81 lines
3.4 KiB
Java
81 lines
3.4 KiB
Java
/*
|
|
* Copyright (c) 2022, 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 8283466
|
|
* @summary When skeleton predicates are not copied into peeled loop and initialized, this can happen:
|
|
* 1. The rangecheck from a load is hoisted outside of the counted loop.
|
|
* 2. The counted loop is peeled (we disable unswitching and unrolling with arguments)
|
|
* 3. The type inside the peeled loop may now be narrower.
|
|
* 4. The dataflow can die when a type becomes impossible.
|
|
* 5. The rangecheck is still before the peeling, and is not copied to the peeled loop. Hence
|
|
* we do not statically realize that the peeled loop can never be entered.
|
|
*
|
|
* @run main/othervm -XX:+UnlockDiagnosticVMOptions -XX:+StressLCM -XX:+StressGCM -XX:+StressCCP -XX:+StressIGVN
|
|
* -Xcomp -XX:-TieredCompilation
|
|
* -XX:LoopMaxUnroll=0 -XX:LoopUnrollLimit=0 -XX:-LoopUnswitching
|
|
* -XX:CompileCommand=compileonly,compiler.loopopts.TestPeelingSkeletonPredicateInitialization::*
|
|
* compiler.loopopts.TestPeelingSkeletonPredicateInitialization
|
|
*/
|
|
|
|
package compiler.loopopts;
|
|
|
|
public class TestPeelingSkeletonPredicateInitialization {
|
|
int N = 400;
|
|
int array[] = new int[N];
|
|
int array2[] = new int[N];
|
|
void run(int X, int inv, boolean b) {
|
|
// have the arguments so the values are unknown to C2, cannot optimize things away
|
|
try {
|
|
// not sure why needed. maybe has sth to do with div_by_zero below?
|
|
int tmp = 1 / 0;
|
|
} catch (ArithmeticException e) {
|
|
}
|
|
for(int i = 2; i > X; i-=3) {
|
|
// potential div_by_zero: somehow the exit is not loop exit but rethrow
|
|
// also: i-1 only works in peeled iteration
|
|
// in peeled loop: ConvI2L dies because it knows dataflow is impossible
|
|
// If skeleton_predicate is missing for peeled loop, then controlflow does not die
|
|
array[i - 1] /= inv;
|
|
// loop invariant check that is not hoisted: this becomes reason for peeling
|
|
array[inv] += 1;
|
|
if (b) {
|
|
// seems to be required for the memory phi so that it can be mangled when data flow dies
|
|
array[inv] += 1;
|
|
array2[inv] += 1;
|
|
}
|
|
}
|
|
}
|
|
public static void main(String[] strArr) {
|
|
try {
|
|
TestPeelingSkeletonPredicateInitialization _instance = new TestPeelingSkeletonPredicateInitialization();
|
|
for (int i = 0; i < 10000; i++) {
|
|
_instance.run(100, 3, false);
|
|
}
|
|
} catch (Exception ex) {
|
|
}
|
|
}
|
|
}
|
|
|