contructed data buffers

A bunch of places were checking that DataBufferHeap::GetBytes returns a
non-null pointer right after constructing it. The only time when
GetBytes returns a null pointer is if it is empty (and I'm not sure that
even this is a good idea), but that is clearly not the case here, as the
buffer was constructed with a non-zero size just a couple of lines back.

llvm-svn: 364754

5 years ago@skipIfXmlSupportMissing TestRecognizeBreakpoint
Pavel Labath [Mon, 1 Jul 2019 13:12:29 +0000 (13:12 +0000)]
@skipIfXmlSupportMissing TestRecognizeBreakpoint

llvm-svn: 364753

5 years ago[ASTImporter] Propagate error from ImportDeclContext
Gabor Marton [Mon, 1 Jul 2019 12:44:39 +0000 (12:44 +0000)]
[ASTImporter] Propagate error from ImportDeclContext

Summary:
During analysis of one project we failed to import one
CXXDestructorDecl. But since we did not propagate the error in
importDeclContext we had a CXXRecordDecl without a destructor. Then the
analyzer engine had a CallEvent where the nonexistent dtor was requested
(crash).

Solution is to propagate the errors we have during importing a
DeclContext.

Reviewers: a_sidorin, a.sidorin, shafik

Subscribers: rnkovacs, dkrupp, Szelethus, gamesh411, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D63603

llvm-svn: 364752

5 years agoRevert "Implement xfer:libraries-svr4:read packet"
Pavel Labath [Mon, 1 Jul 2019 12:41:20 +0000 (12:41 +0000)]
Revert "Implement xfer:libraries-svr4:read packet"

D62502, together with D62503 have broken the builds which have XML
support enabled. Reverting D62503 (r364355) fixed that, but has broken
has left some of the tests introduced by D62502 broken more or less
nondeternimistically (it depended on whether the system happens to place
the library list near unreadable pages of memory). I attempted to make a
partial fix for this in r364748, but Jan Kratochvil pointed out that
this reintroduces the problem which reverting D62503 was trying to
solve.

So instead, I back out the whole thing so we can get back to a clean
slate that works for everyone. We can figure out a way forward from
there.

This reverts r364748, r363772 and r363707.

llvm-svn: 364751

5 years ago[AMDGPU] Call isLoopExiting for blocks in the loop.
Florian Hahn [Mon, 1 Jul 2019 12:36:44 +0000 (12:36 +0000)]
[AMDGPU] Call isLoopExiting for blocks in the loop.

isLoopExiting should only be called for blocks in the loop. A follow
up patch makes this requirement an assertion.

I've updated the usage here, to only match for actual exit blocks. Previously,
it would also match blocks not in the loop.

Reviewers: arsenm, nhaehnle

Reviewed By: nhaehnle

Differential Revision: https://reviews.llvm.org/D63980

llvm-svn: 364750

5 years ago[NFC][InstCombine] Tests for ((~x) + y) + 1 -> y - x fold fold (PR42459)
Roman Lebedev [Mon, 1 Jul 2019 12:22:06 +0000 (12:22 +0000)]
[NFC][InstCombine] Tests for  ((~x) + y) + 1 -> y - x fold fold (PR42459)

To be noted, this pattern is not unhandled by instcombine per-se,
it is somehow does end up being folded when one runs opt -O3,
but not if it's just -instcombine. Regardless, that fold is
indirect, depends on some other folds, and is thus blind
when there are extra uses.

https://bugs.llvm.org/show_bug.cgi?id=42459
https://rise4fun.com/Alive/EPO0

llvm-svn: 364749

5 years agoFix TestGdbRemoteLibrariesSvr4Support
Pavel Labath [Mon, 1 Jul 2019 12:00:25 +0000 (12:00 +0000)]
Fix TestGdbRemoteLibrariesSvr4Support

D62502 had a bug (visible only with D62503 reverted), where it would
error out if attempting to read a string from memory and the memory page
after the string happened to be unmapped.

This fixes the problem by checking for whether ReadMemory read *any*
bytes, instead of checking whether it returned an error. A greater
question is whether ReadMemory should even return an error if it read at
least one byte, but I'm leaving that for a separate patch.

llvm-svn: 364748

5 years ago[clangd] No longer getting template instantiations from header files in Main AST.
Johan Vikstrom [Mon, 1 Jul 2019 11:49:01 +0000 (11:49 +0000)]
[clangd] No longer getting template instantiations from header files in Main AST.

Previous implementation to filter decls not in the main file did not
work in the case where a template was instantiated from a header in the
main file. It would than include that function/class in topLevelDecls.

Differential Revision: https://reviews.llvm.org/D63817

llvm-svn: 364747

5 years ago[RISCV] Add break; to the last switch case
Fangrui Song [Mon, 1 Jul 2019 11:41:07 +0000 (11:41 +0000)]
[RISCV] Add break; to the last switch case

As suggested by jrtc27 in the post-commit review of D60528.

llvm-svn: 364746

5 years ago[X86] CombineShuffleWithExtract - updated description comments. NFCI.
Simon Pilgrim [Mon, 1 Jul 2019 11:33:45 +0000 (11:33 +0000)]
[X86] CombineShuffleWithExtract - updated description comments. NFCI.

