[Tests] Autogen all the LoopPredication tests

I'm about to make some changes to the pass which cause widespread - but uninteresting - test diffs.  Prepare the tests for easy updating.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@357404 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/Transforms/LoopPredication/basic.ll b/test/Transforms/LoopPredication/basic.ll
index 59842902b0e1f8c280cb5811aa49db7dd0c2b592..5575e6ebe13c1eef102ed14137561e82782e5a2d 100644 (file)
--- a/test/Transforms/LoopPredication/basic.ll
+++ b/test/Transforms/LoopPredication/basic.ll
@@ -1,25 +1,46 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -loop-predication < %s 2>&1 | FileCheck %s
 ; RUN: opt -S -passes='require<scalar-evolution>,loop(loop-predication)' < %s 2>&1 | FileCheck %s
 
 declare void @llvm.experimental.guard(i1, ...)
 
 define i32 @unsigned_loop_0_to_n_ult_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @unsigned_loop_0_to_n_ult_check
+; CHECK-LABEL: @unsigned_loop_0_to_n_ult_check(
+; 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:%.*]] = icmp ule i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp ule i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -40,22 +61,42 @@ exit:
 }
 
 define i32 @unsigned_loop_0_to_n_ule_latch_ult_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @unsigned_loop_0_to_n_ule_latch_ult_check
+; CHECK-LABEL: @unsigned_loop_0_to_n_ule_latch_ult_check(
+; 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:%.*]] = icmp ult i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ule i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp ult i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -76,22 +117,42 @@ exit:
 }
 
 define i32 @unsigned_loop_0_to_n_ugt_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @unsigned_loop_0_to_n_ugt_check
+; CHECK-LABEL: @unsigned_loop_0_to_n_ugt_check(
+; 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:%.*]] = icmp ule i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ugt i32 [[LENGTH]], [[I]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp ule i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ugt i32 %length, %i
@@ -112,22 +173,42 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_ult_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_ult_check
+; CHECK-LABEL: @signed_loop_0_to_n_ult_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp sle i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp sle i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -148,22 +229,43 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_ult_check_length_range_known(i32* %array, i32* %length.ptr, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_ult_check_length_range_known
+; CHECK-LABEL: @signed_loop_0_to_n_ult_check_length_range_known(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    [[LENGTH:%.*]] = load i32, i32* [[LENGTH_PTR:%.*]], !range !0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp sle i32 [[N]], [[LENGTH]]
+; CHECK-NEXT:    [[TMP1:%.*]] = and i1 true, [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP1]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   %length = load i32, i32* %length.ptr, !range !{i32 1, i32 2147483648}
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp sle i32 %n, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 true, [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -184,22 +286,42 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_inverse_latch_predicate(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_inverse_latch_predicate
+; CHECK-LABEL: @signed_loop_0_to_n_inverse_latch_predicate(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp slt i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sgt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[EXIT_LOOPEXIT:%.*]], label [[LOOP]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp slt i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -220,22 +342,42 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_sle_latch_ult_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_sle_latch_ult_check
+; CHECK-LABEL: @signed_loop_0_to_n_sle_latch_ult_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp slt i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sle i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp slt i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -256,23 +398,43 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_preincrement_latch_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_preincrement_latch_check
+; CHECK-LABEL: @signed_loop_0_to_n_preincrement_latch_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LENGTH:%.*]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sle i32 [[N]], [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[TMP1]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[length_minus_1:[^ ]+]] = add i32 %length, -1
-; CHECK-NEXT: [[limit_check:[^ ]+]] = icmp sle i32 %n, [[length_minus_1]]
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -293,23 +455,43 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_preincrement_latch_check_postincrement_guard_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_preincrement_latch_check_postincrement_guard_check
+; CHECK-LABEL: @signed_loop_0_to_n_preincrement_latch_check_postincrement_guard_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LENGTH:%.*]], -2
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sle i32 [[N]], [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult i32 1, [[LENGTH]]
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[TMP1]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_NEXT]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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 slt i32 [[I]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[length_minus_2:[^ ]+]] = add i32 %length, -2
-; CHECK-NEXT: [[limit_check:[^ ]+]] = icmp sle i32 %n, [[length_minus_2]]
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 1, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
 
@@ -331,23 +513,44 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_sle_latch_offset_ult_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_sle_latch_offset_ult_check
+; CHECK-LABEL: @signed_loop_0_to_n_sle_latch_offset_ult_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LENGTH:%.*]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[N]], [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult i32 1, [[LENGTH]]
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[TMP1]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_OFFSET:%.*]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_OFFSET]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sle i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[length_minus_1:[^ ]+]] = add i32 %length, -1
-; CHECK-NEXT: [[limit_check:[^ ]+]] = icmp slt i32 %n, [[length_minus_1]]
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 1, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %i.offset = add i32 %i, 1
@@ -369,22 +572,44 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_offset_sle_latch_offset_ult_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_offset_sle_latch_offset_ult_check
+; CHECK-LABEL: @signed_loop_0_to_n_offset_sle_latch_offset_ult_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp slt i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 1, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I_OFFSET:%.*]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_OFFSET]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[I_NEXT_OFFSET:%.*]] = add i32 [[I_NEXT]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sle i32 [[I_NEXT_OFFSET]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp slt i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 1, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %i.offset = add i32 %i, 1
@@ -407,20 +632,39 @@ exit:
 }
 
 define i32 @unsupported_latch_pred_loop_0_to_n(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @unsupported_latch_pred_loop_0_to_n
+; CHECK-LABEL: @unsupported_latch_pred_loop_0_to_n(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH:%.*]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ne i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %within.bounds = icmp ult i32 %i, %length
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -441,20 +685,39 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_unsupported_iv_step(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_unsupported_iv_step
+; CHECK-LABEL: @signed_loop_0_to_n_unsupported_iv_step(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH:%.*]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 2
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %within.bounds = icmp ult i32 %i, %length
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -475,22 +738,44 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_equal_iv_range_check(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_equal_iv_range_check
+; CHECK-LABEL: @signed_loop_0_to_n_equal_iv_range_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp sle i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[J]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add nsw i32 [[J]], 1
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp sle i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %j = phi i32 [ %j.next, %loop ], [ 0, %loop.preheader ]
@@ -514,26 +799,48 @@ exit:
 }
 
 define i32 @signed_loop_start_to_n_offset_iv_range_check(i32* %array, i32 %start.i,
-                                                         i32 %start.j, i32 %length,
-                                                         i32 %n) {
-; CHECK-LABEL: @signed_loop_start_to_n_offset_iv_range_check
+; CHECK-LABEL: @signed_loop_start_to_n_offset_iv_range_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LENGTH:%.*]], [[START_I:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[START_J:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp sle i32 [[N]], [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult i32 [[START_J]], [[LENGTH]]
+; CHECK-NEXT:    [[TMP4:%.*]] = and i1 [[TMP3]], [[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]] ], [ [[START_I]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[LOOP]] ], [ [[START_J]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[J]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP4]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add i32 [[J]], 1
+; CHECK-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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]]
+;
+  i32 %start.j, i32 %length,
+  i32 %n) {
 entry:
   %tmp5 = icmp sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[length_plus_start_i:[^ ]+]] = add i32 %length, %start.i
-; CHECK-NEXT: [[limit:[^ ]+]] = sub i32 [[length_plus_start_i]], %start.j
-; CHECK-NEXT: [[limit_check:[^ ]+]] = icmp sle i32 %n, [[limit]]
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 %start.j, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %start.i, %loop.preheader ]
   %j = phi i32 [ %j.next, %loop ], [ %start.j, %loop.preheader ]
@@ -557,20 +864,41 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_different_iv_types(i32* %array, i16 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_different_iv_types
+; CHECK-LABEL: @signed_loop_0_to_n_different_iv_types(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i16 [ [[J_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i16 [[J]], [[LENGTH:%.*]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add i16 [[J]], 1
+; CHECK-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %within.bounds = icmp ult i16 %j, %length
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %j = phi i16 [ %j.next, %loop ], [ 0, %loop.preheader ]
@@ -594,20 +922,41 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_different_iv_strides(i32* %array, i32 %length, i32 %n) {
-; CHECK-LABEL: @signed_loop_0_to_n_different_iv_strides
+; CHECK-LABEL: @signed_loop_0_to_n_different_iv_strides(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[J]], [[LENGTH:%.*]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add nsw i32 [[J]], 2
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %within.bounds = icmp ult i32 %j, %length
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %j = phi i32 [ %j.next, %loop ], [ 0, %loop.preheader ]
@@ -631,27 +980,52 @@ exit:
 }
 
 define i32 @two_range_checks(i32* %array.1, i32 %length.1,
-                             i32* %array.2, i32 %length.2, i32 %n) {
-; CHECK-LABEL: @two_range_checks
+; CHECK-LABEL: @two_range_checks(
+; 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:%.*]] = icmp ule i32 [[N]], [[LENGTH_2:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH_2]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ule i32 [[N]], [[LENGTH_1:%.*]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ult i32 0, [[LENGTH_1]]
+; CHECK-NEXT:    [[TMP5:%.*]] = and i1 [[TMP4]], [[TMP3]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_1:%.*]] = icmp ult i32 [[I]], [[LENGTH_1]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_2:%.*]] = icmp ult i32 [[I]], [[LENGTH_2]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = and i1 [[WITHIN_BOUNDS_1]], [[WITHIN_BOUNDS_2]]
+; CHECK-NEXT:    [[TMP6:%.*]] = and i1 [[TMP2]], [[TMP5]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP6]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
+; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
+; CHECK-NEXT:    [[ARRAY_2_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_2:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_2_I:%.*]] = load i32, i32* [[ARRAY_2_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_2_I]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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]]
+;
+  i32* %array.2, i32 %length.2, i32 %n) {
 entry:
   %tmp5 = icmp eq i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check_1:[^ ]+]] = icmp ule i32 %n, %length.{{1|2}}
-; CHECK-NEXT: [[first_iteration_check_1:[^ ]+]] = icmp ult i32 0, %length.{{1|2}}
-; CHECK-NEXT: [[wide_cond_1:[^ ]+]] = and i1 [[first_iteration_check_1]], [[limit_check_1]]
-; CHECK-NEXT: [[limit_check_2:[^ ]+]] = icmp ule i32 %n, %length.{{1|2}}
-; CHECK-NEXT: [[first_iteration_check_2:[^ ]+]] = icmp ult i32 0, %length.{{1|2}}
-; CHECK-NEXT: [[wide_cond_2:[^ ]+]] = and i1 [[first_iteration_check_2]], [[limit_check_2]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: [[wide_cond:[^ ]+]] = and i1 [[wide_cond_1]], [[wide_cond_2]]
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds.1 = icmp ult i32 %i, %length.1
@@ -678,32 +1052,62 @@ exit:
 }
 
