llvm/test/Transforms/LoopVectorize/X86/x86-predication.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt < %s -mattr=avx -force-vector-width=2 -force-vector-interleave=1 -passes=loop-vectorize -S | FileCheck %s
; RUN: opt -mcpu=skylake-avx512 -S -force-vector-width=8 -force-vector-interleave=1 -passes=loop-vectorize < %s | FileCheck %s --check-prefix=SINK-GATHER

target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.8.0"

; This test ensures that we don't scalarize the predicated load. Since the load
; can be vectorized with predication, scalarizing it would cause its pointer
; operand to become non-uniform.
;
define i32 @predicated_sdiv_masked_load(ptr %a, ptr %b, i32 %x, i1 %c) {
; CHECK-LABEL: define i32 @predicated_sdiv_masked_load(
; CHECK-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i32 [[X:%.*]], i1 [[C:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> poison, <2 x i32> zeroinitializer
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_SDIV_CONTINUE2:.*]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP11:%.*]], %[[PRED_SDIV_CONTINUE2]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <2 x i32> @llvm.masked.load.v2i32.p0(ptr align 4 [[TMP1]], <2 x i1> [[BROADCAST_SPLAT]], <2 x i32> poison)
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_SDIV_IF:.*]], label %[[PRED_SDIV_CONTINUE:.*]]
; CHECK:       [[PRED_SDIV_IF]]:
; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i32> [[WIDE_MASKED_LOAD]], i64 0
; CHECK-NEXT:    [[TMP3:%.*]] = sdiv i32 [[TMP2]], [[X]]
; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x i32> poison, i32 [[TMP3]], i64 0
; CHECK-NEXT:    br label %[[PRED_SDIV_CONTINUE]]
; CHECK:       [[PRED_SDIV_CONTINUE]]:
; CHECK-NEXT:    [[TMP5:%.*]] = phi <2 x i32> [ poison, %[[VECTOR_BODY]] ], [ [[TMP4]], %[[PRED_SDIV_IF]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_SDIV_IF1:.*]], label %[[PRED_SDIV_CONTINUE2]]
; CHECK:       [[PRED_SDIV_IF1]]:
; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x i32> [[WIDE_MASKED_LOAD]], i64 1
; CHECK-NEXT:    [[TMP7:%.*]] = sdiv i32 [[TMP6]], [[X]]
; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x i32> [[TMP5]], i32 [[TMP7]], i64 1
; CHECK-NEXT:    br label %[[PRED_SDIV_CONTINUE2]]
; CHECK:       [[PRED_SDIV_CONTINUE2]]:
; CHECK-NEXT:    [[TMP9:%.*]] = phi <2 x i32> [ [[TMP5]], %[[PRED_SDIV_CONTINUE]] ], [ [[TMP8]], %[[PRED_SDIV_IF1]] ]
; CHECK-NEXT:    [[TMP10:%.*]] = add nsw <2 x i32> [[TMP9]], [[WIDE_LOAD]]
; CHECK-NEXT:    [[PREDPHI:%.*]] = select i1 [[C]], <2 x i32> [[TMP10]], <2 x i32> [[WIDE_LOAD]]
; CHECK-NEXT:    [[TMP11]] = add <2 x i32> [[VEC_PHI]], [[PREDPHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP11]])
; CHECK-NEXT:    br label %[[FOR_END:.*]]
; CHECK:       [[FOR_END]]:
; CHECK-NEXT:    ret i32 [[TMP13]]
;
; SINK-GATHER-LABEL: define i32 @predicated_sdiv_masked_load(
; SINK-GATHER-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i32 [[X:%.*]], i1 [[C:%.*]]) #[[ATTR0:[0-9]+]] {
; SINK-GATHER-NEXT:  [[ENTRY:.*:]]
; SINK-GATHER-NEXT:    br label %[[VECTOR_PH:.*]]
; SINK-GATHER:       [[VECTOR_PH]]:
; SINK-GATHER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i1> poison, i1 [[C]], i64 0
; SINK-GATHER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i1> [[BROADCAST_SPLATINSERT]], <8 x i1> poison, <8 x i32> zeroinitializer
; SINK-GATHER-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i32> poison, i32 [[X]], i64 0
; SINK-GATHER-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i32> [[BROADCAST_SPLATINSERT1]], <8 x i32> poison, <8 x i32> zeroinitializer
; SINK-GATHER-NEXT:    br label %[[VECTOR_BODY:.*]]
; SINK-GATHER:       [[VECTOR_BODY]]:
; SINK-GATHER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; SINK-GATHER-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP35:%.*]], %[[VECTOR_BODY]] ]
; SINK-GATHER-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; SINK-GATHER-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP0]], align 4
; SINK-GATHER-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX]]
; SINK-GATHER-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr align 4 [[TMP1]], <8 x i1> [[BROADCAST_SPLAT]], <8 x i32> poison)
; SINK-GATHER-NEXT:    [[TMP33:%.*]] = call <8 x i32> @llvm.masked.sdiv.v8i32(<8 x i32> [[WIDE_MASKED_LOAD]], <8 x i32> [[BROADCAST_SPLAT2]], <8 x i1> [[BROADCAST_SPLAT]])
; SINK-GATHER-NEXT:    [[TMP34:%.*]] = add nsw <8 x i32> [[TMP33]], [[WIDE_LOAD]]
; SINK-GATHER-NEXT:    [[PREDPHI:%.*]] = select i1 [[C]], <8 x i32> [[TMP34]], <8 x i32> [[WIDE_LOAD]]
; SINK-GATHER-NEXT:    [[TMP35]] = add <8 x i32> [[VEC_PHI]], [[PREDPHI]]
; SINK-GATHER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; SINK-GATHER-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
; SINK-GATHER-NEXT:    br i1 [[TMP36]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; SINK-GATHER:       [[MIDDLE_BLOCK]]:
; SINK-GATHER-NEXT:    [[TMP37:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP35]])
; SINK-GATHER-NEXT:    br label %[[FOR_END:.*]]
; SINK-GATHER:       [[FOR_END]]:
; SINK-GATHER-NEXT:    ret i32 [[TMP37]]
;
entry:
  br label %for.body

