; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt -passes=loop-vectorize -force-vector-width=2 -S %s | FileCheck %s --check-prefix=VF2
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -S %s | FileCheck %s --check-prefix=VF4
; RUN: opt -passes=loop-vectorize -force-vector-width=8 -S %s | FileCheck %s --check-prefix=VF8

define i32 @clamped_load_reduction_bound4(ptr %A) {
; VF2-LABEL: define i32 @clamped_load_reduction_bound4(
; VF2-SAME: ptr [[A:%.*]]) {
; VF2-NEXT:  [[ENTRY:.*]]:
; VF2-NEXT:    br label %[[LOOP:.*]]
; VF2:       [[LOOP]]:
; VF2-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP]] ]
; VF2-NEXT:    [[SUM:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[SUM_NEXT:%.*]], %[[LOOP]] ]
; VF2-NEXT:    [[CLAMPED:%.*]] = urem i32 [[INDEX]], 4
; VF2-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[CLAMPED]]
; VF2-NEXT:    [[LV:%.*]] = load i32, ptr [[GEP_A]], align 4
; VF2-NEXT:    [[SUM_NEXT]] = add i32 [[SUM]], [[LV]]
; VF2-NEXT:    [[INDEX_NEXT]] = add nuw nsw i32 [[INDEX]], 1
; VF2-NEXT:    [[COND:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
; VF2-NEXT:    br i1 [[COND]], label %[[EXIT:.*]], label %[[LOOP]]
; VF2:       [[EXIT]]:
; VF2-NEXT:    [[R:%.*]] = phi i32 [ [[SUM_NEXT]], %[[LOOP]] ]
; VF2-NEXT:    ret i32 [[R]]
;
; VF4-LABEL: define i32 @clamped_load_reduction_bound4(
; VF4-SAME: ptr [[A:%.*]]) {
; VF4-NEXT:  [[ENTRY:.*]]:
; VF4-NEXT:    br label %[[LOOP:.*]]
; VF4:       [[LOOP]]:
; VF4-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP]] ]
; VF4-NEXT:    [[SUM:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[SUM_NEXT:%.*]], %[[LOOP]] ]
; VF4-NEXT:    [[CLAMPED:%.*]] = urem i32 [[INDEX]], 4
; VF4-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[CLAMPED]]
; VF4-NEXT:    [[LV:%.*]] = load i32, ptr [[GEP_A]], align 4
; VF4-NEXT:    [[SUM_NEXT]] = add i32 [[SUM]], [[LV]]
; VF4-NEXT:    [[INDEX_NEXT]] = add nuw nsw i32 [[INDEX]], 1
; VF4-NEXT:    [[COND:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
; VF4-NEXT:    br i1 [[COND]], label %[[EXIT:.*]], label %[[LOOP]]
; VF4:       [[EXIT]]:
; VF4-NEXT:    [[R:%.*]] = phi i32 [ [[SUM_NEXT]], %[[LOOP]] ]
; VF4-NEXT:    ret i32 [[R]]
;
; VF8-LABEL: define i32 @clamped_load_reduction_bound4(
; VF8-SAME: ptr [[A:%.*]]) {
; VF8-NEXT:  [[ENTRY:.*]]:
; VF8-NEXT:    br label %[[LOOP:.*]]
; VF8:       [[LOOP]]:
; VF8-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP]] ]
; VF8-NEXT:    [[SUM:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[SUM_NEXT:%.*]], %[[LOOP]] ]
; VF8-NEXT:    [[CLAMPED:%.*]] = urem i32 [[INDEX]], 4
; VF8-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[CLAMPED]]
; VF8-NEXT:    [[LV:%.*]] = load i32, ptr [[GEP_A]], align 4
; VF8-NEXT:    [[SUM_NEXT]] = add i32 [[SUM]], [[LV]]
; VF8-NEXT:    [[INDEX_NEXT]] = add nuw nsw i32 [[INDEX]], 1
; VF8-NEXT:    [[COND:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
; VF8-NEXT:    br i1 [[COND]], label %[[EXIT:.*]], label %[[LOOP]]
; VF8:       [[EXIT]]:
; VF8-NEXT:    [[R:%.*]] = phi i32 [ [[SUM_NEXT]], %[[LOOP]] ]
; VF8-NEXT:    ret i32 [[R]]
;
entry:
  br label %loop

loop:
  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
  %sum = phi i32 [ 0, %entry ], [ %sum.next, %loop ]
  %clamped = urem i32 %iv, 4
  %gep.A = getelementptr inbounds i32, ptr %A, i32 %clamped
  %lv = load i32, ptr %gep.A
  %sum.next = add i32 %sum, %lv
  %iv.next = add nuw nsw i32 %iv, 1
  %cond = icmp eq i32 %iv.next, 128
  br i1 %cond, label %exit, label %loop

