# HG changeset patch
# User kvn
# Date 1348613297 25200
# Node ID 06f52c4d0e188e7a54256671d54a679c728ab810
# Parent  0702f188baeb990e4bba2e83af4bde43b7b5ed70
7200264: 7192963 changes disabled shift vectors
Summary: Replaced is_vector_use() call with explicit check for vector shift's count.
Reviewed-by: twisti, roland, dlong, vlivanov

diff -r 0702f188baeb -r 06f52c4d0e18 src/share/vm/opto/superword.cpp
--- a/src/share/vm/opto/superword.cpp	Tue Sep 25 10:41:15 2012 -0700
+++ b/src/share/vm/opto/superword.cpp	Tue Sep 25 15:48:17 2012 -0700
@@ -1088,23 +1088,22 @@
   uint start, end;
   VectorNode::vector_operands(p0, &start, &end);
 
-  // Return false if some input is not vector and inside block
+  // Return false if some inputs are not vectors or vectors with different
+  // size or alignment.
+  // Also, for now, return false if not scalar promotion case when inputs are
+  // the same. Later, implement PackNode and allow differing, non-vector inputs
+  // (maybe just the ones from outside the block.)
   for (uint i = start; i < end; i++) {
-    if (!is_vector_use(p0, i)) {
-      // For now, return false if not scalar promotion case (inputs are the same.)
-      // Later, implement PackNode and allow differing, non-vector inputs
-      // (maybe just the ones from outside the block.)
-      if (!same_inputs(p, i)) {
-        return false;
-      }
-    }
+    if (!is_vector_use(p0, i))
+      return false;
   }
   if (VectorNode::is_shift(p0)) {
-    // For now, return false if shift count is vector because
-    // hw does not support it.
-    if (is_vector_use(p0, 2))
+    // For now, return false if shift count is vector or not scalar promotion
+    // case (different shift counts) because it is not supported yet.
+    Node* cnt = p0->in(2);
+    Node_List* cnt_pk = my_pack(cnt);
+    if (cnt_pk != NULL)
       return false;
-    // For the same reason return false if different shift counts.
     if (!same_inputs(p, 2))
       return false;
   }
@@ -1402,7 +1401,7 @@
         ShouldNotReachHere();
       }
       assert(vn != NULL, "sanity");
-      _phase->_igvn.register_new_node_with_optimizer(vn);
+      _igvn.register_new_node_with_optimizer(vn);
       _phase->set_ctrl(vn, _phase->get_ctrl(p->at(0)));
       for (uint j = 0; j < p->size(); j++) {
         Node* pm = p->at(j);
@@ -1451,9 +1450,9 @@
       } else {
         if (t == NULL || t->_lo < 0 || t->_hi > (int)mask) {
           cnt = ConNode::make(C, TypeInt::make(mask));
-          _phase->_igvn.register_new_node_with_optimizer(cnt);
+          _igvn.register_new_node_with_optimizer(cnt);
           cnt = new (C, 3) AndINode(opd, cnt);
-          _phase->_igvn.register_new_node_with_optimizer(cnt);
+          _igvn.register_new_node_with_optimizer(cnt);
           _phase->set_ctrl(cnt, _phase->get_ctrl(opd));
         }
         assert(opd->bottom_type()->isa_int(), "int type only");
@@ -1461,7 +1460,7 @@
         cnt = new (C, 2) MoveI2FNode(cnt);
       }
       if (cnt != opd) {
-        _phase->_igvn.register_new_node_with_optimizer(cnt);
+        _igvn.register_new_node_with_optimizer(cnt);
         _phase->set_ctrl(cnt, _phase->get_ctrl(opd));
       }
       return cnt;
@@ -1473,7 +1472,7 @@
     const Type* p0_t = velt_type(p0);