CombineShuffleWithExtract no longer requires that both shuffle ops are extract_subvectors, from the same type or from the same size.

llvm-svn: 364745

5 years agoRemove null checks of results of new expressions
Pavel Labath [Mon, 1 Jul 2019 11:09:15 +0000 (11:09 +0000)]
Remove null checks of results of new expressions

operator new doesn't return a null pointer, even if one turns off
exceptions (it calls std::terminate instead). Therefore, all of this is
dead code.

llvm-svn: 364744

5 years ago[SelectionDAG] Do minnum->minimum at legalization time instead of building time
Benjamin Kramer [Mon, 1 Jul 2019 11:00:23 +0000 (11:00 +0000)]
[SelectionDAG] Do minnum->minimum at legalization time instead of building time

The SDAGBuilder behavior stems from the days when we didn't have fast
math flags available in SDAG. We do now and doing the transformation in
the legalizer has the advantage that it also works for vector types.

llvm-svn: 364743

5 years ago[benchmark] Disable CMake get_git_version
Andrew Ng [Mon, 1 Jul 2019 10:58:20 +0000 (10:58 +0000)]
[benchmark] Disable CMake get_git_version

Disabled CMake get_git_version as it is meaningless for this in-tree
build, and hardcoded a null version.

Not using get_git_version avoids a refresh of the git index that is
executed by get_git_version. Refreshing the index can take a
considerable amount of time if the index needs to be refreshed
(particularly with the mono repo). This situation can arise when
building shared source on a host in VMs.

Differential Revision: https://reviews.llvm.org/D63925

llvm-svn: 364742

5 years ago[clangd] Fix unused var from r364735
Sam McCall [Mon, 1 Jul 2019 10:13:03 +0000 (10:13 +0000)]
[clangd] Fix unused var from r364735

llvm-svn: 364741

5 years ago[clangd] Make PreambleStatusCache handle filenames more carefully
Sam McCall [Mon, 1 Jul 2019 10:11:18 +0000 (10:11 +0000)]
[clangd] Make PreambleStatusCache handle filenames more carefully

Summary:
 - when we hit the cache, the reported filename should be that of the
   cache query, not that of the cache store. This matches behaviors of
   common FSes, and avoids triggering difficult edge cases in
   FileManager when files are being moved around concurrently.
 - filename comparisons (both cache queries and == mainfile checks)
   should fold away . and .. in paths. These can appear when relative
   paths occur in compile_commands.json. (gn does this).

Reviewers: kadircet

Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D63931

llvm-svn: 364740

5 years ago[NFC][InstCombine] Tests for x - ~(y) -> x + y + 1 fold (PR42457)
Roman Lebedev [Mon, 1 Jul 2019 09:57:53 +0000 (09:57 +0000)]
[NFC][InstCombine] Tests for x - ~(y)  ->  x + y + 1 fold (PR42457)

https://bugs.llvm.org/show_bug.cgi?id=42457
https://rise4fun.com/Alive/iFhE

llvm-svn: 364739

5 years ago[InstCombine] Omit 'urem' where possible
Roman Lebedev [Mon, 1 Jul 2019 09:41:43 +0000 (09:41 +0000)]
[InstCombine] Omit 'urem' where possible

This was added in D63390 / rL364286 to backend,
but it makes sense to also handle it in middle-end.
https://rise4fun.com/Alive/Zsln

llvm-svn: 364738

5 years ago[NFC][InstCombine] Copy test for omit urem when possible from TargetLowering
Roman Lebedev [Mon, 1 Jul 2019 09:41:27 +0000 (09:41 +0000)]
[NFC][InstCombine] Copy test for omit urem when possible from TargetLowering

Was added in D63390 / rL364286 to backend, but it makes sense to also handle it here.
https://rise4fun.com/Alive/Zsln

llvm-svn: 364737

5 years ago[DebugInfo] Avoid adding too much indirection to pointer-valued variables
Jeremy Morse [Mon, 1 Jul 2019 09:38:23 +0000 (09:38 +0000)]
[DebugInfo] Avoid adding too much indirection to pointer-valued variables

This patch addresses PR41675, where a stack-pointer variable is dereferenced
too many times by its location expression, presenting a value on the stack as
the pointer to the stack.

The difference between a stack *pointer* DBG_VALUE and one that refers to a
value on the stack, is currently the indirect flag. However the DWARF backend
will also try to guess whether something is a memory location or not, based
on whether there is any computation in the location expression. By simply
prepending the stack offset to existing expressions, we can accidentally
convert a register location into a memory location, which introduces a
suprise (and unintended) dereference.

The solution is to add DW_OP_stack_value whenever we add a DIExpression
computation to a stack *pointer*. It's an implicit location computed on the
expression stack, thus needs to be flagged as a stack_value.

