1 files changed, 390 insertions, 0 deletions
diff --git a/llvm/test/Transforms/LoopPredication/reverse.ll b/llvm/test/Transforms/LoopPredication/reverse.ll
new file mode 100644
index 00000000000..95bfe20c04e
--- /dev/null
+++ b/llvm/test/Transforms/LoopPredication/reverse.ll
@@ -0,0 +1,390 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S -loop-predication -loop-predication-enable-count-down-loop=true < %s 2>&1 | FileCheck %s
+; RUN: opt -S -passes='require<scalar-evolution>,loop(loop-predication)' -loop-predication-enable-count-down-loop=true < %s 2>&1 | FileCheck %s
+
+declare void @llvm.experimental.guard(i1, ...)
+
+define i32 @signed_reverse_loop_n_to_lower_limit(i32* %array, i32 %length, i32 %n, i32 %lowerlimit) {
+; CHECK-LABEL: @signed_reverse_loop_n_to_lower_limit(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 [[TMP0]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp sge i32 [[LOWERLIMIT:%.*]], 1
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ [[N]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], -1
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I_NEXT]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sgt i32 [[I]], [[LOWERLIMIT]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
+  %i.next = add nsw i32 %i, -1
+  %within.bounds = icmp ult i32 %i.next, %length
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+  %i.i64 = zext i32 %i.next to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %loop.acc.next = add i32 %loop.acc, %array.i
+  %continue = icmp sgt i32 %i, %lowerlimit
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+
+define i32 @unsigned_reverse_loop_n_to_lower_limit(i32* %array, i32 %length, i32 %n, i32 %lowerlimit) {
+; CHECK-LABEL: @unsigned_reverse_loop_n_to_lower_limit(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 [[TMP0]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp uge i32 [[LOWERLIMIT:%.*]], 1
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ [[N]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], -1
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I_NEXT]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ugt i32 [[I]], [[LOWERLIMIT]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
+  %i.next = add nsw i32 %i, -1
+  %within.bounds = icmp ult i32 %i.next, %length
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+  %i.i64 = zext i32 %i.next to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %loop.acc.next = add i32 %loop.acc, %array.i
+  %continue = icmp ugt i32 %i, %lowerlimit
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+
+
+; if we predicated the loop, the guard will definitely fail and we will
+; deoptimize early on.
+define i32 @unsigned_reverse_loop_n_to_0(i32* %array, i32 %length, i32 %n, i32 %lowerlimit) {
+; CHECK-LABEL: @unsigned_reverse_loop_n_to_0(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 [[TMP0]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], false
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ [[N]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], -1
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I_NEXT]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ugt i32 [[I]], 0
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
+  %i.next = add nsw i32 %i, -1
+  %within.bounds = icmp ult i32 %i.next, %length
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+  %i.i64 = zext i32 %i.next to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %loop.acc.next = add i32 %loop.acc, %array.i
+  %continue = icmp ugt i32 %i, 0
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+
+; do not loop predicate when the range has step -1 and latch has step 1.
+define i32 @reverse_loop_range_step_increment(i32 %n, i32* %array, i32 %length) {
+; CHECK-LABEL: @reverse_loop_range_step_increment(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ [[N]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[IRC:%.*]] = phi i32 [ [[I_INC:%.*]], [[LOOP]] ], [ 1, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_INC]] = add nuw nsw i32 [[IRC]], 1
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[IRC]], [[LENGTH:%.*]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[IRC]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], -1
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ugt i32 [[I]], 65534
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
+  %irc = phi i32 [ %i.inc, %loop ], [ 1, %loop.preheader ]
+  %i.inc = add nuw nsw i32 %irc, 1
+  %within.bounds = icmp ult i32 %irc, %length
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+  %i.i64 = zext i32 %irc to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %i.next = add nsw i32 %i, -1
+  %loop.acc.next = add i32 %loop.acc, %array.i
+  %continue = icmp ugt i32 %i, 65534
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+
+define i32 @signed_reverse_loop_n_to_lower_limit_equal(i32* %array, i32 %length, i32 %n, i32 %lowerlimit) {
+; CHECK-LABEL: @signed_reverse_loop_n_to_lower_limit_equal(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 [[TMP0]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp sgt i32 [[LOWERLIMIT:%.*]], 1
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ [[N]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], -1
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I_NEXT]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sge i32 [[I]], [[LOWERLIMIT]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
+  %i.next = add nsw i32 %i, -1
+  %within.bounds = icmp ult i32 %i.next, %length
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+  %i.i64 = zext i32 %i.next to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %loop.acc.next = add i32 %loop.acc, %array.i
+  %continue = icmp sge i32 %i, %lowerlimit
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+
+define i32 @unsigned_reverse_loop_n_to_lower_limit_equal(i32* %array, i32 %length, i32 %n, i32 %lowerlimit) {
+; CHECK-LABEL: @unsigned_reverse_loop_n_to_lower_limit_equal(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 [[TMP0]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ugt i32 [[LOWERLIMIT:%.*]], 1
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ [[N]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], -1
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I_NEXT]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp uge i32 [[I]], [[LOWERLIMIT]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
+  %i.next = add nsw i32 %i, -1
+  %within.bounds = icmp ult i32 %i.next, %length
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+  %i.i64 = zext i32 %i.next to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %loop.acc.next = add i32 %loop.acc, %array.i
+  %continue = icmp uge i32 %i, %lowerlimit
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+
+
+; if we predicated the loop, the guard will definitely fail and we will
+; deoptimize early on.
+define i32 @unsigned_reverse_loop_n_to_1(i32* %array, i32 %length, i32 %n, i32 %lowerlimit) {
+; CHECK-LABEL: @unsigned_reverse_loop_n_to_1(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 [[TMP0]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], false
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ [[N]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], -1
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I_NEXT]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp uge i32 [[I]], 1
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
+  %i.next = add nsw i32 %i, -1
+  %within.bounds = icmp ult i32 %i.next, %length
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+  %i.i64 = zext i32 %i.next to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %loop.acc.next = add i32 %loop.acc, %array.i
+  %continue = icmp uge i32 %i, 1
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+