Commit Graph

141 Commits

Author SHA1 Message Date
Owen Anderson df7a4f2515 Now with fewer extraneous semicolons!
llvm-svn: 115996
2010-10-07 22:25:06 +00:00
Owen Anderson d31d82d75c Now that PassInfo and Pass::ID have been separated, move the rest of the passes over to the new registration API.
llvm-svn: 111815
2010-08-23 17:52:01 +00:00
Jakob Stoklund Olesen 952a621d93 Preserve subregs on PHI source operands. Patch by Krister Wombell!
llvm-svn: 111366
2010-08-18 16:09:47 +00:00
Evan Cheng 16bfe5b0f5 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
2010-08-17 21:00:37 +00:00
Evan Cheng 647c559172 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
2010-08-17 17:43:50 +00:00
Evan Cheng f259efde47 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
2010-08-17 01:20:36 +00:00
Owen Anderson a7aed18624 Reapply r110396, with fixes to appease the Linux buildbot gods.
llvm-svn: 110460
2010-08-06 18:33:48 +00:00
Owen Anderson bda59bd247 Revert r110396 to fix buildbots.
llvm-svn: 110410
2010-08-06 00:23:35 +00:00
Owen Anderson 755aceb5d0 Don't use PassInfo* as a type identifier for passes. Instead, use the address of the static
ID member as the sole unique type identifier.  Clean up APIs related to this change.

llvm-svn: 110396
2010-08-05 23:42:04 +00:00
Jakob Stoklund Olesen e50d30d586 Emit COPY instructions instead of using copyRegToReg in InstrEmitter,
ScheduleDAGEmit, TwoAddressLowering, and PHIElimination.

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

llvm-svn: 108050
2010-07-10 19:08:25 +00:00
Dan Gohman 3570f81b1e Move PHIElimination's SplitCriticalEdge for MachineBasicBlocks out
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
2010-06-22 17:25:57 +00:00
Stuart Hastings 0125b6410a Add a DebugLoc parameter to TargetInstrInfo::InsertBranch(). This
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
2010-06-17 22:43:56 +00:00
Dan Gohman 779c69bbc5 Add a DebugLoc argument to TargetInstrInfo::copyRegToReg, so that it
doesn't have to guess.

llvm-svn: 103194
2010-05-06 20:33:48 +00:00
Evan Cheng 4b6abd8c2b Move REG_SEQUENCE removal to 2addr pass.
llvm-svn: 103109
2010-05-05 18:45:40 +00:00
Evan Cheng 4c908f4181 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
2010-05-04 20:26:52 +00:00
Evan Cheng a5c0cc329e Rename variables for consistency.
llvm-svn: 103013
2010-05-04 17:12:26 +00:00
Evan Cheng eae56b5717 Avoid being influenced by dbg_value instructions.
llvm-svn: 99465
2010-03-25 01:01:37 +00:00
Evan Cheng 2d23779e7d Remove PHINodeTraits and use MachineInstrExpressionTrait instead.
llvm-svn: 97687
2010-03-03 23:55:49 +00:00
Jakob Stoklund Olesen 38b76e27a7 Keep track of phi join registers explicitly in LiveVariables.
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
2010-02-23 22:43:58 +00:00
Jakob Stoklund Olesen 012d4369f6 Dead code elimination
llvm-svn: 96837
2010-02-23 00:28:53 +00:00
Jakob Stoklund Olesen 5d583da601 Dead code elimination.
llvm-svn: 96496
2010-02-17 19:13:56 +00:00
Chris Lattner b06015aa69 move target-independent opcodes out of TargetInstrInfo
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
2010-02-09 19:54:29 +00:00
David Greene 2555292e44 Change errs() to dbgs().
llvm-svn: 92566
2010-01-05 01:24:24 +00:00
Bill Wendling 819c356a09 Turn off critical edge splitting for landing pads. The introduction of a
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
2009-12-17 23:42:32 +00:00
Jakob Stoklund Olesen ec20a88a57 Reuse lowered phi nodes.
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
2009-12-16 18:55:53 +00:00
Chris Lattner a48f44d9ee improve portability to avoid conflicting with std::next in c++'0x.
Patch by Howard Hinnant!

