LLVM_BUILD_GLOBAL_ISEL

llvm-svn: 260457

llvm-svn: 260456

Summary: Measured to be more performant when reading bitcode.

Reviewers: rafael, dexonsmith

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D16285

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 260455

The Use argument was used to compute the operand number for a fast
path when replacing only one operand. However we always have to go
through all the operands. So the argument number can be recomputed
locally anyway.

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 260454

IMHO this makes the code easier to read and should help with linker
scripts.

This is strongly based on D16575. The main differences are:

We record a range of sections, not every section in a program header.

scanHeaders takes case of deciding what goes in every program header,
including PT_GNU_RELRO

We create dummy sections for the start of the file

With this, program header creation has 3 isolated stages:

Map sections to program headers.
Assign addresses to *sections*
Looking at sections find the address and size of each program header.

Thanks to George Rimar for the initial version.

llvm-svn: 260453

short option as an aid to memory.  Like it's w because of the W in throW.

That helps me remember.  If we are going to take these out we should take them
all out.  But I kind of like them.

llvm-svn: 260452

It looks like clang has a couple of test cases which caught the fact LVI was not slightly more precise after 260439.  When looking at the failures, it struck me as wasteful to be querying nullness of a constant via LVI, so instead of tweaking the clang tests, let's just stop querying constants from this source.

llvm-svn: 260451

The patch adds a parameter in annotateValueSite() to control the max number
of records written to the value profile meta data for each value site. The
default is kept as the current value of 3.

Differential Revision: http://reviews.llvm.org/D17084

llvm-svn: 260450

This caused the compiler to fail with "invalid linkage type
for global declaration" (PR26569).

llvm-svn: 260449

Previously we attempted to be smart; if one job failed, we'd run all
jobs that didn't depend on the failing job.

Problem is, this doesn't work well for e.g. CUDA compilation without
-save-temps.  In this case, the device-side and host-side Assemble
actions (which actually are responsible for preprocess, compile,
backend, and assemble, since we're not saving temps) are necessarily
distinct.  So our clever heuristic doesn't help us, and we repeat every
error message once for host and once for each device arch.

The main effect of this change, other than fixing CUDA, is that if you
pass multiple cc files to one instance of clang and you get a compile
error, we'll stop when the first cc1 job fails.

Reviewers: tra, echristo

Subscribers: jhen, cfe-commits

Differential Revision: http://reviews.llvm.org/D16514

llvm-svn: 260448

Previously the code used getActiveBits() to determine the highest set
bit of each APInt first. However doing so requires the same amount of
memory accesses as simply comparing both numbers right away.

Removing all the active bit checks leads to simpler code and is faster
in my benchmark.

Differential Revision: http://reviews.llvm.org/D16620

llvm-svn: 260447

llvm-svn: 260446

We already do this for Objective-C interfaces, but we never handled protocols
because the DWARF didn't represent them.  Nowadays, though, we can import them
from modules, and we have to mark them properly.

<rdar://problem/24193009>

llvm-svn: 260445

llvm-svn: 260444

llvm-svn: 260443

This restores commit r260408, along with a fix for a bot failure.

The bot failure was caused by dereferencing a unique_ptr in the same
call instruction parameter list where it was passed via std::move.
Apparently due to luck this was not exposed when I built the compiler
with clang, only with gcc.

llvm-svn: 260442

static_cast of a pointer to object before the start of the object's
lifetime has undefined behavior.

This code triggers CFI warnings.

This change replaces C-style casts with reinterpret_cast, which is
fine per the standard, add applies an attribute to silence CFI (which
barks on reinterpret_cast, too).

llvm-svn: 260441

llvm-svn: 260440

There's nothing preventing callers of LVI from asking for lattice values representing a Constant.  In fact, given that several callers are walking back through PHI nodes and trying to simplify predicates, such queries are actually quite common.  This is mostly harmless today, but we start volatiling assertions if we add new calls to getBlockValue in otherwise reasonable places.

