summaryrefslogtreecommitdiffstats
path: root/llvm/test/Transforms/LoopPredication/reverse.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/Transforms/LoopPredication/reverse.ll')
-rw-r--r--llvm/test/Transforms/LoopPredication/reverse.ll235
1 files changed, 193 insertions, 42 deletions
diff --git a/llvm/test/Transforms/LoopPredication/reverse.ll b/llvm/test/Transforms/LoopPredication/reverse.ll
index 3d695d1b4ff..b5b132b0488 100644
--- a/llvm/test/Transforms/LoopPredication/reverse.ll
+++ b/llvm/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 ]
OpenPOWER on IntegriCloud