For the edge case where the offset is zero and the location could be a register
location, DIExpression::prepend will still generate opcodes, and thus
DW_OP_stack_value must still be added.

Differential Revision: https://reviews.llvm.org/D63429

llvm-svn: 364736

5 years ago[clangd] Show better message when we rename macros.
Haojian Wu [Mon, 1 Jul 2019 09:26:48 +0000 (09:26 +0000)]
[clangd] Show better message when we rename macros.

Summary:
Previously, when we rename a macro, we get an error message of "there is
no symbol found".

This patch improves the message of this case (as we don't support macros).

Reviewers: sammccall

Subscribers: ilya-biryukov, MaskRay, jkorous, mgrang, arphaman, kadircet, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D63922

llvm-svn: 364735

5 years ago[SimpleLoopUnswitch] Implement handling of prof branch_weights metadata for SwitchInst
Yevgeny Rouban [Mon, 1 Jul 2019 08:43:53 +0000 (08:43 +0000)]
[SimpleLoopUnswitch] Implement handling of prof branch_weights metadata for SwitchInst

Differential Revision: https://reviews.llvm.org/D60606

llvm-svn: 364734

5 years ago[ARM] WLS/LE Code Generation
Sam Parker [Mon, 1 Jul 2019 08:21:28 +0000 (08:21 +0000)]
[ARM] WLS/LE Code Generation

Backend changes to enable WLS/LE low-overhead loops for armv8.1-m:
1) Use TTI to communicate to the HardwareLoop pass that we should try
   to generate intrinsics that guard the loop entry, as well as setting
   the loop trip count.
2) Lower the BRCOND that uses said intrinsic to an Arm specific node:
   ARMWLS.
3) ISelDAGToDAG the node to a new pseudo instruction:
   t2WhileLoopStart.
4) Add support in ArmLowOverheadLoops to handle the new pseudo
   instruction.

Differential Revision: https://reviews.llvm.org/D63816

llvm-svn: 364733

5 years ago[libcxx] [test] Add void cast to result of compare_exchange_weak to suppress [[nodisc...
Billy Robert O'Neal III [Mon, 1 Jul 2019 08:09:02 +0000 (08:09 +0000)]
[libcxx] [test] Add void cast to result of compare_exchange_weak to suppress [[nodiscard]].

llvm-svn: 364732

5 years ago[clangd] Make FixIt message be consistent with the clang-tidy diagnostic message.
Haojian Wu [Mon, 1 Jul 2019 08:05:53 +0000 (08:05 +0000)]
[clangd] Make FixIt message be consistent with the clang-tidy diagnostic message.

Summary:
We strip the "[clang-tidy-check]" suffix from the clang-tidy diagnostics, we
should be consistent with the message in FixIt (strip the suffix as well).

Reviewers: sammccall

Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, kadircet, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D63926

llvm-svn: 364731

5 years ago[X86] Add more load folding tests for vcvt(t)ps2(u)qq showing missed foldings. NFC
Craig Topper [Mon, 1 Jul 2019 07:59:42 +0000 (07:59 +0000)]
[X86] Add more load folding tests for vcvt(t)ps2(u)qq showing missed foldings. NFC

llvm-svn: 364730

5 years ago[X86] Improve the type checking fast-isel handling of vector bitcasts.
Craig Topper [Mon, 1 Jul 2019 07:09:34 +0000 (07:09 +0000)]
[X86] Improve the type checking fast-isel handling of vector bitcasts.

We had a bunch of vector size legality checks for the source type
based on feature flags, but we didn't check the destination type at
all beyond ensuring that it was a "simple" type. But this allowed
the destination to be i128 which isn't legal.

This commit changes the code to use TLI's isTypeLegal logic in
place of the all the subtarget checks. Then additionally checks
that the source and dest are vectors.

Fixes 42452

llvm-svn: 364729

5 years ago[X86] Add a DAG combine to replace vector loads feeding a v4i32->v2f64 CVTSI2FP/CVTUI...
Craig Topper [Mon, 1 Jul 2019 07:09:31 +0000 (07:09 +0000)]
[X86] Add a DAG combine to replace vector loads feeding a v4i32->v2f64 CVTSI2FP/CVTUI2FP node with a vzload.

But only when the load isn't volatile.

This improves load folding during isel where we only have vzload
and scalar_to_vector+load patterns. We can't have full vector load
isel patterns for the same volatile load issue.

Also add some missing masked cvtsi2fp/cvtui2fp with vzload patterns.

llvm-svn: 364728

5 years ago[X86] Add some additional load folding tests to vec_int_to_fp.ll/vec_int_to_fp-widen...
Craig Topper [Mon, 1 Jul 2019 07:09:26 +0000 (07:09 +0000)]
[X86] Add some additional load folding tests to vec_int_to_fp.ll/vec_int_to_fp-widen.ll and disable the peephole pass.