Note that this change is not NFC.  Specifically:
1) The result returned through getValueAt will now be more precise.  In principle, this could trigger any latent infinite optimization loops in callers, but in practice, we're unlikely to see this.
2) The result returned through getBlockValueAt is potentially weakened for non-constants that were previously queried.  With the old code, you had the possibility that a later query might bypass the cache and discover some information the original query did not.  I can't find a scenario which actually causes this to happen, but it was in principle possible.  On the other hand, this may end up reducing compile time when the same value is queried repeatedly.  

llvm-svn: 260439

llvm-svn: 260438

We have a vector of all of the compact unwind entries anyway, and
its live as long as we need it to be.  So instead of copying from that
vector to another, just take references to the range of the original vector
we need for each compact unwind page.

llvm-svn: 260437

llvm-svn: 260436

Previously LLVM could not process any debug info we produced, so it
didn't make sense to spend time generating it. Now that it has primitive
support for local variable info, it does make sense to generate normal
debug info.

llvm-svn: 260435

Now that SymbolFileDWARF supports having types in completely separate .pcm file with "-fmodules -gmodules", each SymbolFileDWARF can reference module DWARF info by looking in other DWARF files. Then if you have 1000 .o files that each reference one or more .pcm files in their debug info, a simple Module::FindTypes(...) call can end up searching the same .pcm file over and over and over. Now all internal FindTypes methods in classes (ModuleList, Module, SymbolFile) now take an extra argument:

    llvm::DenseSet<lldb_private::SymbolFile *> &searched_symbol_files
    
Each time a SymbolFile::FindTypes() is called, it needs to check the searched_symbol_files list to make sure it hasn't already been asked to find the type and return immediately if it has been checked. This will stop circular dependencies from also crashing LLDB during type queries. 

This has proven to be an issue when debugging large applications on MacOSX that use DWARF in .o files. 

<rdar://problem/24581488>

llvm-svn: 260434

Reviewers: weimingz

Subscribers: cfe-commits

Differential Revision: http://reviews.llvm.org/D17085

llvm-svn: 260433

Summary:
Refactor common value, scope, and label tracking logic out of DwarfDebug
into a common base class called DebugHandlerBase.

Update an old LLVM IR test case to avoid an assertion in LexicalScopes.

Reviewers: dblaikie, majnemer

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D16931

llvm-svn: 260432

This time I kept <ext/hash_map> working!

This patch is the first in a series of patches that's meant to better
support unordered_map. unordered_map has a special "value_type" that
differs from pair<const Key, Value>. In order to meet the EmplaceConstructible
and CopyInsertable requirements we need to teach __hash_table about this
special value_type.

This patch creates a "__hash_node_types" traits class that contains
all of the typedefs needed by the unordered containers and it's iterators.
These typedefs include ones for each node type and  node pointer type,
as well as special typedefs for "unordered_map"'s value type.

As a result of this change all of the unordered containers now all support
incomplete types.

As a drive-by fix I changed the difference_type in __hash_table to always
be ptrdiff_t. There is a corresponding change to size_type but it cannot
take affect until an ABI break.

This patch will be followed up shortly with fixes for various unordered_map
bugs and problems.

llvm-svn: 260431

New option --elf-output-style=LLVM or GNU
Enables -file-headers in readelf style when elf-output-style=GNU

Differential revision: http://reviews.llvm.org/D14128

llvm-svn: 260430

llvm-svn: 260429

llvm-svn: 260428

Summary:
This fixes a crash where subsequent spills would be unable to scavenge
a register. In particular, it fixes a crash in piglit's
spec@glsl-1.50@execution@geometry@max-input-components (the test still
has a shader that fails to compile because of too many SGPR spills, but
at least it doesn't crash any more).

This is a candidate for the release branch.

Reviewers: arsenm, tstellarAMD

Subscribers: qcolombet, arsenm

Differential Revision: http://reviews.llvm.org/D16558

llvm-svn: 260427

llvm-svn: 260426

that's not true in general. Instead, use a preference order to pick the
standard C++ signature 'char*(char*, int)' where possible and fall back to the
C signature 'char*(const char*, int)' only when it's unavailable.

llvm-svn: 260425

The patch adds lazy relocation support for MIPS and R_MIPS_26 relocation
handing.

R_MIPS_26 relocation might require PLT entry creation. In that case it
is fully supported by the patch. But if the relocation target is a local
symbol we need to use a different expression to calculate the relocation
result. This case is not implemented yet because there is no method to
get know the kind of relocation target in the `relocateOne` routine.

Differential Revision: http://reviews.llvm.org/D16982

llvm-svn: 260424

This was supposed to have been removed but made it back in
accidentally.

llvm-svn: 260423

llvm-svn: 260422

Instead of passing varargs directly on the user stack, allocate a buffer in
the caller's stack frame and pass a pointer to it. This simplifies the C
ABI (e.g. non-C callers of C functions do not need to use C's user stack if
they have their own mechanism) and allows further optimizations in the future
(e.g. fewer functions may need to use the stack).

Differential Revision: http://reviews.llvm.org/D17048

llvm-svn: 260421

The following batch files allow to launch scan-build from an arbitrary location if path to clang\tools\scan-build-py\bin is added to %PATH%.

llvm-svn: 260420

llvm-svn: 260419

llvm-svn: 260418