exit:
  %r = phi i32 [ %sum.next, %loop ]
  ret i32 %r
}

define i32 @clamped_load_reduction_bound8(ptr %A) {
; VF2-LABEL: define i32 @clamped_load_reduction_bound8(
; VF2-SAME: ptr [[A:%.*]]) {
; VF2-NEXT:  [[ENTRY:.*]]:
; VF2-NEXT:    br label %[[LOOP:.*]]
; VF2:       [[LOOP]]:
; VF2-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP]] ]
; VF2-NEXT:    [[SUM:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[SUM_NEXT:%.*]], %[[LOOP]] ]
; VF2-NEXT:    [[CLAMPED:%.*]] = urem i32 [[INDEX]], 8
; VF2-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[CLAMPED]]
; VF2-NEXT:    [[LV:%.*]] = load i32, ptr [[GEP_A]], align 4
; VF2-NEXT:    [[SUM_NEXT]] = add i32 [[SUM]], [[LV]]
; VF2-NEXT:    [[INDEX_NEXT]] = add nuw nsw i32 [[INDEX]], 1
; VF2-NEXT:    [[COND:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
; VF2-NEXT:    br i1 [[COND]], label %[[EXIT:.*]], label %[[LOOP]]
; VF2:       [[EXIT]]:
; VF2-NEXT:    [[R:%.*]] = phi i32 [ [[SUM_NEXT]], %[[LOOP]] ]
; VF2-NEXT:    ret i32 [[R]]
;
; VF4-LABEL: define i32 @clamped_load_reduction_bound8(
; VF4-SAME: ptr [[A:%.*]]) {
; VF4-NEXT:  [[ENTRY:.*]]:
; VF4-NEXT:    br label %[[LOOP:.*]]
; VF4:       [[LOOP]]:
; VF4-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP]] ]
; VF4-NEXT:    [[SUM:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[SUM_NEXT:%.*]], %[[LOOP]] ]
; VF4-NEXT:    [[CLAMPED:%.*]] = urem i32 [[INDEX]], 8
; VF4-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[CLAMPED]]
; VF4-NEXT:    [[LV:%.*]] = load i32, ptr [[GEP_A]], align 4
; VF4-NEXT:    [[SUM_NEXT]] = add i32 [[SUM]], [[LV]]
; VF4-NEXT:    [[INDEX_NEXT]] = add nuw nsw i32 [[INDEX]], 1
; VF4-NEXT:    [[COND:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
; VF4-NEXT:    br i1 [[COND]], label %[[EXIT:.*]], label %[[LOOP]]
; VF4:       [[EXIT]]:
; VF4-NEXT:    [[R:%.*]] = phi i32 [ [[SUM_NEXT]], %[[LOOP]] ]
; VF4-NEXT:    ret i32 [[R]]
;
; VF8-LABEL: define i32 @clamped_load_reduction_bound8(
; VF8-SAME: ptr [[A:%.*]]) {
; VF8-NEXT:  [[ENTRY:.*]]:
; VF8-NEXT:    br label %[[LOOP:.*]]
; VF8:       [[LOOP]]:
; VF8-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP]] ]
; VF8-NEXT:    [[SUM:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[SUM_NEXT:%.*]], %[[LOOP]] ]
; VF8-NEXT:    [[CLAMPED:%.*]] = urem i32 [[INDEX]], 8
; VF8-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[CLAMPED]]
; VF8-NEXT:    [[LV:%.*]] = load i32, ptr [[GEP_A]], align 4
; VF8-NEXT:    [[SUM_NEXT]] = add i32 [[SUM]], [[LV]]
; VF8-NEXT:    [[INDEX_NEXT]] = add nuw nsw i32 [[INDEX]], 1
; VF8-NEXT:    [[COND:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
; VF8-NEXT:    br i1 [[COND]], label %[[EXIT:.*]], label %[[LOOP]]
; VF8:       [[EXIT]]:
; VF8-NEXT:    [[R:%.*]] = phi i32 [ [[SUM_NEXT]], %[[LOOP]] ]
; VF8-NEXT:    ret i32 [[R]]
;
entry:
  br label %loop

loop:
  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
  %sum = phi i32 [ 0, %entry ], [ %sum.next, %loop ]
  %clamped = urem i32 %iv, 8
  %gep.A = getelementptr inbounds i32, ptr %A, i32 %clamped
  %lv = load i32, ptr %gep.A
  %sum.next = add i32 %sum, %lv
  %iv.next = add nuw nsw i32 %iv, 1
  %cond = icmp eq i32 %iv.next, 128
  br i1 %cond, label %exit, label %loop

exit:
  %r = phi i32 [ %sum.next, %loop ]
  ret i32 %r
}