llvm-svn: 79622

llvm-svn: 79621

llvm-svn: 79620

Use masks not nodes for vector shuffle predicates. Provide set of 'legal' masks, so legalizer won't infinite cycle

llvm-svn: 79619

llvm-svn: 79617

llvm-svn: 79616

 - Together these form the (Mach-O) back end of the assembler.

 - MCAssembler is the actual assembler backend, which is designed to have a
   reasonable API. This will eventually grow to support multiple object file
   implementations, but for now its Mach-O/i386 only.

 - MCMachOStreamer adapts the MCStreamer "actions" API to the MCAssembler API,
   e.g. converting the various directives into fragments, managing state like
   the current section, and so on.

 - llvm-mc will use the new backend via '-filetype=obj', which may eventually
   be, but is not yet, since I hear that people like assemblers which actually
   assemble.

 - The only thing that works at the moment is changing sections. For the time
   being I have a Python Mach-O dumping tool in test/scripts so this stuff can
   be easily tested, eventually I expect to replace this with a real LLVM tool.

 - More doxyments to come.

I assume that since this stuff doesn't touch any of the things which are part of
2.6 that it is ok to put this in not so long before the freeze, but if someone
objects let me know, I can pull it.

llvm-svn: 79612

llvm-svn: 79603

llvm-svn: 79602

negative checks resemble each other.

llvm-svn: 79595

llvm-svn: 79594

llvm-svn: 79593

llvm-svn: 79592

vector shuffles.  Temporarily remove the tests for these operations until the
new implementation is working.

llvm-svn: 79579

llvm-svn: 79578

and many new unit tests.

llvm-svn: 79574

llvm-svn: 79572

U    include/llvm/Target/TargetData.h
U    lib/Target/TargetData.cpp

Temporarily revert 79555. It was causing hangs and test failures.

llvm-svn: 79568

llvm-svn: 79567

llvm-svn: 79564

llvm-svn: 79563

Clone functions that are shared between the Main thread and Interrupt thread.
CallSites are changed in AsmPrinter currently. A better solution would have been to modify the legalizer (SoftenFloat) to allow targets to change the name of libcalls for float operations. But that currently breaks other targets.
Also, cloing of automatic variables is done AsmPrinter, a better approach would
be to use the ValueMap in CloneFunction itself.

llvm-svn: 79562

Reduce contention on the Attributes lock by using atomic operations for reference counting rather than locking.

llvm-svn: 79560

llvm-svn: 79555

Intel documentation.

llvm-svn: 79554

over absolute addressing even in non-PIC mode (unless the address
has an index or something else incompatible), because it has a
smaller encoding.

llvm-svn: 79553

the ErrorHandler callback.

llvm-svn: 79541

llvm-svn: 79540

and hasNoSignedWrap, for consistency with the nuw and nsw properties.

llvm-svn: 79539

llvm-svn: 79535

llvm-svn: 79533

llvm-svn: 79532

share a landing pad should also use the same call site value.

llvm-svn: 79501

llvm-svn: 79495

llvm-svn: 79494

llvm-svn: 79493

llvm-svn: 79492

llvm-svn: 79486

array member of a struct, it's possible to land in an arbitrary position
inside that struct, such that attempting to find further getelementptr
indices will fail. In such cases, folding cannot be done.

llvm-svn: 79485

llvm-svn: 79484