llvm/llvm/test/CodeGen/X86/x86-win64-shrink-wrapping.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc %s -o - -enable-shrink-wrap=true | FileCheck %s --check-prefix=ENABLE
; RUN: llc %s -o - -enable-shrink-wrap=false | FileCheck %s --check-prefix=DISABLE

target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
target triple = "x86_64--windows-gnu"

; The output of this function with or without shrink-wrapping
; shouldn't change.
; Indeed, the epilogue block would have been if.else, meaning
; after the pops, we will have additional instruction (jump, mov,
; etc.) prior to the return and this is forbidden for Win64.
define i32 @loopInfoSaveOutsideLoop(i32 %cond, i32 %N) #0 {
; ENABLE-LABEL: loopInfoSaveOutsideLoop:
; ENABLE:       # %bb.0: # %entry
; ENABLE-NEXT:    pushq %rbx
; ENABLE-NEXT:    .seh_pushreg %rbx
; ENABLE-NEXT:    .seh_endprologue
; ENABLE-NEXT:    testl %ecx, %ecx
; ENABLE-NEXT:    je .LBB0_4
; ENABLE-NEXT:  # %bb.1: # %for.preheader
; ENABLE-NEXT:    #APP
; ENABLE-NEXT:    nop
; ENABLE-NEXT:    #NO_APP
; ENABLE-NEXT:    xorl %eax, %eax
; ENABLE-NEXT:    movl $10, %ecx
; ENABLE-NEXT:    #APP
; ENABLE-NEXT:    movl $1, %edx
; ENABLE-NEXT:    #NO_APP
; ENABLE-NEXT:    .p2align 4
; ENABLE-NEXT:  .LBB0_2: # %for.body
; ENABLE-NEXT:    # =>This Inner Loop Header: Depth=1
; ENABLE-NEXT:    addl %edx, %eax
; ENABLE-NEXT:    decl %ecx
; ENABLE-NEXT:    jne .LBB0_2
; ENABLE-NEXT:  # %bb.3: # %for.end
; ENABLE-NEXT:    #APP
; ENABLE-NEXT:    nop
; ENABLE-NEXT:    #NO_APP
; ENABLE-NEXT:    shll $3, %eax
; ENABLE-NEXT:    jmp .LBB0_5
; ENABLE-NEXT:  .LBB0_4: # %if.else
; ENABLE-NEXT:    movl %edx, %eax
; ENABLE-NEXT:    addl %edx, %eax
; ENABLE-NEXT:  .LBB0_5: # %if.end
; ENABLE-NEXT:    popq %rbx
; ENABLE-NEXT:    retq
; ENABLE-NEXT:    .seh_endproc
;
; DISABLE-LABEL: loopInfoSaveOutsideLoop:
; DISABLE:       # %bb.0: # %entry
; DISABLE-NEXT:    pushq %rbx
; DISABLE-NEXT:    .seh_pushreg %rbx
; DISABLE-NEXT:    .seh_endprologue
; DISABLE-NEXT:    testl %ecx, %ecx
; DISABLE-NEXT:    je .LBB0_4
; DISABLE-NEXT:  # %bb.1: # %for.preheader
; DISABLE-NEXT:    #APP
; DISABLE-NEXT:    nop
; DISABLE-NEXT:    #NO_APP
; DISABLE-NEXT:    xorl %eax, %eax
; DISABLE-NEXT:    movl $10, %ecx
; DISABLE-NEXT:    #APP
; DISABLE-NEXT:    movl $1, %edx
; DISABLE-NEXT:    #NO_APP
; DISABLE-NEXT:    .p2align 4
; DISABLE-NEXT:  .LBB0_2: # %for.body
; DISABLE-NEXT:    # =>This Inner Loop Header: Depth=1
; DISABLE-NEXT:    addl %edx, %eax
; DISABLE-NEXT:    decl %ecx
; DISABLE-NEXT:    jne .LBB0_2
; DISABLE-NEXT:  # %bb.3: # %for.end
; DISABLE-NEXT:    #APP
; DISABLE-NEXT:    nop
; DISABLE-NEXT:    #NO_APP
; DISABLE-NEXT:    shll $3, %eax
; DISABLE-NEXT:    jmp .LBB0_5
; DISABLE-NEXT:  .LBB0_4: # %if.else
; DISABLE-NEXT:    movl %edx, %eax
; DISABLE-NEXT:    addl %edx, %eax
; DISABLE-NEXT:  .LBB0_5: # %if.end
; DISABLE-NEXT:    popq %rbx
; DISABLE-NEXT:    retq
; DISABLE-NEXT:    .seh_endproc
entry:
  %tobool = icmp eq i32 %cond, 0
  br i1 %tobool, label %if.else, label %for.preheader

for.preheader:                                    ; preds = %entry
  tail call void asm "nop", ""()
  br label %for.body

for.body:                                         ; preds = %for.body, %for.preheader
  %i.05 = phi i32 [ %inc, %for.body ], [ 0, %for.preheader ]
  %sum.04 = phi i32 [ %add, %for.body ], [ 0, %for.preheader ]
  %call = tail call i32 asm "movl $$1, $0", "=r,~{ebx}"()
  %add = add nsw i32 %call, %sum.04
  %inc = add nuw nsw i32 %i.05, 1
  %exitcond = icmp eq i32 %inc, 10
  br i1 %exitcond, label %for.end, label %for.body

for.end:                                          ; preds = %for.body
  tail call void asm "nop", "~{ebx}"()
  %shl = shl i32 %add, 3
  br label %if.end

if.else:                                          ; preds = %entry
  %mul = shl nsw i32 %N, 1
  br label %if.end

if.end:                                           ; preds = %if.else, %for.end
  %sum.1 = phi i32 [ %shl, %for.end ], [ %mul, %if.else ]
  ret i32 %sum.1
}

