llvm/test/Transforms/LoopVectorize/VPlan/vplan-printing-outer-loop-memchecks.ll

; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --filter-out-after "^vector.ph:" --version 6
; RUN: opt -passes=loop-vectorize -enable-vplan-native-path -force-vector-width=4 -vplan-print-after=printOptimizedVPlan -disable-output %s 2>&1 | FileCheck %s --check-prefix=OPTIMIZED
; RUN: opt -passes=loop-vectorize -enable-vplan-native-path -force-vector-width=4 -vplan-print-after=printFinalVPlan -disable-output %s 2>&1 | FileCheck %s --check-prefix=FINAL

; The store may alias the load, but the memory both access over the whole nest
; can be bounded.
define void @may_alias_load(ptr %A, ptr %B, i64 %N, i64 %M) {
; OPTIMIZED-LABEL: VPlan for loop in 'may_alias_load'
; OPTIMIZED:  VPlan ' for VF={4},UF>=1' {
; OPTIMIZED-NEXT:  Live-in vp<[[VP0:%[0-9]+]]> = VF
; OPTIMIZED-NEXT:  Live-in vp<[[VP1:%[0-9]+]]> = VF * UF
; OPTIMIZED-NEXT:  Live-in vp<[[VP2:%[0-9]+]]> = vector-trip-count
; OPTIMIZED-NEXT:  Live-in ir<%N> = original trip-count
; OPTIMIZED-EMPTY:
; OPTIMIZED-NEXT:  ir-bb<entry>:
; OPTIMIZED-NEXT:  Successor(s): scalar.ph, vector.ph
; OPTIMIZED-EMPTY:
; OPTIMIZED-NEXT:  vector.ph:
;
; FINAL-LABEL: VPlan for loop in 'may_alias_load'
; FINAL:  VPlan 'Final VPlan for VF={4},UF={1}' {
; FINAL-NEXT:  Live-in ir<%N> = original trip-count
; FINAL-EMPTY:
; FINAL-NEXT:  ir-bb<entry>:
; FINAL-NEXT:    EMIT vp<%min.iters.check> = icmp ult ir<%N>, ir<4>
; FINAL-NEXT:    EMIT branch-on-cond vp<%min.iters.check>
; FINAL-NEXT:  Successor(s): ir-bb<scalar.ph>, vector.ph
; FINAL-EMPTY:
; FINAL-NEXT:  vector.ph:
;
entry:
  br label %outer.header

outer.header:
  %outer.iv = phi i64 [ 0, %entry ], [ %outer.iv.next, %outer.latch ]
  %outer.iv.mul.M = mul nsw i64 %outer.iv, %M
  br label %inner.body

inner.body:
  %inner.iv = phi i64 [ 0, %outer.header ], [ %inner.iv.next, %inner.body ]
  %sum = phi float [ 0.000000e+00, %outer.header ], [ %sum.next, %inner.body ]
  %idx = add nsw i64 %outer.iv.mul.M, %inner.iv
  %A.ptr = getelementptr inbounds float, ptr %A, i64 %idx
  %A.val = load float, ptr %A.ptr, align 4
  %sum.next = fadd float %sum, %A.val
  %inner.iv.next = add nuw nsw i64 %inner.iv, 1
  %inner.iv.cmp = icmp eq i64 %inner.iv.next, %M
  br i1 %inner.iv.cmp, label %outer.latch, label %inner.body

outer.latch:
  %B.ptr = getelementptr inbounds float, ptr %B, i64 %outer.iv
  store float %sum.next, ptr %B.ptr, align 4
  %outer.iv.next = add nuw nsw i64 %outer.iv, 1
  %outer.iv.cmp = icmp eq i64 %outer.iv.next, %N
  br i1 %outer.iv.cmp, label %exit, label %outer.header, !llvm.loop !0

exit:
  ret void
}

!0 = distinct !{!0, !1}
!1 = !{!"llvm.loop.vectorize.enable"}