Commit e91e1df6 authored by Amara Emerson's avatar Amara Emerson
Browse files

[GlobalISel][Localizer] Enable intra-block localization of already-local uses. 

This changes the localizer to attempt intra-block localizer of instructions
that have local uses. This is useful because sometimes the entry block itself
has many uses of constant-like instructions, which would benefit from shortening
live ranges. Previously if an inst had no non-local uses, we wouldn't add it to
the list of instructions to attempt further intra-block localization.

This gives a 0.7% geomean code size improvement on CTMark.

Differential Revision: https://reviews.llvm.org/D75555
parent 7af4374f

llvm/lib/CodeGen/GlobalISel/Localizer.cpp

+6 −1
Original line number Diff line number Diff line
@@ -86,8 +86,13 @@ bool Localizer::localizeInterBlock(MachineFunction &MF, 
      LLVM_DEBUG(MachineInstr &MIUse = *MOUse.getParent();

                 dbgs() << "Checking use: " << MIUse

                        << " #Opd: " << MIUse.getOperandNo(&MOUse) << '\n');

      if (isLocalUse(MOUse, MI, InsertMBB))

      if (isLocalUse(MOUse, MI, InsertMBB)) {

        // Even if we're in the same block, if the block is very large we could

        // still have many long live ranges. Try to do intra-block localization

        // too.

        LocalizedInstrs.insert(&MI);

        continue;

      }

      LLVM_DEBUG(dbgs() << "Fixing non-local use\n");

      Changed = true;

      auto MBBAndReg = std::make_pair(InsertMBB, Reg);

llvm/test/CodeGen/AArch64/GlobalISel/localizer-arm64-tti.ll

+5 −5
Original line number Diff line number Diff line
@@ -19,14 +19,14 @@ define i32 @foo() { 
  ; CHECK: bb.1.entry:

  ; CHECK:   successors: %bb.2(0x40000000), %bb.3(0x40000000)

  ; CHECK:   [[GV:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var1

  ; CHECK:   [[C:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 1

  ; CHECK:   [[C1:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 2

  ; CHECK:   [[C:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 2

  ; CHECK:   [[GV1:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var2

  ; CHECK:   [[C2:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 3

  ; CHECK:   [[C1:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 3

  ; CHECK:   [[GV2:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var3

  ; CHECK:   [[C3:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 0

  ; CHECK:   [[C2:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 0

  ; CHECK:   [[LOAD:%[0-9]+]]:gpr(s32) = G_LOAD [[GV]](p0) :: (dereferenceable load 4 from @var1)

  ; CHECK:   [[ICMP:%[0-9]+]]:gpr(s32) = G_ICMP intpred(ne), [[LOAD]](s32), [[C]]

  ; CHECK:   [[C3:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 1

  ; CHECK:   [[ICMP:%[0-9]+]]:gpr(s32) = G_ICMP intpred(ne), [[LOAD]](s32), [[C3]]

  ; CHECK:   [[TRUNC:%[0-9]+]]:gpr(s1) = G_TRUNC [[ICMP]](s32)

  ; CHECK:   G_BRCOND [[TRUNC]](s1), %bb.3

  ; CHECK: bb.2.if.then:

llvm/test/CodeGen/AArch64/GlobalISel/localizer.mir

+41 −15
Original line number Diff line number Diff line
@@ -39,6 +39,7 @@ 
  }



  define void @test_inttoptr() { ret void }

  define void @many_local_use_intra_block() { ret void }



...
@@ -335,15 +336,15 @@ body: | 
  ; CHECK-LABEL: name: intrablock_with_globalvalue

  ; CHECK: bb.0.entry:

  ; CHECK:   successors: %bb.1(0x40000000), %bb.2(0x40000000)

  ; CHECK:   [[GV:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var1

  ; CHECK:   [[C:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 1

  ; CHECK:   [[C1:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 2

  ; CHECK:   [[GV1:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var2

  ; CHECK:   [[C2:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 3

  ; CHECK:   [[GV2:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var3

  ; CHECK:   [[C3:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 0

  ; CHECK:   [[LOAD:%[0-9]+]]:gpr(s32) = G_LOAD [[GV]](p0) :: (load 4 from @var1)