Also copy some missing test cases from vec_int_to_fp.ll to vec_int_to_fp-widen.ll

llvm-svn: 364727

5 years ago[X86] Add MOVHPDrm/MOVLPDrm patterns that use VZEXT_LOAD.
Craig Topper [Mon, 1 Jul 2019 07:09:23 +0000 (07:09 +0000)]
[X86] Add MOVHPDrm/MOVLPDrm patterns that use VZEXT_LOAD.

We already had patterns that used scalar_to_vector+load. But we can
also have a vzload.

Found while investigating combining scalar_to_vector+load to vzload.

llvm-svn: 364726

5 years agoVarious tweaks to MSVC natvis visualizers
Mike Spertus [Sun, 30 Jun 2019 22:04:25 +0000 (22:04 +0000)]
Various tweaks to MSVC natvis visualizers

Make more consistent use of na format.
Improve visualization of deduction guides.
Add visualizer for explicit specifier (including conditionally explicit)
Fix some typos

llvm-svn: 364724

5 years agoClean up MSVC visualization of LLVM pointer types
Mike Spertus [Sun, 30 Jun 2019 21:54:34 +0000 (21:54 +0000)]
Clean up MSVC visualization of LLVM pointer types

Create separate natvis ptr and int views for PointerIntPair.
These are convenient in watch Windows and will be used by
Clang visualizers to be checked in shortly

Also, removed deref views as the MSVC na format has
done the same thing natively since MSVC2013.

llvm-svn: 364723

5 years agoRevert "[GDBRemote] Remove code that flushes GDB remote packets"
Jonas Devlieghere [Sun, 30 Jun 2019 19:00:09 +0000 (19:00 +0000)]
Revert "[GDBRemote] Remove code that flushes GDB remote packets"

Reverting this again as it doesn't appear to solve the flakiness on the
LLDB standalone bot.

llvm-svn: 364722

5 years ago[InstCombine] canonicalize fcmp+select to minnum/maxnum intrinsics
Sanjay Patel [Sun, 30 Jun 2019 13:40:31 +0000 (13:40 +0000)]
[InstCombine] canonicalize fcmp+select to minnum/maxnum intrinsics

This is the opposite direction of D62158 (we have to choose 1 form or the other).
Now that we have FMF on the select, this becomes more palatable. And the benefits
of having a single IR instruction for this operation (less chances of missing folds
based on extra uses, etc) overcome my previous comments about the potential advantage
of larger pattern matching/analysis.

Differential Revision: https://reviews.llvm.org/D62414

llvm-svn: 364721

5 years agoCleanup: llvm::bsearch -> llvm::partition_point after r364719
Fangrui Song [Sun, 30 Jun 2019 11:19:56 +0000 (11:19 +0000)]
Cleanup: llvm::bsearch -> llvm::partition_point after r364719

llvm-svn: 364720

5 years ago[ADT] Implement llvm::bsearch() with std::partition_point()
Fangrui Song [Sun, 30 Jun 2019 09:17:59 +0000 (09:17 +0000)]
[ADT] Implement llvm::bsearch() with std::partition_point()

Summary:
Delete the begin-end form because the standard std::partition_point
can be easily used as a replacement.

The ranges-style llvm::bsearch will be renamed to llvm::partition_point
in the next clean-up patch.

The name "bsearch" doesn't meet people's expectation because in C:

> If two or more members compare equal, which member is returned is unspecified.

Reviewed By: sammccall

Differential Revision: https://reviews.llvm.org/D63718

llvm-svn: 364719

5 years ago[X86] Custom lower AVX masked loads to masked load and vselect instead of selecting...
Craig Topper [Sun, 30 Jun 2019 06:46:37 +0000 (06:46 +0000)]
[X86] Custom lower AVX masked loads to masked load and vselect instead of selecting a maskmov+vblend during isel.

AVX masked loads only support 0 as the value for masked off elements.
So we need an extra blend to support other values. Previously we
expanded the masked load to two instructions with isel patterns.
With this patch we now insert the vselect during lowering and it
will be separately selected as a blend.

llvm-svn: 364718

5 years ago[SelectionDAG] Use the memory VT instead of result VT for FoldingSet profiling in...
Craig Topper [Sun, 30 Jun 2019 06:46:33 +0000 (06:46 +0000)]
[SelectionDAG] Use the memory VT instead of result VT for FoldingSet profiling in getMaskedLoad/getMaskedStore.