     VectorNode* vn = VectorNode::scalar2vector(_phase->C, opd, vlen, p0_t);
 
-    _phase->_igvn.register_new_node_with_optimizer(vn);
+    _igvn.register_new_node_with_optimizer(vn);
     _phase->set_ctrl(vn, _phase->get_ctrl(opd));
 #ifdef ASSERT
     if (TraceNewVectors) {
@@ -1496,7 +1495,7 @@
     assert(opd_bt == in->bottom_type()->basic_type(), "all same type");
     pk->add_opd(in);
   }
-  _phase->_igvn.register_new_node_with_optimizer(pk);
+  _igvn.register_new_node_with_optimizer(pk);
   _phase->set_ctrl(pk, _phase->get_ctrl(opd));
 #ifdef ASSERT
   if (TraceNewVectors) {
@@ -1543,7 +1542,7 @@
     int def_pos = alignment(def) / data_size(def);
 
     Node* ex = ExtractNode::make(_phase->C, def, def_pos, velt_basic_type(def));
-    _phase->_igvn.register_new_node_with_optimizer(ex);
+    _igvn.register_new_node_with_optimizer(ex);
     _phase->set_ctrl(ex, _phase->get_ctrl(def));
     _igvn.replace_input_of(use, idx, ex);
     _igvn._worklist.push(def);
@@ -2023,33 +2022,33 @@
     // incorporate any extra invariant piece producing (offset +/- invar) >>> log2(elt)
     Node* log2_elt = _igvn.intcon(exact_log2(elt_size));
     Node* aref     = new (_phase->C, 3) URShiftINode(align_to_ref_p.invar(), log2_elt);
-    _phase->_igvn.register_new_node_with_optimizer(aref);
+    _igvn.register_new_node_with_optimizer(aref);
     _phase->set_ctrl(aref, pre_ctrl);
     if (align_to_ref_p.negate_invar()) {
       e = new (_phase->C, 3) SubINode(e, aref);
     } else {
       e = new (_phase->C, 3) AddINode(e, aref);
     }
-    _phase->_igvn.register_new_node_with_optimizer(e);
+    _igvn.register_new_node_with_optimizer(e);
     _phase->set_ctrl(e, pre_ctrl);
   }
   if (vw > ObjectAlignmentInBytes) {
     // incorporate base e +/- base && Mask >>> log2(elt)
     Node* xbase = new(_phase->C, 2) CastP2XNode(NULL, align_to_ref_p.base());
-    _phase->_igvn.register_new_node_with_optimizer(xbase);
+    _igvn.register_new_node_with_optimizer(xbase);
 #ifdef _LP64
     xbase  = new (_phase->C, 2) ConvL2INode(xbase);
-    _phase->_igvn.register_new_node_with_optimizer(xbase);
+    _igvn.register_new_node_with_optimizer(xbase);
 #endif
     Node* mask = _igvn.intcon(vw-1);
     Node* masked_xbase  = new (_phase->C, 3) AndINode(xbase, mask);
-    _phase->_igvn.register_new_node_with_optimizer(masked_xbase);
+    _igvn.register_new_node_with_optimizer(masked_xbase);
     Node* log2_elt = _igvn.intcon(exact_log2(elt_size));
     Node* bref     = new (_phase->C, 3) URShiftINode(masked_xbase, log2_elt);
-    _phase->_igvn.register_new_node_with_optimizer(bref);
+    _igvn.register_new_node_with_optimizer(bref);
     _phase->set_ctrl(bref, pre_ctrl);
     e = new (_phase->C, 3) AddINode(e, bref);
-    _phase->_igvn.register_new_node_with_optimizer(e);
+    _igvn.register_new_node_with_optimizer(e);
     _phase->set_ctrl(e, pre_ctrl);
   }
 
@@ -2059,20 +2058,20 @@
   } else {
     e = new (_phase->C, 3) AddINode(e, lim0);
   }
-  _phase->_igvn.register_new_node_with_optimizer(e);
+  _igvn.register_new_node_with_optimizer(e);
   _phase->set_ctrl(e, pre_ctrl);
 
   if (stride * scale > 0) {
     // compute V - (e +/- lim0)
     Node* va  = _igvn.intcon(v_align);
     e = new (_phase->C, 3) SubINode(va, e);
-    _phase->_igvn.register_new_node_with_optimizer(e);
