This is handled at node creation time (similar to X/1.0)
after:
rL357029
(no fast-math-flags needed)

llvm-svn: 368443

llvm-svn: 368442

Summary:
This is exposed by adding a new testcase in PowerPC in
https://reviews.llvm.org/rL367732

The testcase got different output on different platform, hence breaking
buildbots.

The problem is that we get differnt FuncUnitOrder when calculateResMII.

The root cause is:
1. Two MachineInstr might get SAME priority(MFUsx) from minFuncUnits.
2. Current comparison operator() will return `MFUs1 > MFUs2`.
3. We use iterators for MachineInstr, so the input to FuncUnitSorter
   might be different on differnt platform due to the iterator nature.

So for two MI with same MFU, their order is actually depends on the
iterator order, which is platform (implemtation) dependent.

This is risky, and may cause cross-compiling problems.

The fix is to check make sure we assign a determine order when they are
equal.

Reviewers: bcahoon, hfinkel, jmolloy

Subscribers: nemanjai, hiraditya, MaskRay, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D65992

llvm-svn: 368441

llvm-svn: 368440

Summary: Implement a new analysis to estimate the number of cache lines
required by a loop nest.
The analysis is largely based on the following paper:

Compiler Optimizations for Improving Data Locality
By: Steve Carr, Katherine S. McKinley, Chau-Wen Tseng
http://www.cs.utexas.edu/users/mckinley/papers/asplos-1994.pdf
The analysis considers temporal reuse (accesses to the same memory
location) and spatial reuse (accesses to memory locations within a cache
line). For simplicity the analysis considers memory accesses in the
innermost loop in a loop nest, and thus determines the number of cache
lines used when the loop L in loop nest LN is placed in the innermost
position.

The result of the analysis can be used to drive several transformations.
As an example, loop interchange could use it determine which loops in a
perfect loop nest should be interchanged to maximize cache reuse.
Similarly, loop distribution could be enhanced to take into
consideration cache reuse between arrays when distributing a loop to
eliminate vectorization inhibiting dependencies.

The general approach taken to estimate the number of cache lines used by
the memory references in the inner loop of a loop nest is:

Partition memory references that exhibit temporal or spatial reuse into
reference groups.
For each loop L in the a loop nest LN: a. Compute the cost of the
reference group b. Compute the 'cache cost' of the loop nest by summing
up the reference groups costs
For further details of the algorithm please refer to the paper.
Authored By: etiotto
Reviewers: hfinkel, Meinersbur, jdoerfert, kbarton, bmahjour, anemet,
fhahn
Reviewed By: Meinersbur
Subscribers: reames, nemanjai, MaskRay, wuzish, Hahnfeld, xusx595,
venkataramanan.kumar.llvm, greened, dmgreen, steleman, fhahn, xblvaOO,
Whitney, mgorny, hiraditya, mgrang, jsji, llvm-commits
Tag: LLVM
Differential Revision: https://reviews.llvm.org/D63459

llvm-svn: 368439

As discussed on PR42825, if we are inverting the selection mask we can just swap the inputs and avoid the inversion.

Differential Revision: https://reviews.llvm.org/D65522

llvm-svn: 368438

This is a minimal fix (copy the predicate for the assert) to
prevent the crashing seen in:
https://bugs.llvm.org/show_bug.cgi?id=42932
...when converting a constant integer of arbitrary width to uint64_t.

Differential Revision: https://reviews.llvm.org/D65970

llvm-svn: 368437

MSVC (19.16) wants to see the definition of Instruction in
`std::pair<unsigned, const Instruction &> SourceRef` to decide
if it is assignable.

Patch by Orivej Desh.

Differential Revision: https://reviews.llvm.org/D65844

llvm-svn: 368436

This patch addresses two closely related bugs:
https://bugs.llvm.org/show_bug.cgi?id=42930 and
https://bugs.llvm.org/show_bug.cgi?id=42931.

GNU readelf prints the file name for every input unless there is only
one input and that input is not an archive. This patch adds the printing
for multiple inputs. A previous change did it for archives, but
introduced a regression with GNU compatibility for single-output
formatting, resulting in a spurious initial blank line. This is fixed in
this patch too.

Reviewed by: grimar, MaskRay

Differential Revision: https://reviews.llvm.org/D65953

llvm-svn: 368435

Summary: Constructor initializers were not being highlighted. This adds highlighting for them by using TraverseConstructorInitializer. Uses the Traverse* because there is no visit for CXXCtorInitializer.

Reviewers: hokein, ilya-biryukov

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D66001

llvm-svn: 368434

Fast-isel was picking AFGR64 register class for processing call
arguments when +fp64 options was used. We simply check is option +fp64
is used and pick appropriate register.

Patch by Mirko Brkusanin.

Differential Revision: https://reviews.llvm.org/D65886

llvm-svn: 368433

Flag -show-encoding enables the printing of instruction encodings as part of the
the instruction info view.