This matches what is done by the Profile function. Otherwise CSE
won't work properly.

llvm-svn: 364717

5 years agoReplace tabs with spaces.
Adrian Prantl [Sat, 29 Jun 2019 18:32:16 +0000 (18:32 +0000)]
Replace tabs with spaces.

llvm-svn: 364716

5 years ago[LFTR] Rephrase getLoopTest into "based-on" check; NFCI
Nikita Popov [Sat, 29 Jun 2019 15:12:59 +0000 (15:12 +0000)]
[LFTR] Rephrase getLoopTest into "based-on" check; NFCI

What we want to know here is whether we're already using this value
for the loop condition, so make the query about that. We can extend
this to a more general "based-on" relationship, rather than a direct
icmp use later.

llvm-svn: 364715

5 years ago[InstCombine] canonicalize fmin/fmax to LLVM intrinsics minnum/maxnum
Sanjay Patel [Sat, 29 Jun 2019 14:28:54 +0000 (14:28 +0000)]
[InstCombine] canonicalize fmin/fmax to LLVM intrinsics minnum/maxnum

This transform came up in D62414, but we should deal with it first.
We have LLVM intrinsics that correspond exactly to libm calls (unlike
most libm calls, these libm calls never set errno).
This holds without any fast-math-flags, so we should always canonicalize
to those intrinsics directly for better optimization.
Currently, we convert to fcmp+select only when we have FMF (nnan) because
fcmp+select does not preserve the semantics of the call in the general case.

Differential Revision: https://reviews.llvm.org/D63214

llvm-svn: 364714

5 years ago[LFTR] Remove unnecessary latch check; NFCI
Nikita Popov [Sat, 29 Jun 2019 12:41:02 +0000 (12:41 +0000)]
[LFTR] Remove unnecessary latch check; NFCI

The whole indvars pass works on loops in simplified form, so there
is always a unique latch. Convert the condition into an assertion
in needsLFTR (though we also assert this in later LFTR functions).

Additionally update the comment on getLoopTest() now that we are
dealing with multiple exits.

llvm-svn: 364713

5 years ago[InstCombine] Shift amount reassociation (PR42391)
Roman Lebedev [Sat, 29 Jun 2019 11:51:50 +0000 (11:51 +0000)]
[InstCombine] Shift amount reassociation (PR42391)

Summary:
Given pattern:
`(x shiftopcode Q) shiftopcode K`
we should rewrite it as
`x shiftopcode (Q+K)`  iff `(Q+K) u< bitwidth(x)`
This is valid for any shift, but they must be identical.

* https://rise4fun.com/Alive/9E2
* exact on both lshr => exact https://rise4fun.com/Alive/plHk
* exact on both ashr => exact https://rise4fun.com/Alive/QDAA
* nuw on both shl => nuw https://rise4fun.com/Alive/5Uk
* nsw on both shl => nsw https://rise4fun.com/Alive/0plg

Should fix [[ https://bugs.llvm.org/show_bug.cgi?id=42391 | PR42391]].

Reviewers: spatel, nikic, RKSimon

Reviewed By: nikic

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63812

llvm-svn: 364712

5 years ago[IR][Patternmatch] Add m_SpecificInt_ULT() predicate
Roman Lebedev [Sat, 29 Jun 2019 11:51:37 +0000 (11:51 +0000)]
[IR][Patternmatch] Add m_SpecificInt_ULT() predicate

Summary:
Match an integer or vector with every element unsigned less than the
Threshold. For vectors, this includes constants with undefined elements.

FIXME: is it worth generalizing this to simply take ICmpInst::Predicate?

Reviewers: craig.topper, spatel, nikic

Reviewed By: spatel

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63811

llvm-svn: 364711

5 years ago[APInt] Fix getBitsNeeded for INT_MIN values
Dmitry Venikov [Sat, 29 Jun 2019 11:38:12 +0000 (11:38 +0000)]
[APInt] Fix getBitsNeeded for INT_MIN values

Summary: This patch fixes behaviour of APInt::getBitsNeeded for INT_MIN 10 bits values.

Reviewers: regehr, RKSimon

Reviewed By: RKSimon

Subscribers: grandinj, dexonsmith, kristina, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63691

llvm-svn: 364710

5 years ago[LFTR] Fix post-inc pointer IV with truncated exit count (PR41998)
Nikita Popov [Sat, 29 Jun 2019 09:24:12 +0000 (09:24 +0000)]
[LFTR] Fix post-inc pointer IV with truncated exit count (PR41998)

Fixes https://bugs.llvm.org/show_bug.cgi?id=41998. Usually when we
have a truncated exit count we'll truncate the IV when comparing
against the limit, in which case exit count overflow in post-inc
form doesn't matter. However, for pointer IVs we don't do that, so
we have to be careful about incrementing the IV in the wide type.

I'm fixing this by removing the IVCount variable (which was
ExitCount or ExitCount+1) and replacing it with a UsePostInc flag,
and then moving the actual limit adjustment to the individual cases
(which are: pointer IV where we add to the wide type, integer IV
where we add to the narrow type, and constant integer IV where we
add to the wide type).

Differential Revision: https://reviews.llvm.org/D63686

llvm-svn: 364709

5 years agoPartial revert of "[llvm-ar] Document response file support in --help"
Sam Clegg [Sat, 29 Jun 2019 01:53:26 +0000 (01:53 +0000)]
Partial revert of "[llvm-ar] Document response file support in --help"

This is partial revert of 70a8027c60fe1f95e8a8a1ff6575ebf8778d3544.