 define i32 @three_range_checks(i32* %array.1, i32 %length.1,
-                               i32* %array.2, i32 %length.2,
-                               i32* %array.3, i32 %length.3, i32 %n) {
-; CHECK-LABEL: @three_range_checks
+; CHECK-LABEL: @three_range_checks(
+; 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:%.*]] = icmp ule i32 [[N]], [[LENGTH_3:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH_3]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ule i32 [[N]], [[LENGTH_2:%.*]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ult i32 0, [[LENGTH_2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = and i1 [[TMP4]], [[TMP3]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp ule i32 [[N]], [[LENGTH_1:%.*]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp ult i32 0, [[LENGTH_1]]
+; CHECK-NEXT:    [[TMP8:%.*]] = and i1 [[TMP7]], [[TMP6]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_1:%.*]] = icmp ult i32 [[I]], [[LENGTH_1]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_2:%.*]] = icmp ult i32 [[I]], [[LENGTH_2]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_3:%.*]] = icmp ult i32 [[I]], [[LENGTH_3]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_1_AND_2:%.*]] = and i1 [[WITHIN_BOUNDS_1]], [[WITHIN_BOUNDS_2]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = and i1 [[WITHIN_BOUNDS_1_AND_2]], [[WITHIN_BOUNDS_3]]
+; CHECK-NEXT:    [[TMP9:%.*]] = and i1 [[TMP2]], [[TMP5]]
+; CHECK-NEXT:    [[TMP10:%.*]] = and i1 [[TMP9]], [[TMP8]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP10]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
+; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
+; CHECK-NEXT:    [[ARRAY_2_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_2:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_2_I:%.*]] = load i32, i32* [[ARRAY_2_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_2:%.*]] = add i32 [[LOOP_ACC_1]], [[ARRAY_2_I]]
+; CHECK-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_2]], [[ARRAY_3_I]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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]]
+;
+  i32* %array.2, i32 %length.2,
+  i32* %array.3, i32 %length.3, i32 %n) {
 entry:
   %tmp5 = icmp eq i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check_1:[^ ]+]] = icmp ule i32 %n, %length.{{1|2|3}}
-; CHECK-NEXT: [[first_iteration_check_1:[^ ]+]] = icmp ult i32 0, %length.{{1|2|3}}
-; CHECK-NEXT: [[wide_cond_1:[^ ]+]] = and i1 [[first_iteration_check_1]], [[limit_check_1]]
-; CHECK-NEXT: [[limit_check_2:[^ ]+]] = icmp ule i32 %n, %length.{{1|2|3}}
-; CHECK-NEXT: [[first_iteration_check_2:[^ ]+]] = icmp ult i32 0, %length.{{1|2|3}}
-; CHECK-NEXT: [[wide_cond_2:[^ ]+]] = and i1 [[first_iteration_check_2]], [[limit_check_2]]
-; CHECK-NEXT: [[limit_check_3:[^ ]+]] = icmp ule i32 %n, %length.{{1|2|3}}
-; CHECK-NEXT: [[first_iteration_check_3:[^ ]+]] = icmp ult i32 0, %length.{{1|2|3}}
-; CHECK-NEXT: [[wide_cond_3:[^ ]+]] = and i1 [[first_iteration_check_3]], [[limit_check_3]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: [[wide_cond_and:[^ ]+]] = and i1 [[wide_cond_1]], [[wide_cond_2]]
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[wide_cond_and]], [[wide_cond_3]]
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds.1 = icmp ult i32 %i, %length.1
@@ -736,32 +1140,60 @@ exit:
 }
 
 define i32 @three_guards(i32* %array.1, i32 %length.1,
-                         i32* %array.2, i32 %length.2,
-                         i32* %array.3, i32 %length.3, i32 %n) {
-; CHECK-LABEL: @three_guards
+; CHECK-LABEL: @three_guards(
+; 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:%.*]] = icmp ule i32 [[N]], [[LENGTH_1:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH_1]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ule i32 [[N]], [[LENGTH_2:%.*]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ult i32 0, [[LENGTH_2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = and i1 [[TMP4]], [[TMP3]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp ule i32 [[N]], [[LENGTH_3:%.*]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp ult i32 0, [[LENGTH_3]]
+; CHECK-NEXT:    [[TMP8:%.*]] = and i1 [[TMP7]], [[TMP6]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_1:%.*]] = icmp ult i32 [[I]], [[LENGTH_1]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
+; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_2:%.*]] = icmp ult i32 [[I]], [[LENGTH_2]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP5]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[ARRAY_2_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_2:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_2_I:%.*]] = load i32, i32* [[ARRAY_2_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_2:%.*]] = add i32 [[LOOP_ACC_1]], [[ARRAY_2_I]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS_3:%.*]] = icmp ult i32 [[I]], [[LENGTH_3]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP8]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_2]], [[ARRAY_3_I]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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]]
+;
+  i32* %array.2, i32 %length.2,
+  i32* %array.3, i32 %length.3, i32 %n) {
 entry:
   %tmp5 = icmp eq i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check_1:[^ ]+]] = icmp ule i32 %n, %length.{{1|2|3}}
-; CHECK-NEXT: [[first_iteration_check_1:[^ ]+]] = icmp ult i32 0, %length.{{1|2|3}}
-; CHECK-NEXT: [[wide_cond_1:[^ ]+]] = and i1 [[first_iteration_check_1]], [[limit_check_1]]
-; CHECK-NEXT: [[limit_check_2:[^ ]+]] = icmp ule i32 %n, %length.{{1|2|3}}
-; CHECK-NEXT: [[first_iteration_check_2:[^ ]+]] = icmp ult i32 0, %length.{{1|2|3}}
-; CHECK-NEXT: [[wide_cond_2:[^ ]+]] = and i1 [[first_iteration_check_2]], [[limit_check_2]]
-; CHECK-NEXT: [[limit_check_3:[^ ]+]] = icmp ule i32 %n, %length.{{1|2|3}}
-; CHECK-NEXT: [[first_iteration_check_3:[^ ]+]] = icmp ult i32 0, %length.{{1|2|3}}
-; CHECK-NEXT: [[wide_cond_3:[^ ]+]] = and i1 [[first_iteration_check_3]], [[limit_check_3]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond_1]], i32 9) [ "deopt"() ]
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond_2]], i32 9) [ "deopt"() ]
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond_3]], i32 9) [ "deopt"() ]
 
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
@@ -798,24 +1230,45 @@ exit:
 }
 
 define i32 @unsigned_loop_0_to_n_unrelated_condition(i32* %array, i32 %length, i32 %n, i32 %x) {
-; CHECK-LABEL: @unsigned_loop_0_to_n_unrelated_condition
+; CHECK-LABEL: @unsigned_loop_0_to_n_unrelated_condition(
+; 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:%.*]] = icmp ule i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    [[UNRELATED_COND:%.*]] = icmp ult i32 [[X:%.*]], [[LENGTH]]
+; CHECK-NEXT:    [[GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[UNRELATED_COND]]
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[UNRELATED_COND]], [[TMP2]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp ule i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %unrelated.cond = icmp ult i32 %x, %length
-; CHECK: [[guard_cond:[^ ]+]] = and i1 %unrelated.cond, [[wide_cond]]
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[guard_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -839,24 +1292,43 @@ exit:
 
 ; Don't change the guard condition if there were no widened subconditions
 define i32 @test_no_widened_conditions(i32* %array, i32 %length, i32 %n, i32 %x1, i32 %x2, i32 %x3) {
-; CHECK-LABEL: @test_no_widened_conditions
+; CHECK-LABEL: @test_no_widened_conditions(
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[UNRELATED_COND_1:%.*]] = icmp eq i32 [[X1:%.*]], [[I]]
+; CHECK-NEXT:    [[UNRELATED_COND_2:%.*]] = icmp eq i32 [[X2:%.*]], [[I]]
+; CHECK-NEXT:    [[UNRELATED_COND_3:%.*]] = icmp eq i32 [[X3:%.*]], [[I]]
+; CHECK-NEXT:    [[UNRELATED_COND_AND_1:%.*]] = and i1 [[UNRELATED_COND_1]], [[UNRELATED_COND_2]]
+; CHECK-NEXT:    [[GUARD_COND:%.*]] = and i1 [[UNRELATED_COND_AND_1]], [[UNRELATED_COND_3]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[GUARD_COND]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %unrelated.cond.1 = icmp eq i32 %x1, %i
-; CHECK-NEXT: %unrelated.cond.2 = icmp eq i32 %x2, %i
-; CHECK-NEXT: %unrelated.cond.3 = icmp eq i32 %x3, %i
-; CHECK-NEXT: %unrelated.cond.and.1 = and i1 %unrelated.cond.1, %unrelated.cond.2
-; CHECK-NEXT: %guard.cond = and i1 %unrelated.cond.and.1, %unrelated.cond.3
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %guard.cond, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %unrelated.cond.1 = icmp eq i32 %x1, %i
@@ -882,21 +1354,40 @@ exit:
 }
 
 define i32 @signed_loop_start_to_n_loop_variant_bound(i32* %array, i32 %x, i32 %start, i32 %n) {
-; CHECK-LABEL: @signed_loop_start_to_n_loop_variant_bound
+; CHECK-LABEL: @signed_loop_start_to_n_loop_variant_bound(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle 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]] ], [ [[START:%.*]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BOUND:%.*]] = add i32 [[I]], [[X:%.*]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[BOUND]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %bound = add i32 %i, %x
-; CHECK-NEXT: %within.bounds = icmp ult i32 %i, %bound
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %start, %loop.preheader ]
   %bound = add i32 %i, %x
@@ -918,20 +1409,39 @@ exit:
 }
 
 define i32 @signed_loop_start_to_n_non_monotonic_predicate(i32* %array, i32 %x, i32 %start, i32 %n) {
-; CHECK-LABEL: @signed_loop_start_to_n_non_monotonic_predicate
+; CHECK-LABEL: @signed_loop_start_to_n_non_monotonic_predicate(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle 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]] ], [ [[START:%.*]], [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[GUARD_COND:%.*]] = icmp eq i32 [[I]], [[X:%.*]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[GUARD_COND]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
 
 loop.preheader:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %guard.cond = icmp eq i32 %i, %x
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %guard.cond, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %start, %loop.preheader ]
   %guard.cond = icmp eq i32 %i, %x
@@ -952,23 +1462,44 @@ exit:
 }
 
 define i32 @unsigned_loop_0_to_n_hoist_length(i32* %array, i16 %length.i16, i32 %n) {
-; CHECK-LABEL: @unsigned_loop_0_to_n_hoist_length
+; CHECK-LABEL: @unsigned_loop_0_to_n_hoist_length(
+; 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:%.*]] = zext i16 [[LENGTH_I16:%.*]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule i32 [[N]], [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult i32 0, [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[TMP1]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[LENGTH:%.*]] = zext i16 [[LENGTH_I16]] to i32
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK: [[length:[^ ]+]] = zext i16 %length.i16 to i32
-; CHECK-NEXT: [[limit_check:[^ ]+]] = icmp ule i32 %n, [[length]]
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, [[length]]
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %length = zext i16 %length.i16 to i32
@@ -990,23 +1521,40 @@ exit:
 }
 
 define i32 @unsigned_loop_0_to_n_cant_hoist_length(i32* %array, i32 %length, i32 %divider, i32 %n) {
-; CHECK-LABEL: @unsigned_loop_0_to_n_cant_hoist_length
+; CHECK-LABEL: @unsigned_loop_0_to_n_cant_hoist_length(
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[LENGTH_UDIV:%.*]] = udiv i32 [[LENGTH:%.*]], [[DIVIDER:%.*]]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH_UDIV]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK-NEXT: %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
-; CHECK-NEXT: %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
-; CHECK-NEXT: %length.udiv = udiv i32 %length, %divider
-; CHECK-NEXT: %within.bounds = icmp ult i32 %i, %length.udiv
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %length.udiv = udiv i32 %length, %divider