Example (with flags -mtriple=x86_64--  -mcpu=btver2):

Instruction Info:
[1]: #uOps
[2]: Latency
[3]: RThroughput
[4]: MayLoad
[5]: MayStore
[6]: HasSideEffects (U)
[7]: Encoding Size

[1]    [2]    [3]    [4]    [5]    [6]    [7]    Encodings:     Instructions:
 1      2     1.00                         4     c5 f0 59 d0    vmulps   %xmm0, %xmm1, %xmm2
 1      4     1.00                         4     c5 eb 7c da    vhaddps  %xmm2, %xmm2, %xmm3
 1      4     1.00                         4     c5 e3 7c e3    vhaddps  %xmm3, %xmm3, %xmm4

In this example, column Encoding Size is the size in bytes of the instruction
encoding. Column Encodings reports the actual instruction encodings as byte
sequences in hex (objdump style).

The computation of encodings is done by a utility class named mca::CodeEmitter.

In future, I plan to expose the CodeEmitter to the instruction builder, so that
information about instruction encoding sizes can be used by the simulator. That
would be a first step towards simulating the throughput from the decoders in the
hardware frontend.

Differential Revision: https://reviews.llvm.org/D65948

llvm-svn: 368432

Summary:
The Arm Neoverse E1 and Cortex-A65 Software Optimization Guide [1][2],
Section "4.7 Branch instruction alignment" state:

"It is preferable for branch targets, including subroutine entry points,
to be placed on aligned 64-bit boundaries to maximize instruction fetch
efficiency."

This patch sets the preferred function alignment on Neoverse E1 and
Cortex-A65 to 2^3=8B. This was already the case in some Cortex-A CPUs
such as Cortex-A53.

[1] https://developer.arm.com/docs/swog466751/latest/arm-neoversetm-e1-core-software-optimization-guide
[2] https://developer.arm.com/docs/swog010045/latest/arm-cortex-a65-core-software-optimization-guide

Reviewers: dmgreen, fhahn, samparker

Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D65937

llvm-svn: 368431

This is a bit strange method. It works like a unwrapOrError,
but named error. It does not report an Input name.
I removed it.

Differential revision: https://reviews.llvm.org/D66000

llvm-svn: 368430

Summary:
The API provided by refactoring lib doesn't provide enough flexibility
to get clangd's rename to behave as we expect. Instead, we replace it
with the low-level rename functions, which give us more control.

Bonus:
- performance, previously we visit the TU to find all occurrences,
  now we just visit top-level decls from main file;
- fix a bug where we wrongly filter out the main file replacement due to the
  different relative/absolute file path;

Reviewers: sammccall

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D65936

llvm-svn: 368429

This patch changes the code to use a modern unwrapOrError(StringRef Input, Expected<T> EO)
version that contains the input source name and removes the deprecated version.

Differential revision: https://reviews.llvm.org/D65946

llvm-svn: 368428

llvm-svn: 368425

These types were recently added in D62960 but it seems the patch didn't
consider LLDB which causes a bunch of compiler warnings about
missing enum values. It seems this feature isn't fully implemented yet,
so I don't think we can write any test for this. For now lets just add
the missing types to our usual list of unsupported types.

llvm-svn: 368424

All TLS access on Darwin is in the "general dynamic" form where we call
a function to resolve the address, so implementation is pretty simple.

llvm-svn: 368418

Not NFC as this will probably fix a wrong guard variable check
on Windows. Not sure though what Windows test can now be safely
enabled.

llvm-svn: 368417

Summary:
When opening a minidump, we were failing to find an executable because
we were searching for i386-unknown-windows, whereas we recognize the
pe/coff files as i386-pc-windows. This fixes the triple computation code
in the minidump parser to match pe/coff, and adds an appropriate test.

NB: I'm not sure setting the vendor to "pc" is really correct for
arm(64) windows, but right now that seems to match what we do in the
pe/coff case (ArchSpec.cpp:935).

Reviewers: clayborg, amccarth

Subscribers: javed.absar, kristof.beyls, rnk, markmentovai, lldb-commits

Differential Revision: https://reviews.llvm.org/D65955

llvm-svn: 368416

llvm-svn: 368415

This patch adds the SVE built-in types defined by the Procedure Call
Standard for the Arm Architecture:

   https://developer.arm.com/docs/100986/0000

It handles the types in all relevant places that deal with built-in types.
At the moment, some of these places bail out with an error, including:

   (1) trying to generate LLVM IR for the types
   (2) trying to generate debug info for the types
   (3) trying to mangle the types using the Microsoft C++ ABI
   (4) trying to @encode the types in Objective C

(1) and (2) are fixed by follow-on patches but (unlike this patch)
they deal mostly with target-specific LLVM details, so seemed like
a logically separate change.  There is currently no spec for (3) and
(4), so reporting an error seems like the correct behaviour for now.

