I accidentally removed this check in an earlier commit, which cause breakage in the pre alloc splitter.

llvm-svn: 62678

Also a few signed comparison fixes.

llvm-svn: 62665

- Ensure that (operation) legalization emits proper FDIV libcall when needed.
- Fix various bugs encountered during llvm-spu-gcc build, along with various
  cleanups.
- Start supporting double precision comparisons for remaining libgcc2 build.
  Discovered interesting DAGCombiner feature, which is currently solved via
  custom lowering (64-bit constants are not legal on CellSPU, but DAGCombiner
  insists on inserting one anyway.)
- Update README.

llvm-svn: 62664

Allow targets to legalize operations (with illegal operands) that produces multiple values. For example, a load with an illegal operand (a load produces two values, a value and chain).

llvm-svn: 62663

unsigned test(unsigned a) {
  return ~a;
}
llvm used to generate:
movl    $4294967295, %eax
xorl    4(%esp), %eax

Now it generates:
movl      4(%esp), %eax
notl      %eax

It's 3 bytes shorter.

llvm-svn: 62661

llvm-svn: 62660

Besides APFloat, this involved removing code
from two places that thought they knew the
result of frem(0., x) but were wrong.

llvm-svn: 62645

llvm-svn: 62639

llvm-svn: 62638

llvm-svn: 62625

llvm-svn: 62624

invoking the host fmod, not by lowering to frem and
constant-folding that.  Fix this so it tests what I
want to test.

llvm-svn: 62622

for whatever llc defaults to.  This fixes PR3363

llvm-svn: 62619

Fix PR3243: a LiveVariables bug. When HandlePhysRegKill is checking whether the last reference is also the last def (i.e. dead def), it should also check if last reference is the current machine instruction being processed. This can happen when it is processing a physical register use and setting the current machine instruction as sub-register's last ref.

llvm-svn: 62617

llvm-svn: 62616

causing the limited precision stuff to produce the wrong result for values in
the range [0, 1).

llvm-svn: 62615

llvm-svn: 62610

llvm-svn: 62604

llvm-svn: 62603

llvm-svn: 62602

llvm-svn: 62600

llvm-svn: 62598

llvm-svn: 62596

llvm-svn: 62594

fully implemented yet and not used.  This is
mainly to clarify that APFloat::mod implements
C fmod, not remainder.

llvm-svn: 62593

Patrick Boettcher!

llvm-svn: 62592

llvm-svn: 62591

llvm-svn: 62590

llvm-svn: 62589

accidentally reverted again.

llvm-svn: 62587

llvm-svn: 62584

llvm-svn: 62578

elements, even if it is only to take the address.  Test:
break-anti-dependencies.ll with ENABLE_EXPENSIVE_CHECKS.

llvm-svn: 62576

llvm-svn: 62573

llvm-svn: 62572

llvm-svn: 62571

llvm-svn: 62561

llvm-svn: 62558

as its comment says, even in the case where it will be generating
extending loads. This fixes PR3216.

llvm-svn: 62557

Do not use DenseMap because the iterator is invalidated while constructing types. After all there was a reason why std::map was used initially!

llvm-svn: 62555