diff --git a/test/Transforms/LoopPredication/basic_widenable_branch_guards.ll b/test/Transforms/LoopPredication/basic_widenable_branch_guards.ll
index c836e5f4d835965aff1c9c62a16c58905b04c3b1..a2d4b07a1f93cd3db984902ec4bc76fe6fa82db7 100644 (file)
--- a/test/Transforms/LoopPredication/basic_widenable_branch_guards.ll
+++ b/test/Transforms/LoopPredication/basic_widenable_branch_guards.ll
@@ -22,6 +22,23 @@ define i32 @unsigned_loop_0_to_n_ult_check(i32* %array, i32 %length, i32 %n) {
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -74,6 +91,23 @@ define i32 @unsigned_loop_0_to_n_ule_latch_ult_check(i32* %array, i32 %length, i
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ule i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -126,6 +160,23 @@ define i32 @unsigned_loop_0_to_n_ugt_check(i32* %array, i32 %length, i32 %n) {
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -178,8 +229,25 @@ define i32 @signed_loop_0_to_n_ult_check(i32* %array, i32 %length, i32 %n) {
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
-
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
   br i1 %tmp5, label %exit, label %loop.preheader
@@ -231,6 +299,23 @@ define i32 @signed_loop_0_to_n_ult_check_length_range_known(i32* %array, i32* %l
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -284,6 +369,23 @@ define i32 @signed_loop_0_to_n_inverse_latch_predicate(i32* %array, i32 %length,
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sgt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[EXIT_LOOPEXIT:%.*]], label [[LOOP]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -336,6 +438,23 @@ define i32 @signed_loop_0_to_n_sle_latch_ult_check(i32* %array, i32 %length, i32
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sle i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -389,6 +508,23 @@ define i32 @signed_loop_0_to_n_preincrement_latch_check(i32* %array, i32 %length
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = and i1 [[TMP3]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -443,6 +579,22 @@ define i32 @signed_loop_0_to_n_preincrement_latch_check_postincrement_guard_chec
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = and i1 [[TMP3]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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 slt i32 [[I]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -497,6 +649,23 @@ define i32 @signed_loop_0_to_n_sle_latch_offset_ult_check(i32* %array, i32 %leng
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = and i1 [[TMP3]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sle i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -551,6 +720,24 @@ define i32 @signed_loop_0_to_n_offset_sle_latch_offset_ult_check(i32* %array, i3
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[I_NEXT_OFFSET:%.*]] = add i32 [[I_NEXT]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp sle i32 [[I_NEXT_OFFSET]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -601,6 +788,23 @@ define i32 @unsupported_latch_pred_loop_0_to_n(i32* %array, i32 %length, i32 %n)
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ne i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -649,6 +853,23 @@ define i32 @signed_loop_0_to_n_unsupported_iv_step(i32* %array, i32 %length, i32
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 2
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -702,6 +923,24 @@ define i32 @signed_loop_0_to_n_equal_iv_range_check(i32* %array, i32 %length, i3
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add nsw i32 [[J]], 1
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -759,6 +998,24 @@ define i32 @signed_loop_start_to_n_offset_iv_range_check(i32* %array, i32 %start
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = and i1 [[TMP4]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add i32 [[J]], 1
+; CHECK-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -810,6 +1067,24 @@ define i32 @signed_loop_0_to_n_different_iv_types(i32* %array, i16 %length, i32
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add i16 [[J]], 1
+; CHECK-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -861,6 +1136,24 @@ define i32 @signed_loop_0_to_n_different_iv_strides(i32* %array, i32 %length, i3
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[J_NEXT]] = add nsw i32 [[J]], 2
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -921,6 +1214,26 @@ define i32 @two_range_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32 %l
 ; CHECK-NEXT:    [[TMP6:%.*]] = and i1 [[TMP2]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = and i1 [[TMP6]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
+; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
+; CHECK-NEXT:    [[ARRAY_2_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_2:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_2_I:%.*]] = load i32, i32* [[ARRAY_2_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_2_I]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -990,6 +1303,29 @@ define i32 @three_range_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32
 ; CHECK-NEXT:    [[TMP10:%.*]] = and i1 [[TMP9]], [[TMP8]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = and i1 [[TMP10]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
+; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
+; CHECK-NEXT:    [[ARRAY_2_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_2:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_2_I:%.*]] = load i32, i32* [[ARRAY_2_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_2:%.*]] = add i32 [[LOOP_ACC_1]], [[ARRAY_2_I]]
+; CHECK-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_2]], [[ARRAY_3_I]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -1180,6 +1516,23 @@ define i32 @unsigned_loop_0_to_n_unrelated_condition(i32* %array, i32 %length, i
 ; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[UNRELATED_COND]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = and i1 [[TMP3]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -1234,6 +1587,23 @@ define i32 @test_no_widened_conditions(i32* %array, i32 %length, i32 %n, i32 %x1
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[GUARD_COND]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -1287,6 +1657,23 @@ define i32 @signed_loop_start_to_n_loop_variant_bound(i32* %array, i32 %x, i32 %
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -1336,6 +1723,23 @@ define i32 @signed_loop_start_to_n_non_monotonic_predicate(i32* %array, i32 %x,
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[GUARD_COND]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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 sle i32 %n, 0
@@ -1390,6 +1794,23 @@ define i32 @unsigned_loop_0_to_n_hoist_length(i32* %array, i16 %length.i16, i32
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = and i1 [[TMP3]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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
@@ -1440,6 +1861,23 @@ define i32 @unsigned_loop_0_to_n_cant_hoist_length(i32* %array, i32 %length, i32
 ; CHECK-NEXT:    [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
 ; CHECK-NEXT:    [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[WITHIN_BOUNDS]], [[WIDENABLE_COND]]
 ; CHECK-NEXT:    br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK:       deopt:
+; CHECK-NEXT:    [[DEOPTCALL:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 9) [ "deopt"() ]
+; CHECK-NEXT:    ret i32 [[DEOPTCALL]]
+; CHECK:       guarded:
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
+; 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

diff --git a/test/Transforms/LoopPredication/nested.ll b/test/Transforms/LoopPredication/nested.ll
index 8a43b058a8725e47bc0ff0b69a24735a817285c9..27ad2cee263f824fcdcbc07f36d3fd092b8697b2 100644 (file)
--- a/test/Transforms/LoopPredication/nested.ll
+++ b/test/Transforms/LoopPredication/nested.ll
@@ -1,10 +1,53 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -loop-predication < %s 2>&1 | FileCheck %s
 ; RUN: opt -S -passes='require<scalar-evolution>,loop(loop-predication)' < %s 2>&1 | FileCheck %s
 
 declare void @llvm.experimental.guard(i1, ...)
 
 define i32 @signed_loop_0_to_n_nested_0_to_l_inner_index_check(i32* %array, i32 %length, i32 %n, i32 %l) {
-; CHECK-LABEL: @signed_loop_0_to_n_nested_0_to_l_inner_index_check
+; CHECK-LABEL: @signed_loop_0_to_n_nested_0_to_l_inner_index_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[OUTER_LOOP_PREHEADER:%.*]]
+; CHECK:       outer.loop.preheader:
+; CHECK-NEXT:    br label [[OUTER_LOOP:%.*]]
+; CHECK:       outer.loop:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT:%.*]], [[OUTER_LOOP_INC:%.*]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[OUTER_LOOP_INC]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp sle i32 [[L:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP6]], label [[OUTER_LOOP_INC]], label [[INNER_LOOP_PREHEADER:%.*]]
+; CHECK:       inner.loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp sle i32 [[L]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
+; CHECK:       inner.loop:
+; CHECK-NEXT:    [[INNER_LOOP_ACC:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT:%.*]], [[INNER_LOOP]] ], [ [[OUTER_LOOP_ACC]], [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[INNER_LOOP]] ], [ 0, [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[J]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[J_I64:%.*]] = zext i32 [[J]] to i64
+; CHECK-NEXT:    [[ARRAY_J_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[J_I64]]
+; CHECK-NEXT:    [[ARRAY_J:%.*]] = load i32, i32* [[ARRAY_J_PTR]], align 4
+; CHECK-NEXT:    [[INNER_LOOP_ACC_NEXT]] = add i32 [[INNER_LOOP_ACC]], [[ARRAY_J]]
+; CHECK-NEXT:    [[J_NEXT]] = add nsw i32 [[J]], 1
+; CHECK-NEXT:    [[INNER_CONTINUE:%.*]] = icmp slt i32 [[J_NEXT]], [[L]]
+; CHECK-NEXT:    br i1 [[INNER_CONTINUE]], label [[INNER_LOOP]], label [[OUTER_LOOP_INC_LOOPEXIT:%.*]]
+; CHECK:       outer.loop.inc.loopexit:
+; CHECK-NEXT:    [[INNER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT]], [[INNER_LOOP]] ]
+; CHECK-NEXT:    br label [[OUTER_LOOP_INC]]
+; CHECK:       outer.loop.inc:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT]] = phi i32 [ [[OUTER_LOOP_ACC]], [[OUTER_LOOP]] ], [ [[INNER_LOOP_ACC_NEXT_LCSSA]], [[OUTER_LOOP_INC_LOOPEXIT]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[OUTER_CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[OUTER_CONTINUE]], label [[OUTER_LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT]], [[OUTER_LOOP_INC]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[OUTER_LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
 entry:
   %tmp5 = icmp sle i32 %n, 0
   br i1 %tmp5, label %exit, label %outer.loop.preheader
@@ -17,24 +60,17 @@ outer.loop:
   %i = phi i32 [ %i.next, %outer.loop.inc ], [ 0, %outer.loop.preheader ]
   %tmp6 = icmp sle i32 %l, 0
   br i1 %tmp6, label %outer.loop.inc, label %inner.loop.preheader
-  
+
 inner.loop.preheader:
-; CHECK: inner.loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp sle i32 %l, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %inner.loop
   br label %inner.loop
 
 inner.loop:
-; CHECK: inner.loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %inner.loop.acc = phi i32 [ %inner.loop.acc.next, %inner.loop ], [ %outer.loop.acc, %inner.loop.preheader ]
   %j = phi i32 [ %j.next, %inner.loop ], [ 0, %inner.loop.preheader ]
 
