.zerofill directive.  Streamerize its generation.

llvm-svn: 93868

MDNodes are not expected to disappear or replaced by another MDNode, so there is no need to pay the cost of WeakVH and ValueMaps.

llvm-svn: 93865

llvm-svn: 93864

llvm-svn: 93863

llvm-svn: 93860

llvm-svn: 93859

doing global variable classification anymore) and hookized, sink almost
all target targets global variable emission code into AsmPrinter and out
of each target.

Some notes:

1. PIC16 does completely custom and crazy stuff, so it is not changed.
2. XCore has some custom handling for extra directives.  I'll look at it next.
3. This switches linux/ppc to use .globl instead of .global.  If .globl is
   actually wrong, let me know and I'll fix it.
4. This makes linux/ppc get a lot of random cases right which were obviously
   wrong before, it is probably now a bit healthier.
5. Blackfin will probably start getting .comm and other things that it didn't
   before.  If this is undesirable, it should explicitly opt out of these
   things by clearing the relevant fields of MCAsmInfo.

This leads to a nice diffstat:
 14 files changed, 127 insertions(+), 830 deletions(-)

llvm-svn: 93858

llvm-svn: 93857

just like all other elf targets.  Bruno, if this isn't right, please
let me know + why :)

llvm-svn: 93856

llvm-svn: 93855

I'm not sure that this is correct, but it causes no test failures,
and just emitting a .comm without protecting its linkage somehow
is surely not right.

llvm-svn: 93854

llvm-svn: 93853

llvm-svn: 93852

This makes a similar code dead in all the other targets, I'll clean it up
in a bit.

This also moves handling of lcomm up before acquisition of a section,
since lcomm never needs a section.

llvm-svn: 93851

darwin into common code.

llvm-svn: 93849

are the same.  I had already fixed a similar problem where the source and
destination were different bitcasts derived from the same alloca, but the
previous fix still did not handle the case where both operands are exactly
the same value.  Radar 7552893.

llvm-svn: 93848

duplicating the logic (differently) in lots of different targets.

llvm-svn: 93847

and everything else (weak).

llvm-svn: 93846

llvm-svn: 93845

GCC would put weak zero initialized mutable data in the .bss section,
we would put it into a crasy '.gnu.linkonce.b.test,"aw",@nobits' 
section.  Fixing this will allow simplifications next up.

llvm-svn: 93844

simplify and commonize some of the asmprinter logic for globals.

This also avoids printing the MCSection for .zerofill, which broke
the llvm-gcc build.

llvm-svn: 93843

llvm-svn: 93841

llvm-svn: 93839

1. TargetLoweringObjectFileMachO should decide if something
   goes in zerofill instead of having every target do it.
2. TargetLoweringObjectFileMachO should assign said symbols to
   the right MCSection, the asmprinters should just emit to the
   right section.
3. Since all zerofill stuff goes through mcstreamer anymore,
   MAI can have a bool "haszerofill" instead of having the textual
   directive to emit.

llvm-svn: 93838

llvm-svn: 93835

llvm-svn: 93832

llvm-svn: 93831

Fix r93758. Use isel patterns instead of c++ selection code to select rbit and make sure we pick different instructions for ARM vs. Thumb2.

llvm-svn: 93829

llvm-svn: 93818

comments (fast isel, X86).  This doesn't seem
to break any functionality, but will introduce
cases where -g affects the generated code.  I'll
be fixing that.

llvm-svn: 93811

Add comment that MDNode::getFunction() is not to be used by performance-critical code (currently only used by AsmWriter)

llvm-svn: 93802

llvm-svn: 93797

- Greatly improve the comments to the getLSDAEncoding method.

llvm-svn: 93796

aggressive changed the canonical form from sext(trunc(x)) to ashr(lshr(x)),
make sure to transform a couple more things into that canonical form,
and catch a case where we missed turning zext/shl/ashr into a single sext.

llvm-svn: 93787

llvm-svn: 93781

printing via <<. Otherwise we just print the pointer value.

llvm-svn: 93777

Instcombine does this but apparently there are situations where this pattern will escape the optimizer and / or created by isel. Here is a case that's seen in JavaScriptCore:
  %t1 = sub i32 0, %a
  %t2 = add i32 %t1, -1
The dag combiner pattern: ((c1-A)+c2) -> (c1+c2)-A
will fold it to -1 - %a.

llvm-svn: 93773

llvm-svn: 93764

Simplify MDNode::getFunction() and assertLocalFunction() by avoiding extra Function* variable and smallptrset since function-local metadata cannot be cyclic

llvm-svn: 93762

the various MOV (register) instructions (16-bit Thumb), including tBRIND (the
indirect branch).  Instead of '1', it should be specified as '?', because GPR
only specifies the register class, which includes both hi-and-lo registers.

llvm-svn: 93759