for.body:
  %i = phi i64 [ 0, %entry ], [ %i.next, %for.inc ]
  %r = phi i32 [ 0, %entry ], [ %t7, %for.inc ]
  %t0 = getelementptr inbounds i32, ptr %a, i64 %i
  %t1 = load i32, ptr %t0, align 4
  br i1 %c, label %if.then, label %for.inc

if.then:
  %t2 = getelementptr inbounds i32, ptr %b, i64 %i
  %t3 = load i32, ptr %t2, align 4
  %t4 = sdiv i32 %t3, %x
  %t5 = add nsw i32 %t4, %t1
  br label %for.inc

for.inc:
  %t6 = phi i32 [ %t1, %for.body ], [ %t5, %if.then]
  %t7 = add i32 %r, %t6
  %i.next = add nuw nsw i64 %i, 1
  %cond = icmp eq i64 %i.next, 10000
  br i1 %cond, label %for.end, label %for.body

for.end:
  %t8 = phi i32 [ %t7, %for.inc ]
  ret i32 %t8
}

; This test ensures that a load, which would have been widened otherwise is
; instead scalarized if Cost-Model so decided as part of its
; sink-scalar-operands optimization for predicated instructions.
define i32 @scalarize_and_sink_gather(ptr %a, i1 %c, i32 %x, i64 %n) {
; CHECK-LABEL: define i32 @scalarize_and_sink_gather(
; CHECK-SAME: ptr [[A:%.*]], i1 [[C:%.*]], i32 [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 1)
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[SMAX]], 1
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[X]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_UDIV_CONTINUE2:.*]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP14:%.*]], %[[PRED_UDIV_CONTINUE2]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_UDIV_IF:.*]], label %[[PRED_UDIV_CONTINUE:.*]]
; CHECK:       [[PRED_UDIV_IF]]:
; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[INDEX]], 777
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP1]]
; CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
; CHECK-NEXT:    [[TMP4:%.*]] = udiv i32 [[TMP3]], [[X]]
; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i64 0
; CHECK-NEXT:    br label %[[PRED_UDIV_CONTINUE]]
; CHECK:       [[PRED_UDIV_CONTINUE]]:
; CHECK-NEXT:    [[TMP6:%.*]] = phi <2 x i32> [ poison, %[[VECTOR_BODY]] ], [ [[TMP5]], %[[PRED_UDIV_IF]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_UDIV_IF1:.*]], label %[[PRED_UDIV_CONTINUE2]]
; CHECK:       [[PRED_UDIV_IF1]]:
; CHECK-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 1
; CHECK-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 777
; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP8]]
; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
; CHECK-NEXT:    [[TMP11:%.*]] = udiv i32 [[TMP10]], [[X]]
; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP11]], i64 1
; CHECK-NEXT:    br label %[[PRED_UDIV_CONTINUE2]]
; CHECK:       [[PRED_UDIV_CONTINUE2]]:
; CHECK-NEXT:    [[TMP13:%.*]] = phi <2 x i32> [ [[TMP6]], %[[PRED_UDIV_CONTINUE]] ], [ [[TMP12]], %[[PRED_UDIV_IF1]] ]
; CHECK-NEXT:    [[PREDPHI:%.*]] = select i1 [[C]], <2 x i32> [[TMP13]], <2 x i32> [[BROADCAST_SPLAT]]
; CHECK-NEXT:    [[TMP14]] = add <2 x i32> [[VEC_PHI]], [[PREDPHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP15]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP14]])