   %within.bounds = icmp ult i32 %j, %length
   call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
-  
+
   %j.i64 = zext i32 %j to i64
   %array.j.ptr = getelementptr inbounds i32, i32* %array, i64 %j.i64
   %array.j = load i32, i32* %array.j.ptr, align 4
@@ -56,17 +92,54 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_nested_0_to_l_outer_index_check(i32* %array, i32 %length, i32 %n, i32 %l) {
-; CHECK-LABEL: @signed_loop_0_to_n_nested_0_to_l_outer_index_check
+; CHECK-LABEL: @signed_loop_0_to_n_nested_0_to_l_outer_index_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[OUTER_LOOP_PREHEADER:%.*]]
+; CHECK:       outer.loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp sle i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    br label [[OUTER_LOOP:%.*]]
+; CHECK:       outer.loop:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT:%.*]], [[OUTER_LOOP_INC:%.*]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[OUTER_LOOP_INC]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp sle i32 [[L:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP6]], label [[OUTER_LOOP_INC]], label [[INNER_LOOP_PREHEADER:%.*]]
+; CHECK:       inner.loop.preheader:
+; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
+; CHECK:       inner.loop:
+; CHECK-NEXT:    [[INNER_LOOP_ACC:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT:%.*]], [[INNER_LOOP]] ], [ [[OUTER_LOOP_ACC]], [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[INNER_LOOP]] ], [ 0, [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[INNER_LOOP_ACC_NEXT]] = add i32 [[INNER_LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[J_NEXT]] = add nsw i32 [[J]], 1
+; CHECK-NEXT:    [[INNER_CONTINUE:%.*]] = icmp slt i32 [[J_NEXT]], [[L]]
+; CHECK-NEXT:    br i1 [[INNER_CONTINUE]], label [[INNER_LOOP]], label [[OUTER_LOOP_INC_LOOPEXIT:%.*]]
+; CHECK:       outer.loop.inc.loopexit:
+; CHECK-NEXT:    [[INNER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT]], [[INNER_LOOP]] ]
+; CHECK-NEXT:    br label [[OUTER_LOOP_INC]]
+; CHECK:       outer.loop.inc:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT]] = phi i32 [ [[OUTER_LOOP_ACC]], [[OUTER_LOOP]] ], [ [[INNER_LOOP_ACC_NEXT_LCSSA]], [[OUTER_LOOP_INC_LOOPEXIT]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[OUTER_CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[OUTER_CONTINUE]], label [[OUTER_LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT]], [[OUTER_LOOP_INC]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[OUTER_LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
 entry:
   %tmp5 = icmp sle i32 %n, 0
   br i1 %tmp5, label %exit, label %outer.loop.preheader
 
 outer.loop.preheader:
-; CHECK: outer.loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp sle i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %outer.loop
   br label %outer.loop
 
 outer.loop:
@@ -74,20 +147,18 @@ outer.loop:
   %i = phi i32 [ %i.next, %outer.loop.inc ], [ 0, %outer.loop.preheader ]
   %tmp6 = icmp sle i32 %l, 0
   br i1 %tmp6, label %outer.loop.inc, label %inner.loop.preheader
-  
+
 inner.loop.preheader:
   br label %inner.loop
 
 inner.loop:
-; CHECK: inner.loop:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 
   %inner.loop.acc = phi i32 [ %inner.loop.acc.next, %inner.loop ], [ %outer.loop.acc, %inner.loop.preheader ]
   %j = phi i32 [ %j.next, %inner.loop ], [ 0, %inner.loop.preheader ]
 
   %within.bounds = icmp ult i32 %i, %length
   call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
-  
+
   %i.i64 = zext i32 %i to i64
   %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
   %array.i = load i32, i32* %array.i.ptr, align 4
@@ -109,42 +180,76 @@ exit:
 }
 
 define i32 @signed_loop_0_to_n_nested_i_to_l_inner_index_check(i32* %array, i32 %length, i32 %n, i32 %l) {
-; CHECK-LABEL: @signed_loop_0_to_n_nested_i_to_l_inner_index_check
+; CHECK-LABEL: @signed_loop_0_to_n_nested_i_to_l_inner_index_check(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[OUTER_LOOP_PREHEADER:%.*]]
+; CHECK:       outer.loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp sle i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    br label [[OUTER_LOOP:%.*]]
+; CHECK:       outer.loop:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT:%.*]], [[OUTER_LOOP_INC:%.*]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[OUTER_LOOP_INC]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp sle i32 [[L:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP6]], label [[OUTER_LOOP_INC]], label [[INNER_LOOP_PREHEADER:%.*]]
+; CHECK:       inner.loop.preheader:
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp sle i32 [[L]], [[LENGTH]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    [[TMP5:%.*]] = and i1 [[TMP4]], [[TMP3]]
+; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
+; CHECK:       inner.loop:
+; CHECK-NEXT:    [[INNER_LOOP_ACC:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT:%.*]], [[INNER_LOOP]] ], [ [[OUTER_LOOP_ACC]], [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[INNER_LOOP]] ], [ [[I]], [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[J]], [[LENGTH]]
+; CHECK-NEXT:    [[TMP6:%.*]] = and i1 [[TMP3]], [[TMP2]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP6]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[J_I64:%.*]] = zext i32 [[J]] to i64
+; CHECK-NEXT:    [[ARRAY_J_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[J_I64]]
+; CHECK-NEXT:    [[ARRAY_J:%.*]] = load i32, i32* [[ARRAY_J_PTR]], align 4
+; CHECK-NEXT:    [[INNER_LOOP_ACC_NEXT]] = add i32 [[INNER_LOOP_ACC]], [[ARRAY_J]]
+; CHECK-NEXT:    [[J_NEXT]] = add nsw i32 [[J]], 1
+; CHECK-NEXT:    [[INNER_CONTINUE:%.*]] = icmp slt i32 [[J_NEXT]], [[L]]
+; CHECK-NEXT:    br i1 [[INNER_CONTINUE]], label [[INNER_LOOP]], label [[OUTER_LOOP_INC_LOOPEXIT:%.*]]
+; CHECK:       outer.loop.inc.loopexit:
+; CHECK-NEXT:    [[INNER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT]], [[INNER_LOOP]] ]
+; CHECK-NEXT:    br label [[OUTER_LOOP_INC]]
+; CHECK:       outer.loop.inc:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT]] = phi i32 [ [[OUTER_LOOP_ACC]], [[OUTER_LOOP]] ], [ [[INNER_LOOP_ACC_NEXT_LCSSA]], [[OUTER_LOOP_INC_LOOPEXIT]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[OUTER_CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[OUTER_CONTINUE]], label [[OUTER_LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT]], [[OUTER_LOOP_INC]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[OUTER_LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
 entry:
   %tmp5 = icmp sle i32 %n, 0
   br i1 %tmp5, label %exit, label %outer.loop.preheader
 
 outer.loop.preheader:
-; CHECK: outer.loop.preheader:
-; CHECK-NEXT: [[limit_check_outer:[^ ]+]] = icmp sle i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check_outer:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond_outer:[^ ]+]] = and i1 [[first_iteration_check_outer]], [[limit_check_outer]]
-; CHECK-NEXT: br label %outer.loop
   br label %outer.loop
 
 outer.loop:
-; CHECK: outer.loop:
   %outer.loop.acc = phi i32 [ %outer.loop.acc.next, %outer.loop.inc ], [ 0, %outer.loop.preheader ]
   %i = phi i32 [ %i.next, %outer.loop.inc ], [ 0, %outer.loop.preheader ]
   %tmp6 = icmp sle i32 %l, 0
   br i1 %tmp6, label %outer.loop.inc, label %inner.loop.preheader
-  
+
 inner.loop.preheader:
-; CHECK: inner.loop.preheader:
-; CHECK: [[limit_check_inner:[^ ]+]] = icmp sle i32 %l, %length
-; CHECK: br label %inner.loop
   br label %inner.loop
 
 inner.loop:
-; CHECK: inner.loop:
-; CHECK: [[wide_cond:[^ ]+]] = and i1 [[limit_check_inner]], [[wide_cond_outer]]
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
   %inner.loop.acc = phi i32 [ %inner.loop.acc.next, %inner.loop ], [ %outer.loop.acc, %inner.loop.preheader ]
   %j = phi i32 [ %j.next, %inner.loop ], [ %i, %inner.loop.preheader ]
 
   %within.bounds = icmp ult i32 %j, %length
   call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
-  
+
   %j.i64 = zext i32 %j to i64
   %array.j.ptr = getelementptr inbounds i32, i32* %array, i64 %j.i64
   %array.j = load i32, i32* %array.j.ptr, align 4
@@ -166,7 +271,47 @@ exit:
 }
 
 define i32 @cant_expand_guard_check_start(i32* %array, i32 %length, i32 %n, i32 %l, i32 %maybezero) {
-; CHECK-LABEL: @cant_expand_guard_check_start
+; CHECK-LABEL: @cant_expand_guard_check_start(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp sle i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[OUTER_LOOP_PREHEADER:%.*]]
+; CHECK:       outer.loop.preheader:
+; CHECK-NEXT:    br label [[OUTER_LOOP:%.*]]
+; CHECK:       outer.loop:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT:%.*]], [[OUTER_LOOP_INC:%.*]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[OUTER_LOOP_INC]] ], [ 0, [[OUTER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp sle i32 [[L:%.*]], 0
+; CHECK-NEXT:    [[DIV:%.*]] = udiv i32 [[I]], [[MAYBEZERO:%.*]]
+; CHECK-NEXT:    br i1 [[TMP6]], label [[OUTER_LOOP_INC]], label [[INNER_LOOP_PREHEADER:%.*]]
+; CHECK:       inner.loop.preheader:
+; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
+; CHECK:       inner.loop:
+; CHECK-NEXT:    [[INNER_LOOP_ACC:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT:%.*]], [[INNER_LOOP]] ], [ [[OUTER_LOOP_ACC]], [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[INNER_LOOP]] ], [ [[DIV]], [[INNER_LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[J]], [[LENGTH:%.*]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[J_I64:%.*]] = zext i32 [[J]] to i64
+; CHECK-NEXT:    [[ARRAY_J_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[J_I64]]
+; CHECK-NEXT:    [[ARRAY_J:%.*]] = load i32, i32* [[ARRAY_J_PTR]], align 4
+; CHECK-NEXT:    [[INNER_LOOP_ACC_NEXT]] = add i32 [[INNER_LOOP_ACC]], [[ARRAY_J]]
+; CHECK-NEXT:    [[J_NEXT]] = add nsw i32 [[J]], 1
+; CHECK-NEXT:    [[INNER_CONTINUE:%.*]] = icmp slt i32 [[J_NEXT]], [[L]]
+; CHECK-NEXT:    br i1 [[INNER_CONTINUE]], label [[INNER_LOOP]], label [[OUTER_LOOP_INC_LOOPEXIT:%.*]]
+; CHECK:       outer.loop.inc.loopexit:
+; CHECK-NEXT:    [[INNER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[INNER_LOOP_ACC_NEXT]], [[INNER_LOOP]] ]
+; CHECK-NEXT:    br label [[OUTER_LOOP_INC]]
+; CHECK:       outer.loop.inc:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT]] = phi i32 [ [[OUTER_LOOP_ACC]], [[OUTER_LOOP]] ], [ [[INNER_LOOP_ACC_NEXT_LCSSA]], [[OUTER_LOOP_INC_LOOPEXIT]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[OUTER_CONTINUE:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[OUTER_CONTINUE]], label [[OUTER_LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[OUTER_LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[OUTER_LOOP_ACC_NEXT]], [[OUTER_LOOP_INC]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[OUTER_LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
 entry:
   %tmp5 = icmp sle i32 %n, 0
   br i1 %tmp5, label %exit, label %outer.loop.preheader
@@ -180,22 +325,17 @@ outer.loop:
   %tmp6 = icmp sle i32 %l, 0
   %div = udiv i32 %i, %maybezero
   br i1 %tmp6, label %outer.loop.inc, label %inner.loop.preheader
-  
+
 inner.loop.preheader:
-; CHECK: inner.loop.preheader:
-; CHECK: br label %inner.loop
   br label %inner.loop
 
 inner.loop:
-; CHECK: inner.loop:
-; CHECK: %within.bounds = icmp ult i32 %j, %length
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
   %inner.loop.acc = phi i32 [ %inner.loop.acc.next, %inner.loop ], [ %outer.loop.acc, %inner.loop.preheader ]
   %j = phi i32 [ %j.next, %inner.loop ], [ %div, %inner.loop.preheader ]
 
