1. MachineJumpTableInfo is now created lazily for a function the first time
   it actually makes a jump table instead of for every function.
2. The encoding of jump table entries is now described by the
   MachineJumpTableInfo::JTEntryKind enum.  This enum is determined by the
   TLI::getJumpTableEncoding() hook, instead of by lots of code scattered
   throughout the compiler that "knows" that jump table entries are always
   32-bits in pic mode (for example).
3. The size and alignment of jump table entries is now calculated based on
   their kind, instead of at machinefunction creation time.

Future work includes using the EntryKind in more places in the compiler,
eliminating other logic that "knows" the layout of jump tables in various
situations.

llvm-svn: 94470

a null pointer for functions with no jump tables.  No functionality
change.

llvm-svn: 94469

when we don't have one laying around.  Useful if you don't
have an llvmcontext handy.

llvm-svn: 94468

llvm-svn: 94465

llvm-svn: 94464

llvm-svn: 94457

TargetAsmLexer.  Dialect-specific lexing code will
be placed in the functions LexTokenATT() and
LexTokenIntel().

llvm-svn: 94456

llvm-svn: 94455

of a forward-reference, which doesn't use an "abbrev" encoding.

llvm-svn: 94454

llvm-svn: 94452

already checked that TmpBB->getSinglePredecessor() is non-null.

llvm-svn: 94451

llvm-svn: 94450

llvm-svn: 94449

make it clear what it is, instead of how it is used.

llvm-svn: 94448

and MCize the non-pic case.  Now printPICJumpTableEntry really
is just about printing PIC entries.

llvm-svn: 94446

llvm-svn: 94445

llvm-svn: 94443

llvm-svn: 94441

rename it to avoid shadowing.

llvm-svn: 94440

even on x86-32/elf which uses a GOT.

llvm-svn: 94439

llvm-svn: 94438

normal form of .file would fail if the filename had a weird character
in it.

llvm-svn: 94437

llvm-svn: 94436

llvm-svn: 94435

parameter with a default value, instead of just hardcoding it in the
implementation.  The limit of MaxLookup = 6 was introduced in r69151 to fix
a performance problem with O(n^2) behavior in instcombine, but the scalarrepl
pass is relying on getUnderlyingObject to go all the way back to an AllocaInst.
Making the limit part of the method signature makes it clear that by default
the result is limited and should help avoid similar problems in the future.
This fixes pr6126.

llvm-svn: 94433

llvm-svn: 94432

llvm-svn: 94421

llvm-svn: 94417

mcstreamer.

llvm-svn: 94416

llvm-svn: 94414

llvm-svn: 94413

Previously we would just silently miscompile code that used aligned
common's, now at least you'll get a build error.  tiger-ppc already
triggered the build error because it didn't have a version of this
logic.

llvm-svn: 94412

llvm-svn: 94411

llvm-svn: 94410

llvm-svn: 94408

Patch by Torvald Riegel!

llvm-svn: 94405

We are not emitting alignments on Darwin for "bar". Not sure what is the
correct way to do it.

llvm-svn: 94400

llvm-svn: 94399

llvm-svn: 94398

It looks like linux/arm and linux/mips have the same setting, which 
are probably wrong.  Someone who cares about ARM and MIPS should 
investigate with the testcase in PR6129.

llvm-svn: 94381