417e8e449d
Bail out in PhaseIdealLoop:split_thru_phi when trying to split a Div or ModNode iv phi whose zero check was removed but could potentially still be zero based on type information. Reviewed-by: kvn, thartmann
96 lines
3.8 KiB
Java
96 lines
3.8 KiB
Java
/*
|
|
* Copyright (c) 2020, 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 8248552
|
|
* @summary A Division/modulo node whose zero check was removed is split through an induction variable phi and executed before
|
|
* the loop limit check resulting in a SIGFPE because the divisor is zero.
|
|
*
|
|
* @run main/othervm -XX:CompileCommand=dontinline,compiler.c2.loopopts.TestSplitThruPhiDivMod::test* compiler.c2.loopopts.TestSplitThruPhiDivMod
|
|
*/
|
|
package compiler.c2.loopopts;
|
|
|
|
public class TestSplitThruPhiDivMod {
|
|
|
|
int x;
|
|
|
|
public int testMod() {
|
|
int i1 = 2;
|
|
for (int i = 5; i < 25; i++) {
|
|
for (int j = 50; j > 1; j -= 2) {
|
|
/*
|
|
* Zero check is removed based on the type of the induction variable phi (variable j) since its always between 1 and 50.
|
|
* However, when splitting the modulo node through the phi, it can be executed right after the subtraction j-2 which can be
|
|
* 0 before evaluation the loop limit condition in the last iteration when j is 2: j-2 = 2-2 = 0. This results in a SIGFPE.
|
|
* The fix is to not split a division or modulo node 'n' through the induction variable phi if the zero check was removed
|
|
* earlier and the new inputs of the clones of 'n' after the split could be zero (i.e. the type of the clones of 'n' include 0).
|
|
*/
|
|
x = (20 % j); // Problematic division as part of modulo. Results in a SIGFPE, even though j is always non-zero.
|
|
i1 = (i1 / i);
|
|
for (int k = 3; k > 1; k--) {
|
|
switch ((i % 4) + 22) {
|
|
case 22:
|
|
switch (j % 10) {
|
|
case 83:
|
|
x += 5;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
return i1;
|
|
}
|
|
|
|
public int testDiv() {
|
|
int i1 = 2;
|
|
for (int i = 5; i < 25; i++) {
|
|
for (int j = 50; j > 1; j -= 2) {
|
|
// Same issue as above but with a division node. See explanation above.
|
|
x = (20 / j); // Problematic division. Results in a SIGFPE, even though j is always non-zero.
|
|
i1 = (i1 / i);
|
|
for (int k = 3; k > 1; k--) {
|
|
switch ((i % 4) + 22) {
|
|
case 22:
|
|
switch (j % 10) {
|
|
case 83:
|
|
x += 5;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
return i1;
|
|
}
|
|
|
|
public static void main(String[] strArr) {
|
|
TestSplitThruPhiDivMod t = new TestSplitThruPhiDivMod();
|
|
for (int i = 0; i < 10000; i++) {
|
|
t.testDiv();
|
|
t.testMod();
|
|
}
|
|
}
|
|
}
|