+    _igvn.register_new_node_with_optimizer(e);
     _phase->set_ctrl(e, pre_ctrl);
   }
   // compute N = (exp) % V
   Node* va_msk = _igvn.intcon(v_align - 1);
   Node* N = new (_phase->C, 3) AndINode(e, va_msk);
-  _phase->_igvn.register_new_node_with_optimizer(N);
+  _igvn.register_new_node_with_optimizer(N);
   _phase->set_ctrl(N, pre_ctrl);
 
   //   substitute back into (1), so that new limit
@@ -2083,12 +2082,12 @@
   } else {
     lim = new (_phase->C, 3) AddINode(lim0, N);
   }
-  _phase->_igvn.register_new_node_with_optimizer(lim);
+  _igvn.register_new_node_with_optimizer(lim);
   _phase->set_ctrl(lim, pre_ctrl);
   Node* constrained =
     (stride > 0) ? (Node*) new (_phase->C,3) MinINode(lim, orig_limit)
                  : (Node*) new (_phase->C,3) MaxINode(lim, orig_limit);
-  _phase->_igvn.register_new_node_with_optimizer(constrained);
+  _igvn.register_new_node_with_optimizer(constrained);
   _phase->set_ctrl(constrained, pre_ctrl);
   _igvn.hash_delete(pre_opaq);
   pre_opaq->set_req(1, constrained);
diff -r 0702f188baeb -r 06f52c4d0e18 test/compiler/7200264/Test7200264.sh
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/compiler/7200264/Test7200264.sh	Tue Sep 25 15:48:17 2012 -0700
@@ -0,0 +1,150 @@
+#!/bin/sh
+# 
+# Copyright (c) 2012, Oracle and/or its affiliates. All rights reserved.
+# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+# 
+# This code is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License version 2 only, as
+# published by the Free Software Foundation.
+# 
+# This code is distributed in the hope that it will be useful, but WITHOUT
+# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+# FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+# version 2 for more details (a copy is included in the LICENSE file that
+# accompanied this code).
+# 
+# You should have received a copy of the GNU General Public License version
+# 2 along with this work; if not, write to the Free Software Foundation,
+# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+# 
+# Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+# or visit www.oracle.com if you need additional information or have any
+# questions.
+# 
+# 
+
+if [ "${TESTSRC}" = "" ]
+then
+  echo "TESTSRC not set.  Test cannot execute.  Failed."
+  exit 1
+fi
+echo "TESTSRC=${TESTSRC}"
+if [ "${TESTJAVA}" = "" ]
+then
+  echo "TESTJAVA not set.  Test cannot execute.  Failed."
+  exit 1
+fi
+echo "TESTJAVA=${TESTJAVA}"
+if [ "${TESTCLASSES}" = "" ]
+then
+  echo "TESTCLASSES not set.  Test cannot execute.  Failed."
+  exit 1
+fi
+echo "TESTCLASSES=${TESTCLASSES}"
+echo "CLASSPATH=${CLASSPATH}"
+
+# set platform-dependent variables
+OS=`uname -s`
+case "$OS" in
+  SunOS | Linux | Darwin )
+    NULL=/dev/null
+    PS=":"
+    FS="/"
+    ;;
+  Windows_* )
+    NULL=NUL
+    PS=";"
+    FS="\\"
+    ;;
+  CYGWIN_* )
+    NULL=/dev/null
+    PS=";"
+    FS="/"
+    ;;
+  * )
+    echo "Unrecognized system!"
+    exit 1;
+    ;;
+esac
+
+
+${TESTJAVA}${FS}bin${FS}java ${TESTVMOPTS} -Xinternalversion | sed 's/amd64/x86/' | grep "x86" | grep "Server VM" | grep "debug"
+
+# Only test fastdebug Server VM on x86
+if [ $? != 0 ]
+then
+    echo "Test Passed"
+    exit 0
+fi
+
+cp ${TESTSRC}${FS}TestIntVect.java .