   %within.bounds = icmp ult i32 %j, %length
   call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
-  
+
   %j.i64 = zext i32 %j to i64
   %array.j.ptr = getelementptr inbounds i32, i32* %array, i64 %j.i64
   %array.j = load i32, i32* %array.j.ptr, align 4
@@ -214,4 +354,4 @@ outer.loop.inc:
 exit:
   %result = phi i32 [ 0, %entry ], [ %outer.loop.acc.next, %outer.loop.inc ]
   ret i32 %result
-}
\ No newline at end of file
+}

diff --git a/test/Transforms/LoopPredication/profitability.ll b/test/Transforms/LoopPredication/profitability.ll
index ce01d3c185360071aaaa91ad574efdd84e9300b7..8fb2263e8e92923fd23475e92dc333b18a832df6 100644 (file)
--- a/test/Transforms/LoopPredication/profitability.ll
+++ b/test/Transforms/LoopPredication/profitability.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -loop-predication -loop-predication-skip-profitability-checks=false < %s 2>&1 | FileCheck %s
 ; RUN: opt -S -loop-predication-skip-profitability-checks=false -passes='require<scalar-evolution>,require<branch-prob>,loop(loop-predication)' < %s 2>&1 | FileCheck %s
 
@@ -7,15 +8,35 @@
 ; LatchExitProbability: 0x04000000 / 0x80000000 = 3.12%
 ; ExitingBlockProbability: 0x7ffa572a / 0x80000000 = 99.98%
 define i64 @donot_predicate(i64* nocapture readonly %arg, i32 %length, i64* nocapture readonly %arg2, i64* nocapture readonly %n_addr, i64 %i) {
-; CHECK-LABEL: donot_predicate(
+; CHECK-LABEL: @donot_predicate(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[LENGTH_EXT:%.*]] = zext i32 [[LENGTH:%.*]] to i64
+; CHECK-NEXT:    [[N_PRE:%.*]] = load i64, i64* [[N_ADDR:%.*]], align 4
+; CHECK-NEXT:    br label [[HEADER:%.*]]
+; CHECK:       Header:
+; CHECK-NEXT:    [[RESULT_IN3:%.*]] = phi i64* [ [[ARG2:%.*]], [[ENTRY:%.*]] ], [ [[ARG:%.*]], [[LATCH:%.*]] ]
+; CHECK-NEXT:    [[J2:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[J_NEXT:%.*]], [[LATCH]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i64 [[J2]], [[LENGTH_EXT]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[INNERCMP:%.*]] = icmp eq i64 [[J2]], [[N_PRE]]
+; CHECK-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J2]], 1
+; CHECK-NEXT:    br i1 [[INNERCMP]], label [[LATCH]], label [[EXIT:%.*]], !prof !0
+; CHECK:       Latch:
+; CHECK-NEXT:    [[SPECULATE_TRIP_COUNT:%.*]] = icmp ult i64 [[J_NEXT]], 1048576
+; CHECK-NEXT:    br i1 [[SPECULATE_TRIP_COUNT]], label [[HEADER]], label [[DEOPT:%.*]]
+; CHECK:       deopt:
+; CHECK-NEXT:    [[COUNTED_SPECULATION_FAILED:%.*]] = call i64 (...) @llvm.experimental.deoptimize.i64(i64 30) [ "deopt"(i32 0) ]
+; CHECK-NEXT:    ret i64 [[COUNTED_SPECULATION_FAILED]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT_IN3_LCSSA:%.*]] = phi i64* [ [[RESULT_IN3]], [[HEADER]] ]
+; CHECK-NEXT:    [[RESULT_LE:%.*]] = load i64, i64* [[RESULT_IN3_LCSSA]], align 8
+; CHECK-NEXT:    ret i64 [[RESULT_LE]]
+;
 entry:
   %length.ext = zext i32 %length to i64
   %n.pre = load i64, i64* %n_addr, align 4
   br label %Header
 
-; CHECK-LABEL: Header:
-; CHECK:         %within.bounds = icmp ult i64 %j2, %length.ext
-; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9)
 Header:                                          ; preds = %entry, %Latch
   %result.in3 = phi i64* [ %arg2, %entry ], [ %arg, %Latch ]
   %j2 = phi i64 [ 0, %entry ], [ %j.next, %Latch ]
@@ -43,18 +64,37 @@ exit:                                             ; preds = %Header
 ; predicate loop since there's no profile information and BPI concluded all
 ; exiting blocks have same probability of exiting from loop.
 define i64 @predicate(i64* nocapture readonly %arg, i32 %length, i64* nocapture readonly %arg2, i64* nocapture readonly %n_addr, i64 %i) {
-; CHECK-LABEL: predicate(
-; CHECK-LABEL: entry:
-; CHECK:           [[limit_check:[^ ]+]] = icmp ule i64 1048576, %length.ext
-; CHECK-NEXT:      [[first_iteration_check:[^ ]+]] = icmp ult i64 0, %length.ext
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
+; CHECK-LABEL: @predicate(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[LENGTH_EXT:%.*]] = zext i32 [[LENGTH:%.*]] to i64
+; CHECK-NEXT:    [[N_PRE:%.*]] = load i64, i64* [[N_ADDR:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp ule i64 1048576, [[LENGTH_EXT]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i64 0, [[LENGTH_EXT]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    br label [[HEADER:%.*]]
+; CHECK:       Header:
+; CHECK-NEXT:    [[RESULT_IN3:%.*]] = phi i64* [ [[ARG2:%.*]], [[ENTRY:%.*]] ], [ [[ARG:%.*]], [[LATCH:%.*]] ]
+; CHECK-NEXT:    [[J2:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[J_NEXT:%.*]], [[LATCH]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i64 [[J2]], [[LENGTH_EXT]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP2]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[INNERCMP:%.*]] = icmp eq i64 [[J2]], [[N_PRE]]
+; CHECK-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J2]], 1
+; CHECK-NEXT:    br i1 [[INNERCMP]], label [[LATCH]], label [[EXIT:%.*]]
+; CHECK:       Latch:
+; CHECK-NEXT:    [[SPECULATE_TRIP_COUNT:%.*]] = icmp ult i64 [[J_NEXT]], 1048576
+; CHECK-NEXT:    br i1 [[SPECULATE_TRIP_COUNT]], label [[HEADER]], label [[EXITLATCH:%.*]]
+; CHECK:       exitLatch:
+; CHECK-NEXT:    ret i64 1
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT_IN3_LCSSA:%.*]] = phi i64* [ [[RESULT_IN3]], [[HEADER]] ]
+; CHECK-NEXT:    [[RESULT_LE:%.*]] = load i64, i64* [[RESULT_IN3_LCSSA]], align 8
+; CHECK-NEXT:    ret i64 [[RESULT_LE]]
+;
 entry:
   %length.ext = zext i32 %length to i64
   %n.pre = load i64, i64* %n_addr, align 4
   br label %Header
 
-; CHECK-LABEL: Header:
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 Header:                                          ; preds = %entry, %Latch
   %result.in3 = phi i64* [ %arg2, %entry ], [ %arg, %Latch ]
   %j2 = phi i64 [ 0, %entry ], [ %j.next, %Latch ]
@@ -82,16 +122,34 @@ exit:                                             ; preds = %Header
 ; LatchExitProbability: 0x000020e1 / 0x80000000 = 0.00%
 ; ExitingBlockProbability: 0x7ffcbb86 / 0x80000000 = 99.99%
 define i64 @donot_predicate_prof(i64* nocapture readonly %arg, i32 %length, i64* nocapture readonly %arg2, i64* nocapture readonly %n_addr, i64 %i) {
-; CHECK-LABEL: donot_predicate_prof(
-; CHECK-LABEL: entry:
+; CHECK-LABEL: @donot_predicate_prof(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[LENGTH_EXT:%.*]] = zext i32 [[LENGTH:%.*]] to i64
+; CHECK-NEXT:    [[N_PRE:%.*]] = load i64, i64* [[N_ADDR:%.*]], align 4
+; CHECK-NEXT:    br label [[HEADER:%.*]]
+; CHECK:       Header:
+; CHECK-NEXT:    [[RESULT_IN3:%.*]] = phi i64* [ [[ARG2:%.*]], [[ENTRY:%.*]] ], [ [[ARG:%.*]], [[LATCH:%.*]] ]
+; CHECK-NEXT:    [[J2:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[J_NEXT:%.*]], [[LATCH]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i64 [[J2]], [[LENGTH_EXT]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[INNERCMP:%.*]] = icmp eq i64 [[J2]], [[N_PRE]]
+; CHECK-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J2]], 1
+; CHECK-NEXT:    br i1 [[INNERCMP]], label [[LATCH]], label [[EXIT:%.*]], !prof !1
+; CHECK:       Latch:
+; CHECK-NEXT:    [[SPECULATE_TRIP_COUNT:%.*]] = icmp ult i64 [[J_NEXT]], 1048576
+; CHECK-NEXT:    br i1 [[SPECULATE_TRIP_COUNT]], label [[HEADER]], label [[EXITLATCH:%.*]], !prof !2
+; CHECK:       exitLatch:
+; CHECK-NEXT:    ret i64 1
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT_IN3_LCSSA:%.*]] = phi i64* [ [[RESULT_IN3]], [[HEADER]] ]
+; CHECK-NEXT:    [[RESULT_LE:%.*]] = load i64, i64* [[RESULT_IN3_LCSSA]], align 8
+; CHECK-NEXT:    ret i64 [[RESULT_LE]]
+;
 entry:
   %length.ext = zext i32 %length to i64
   %n.pre = load i64, i64* %n_addr, align 4
   br label %Header
 
-; CHECK-LABEL: Header:
-; CHECK:         %within.bounds = icmp ult i64 %j2, %length.ext
-; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9)
 Header:                                          ; preds = %entry, %Latch
   %result.in3 = phi i64* [ %arg2, %entry ], [ %arg, %Latch ]
   %j2 = phi i64 [ 0, %entry ], [ %j.next, %Latch ]

diff --git a/test/Transforms/LoopPredication/reverse.ll b/test/Transforms/LoopPredication/reverse.ll
index 3d695d1b4ff660cfdf2bed5d25f04de3699ac37a..b5b132b048888170b558d89d194d2d53721084a8 100644 (file)
--- a/test/Transforms/LoopPredication/reverse.ll
+++ b/test/Transforms/LoopPredication/reverse.ll
@@ -1,3 +1,4 @@
+; 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
 
@@ -5,20 +6,41 @@ 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:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_NEXT]], [[LENGTH]]
+; 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
 
-; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[range_start:%.*]] = add i32 %n, -1
-; CHECK-NEXT:    [[first_iteration_check:%.*]] = icmp ult i32 [[range_start]], %length
-; CHECK-NEXT:    [[no_wrap_check:%.*]] = icmp sge i32 %lowerlimit, 1
-; CHECK-NEXT:    [[wide_cond:%.*]] = and i1 [[first_iteration_check]], [[no_wrap_check]]
 loop.preheader:
   br label %loop
 
-; CHECK: loop:
-; CHECK:    call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 loop:
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
@@ -39,20 +61,41 @@ exit:
 
 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:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_NEXT]], [[LENGTH]]
+; 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
 
-; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[range_start:%.*]] = add i32 %n, -1
-; CHECK-NEXT:    [[first_iteration_check:%.*]] = icmp ult i32 [[range_start]], %length
-; CHECK-NEXT:    [[no_wrap_check:%.*]] = icmp uge i32 %lowerlimit, 1
-; CHECK-NEXT:    [[wide_cond:%.*]] = and i1 [[first_iteration_check]], [[no_wrap_check]]
 loop.preheader:
   br label %loop
 
-; CHECK: loop:
-; CHECK:    call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 loop:
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
@@ -76,19 +119,40 @@ exit:
 ; 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:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_NEXT]], [[LENGTH]]
+; 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
 
-; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[range_start:%.*]] = add i32 %n, -1
-; CHECK-NEXT:    [[first_iteration_check:%.*]] = icmp ult i32 [[range_start]], %length
-; CHECK-NEXT:    [[wide_cond:%.*]] = and i1 [[first_iteration_check]], false
 loop.preheader:
   br label %loop
 
-; CHECK: loop:
-; CHECK:    call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 loop:
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
@@ -110,6 +174,32 @@ exit:
 ; 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
@@ -117,8 +207,6 @@ entry:
 loop.preheader:
   br label %loop
 
-; CHECK: loop:
-; CHECK: llvm.experimental.guard(i1 %within.bounds, i32 9)
 loop:
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
@@ -141,20 +229,41 @@ exit:
 
 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:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_NEXT]], [[LENGTH]]
+; 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
 
-; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[range_start:%.*]] = add i32 %n, -1
-; CHECK-NEXT:    [[first_iteration_check:%.*]] = icmp ult i32 [[range_start]], %length
-; CHECK-NEXT:    [[no_wrap_check:%.*]] = icmp sgt i32 %lowerlimit, 1
-; CHECK-NEXT:    [[wide_cond:%.*]] = and i1 [[first_iteration_check]], [[no_wrap_check]]
 loop.preheader:
   br label %loop
 
-; CHECK: loop:
-; CHECK:    call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 loop:
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
@@ -175,20 +284,41 @@ exit:
 
 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:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_NEXT]], [[LENGTH]]
+; 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
 
-; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[range_start:%.*]] = add i32 %n, -1
-; CHECK-NEXT:    [[first_iteration_check:%.*]] = icmp ult i32 [[range_start]], %length
-; CHECK-NEXT:    [[no_wrap_check:%.*]] = icmp ugt i32 %lowerlimit, 1
-; CHECK-NEXT:    [[wide_cond:%.*]] = and i1 [[first_iteration_check]], [[no_wrap_check]]
 loop.preheader:
   br label %loop
 
-; CHECK: loop:
-; CHECK:    call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 loop:
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]
@@ -212,19 +342,40 @@ exit:
 ; 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:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I_NEXT]], [[LENGTH]]
+; 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
 
-; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[range_start:%.*]] = add i32 %n, -1
-; CHECK-NEXT:    [[first_iteration_check:%.*]] = icmp ult i32 [[range_start]], %length
-; CHECK-NEXT:    [[wide_cond:%.*]] = and i1 [[first_iteration_check]], false
 loop.preheader:
   br label %loop
 
-; CHECK: loop:
-; CHECK:    call void (i1, ...) @llvm.experimental.guard(i1 [[wide_cond]], i32 9) [ "deopt"() ]
 loop:
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ %n, %loop.preheader ]

diff --git a/test/Transforms/LoopPredication/visited.ll b/test/Transforms/LoopPredication/visited.ll
index 8c3d54773ec0a9bb51c95025ae13c39fe8bc2e90..b4b291fe59ac65bb62a163b16846861019a297ba 100644 (file)
--- a/test/Transforms/LoopPredication/visited.ll
+++ b/test/Transforms/LoopPredication/visited.ll
@@ -1,27 +1,146 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -loop-predication < %s 2>&1 | FileCheck %s
 ; RUN: opt -S -passes='require<scalar-evolution>,loop(loop-predication)' < %s 2>&1 | FileCheck %s
 
 declare void @llvm.experimental.guard(i1, ...)
 
 define i32 @test_visited(i32* %array, i32 %length, i32 %n, i32 %x) {
-; CHECK-LABEL: @test_visited
+; CHECK-LABEL: @test_visited(
+; 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:%.*]] = icmp ule i32 [[N]], [[LENGTH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 0, [[LENGTH]]
+; CHECK-NEXT:    [[TMP2:%.*]] = and i1 [[TMP1]], [[TMP0]]
+; 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]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH]]
+; CHECK-NEXT:    [[UNRELATED_COND:%.*]] = icmp eq i32 [[X:%.*]], [[I]]
+; CHECK-NEXT:    [[GUARD_COND_2:%.*]] = and i1 [[WITHIN_BOUNDS]], [[UNRELATED_COND]]
+; CHECK-NEXT:    [[GUARD_COND_3:%.*]] = and i1 [[GUARD_COND_2]], [[UNRELATED_COND]]
+; CHECK-NEXT:    [[GUARD_COND_4:%.*]] = and i1 [[GUARD_COND_3]], [[GUARD_COND_2]]
+; CHECK-NEXT:    [[GUARD_COND_5:%.*]] = and i1 [[GUARD_COND_4]], [[GUARD_COND_3]]
+; CHECK-NEXT:    [[GUARD_COND_6:%.*]] = and i1 [[GUARD_COND_5]], [[GUARD_COND_4]]
+; CHECK-NEXT:    [[GUARD_COND_7:%.*]] = and i1 [[GUARD_COND_6]], [[GUARD_COND_5]]
+; CHECK-NEXT:    [[GUARD_COND_8:%.*]] = and i1 [[GUARD_COND_7]], [[GUARD_COND_6]]
+; CHECK-NEXT:    [[GUARD_COND_9:%.*]] = and i1 [[GUARD_COND_8]], [[GUARD_COND_7]]
+; CHECK-NEXT:    [[GUARD_COND_10:%.*]] = and i1 [[GUARD_COND_9]], [[GUARD_COND_8]]
+; CHECK-NEXT:    [[GUARD_COND_11:%.*]] = and i1 [[GUARD_COND_10]], [[GUARD_COND_9]]
+; CHECK-NEXT:    [[GUARD_COND_12:%.*]] = and i1 [[GUARD_COND_11]], [[GUARD_COND_10]]
+; CHECK-NEXT:    [[GUARD_COND_13:%.*]] = and i1 [[GUARD_COND_12]], [[GUARD_COND_11]]
+; CHECK-NEXT:    [[GUARD_COND_14:%.*]] = and i1 [[GUARD_COND_13]], [[GUARD_COND_12]]
+; CHECK-NEXT:    [[GUARD_COND_15:%.*]] = and i1 [[GUARD_COND_14]], [[GUARD_COND_13]]
+; CHECK-NEXT:    [[GUARD_COND_16:%.*]] = and i1 [[GUARD_COND_15]], [[GUARD_COND_14]]
+; CHECK-NEXT:    [[GUARD_COND_17:%.*]] = and i1 [[GUARD_COND_16]], [[GUARD_COND_15]]
+; CHECK-NEXT:    [[GUARD_COND_18:%.*]] = and i1 [[GUARD_COND_17]], [[GUARD_COND_16]]
+; CHECK-NEXT:    [[GUARD_COND_19:%.*]] = and i1 [[GUARD_COND_18]], [[GUARD_COND_17]]
+; CHECK-NEXT:    [[GUARD_COND_20:%.*]] = and i1 [[GUARD_COND_19]], [[GUARD_COND_18]]
+; CHECK-NEXT:    [[GUARD_COND_21:%.*]] = and i1 [[GUARD_COND_20]], [[GUARD_COND_19]]
+; CHECK-NEXT:    [[GUARD_COND_22:%.*]] = and i1 [[GUARD_COND_21]], [[GUARD_COND_20]]
+; CHECK-NEXT:    [[GUARD_COND_23:%.*]] = and i1 [[GUARD_COND_22]], [[GUARD_COND_21]]
+; CHECK-NEXT:    [[GUARD_COND_24:%.*]] = and i1 [[GUARD_COND_23]], [[GUARD_COND_22]]
+; CHECK-NEXT:    [[GUARD_COND_25:%.*]] = and i1 [[GUARD_COND_24]], [[GUARD_COND_23]]
+; CHECK-NEXT:    [[GUARD_COND_26:%.*]] = and i1 [[GUARD_COND_25]], [[GUARD_COND_24]]
+; CHECK-NEXT:    [[GUARD_COND_27:%.*]] = and i1 [[GUARD_COND_26]], [[GUARD_COND_25]]
+; CHECK-NEXT:    [[GUARD_COND_28:%.*]] = and i1 [[GUARD_COND_27]], [[GUARD_COND_26]]
+; CHECK-NEXT:    [[GUARD_COND_29:%.*]] = and i1 [[GUARD_COND_28]], [[GUARD_COND_27]]
+; CHECK-NEXT:    [[GUARD_COND_30:%.*]] = and i1 [[GUARD_COND_29]], [[GUARD_COND_28]]
+; CHECK-NEXT:    [[GUARD_COND_31:%.*]] = and i1 [[GUARD_COND_30]], [[GUARD_COND_29]]
+; CHECK-NEXT:    [[GUARD_COND_32:%.*]] = and i1 [[GUARD_COND_31]], [[GUARD_COND_30]]
+; CHECK-NEXT:    [[GUARD_COND_33:%.*]] = and i1 [[GUARD_COND_32]], [[GUARD_COND_31]]
+; CHECK-NEXT:    [[GUARD_COND_34:%.*]] = and i1 [[GUARD_COND_33]], [[GUARD_COND_32]]
+; CHECK-NEXT:    [[GUARD_COND_35:%.*]] = and i1 [[GUARD_COND_34]], [[GUARD_COND_33]]
+; CHECK-NEXT:    [[GUARD_COND_36:%.*]] = and i1 [[GUARD_COND_35]], [[GUARD_COND_34]]
+; CHECK-NEXT:    [[GUARD_COND_37:%.*]] = and i1 [[GUARD_COND_36]], [[GUARD_COND_35]]
+; CHECK-NEXT:    [[GUARD_COND_38:%.*]] = and i1 [[GUARD_COND_37]], [[GUARD_COND_36]]
+; CHECK-NEXT:    [[GUARD_COND_39:%.*]] = and i1 [[GUARD_COND_38]], [[GUARD_COND_37]]
+; CHECK-NEXT:    [[GUARD_COND_40:%.*]] = and i1 [[GUARD_COND_39]], [[GUARD_COND_38]]
+; CHECK-NEXT:    [[GUARD_COND_41:%.*]] = and i1 [[GUARD_COND_40]], [[GUARD_COND_39]]
+; CHECK-NEXT:    [[GUARD_COND_42:%.*]] = and i1 [[GUARD_COND_41]], [[GUARD_COND_40]]
+; CHECK-NEXT:    [[GUARD_COND_43:%.*]] = and i1 [[GUARD_COND_42]], [[GUARD_COND_41]]
+; CHECK-NEXT:    [[GUARD_COND_44:%.*]] = and i1 [[GUARD_COND_43]], [[GUARD_COND_42]]
+; CHECK-NEXT:    [[GUARD_COND_45:%.*]] = and i1 [[GUARD_COND_44]], [[GUARD_COND_43]]
+; CHECK-NEXT:    [[GUARD_COND_46:%.*]] = and i1 [[GUARD_COND_45]], [[GUARD_COND_44]]
+; CHECK-NEXT:    [[GUARD_COND_47:%.*]] = and i1 [[GUARD_COND_46]], [[GUARD_COND_45]]
+; CHECK-NEXT:    [[GUARD_COND_48:%.*]] = and i1 [[GUARD_COND_47]], [[GUARD_COND_46]]
+; CHECK-NEXT:    [[GUARD_COND_49:%.*]] = and i1 [[GUARD_COND_48]], [[GUARD_COND_47]]
+; CHECK-NEXT:    [[GUARD_COND_50:%.*]] = and i1 [[GUARD_COND_49]], [[GUARD_COND_48]]
+; CHECK-NEXT:    [[GUARD_COND_51:%.*]] = and i1 [[GUARD_COND_50]], [[GUARD_COND_49]]
+; CHECK-NEXT:    [[GUARD_COND_52:%.*]] = and i1 [[GUARD_COND_51]], [[GUARD_COND_50]]
+; CHECK-NEXT:    [[GUARD_COND_53:%.*]] = and i1 [[GUARD_COND_52]], [[GUARD_COND_51]]
+; CHECK-NEXT:    [[GUARD_COND_54:%.*]] = and i1 [[GUARD_COND_53]], [[GUARD_COND_52]]
+; CHECK-NEXT:    [[GUARD_COND_55:%.*]] = and i1 [[GUARD_COND_54]], [[GUARD_COND_53]]
+; CHECK-NEXT:    [[GUARD_COND_56:%.*]] = and i1 [[GUARD_COND_55]], [[GUARD_COND_54]]
+; CHECK-NEXT:    [[GUARD_COND_57:%.*]] = and i1 [[GUARD_COND_56]], [[GUARD_COND_55]]
+; CHECK-NEXT:    [[GUARD_COND_58:%.*]] = and i1 [[GUARD_COND_57]], [[GUARD_COND_56]]
+; CHECK-NEXT:    [[GUARD_COND_59:%.*]] = and i1 [[GUARD_COND_58]], [[GUARD_COND_57]]
+; CHECK-NEXT:    [[GUARD_COND_60:%.*]] = and i1 [[GUARD_COND_59]], [[GUARD_COND_58]]
+; CHECK-NEXT:    [[GUARD_COND_61:%.*]] = and i1 [[GUARD_COND_60]], [[GUARD_COND_59]]
+; CHECK-NEXT:    [[GUARD_COND_62:%.*]] = and i1 [[GUARD_COND_61]], [[GUARD_COND_60]]
+; CHECK-NEXT:    [[GUARD_COND_63:%.*]] = and i1 [[GUARD_COND_62]], [[GUARD_COND_61]]
+; CHECK-NEXT:    [[GUARD_COND_64:%.*]] = and i1 [[GUARD_COND_63]], [[GUARD_COND_62]]
+; CHECK-NEXT:    [[GUARD_COND_65:%.*]] = and i1 [[GUARD_COND_64]], [[GUARD_COND_63]]
+; CHECK-NEXT:    [[GUARD_COND_66:%.*]] = and i1 [[GUARD_COND_65]], [[GUARD_COND_64]]
+; CHECK-NEXT:    [[GUARD_COND_67:%.*]] = and i1 [[GUARD_COND_66]], [[GUARD_COND_65]]
+; CHECK-NEXT:    [[GUARD_COND_68:%.*]] = and i1 [[GUARD_COND_67]], [[GUARD_COND_66]]
+; CHECK-NEXT:    [[GUARD_COND_69:%.*]] = and i1 [[GUARD_COND_68]], [[GUARD_COND_67]]
+; CHECK-NEXT:    [[GUARD_COND_70:%.*]] = and i1 [[GUARD_COND_69]], [[GUARD_COND_68]]
+; CHECK-NEXT:    [[GUARD_COND_71:%.*]] = and i1 [[GUARD_COND_70]], [[GUARD_COND_69]]
+; CHECK-NEXT:    [[GUARD_COND_72:%.*]] = and i1 [[GUARD_COND_71]], [[GUARD_COND_70]]
+; CHECK-NEXT:    [[GUARD_COND_73:%.*]] = and i1 [[GUARD_COND_72]], [[GUARD_COND_71]]
+; CHECK-NEXT:    [[GUARD_COND_74:%.*]] = and i1 [[GUARD_COND_73]], [[GUARD_COND_72]]
+; CHECK-NEXT:    [[GUARD_COND_75:%.*]] = and i1 [[GUARD_COND_74]], [[GUARD_COND_73]]
+; CHECK-NEXT:    [[GUARD_COND_76:%.*]] = and i1 [[GUARD_COND_75]], [[GUARD_COND_74]]
+; CHECK-NEXT:    [[GUARD_COND_77:%.*]] = and i1 [[GUARD_COND_76]], [[GUARD_COND_75]]
+; CHECK-NEXT:    [[GUARD_COND_78:%.*]] = and i1 [[GUARD_COND_77]], [[GUARD_COND_76]]
+; CHECK-NEXT:    [[GUARD_COND_79:%.*]] = and i1 [[GUARD_COND_78]], [[GUARD_COND_77]]
+; CHECK-NEXT:    [[GUARD_COND_80:%.*]] = and i1 [[GUARD_COND_79]], [[GUARD_COND_78]]
+; CHECK-NEXT:    [[GUARD_COND_81:%.*]] = and i1 [[GUARD_COND_80]], [[GUARD_COND_79]]
+; CHECK-NEXT:    [[GUARD_COND_82:%.*]] = and i1 [[GUARD_COND_81]], [[GUARD_COND_80]]
+; CHECK-NEXT:    [[GUARD_COND_83:%.*]] = and i1 [[GUARD_COND_82]], [[GUARD_COND_81]]
+; CHECK-NEXT:    [[GUARD_COND_84:%.*]] = and i1 [[GUARD_COND_83]], [[GUARD_COND_82]]
+; CHECK-NEXT:    [[GUARD_COND_85:%.*]] = and i1 [[GUARD_COND_84]], [[GUARD_COND_83]]
+; CHECK-NEXT:    [[GUARD_COND_86:%.*]] = and i1 [[GUARD_COND_85]], [[GUARD_COND_84]]
+; CHECK-NEXT:    [[GUARD_COND_87:%.*]] = and i1 [[GUARD_COND_86]], [[GUARD_COND_85]]
+; CHECK-NEXT:    [[GUARD_COND_88:%.*]] = and i1 [[GUARD_COND_87]], [[GUARD_COND_86]]
+; CHECK-NEXT:    [[GUARD_COND_89:%.*]] = and i1 [[GUARD_COND_88]], [[GUARD_COND_87]]
+; CHECK-NEXT:    [[GUARD_COND_90:%.*]] = and i1 [[GUARD_COND_89]], [[GUARD_COND_88]]
+; CHECK-NEXT:    [[GUARD_COND_91:%.*]] = and i1 [[GUARD_COND_90]], [[GUARD_COND_89]]
+; CHECK-NEXT:    [[GUARD_COND_92:%.*]] = and i1 [[GUARD_COND_91]], [[GUARD_COND_90]]
+; CHECK-NEXT:    [[GUARD_COND_93:%.*]] = and i1 [[GUARD_COND_92]], [[GUARD_COND_91]]
+; CHECK-NEXT:    [[GUARD_COND_94:%.*]] = and i1 [[GUARD_COND_93]], [[GUARD_COND_92]]
+; CHECK-NEXT:    [[GUARD_COND_95:%.*]] = and i1 [[GUARD_COND_94]], [[GUARD_COND_93]]
+; CHECK-NEXT:    [[GUARD_COND_96:%.*]] = and i1 [[GUARD_COND_95]], [[GUARD_COND_94]]
+; CHECK-NEXT:    [[GUARD_COND_97:%.*]] = and i1 [[GUARD_COND_96]], [[GUARD_COND_95]]
+; CHECK-NEXT:    [[GUARD_COND_98:%.*]] = and i1 [[GUARD_COND_97]], [[GUARD_COND_96]]
+; CHECK-NEXT:    [[GUARD_COND_99:%.*]] = and i1 [[GUARD_COND_98]], [[GUARD_COND_97]]
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[UNRELATED_COND]], [[TMP2]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP3]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] 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:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; 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:
-; CHECK: loop.preheader:
-; CHECK: [[limit_check:[^ ]+]] = icmp ule i32 %n, %length
-; CHECK-NEXT: [[first_iteration_check:[^ ]+]] = icmp ult i32 0, %length
-; CHECK-NEXT: [[wide_cond:[^ ]+]] = and i1 [[first_iteration_check]], [[limit_check]]
-; CHECK-NEXT: br label %loop
   br label %loop
 
