8311932: Suboptimal compiled code of nested loop over memory segment

Reviewed-by: thartmann, chagedorn

src/hotspot/share/opto/loopnode.cpp

@@ -855,6 +855,28 @@ bool PhaseIdealLoop::create_loop_nest(IdealLoopTree* loop, Node_List &old_new) {
     // not a loop after all
     return false;
   }
+
+  if (range_checks.size() > 0) {
+    // This transformation requires peeling one iteration. Also, if it has range checks and they are eliminated by Loop
+    // Predication, then 2 Hoisted Check Predicates are added for one range check. Finally, transforming a long range
+    // check requires extra logic to be executed before the loop is entered and for the outer loop. As a result, the
+    // transformations can't pay off for a small number of iterations: roughly, if the loop runs for 3 iterations, it's
+    // going to execute as many range checks once transformed with range checks eliminated (1 peeled iteration with
+    // range checks + 2 predicates per range checks) as it would have not transformed. It also has to pay for the extra
+    // logic on loop entry and for the outer loop.
+    loop->compute_trip_count(this);
+    if (head->is_CountedLoop() && head->as_CountedLoop()->has_exact_trip_count()) {
+      if (head->as_CountedLoop()->trip_count() <= 3) {
+        return false;
+      }
+    } else {
+      loop->compute_profile_trip_cnt(this);
+      if (!head->is_profile_trip_failed() && head->profile_trip_cnt() <= 3) {
+        return false;
+      }
+    }
+  }
+
   julong orig_iters = (julong)hi->hi_as_long() - lo->lo_as_long();
   iters_limit = checked_cast<int>(MIN2((julong)iters_limit, orig_iters));
 

test/hotspot/jtreg/compiler/c2/irTests/TestLongRangeChecks.java

@@ -29,7 +29,7 @@ import java.util.Objects;
 
 /*
  * @test
- * @bug 8259609 8276116
+ * @bug 8259609 8276116 8311932
  * @summary C2: optimize long range checks in long counted loops
  * @library /test/lib /
  * @requires vm.compiler2.enabled
@@ -38,9 +38,9 @@ import java.util.Objects;
 
 public class TestLongRangeChecks {
     public static void main(String[] args) {
-        TestFramework.runWithFlags("-XX:-UseCountedLoopSafepoints");
+        TestFramework.runWithFlags("-XX:+TieredCompilation", "-XX:-UseCountedLoopSafepoints", "-XX:LoopUnrollLimit=0");
-        TestFramework.runWithFlags("-XX:+UseCountedLoopSafepoints", "-XX:LoopStripMiningIter=1");
+        TestFramework.runWithFlags("-XX:+TieredCompilation", "-XX:+UseCountedLoopSafepoints", "-XX:LoopStripMiningIter=1", "-XX:LoopUnrollLimit=0");
-        TestFramework.runWithFlags("-XX:+UseCountedLoopSafepoints", "-XX:LoopStripMiningIter=1000");
+        TestFramework.runWithFlags("-XX:+TieredCompilation", "-XX:+UseCountedLoopSafepoints", "-XX:LoopStripMiningIter=1000", "-XX:LoopUnrollLimit=0");
     }
 
 
@@ -246,4 +246,58 @@ public class TestLongRangeChecks {
     private void testStridePosScaleNegInIntLoop2_runner() {
         testStridePosScaleNegInIntLoop2(0, 100, 200, 198);
     }
+
+    @Test
+    @IR(counts = { IRNode.LONG_COUNTED_LOOP, "1" })
+    @IR(failOn = { IRNode.COUNTED_LOOP, IRNode.LOOP })
+    public static void testStridePosScalePosShortLoop(long start, long stop, long length, long offset) {
+        final long scale = 1;
+        final long stride = 1;
+
+        // Loop runs for too few iterations. Transforming it wouldn't pay off.
+        for (long i = start; i < stop; i += stride) {
+            Objects.checkIndex(scale * i + offset, length);
+        }
+    }
+
+    @Run(test = "testStridePosScalePosShortLoop")
+    private void testStridePosScalePosShortLoop_runner() {
+        testStridePosScalePosShortLoop(0, 2, 2, 0);
+    }
+
+    @Test
+    @IR(counts = { IRNode.COUNTED_LOOP, "1" })
+    @IR(failOn = { IRNode.LOOP })
+    public static void testStridePosScalePosInIntLoopShortLoop1(int start, int stop, long length, long offset) {
+        final long scale = 2;
+        final int stride = 1;
+
+        // Same but with int loop
+        for (int i = start; i < stop; i += stride) {
+            Objects.checkIndex(scale * i + offset, length);
+        }
+    }
+
+    @Run(test = "testStridePosScalePosInIntLoopShortLoop1")
+    private void testStridePosScalePosInIntLoopShortLoop1_runner() {
+        testStridePosScalePosInIntLoopShortLoop1(0, 2, 4, 0);
+    }
+
+    @Test
+    @IR(counts = { IRNode.COUNTED_LOOP, "1" })
+    @IR(failOn = { IRNode.LOOP })
+    public static void testStridePosScalePosInIntLoopShortLoop2(long length, long offset) {
+        final long scale = 2;
+        final int stride = 1;
+
+        // Same but with int loop
+        for (int i = 0; i < 3; i += stride) {
+            Objects.checkIndex(scale * i + offset, length);
+        }
+    }
+
+    @Run(test = "testStridePosScalePosInIntLoopShortLoop2")
+    private void testStridePosScalePosInIntLoopShortLoop2_runner() {
+        testStridePosScalePosInIntLoopShortLoop2(6, 0);
+    }
 }