+${TESTJAVA}${FS}bin${FS}javac -d . TestIntVect.java
+
+${TESTJAVA}${FS}bin${FS}java ${TESTVMOPTS} -Xbatch -XX:-TieredCompilation -XX:CICompilerCount=1 -XX:+PrintCompilation -XX:+TraceNewVectors TestIntVect > test.out 2>&1
+
+COUNT=`grep AddVI test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 4 ]
+then
+    echo "Test Failed: AddVI $COUNT < 4"
+    exit 1
+fi
+
+# AddVI is generated for test_subc
+COUNT=`grep SubVI test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 4 ]
+then
+    echo "Test Failed: SubVI $COUNT < 4"
+    exit 1
+fi
+
+# LShiftVI+SubVI is generated for test_mulc
+COUNT=`grep MulVI test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 2 ]
+then
+    echo "Test Failed: MulVI $COUNT < 2"
+    exit 1
+fi
+
+COUNT=`grep AndV test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 3 ]
+then
+    echo "Test Failed: AndV $COUNT < 3"
+    exit 1
+fi
+
+COUNT=`grep OrV test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 3 ]
+then
+    echo "Test Failed: OrV $COUNT < 3"
+    exit 1
+fi
+
+COUNT=`grep XorV test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 3 ]
+then
+    echo "Test Failed: XorV $COUNT < 3"
+    exit 1
+fi
+
+COUNT=`grep LShiftVI test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 5 ]
+then
+    echo "Test Failed: LShiftVI $COUNT < 5"
+    exit 1
+fi
+
+# RShiftVI + URShiftVI
+COUNT=`grep RShiftVI test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 6 ]
+then
+    echo "Test Failed: RShiftVI $COUNT < 6"
+    exit 1
+fi
+
+COUNT=`grep URShiftVI test.out | wc -l | awk '{print $1}'`
+if [ $COUNT -lt 3 ]
+then
+    echo "Test Failed: URShiftVI $COUNT < 3"
+    exit 1
+fi
+
diff -r 0702f188baeb -r 06f52c4d0e18 test/compiler/7200264/TestIntVect.java
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/compiler/7200264/TestIntVect.java	Tue Sep 25 15:48:17 2012 -0700
@@ -0,0 +1,650 @@
+/*
+ * Copyright (c) 2012, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+/**
+ * @test
+ * @bug 7200264
+ * @summary 7192963 changes disabled shift vectors
+ *
+ * @run shell Test7200264.sh
+ */
+
+/*
+ * Copy of test/compiler/6340864/TestIntVect.java without performance tests.
+ */
+public class TestIntVect {
+  private static final int ARRLEN = 997;
+  private static final int ITERS  = 11000;
+  private static final int ADD_INIT = Integer.MAX_VALUE-500;
+  private static final int BIT_MASK = 0xEC80F731;
+  private static final int VALUE = 15;
+  private static final int SHIFT = 32;
+
+  public static void main(String args[]) {
+    System.out.println("Testing Integer vectors");
+    int errn = test();
+    if (errn > 0) {
+      System.err.println("FAILED: " + errn + " errors");
+      System.exit(97);
+    }
+    System.out.println("PASSED");
+  }
+
+  static int test() {
+    int[] a0 = new int[ARRLEN];
+    int[] a1 = new int[ARRLEN];
+    int[] a2 = new int[ARRLEN];
+    int[] a3 = new int[ARRLEN];
+    int[] a4 = new int[ARRLEN];
+    long[] p2 = new long[ARRLEN/2];
+    // Initialize
+    int gold_sum = 0;
+    for (int i=0; i<ARRLEN; i++) {
+      int val = (int)(ADD_INIT+i);
+      gold_sum += val;
+      a1[i] = val;
+      a2[i] = (int)VALUE;
+      a3[i] = (int)-VALUE;
+      a4[i] = (int)BIT_MASK;
+    }
+    System.out.println("Warmup");
+    for (int i=0; i<ITERS; i++) {
+      test_sum(a1);
+      test_addc(a0, a1);
+      test_addv(a0, a1, (int)VALUE);
+      test_adda(a0, a1, a2);
+      test_subc(a0, a1);
+      test_subv(a0, a1, (int)VALUE);
+      test_suba(a0, a1, a2);
+      test_mulc(a0, a1);
+      test_mulv(a0, a1, (int)VALUE);
+      test_mula(a0, a1, a2);
+      test_divc(a0, a1);
+      test_divv(a0, a1, (int)VALUE);
+      test_diva(a0, a1, a2);
+      test_mulc_n(a0, a1);
+      test_mulv(a0, a1, (int)-VALUE);
+      test_mula(a0, a1, a3);