The test apparently failed on win32 bots due to the way slashes in
pathnames are handled.

llvm-svn: 364705

5 years ago[Driver] Fix style issues of --print-supported-cpus after D63105
Fangrui Song [Sat, 29 Jun 2019 01:24:36 +0000 (01:24 +0000)]
[Driver] Fix style issues of --print-supported-cpus after D63105

Reviewed By: ziangwan

Differential Revision: https://reviews.llvm.org/D63822

llvm-svn: 364704

5 years agoAMDGPU/GlobalISel: Add some more tests for icmp select
Matt Arsenault [Sat, 29 Jun 2019 00:55:16 +0000 (00:55 +0000)]
AMDGPU/GlobalISel: Add some more tests for icmp select

llvm-svn: 364703

5 years agoUse const auto *
Fangrui Song [Sat, 29 Jun 2019 00:55:13 +0000 (00:55 +0000)]
Use const auto *

llvm-svn: 364702

5 years agoAMDGPU/GlobalISel: RegBankSelect for update.dpp
Matt Arsenault [Sat, 29 Jun 2019 00:44:36 +0000 (00:44 +0000)]
AMDGPU/GlobalISel: RegBankSelect for update.dpp

llvm-svn: 364701

5 years ago[TSan] Attempt to fix linker error for Linux on AArch64
Julian Lettner [Sat, 29 Jun 2019 00:40:45 +0000 (00:40 +0000)]
[TSan] Attempt to fix linker error for Linux on AArch64

Introduced in 5be69ebe121d5b6ea284c3dc6d1cd5098c94f353.

llvm-svn: 364700

5 years agoAMDGPU/GlobalISel: RegBankSelect for atomic.inc/atomic.dec
Matt Arsenault [Sat, 29 Jun 2019 00:39:20 +0000 (00:39 +0000)]
AMDGPU/GlobalISel: RegBankSelect for atomic.inc/atomic.dec

llvm-svn: 364699

5 years agoAMDGPU/GlobalISel: RegBankSelect for some DS intrinsics
Matt Arsenault [Sat, 29 Jun 2019 00:33:13 +0000 (00:33 +0000)]
AMDGPU/GlobalISel: RegBankSelect for some DS intrinsics

llvm-svn: 364698

5 years agoAMDGPU/GlobalISel: RegBankSelect for some easy intrinsics
Matt Arsenault [Sat, 29 Jun 2019 00:29:56 +0000 (00:29 +0000)]
AMDGPU/GlobalISel: RegBankSelect for some easy intrinsics

llvm-svn: 364697

5 years agoAMDGPU/GlobalISel: RegBankSelect for icmp/fcmp intrinsics
Matt Arsenault [Sat, 29 Jun 2019 00:28:52 +0000 (00:28 +0000)]
AMDGPU/GlobalISel: RegBankSelect for icmp/fcmp intrinsics

llvm-svn: 364696

5 years agoAMDGPU/GlobalISel: RegBankSelect for amdgcn.div.fmas
Matt Arsenault [Sat, 29 Jun 2019 00:25:53 +0000 (00:25 +0000)]
AMDGPU/GlobalISel: RegBankSelect for amdgcn.div.fmas

llvm-svn: 364695

5 years agoAMDGPU/GlobalISel: RegBankSelect for some simple leaf intrinsics
Matt Arsenault [Sat, 29 Jun 2019 00:22:28 +0000 (00:22 +0000)]
AMDGPU/GlobalISel: RegBankSelect for some simple leaf intrinsics

llvm-svn: 364694

5 years ago[IndVars] Remove a bit of manual constant folding [NFC]
Philip Reames [Sat, 29 Jun 2019 00:19:31 +0000 (00:19 +0000)]
[IndVars] Remove a bit of manual constant folding [NFC]

SCEV is more than capable of folding (add x, trunc(0)) to x.

llvm-svn: 364693

5 years agoRevert "[clang][NewPM] Fix broken profile test"
Leonard Chan [Sat, 29 Jun 2019 00:10:22 +0000 (00:10 +0000)]
Revert "[clang][NewPM] Fix broken profile test"

This reverts commit ab2c0ed01edfec9a9402d03bdf8633b34b73f3a7.

See https://reviews.llvm.org/D63155

llvm-svn: 364692

5 years agoAMDGPU: Add baseline test for packed shufflevector
Matt Arsenault [Fri, 28 Jun 2019 23:43:40 +0000 (23:43 +0000)]
AMDGPU: Add baseline test for packed shufflevector

llvm-svn: 364691

5 years ago[clang][test][NFC] Explicitly specify clang ABI in AST Dumper test
Jan Korous [Fri, 28 Jun 2019 22:37:31 +0000 (22:37 +0000)]
[clang][test][NFC] Explicitly specify clang ABI in AST Dumper test

Clang <= 4 used the pre-C++11 rule about which structures can be passed in registers.

llvm-svn: 364690

5 years ago[WebAssembly] Assembler: support .int16/32/64 directives.
Wouter van Oortmerssen [Fri, 28 Jun 2019 22:20:33 +0000 (22:20 +0000)]
[WebAssembly] Assembler: support .int16/32/64 directives.