 loop:
-; CHECK: loop:
-; CHECK: %unrelated.cond = icmp eq i32 %x, %i
-; CHECK: [[guard_cond:[^ ]+]] = and i1 %unrelated.cond, [[wide_cond]]
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[guard_cond]], i32 9) [ "deopt"() ]
   %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
   %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
   %within.bounds = icmp ult i32 %i, %length
@@ -138,4 +257,4 @@ loop:
 exit:
   %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
   ret i32 %result
-}
\ No newline at end of file
+}

diff --git a/test/Transforms/LoopPredication/widened.ll b/test/Transforms/LoopPredication/widened.ll
index 33c4e27061333cff9386b190e53ce19ab4ea9753..374c74c1af582bc512ed7ed4066ed3d78299d097 100644 (file)
--- a/test/Transforms/LoopPredication/widened.ll
+++ b/test/Transforms/LoopPredication/widened.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -loop-predication -loop-predication-enable-iv-truncation=true < %s 2>&1 | FileCheck %s
 declare void @llvm.experimental.guard(i1, ...)
 
@@ -7,25 +8,49 @@ declare i16 @short_length(i8*)
 ; Consider range check of type i16 and i32, while IV is of type i64
 ; We can loop predicate this because the IV range is within i16 and within i32.
 define i64 @iv_wider_type_rc_two_narrow_types(i32 %offA, i16 %offB, i8* %arrA, i8* %arrB) {
-; CHECK-LABEL: iv_wider_type_rc_two_narrow_types
+; CHECK-LABEL: @iv_wider_type_rc_two_narrow_types(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[LENGTHA:%.*]] = call i32 @length(i8* [[ARRA:%.*]])
+; CHECK-NEXT:    [[LENGTHB:%.*]] = call i16 @short_length(i8* [[ARRB:%.*]])
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i16 [[LENGTHB]], [[OFFB:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule i16 16, [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult i16 [[OFFB]], [[LENGTHB]]
+; CHECK-NEXT:    [[TMP3:%.*]] = and i1 [[TMP2]], [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = sub i32 [[LENGTHA]], [[OFFA:%.*]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp ule i32 16, [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp ult i32 [[OFFA]], [[LENGTHA]]
+; CHECK-NEXT:    [[TMP7:%.*]] = and i1 [[TMP6]], [[TMP5]]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_TRUNC_32:%.*]] = trunc i64 [[IV]] to i32
+; CHECK-NEXT:    [[IV_TRUNC_16:%.*]] = trunc i64 [[IV]] to i16
+; CHECK-NEXT:    [[INDEXA:%.*]] = add i32 [[IV_TRUNC_32]], [[OFFA]]
+; CHECK-NEXT:    [[INDEXB:%.*]] = add i16 [[IV_TRUNC_16]], [[OFFB]]
+; CHECK-NEXT:    [[RCA:%.*]] = icmp ult i32 [[INDEXA]], [[LENGTHA]]
+; CHECK-NEXT:    [[RCB:%.*]] = icmp ult i16 [[INDEXB]], [[LENGTHB]]
+; CHECK-NEXT:    [[WIDE_CHK:%.*]] = and i1 [[RCA]], [[RCB]]
+; CHECK-NEXT:    [[TMP8:%.*]] = and i1 [[TMP3]], [[TMP7]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP8]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[INDEXA_EXT:%.*]] = zext i32 [[INDEXA]] to i64
+; CHECK-NEXT:    [[ADDRA:%.*]] = getelementptr inbounds i8, i8* [[ARRA]], i64 [[INDEXA_EXT]]
+; CHECK-NEXT:    [[ELTA:%.*]] = load i8, i8* [[ADDRA]]
+; CHECK-NEXT:    [[INDEXB_EXT:%.*]] = zext i16 [[INDEXB]] to i64
+; CHECK-NEXT:    [[ADDRB:%.*]] = getelementptr inbounds i8, i8* [[ARRB]], i64 [[INDEXB_EXT]]
+; CHECK-NEXT:    store i8 [[ELTA]], i8* [[ADDRB]]
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[LATCH_CHECK:%.*]] = icmp ult i64 [[IV_NEXT]], 16
+; CHECK-NEXT:    br i1 [[LATCH_CHECK]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[IV_LCSSA:%.*]] = phi i64 [ [[IV]], [[LOOP]] ]
+; CHECK-NEXT:    ret i64 [[IV_LCSSA]]
+;
 entry:
-; CHECK-LABEL: entry:
-; CHECK: [[idxB:[^ ]+]] = sub i16 %lengthB, %offB
-; CHECK-NEXT: [[limit_checkB:[^ ]+]] = icmp ule i16 16, [[idxB]]
-; CHECK-NEXT: [[first_iteration_checkB:[^ ]+]] = icmp ult i16 %offB, %lengthB
-; CHECK-NEXT: [[WideChkB:[^ ]+]] = and i1 [[first_iteration_checkB]], [[limit_checkB]]
-; CHECK-NEXT: [[idxA:[^ ]+]] = sub i32 %lengthA, %offA
-; CHECK-NEXT: [[limit_checkA:[^ ]+]] = icmp ule i32 16, [[idxA]]
-; CHECK-NEXT: [[first_iteration_checkA:[^ ]+]] = icmp ult i32 %offA, %lengthA
-; CHECK-NEXT: [[WideChkA:[^ ]+]] = and i1 [[first_iteration_checkA]], [[limit_checkA]]
   %lengthA = call i32 @length(i8* %arrA)
   %lengthB = call i16 @short_length(i8* %arrB)
-   br label %loop
+  br label %loop
 