llvm-svn: 90365
2009-12-03 00:50:42 +00:00
Jakob Stoklund Olesen defc47088a Move PHIElimination::isLiveOut method to LiveVariables.
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
2009-12-01 17:13:31 +00:00
Jakob Stoklund Olesen bc630aca39 Be more clever about calculating live variables through new basic blocks.
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
2009-11-21 02:05:21 +00:00
Jakob Stoklund Olesen 331e6880a5 Place new basic blocks immediately after their predecessor when splitting
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
2009-11-19 19:42:16 +00:00
Jakob Stoklund Olesen 736888f7fa Don't require LiveVariables for PHIElimination. Enable critical edge splitting
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
2009-11-18 18:01:35 +00:00
Jakob Stoklund Olesen 4797e58d6b Fix inverted test and add testcase from failing self-host.
llvm-svn: 89167
2009-11-18 00:02:18 +00:00
Jakob Stoklund Olesen 31e24b2d0c Disable -split-phi-edges to unbreak the buildbots
llvm-svn: 89142
2009-11-17 21:23:49 +00:00
Jakob Stoklund Olesen f992022e54 Never call UpdateTerminator() when AnalyzeBranch would fail.
llvm-svn: 89139
2009-11-17 20:46:00 +00:00
Jakob Stoklund Olesen fffff88a3c Enable -split-phi-edges by default, except when -regalloc=local.
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
2009-11-17 19:15:50 +00:00
Evan Cheng 84efacfaad Revert 89021. It's miscompiling llvm-gcc driver driver at -O0.
llvm-svn: 89082
2009-11-17 09:55:52 +00:00
Jakob Stoklund Olesen 9f0d55d8d8 Enable -split-phi-edges by default
llvm-svn: 89021
2009-11-17 01:07:22 +00:00
Lang Hames 968bd90f88 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
2009-11-16 02:00:09 +00:00
Jakob Stoklund Olesen 24b027401e Fix bug in -split-phi-edges.
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
2009-11-14 00:38:13 +00:00
Jakob Stoklund Olesen 15ca009ee7 Update MachineDominator information
llvm-svn: 88727
2009-11-14 00:38:06 +00:00
Jakob Stoklund Olesen ad205d6102 Fix PHIElimination optimization that uses MBB->getBasicBlock.
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
2009-11-13 21:56:15 +00:00
Daniel Dunbar 5026bc3a86 Fix -Asserts warning.
llvm-svn: 87024
2009-11-12 20:53:43 +00:00
Jakob Stoklund Olesen 4f7fd3baeb Fix liveness calculation when splitting critical edges during PHI elimination.
- Edges are split before any phis are eliminated, so the code is SSA.

- Create a proper IR BasicBlock for the split edges.

- LiveVariables::addNewBlock now has same syntax as
  MachineDominatorTree::addNewBlock. Algorithm calculates predecessor live-out
  set rather than successor live-in set.

This feature still causes some miscompilations.

llvm-svn: 86867
2009-11-11 19:31:31 +00:00
Jakob Stoklund Olesen 4453dc976b Teach PHIElimination to split critical edges when -split-phi-edges is enabled.
Critical edges leading to a PHI node are split when the PHI source variable is
live out from the predecessor block. This help the coalescer eliminate more
PHI joins.

llvm-svn: 86725
2009-11-10 22:01:05 +00:00
Jakob Stoklund Olesen 19f235e455 Refactoring: Extract method PHIElimination::isLiveOut().
Clean up some whitespace.
No functional changes.

llvm-svn: 86724
2009-11-10 22:00:56 +00:00
Dan Gohman 0402315d41 Use setPreservesAll and setPreservesCFG in CodeGen passes.
llvm-svn: 77754
2009-07-31 23:37:33 +00:00
Lang Hames 077415e84b For real this time: PHI Def & Kill tracking added to PHIElimination.
llvm-svn: 76865
2009-07-23 05:44:24 +00:00
Lang Hames a77a3c3782 Added PHI Def & Kill tracking to PHIElimination pass.
llvm-svn: 76849
2009-07-23 04:34:03 +00:00
Lang Hames aa037759b2 Exposed PHIElimination pass within CodeGen.
llvm-svn: 76688
2009-07-21 23:47:33 +00:00
Evan Cheng 0878f15fa1 Revert 74898. It broke several tests.
llvm-svn: 74925
2009-07-07 17:50:43 +00:00
Sanjiv Gupta 031bbef528 if the terminator is a branch depending upon the side effects of a
previous cmp; a copy can not be inserted here if the copy insn also has
 side effects. We don't have access to the attributes of copy insn here;
 so just play safe by finding a safe locations for branch terminators.

llvm-svn: 74898
2009-07-07 08:04:51 +00:00