+      test_divc_n(a0, a1);
+      test_divv(a0, a1, (int)-VALUE);
+      test_diva(a0, a1, a3);
+      test_andc(a0, a1);
+      test_andv(a0, a1, (int)BIT_MASK);
+      test_anda(a0, a1, a4);
+      test_orc(a0, a1);
+      test_orv(a0, a1, (int)BIT_MASK);
+      test_ora(a0, a1, a4);
+      test_xorc(a0, a1);
+      test_xorv(a0, a1, (int)BIT_MASK);
+      test_xora(a0, a1, a4);
+      test_sllc(a0, a1);
+      test_sllv(a0, a1, VALUE);
+      test_srlc(a0, a1);
+      test_srlv(a0, a1, VALUE);
+      test_srac(a0, a1);
+      test_srav(a0, a1, VALUE);
+      test_sllc_n(a0, a1);
+      test_sllv(a0, a1, -VALUE);
+      test_srlc_n(a0, a1);
+      test_srlv(a0, a1, -VALUE);
+      test_srac_n(a0, a1);
+      test_srav(a0, a1, -VALUE);
+      test_sllc_o(a0, a1);
+      test_sllv(a0, a1, SHIFT);
+      test_srlc_o(a0, a1);
+      test_srlv(a0, a1, SHIFT);
+      test_srac_o(a0, a1);
+      test_srav(a0, a1, SHIFT);
+      test_sllc_on(a0, a1);
+      test_sllv(a0, a1, -SHIFT);
+      test_srlc_on(a0, a1);
+      test_srlv(a0, a1, -SHIFT);
+      test_srac_on(a0, a1);
+      test_srav(a0, a1, -SHIFT);
+      test_pack2(p2, a1);
+      test_unpack2(a0, p2);
+      test_pack2_swap(p2, a1);
+      test_unpack2_swap(a0, p2);
+    }
+    // Test and verify results
+    System.out.println("Verification");
+    int errn = 0;
+    {
+      int sum = test_sum(a1);
+      if (sum != gold_sum) {
+        System.err.println("test_sum:  " + sum + " != " + gold_sum);
+        errn++;
+      }
+
+      test_addc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_addc: ", i, a0[i], (int)((int)(ADD_INIT+i)+VALUE));
+      }
+      test_addv(a0, a1, (int)VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_addv: ", i, a0[i], (int)((int)(ADD_INIT+i)+VALUE));
+      }
+      test_adda(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_adda: ", i, a0[i], (int)((int)(ADD_INIT+i)+VALUE));
+      }
+
+      test_subc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_subc: ", i, a0[i], (int)((int)(ADD_INIT+i)-VALUE));
+      }
+      test_subv(a0, a1, (int)VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_subv: ", i, a0[i], (int)((int)(ADD_INIT+i)-VALUE));
+      }
+      test_suba(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_suba: ", i, a0[i], (int)((int)(ADD_INIT+i)-VALUE));
+      }
+
+      test_mulc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulc: ", i, a0[i], (int)((int)(ADD_INIT+i)*VALUE));
+      }
+      test_mulv(a0, a1, (int)VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulv: ", i, a0[i], (int)((int)(ADD_INIT+i)*VALUE));
+      }
+      test_mula(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mula: ", i, a0[i], (int)((int)(ADD_INIT+i)*VALUE));
+      }
+
+      test_divc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divc: ", i, a0[i], (int)((int)(ADD_INIT+i)/VALUE));
+      }
+      test_divv(a0, a1, (int)VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divv: ", i, a0[i], (int)((int)(ADD_INIT+i)/VALUE));
+      }
+      test_diva(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_diva: ", i, a0[i], (int)((int)(ADD_INIT+i)/VALUE));
+      }
+
+      test_mulc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulc_n: ", i, a0[i], (int)((int)(ADD_INIT+i)*(-VALUE)));
+      }
+      test_mulv(a0, a1, (int)-VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulv_n: ", i, a0[i], (int)((int)(ADD_INIT+i)*(-VALUE)));
+      }
+      test_mula(a0, a1, a3);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mula_n: ", i, a0[i], (int)((int)(ADD_INIT+i)*(-VALUE)));
+      }
+
+      test_divc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divc_n: ", i, a0[i], (int)((int)(ADD_INIT+i)/(-VALUE)));
+      }