; When we can sink the epilogue of the function into an existing exit block,
; this is Ok for shrink-wrapping to kicks in.
define i32 @loopInfoSaveOutsideLoop2(i32 %cond, i32 %N) #0 {
; ENABLE-LABEL: loopInfoSaveOutsideLoop2:
; ENABLE:       # %bb.0: # %entry
; ENABLE-NEXT:    testl %ecx, %ecx
; ENABLE-NEXT:    je .LBB1_4
; ENABLE-NEXT:  # %bb.1: # %for.preheader
; ENABLE-NEXT:    pushq %rbx
; ENABLE-NEXT:    .seh_pushreg %rbx
; ENABLE-NEXT:    .seh_endprologue
; ENABLE-NEXT:    #APP
; ENABLE-NEXT:    nop
; ENABLE-NEXT:    #NO_APP
; ENABLE-NEXT:    xorl %eax, %eax
; ENABLE-NEXT:    movl $10, %ecx
; ENABLE-NEXT:    .p2align 4
; ENABLE-NEXT:  .LBB1_2: # %for.body
; ENABLE-NEXT:    # =>This Inner Loop Header: Depth=1
; ENABLE-NEXT:    #APP
; ENABLE-NEXT:    movl $1, %edx
; ENABLE-NEXT:    #NO_APP
; ENABLE-NEXT:    addl %edx, %eax
; ENABLE-NEXT:    decl %ecx
; ENABLE-NEXT:    jne .LBB1_2
; ENABLE-NEXT:  # %bb.3: # %for.end
; ENABLE-NEXT:    #APP
; ENABLE-NEXT:    nop
; ENABLE-NEXT:    #NO_APP
; ENABLE-NEXT:    shll $3, %eax
; ENABLE-NEXT:    popq %rbx
; ENABLE-NEXT:    retq
; ENABLE-NEXT:  .LBB1_4: # %if.else
; ENABLE-NEXT:    addl %edx, %edx
; ENABLE-NEXT:    movl %edx, %eax
; ENABLE-NEXT:    retq
; ENABLE-NEXT:    .seh_endproc
;
; DISABLE-LABEL: loopInfoSaveOutsideLoop2:
; DISABLE:       # %bb.0: # %entry
; DISABLE-NEXT:    pushq %rbx
; DISABLE-NEXT:    .seh_pushreg %rbx
; DISABLE-NEXT:    .seh_endprologue
; DISABLE-NEXT:    testl %ecx, %ecx
; DISABLE-NEXT:    je .LBB1_4
; DISABLE-NEXT:  # %bb.1: # %for.preheader
; DISABLE-NEXT:    #APP
; DISABLE-NEXT:    nop
; DISABLE-NEXT:    #NO_APP
; DISABLE-NEXT:    xorl %eax, %eax
; DISABLE-NEXT:    movl $10, %ecx
; DISABLE-NEXT:    .p2align 4
; DISABLE-NEXT:  .LBB1_2: # %for.body
; DISABLE-NEXT:    # =>This Inner Loop Header: Depth=1
; DISABLE-NEXT:    #APP
; DISABLE-NEXT:    movl $1, %edx
; DISABLE-NEXT:    #NO_APP
; DISABLE-NEXT:    addl %edx, %eax
; DISABLE-NEXT:    decl %ecx
; DISABLE-NEXT:    jne .LBB1_2
; DISABLE-NEXT:  # %bb.3: # %for.end
; DISABLE-NEXT:    #APP
; DISABLE-NEXT:    nop
; DISABLE-NEXT:    #NO_APP
; DISABLE-NEXT:    shll $3, %eax
; DISABLE-NEXT:    jmp .LBB1_5
; DISABLE-NEXT:  .LBB1_4: # %if.else
; DISABLE-NEXT:    addl %edx, %edx
; DISABLE-NEXT:    movl %edx, %eax
; DISABLE-NEXT:  .LBB1_5: # %if.end
; DISABLE-NEXT:    popq %rbx
; DISABLE-NEXT:    retq
; DISABLE-NEXT:    .seh_endproc
entry:
  %tobool = icmp eq i32 %cond, 0
  br i1 %tobool, label %if.else, label %for.preheader

for.preheader:                                    ; preds = %entry
  tail call void asm "nop", ""()
  br label %for.body

for.body:                                         ; preds = %for.body, %for.preheader
  %i.05 = phi i32 [ %inc, %for.body ], [ 0, %for.preheader ]
  %sum.04 = phi i32 [ %add, %for.body ], [ 0, %for.preheader ]
  %call = tail call i32 asm sideeffect "movl $$1, $0", "=r,~{ebx}"()
  %add = add nsw i32 %call, %sum.04
  %inc = add nuw nsw i32 %i.05, 1
  %exitcond = icmp eq i32 %inc, 10
  br i1 %exitcond, label %for.end, label %for.body

for.end:                                          ; preds = %for.body
  tail call void asm "nop", "~{ebx}"()
  %shl = shl i32 %add, 3
  ret i32 %shl

if.else:                                          ; preds = %entry
  %mul = shl nsw i32 %N, 1
  br label %if.end

if.end:                                           ; preds = %if.else, %for.end
  ret i32 %mul
}

attributes #0 = { uwtable }