llvm-svn: 233763

llvm-svn: 233762

llvm-svn: 233761

replaceWithUniquedUnresolved
replaceWithUniquedUnresolvedChangedOperand
 =>
replaceWithUniquedResolvingOperand
replaceWithUniquedChangingOperand

I find the new names less confusing; they're also more accurate.  Sorry
for the churn.

llvm-svn: 233759

In an effort to reduce binary size for components not wishing to
link against all of LLDB, as well as a parallel effort to reduce
link dependencies on Python, this patch splits out the notion of
LLDB initialization into "full" and "common" initialization.

All code related to initializing the full LLDB suite lives directly
in API now.  Previously it was only referenced from API, but because
it was defined in lldbCore, it would get implicitly linked against
by everything including lldb-server, causing a considerable
increase in binary size.

By moving this to the API layer, it also creates a better layering
for the ongoing effort to make the embedded interpreter replacable
with one from a different language (or even be completely removeable).

One semantic change necessary to get this all working was to remove
the notion of a shared debugger refcount.  The debugger is either
initialized or uninitialized now, and calling Initialize() multiple
times will simply have no effect, while the first Terminate() will
now shut it down no matter how many times Initialize() was called.
This behaves nicely with all of our supported usage patterns though,
and allows us to fix a number of nasty hacks from before.

Differential Revision: http://reviews.llvm.org/D8462

llvm-svn: 233758

I am fixing this by:
1 - make sure we aren't trying to set the symbol file for a module to the same thing it already has and leaving it alone if it is the same
2 - keep all old symbol files around in the module in case there are any outstanding type references

<rdar://problem/18029116>

llvm-svn: 233757

r233664 fixed the `Verifier` so that it doesn't crash on bad type refs.
This deserves a test!

llvm-svn: 233756

Even at -O0, we fall back to SDAG when we hit intrinsics, and if the intrinsic
is a memset/memcpy/etc. we might normally use vector types. At -O0, this is
probably not a good idea (because, if there is a bug in the lowering code,
there would be no good way to turn it off). At -O0, only use scalar preferred
types.

Related to PR22754.

llvm-svn: 233755

llvm-svn: 233754

extended loads.
Implement the related target lowering hook so that the optimization has a better
estimation of the cost of an extension.

rdar://problem/19267165

llvm-svn: 233753

llvm-svn: 233752

Uniqued nodes have more complete registration with
`ReplaceableMetadataImpl` so that they can update themselves when
operands change.  Fix a bug where `MDNode::replaceWithUniqued()` wasn't
enabling these callbacks.

The two most obvious ways missing callbacks causes problems is that
auto-resolution fails and re-uniquing (on changed operands) just doesn't
happen.  I've added tests for both -- in both cases, I confirmed that
the final check was failing before the fix.

rdar://problem/20365935

llvm-svn: 233751

Commit r233747 fixed the issue that had been blocking this.

llvm-svn: 233750

The existing code in getMemsetValue only handled integer-preferred types when
the fill value was not a constant. Make this more robust in two ways:

  1. If the preferred type is a floating-point value, do the mul-splat trick on
     the corresponding integer type and then bitcast.
  2. If the preferred type is a vector, do the mul-splat trick on one vector
     element, and then build a vector out of them.

Fixes PR22754 (although, we should also turn off use of vector types at -O0).

llvm-svn: 233749

Only MIPS used that member function, and by removing the use of the
function, I removed a static_cast. Seems like it's a win.

llvm-svn: 233748

This patch fixes MCJIT::addGlobalMapping by changing the implementation of the
ExecutionEngineState class. The new implementation maintains a bidirectional
mapping between symbol names (std::strings) and addresses (uint64_ts), rather
than a mapping between Value*s and void*s.

This has fix has been made for backwards compatibility, however the strongly
preferred way to resolve unknown symbols is by writing a custom
RuntimeDyld::SymbolResolver (formerly RTDyldMemoryManager) and overriding the
findSymbol method. The addGlobalMapping method is a hangover from the legacy JIT
(which has was removed in 3.6), and may be deprecated in a future release as
part of a clean-up of the ExecutionEngine interface.

Patch by Murat Bolat. Thanks Murat!

llvm-svn: 233747

At least in Mips we don't have a prefix for member variables.
Repeating the architecture is verbose.

llvm-svn: 233746

llvm-svn: 233745

llvm-svn: 233744

Specify an allocation order with a register class. This is used by register
allocators with a greedy heuristic. This is usefull as it is sometimes
beneficial to color more constrained classes first.

Differential Revision: http://reviews.llvm.org/D8626

llvm-svn: 233743

When allocating live intervals in linear order and all of them are local
to a single basic block you get an optimal coloring. This is also true
if you reverse the order, but it is not true if you sort live ranges
beginnings in reverse order, change to sort live range endings in
reverse order. Take the following live ranges for example:

   |---| |--------|
|----------| |-------|

They get colored suboptimally with 3 registers if you sort the live range
starting points in reverse order (but optimally with live range begins in order,
or live range ends in reverse order).

Apparently the previous strategy was intentional because of allocation
time considerations. I am having a hard time replicating these effects,
while I see substantial improvements in allocation quality with this
change.

No testcase as none of the (in tree) targets use reverse order mode.

Differential Revision: http://reviews.llvm.org/D8625

llvm-svn: 233742

Apparently they are copy-pastes. They need to be merged, or otherwise
they will diverge needlessly as I did in r233723...

llvm-svn: 233741

No functional changes.

llvm-svn: 233740

llvm-svn: 233739

This change should have been done in r233737, but I made a mistake to
not include into that.

llvm-svn: 233738

Identifiers starting with _[A-Z] is reserved for the language.
User programs shouldn't use such identifiers.

llvm-svn: 233737

Change lowerCTPOP to:
- Gracefully handle a known-zero input value
- Simplify computation of significant bit size

Thanks to Jay Foad for the review!

llvm-svn: 233736

llvm-svn: 233735

No functional changes.

llvm-svn: 233727

Lets us fuse some branches into bit tests downstream. NFC.

llvm-svn: 233725

This is a follow-on to r233704 and another partial fix for PR22685:
https://llvm.org/bugs/show_bug.cgi?id=22685

llvm-svn: 233724

llvm-svn: 233723

These regression tests are supposed to test small code model support, but have
been XFAIL'd because we don't have an in-tree memory manager that can guarantee
a small-code-model compatible memory layout. Unfortunately, they can
occasionally pass if they get lucky with memory allocation, causing unexpected
passes on the bots. That's not very helpful.

I'm going to remove these until we have the infrastructure (small-code-model
compatible memory manager) to run them properly.

llvm-svn: 233722

llvm-svn: 233721

See https://code.google.com/p/address-sanitizer/issues/detail?id=385.

llvm-svn: 233720

FINDV4BITMASK macro is defined as a macro so that the macro body is inlined.
We should use inlined functions instead of macros.

llvm-svn: 233719

Multiple inheritance is casually used here. Rewriting to not
using multiple inheritance reduces the complexity of the code
and also makes it shorter.

llvm-svn: 233718

Works with x86_64 inferior, fails w/i386 inferior - updated test to
reflect

llvm-svn: 233717

Removed expectedFailureLinux from failures that I was unable to
reproduce, updated and improved some other comments near XFAIL tests

Differential Revision: http://reviews.llvm.org/D8676

llvm-svn: 233716

Support for this target was added in LLVM r233575 and r233583

llvm-svn: 233715