llvm-svn: 85796

function to calls of that function, regardless of whether it has local
linkage or has its address taken.  Not escaping should only affect 
whether we make an aggressive assumption about the arguments to a 
function, not whether we can track the result of it.

llvm-svn: 85795

llvm-svn: 85794

llvm-svn: 85793

a shared place instead of duplicating it 4 times.

llvm-svn: 85792

"multiple return values" but not "first class aggregates"

llvm-svn: 85791

llvm-svn: 85790

instead of reinventing SCCP-specific logic.  This gives us
new powers.

llvm-svn: 85789

a DenseMap.  Doing this required being aware of subtle iterator
invalidation issues, but it provides a big speedup.  In a 
release-asserts build, this sped up optimizing 403.gcc from
1.34s -> 0.79s (IPSCCP) and 1.11s -> 0.44s (SCCP).

This commit also conflates in a bunch of general cleanups, sorry.

llvm-svn: 85788

llvm-svn: 85787

llvm-svn: 85786

llvm-svn: 85785

an assertion on the buildbot.

llvm-svn: 85784

simplify some code.

llvm-svn: 85783

llvm-svn: 85782

llvm-svn: 85781

llvm-svn: 85780

llvm-svn: 85779

llvm-svn: 85778

llvm-svn: 85777

function, eliminate temporary (and pointless) smallvector.

llvm-svn: 85776

llvm-svn: 85775

llvm-svn: 85774

llvm-svn: 85773

of the two loads agree.  Propagate that onto the new store.

llvm-svn: 85772

llvm-svn: 85771

llvm-svn: 85770

llvm-svn: 85769

llvm-svn: 85768

llvm-svn: 85767

Do not infer the target type for COPY_TO_REGCLASS from dest regclass, this won't work if it can contain several types. Require explicit result type for the node for now. This fixes PR5364.

PS: It seems that blackfin usage of copy_to_regclass is completely bogus!
llvm-svn: 85766

llvm-svn: 85765

Use NEON reg-reg moves, where profitable. This reduces "domain-cross" stalls, when we used to mix vfp and neon code (the former were used for reg-reg moves)

llvm-svn: 85764

llvm-svn: 85762

llvm-svn: 85748

llvm-svn: 85747

llvm-svn: 85746

llvm-svn: 85743

1. we'd run simplifycfg at the very start, even though
the per function passes have already cleaned this up.

2. In the main per-function pipeline that is interlaced with inlining 
   etc, we would do instcombine, jump threading, simplifycfg *before*
   doing SROA.  SROA is much more likely to expose opportunities for
   these passes than they are for SROA, so move SRoA up earlier.

also add some comments.

llvm-svn: 85742

integral constant expression, make sure to find where the initializer
was provided---inside or outside the class definition---since that can
affect whether we have an integral constant expression (and, we need
to see the initializer itself).

llvm-svn: 85741