8348263: C2 SuperWord: TestMemorySegment.java has failing IR rules with AlignVector after JDK-8343685

Reviewed-by: chagedorn, thartmann

Author: eme64

test/hotspot/jtreg/compiler/loopopts/superword/TestMemorySegment.java

@@ -39,6 +39,14 @@
  * @run driver compiler.loopopts.superword.TestMemorySegment ByteArray
  */
 
+/*
+ * @test id=byte-array-AlignVector
+ * @bug 8329273 8348263
+ * @summary Test vectorization of loops over MemorySegment
+ * @library /test/lib /
+ * @run driver compiler.loopopts.superword.TestMemorySegment ByteArray AlignVector
+ */
+
 /*
  * @test id=char-array
  * @bug 8329273
@@ -63,6 +71,14 @@
  * @run driver compiler.loopopts.superword.TestMemorySegment IntArray
  */
 
+/*
+ * @test id=int-array-AlignVector
+ * @bug 8329273 8348263
+ * @summary Test vectorization of loops over MemorySegment
+ * @library /test/lib /
+ * @run driver compiler.loopopts.superword.TestMemorySegment IntArray AlignVector
+ */
+
 /*
  * @test id=long-array
  * @bug 8329273
@@ -71,6 +87,14 @@
  * @run driver compiler.loopopts.superword.TestMemorySegment LongArray
  */
 
+/*
+ * @test id=long-array-AlignVector
+ * @bug 8329273 8348263
+ * @summary Test vectorization of loops over MemorySegment
+ * @library /test/lib /
+ * @run driver compiler.loopopts.superword.TestMemorySegment LongArray AlignVector
+ */
+
 /*
  * @test id=float-array
  * @bug 8329273
@@ -111,6 +135,14 @@
  * @run driver compiler.loopopts.superword.TestMemorySegment Native
  */
 
+/*
+ * @test id=native-AlignVector
+ * @bug 8329273 8348263
+ * @summary Test vectorization of loops over MemorySegment
+ * @library /test/lib /
+ * @run driver compiler.loopopts.superword.TestMemorySegment Native AlignVector
+ */
+
 // FAILS: mixed providers currently do not vectorize. Maybe there is some inlining issue.
 // /*
 //  * @test id=mixed-array
@@ -140,6 +172,9 @@ public class TestMemorySegment {
     public static void main(String[] args) {
         TestFramework framework = new TestFramework(TestMemorySegmentImpl.class);
         framework.addFlags("-DmemorySegmentProviderNameForTestVM=" + args[0]);
+        if (args.length > 1 && args[1].equals("AlignVector")) {
+            framework.addFlags("-XX:+AlignVector");
+        }
         framework.setDefaultWarmup(100);
         framework.start();
     }
@@ -556,6 +591,7 @@ static Object[] testIntLoop_longIndex_longInvar_sameAdr_int(MemorySegment a, lon
                   IRNode.ADD_VI,        "> 0",
                   IRNode.STORE_VECTOR,  "> 0"},
         applyIfPlatform = {"64-bit", "true"},
+        applyIf = {"AlignVector", "false"},
         applyIfCPUFeatureOr = {"sse4.1", "true", "asimd", "true"})
     static Object[] testIntLoop_longIndex_intInvar_int(MemorySegment a, int invar) {
         for (int i = 0; i < (int)a.byteSize()/4; i++) {
@@ -572,6 +608,7 @@ static Object[] testIntLoop_longIndex_intInvar_int(MemorySegment a, int invar) {
                   IRNode.ADD_VI,        "> 0",
                   IRNode.STORE_VECTOR,  "> 0"},
         applyIfPlatform = {"64-bit", "true"},
+        applyIf = {"AlignVector", "false"},
         applyIfCPUFeatureOr = {"sse4.1", "true", "asimd", "true"})
     static Object[] testIntLoop_longIndex_longInvar_int(MemorySegment a, long invar) {
         for (int i = 0; i < (int)a.byteSize()/4; i++) {
@@ -705,14 +742,12 @@ static Object[] testLongLoop_iv_int(MemorySegment a) {
     }
 
     @Test
-    //@IR(counts = {IRNode.LOAD_VECTOR_I, "> 0",
-    //              IRNode.ADD_VI,        "> 0",
-    //              IRNode.STORE_VECTOR,  "> 0"},
-    //    applyIfPlatform = {"64-bit", "true"},
-    //    applyIf = {"AlignVector", "false"},
-    //    applyIfCPUFeatureOr = {"sse4.1", "true", "asimd", "true"})
-    // FAILS: for native memory. I think it is because of invariants, but need investigation.
-    //        The long -> int loop conversion introduces extra invariants.
+    @IR(counts = {IRNode.LOAD_VECTOR_I, "> 0",
+                  IRNode.ADD_VI,        "> 0",
+                  IRNode.STORE_VECTOR,  "> 0"},
+        applyIfPlatform = {"64-bit", "true"},
+        applyIf = {"AlignVector", "false"},
+        applyIfCPUFeatureOr = {"sse4.1", "true", "asimd", "true"})
     static Object[] testLongLoop_longIndex_intInvar_sameAdr_int(MemorySegment a, int invar) {
         for (long i = 0; i < a.byteSize()/4; i++) {
             long adr = 4L * (long)(i) + 4L * (long)(invar);
@@ -723,14 +758,12 @@ static Object[] testLongLoop_longIndex_intInvar_sameAdr_int(MemorySegment a, int
     }
 
     @Test
-    //@IR(counts = {IRNode.LOAD_VECTOR_I, "> 0",
-    //              IRNode.ADD_VI,        "> 0",
-    //              IRNode.STORE_VECTOR,  "> 0"},
-    //    applyIfPlatform = {"64-bit", "true"},
-    //    applyIf = {"AlignVector", "false"},
-    //    applyIfCPUFeatureOr = {"sse4.1", "true", "asimd", "true"})
-    // FAILS: for native memory. I think it is because of invariants, but need investigation.
-    //        The long -> int loop conversion introduces extra invariants.
+    @IR(counts = {IRNode.LOAD_VECTOR_I, "> 0",
+                  IRNode.ADD_VI,        "> 0",
+                  IRNode.STORE_VECTOR,  "> 0"},
+        applyIfPlatform = {"64-bit", "true"},
+        applyIf = {"AlignVector", "false"},
+        applyIfCPUFeatureOr = {"sse4.1", "true", "asimd", "true"})
     static Object[] testLongLoop_longIndex_longInvar_sameAdr_int(MemorySegment a, long invar) {
         for (long i = 0; i < a.byteSize()/4; i++) {
             long adr = 4L * (long)(i) + 4L * (long)(invar);