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP16]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
; CHECK:       [[FOR_BODY]]:
; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[I_NEXT:%.*]], %[[FOR_INC:.*]] ]
; CHECK-NEXT:    [[R:%.*]] = phi i32 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[T6:%.*]], %[[FOR_INC]] ]
; CHECK-NEXT:    [[I7:%.*]] = mul i64 [[I]], 777
; CHECK-NEXT:    br i1 [[C]], label %[[IF_THEN:.*]], label %[[FOR_INC]]
; CHECK:       [[IF_THEN]]:
; CHECK-NEXT:    [[T0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I7]]
; CHECK-NEXT:    [[T2:%.*]] = load i32, ptr [[T0]], align 4
; CHECK-NEXT:    [[T4:%.*]] = udiv i32 [[T2]], [[X]]
; CHECK-NEXT:    br label %[[FOR_INC]]
; CHECK:       [[FOR_INC]]:
; CHECK-NEXT:    [[T5:%.*]] = phi i32 [ [[X]], %[[FOR_BODY]] ], [ [[T4]], %[[IF_THEN]] ]
; CHECK-NEXT:    [[T6]] = add i32 [[R]], [[T5]]
; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[COND]], label %[[FOR_BODY]], label %[[FOR_END]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       [[FOR_END]]:
; CHECK-NEXT:    [[T7:%.*]] = phi i32 [ [[T6]], %[[FOR_INC]] ], [ [[TMP16]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret i32 [[T7]]
;
; SINK-GATHER-LABEL: define i32 @scalarize_and_sink_gather(
; SINK-GATHER-SAME: ptr [[A:%.*]], i1 [[C:%.*]], i32 [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
; SINK-GATHER-NEXT:  [[ENTRY:.*]]:
; SINK-GATHER-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 1)
; SINK-GATHER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 8
; SINK-GATHER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; SINK-GATHER:       [[VECTOR_PH]]:
; SINK-GATHER-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[SMAX]], 7
; SINK-GATHER-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; SINK-GATHER-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i1> poison, i1 [[C]], i64 0
; SINK-GATHER-NEXT:    [[BROADCAST_SPLAT1:%.*]] = shufflevector <8 x i1> [[BROADCAST_SPLATINSERT1]], <8 x i1> poison, <8 x i32> zeroinitializer
; SINK-GATHER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[X]], i64 0
; SINK-GATHER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i32> [[BROADCAST_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
; SINK-GATHER-NEXT:    br label %[[VECTOR_BODY:.*]]
; SINK-GATHER:       [[VECTOR_BODY]]:
; SINK-GATHER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; SINK-GATHER-NEXT:    [[VEC_IND:%.*]] = phi <8 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; SINK-GATHER-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP49:%.*]], %[[VECTOR_BODY]] ]
; SINK-GATHER-NEXT:    [[TMP0:%.*]] = mul <8 x i64> [[VEC_IND]], splat (i64 777)
; SINK-GATHER-NEXT:    [[WIDE_GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], <8 x i64> [[TMP0]]
; SINK-GATHER-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <8 x i32> @llvm.masked.gather.v8i32.v8p0(<8 x ptr> align 4 [[WIDE_GEP]], <8 x i1> [[BROADCAST_SPLAT1]], <8 x i32> poison)
; SINK-GATHER-NEXT:    [[TMP48:%.*]] = call <8 x i32> @llvm.masked.udiv.v8i32(<8 x i32> [[WIDE_MASKED_GATHER]], <8 x i32> [[BROADCAST_SPLAT]], <8 x i1> [[BROADCAST_SPLAT1]])
