llvm/test/Transforms/LoopVectorize/div-exact.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt -S < %s -p loop-vectorize -force-vector-width=4 | FileCheck %s

define void @unknown_divisor(ptr noalias %p, i32 %n, i1 %c) {
; CHECK-LABEL: define void @unknown_divisor(
; CHECK-SAME: ptr noalias [[P:%.*]], i32 [[N:%.*]], i1 [[C:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
; CHECK:       [[VECTOR_SCEVCHECK]]:
; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
; CHECK-NEXT:    br i1 [[TMP1]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = and i32 [[N]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[N]], [[N_MOD_VF]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE14:.*]] ]
; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 1
; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[INDEX]], 2
; CHECK-NEXT:    [[TMP4:%.*]] = add i32 [[INDEX]], 3
; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr [[P]], i32 [[INDEX]]
; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[P]], i32 [[TMP2]]
; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[P]], i32 [[TMP3]]
; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[P]], i32 [[TMP4]]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF:.*]], label %[[PRED_LOAD_CONTINUE:.*]]
; CHECK:       [[PRED_LOAD_IF]]:
; CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP7]], align 4
; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> poison, i32 [[TMP12]], i64 0
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE]]
; CHECK:       [[PRED_LOAD_CONTINUE]]:
; CHECK-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ poison, %[[VECTOR_BODY]] ], [ [[TMP13]], %[[PRED_LOAD_IF]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF3:.*]], label %[[PRED_LOAD_CONTINUE4:.*]]
; CHECK:       [[PRED_LOAD_IF3]]:
; CHECK-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP8]], align 4
; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP16]], i64 1
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE4]]
; CHECK:       [[PRED_LOAD_CONTINUE4]]:
; CHECK-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP14]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP17]], %[[PRED_LOAD_IF3]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF5:.*]], label %[[PRED_LOAD_CONTINUE6:.*]]
; CHECK:       [[PRED_LOAD_IF5]]:
; CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP9]], align 4
; CHECK-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP20]], i64 2
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE6]]
; CHECK:       [[PRED_LOAD_CONTINUE6]]:
; CHECK-NEXT:    [[TMP22:%.*]] = phi <4 x i32> [ [[TMP18]], %[[PRED_LOAD_CONTINUE4]] ], [ [[TMP21]], %[[PRED_LOAD_IF5]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF7:.*]], label %[[PRED_LOAD_CONTINUE8:.*]]
; CHECK:       [[PRED_LOAD_IF7]]:
; CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP10]], align 4
; CHECK-NEXT:    [[TMP25:%.*]] = insertelement <4 x i32> [[TMP22]], i32 [[TMP24]], i64 3
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE8]]
; CHECK:       [[PRED_LOAD_CONTINUE8]]:
; CHECK-NEXT:    [[TMP23:%.*]] = phi <4 x i32> [ [[TMP22]], %[[PRED_LOAD_CONTINUE6]] ], [ [[TMP25]], %[[PRED_LOAD_IF7]] ]
; CHECK-NEXT:    [[TMP28:%.*]] = call <4 x i32> @llvm.masked.udiv.v4i32(<4 x i32> splat (i32 100), <4 x i32> [[TMP23]], <4 x i1> [[BROADCAST_SPLAT]])
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
; CHECK:       [[PRED_STORE_IF]]:
; CHECK-NEXT:    [[TMP29:%.*]] = extractelement <4 x i32> [[TMP28]], i64 0
; CHECK-NEXT:    store i32 [[TMP29]], ptr [[TMP7]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE]]
; CHECK:       [[PRED_STORE_CONTINUE]]:
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF9:.*]], label %[[PRED_STORE_CONTINUE10:.*]]
; CHECK:       [[PRED_STORE_IF9]]:
; CHECK-NEXT:    [[TMP30:%.*]] = extractelement <4 x i32> [[TMP28]], i64 1
; CHECK-NEXT:    store i32 [[TMP30]], ptr [[TMP8]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE10]]
; CHECK:       [[PRED_STORE_CONTINUE10]]:
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF11:.*]], label %[[PRED_STORE_CONTINUE12:.*]]
; CHECK:       [[PRED_STORE_IF11]]:
; CHECK-NEXT:    [[TMP31:%.*]] = extractelement <4 x i32> [[TMP28]], i64 2
; CHECK-NEXT:    store i32 [[TMP31]], ptr [[TMP9]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE12]]
; CHECK:       [[PRED_STORE_CONTINUE12]]:
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF13:.*]], label %[[PRED_STORE_CONTINUE14]]
; CHECK:       [[PRED_STORE_IF13]]:
; CHECK-NEXT:    [[TMP32:%.*]] = extractelement <4 x i32> [[TMP28]], i64 3
; CHECK-NEXT:    store i32 [[TMP32]], ptr [[TMP10]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE14]]
; CHECK:       [[PRED_STORE_CONTINUE14]]:
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT:    [[TMP33:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP33]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LATCH:.*]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[IF:.*]], label %[[LATCH]]
; CHECK:       [[IF]]:
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[P]], i32 [[IV]]
; CHECK-NEXT:    [[X:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT:    [[Y:%.*]] = udiv exact i32 100, [[X]]
; CHECK-NEXT:    store i32 [[Y]], ptr [[GEP]], align 4
; CHECK-NEXT:    br label %[[LATCH]]
; CHECK:       [[LATCH]]:
; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret void
;
entry:
  br label %loop