+      test_divv(a0, a1, (int)-VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divv_n: ", i, a0[i], (int)((int)(ADD_INIT+i)/(-VALUE)));
+      }
+      test_diva(a0, a1, a3);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_diva_n: ", i, a0[i], (int)((int)(ADD_INIT+i)/(-VALUE)));
+      }
+
+      test_andc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_andc: ", i, a0[i], (int)((int)(ADD_INIT+i)&BIT_MASK));
+      }
+      test_andv(a0, a1, (int)BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_andv: ", i, a0[i], (int)((int)(ADD_INIT+i)&BIT_MASK));
+      }
+      test_anda(a0, a1, a4);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_anda: ", i, a0[i], (int)((int)(ADD_INIT+i)&BIT_MASK));
+      }
+
+      test_orc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_orc: ", i, a0[i], (int)((int)(ADD_INIT+i)|BIT_MASK));
+      }
+      test_orv(a0, a1, (int)BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_orv: ", i, a0[i], (int)((int)(ADD_INIT+i)|BIT_MASK));
+      }
+      test_ora(a0, a1, a4);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_ora: ", i, a0[i], (int)((int)(ADD_INIT+i)|BIT_MASK));
+      }
+
+      test_xorc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_xorc: ", i, a0[i], (int)((int)(ADD_INIT+i)^BIT_MASK));
+      }
+      test_xorv(a0, a1, (int)BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_xorv: ", i, a0[i], (int)((int)(ADD_INIT+i)^BIT_MASK));
+      }
+      test_xora(a0, a1, a4);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_xora: ", i, a0[i], (int)((int)(ADD_INIT+i)^BIT_MASK));
+      }
+
+      test_sllc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc: ", i, a0[i], (int)((int)(ADD_INIT+i)<<VALUE));
+      }
+      test_sllv(a0, a1, VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv: ", i, a0[i], (int)((int)(ADD_INIT+i)<<VALUE));
+      }
+
+      test_srlc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>VALUE));
+      }
+      test_srlv(a0, a1, VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>VALUE));
+      }
+
+      test_srac(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac: ", i, a0[i], (int)((int)(ADD_INIT+i)>>VALUE));
+      }
+      test_srav(a0, a1, VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav: ", i, a0[i], (int)((int)(ADD_INIT+i)>>VALUE));
+      }
+
+      test_sllc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_n: ", i, a0[i], (int)((int)(ADD_INIT+i)<<(-VALUE)));
+      }
+      test_sllv(a0, a1, -VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_n: ", i, a0[i], (int)((int)(ADD_INIT+i)<<(-VALUE)));
+      }
+
+      test_srlc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_n: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>(-VALUE)));
+      }
+      test_srlv(a0, a1, -VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_n: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>(-VALUE)));
+      }
+
+      test_srac_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_n: ", i, a0[i], (int)((int)(ADD_INIT+i)>>(-VALUE)));
+      }
+      test_srav(a0, a1, -VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_n: ", i, a0[i], (int)((int)(ADD_INIT+i)>>(-VALUE)));
+      }
+
+      test_sllc_o(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_o: ", i, a0[i], (int)((int)(ADD_INIT+i)<<SHIFT));
+      }
+      test_sllv(a0, a1, SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_o: ", i, a0[i], (int)((int)(ADD_INIT+i)<<SHIFT));
+      }
+
+      test_srlc_o(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_o: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>SHIFT));
+      }
+      test_srlv(a0, a1, SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_o: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>SHIFT));
+      }
+
+      test_srac_o(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_o: ", i, a0[i], (int)((int)(ADD_INIT+i)>>SHIFT));
