Quentin Colombet 9d60e0ff0a [RegAllocGreedy] Introduce a late pass to repair broken hints.
A broken hint is a copy where both ends are assigned different colors. When a
variable gets evicted in the neighborhood of such copies, it is likely we can
reconcile some of them.


** Context **

Copies are inserted during the register allocation via splitting. These split
points are required to relax the constraints on the allocation problem. When
such a point is inserted, both ends of the copy would not share the same color
with respect to the current allocation problem. When variables get evicted,
the allocation problem becomes different and some split point may not be
required anymore. However, the related variables may already have been colored.

This usually shows up in the assembly with pattern like this:
def A
...
save A to B
def A
use A
restore A from B
...
use B

Whereas we could simply have done:
def B
...
def A
use A
...
use B


** Proposed Solution **

A variable having a broken hint is marked for late recoloring if and only if
selecting a register for it evict another variable. Indeed, if no eviction
happens this is pointless to look for recoloring opportunities as it means the
situation was the same as the initial allocation problem where we had to break
the hint.

Finally, when everything has been allocated, we look for recoloring
opportunities for all the identified candidates.
The recoloring is performed very late to rely on accurate copy cost (all
involved variables are allocated).
The recoloring is simple unlike the last change recoloring. It propagates the
color of the broken hint to all its copy-related variables. If the color is
available for them, the recoloring uses it, otherwise it gives up on that hint
even if a more complex coloring would have worked.

The recoloring happens only if it is profitable. The profitability is evaluated
using the expected frequency of the copies of the currently recolored variable
with a) its current color and b) with the target color. If a) is greater or
equal than b), then it is profitable and the recoloring happen.


** Example **

Consider the following example:
BB1:
  a =
  b =
BB2:
  ...
   = b
   = a
Let us assume b gets split:
BB1:
  a =
  b =
BB2:
  c = b
  ...
  d = c
  = d
  = a
Because of how the allocation work, b, c, and d may be assigned different
colors. Now, if a gets evicted to make room for c, assuming b and d were
assigned to something different than a.
We end up with:
BB1:
  a =
  st a, SpillSlot
  b =
BB2:
  c = b
  ...
  d = c
  = d
  e = ld SpillSlot
  = e
This is likely that we can assign the same register for b, c, and d,
getting rid of 2 copies.


** Performances **

Both ARM64 and x86_64 show performance improvements of up to 3% for the
llvm-testsuite + externals with Os and O3. There are a few regressions too that
comes from the (in)accuracy of the block frequency estimate.

<rdar://problem/18312047>


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225422 91177308-0d34-0410-b5e6-96231b3b80d8
2015-01-08 01:16:39 +00:00
..
2015-01-07 21:35:13 +00:00
2014-12-09 18:38:53 +00:00
2014-10-29 15:23:11 +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 r4, [sp, #+52]
        ldr lr, [r1, #+32]
        sxth r3, r3
        mla r4, r3, lr, r4

It also increase the likelihood the store may become dead.

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

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 improvements:

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

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

The scheduler should be able to sort nearby instructions by their address. For
example, in an expanded memset sequence it's not uncommon to see code like this:

  movl $0, 4(%rdi)
  movl $0, 8(%rdi)
  movl $0, 12(%rdi)
  movl $0, 0(%rdi)

Each of the stores is independent, and the scheduler is currently making an
arbitrary decision about the order.

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

Another opportunitiy in this code is that the $0 could be moved to a register:

  movl $0, 4(%rdi)
  movl $0, 8(%rdi)
  movl $0, 12(%rdi)
  movl $0, 0(%rdi)

This would save substantial code size, especially for longer sequences like
this. It would be easy to have a rule telling isel to avoid matching MOV32mi
if the immediate has more than some fixed number of uses. It's more involved
to teach the register allocator how to do late folding to recover from
excessive register pressure.