llvm/test/Transforms/LoopVectorize/SystemZ/branch-for-predicated-block.ll

; REQUIRES: asserts
; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z13 -passes=loop-vectorize \
; RUN:   -force-vector-width=2 -debug-only=loop-vectorize \
; RUN:   -disable-output < %s 2>&1 | FileCheck %s

; Check costs for branches inside a vectorized loop around predicated
; blocks. Each such branch will be guarded with an extractelement from the
; vector compare plus a test under mask instruction. This cost is modelled on
; the extractelement of i1.

define void @fun(ptr %arr, i64 %trip.count) {
entry:
  br label %for.body

for.body:
  %indvars.iv = phi i64 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
  %arrayidx = getelementptr inbounds i32, ptr %arr, i64 %indvars.iv
  %l = load i32, ptr %arrayidx, align 4
  %cmp55 = icmp sgt i32 %l, 0
  br i1 %cmp55, label %if.then, label %for.inc

if.then:
  %sub = sub nsw i32 0, %l
  store i32 %sub, ptr %arrayidx, align 4
  br label %for.inc

for.inc:
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %exitcond = icmp eq i64 %indvars.iv.next, %trip.count
  br i1 %exitcond, label %for.end.loopexit, label %for.body

for.end.loopexit:
  ret void

; CHECK: Cost of 1 for VF 2: profitable to scalarize   store i32 %sub, ptr %arrayidx, align 4
; CHECK: Cost of 2.5 for VF 2: profitable to scalarize   %sub = sub nsw i32 0, %l
; CHECK: Cost of 1 for VF 2: WIDEN ir<%cmp55> = icmp sgt ir<%l>, ir<0>
; CHECK: Cost of 0 for VF 2: vector loop backedge
}