Reviewers: sbc100

Subscribers: dschuff, jgravelle-google, aheejin, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63959

llvm-svn: 364689

5 years ago[WebAssembly] Allow @object in .type directives.
Wouter van Oortmerssen [Fri, 28 Jun 2019 21:53:11 +0000 (21:53 +0000)]
[WebAssembly] Allow @object in .type directives.

Reviewers: sbc100

Subscribers: dschuff, jgravelle-google, aheejin, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63955

llvm-svn: 364688

5 years ago[x86] remove stale comment about cmov; NFC
Sanjay Patel [Fri, 28 Jun 2019 21:45:55 +0000 (21:45 +0000)]
[x86] remove stale comment about cmov; NFC

The cmov node used to sometimes return a glue result (and that's what
'flag' meant in this context), but that was removed with D38664.

llvm-svn: 364687

5 years agoGet the expression parser to handle missing weak symbols.
Jim Ingham [Fri, 28 Jun 2019 21:40:05 +0000 (21:40 +0000)]
Get the expression parser to handle missing weak symbols.
MachO only for this patch.

Differential Revision: https://reviews.llvm.org/D63914

<rdar://problem/51463642>

llvm-svn: 364686

5 years ago[NFC][NewGVN] Explicitly check fpmath metadata in fpmath.ll
Cameron McInally [Fri, 28 Jun 2019 21:39:08 +0000 (21:39 +0000)]
[NFC][NewGVN] Explicitly check fpmath metadata in fpmath.ll

Suggested in D63933.

llvm-svn: 364685

5 years ago[Lanai] auto-generate complete test checks; NFC
Sanjay Patel [Fri, 28 Jun 2019 20:45:32 +0000 (20:45 +0000)]
[Lanai] auto-generate complete test checks; NFC

This file will fail with a common codegen transform that
I'm looking at, and I can't tell if that's an improvement
or regression based on the sparse checking.

llvm-svn: 364684

5 years ago[OPENMP]Improve analysis of implicit captures.
Alexey Bataev [Fri, 28 Jun 2019 20:45:14 +0000 (20:45 +0000)]
[OPENMP]Improve analysis of implicit captures.

If the variable is used in the OpenMP region implicitly, we need to
check the data-sharing attributes for such variables and generate
implicit clauses for them. Patch improves analysis of such variables for
better handling of data-sharing rules.

llvm-svn: 364683

5 years ago[WebAssembly] Assembler: Allow offsets and p2align in symbol load.
Wouter van Oortmerssen [Fri, 28 Jun 2019 20:31:13 +0000 (20:31 +0000)]
[WebAssembly] Assembler: Allow offsets and p2align in symbol load.

Reviewers: sbc100

Subscribers: dschuff, jgravelle-google, aheejin, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63951

llvm-svn: 364682

5 years ago[WebAssembly] Assembler: Improve section parsing.
Wouter van Oortmerssen [Fri, 28 Jun 2019 20:29:16 +0000 (20:29 +0000)]
[WebAssembly] Assembler: Improve section parsing.

Reviewers: sbc100

Subscribers: dschuff, jgravelle-google, aheejin, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63947

llvm-svn: 364681

5 years ago[NewGVN] Add unary FNeg support to NewGVN pass
Cameron McInally [Fri, 28 Jun 2019 20:09:32 +0000 (20:09 +0000)]
[NewGVN] Add unary FNeg support to NewGVN pass

Differential Revision: https://reviews.llvm.org/D63933

llvm-svn: 364680

5 years agoRevert enabling frame pointer elimination on OpenBSD for now.
Brad Smith [Fri, 28 Jun 2019 19:57:51 +0000 (19:57 +0000)]
Revert enabling frame pointer elimination on OpenBSD for now.

llvm-svn: 364679

5 years ago[GVNSink] Add unary FNeg support to GVNSink pass
Cameron McInally [Fri, 28 Jun 2019 19:57:31 +0000 (19:57 +0000)]
[GVNSink] Add unary FNeg support to GVNSink pass

Differential Revision: https://reviews.llvm.org/D63900

llvm-svn: 364678

5 years ago[demangle] Support for C++2a char8_t
Erik Pilkington [Fri, 28 Jun 2019 19:54:19 +0000 (19:54 +0000)]
[demangle] Support for C++2a char8_t

llvm-svn: 364677

5 years ago[TSan] Fix build build breakage on Android
Julian Lettner [Fri, 28 Jun 2019 19:49:14 +0000 (19:49 +0000)]
[TSan] Fix build build breakage on Android

Introduced in 5be69ebe121d5b6ea284c3dc6d1cd5098c94f353.

llvm-svn: 364676

5 years agoDefault to Secure PLT on PPC for musl libc.
Brad Smith [Fri, 28 Jun 2019 19:48:31 +0000 (19:48 +0000)]
Default to Secure PLT on PPC for musl libc.