+      }
+      test_srav(a0, a1, SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_o: ", i, a0[i], (int)((int)(ADD_INIT+i)>>SHIFT));
+      }
+
+      test_sllc_on(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_on: ", i, a0[i], (int)((int)(ADD_INIT+i)<<(-SHIFT)));
+      }
+      test_sllv(a0, a1, -SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_on: ", i, a0[i], (int)((int)(ADD_INIT+i)<<(-SHIFT)));
+      }
+
+      test_srlc_on(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_on: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>(-SHIFT)));
+      }
+      test_srlv(a0, a1, -SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_on: ", i, a0[i], (int)((int)(ADD_INIT+i)>>>(-SHIFT)));
+      }
+
+      test_srac_on(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_on: ", i, a0[i], (int)((int)(ADD_INIT+i)>>(-SHIFT)));
+      }
+      test_srav(a0, a1, -SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_on: ", i, a0[i], (int)((int)(ADD_INIT+i)>>(-SHIFT)));
+      }
+
+      test_pack2(p2, a1);
+      for (int i=0; i<ARRLEN/2; i++) {
+        errn += verify("test_pack2: ", i, p2[i], ((long)(ADD_INIT+2*i) & 0xFFFFFFFFl) | ((long)(ADD_INIT+2*i+1) << 32));
+      }
+      for (int i=0; i<ARRLEN; i++) {
+        a0[i] = -1;
+      }
+      test_unpack2(a0, p2);
+      for (int i=0; i<(ARRLEN&(-2)); i++) {
+        errn += verify("test_unpack2: ", i, a0[i], (ADD_INIT+i));
+      }
+
+      test_pack2_swap(p2, a1);
+      for (int i=0; i<ARRLEN/2; i++) {
+        errn += verify("test_pack2_swap: ", i, p2[i], ((long)(ADD_INIT+2*i+1) & 0xFFFFFFFFl) | ((long)(ADD_INIT+2*i) << 32));
+      }
+      for (int i=0; i<ARRLEN; i++) {
+        a0[i] = -1;
+      }
+      test_unpack2_swap(a0, p2);
+      for (int i=0; i<(ARRLEN&(-2)); i++) {
+        errn += verify("test_unpack2_swap: ", i, a0[i], (ADD_INIT+i));
+      }
+
+    }
+
+    return errn;
+  }
+
+  static int test_sum(int[] a1) {
+    int sum = 0;
+    for (int i = 0; i < a1.length; i+=1) {
+      sum += a1[i];
+    }
+    return sum;
+  }
+
+  static void test_addc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]+VALUE);
+    }
+  }
+  static void test_addv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]+b);
+    }
+  }
+  static void test_adda(int[] a0, int[] a1, int[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]+a2[i]);
+    }
+  }
+
+  static void test_subc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]-VALUE);
+    }
+  }
+  static void test_subv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]-b);
+    }
+  }
+  static void test_suba(int[] a0, int[] a1, int[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]-a2[i]);
+    }
+  }
+
+  static void test_mulc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]*VALUE);
+    }
+  }
+  static void test_mulc_n(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]*(-VALUE));
+    }
+  }
+  static void test_mulv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]*b);
+    }
+  }
+  static void test_mula(int[] a0, int[] a1, int[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]*a2[i]);
+    }
+  }
+
+  static void test_divc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]/VALUE);
+    }
+  }
+  static void test_divc_n(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]/(-VALUE));
+    }
+  }
+  static void test_divv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]/b);
+    }
+  }
+  static void test_diva(int[] a0, int[] a1, int[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]/a2[i]);
+    }
+  }
+
+  static void test_andc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]&BIT_MASK);
+    }
+  }
+  static void test_andv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]&b);