loop:
  %iv = phi i32 [0, %entry], [%iv.next, %latch]
  br i1 %c, label %if, label %latch

if:
  %gep = getelementptr i32, ptr %p, i32 %iv
  %x = load i32, ptr %gep
  %y = udiv exact i32 100, %x
  store i32 %y, ptr %gep
  br label %latch

latch:
  %iv.next = add i32 %iv, 1
  %ec = icmp eq i32 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  ret void
}

define void @unknown_dividend(ptr noalias %p, i32 %n, i1 %c) {
; CHECK-LABEL: define void @unknown_dividend(
; CHECK-SAME: ptr noalias [[P:%.*]], i32 [[N:%.*]], i1 [[C:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
; CHECK:       [[VECTOR_SCEVCHECK]]:
; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
; CHECK-NEXT:    br i1 [[TMP1]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = and i32 [[N]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[N]], [[N_MOD_VF]]
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE14:.*]] ]
; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 1
; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[INDEX]], 2
; CHECK-NEXT:    [[TMP4:%.*]] = add i32 [[INDEX]], 3
; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr [[P]], i32 [[INDEX]]
; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[P]], i32 [[TMP2]]
; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[P]], i32 [[TMP3]]
; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[P]], i32 [[TMP4]]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF:.*]], label %[[PRED_LOAD_CONTINUE:.*]]
; CHECK:       [[PRED_LOAD_IF]]:
; CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP7]], align 4
; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> poison, i32 [[TMP12]], i64 0
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE]]
; CHECK:       [[PRED_LOAD_CONTINUE]]:
; CHECK-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ poison, %[[VECTOR_BODY]] ], [ [[TMP13]], %[[PRED_LOAD_IF]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF3:.*]], label %[[PRED_LOAD_CONTINUE4:.*]]
; CHECK:       [[PRED_LOAD_IF3]]:
; CHECK-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP8]], align 4
; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP16]], i64 1
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE4]]
; CHECK:       [[PRED_LOAD_CONTINUE4]]:
; CHECK-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP14]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP17]], %[[PRED_LOAD_IF3]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF5:.*]], label %[[PRED_LOAD_CONTINUE6:.*]]
; CHECK:       [[PRED_LOAD_IF5]]:
; CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP9]], align 4
; CHECK-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP20]], i64 2
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE6]]
; CHECK:       [[PRED_LOAD_CONTINUE6]]:
; CHECK-NEXT:    [[TMP22:%.*]] = phi <4 x i32> [ [[TMP18]], %[[PRED_LOAD_CONTINUE4]] ], [ [[TMP21]], %[[PRED_LOAD_IF5]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_LOAD_IF7:.*]], label %[[PRED_LOAD_CONTINUE8:.*]]
