Sanjay Patel 8f61a6eb1f [DAGCombiner] use narrow vector ops to eliminate concat/extract (PR32790)
In the best case:
extract (binop (concat X1, X2), (concat Y1, Y2)), N --> binop XN, YN
...we kill all of the extract/concat and just have narrow binops remaining.

If only one of the binop operands is amenable, this transform is still
worthwhile because we kill some of the extract/concat.

Optional bitcasting makes the code more complicated, but there doesn't
seem to be a way to avoid that.

The TODO about extending to more than bitwise logic is there because we really
will regress several x86 tests including madd, psad, and even a plain
integer-multiply-by-2 or shift-left-by-1. I don't think there's anything
fundamentally wrong with this patch that would cause those regressions; those
folds are just missing or brittle.

If we extend to more binops, I found that this patch will fire on at least one
non-x86 regression test. There's an ARM NEON test in
test/CodeGen/ARM/coalesce-subregs.ll with a pattern like:

            t5: v2f32 = vector_shuffle<0,3> t2, t4
          t6: v1i64 = bitcast t5
          t8: v1i64 = BUILD_VECTOR Constant:i64<0>
        t9: v2i64 = concat_vectors t6, t8
      t10: v4f32 = bitcast t9
    t12: v4f32 = fmul t11, t10
  t13: v2i64 = bitcast t12
t16: v1i64 = extract_subvector t13, Constant:i32<0>

There was no functional change in the codegen from this transform from what I
could see though.

For the x86 test changes:

1. PR32790() is the closest call. We don't reduce the AVX1 instruction count in that case,
   but we improve throughput. Also, on a core like Jaguar that double-pumps 256-bit ops,
   there's an unseen win because two 128-bit ops have the same cost as the wider 256-bit op.
   SSE/AVX2/AXV512 are not affected which is expected because only AVX1 has the extract/concat
   ops to match the pattern.
2. do_not_use_256bit_op() is the best case. Everyone wins by avoiding the concat/extract.
   Related bug for IR filed as: https://bugs.llvm.org/show_bug.cgi?id=33026
3. The SSE diffs in vector-trunc-math.ll are just scheduling/RA, so nothing real AFAICT.
4. The AVX1 diffs in vector-tzcnt-256.ll are all the same pattern: we reduced the instruction
   count by one in each case by eliminating two insert/extract while adding one narrower logic op.

https://bugs.llvm.org/show_bug.cgi?id=32790

Differential Revision: https://reviews.llvm.org/D33137



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@303997 91177308-0d34-0410-b5e6-96231b3b80d8
2017-05-26 15:33:18 +00:00
..
2017-01-31 17:00:27 +00:00
2017-05-13 10:59:05 +00:00
2016-06-29 20:37:43 +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.