This matches the default settings of clang.

llvm-svn: 364675

5 years ago[clang-doc] Handle anonymous namespaces
Julie Hockett [Fri, 28 Jun 2019 19:07:56 +0000 (19:07 +0000)]
[clang-doc] Handle anonymous namespaces

Improves output for anonymous decls, and updates the '--public' flag to exclude everything under an anonymous namespace.

Differential Revision: https://reviews.llvm.org/D52847

llvm-svn: 364674

5 years ago[llvm-ar] Document response file support in --help
Sam Clegg [Fri, 28 Jun 2019 18:48:05 +0000 (18:48 +0000)]
[llvm-ar] Document response file support in --help

Also a test for this.

Differential Revision: https://reviews.llvm.org/D63836

llvm-svn: 364673

5 years agoRe-apply r364600 with fixes.
Lang Hames [Fri, 28 Jun 2019 18:36:59 +0000 (18:36 +0000)]
Re-apply r364600 with fixes.

Fix: MachO/X86_64_RELOC_GOT is a 32-bit reloc, so only compare 32 bits.
llvm-svn: 364672

5 years ago[unittests][Support] Fix LLVM-Unit :: Support/./SupportTests/FileSystemTest.permissio...
Rainer Orth [Fri, 28 Jun 2019 18:29:18 +0000 (18:29 +0000)]
[unittests][Support] Fix LLVM-Unit :: Support/./SupportTests/FileSystemTest.permissions on Solaris

LLVM-Unit :: Support/./SupportTests/FileSystemTest.permissions currently
FAILs on Solaris:

  FAIL: LLVM-Unit :: Support/./SupportTests/FileSystemTest.permissions (2940 of 51555)
  ******************** TEST 'LLVM-Unit :: Support/./SupportTests/FileSystemTest.permissions' FAILED ********************
  Note: Google Test filter = FileSystemTest.permissions
  [==========] Running 1 test from 1 test case.
  [----------] Global test environment set-up.
  [----------] 1 test from FileSystemTest
  [ RUN      ] FileSystemTest.permissions
  /opt/llvm-buildbot/obj/llvm/llvm.src/unittests/Support/Path.cpp:1705: Failure
  Value of: CheckPermissions(fs::sticky_bit)
    Actual: false
  Expected: true
  /opt/llvm-buildbot/obj/llvm/llvm.src/unittests/Support/Path.cpp:1712: Failure
  Value of: CheckPermissions(fs::set_uid_on_exe | fs::set_gid_on_exe | fs::sticky_bit)
    Actual: false
  Expected: true
  /opt/llvm-buildbot/obj/llvm/llvm.src/unittests/Support/Path.cpp:1719: Failure
  Value of: CheckPermissions(fs::all_read | fs::set_uid_on_exe | fs::set_gid_on_exe | fs::sticky_bit)
    Actual: false
  Expected: true
  /opt/llvm-buildbot/obj/llvm/llvm.src/unittests/Support/Path.cpp:1722: Failure
  Value of: CheckPermissions(fs::all_perms)
    Actual: false
  Expected: true
  [  FAILED  ] FileSystemTest.permissions (0 ms)
  [----------] 1 test from FileSystemTest (0 ms total)

  [----------] Global test environment tear-down
  [==========] 1 test from 1 test case ran. (1 ms total)
  [  PASSED  ] 0 tests.
  [  FAILED  ] 1 test, listed below:
  [  FAILED  ] FileSystemTest.permissions

   1 FAILED TEST

Checking with truss reveals that this is the same issue as on AIX and
documented in chmod(2):

  If the process is not a privileged process and the file is not a direc-
  tory, mode bit 01000 (S_ISVTX, the sticky bit) is cleared.

The following patch fixes this in the same way.  Tested on amd64-pc-solaris2.11.

Differential Revision: https://reviews.llvm.org/D63598

llvm-svn: 364671

5 years ago[clang-doc] De-duplicate comments and locations
Julie Hockett [Fri, 28 Jun 2019 18:17:58 +0000 (18:17 +0000)]
[clang-doc] De-duplicate comments and locations

De-duplicate comments and declaration locations in reduce function.
When two files include the same header file, this file's content is mapped
twice causing comments and locations to be duplicated after the reduce stage.

Committed on behalf of Diego Astiazarán (diegoaat97@gmail.com).

Differential Revision: https://reviews.llvm.org/D62970

llvm-svn: 364670

5 years ago[GDBRemote] Remove code that flushes GDB remote packets
Jonas Devlieghere [Fri, 28 Jun 2019 18:14:27 +0000 (18:14 +0000)]
[GDBRemote] Remove code that flushes GDB remote packets

The arbitrary timeout when flushing GDB remote packets caused
non-determinism and flakiness between test runs. I suspect it is what's
causing the flakiness of the reproducer tests on GreenDragon, and want
to see if removing it causes that to go away.

This change was originally introduced in r197579 to discard a
`$T02thread:01;#4` that QEMU was sending. If anybody knows how to test
that this continues working after removing this code, I'd love to hear
it.

llvm-svn: 364669