  ; CHECK:   [[ICMP:%[0-9]+]]:gpr(s32) = G_ICMP intpred(eq), [[LOAD]](s32), [[C]]

  ; CHECK:   [[C:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 2

  ; CHECK:   [[GV:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var2

  ; CHECK:   [[C1:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 3

  ; CHECK:   [[GV1:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var3

  ; CHECK:   [[C2:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 0

  ; CHECK:   [[GV2:%[0-9]+]]:gpr(p0) = G_GLOBAL_VALUE @var1

  ; CHECK:   [[LOAD:%[0-9]+]]:gpr(s32) = G_LOAD [[GV2]](p0) :: (load 4 from @var1)

  ; CHECK:   [[C3:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 1

  ; CHECK:   [[ICMP:%[0-9]+]]:gpr(s32) = G_ICMP intpred(eq), [[LOAD]](s32), [[C3]]

  ; CHECK:   [[TRUNC:%[0-9]+]]:gpr(s1) = G_TRUNC [[ICMP]](s32)

  ; CHECK:   G_BRCOND [[TRUNC]](s1), %bb.1

  ; CHECK:   G_BR %bb.2
@@ -403,12 +404,12 @@ body: | 
  ; CHECK:   liveins: $w0, $x1

  ; CHECK:   [[COPY:%[0-9]+]]:gpr(s32) = COPY $w0

  ; CHECK:   [[COPY1:%[0-9]+]]:gpr(p0) = COPY $x1

  ; CHECK:   [[C:%[0-9]+]]:gpr(s64) = G_CONSTANT i64 128

  ; CHECK:   [[C1:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 0

  ; CHECK:   [[C2:%[0-9]+]]:gpr(s64) = G_CONSTANT i64 0

  ; CHECK:   [[INTTOPTR:%[0-9]+]]:gpr(p0) = G_INTTOPTR [[C2]](s64)

  ; CHECK:   [[INTTOPTR1:%[0-9]+]]:gpr(p0) = G_INTTOPTR [[C]](s64)

  ; CHECK:   [[ICMP:%[0-9]+]]:gpr(s32) = G_ICMP intpred(eq), [[COPY]](s32), [[C1]]

  ; CHECK:   [[C:%[0-9]+]]:gpr(s64) = G_CONSTANT i64 0

  ; CHECK:   [[INTTOPTR:%[0-9]+]]:gpr(p0) = G_INTTOPTR [[C]](s64)

  ; CHECK:   [[C1:%[0-9]+]]:gpr(s64) = G_CONSTANT i64 128

  ; CHECK:   [[INTTOPTR1:%[0-9]+]]:gpr(p0) = G_INTTOPTR [[C1]](s64)

  ; CHECK:   [[C2:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 0

  ; CHECK:   [[ICMP:%[0-9]+]]:gpr(s32) = G_ICMP intpred(eq), [[COPY]](s32), [[C2]]

  ; CHECK:   [[TRUNC:%[0-9]+]]:gpr(s1) = G_TRUNC [[ICMP]](s32)

  ; CHECK:   G_BRCOND [[TRUNC]](s1), %bb.1

  ; CHECK:   G_BR %bb.2
@@ -450,3 +451,28 @@ body: | 
    RET_ReallyLR implicit $x0



...



---

name:            many_local_use_intra_block

legalized:       true

regBankSelected: true

body:             |

  bb.0:

    ; CHECK-LABEL: name: many_local_use_intra_block

    ; CHECK: [[C:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 1

    ; CHECK: [[ADD:%[0-9]+]]:gpr(s32) = G_ADD [[C]], [[C]]

    ; CHECK: [[ADD1:%[0-9]+]]:gpr(s32) = G_ADD [[C]], [[C]]

    ; CHECK: [[ADD2:%[0-9]+]]:gpr(s32) = G_ADD [[C]], [[C]]

    ; CHECK: [[ADD3:%[0-9]+]]:gpr(s32) = G_ADD [[C]], [[C]]

    ; CHECK: [[ADD4:%[0-9]+]]:gpr(s32) = G_ADD [[C]], [[C]]