+    }
+  }
+  static void test_anda(int[] a0, int[] a1, int[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]&a2[i]);
+    }
+  }
+
+  static void test_orc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]|BIT_MASK);
+    }
+  }
+  static void test_orv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]|b);
+    }
+  }
+  static void test_ora(int[] a0, int[] a1, int[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]|a2[i]);
+    }
+  }
+
+  static void test_xorc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]^BIT_MASK);
+    }
+  }
+  static void test_xorv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]^b);
+    }
+  }
+  static void test_xora(int[] a0, int[] a1, int[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]^a2[i]);
+    }
+  }
+
+  static void test_sllc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]<<VALUE);
+    }
+  }
+  static void test_sllc_n(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]<<(-VALUE));
+    }
+  }
+  static void test_sllc_o(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]<<SHIFT);
+    }
+  }
+  static void test_sllc_on(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]<<(-SHIFT));
+    }
+  }
+  static void test_sllv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]<<b);
+    }
+  }
+
+  static void test_srlc(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>>VALUE);
+    }
+  }
+  static void test_srlc_n(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>>(-VALUE));
+    }
+  }
+  static void test_srlc_o(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>>SHIFT);
+    }
+  }
+  static void test_srlc_on(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>>(-SHIFT));
+    }
+  }
+  static void test_srlv(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>>b);
+    }
+  }
+
+  static void test_srac(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>VALUE);
+    }
+  }
+  static void test_srac_n(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>(-VALUE));
+    }
+  }
+  static void test_srac_o(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>SHIFT);
+    }
+  }
+  static void test_srac_on(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>(-SHIFT));
+    }
+  }
+  static void test_srav(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)(a1[i]>>b);
+    }
+  }
+
+  static void test_pack2(long[] p2, int[] a1) {
+    if (p2.length*2 > a1.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      long l0 = (long)a1[i*2+0];
+      long l1 = (long)a1[i*2+1];
+      p2[i] = (l1 << 32) | (l0 & 0xFFFFFFFFl);
+    }
+  }
+  static void test_unpack2(int[] a0, long[] p2) {
+    if (p2.length*2 > a0.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      long l = p2[i];
+      a0[i*2+0] = (int)(l & 0xFFFFFFFFl);
+      a0[i*2+1] = (int)(l >> 32);
+    }
+  }
+  static void test_pack2_swap(long[] p2, int[] a1) {
+    if (p2.length*2 > a1.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      long l0 = (long)a1[i*2+0];
+      long l1 = (long)a1[i*2+1];
+      p2[i] = (l0 << 32) | (l1 & 0xFFFFFFFFl);
+    }
+  }
+  static void test_unpack2_swap(int[] a0, long[] p2) {
+    if (p2.length*2 > a0.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      long l = p2[i];
+      a0[i*2+0] = (int)(l >> 32);
+      a0[i*2+1] = (int)(l & 0xFFFFFFFFl);
+    }
+  }
+
+  static int verify(String text, int i, int elem, int val) {
+    if (elem != val) {
+      System.err.println(text + "[" + i + "] = " + elem + " != " + val);
+      return 1;
+    }
+    return 0;
+  }
+
+  static int verify(String text, int i, long elem, long val) {
+    if (elem != val) {
+      System.err.println(text + "[" + i + "] = " + Long.toHexString(elem) + " != " + Long.toHexString(val));
+      return 1;
+    }
+    return 0;
+  }
+}