; CHECK:       [[PRED_LOAD_IF7]]:
; CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP10]], align 4
; CHECK-NEXT:    [[TMP25:%.*]] = insertelement <4 x i32> [[TMP22]], i32 [[TMP24]], i64 3
; CHECK-NEXT:    br label %[[PRED_LOAD_CONTINUE8]]
; CHECK:       [[PRED_LOAD_CONTINUE8]]:
; CHECK-NEXT:    [[TMP23:%.*]] = phi <4 x i32> [ [[TMP22]], %[[PRED_LOAD_CONTINUE6]] ], [ [[TMP25]], %[[PRED_LOAD_IF7]] ]
; CHECK-NEXT:    [[TMP27:%.*]] = udiv exact <4 x i32> [[TMP23]], splat (i32 100)
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
; CHECK:       [[PRED_STORE_IF]]:
; CHECK-NEXT:    [[TMP28:%.*]] = extractelement <4 x i32> [[TMP27]], i64 0
; CHECK-NEXT:    store i32 [[TMP28]], ptr [[TMP7]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE]]
; CHECK:       [[PRED_STORE_CONTINUE]]:
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF9:.*]], label %[[PRED_STORE_CONTINUE10:.*]]
; CHECK:       [[PRED_STORE_IF9]]:
; CHECK-NEXT:    [[TMP29:%.*]] = extractelement <4 x i32> [[TMP27]], i64 1
; CHECK-NEXT:    store i32 [[TMP29]], ptr [[TMP8]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE10]]
; CHECK:       [[PRED_STORE_CONTINUE10]]:
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF11:.*]], label %[[PRED_STORE_CONTINUE12:.*]]
; CHECK:       [[PRED_STORE_IF11]]:
; CHECK-NEXT:    [[TMP30:%.*]] = extractelement <4 x i32> [[TMP27]], i64 2
; CHECK-NEXT:    store i32 [[TMP30]], ptr [[TMP9]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE12]]
; CHECK:       [[PRED_STORE_CONTINUE12]]:
; CHECK-NEXT:    br i1 [[C]], label %[[PRED_STORE_IF13:.*]], label %[[PRED_STORE_CONTINUE14]]
; CHECK:       [[PRED_STORE_IF13]]:
; CHECK-NEXT:    [[TMP31:%.*]] = extractelement <4 x i32> [[TMP27]], i64 3
; CHECK-NEXT:    store i32 [[TMP31]], ptr [[TMP10]], align 4
; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE14]]
; CHECK:       [[PRED_STORE_CONTINUE14]]:
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP32]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LATCH:.*]] ]
; CHECK-NEXT:    br i1 [[C]], label %[[IF:.*]], label %[[LATCH]]
; CHECK:       [[IF]]:
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[P]], i32 [[IV]]
; CHECK-NEXT:    [[X:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT:    [[Y:%.*]] = udiv exact i32 [[X]], 100
; CHECK-NEXT:    store i32 [[Y]], ptr [[GEP]], align 4
; CHECK-NEXT:    br label %[[LATCH]]
; CHECK:       [[LATCH]]:
; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret void
;
entry:
  br label %loop

loop:
  %iv = phi i32 [0, %entry], [%iv.next, %latch]
  br i1 %c, label %if, label %latch

if:
  %gep = getelementptr i32, ptr %p, i32 %iv
  %x = load i32, ptr %gep
  %y = udiv exact i32 %x, 100
  store i32 %y, ptr %gep
  br label %latch

latch:
  %iv.next = add i32 %iv, 1
  %ec = icmp eq i32 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  ret void
}