llvm-svn: 196732

in the td file.

llvm-svn: 196731

here is to make save/restore into variable number of argument instructions.

llvm-svn: 196726

When trying to eliminate an "sub sp, sp, #N" instruction by folding
it into an existing push/pop using dummy registers, we need to account
for the fact that this might affect precisely how "fp" gets set in the
prologue.

We were attempting this, but assuming that *whenever* we performed a
fold it would make a difference. This is false, for example, in:
    push {r4, r7, lr}
    add fp, sp, #4
    vpush {d8}
    sub sp, sp, #8

we can fold the "sub" into the "vpush", forming "vpush {d7, d8}".
However, in that case the "add fp" instruction mustn't change, which
we were getting wrong before.

Should fix PR18160.

llvm-svn: 196725

llvm-svn: 196719

llvm-svn: 196718

No functionality change.  Changing comments to match code.

llvm-svn: 196713

This fixes PR17872.  This bug can lead to C++ destructors not being
called when they should be, when an exception is thrown.

llvm-svn: 196711

Before this change, inlining one "invoke" into an outer "invoke" call
site can lead to the outer landingpad's catch/filter clauses being
copied multiple times into the resulting landingpad.  This happens:

 * when the inlined function contains multiple "resume" instructions,
   because forwardResume() copies the clauses but is called multiple
   times;

 * when the inlined function contains a "resume" and a "call", because
   HandleCallsInBlockInlinedThroughInvoke() copies the clauses but is
   redundant with forwardResume().

Fix this by deduplicating the code.

This problem doesn't lead to any incorrect execution; it's only
untidy.

This change will make fixing PR17872 a little easier.

llvm-svn: 196710

llvm-svn: 196669

overall time of LLVM compilation by ~1%.

llvm-svn: 196667

They were out of place since the introduction of arbitrary precision integer
types.

This also synchronizes the documentation to Types.h, so it refers to first class
types and single value types.

llvm-svn: 196661

llvm-svn: 196654

llvm-svn: 196645

These helper classes take care of the book-keeping the drives the
GenericScheduler heuristics. It is likely that developers writing
target-specific schedulers that work similarly to GenericScheduler
will want to use these helpers too. The immediate goal is to develop a
GenericPostScheduler that can run in place of the old PostRAScheduler,
but will use the new machine model.

No functionality change intended.

llvm-svn: 196643

llvm-svn: 196641

llvm-svn: 196640

Patch by Andy Kaylor, with minor edits to resolve merge conflicts.

llvm-svn: 196639

llvm-svn: 196638

llvm-svn: 196637

Indent the command lines to indicate they continue from previous lines. Also,
fix incorrect uses of CHECK-DAG and CHECK-NOT.

llvm-svn: 196636

llvm-svn: 196635

llvm-svn: 196634

llvm-svn: 196633

llvm-svn: 196632

The sefault occurs due to an infinite loop when the verifier tries to
determine the size of a type of the form "%rt = type { %rt }" while
checking an alloca of the type.

llvm-svn: 196626

lib/Transforms/Instrumentation/AddressSanitizer.cpp:1405:36: error: non-constant-expression cannot be narrowed from type 'uint64_t' (aka 'unsigned long long') to 'size_t' (aka 'unsigned int') in initializer list [-Wc++11-narrowing]
    getAllocaSizeInBytes(AI),
    ^~~~~~~~~~~~~~~~~~~~~~~~

llvm-svn: 196623

This commit caches the value of the AllowAtInIdentifier variable as
a class variable in AsmLexer. We do this to avoid repeated MAI
queries and string comparisons each time we lex an identifier.

llvm-svn: 196622

Command line arguments that begin with @ but aren't a path to an
existing file currently cause later @file arguments to be ignored.

Correctly skip over these arguments instead of trying to read a
non-existent file 20 times and giving up.

Since the problem manifests in the clang driver, the test is in that
repository.

Fixes rdar://problem/15590906

llvm-svn: 196620

- krait processor currently modeled with the same features as A9.
- Krait processor additionally has VFP4 (fused multiply add/sub)
and hardware division features enabled.
- krait has currently the same Schedule model as A9
- krait cpu flag is not recognized by the GNU assembler yet,
it is replaced with march=armv7-a to avoid a lower march
from being used.

llvm-svn: 196619

This removes another case of spooky action at a distance (building the
same label names in multiple places creating an implicit dependency
between those places) and helps pave the way for type units.

llvm-svn: 196617

This is a precursor to moving type units into the correct (debug_types)
section with comdat groups and full type unit headers.

llvm-svn: 196615

ConstantExpr can evaluate to false even when isNullValue gives false.

Fixes PR18143.

llvm-svn: 196611

llvm-svn: 196609

The integrated assembler fails to properly lex arm comments when
they are adjacent to an identifier in the input stream. The reason
is that the arm comment symbol '@' is also used as symbol variant in
other assembly languages so when lexing an identifier it allows the
'@' symbol as part of the identifier.

Example:
  $ cat comment.s
  foo:
    add r0, r0@got to parse this as a comment

  $ llvm-mc -triple armv7 comment.s
  comment.s:4:18: error: unexpected token in argument list
    add r0, r0@got to parse this as a comment
                   ^
This should be parsed as correctly as `add r0, r0`.

This commit modifes the assembly lexer to not include the '@' symbol
in identifiers when lexing for targets that use '@' for comments.

llvm-svn: 196607

llvm-svn: 196604

This more accurately represents the actual walk - pubnames/pubtypes are
emitted into the .o, not the .dwo, and reference the skeletons not the
full units.

Use the newly established ID->index invariant to lookup the underlying
full unit to retrieve its public names and types.

llvm-svn: 196601

This simplifies reasoning about the code and enables simple navigation
from a skeleton to its full unit. (currently there are no type unit
skeletons, so the skeleton list doesn't have the same ID == index
property)

Eventually we should get rid of this ID and just store the labels we
need as the IDs are allowing this code to create difficult to
manage/understand associations (loops over non-skeletal units are
implicitly referencing their skeletal units during pub* emission, for
example). It may be necessary to have some kind of skeleton->full unit
association and a more direct pointer or similar device would be
preferable than an index.

llvm-svn: 196600

The current peephole optimizing for compare inst assumes an instr that
uses CPSR has an MO for ARM Cond code.However, for VSEL instructions
(vseqeq, vselgt, vselgt, vselvs), there is no such operand nor do
they support the modification of Cond Code.

llvm-svn: 196588

llvm-svn: 196585