llvm-svn: 115996

Now that PassInfo and Pass::ID have been separated, move the rest of the passes over to the new registration API.

llvm-svn: 111815

llvm-svn: 111366

PHI elimination shouldn't require machineloopinfo since it's used at -O0. Move the requirement to LiveIntervalAnalysis instead. Note this does not change the number of times machineloopinfo is computed.

llvm-svn: 111285

Move the decision logic whether it's a good idea to split a critical edge to clients. Also fixed an erroneous check. An edge is only a back edge when the from and to blocks are in the same loop.

llvm-svn: 111256

PHI elimination should not break back edge. It can cause some significant code placement issues. rdar://8263994

good:
LBB0_2:
  mov     r2, r0
  . . .
  mov     r1, r2
  bne     LBB0_2

bad:
LBB0_2:
  mov     r2, r0
  . . .
@ BB#3:
  mov     r1, r2
  b       LBB0_2

llvm-svn: 111221

llvm-svn: 110460

llvm-svn: 110410

ID member as the sole unique type identifier.  Clean up APIs related to this change.

llvm-svn: 110396

ScheduleDAGEmit, TwoAddressLowering, and PHIElimination.

This switches the bulk of register copies to using COPY, but many less used
copyRegToReg calls remain.

llvm-svn: 108050

into a utility routine, teach it how to update MachineLoopInfo, and
make use of it in MachineLICM to split critical edges on demand.

llvm-svn: 106555

addresses a longstanding deficiency noted in many FIXMEs scattered
across all the targets.

This effectively moves the problem up one level, replacing eleven
FIXMEs in the targets with eight FIXMEs in CodeGen, plus one path
through FastISel where we actually supply a DebugLoc, fixing Radar
7421831.

llvm-svn: 106243

doesn't have to guess.

llvm-svn: 103194

llvm-svn: 103109

Teach PHI elimination to remove REG_SEQUENCE instructions and update references of the source operands with references of the destination with subreg indices. e.g.
%reg1029<def>, %reg1030<def> = VLD1q16 %reg1024<kill>, ...
%reg1031<def> = REG_SEQUENCE %reg1029<kill>, 5, %reg1030<kill>, 6
=>
%reg1031:5<def>, %reg1031:6<def> = VLD1q16 %reg1024<kill>, ...

PHI elimination now does more than phi elimination. It is really a de-SSA pass.

llvm-svn: 103039

llvm-svn: 103013

llvm-svn: 99465

llvm-svn: 97687

Previously, LiveIntervalAnalysis would infer phi joins by looking for multiply
defined registers. That doesn't work if the phi join is implicitly defined in
all but one of the predecessors.

llvm-svn: 96994

llvm-svn: 96837

llvm-svn: 96496

into TargetOpcodes.h.  #include the new TargetOpcodes.h
into MachineInstr.  Add new inline accessors (like isPHI())
to MachineInstr, and start using them throughout the 
codebase.

llvm-svn: 95687

llvm-svn: 92566

non-landing pad basic block as the successor to a block that ends in an
unconditional jump will cause block folding to remove the added block as a
successor. Thus eventually removing it AND the landing pad entirely. Critical
edge splitting is an optimization, so we can safely turn it off when dealing
with landing pads.

llvm-svn: 91634

Tail duplication produces lots of identical phi nodes in different basic
blocks. Teach PHIElimination to reuse the join registers when lowering a phi
node that is identical to an already lowered node. This saves virtual
registers, and more importantly it avoids creating copies the the coalescer
doesn't know how to eliminate.

Teach LiveIntervalAnalysis about the phi joins with multiple uses.

This patch significantly reduces code size produced by -pre-regalloc-taildup.

llvm-svn: 91549

Patch by Howard Hinnant!

llvm-svn: 90365

We want LiveVariables clients to use methods rather than accessing the
getVarInfo data structure directly. That way it will be possible to change the
LiveVariables representation.

llvm-svn: 90240

When splitting a critical edge, the registers live through the edge are:

- Used in a PHI instruction, or
- Live out from the predecessor, and
- Live in to the successor.

This allows the coalescer to eliminate even more phi joins.

llvm-svn: 89530

critical edges in PHIElimination.

This has a huge impact on regalloc performance, and we recover almost all of
the 10% compile time regression that edge splitting introduced.

llvm-svn: 89381

when LiveVariables is available.

The -split-phi-edges is now gone, and so is the hack to disable it when using
the local register allocator. The PHIElimination pass no longer has
LiveVariables as a prerequisite - that is what broke the local allocator.
Instead we do critical edge splitting when possible - that is when
LiveVariables is available.

llvm-svn: 89213

llvm-svn: 89167

llvm-svn: 89142

llvm-svn: 89139

The local register allocator doesn't like it when LiveVariables is run.
We should also disable edge splitting under -O0, but that has to wait a bit.

llvm-svn: 89125

llvm-svn: 89082

llvm-svn: 89021

Fix for the original bug in PR5495 - Look at uses as well as defs when determining the PHI-copy insert point.

- Patch by Andrew Canis!

llvm-svn: 88880

When splitting an edge after a machine basic block with fall-through, we
forgot to insert a jump instruction. Fix this by calling updateTerminator() on
the fall-through block when relevant.

Also be more precise in PHIElimination::isLiveIn.

llvm-svn: 88728

llvm-svn: 88727

The BasicBlock associated with a MachineBasicBlock does not necessarily
correspond to the code in the MBB.

Don't insert a new IR BasicBlock when splitting critical edges. We are not
supposed to modify the IR during codegen, and we should be able to do just
fine with a NULL BB.

llvm-svn: 88707