; SINK-GATHER-NEXT:    [[PREDPHI:%.*]] = select i1 [[C]], <8 x i32> [[TMP48]], <8 x i32> [[BROADCAST_SPLAT]]
; SINK-GATHER-NEXT:    [[TMP49]] = add <8 x i32> [[VEC_PHI]], [[PREDPHI]]
; SINK-GATHER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; SINK-GATHER-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <8 x i64> [[VEC_IND]], splat (i64 8)
; SINK-GATHER-NEXT:    [[TMP50:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-GATHER-NEXT:    br i1 [[TMP50]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; SINK-GATHER:       [[MIDDLE_BLOCK]]:
; SINK-GATHER-NEXT:    [[TMP51:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP49]])
; SINK-GATHER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; SINK-GATHER-NEXT:    br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; SINK-GATHER:       [[SCALAR_PH]]:
; SINK-GATHER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; SINK-GATHER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP51]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; SINK-GATHER-NEXT:    br label %[[FOR_BODY:.*]]
; SINK-GATHER:       [[FOR_BODY]]:
; SINK-GATHER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[I_NEXT:%.*]], %[[FOR_INC:.*]] ]
; SINK-GATHER-NEXT:    [[R:%.*]] = phi i32 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[T6:%.*]], %[[FOR_INC]] ]
; SINK-GATHER-NEXT:    [[I7:%.*]] = mul i64 [[I]], 777
; SINK-GATHER-NEXT:    br i1 [[C]], label %[[IF_THEN:.*]], label %[[FOR_INC]]
; SINK-GATHER:       [[IF_THEN]]:
; SINK-GATHER-NEXT:    [[T0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I7]]
; SINK-GATHER-NEXT:    [[T2:%.*]] = load i32, ptr [[T0]], align 4
; SINK-GATHER-NEXT:    [[T4:%.*]] = udiv i32 [[T2]], [[X]]
; SINK-GATHER-NEXT:    br label %[[FOR_INC]]
; SINK-GATHER:       [[FOR_INC]]:
; SINK-GATHER-NEXT:    [[T5:%.*]] = phi i32 [ [[X]], %[[FOR_BODY]] ], [ [[T4]], %[[IF_THEN]] ]
; SINK-GATHER-NEXT:    [[T6]] = add i32 [[R]], [[T5]]
; SINK-GATHER-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; SINK-GATHER-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; SINK-GATHER-NEXT:    br i1 [[COND]], label %[[FOR_BODY]], label %[[FOR_END]], !llvm.loop [[LOOP4:![0-9]+]]
; SINK-GATHER:       [[FOR_END]]:
; SINK-GATHER-NEXT:    [[T7:%.*]] = phi i32 [ [[T6]], %[[FOR_INC]] ], [ [[TMP51]], %[[MIDDLE_BLOCK]] ]
; SINK-GATHER-NEXT:    ret i32 [[T7]]
;
entry:
  br label %for.body

for.body:
  %i = phi i64 [ 0, %entry ], [ %i.next, %for.inc ]
  %r = phi i32 [ 0, %entry ], [ %t6, %for.inc ]
  %i7 = mul i64 %i, 777
  br i1 %c, label %if.then, label %for.inc

if.then:
  %t0 = getelementptr inbounds i32, ptr %a, i64 %i7
  %t2 = load i32, ptr %t0, align 4
  %t4 = udiv i32 %t2, %x
  br label %for.inc

for.inc:
  %t5 = phi i32 [ %x, %for.body ], [ %t4, %if.then]
  %t6 = add i32 %r, %t5
  %i.next = add nuw nsw i64 %i, 1
  %cond = icmp slt i64 %i.next, %n
  br i1 %cond, label %for.body, label %for.end

for.end:
  %t7 = phi i32 [ %t6, %for.inc ]
  ret i32 %t7
}