 loop:
-; CHECK-LABEL: loop:
-; CHECK: [[invariant_check:[^ ]+]] = and i1 [[WideChkB]], [[WideChkA]]
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 [[invariant_check]], i32 9)
   %iv = phi i64 [0, %entry ], [ %iv.next, %loop ]
   %iv.trunc.32 = trunc i64 %iv to i32
   %iv.trunc.16 = trunc i64 %iv to i16
@@ -46,40 +71,67 @@ loop:
   br i1 %latch.check, label %loop, label %exit
 
 exit:
- ret i64 %iv
+  ret i64 %iv
 }
 
 
 ; Consider an IV of type long and an array access into int array.
 ; IV is of type i64 while the range check operands are of type i32 and i64.
 define i64 @iv_rc_different_types(i32 %offA, i32 %offB, i8* %arrA, i8* %arrB, i64 %max)
+; CHECK-LABEL: @iv_rc_different_types(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[LENGTHA:%.*]] = call i32 @length(i8* [[ARRA:%.*]])
+; CHECK-NEXT:    [[LENGTHB:%.*]] = call i32 @length(i8* [[ARRB:%.*]])
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LENGTHB]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[OFFB:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ule i32 15, [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult i32 [[OFFB]], [[LENGTHB]]
+; CHECK-NEXT:    [[TMP4:%.*]] = and i1 [[TMP3]], [[TMP2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[MAX:%.*]], -1
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp ule i64 15, [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp ult i64 0, [[MAX]]
+; CHECK-NEXT:    [[TMP8:%.*]] = and i1 [[TMP7]], [[TMP6]]
+; CHECK-NEXT:    [[TMP9:%.*]] = add i32 [[LENGTHA]], -1
+; CHECK-NEXT:    [[TMP10:%.*]] = sub i32 [[TMP9]], [[OFFA:%.*]]
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp ule i32 15, [[TMP10]]
+; CHECK-NEXT:    [[TMP12:%.*]] = icmp ult i32 [[OFFA]], [[LENGTHA]]
+; CHECK-NEXT:    [[TMP13:%.*]] = and i1 [[TMP12]], [[TMP11]]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_TRUNC:%.*]] = trunc i64 [[IV]] to i32
+; CHECK-NEXT:    [[INDEXA:%.*]] = add i32 [[IV_TRUNC]], [[OFFA]]
+; CHECK-NEXT:    [[INDEXB:%.*]] = add i32 [[IV_TRUNC]], [[OFFB]]
+; CHECK-NEXT:    [[RCA:%.*]] = icmp ult i32 [[INDEXA]], [[LENGTHA]]
+; CHECK-NEXT:    [[RCIV:%.*]] = icmp ult i64 [[IV]], [[MAX]]
+; CHECK-NEXT:    [[WIDE_CHK:%.*]] = and i1 [[RCA]], [[RCIV]]
+; CHECK-NEXT:    [[RCB:%.*]] = icmp ult i32 [[INDEXB]], [[LENGTHB]]
+; CHECK-NEXT:    [[WIDE_CHK_FINAL:%.*]] = and i1 [[WIDE_CHK]], [[RCB]]
+; CHECK-NEXT:    [[TMP14:%.*]] = and i1 [[TMP4]], [[TMP8]]
+; CHECK-NEXT:    [[TMP15:%.*]] = and i1 [[TMP14]], [[TMP13]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[TMP15]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[INDEXA_EXT:%.*]] = zext i32 [[INDEXA]] to i64
+; CHECK-NEXT:    [[ADDRA:%.*]] = getelementptr inbounds i8, i8* [[ARRA]], i64 [[INDEXA_EXT]]
+; CHECK-NEXT:    [[ELTA:%.*]] = load i8, i8* [[ADDRA]]
+; CHECK-NEXT:    [[INDEXB_EXT:%.*]] = zext i32 [[INDEXB]] to i64
+; CHECK-NEXT:    [[ADDRB:%.*]] = getelementptr inbounds i8, i8* [[ARRB]], i64 [[INDEXB_EXT]]
+; CHECK-NEXT:    [[ELTB:%.*]] = load i8, i8* [[ADDRB]]
+; CHECK-NEXT:    [[RESULT:%.*]] = xor i8 [[ELTA]], [[ELTB]]
+; CHECK-NEXT:    store i8 [[RESULT]], i8* [[ADDRA]]
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[LATCH_CHECK:%.*]] = icmp ult i64 [[IV]], 15
+; CHECK-NEXT:    br i1 [[LATCH_CHECK]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[IV_LCSSA:%.*]] = phi i64 [ [[IV]], [[LOOP]] ]
+; CHECK-NEXT:    ret i64 [[IV_LCSSA]]
+;
 {
-; CHECK-LABEL: iv_rc_different_types
 entry:
-; CHECK-LABEL: entry:
-; CHECK: [[lenB:[^ ]+]] = add i32 %lengthB, -1
-; CHECK-NEXT: [[idxB:[^ ]+]] = sub i32 [[lenB]], %offB
-; CHECK-NEXT: [[limit_checkB:[^ ]+]] = icmp ule i32 15, [[idxB]]
-; CHECK-NEXT: [[first_iteration_checkB:[^ ]+]] = icmp ult i32 %offB, %lengthB
-; CHECK-NEXT: [[WideChkB:[^ ]+]] = and i1 [[first_iteration_checkB]], [[limit_checkB]]
-; CHECK-NEXT: [[maxMinusOne:[^ ]+]] = add i64 %max, -1
-; CHECK-NEXT: [[limit_checkMax:[^ ]+]] = icmp ule i64 15, [[maxMinusOne]]
-; CHECK-NEXT: [[first_iteration_checkMax:[^ ]+]] = icmp ult  i64 0, %max
-; CHECK-NEXT: [[WideChkMax:[^ ]+]] = and i1 [[first_iteration_checkMax]], [[limit_checkMax]]
-; CHECK-NEXT: [[lenA:[^ ]+]] = add i32 %lengthA, -1
-; CHECK-NEXT: [[idxA:[^ ]+]] = sub i32 [[lenA]], %offA
-; CHECK-NEXT: [[limit_checkA:[^ ]+]] = icmp ule i32 15, [[idxA]]
-; CHECK-NEXT: [[first_iteration_checkA:[^ ]+]] = icmp ult i32 %offA, %lengthA
-; CHECK-NEXT: [[WideChkA:[^ ]+]] = and i1 [[first_iteration_checkA]], [[limit_checkA]]
   %lengthA = call i32 @length(i8* %arrA)
   %lengthB = call i32 @length(i8* %arrB)
   br label %loop
 
 loop:
-; CHECK-LABEL: loop:
-; CHECK: [[BandMax:[^ ]+]] = and i1 [[WideChkB]], [[WideChkMax]]
-; CHECK: [[ABandMax:[^ ]+]] = and i1 [[BandMax]], [[WideChkA]]
-; CHECK: call void (i1, ...) @llvm.experimental.guard(i1 [[ABandMax]], i32 9)
   %iv = phi i64 [0, %entry ], [ %iv.next, %loop ]
   %iv.trunc = trunc i64 %iv to i32
   %indexA = add i32 %iv.trunc, %offA
@@ -110,15 +162,33 @@ exit:
 ; for (i64 i= 5; i>= 2; i++)
 ; this loop wraps around after reaching 2^64.
 define i64 @iv_rc_different_type(i32 %offA, i8* %arrA) {
-; CHECK-LABEL: iv_rc_different_type
+; CHECK-LABEL: @iv_rc_different_type(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[LENGTHA:%.*]] = call i32 @length(i8* [[ARRA:%.*]])
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 5, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_TRUNC_32:%.*]] = trunc i64 [[IV]] to i32
+; CHECK-NEXT:    [[INDEXA:%.*]] = add i32 [[IV_TRUNC_32]], [[OFFA:%.*]]
+; CHECK-NEXT:    [[RCA:%.*]] = icmp ult i32 [[INDEXA]], [[LENGTHA]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[RCA]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[INDEXA_EXT:%.*]] = zext i32 [[INDEXA]] to i64
+; CHECK-NEXT:    [[ADDRA:%.*]] = getelementptr inbounds i8, i8* [[ARRA]], i64 [[INDEXA_EXT]]
+; CHECK-NEXT:    [[ELTA:%.*]] = load i8, i8* [[ADDRA]]
+; CHECK-NEXT:    [[RES:%.*]] = add i8 [[ELTA]], 2
+; CHECK-NEXT:    store i8 [[ELTA]], i8* [[ADDRA]]
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[LATCH_CHECK:%.*]] = icmp sge i64 [[IV_NEXT]], 2
+; CHECK-NEXT:    br i1 [[LATCH_CHECK]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[IV_LCSSA:%.*]] = phi i64 [ [[IV]], [[LOOP]] ]
+; CHECK-NEXT:    ret i64 [[IV_LCSSA]]
+;
 entry:
   %lengthA = call i32 @length(i8* %arrA)
   br label %loop
 
 loop:
-; CHECK-LABEL: loop:
-; CHECK: %rcA = icmp ult i32 %indexA, %lengthA
-; CHECK-NEXT: call void (i1, ...) @llvm.experimental.guard(i1 %rcA, i32 9)
   %iv = phi i64 [ 5, %entry ], [ %iv.next, %loop ]
   %iv.trunc.32 = trunc i64 %iv to i32
   %indexA = add i32 %iv.trunc.32, %offA
@@ -134,5 +204,5 @@ loop:
   br i1 %latch.check, label %loop, label %exit
 
 exit:
- ret i64 %iv
+  ret i64 %iv
 }