llvm-svn: 148173

llvm-svn: 148172

llvm-svn: 148171

llvm-svn: 148170

live across BBs before register allocation. This miscompiled 197.parser
when a cmp + b are optimized to a cbnz instruction even though the CPSR def
is live-in a successor.
        cbnz    r6, LBB89_12
...
LBB89_12:
        ble     LBB89_1

The fix consists of two parts. 1) Teach LiveVariables that some unallocatable
registers might be liveouts so don't mark their last use as kill if they are.
2) ARM constantpool island pass shouldn't form cbz / cbnz if the conditional
branch does not kill CPSR.

rdar://10676853

llvm-svn: 148168

llvm-svn: 148167

llvm-svn: 148166

The hook returns a bit-mask of call-preserved registers that will
eventually replace the current list of implicit defs on call
instructions.  This will make it possible to support multiple calling
conventions without duplicating call instruction descriptors.

The call-preserved mask is slightly different from the list returned by
the getCalleeSavedRegs() hook, it includes all aliases that are
preserved by calls.

The hook takes a CallingConv::ID argument instead of a MachineFunction
pointer, so it can provide information about calls to extern functions,
and even indirect function calls.

TRI::getCalleeSavedRegs() returns information about the function
currently being compiled. TRI::getCallPreservedMask() returns
information about the functions it is calling.

llvm-svn: 148165

llvm-svn: 148164

llvm-svn: 148161

llvm-svn: 148156

The QQ and QQQQ registers are not 'real', they are pseudo-registers used
to model some vld and vst instructions.

This makes the call clobber lists longer, but I intend to get rid of
those soon.

llvm-svn: 148151

llvm-svn: 148150

llvm-svn: 148149

llvm-svn: 148147

llvm-svn: 148143

llvm-svn: 148136

Revert r148131, it was committed before it was ready.

llvm-svn: 148134

llvm-svn: 148133

LoopUnswitch: All helper data that is collected during loop-unswitch iterations was moved to separated class (LUAnalysisCache). 

llvm-svn: 148132

llvm-svn: 148131

llvm-svn: 148128

llvm-svn: 148123

Convert SHUFPD with the same register for both sources to PSHUFD if it would prevent a register copy. Similar to SHUFPS, but requires the mask to be converted.

llvm-svn: 148112

llvm-svn: 148109

Make X86 instruction selection use 256-bit VPXOR for build_vector of all ones if AVX2 is enabled. This gives the ExeDepsFix pass a chance to choose FP vs int as appropriate. Also use v8i32 as the type for getZeroVector if AVX2 is enabled. This is consistent with SSE2 using prefering v4i32.

llvm-svn: 148108

llvm-svn: 148107

llvm-svn: 148106

llvm-svn: 148105

llvm-svn: 148104

llvm-svn: 148103

llvm-svn: 148102

llvm-svn: 148101

Fix typo in PerformAddCombine that caused any vector type to be checked for horizontal add/sub if AVX2 is enabled. This caused an assert to fail for non 128/256-bit vectors when done before type legalizing. Fixes PR11749.

llvm-svn: 148096

The code type was always identical to a string anyway. Now it is simply
a synonym. The code literal syntax [{...}] is still valid.

llvm-svn: 148092

This avoids a gazillion StringMap and dynamic_cast calls, making
TableGen run 3x faster.

llvm-svn: 148091

overly conservative. It was concerned about cases where it would prohibit
folding simple [r, c] addressing modes. e.g.
  ldr r0, [r2]
  ldr r1, [r2, #4]
=>
  ldr r0, [r2], #4
  ldr r1, [r2]
Change the logic to look for such cases which allows it to form indexed memory
ops more aggressively.

rdar://10674430

llvm-svn: 148086

llvm-svn: 148077

the optimizer doesn't eliminate objc_retainBlock calls which are needed
for their side effect of copying blocks onto the heap.
This implements rdar://10361249.

llvm-svn: 148076

llvm-svn: 148067