Previously we allowed in access functions only a single SCEVUnknown, which later
became the base address. We now use getPointerBase() to derive the base address
and all remaining unknowns are handled as parameters. This allows us to handle
cases like A[b+c];

llvm-svn: 144278

llvm-svn: 144277

need to provide a 'dominating IP' which is guaranteed to
dominate the (de)activation point but which cannot be avoided
along any execution path from the (de)activation point to
the push-point of the cleanup.  Using the entry block is
bad mojo.

llvm-svn: 144276

llvm-svn: 144275

llvm-svn: 144274

llvm-svn: 144273

llvm-svn: 144272

a previously-inactive cleanup, not only do we need a
flag variable, but we should also force the cleanup to
query the flag variable.  However, we only need to do
this when we're activating in a context that's
conditionally executed;  otherwise, we may safely
assume that the cleanup is dominated by the activation
point.

llvm-svn: 144271

defaulted or deleted.

llvm-svn: 144270

I've tested it on simple cases and it works.  Test cases to follow as well as a few tweaks.

llvm-svn: 144269

full-expression.  Naturally they're inactive before we enter
the block literal expression.  This restores the intended
behavior that blocks belong to their enclosing scope.

There's a useful -O0 / compile-time optimization that we're
missing here with activating cleanups following straight-line
code from their inactive beginnings.

llvm-svn: 144268

instruction lower optimization" in the pre-RA scheduler.

The optimization, rather the hack, was done before MI use-list was available.
Now we should be able to implement it in a better way, perhaps in the
two-address pass until a MI scheduler is available.

Now that the scheduler has to backtrack to handle call sequences. Adding
artificial scheduling constraints is just not safe. Furthermore, the hack
is not taking all the other scheduling decisions into consideration so it's just
as likely to pessimize code. So I view disabling this optimization goodness
regardless of PR11314.

llvm-svn: 144267

Note: These patterns only works in some cases because
many times the load sd node is bitcasted from a load
node of a different type.

llvm-svn: 144266

literal types, as well as derived-to-base casts for lvalues and
derived-to-virtual-base casts.

llvm-svn: 144265

llvm-svn: 144264

but it is sometimes useful to track blocks.  Do so.  Also
optimize the storage of these expressions.

llvm-svn: 144263

things out correctly again.

llvm-svn: 144261

is currently too inefficient to allow us to use it for array initializers, but
fortunately we usually don't yet need to evaluate such initializers.

llvm-svn: 144260

Fixed an issue where if you had an initialized global variable, we would not
link it up correctly in the debug info if the .o file had the symbols as
UNDF + EXT (undefined external). We now properly link the globals.

llvm-svn: 144259

determine if the value is negative and flip the sign accordingly.
rdar://10422026

llvm-svn: 144258

llvm-svn: 144257

modules first in the target, then fall back to the global shared module
cache, then fall back to the global module list.

llvm-svn: 144256

options to llvm-build, so the all-targets etc. components are defined properly.

llvm-svn: 144255

Using the wrong type for the break id's (user_id_t is an unsigned int, but internal breakpoints can be negative, and anyway it is a good idea to use break_id_t for breakpoints, no?)

llvm-svn: 144254

handle defining the "magic" target related components (like native,
nativecodegen, and engine).
 - We still require these components to be in the project (currently in
   lib/Target) so that we have a place to document them and hopefully make it
   more obvious that they are "magic".

llvm-svn: 144253

llvm-svn: 144252

change the generated library .a file name once we fully switch over, but
simplifies how we treat these targets without requiring more special casing
(since their library group name and the codegen library name currently map to
the same "llvm-config" style component name).

llvm-svn: 144251

 - Gives us a place to hang target specific metadata (like whether the target has a JIT).

llvm-svn: 144250

llvm-svn: 144249

"TypedefContext". No functionality change.

llvm-svn: 144248

The TII.foldMemoryOperand hook preserves implicit operands from the
original instruction.  This is not what we want when those implicit
operands refer to the register being spilled.

Implicit operands referring to other registers are preserved.

This fixes PR11347.

llvm-svn: 144247

Fix some additional x86 intrinsics to use "I" (ICE) markings.  Fix *mmintrin.h to take them into account.

<rdar://problem/10341145>

llvm-svn: 144246

llvm-svn: 144245

llvm-svn: 144244

llvm-svn: 144243

rdar://10422955

llvm-svn: 144242

llvm-svn: 144241

Fixed an issue with the gdb format stuff for any aliases that expand to
contain a "--".

llvm-svn: 144240

Currently checks alignment and killing stores on a power of 2 boundary as this is likely
to trim the size of the earlier store without breaking large vector stores into scalar ones.

Fixes <rdar://problem/10140300>

llvm-svn: 144239

llvm-svn: 144237