[PeepholeOptimizer] Look through PHIs to find additional register sources

Reintroduce r245442. Remove an overly conservative assertion introduced
in r245442. We could replace the assertion to use `shareSameRegisterFile`
instead, but in that point in `insertPHI` we already lost the original
Def subreg to check against. So drop the assertion completely.

Original commit message:

- Teaches the ValueTracker in the PeepholeOptimizer to look through PHI
instructions.
- Add findNextSourceAndRewritePHI method to lookup into multiple sources
returnted by the ValueTracker and rewrite PHIs with new sources.

With these changes we can find more register sources and rewrite more
copies to allow coaslescing of bitcast instructions. Hence, we eliminate
unnecessary VR64 <-> GR64 copies in x86, but it could be extended to
other archs by marking "isBitcast" on target specific instructions. The
x86 example follows:

A:
  psllq %mm1, %mm0
  movd  %mm0, %r9
  jmp C

B:
  por %mm1, %mm0
  movd  %mm0, %r9
  jmp C

C:
  movd  %r9, %mm0
  pshufw  $238, %mm0, %mm0

Becomes:

A:
  psllq %mm1, %mm0
  jmp C

B:
  por %mm1, %mm0
  jmp C

C:
  pshufw  $238, %mm0, %mm0

Differential Revision: http://reviews.llvm.org/D11197
rdar://problem/20404526

1 files changed, 2 insertions, 1 deletions

diff --git a/llvm/lib/Target/X86/X86InstrMMX.td b/llvm/lib/Target/X86/X86InstrMMX.td
index eaa7894004c..ab301adca48 100644
--- a/llvm/lib/Target/X86/X86InstrMMX.td
+++ b/llvm/lib/Target/X86/X86InstrMMX.td
@@ -249,6 +249,7 @@ def MMX_MOVD64grr : MMXI<0x7E, MRMDestReg, (outs GR32:$dst), (ins VR64:$src),
                           (MMX_X86movd2w (x86mmx VR64:$src)))],
                           IIC_MMX_MOV_REG_MM>, Sched<[WriteMove]>;
 
+let isBitcast = 1 in
 def MMX_MOVD64to64rr : MMXRI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR64:$src),
                              "movd\t{$src, $dst|$dst, $src}",
                              [(set VR64:$dst, (bitconvert GR64:$src))],
@@ -262,7 +263,7 @@ def MMX_MOVD64to64rm : MMXRI<0x6E, MRMSrcMem, (outs VR64:$dst),
 // These are 64 bit moves, but since the OS X assembler doesn't
 // recognize a register-register movq, we write them as
 // movd.
-let SchedRW = [WriteMove] in {
+let SchedRW = [WriteMove], isBitcast = 1 in {
 def MMX_MOVD64from64rr : MMXRI<0x7E, MRMDestReg,
                                (outs GR64:$dst), (ins VR64:$src),
                                "movd\t{$src, $dst|$dst, $src}",