The intention is that the types will become sizeless types:

   http://lists.llvm.org/pipermail/cfe-dev/2019-June/062523.html

The main purpose of the sizeless type extension is to diagnose
impossible or dangerous uses of the types, such as any that would
require sizeof to have a meaningful defined value.

Until then, the patch sets the alignments of the types to the values
specified in the link above.  It also sets the sizes of the types to
zero, which is chosen to be consistently wrong and shouldn't affect
correctly-written code (i.e. code that would compile even with the
sizeless type extension).

The patch adds the common subset of functionality needed to test the
sizeless type extension on the one hand and to provide SVE intrinsic
functions on the other.  After this patch, the two pieces of work are
essentially independent.

The patch is based on one by Graham Hunter:

   https://reviews.llvm.org/D59245

Differential Revision: https://reviews.llvm.org/D62960

llvm-svn: 368413

It is outdated. Using of Expected<> is preferred, also it does
not provide a way to report a file name.

I updated the code to use the modern version of unwrapOrError instead.

Differential revision: https://reviews.llvm.org/D65951

llvm-svn: 368410

The filename part in the message header is used by Visual Studio
to fill Error List so that a user can click on an item and jump
to the mentioned location. If we use only the name of a source file
and not the full path, Visual Studio might be unable to find the right
file or, even worse, show a wrong one.

Differential Revision: https://reviews.llvm.org/D65875

llvm-svn: 368409

I've now needed to add an extra parameter to this call twice recently. Not only
is the signature getting extremely unwieldy, but just updating all of the
callsites and implementations is a pain. Putting the parameters in a struct
sidesteps both issues.

llvm-svn: 368408

llvm-svn: 368406

llvm-svn: 368405

As loads are combined and widened, we replaced their sext users
operands whereas we should have been replacing the uses of the sext.
I've added a load of tests, with only a few of them originally
causing assertion failures, the rest improve pattern coverage.

Differential Revision: https://reviews.llvm.org/D65740

llvm-svn: 368404

[clangd] Fixed printTemplateSpecializationArgs not printing partial variable specialization arguments.

Summary:
printTemplateSpecializationArgs was not printing partial variable specialization args. This adds an additional If clause where we check if it's a VariableTemplatePartialSpecializationDecl and returns the ArgumentLocs if that's the case.
Also adds tests for printTemplateSpecializationArgs in ASTTests.cpp.

Reviewers: hokein, ilya-biryukov

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

Tags: #clang

Differential Revision: https://reviews.llvm.org/D65926

llvm-svn: 368403

Summary:
Clang generates function bodies and puts them in the AST for default methods if it is defaulted outside the class definition.

`
struct A {
   A &operator=(A &&O);
};

A &A::operator=(A &&O) = default;
`

This will generate a function body for the `A &A::operator=(A &&O)` and put it in the AST. This body should not be visited if implicit code is not visited as it is implicit.

This was causing SemanticHighlighting in clangd to generate duplicate tokens and putting them in weird places.

Reviewers: hokein, ilya-biryukov, gribozavr

Subscribers: mgorny, jkorous, arphaman, kadircet, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D65938

llvm-svn: 368402

Summary:
Make sure that we report that changes has been made
by InstSimplify also in situations when only trivially
dead instructions has been removed. If for example a call
is removed the call graph must be updated.

Bug seem to have been introduced by llvm-svn r367173
(commit 02b9e45a), since the code in question
was rewritten in that commit.

Reviewers: spatel, chandlerc, foad

Reviewed By: spatel

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D65973

llvm-svn: 368401

We shouldn't form trunc stores that need to be expanded now that
we are using widening legalization.

llvm-svn: 368400

Summary: ASSERT_THAT_ERROR looks like the intended helper for use in tests.

Reviewers: plotfi, jkorous, compnerd

Subscribers: mgorny, dexonsmith, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D65853

llvm-svn: 368399

Summary: libs can be installed to ../lib64.

Subscribers: mgorny, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D65972

llvm-svn: 368398

I believe PR34584 was tracking that FIXME, but its since been
closed and a test case was added.

llvm-svn: 368397

The only way to generate these was through promoting legalization
of narrow vectors, but we widen those types now. So we shouldn't
produce these nodes.

llvm-svn: 368396

[X86] Remove handler for (U/S)(ADD/SUB)SAT from ReplaceNodeResults. Remove TypeWidenVector check from code that handles X86ISD::VPMADDWD and X86ISD::AVG.

More unneeded code since we now legalize narrow vectors by widening.

llvm-svn: 368395

This is no longer needed since we widen v2i32 instead of promoting.

llvm-svn: 368394

I am working to remove this concept of the "FinalPhase" in the clang driver,
but it is used in a lot of different places to do argument handling for
different combinations of phase pipelines and arguments. I am trying to
consolidate most of the uses of "FinalPhase" into its own separate scope.
Eventually, in a subsequent patch I will move all of this stuff to a separate
function, and have more of the complication phase list construction setup into
types::getComplicationPhases.

Differential Revision: https://reviews.llvm.org/D65969

llvm-svn: 368393