    ; CHECK: [[C1:%[0-9]+]]:gpr(s32) = G_CONSTANT i32 2

    ; CHECK: [[ADD5:%[0-9]+]]:gpr(s32) = G_ADD [[C1]], [[C1]]

    %0:gpr(s32) = G_CONSTANT i32 1

    %1:gpr(s32) = G_CONSTANT i32 2

    %2:gpr(s32) = G_ADD %0, %0

    %3:gpr(s32) = G_ADD %0, %0

    %4:gpr(s32) = G_ADD %0, %0

    %5:gpr(s32) = G_ADD %0, %0

    %6:gpr(s32) = G_ADD %0, %0

    %7:gpr(s32) = G_ADD %1, %1

...

llvm/test/CodeGen/AArch64/GlobalISel/swifterror.ll

+3 −3
Original line number Diff line number Diff line
@@ -8,9 +8,9 @@ declare void @free(i8*) 
; that takes a swifterror parameter and "caller" is the caller of "foo".

define float @foo(%swift_error** swifterror %error_ptr_ref) {

; CHECK-LABEL: foo:

; CHECK: mov [[ID:w[0-9]+]], #1

; CHECK: mov w0, #16

; CHECK: malloc

; CHECK: mov [[ID:w[0-9]+]], #1

; CHECK: strb [[ID]], [x0, #8]

; CHECK: mov x21, x0

; CHECK-NOT: x21
@@ -164,9 +164,9 @@ bb_end: 
define void @foo_sret(%struct.S* sret %agg.result, i32 %val1, %swift_error** swifterror %error_ptr_ref) {

; CHECK-LABEL: foo_sret:

; CHECK: mov [[SRET:x[0-9]+]], x8

; CHECK: mov [[ID:w[0-9]+]], #1

; CHECK: mov w0, #16

; CHECK: malloc

; CHECK: mov [[ID:w[0-9]+]], #1

; CHECK: strb [[ID]], [x0, #8]

; CHECK: str w{{.*}}, [{{.*}}[[SRET]], #4]

; CHECK: mov x21, x0
@@ -220,9 +220,9 @@ handler: 
declare void @llvm.va_start(i8*) nounwind

define float @foo_vararg(%swift_error** swifterror %error_ptr_ref, ...) {

; CHECK-LABEL: foo_vararg:

; CHECK-DAG: mov [[ID:w[0-9]+]], #1

; CHECK: mov w0, #16

; CHECK: malloc

; CHECK-DAG: mov [[ID:w[0-9]+]], #1

; CHECK-DAG: strb [[ID]], [x0, #8]



; First vararg

llvm/test/CodeGen/AArch64/GlobalISel/translate-constant-dag.ll

+2 −2
Original line number Diff line number Diff line
@@ -9,12 +9,12 @@ define void @test_const(%dag* %dst) { 
  ; CHECK:   liveins: $x0

  ; CHECK:   [[COPY:%[0-9]+]]:gpr64sp = COPY $x0

  ; CHECK:   [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 10

  ; CHECK:   [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 20

  ; CHECK:   [[MOVi32imm2:%[0-9]+]]:gpr32 = MOVi32imm 50

  ; CHECK:   STRBBui [[MOVi32imm]], [[COPY]], 0 :: (store 1 into %ir.dst)

  ; CHECK:   [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 20

  ; CHECK:   STRBBui [[MOVi32imm1]], [[COPY]], 1 :: (store 1 into %ir.dst + 1)

  ; CHECK:   STRBBui [[MOVi32imm]], [[COPY]], 2 :: (store 1 into %ir.dst + 2)

  ; CHECK:   STRBBui [[MOVi32imm1]], [[COPY]], 3 :: (store 1 into %ir.dst + 3)

  ; CHECK:   [[MOVi32imm2:%[0-9]+]]:gpr32 = MOVi32imm 50

  ; CHECK:   STRBBui [[MOVi32imm2]], [[COPY]], 4 :: (store 1 into %ir.dst + 4)

  ; CHECK:   STRBBui [[MOVi32imm]], [[COPY]], 5 :: (store 1 into %ir.dst + 5)

  ; CHECK:   STRBBui [[MOVi32imm1]], [[COPY]], 6 :: (store 1 into %ir.dst + 6)