llvm-6502/lib/CodeGen
2008-07-09 21:53:02 +00:00
..
SelectionDAG Move the IsVolatile and SVOffset fields into the MemSDNode base 2008-07-09 21:23:02 +00:00
AsmPrinter.cpp Switch to new section name handling facility 2008-07-09 13:27:16 +00:00
BranchFolding.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
Collector.cpp Use of operator* is redundant and confusing here. 2008-07-07 20:08:05 +00:00
CollectorMetadata.cpp Add a new (simple) StringMap::clear method, patch by Pratik 2008-07-02 05:26:32 +00:00
Collectors.cpp
DwarfWriter.cpp Avoid creating expensive comment string if it's not going to be printed. 2008-07-09 21:53:02 +00:00
ELFWriter.cpp
ELFWriter.h
IfConversion.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
IntrinsicLowering.cpp
LiveInterval.cpp Instead of adding an isSS field to LiveInterval to denote stack slot. Use top bit of 'reg' instead. If the top bit is set, than the LiveInterval represents a stack slot live interval. 2008-06-23 21:03:19 +00:00
LiveIntervalAnalysis.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
LiveStackAnalysis.cpp Add a stack slot coloring pass. Not yet enabled. 2008-06-04 09:18:41 +00:00
LiveVariables.cpp - Remove calls to copyKillDeadInfo which is an N^2 function. Instead, propagate kill / dead markers as new instructions are constructed in foldMemoryOperand, convertToThressAddress, etc. 2008-07-03 09:09:37 +00:00
LLVMTargetMachine.cpp Don't run stack slot coloring if -fast. 2008-06-30 22:33:16 +00:00
LoopAligner.cpp
LowerSubregs.cpp It's not safe to remove SUBREG_TO_REG that looks like identity copies, e.g. movl %eax, %eax on x86-64 actually does a zero-extend. 2008-06-17 17:59:16 +00:00
MachineBasicBlock.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
MachineDominators.cpp
MachineFunction.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
MachineInstr.cpp It's no longer necessary to test if a MachineBasicBlock's 2008-07-08 23:59:09 +00:00
MachineLICM.cpp Prevent option name conflict. 2008-07-07 05:42:27 +00:00
MachineLoopInfo.cpp
MachineModuleInfo.cpp Make the DICountVisitor not a visitor. This keeps us from calling virtual 2008-07-09 06:02:33 +00:00
MachinePassRegistry.cpp
MachineRegisterInfo.cpp Assert that all MachineInstrs update PhysRegUseDefLists in 2008-07-07 19:55:35 +00:00
MachineSink.cpp
MachOWriter.cpp
MachOWriter.h
Makefile
OcamlCollector.cpp
Passes.cpp
PHIElimination.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
PhysRegTracker.h
PostRASchedulerList.cpp
PrologEpilogInserter.cpp No need to align the stack if there are no stack 2008-06-30 20:40:16 +00:00
PseudoSourceValue.cpp
README.txt Enable stack coloring by default. 2008-06-06 19:52:44 +00:00
RegAllocBigBlock.cpp
RegAllocLinearScan.cpp A brief survey of priority_queue usage in the tree turned this up 2008-06-23 23:51:16 +00:00
RegAllocLocal.cpp Loosen our check here. Local regalloc only cares that the reg is used and def'd by the same instruction, but about the details of 2008-07-09 21:34:36 +00:00
RegAllocSimple.cpp Reuse the MO variable instead of recomputing it in RegAllocLocal. 2008-07-09 20:12:26 +00:00
RegisterCoalescer.cpp
RegisterScavenging.cpp Minor const-correctness fixes. 2008-07-07 20:06:06 +00:00
ShadowStackCollector.cpp
SimpleRegisterCoalescing.cpp Use getMBBEndIdx rather than assuming that the end is right after the last instruction in the block. 2008-06-23 22:12:23 +00:00
SimpleRegisterCoalescing.h Coalesce copy from one register class to a sub register class. e.g. X86::MOV16to16_. 2008-06-19 01:39:21 +00:00
StackSlotColoring.cpp Remove option used to debug stack coloring bugs. It's no longer needed since stack coloring is now bug free. 2008-06-23 21:24:32 +00:00
StrongPHIElimination.cpp Remove debugging code. 2008-06-05 18:43:34 +00:00
TargetInstrInfoImpl.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
TwoAddressInstructionPass.cpp - Remove calls to copyKillDeadInfo which is an N^2 function. Instead, propagate kill / dead markers as new instructions are constructed in foldMemoryOperand, convertToThressAddress, etc. 2008-07-03 09:09:37 +00:00
UnreachableBlockElim.cpp
VirtRegMap.cpp Pool-allocation for MachineInstrs, MachineBasicBlocks, and 2008-07-07 23:14:23 +00:00
VirtRegMap.h Move #include to right place. 2008-06-04 09:16:33 +00:00

//===---------------------------------------------------------------------===//

Common register allocation / spilling problem:

        mul lr, r4, lr
        str lr, [sp, #+52]
        ldr lr, [r1, #+32]
        sxth r3, r3
        ldr r4, [sp, #+52]
        mla r4, r3, lr, r4

can be:

        mul lr, r4, lr
        mov r4, lr
        str lr, [sp, #+52]
        ldr lr, [r1, #+32]
        sxth r3, r3
        mla r4, r3, lr, r4

and then "merge" mul and mov:

        mul r4, r4, lr
        str lr, [sp, #+52]
        ldr lr, [r1, #+32]
        sxth r3, r3
        mla r4, r3, lr, r4

It also increase the likelyhood the store may become dead.

//===---------------------------------------------------------------------===//

I think we should have a "hasSideEffects" flag (which is automatically set for
stuff that "isLoad" "isCall" etc), and the remat pass should eventually be able
to remat any instruction that has no side effects, if it can handle it and if
profitable.

