[llvm] [VPlan] Collect FOR PHIs before sinking/hoisting recurrence users (PR #194671)

[Florian Hahn via llvm-commits]

================
@@ -0,0 +1,99 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 6
+; RUN: opt < %s -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -vplan-print-after=printOptimizedVPlan -disable-output 2>&1 | FileCheck %s
+
+define i32 @test(i32 %limit, i1 %guard, i32 %divisor) {
+; CHECK-LABEL: VPlan for loop in 'test'
+; CHECK:  VPlan 'Initial VPlan for VF={4},UF>=1' {
+; CHECK-NEXT:  Live-in vp<[[VP0:%[0-9]+]]> = VF
+; CHECK-NEXT:  Live-in vp<[[VP1:%[0-9]+]]> = VF * UF
+; CHECK-NEXT:  Live-in vp<[[VP2:%[0-9]+]]> = vector-trip-count
+; CHECK-NEXT:  vp<[[VP3:%[0-9]+]]> = original trip-count
+; CHECK-EMPTY:
+; CHECK-NEXT:  ir-bb<entry>:
+; CHECK-NEXT:    EMIT vp<[[VP3]]> = EXPAND SCEV (1 + (0 smax %limit))<nuw>
+; CHECK-NEXT:  Successor(s): scalar.ph, vector.ph
+; CHECK-EMPTY:
+; CHECK-NEXT:  vector.ph:
+; CHECK-NEXT:    EMIT vp<[[VP4:%[0-9]+]]> = reduction-start-vector ir<0>, ir<0>, ir<1>
+; CHECK-NEXT:  Successor(s): vector loop
+; CHECK-EMPTY:
+; CHECK-NEXT:  <x1> vector loop: {
+; CHECK-NEXT:  vp<[[VP5:%[0-9]+]]> = CANONICAL-IV
+; CHECK-EMPTY:
+; CHECK-NEXT:    vector.body:
+; CHECK-NEXT:      ir<%iv> = WIDEN-INDUCTION nuw nsw ir<0>, ir<1>, vp<[[VP0]]>
+; CHECK-NEXT:      WIDEN-REDUCTION-PHI ir<%sum> = phi vp<[[VP4]]>, ir<%sum.next>
+; CHECK-NEXT:      WIDEN ir<%iv.nonzero> = icmp ne ir<%iv>, ir<0>
+; CHECK-NEXT:      WIDEN ir<%or.cond> = select ir<%iv.nonzero>, ir<%guard>, ir<false>
+; CHECK-NEXT:      EMIT vp<[[VP6:%[0-9]+]]> = select ir<%or.cond>, ir<1>, ir<%divisor>
+; CHECK-NEXT:      WIDEN ir<%rem> = srem ir<%iv>, vp<[[VP6]]>
+; CHECK-NEXT:      BLEND ir<%cond> = ir<%rem> ir<0>/ir<%or.cond>
+; CHECK-NEXT:      WIDEN ir<%sum.next> = add ir<%sum>, ir<%cond>
+; CHECK-NEXT:      EMIT vp<%index.next> = add nuw vp<[[VP5]]>, vp<[[VP1]]>
+; CHECK-NEXT:      EMIT branch-on-count vp<%index.next>, vp<[[VP2]]>
+; CHECK-NEXT:    No successors
+; CHECK-NEXT:  }
+; CHECK-NEXT:  Successor(s): middle.block
+; CHECK-EMPTY:
+; CHECK-NEXT:  middle.block:
+; CHECK-NEXT:    WIDEN ir<%prev.iv> = add nsw ir<%iv>, ir<-1>
+; CHECK-NEXT:    EMIT vp<[[VP8:%[0-9]+]]> = compute-reduction-result (add) ir<%sum.next>
+; CHECK-NEXT:    EMIT vp<[[VP9:%[0-9]+]]> = extract-last-part ir<%prev.iv>
+; CHECK-NEXT:    EMIT vp<%vector.recur.extract> = extract-last-lane vp<[[VP9]]>
+; CHECK-NEXT:    EMIT vp<%cmp.n> = icmp eq vp<[[VP3]]>, vp<[[VP2]]>
+; CHECK-NEXT:    EMIT branch-on-cond vp<%cmp.n>
+; CHECK-NEXT:  Successor(s): ir-bb<exit>, scalar.ph
+; CHECK-EMPTY:
+; CHECK-NEXT:  ir-bb<exit>:
+; CHECK-NEXT:    IR   %sum.next.lcssa = phi i32 [ %sum.next, %loop.latch ] (extra operand: vp<[[VP8]]> from middle.block)
+; CHECK-NEXT:  No successors
+; CHECK-EMPTY:
+; CHECK-NEXT:  scalar.ph:
+; CHECK-NEXT:    EMIT-SCALAR vp<%bc.resume.val> = phi [ vp<[[VP2]]>, middle.block ], [ ir<0>, ir-bb<entry> ]
+; CHECK-NEXT:    EMIT-SCALAR vp<%scalar.recur.init> = phi [ vp<%vector.recur.extract>, middle.block ], [ ir<0>, ir-bb<entry> ]
+; CHECK-NEXT:    EMIT-SCALAR vp<%bc.merge.rdx> = phi [ vp<[[VP8]]>, middle.block ], [ ir<0>, ir-bb<entry> ]
+; CHECK-NEXT:  Successor(s): ir-bb<loop.header>
+; CHECK-EMPTY:
+; CHECK-NEXT:  ir-bb<loop.header>:
----------------
fhahn wrote:

can pass `-filter-out-after=\ir-bb<loop.header>:'` or something like that to filter out this bit 

https://github.com/llvm/llvm-project/pull/194671