follow the rules for undef used in instcombine.

llvm-svn: 37851

model to include f32 variants.  Some factoring
improvments forthcoming.

llvm-svn: 37847

the recent lowering changes, this allows types like <4 x double> to
be returned, using two vector registers.

llvm-svn: 37844

refactored getCopyFromParts and getCopyToParts, which are more general.
This effectively adds support for lowering illegal by-val vector call
arguments.

llvm-svn: 37843

visitFSUB to fold 0-B to -B in UnsafeFPMath mode. Also change visitFNEG
to use isNegatibleForFree/GetNegatedExpression instead of doing a subset
of the same thing manually.

This fixes test/CodeGen/X86/negative-sin.ll.

llvm-svn: 37842

llvm-svn: 37839

llvm-svn: 37829

llvm-svn: 37825

llvm-svn: 37822

llvm-svn: 37820

llvm-svn: 37818

llvm-svn: 37816

llvm-svn: 37814

llvm-svn: 37801

llvm-svn: 37799

by _GLIBCXX_DEBUG.

llvm-svn: 37796

Uncovered by _GLIBCXX_DEBUG.

llvm-svn: 37795

ordering and thus violated the strict weak ordering requirement of
priority_queue.  Uncovered by _GLIBCXX_DEBUG.

llvm-svn: 37794

_GLIBCXX_DEBUG.

llvm-svn: 37793

llvm-svn: 37792

llvm-svn: 37790

llvm-svn: 37789

llvm-svn: 37788

llvm-svn: 37786

vectors are split down to single elements as part of legalization.

llvm-svn: 37785

llvm-svn: 37784

Add support for performing GVNPRE on select instructions.  This fixes test/Transforms/GVNPRE/select.ll.

llvm-svn: 37783

illegal value type will be transformed to, for code that needs the
register type after all transformations instead of just after the first
transformation.

Factor out the code that uses this information to do copy-from-regs and
copy-to-regs for various purposes into separate functions so that they
are done consistently.

llvm-svn: 37781

llvm-svn: 37778

llvm-svn: 37776

loop.
- Avoid loop unswich for loop header branch.
- While cloning dominators fix typo and handle self dominating blocks.

llvm-svn: 37772

llvm-svn: 37771

to loop unswitch candidate for the loop.

llvm-svn: 37770

Anton's testcase from 35.5s to 34.7s.

llvm-svn: 37769

llvm-svn: 37768

llvm-svn: 37765

llvm-svn: 37764

Partial fix for PR1502: If a EH register is needed in a successor of landing pad, add it as livein to all the blocks in the paths between the landing pad and the specified block.

llvm-svn: 37763

Use cached information that has already been computed to make clean() simpler and faster.  This is a small speedup on most cases.

llvm-svn: 37761

This saves many lines of code duplication.  No functionality change.

llvm-svn: 37759