For now, I'd suggest having the remat stuff work like this:

1. I need to spill/reload this thing.
2. Check to see if it has side effects.
3. Check to see if it is simple enough: e.g. it only has one register
destination and no register input.
4. If so, clone the instruction, do the xform, etc.

Advantages of this are:

1. the .td file describes the behavior of the instructions, not the way the
   algorithm should work.
2. as remat gets smarter in the future, we shouldn't have to be changing the .td
   files.
3. it is easier to explain what the flag means in the .td file, because you
   don't have to pull in the explanation of how the current remat algo works.

Some potential added complexities:

1. Some instructions have to be glued to it's predecessor or successor. All of
   the PC relative instructions and condition code setting instruction. We could
   mark them as hasSideEffects, but that's not quite right. PC relative loads
   from constantpools can be remat'ed, for example. But it requires more than
   just cloning the instruction. Some instructions can be remat'ed but it
   expands to more than one instruction. But allocator will have to make a
   decision.

4. As stated in 3, not as simple as cloning in some cases. The target will have
   to decide how to remat it. For example, an ARM 2-piece constant generation
   instruction is remat'ed as a load from constantpool.

//===---------------------------------------------------------------------===//

bb27 ...
        ...
        %reg1037 = ADDri %reg1039, 1
        %reg1038 = ADDrs %reg1032, %reg1039, %NOREG, 10
    Successors according to CFG: 0x8b03bf0 (#5)

bb76 (0x8b03bf0, LLVM BB @0x8b032d0, ID#5):
    Predecessors according to CFG: 0x8b0c5f0 (#3) 0x8b0a7c0 (#4)
        %reg1039 = PHI %reg1070, mbb<bb76.outer,0x8b0c5f0>, %reg1037, mbb<bb27,0x8b0a7c0>

Note ADDri is not a two-address instruction. However, its result %reg1037 is an
operand of the PHI node in bb76 and its operand %reg1039 is the result of the
PHI node. We should treat it as a two-address code and make sure the ADDri is
scheduled after any node that reads %reg1039.

//===---------------------------------------------------------------------===//

Use local info (i.e. register scavenger) to assign it a free register to allow
reuse:
	ldr r3, [sp, #+4]
	add r3, r3, #3
	ldr r2, [sp, #+8]
	add r2, r2, #2
	ldr r1, [sp, #+4]  <==
	add r1, r1, #1
	ldr r0, [sp, #+4]
	add r0, r0, #2

//===---------------------------------------------------------------------===//

LLVM aggressively lift CSE out of loop. Sometimes this can be negative side-
effects:

R1 = X + 4
R2 = X + 7
R3 = X + 15

loop:
load [i + R1]
...
load [i + R2]
...
load [i + R3]

Suppose there is high register pressure, R1, R2, R3, can be spilled. We need
to implement proper re-materialization to handle this:

R1 = X + 4
R2 = X + 7
R3 = X + 15

loop:
R1 = X + 4  @ re-materialized
load [i + R1]
...
R2 = X + 7 @ re-materialized
load [i + R2]
...
R3 = X + 15 @ re-materialized
load [i + R3]

Furthermore, with re-association, we can enable sharing:

R1 = X + 4
R2 = X + 7
R3 = X + 15

loop:
T = i + X
load [T + 4]
...
load [T + 7]
...
load [T + 15]
//===---------------------------------------------------------------------===//

It's not always a good idea to choose rematerialization over spilling. If all
the load / store instructions would be folded then spilling is cheaper because
it won't require new live intervals / registers. See 2003-05-31-LongShifts for
an example.

//===---------------------------------------------------------------------===//

With a copying garbage collector, derived pointers must not be retained across
collector safe points; the collector could move the objects and invalidate the
derived pointer. This is bad enough in the first place, but safe points can
crop up unpredictably. Consider:

        %array = load { i32, [0 x %obj] }** %array_addr
        %nth_el = getelementptr { i32, [0 x %obj] }* %array, i32 0, i32 %n
        %old = load %obj** %nth_el
        %z = div i64 %x, %y
        store %obj* %new, %obj** %nth_el

If the i64 division is lowered to a libcall, then a safe point will (must)
appear for the call site. If a collection occurs, %array and %nth_el no longer
point into the correct object.

The fix for this is to copy address calculations so that dependent pointers
are never live across safe point boundaries. But the loads cannot be copied
like this if there was an intervening store, so may be hard to get right.

Only a concurrent mutator can trigger a collection at the libcall safe point.
So single-threaded programs do not have this requirement, even with a copying
collector. Still, LLVM optimizations would probably undo a front-end's careful
work.

//===---------------------------------------------------------------------===//

The ocaml frametable structure supports liveness information. It would be good
to support it.

//===---------------------------------------------------------------------===//

The FIXME in ComputeCommonTailLength in BranchFolding.cpp needs to be
revisited. The check is there to work around a misuse of directives in inline
assembly.

//===---------------------------------------------------------------------===//

It would be good to detect collector/target compatibility instead of silently
doing the wrong thing.

//===---------------------------------------------------------------------===//

It would be really nice to be able to write patterns in .td files for copies,
which would eliminate a bunch of explicit predicates on them (e.g. no side 
effects).  Once this is in place, it would be even better to have tblgen 
synthesize the various copy insertion/inspection methods in TargetInstrInfo.

//===---------------------------------------------------------------------===//

Stack coloring improvments:

1. Do proper LiveStackAnalysis on all stack objects including those which are
   not spill slots.
2. Reorder objects to fill in gaps between objects.
   e.g. 4, 1, <gap>, 4, 1, 1, 1, <gap>, 4 => 4, 1, 1, 1, 1, 4, 4