shift.
- Add a readme entry for a missing vector_shuffle optimization that results in
  awful codegen.

llvm-svn: 52740

For this it is convenient to permit floats to
be used with EXTRACT_ELEMENT, so I tweaked
things to allow that.  I also added libcalls
for ppcf128 to i32 forms of FP_TO_XINT, since
they exist in libgcc and this case can certainly
occur (and does occur in the testsuite) - before
the i64 libcall was being used.  Also, the
XINT_TO_FP result seemed to be wrong when
the argument is an i128: the wrong fudge
factor was added (the i32 and i64 cases were
handled directly, but the i128 code fell
through to some generic softening code which
seemed to think it was i64 to f32!).  So I
fixed it by adding a fudge factor that I
found in my breakfast cereal.

llvm-svn: 52739

llvm-svn: 52738

llvm-svn: 52736

InvalidateInstructionCache method instead of calling through
a hook on the JIT.  This is a host feature, not a target feature.

llvm-svn: 52734

llvm-svn: 52733

llvm-svn: 52731

llvm-svn: 52725

select_cc and friends.  This code could be
factorized a bit but I'm not sure that it's
worth it.

llvm-svn: 52724

llvm-svn: 52723

the base SDNode's VTList.

llvm-svn: 52722

Added abstract class MemSDNode for any Node that have an associated MemOperand
Changed atomic.lcs => atomic.cmp.swap, atomic.las => atomic.load.add, and
atomic.lss => atomic.load.sub

llvm-svn: 52706

llvm-svn: 52705

llvm-svn: 52704

- Avoid speculatively execute vector ops.

llvm-svn: 52703

llvm-svn: 52702

llvm-svn: 52701

Use push_back rather than operator[], which is incorrect in this cases. Unfortunately, this slow the testcase down a little bit,
but only marginally.

llvm-svn: 52700

of after, so that any reallocation it does doesn't get counted for the pass
being timed.  This probably doesn't account for a timing discrepancy I was
looking into, but I'm fixing it anyway.

llvm-svn: 52693

load,store,call,return,bitcast.  This is enough to
make call and return work.

llvm-svn: 52691

In ConstantArray::getAsString(), we know the size of the resultant string in advance so we can pre-allocate it and just fill in
the entries.  This improves the time for the AsmPrinter on InstructionCombining.cpp from 0.4248s to 0.3370s.

llvm-svn: 52690

llvm-svn: 52689

llvm-svn: 52688

llvm-svn: 52687

llvm-svn: 52684

Also add a testcase to reproduce the gcc bootstrap failure in very much reduced form.

llvm-svn: 52677

llvm-svn: 52672

If it's determined safe, remat MOV32r0 (i.e. xor r, r) and others as it is instead of using the longer MOV32ri instruction.

llvm-svn: 52670

llvm-svn: 52664

llvm-svn: 52663

general solution possible, but it's a fairly simple one.
Based on a patch from the OpenGTL project!

llvm-svn: 52662

as a questionable case, but the code isn't actually needed.

llvm-svn: 52657

    ,------.
    |      |
    |      v
    |   t2 = phi ... t1 ...
    |      |
    |      v
    |   t1 = ...
    |  ... = ... t1 ...
    |      |
    `------'

where there is a use in a PHI node that's a predecessor to the defining
block. We don't want to mark all predecessors as having the value "alive" in
this case. Also, the assert was too restrictive and didn't handle this case.

llvm-svn: 52655

needs arbitrary-element removal.

llvm-svn: 52654

llvm-svn: 52651

Use getMBBEndIdx rather than assuming that the end is right after the last instruction in the block.

llvm-svn: 52649

llvm-svn: 52648

in the presence of out-of-loop users of in-loop values and the trip
count is not a known multiple of the unroll count, and to be a bit
simpler overall. This fixes PR2253.

llvm-svn: 52645

Remove option used to debug stack coloring bugs. It's no longer needed since stack coloring is now bug free.

llvm-svn: 52644

llvm-svn: 52643